UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

The effects of prenatal exposure to alcohol, tobacco and other risks on children’s health, behaviour… George, Mary Anne 2001

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
831-ubc_2001-610934.pdf [ 16.34MB ]
Metadata
JSON: 831-1.0076855.json
JSON-LD: 831-1.0076855-ld.json
RDF/XML (Pretty): 831-1.0076855-rdf.xml
RDF/JSON: 831-1.0076855-rdf.json
Turtle: 831-1.0076855-turtle.txt
N-Triples: 831-1.0076855-rdf-ntriples.txt
Original Record: 831-1.0076855-source.json
Full Text
831-1.0076855-fulltext.txt
Citation
831-1.0076855.ris

Full Text

THE EFFECTS OF PRENATAL EXPOSURE TO ALCOHOL, TOBACCO AND OTHER RISKS ON CHILDREN'S HEALTH, BEHAVIOUR AND ACADEMIC ABILITIES by MARY ANNE GEORGE B.A., Acadia University, 1966 M.L.S., The University of British Columbia, 1989 A THESIS SUBMITTED IN PARTIAL FULFILMENT OF THE REQUIREMENT FOR THE DEGREE OF DOCTOR OF PHILOSOPHY in THE FACULTY OF GRADUATE STUDIES Individual Interdisciplinary Graduate Studies Program We accept this thesis as confirming to the required standard  THE UNIVERSITY OF BRITISH COLUMBIA April 2001 © Mary Anne George, 2001  In  presenting  this  thesis in  degree at the University of  partial  fulfilment of  the  requirements  for  an advanced  British Columbia, I agree that the Library shall make it  freely available for reference and study. I further agree that permission for extensive copying of this thesis for department  or  by  his  or  scholarly purposes may be granted her  representatives.  It  is  by the head of  understood  that  copying  my or  publication of this thesis for financial gain shall not be allowed without my written permission.  DE-6 (2/88)  Abstract Maternal behaviours during pregnancy may contribute to unhealthy child development. Previous research has linked adverse outcomes to prenatal exposure to heavy alcohol use, to tobacco use, and to socioeconomic background. Health professionals have been responsive because of the preventable nature of these risks. This research examined the impact on childrenfromdiverse backgrounds who had been exposed prenatally to varying amounts of alcohol and tobacco, or both. It further assessed the additive effects of these risks. The Vancouver Island Pregnancy Follow-up Study is population-based. It examined the health, academic abilities and behaviours of 8-year old children born over a 1-year period in a pre-defined area of British Columbia (n=3,659). It attempted tofindthe total randomly selected sample (n=403), with the result that 65.7% families (n=265) participated, 25.8% were not located, 5.5% refused, and 3.0% were not available. The data sources were children, who were tested on their academic abilities using the WRAT-3, and their parents and teachers who were interviewed regarding the child's health, behaviour and school performance. Results with respect to academic abilities showed that only reading scores were associated with prenatal alcohol exposure. Reading and spelling ability scores were associated with the combination of being prenatally exposed to alcohol and having a father with low education. Poorer health ratings were associated with prenatal alcohol exposure, with the combination of pre- and postnatal tobacco exposure, and with lower socioeconomic background. All three independent variables were associated with behaviour scores. Children of heavy drinkers had higher scores on the total behaviour scale and onfivesubscales. Prenatal tobacco exposure was associated with higher Conduct Disorder scores. The findings are consistent with other longitudinal studies. In the absence of other risk factors, there was little evidence of associations between prenatal exposure to moderate alcohol use or risk level and academic abilities, health ratings, or behavioural problems. It may be productive to take ecological approaches to understanding children's behavioural, academic and health problems in the context of their socioeconomic conditions and their exposure to alcohol prenatally and to tobacco both prenatally and postnatally.  ii  Table of Contents Abstract  ii  List of Figures  vi  List of Tables  vii  Acknowledgments  ix  Chapter I Introduction 1.1 Introduction 1.2 Definitions  1 1 5  Chapter II Literature Review 2.1 Introduction and methods 2.2 Prenatal effects of alcohol 2.2.1 Physical and health effects 2.2.2 Cognitive and behavioural effects of children prenatally exposed to alcohol 2.2.2.i Cognitive development, academic abilities or learning problems 2.2.2.ii Behavioural effects of prenatal alcohol exposure 2.2.3 Summary and discussion of prenatal alcohol exposure research 2.3 Prenatal and postnatal exposure to tobacco 2.3.1 Physical effects of prenatal tobacco exposure 2.3.2 Cognitive and behavioural effects of prenatal tobacco exposure 2.3.2.i Behavioural effects of prenatal tobacco exposure 2.3.2ii Cognitive development, academic abilities or learning problems 2.3.3 Summary and discussion of prenatal tobacco exposure research 2.3.4 Postnatal exposure to tobacco 2.4 Socioeconomic background 2.5 Other potentially confounding factors 2.5.1 Other maternal and child factors 2.5.2 Other drugs 2.6 Summary 2.7 Theoretical guides Chapter III Research Questions and Methods 3.1 Research Questions 3.2 Methods 3.2.1 Ethics review 3.2.2 Stages of data collection 3.2.3 Population and sample iii  6 .6 8 11 16 . 16 31 39 49 51 57 58 . 77 89 93 104 110 112 113 116 120  127 127 .130 130 131 132  3.2.3.i Study population and procedures for locating subjects 3.2.3.ii Study sample 3.2.3.ii (a) Refusal rate 3.2.3.ii (b) Teachers' participation 3.2.4 Variables and measures 3.2.4.i Independent Variables 3.2.4.i (a) Prenatal alcohol exposure 3.2.4.i(b) Prenatal and postnatal tobacco exposure 3.2.4.i (c) Socioeconomic background 3.2.4.H Dependent Variables 3.2.4.ii (a) Health 3.2.4.ii (b) Academic abilities 3.2.4.U (c) Behaviour 3.2.4.iii Other variables 3.2.5 Study instruments and means of administration 3.2.5.i Parents' questionnaire 3.2.5.ii Teachers' questionnaire 3.2.6 Analytic procedures ..." Chapter IV Results 4.1 Sample 4.1.1 Participation in the follow-up study 4.1.2 Demographic characteristics of sample participating at Stage 3 4.2 Independent and dependent variables 4.2.1 Independent variables 4.1.2.i Alcohol use during pregnancy 4.1.2.H Tobacco use during pregnancy 4.1.2.iii Socioeconomic status 4.2.2 Outcome variables, academic abilities, health and behaviour . 4.2.2.i Academic ability scores of study sample 4.2.2.H Health 4.2.2.iii Behaviour 4.3 Relationship between independent and dependent variables 4.3.1 Academic Abilities 4.3. l.i Prenatal alcohol exposure and WRAT-3 scores 4.3.1 .ii Prenatal tobacco exposure and WRAT-3 scores 4.3.1.iii Socioeconomic background and WRAT-3 scores 4.3.1. iv Comparing risk levels of a combination of independent variables on WRAT-3 scores 4.3.1 .iv (a) Comparisons of academic scores according to level of risk 4.3.1 .iv (b) Comparisons between the most at-risk groups and least at-risk groups iv  132 135 139 139 141 142 T42 147 148 150 150 153 156 158 159 159 161 163 173 173 173 176 178 178 179 181 182 183 183 186 192 194 195 196 197 198 200 200 202  . 4.3.2 Health 4.3.2.i Birth weight 4.3.2.ii Prenatal alcohol exposure and health 4.3.2.iii Prenatal and postnatal tobacco exposure and health 4.3.2.iv Socioeconomic background and health 4.3.2. V Combined effects of independent variables on health 4.3.3 Behaviour 4.3.3.i Prenatal alcohol exposure and risk and behaviour 4.3.3. H Prenatal and postnatal tobacco exposure and behaviour 4.3.3.iii Socioeconomic background and behaviour 4.3.3.iv Comparison of risk levels of a combination of independent variables on behaviour 4.3.3.iv (a) Comparisons of behavioural scores according to levels of risks 4.3.3.iv (b) Comparisons between the most at-risk groups and least at-risk groups  204 206 207 208 212 213 215 217 225 229 236 236 238  Chapter V  Conclusions  244  Chapter VI  Implications  .258  Chapter VII  Limitations of the study  262  Bibliography  265  Appendices  334  Appendix A: Appendix B.  Questions relating to alcohol use appended to the Provincial Prenatal Record Parents' and teachers' questionnaires  335 336  List of Figures Figure 2.1 Conceptualization of the interacting variables influencing birth outcomes 121 Figure 2.2 Schematic representation of summary of relationships among permissive and provocative maternal risk factors underlying FAS/ARBDs (alcohol-related birth defects) 122 Figure 3.1. Conceptual model of relationships between independent variables (prenatal alcohol exposure, prenatal tobacco exposure and socioeconomic status) and outcome variables (academic abilities, health and behaviour) 129 Figure 4.1 Relationships and correlation coefficients between independent variables (socioeconomic background, tobacco use and alcohol use during pregnancy) and academic abilities as measured by WRAT-3, including p values 195 Figure 4.2 Correlation coefficients (bivariate analysis) between independent variables (alcohol and tobacco exposure during pregnancy and socioeconomic background) and health, as measured by parents' rating of child's health and number of visits to medical professionals, including p values 205 Figure 4.3 Correlation coefficients (bivariate anlyses) between maternal education, prenatal and postnatal exposure to smoking and child's health 210 Figure 4.4 Correlation coefficients (bivariate analyses) between independent variables (socioeconomic background, tobacco use and alcohol use during pregnancy) and behaviour, as measured by the Teachers' Behaviour Rating Scale, including p values .216  J  •  •  •  .  vi  •  •  •  List of Tables Table 2.1 Studies of cognitive and behavioural outcomes associated with prenatal alcohol exposure, listed in alphabetical order byfirstauthor 44 Table 2.2 Studies of cognitive and behavioural outcomes associated with prenatal tobacco exposure, listed in alphabetical order byfirstauthor 96 Table 3.1 Selected maternal demographic variables of study population of women with live births in central and northern Vancouver Island, 1990-91 data collection 134 Table 3.2 Methods of attempting to locate subjects, with refusal and participation rates 138 Table 3.3 Distribution of T-ACE risk levels for population (Stage 1), random sample (Stage 2) and sample found at Stage 3. . . . . . 145 Table 3.4 Data collected during home interview regarding alcohol consumption 146 Table 3.5 Sources of data for dependent variables at Stage 3 data collection 150 Table 3.6 Variables, variable type and source of data collection 171 Table 4.1 Frequency and percentage of parents in sample with various demographic and lifestyle characteristics 177 Table 4.2 Correlation coefficients (Spearman's rho) between T-ACE risk as measured at Stage 1, T-ACE risk as measured at Stage 2 and alcohol use (frequency times quantity) during pregnancy, including p values 179 Table 4.3 Number and percentage of women who reported levels of drinking 181 Table 4.4 Number (and percentage) of smokers by trimester and categorized according to amount of daily smoking 182 Table 4.5 Results of WRAT-3 scores (n=252), including mean, minimum and maximum scores, standard deviation, correlations between reading, spelling and arithmetic, and compared with U.S. national standardized scores 184 Table 4.6 Correlational relationships (Spearman's rho) between WRAT-3 scores on reading, spelling and arithmetic tests, comparisons with teachers' ratings of reading and arithmetic and with parents' assessments of their child's difficulties with these subjects 186 Table 4.7 Parents' ratings of their child's health, their own health and their youngest and oldest children's health, on a 5-point scale, and correlations among these ratings 187 Table 4.8 Number of children with specific chronic illnesses or conditions . 189 Table 4.9 Frequency of child's visits with various health professionals in the past year, including mean number of visits 190 Table 4.10 Correlational relationships, including p values, among four measures of health: rating by parents, visits with various medical and nurse professionals, amount of time child has been in good health in recent months, and school absenteeism 191 Table 4.11 Correlational relationships among three measures of behaviour: Parents' Behaviour Rating Scale, Teachers' Behaviour Rating Scale, and parents' responses to single-item question on behavioural problems 193 Table 4.12 Frequency of categorical variables used in ANOVA 201  vii  Table 4.13 Contrasts between scores on WRAT-3 tests for reading, spelling and arithmetic for children exposed to high, medium or lowrisksof maternal alcohol consumption, T-ACEriskand fathers' education level . . . . 203 Table 4.14 Teachers' Behaviour Rating Scale ratings, according to T-ACEriskscores (Stage 2 administration), including mean, standard deviation (S.D.) and range of scores according to each T-ACErisklevel 217 Table 4.15 Teachers' Behaviour Rating Scale ratings, according alcohol use levels, including mean, standard deviation (S.D.) and range of scores according to each alcohol level 218 Table 4.16 Correlation coefficients (Spearman's rho) and F tests between Parents' and Teachers' Behaviour Rating Scales for total behaviour score and each of the subscales, according to T-ACErisk(Stage 1 administration) and to alcohol consumption during pregnancy 219 Table 4.17 ANOVA results for Teachers' rating scale using both T-ACE risk (Stage 1) and alcohol consumption during pregnancy, by gender 221 Table 4.18 Comparisons of Teachers' Behaviour Rating Scale scores between children of abstainers and those prenatally exposed to heavy drinking and to social drinking levels 223 Table 4.19 Test statistics for Parents' and Teachers' rating of behaviour subscale Conduct Disorder scores, and combined risks of prenatal tobacco exposure, by each trimester, and current second-hand smoke exposure, for total sample, boys only and girls only, including F test result, degrees of freedom (df) and p value 228 Table 4.20 Correlation coefficients (Spearman's rho) and F tests for total scores on Teachers' Behaviour Rating Scale and total scores on Parents' Behaviour Rating Scale and each of five socioeconomic indicators 230 Table 4.21 ANOVA F test coefficients and p values for associations with teachers' rating for total Behaviour score and subscales and three socioeconomic indicators, household income, and mothers' and fathers' occupation levels 231 Table 4.22 Number and percentage of children wh experienced major worries during their lifetime, according to income category of parents 235 Table 4.23 Number of children who experienced major worries, according to cause of the worry and income category of parents 236 Table 4.24 Scores of children exposed torisksof both tobacco use during 2 and 3 trimester and T-ACErisk,compared with those with no tobacco exposure and low T-ACE risk 239 nd  rd  viii  Acknowledgments Many dedicated, talented and patient people have contributed to this research: those who discussed theoretical issues relating to women and lifestyles; those who provided essential funding; those who offered practical and methodological advice based on their many years of experience; those who carried out the essential tasks of a large project; the hundreds of families who willingly participated as subjects; and those who were simply supportive.  Five people served as members of my doctoral supervisory committee and each deserves my special gratitude. Dr. Lawrence Green served as Supervisor for the initial years of this research and, on his departure to his homeland, passed the torch to Dr. Robert Armstrong. I greatly appreciate both supervisors being available whenever I called upon them. Also on the Committee from its beginning were Dr. Heather Clarke and Dr. Judith Ottoson. When Dr. Ottoson was offered an academic position elsewhere, I appreciated Dr. Julianne Corny's willingness to step in mid-way through the project. Dr. Green, Dr. Armstrong, Dr. Clarke, Dr. Ottoson and Dr. Conry gave willingly of their time. They contributed intellectually to the research design and gave editorial assistance and suggestions regarding methods and analyses to the project and its reporting. They provided mentorshipfromtheir vast knowledge and experience to my doctoral education. Gatheredfromthefieldsor disciplines of Health Promotion, Medicine, Nursing, Adult Education and Educational Psychology, this multidisciplinary team of advisors has shaped this project and my education in numerous ways. I am deeply grateful to each of them.  I would like to thank three groups of people for their roles in the initial stages of this research a ix  decade ago. The current research would not have been possible without the prior work of Principal Investigators, Dr. Robert Armstrong, Dr. Geoffry Robinson and Dr. Christine Loock, who conceptualized a strengthened role for primary care physicians in preventing alcohol-related birth defects. They designed and conducted the original research on Vancouver Island, collecting data that is the foundation of the present study. I thank them for the study. Secondly, Lisa Lund and Norma Lupton contributed in many capacities, including project co-ordination, data entry and interviewing on that original 1990-1991 data collection. Their obvious dedication to this work and to their particular tasks left a collection of excellent data with which to work. Lisa Lund and Norma Lupton were also valuable memory banks for the current project. I thank them both for an untiring willingness to answer questions about work they had done 10 years earlier. Finally, I would like to thank Dr. Chris Lalonde, now of the University of Victoria, for his patience as the keeper of the data for many years and for his contributions in analyzing and interpreting the data.  Majorfinancialsupport for this study was provided by the Scottish Rite Charitable Foundation, an important contributor to research on the prevention and treatment of disabilities. The Foundation contributed three years of funding under their "major grants" awards that was invaluable to the success of this project. A doctoral scholarshipfromthe Lion's Gate Medical Research Foundation and a Joint Studentship awardfromthe B.C. Health Research Foundation and the B.C. Centre for Excellence in Women's Health allowed me the opportunity to dedicate full-time efforts during various stages of the research. I am most appreciative of the financial and in-kind support provided by each of these groups.  x  Hundreds of families contributed by participating as subjects in this study. In 1990-1991, 403 women answered questions about themselves and their newborns, and of these, 265 participated nine years later by completing questionnaires and giving permission for their children to be tested and their children's teachers to be contacted. I appreciate their time and effort and the trust they had in the project. I am also grateful to the hundreds of teachers who willingly participated and took, the time during their professional lives to complete a questionnaire about a student in their class, and in some cases, more than one student. I also appreciate the school boards on Vancouver Island and a local professional teachers' association, that agreed to teachers' participation.  People who are employed in our health systems were also helpful in the follow-up of subjects. Two public health nursing managers, Allison Cutler and Marg Wooton, as well as other public health nurses assisted by helping to notify subjects of our interest in contacting them. Among the public health nurses who contributed to the present study were Katie Hine and Norma Lupton, both involved as interviewers during the 1990-91 data collection. I appreciate the interest and effortfromall the public health nursing staff who contributed to this study.  Staff at the University of British Columbia and at the Ministry of Health were helpful in permitting the use of B.C.'s linked database to assist in locating subjects, after all other avenues of finding the 403 women of the 1990-1991 study had been exhausted. In particular, I would like to thank Dr. Allan Thomson, Ron Strohmaier and Al Cassidy. Forty subjects were found through the Linked Database, each one important to the study.  xi  I thank the 13 parents and 6 teachers who kindly agreed to pre-test questionnaires, all taking time from their already busy schedules.  The interest of staff and colleagues at two U.B.C. centres, the Institute of Health Promotion Research (DHPR) and the Centre for Community Health and Health Evaluation Research (CCHHER) has also contributed to the success of this project. Dr. Green, former Director of IHPR and Dr. Armstrong, Director of CCHHER, were supportive in offering space to carry out the work, generous with their time and patient in discussing questions raised by my research. Administrative and executive assistant staff, Mary Sun, Allison Bell and Lisa Lund, helped with the onerous tasks of paying the bills or scheduling meetings for a group with exceptionally full calendars. I thank each of them for helping to keep the project running smoothly.  Many others have contributed their particular skills to the project, thatfromtime to time needed very skilful people. Sherry Batdorf, Debra Cowan, Donna McNeill, Sherry Miyamoto, JoAnne Palin and Maureen Shiner served as excellent interviewers, interested in collecting the most reliable data. They drove to subjects' homes, often many miles from their own homes and difficult to find, to interview women and their children, always maintaining a professional attitude, good humour and patience. As well as interviewing, Sherry Miyamoto contributed to the study by spending many hours methodically and efficiently entering the data into database format. May Lee, and her supervisor Dr. Ying McNab, contributed their expertise as statisticians. In addition, six undergraduate students, Sara Sturrock, Jon Grenke, Colleen French, Meredith Burgon, Mary George and Marcia George, helped throughout the project with various tasks  xii  essential to the progress of the research.  I also thank two editors, Joy Davidson and Jane Srivastava.  Finally, I acknowledge and thank my two daughters Marcia and Mary George, my parents Edouard and Phyllis Gautschi and my sisters Jane Underwood and Marcia Rioux, all of whom were supportive and encouraging with their friendly urging and good humour.  xiii  Chapter I  Introduction  1.1 Introduction Maternal behaviour during pregnancy may contribute to unhealthy fetal and child development. This research examined two maternal behaviours, alcohol and tobacco use, during pregnancy. Previous research has linked heavy alcohol use with fetal alcohol syndrome and tobacco use with various health problems for neonates. This has concerned health professionals in Canada and elsewhere where alcohol and tobacco are commonly used by women. Health policy makers and practitioners have been responsive, particularly because of the preventable nature of theserisksto healthy child development.  It is known that excessive alcohol use during pregnancy can result in a range of disabilities and academic, health or behavioural problems (Abel, 1998b). The association between alcohol and adverse birth outcomes had been suspected for at least the past century, but until 1941, no attention had been given by the scientific community to links between environmental or lifestyle factors and malformations in human embryos (Naeye, 1992b). Our scientific knowledge regarding alcohol as a teratogen has developed only over the past 30 years after French (Lemoine et al., 1968) and American researchers (Jones & Smith, 1973; Jones et al., 1973; Ulleland, 1972) reported on patterns of malformations in infants born to mothers who drank excessively during pregnancy. Various case studies described children who were severely affected by alcohol (Armstrong, 1998) with little analytic control in the study design of other confounding risk behaviours or social environments. From these case and longitudinal studies following children severely affected, it is known that prenatal alcohol exposure can result in a spectrum of adverse  1  effects, with Fetal Alcohol Syndrome (FAS) being the most extreme. Fewer or less severe effects have been termed Fetal Alcohol Effects (FAE) or more recently termed "partial FAS". Children with FAE often do not have the distinguishing facial features of those with full FAS, but they may have a range of behavioural and cognitive effects, such as attention deficits or difficulty understanding concepts, that in rum can cause secondary effects of social and behavioural problems. FAS incidence estimates, derivedfromdatafroma number of countries, range from 0.33 to 0.97 per 1,000 live births (Single et al., 1999; Abel, 1995) for western countries.  Since those early reports in the late 1960's and early 1970's, over 6,000 journal articles have reported animal or behavioural research describing the prenatal effects of alcohol, with the cumulative evidence leaving little doubt that alcohol has teratogenic effects. These studies have broadened our understanding of the effects of alcohol exposure but much remains to be learned. Unique patterns of alcohol use (ranging in quantity and frequency), the stages of pregnancy during which women consume alcohol and the individuality of each mother's underlying health has made it difficult to complete the understanding of the effects of maternal alcohol use on humans. Because of the complex nature and the variety of the anomalies of alcohol-related birth defects, even diagnosis is not fully assured unless there is confirmation of the mother's use. Twelve longitudinal studies in the English-language journals have reported the follow-up of alcohol-exposed children beyond newborn stage, and many of these followed children with FAS. Few longitudinal studies have followed children prenatally exposed to alcohol, but not having FAS. Prior to the present study, there has been only one Canadian longitudinal study (Fried, O'Connell & Watkinson, 1992; Fried, Watkinson and Gray, 1992; Fried & Watkinson, 1990;  2  Fried & Makin, 1987; Fried & O'Connell, 1987). Longitudinal studies on the outcomes associated with prenatal alcohol or tobacco exposure are needed since many of the behaviours or learning problems do not become evident many years after birth. An inability to understand concepts, and ability to do arithmetic are two such outcomes often associated with children with FAS that are not noticeable until early school-age.  Smoking tobacco during pregnancy has detrimental effects on the health of the mother, fetus, newborn and child. Edwards, Sims-Jones and Hotz (1996) found in their review of the literature that most international surveys indicated prevalence rates of smoking by pregnant women ranged from 25% to 40%. A 1996-1997 Canadian national survey found that 21.3% of children under age 3 had mothers who reported that they smoked during pregnancy, with the rate being smaller in British Columbia (18.6%) (Canada. Health Canada, 2000a). Data collected during the home interviewsfromthe first stage of this present study (1990-1991) indicated that 37.5% of the women smoked at some time during their pregnancy.  Considerable evidence has shown the importance of socioeconomic inequalities in contributing to both child and adult poor health. Socioeconomic background was included in this study because the effects of the social and economic context in which children are raised may modify the effects of otherrisksincluding prenatal alcohol and tobacco exposure.  The study described in this thesis, referred to as the Vancouver Island Pregnancy Follow-up Study, examined the outcomes of children exposed to various levels of maternal alcohol and  3  tobacco use. It followed the children of an earlier study, the Vancouver Island Pregnancy Study (Armstrong et al., 2001a, 2001b, 1994) that was a population-based study testing the use of an alcohol screening tool by primary care practitioners and examining the prevalence of alcohol use amongst a general population. The purpose of the Vancouver Island Pregnancy Follow-up Study was to examine the outcomes of children exposed to various levels of maternal alcohol or tobacco use, or both, rather than study the specific sub-population of children with FAS. It was not a purpose of either study to determine the incidence of FAS or FAE in this population, nor was there any attempt in the Vancouver Island Pregnancy Follow-up Study to diagnose FAS or FAE.  Datafromthe first stage of this research (1990-1991 data collection) indicated that prior to pregnancy, 14.2% of women (n=3,059) were in a highriskcategory of alcohol use, 36.4% were in a mediumriskcategory (Armstrong et al., 2001a), as measured with the alcohol screening tool T-ACE scale (Sokol et al., 1989). Approximately one-half (51.7%) of the women drank alcohol at some time during their pregnancies. Bothfrequentdrinking of alcohol and binge drinking are considered to put the fetus atriskand datafromthis sample indicated that at least 6.5% of the women consumed five or more drinks during a day (considered binge drinking) on some occasions during pregnancy. These rates of use are considerably higher than those reported by a 1996-1997 national study, that found that 16.6% of children under the age of 3 years had mothers who reported that they drank alcohol during their pregnancies, and 14.9% of 3 year olds in British Columbia had mothers who reported drinking during their pregnancies (Canada. Health Canada, 2000a).  4  1.2 Definitions This study focused on prenatal exposure to alcohol and to tobacco, referring to maternal use of alcohol and tobacco while the mother was pregnant. Pre-conception exposure because of paternal use of alcohol was not a topic of this study, nor was prenatal exposure to smokefromthe fathers' or others' tobacco use. Unless otherwise noted, any reference to "prenatal alcohol exposure" or "prenatal tobacco exposure" refers to maternal use during pregnancy. Postnatal exposure to tobacco refers to maternal or others' use of tobacco, that is, second-hand or environmental exposure of the same child.  Both definitional and measurement problems exist with the terms "moderate", "social", "light" or "heavy" alcohol use, since the terms do not have common usage in the research literature. For the purposes of this research, maternal drinking behavior is defined accordingly: abstainers (no use throughout pregnancy), social drinking (one or fewer drinks per day), and heavy drinking (2 or more drinks per day and/or any occasions of binge drinking, defined as at least 5 drinks per occasion). Both the 2 drinks per day cut-off for infrequent drinking and the definition of 5 drinks per occasion for binge drinking were often, but not always, found in the research literature. No women in this study said they drank between 1 and 2 drinks, that is 1 V2 drinks, therefore, all drinkers fall into the categories of either social drinkers (1 or fewer drinks) or heavy drinkers (2 or more drinks). The literature review uses the term "moderate" drinking, since it is often used in research studies, although number of drinks changes from study to study. It is consistent that the term "moderate" falls somewhere between abstainer and heavy drinking. The actual amount of alcohol referred to as moderate is given if it is defined in the literature cited.  5  Chapter JJ  Literature Review  2.1 Introduction and methods The research literature on the topic of prenatal alcohol exposure is enormous, to the extent that the prolific writer Ernest L. Abel took 3 years to write his third book on the topic, rather than the planned one year because he "was unable to reconcile the many contradictions and new ideas that now characterize this literature" (Abel, 1998b, p. 1).  The purpose of the Vancouver Island Pregnancy Follow-up Study was to examine the outcomes of children exposed to theriskbehaviours of maternal alcohol or tobacco use, or both, rather than specifically study a sub-population of children with FAS. FAS, however, is included in the literature review because it informs the general topic of the effects of prenatal alcohol exposure. The present literature review focuses on the outcomes of prenatal alcohol exposure, prenatal tobacco exposure, and more briefly on socioeconomic background. It is selective, based on aspects most salient to the thesis, rather than a comprehensive review of all the effects of prenatal alcohol or tobacco exposure or on the associations between the socioeconomic gradient and health status.  With respect to prenatal alcohol exposure, the literature review for this thesis concentrates on studies describing the effects of a range of alcohol consumption, but emphasizes the impact of social or moderate alcohol consumption, rather than describing children with FAS or FAE. The majority of women in this research did not drink heavily and therefore the review focuses on the effects of moderate alcohol use. It focuses on human longitudinal studies. It includes longitudinal 6  studies following children with FAS or FAE because it was found that these informed the literature regarding the effects of prenatal alcohol use in general even though descriptions of children with FAS or FAE are not the focus of this thesis. Studies that were not longitudinal in design, but that inform the literature on outcomes of interest to this study, such as effects on academic performance, are included.  Systematic searches for pertinent literature were conducted using four electronic databases, MEDLINE, PsycINFO, Sociological Abstracts and ERIC. Because of the organizational structure of the particular database, and because of the relevance to the topics being searched, the date range differed slightly for each database. The date range searched for MEDLINE and for ERIC was from 1966 to present, for PsychlNFOfrom1967 to present and for Sociological Abstractsfrom1963 to present. The majority of the literature was found through MEDLINE searches. Keywords used in the searchers were: fetal alcohol, maternal smoking, prenatal alcohol, prenatal tobacco, maternal alcohol, pregnancy + alcohol use, pregnancy + tobacco use, cognitive, developmental, neuropsychological, socioeconomic, and SES. To ensure comprehensiveness of specific topics, searches were conducted when certain subjects were found to have less commonly used keywords (e.g., Conduct Disorder and tobacco). All longitudinal studies were included in the literature review for both the topics of prenatal alcohol exposure and for behavioural and effects of tobacco exposure. There was an attempt to be comprehensive in reporting on all longitudinal studies in the English-language literature on these topics. Cross-sectional studies were included only if there were pertinent to the topic and age-group of the study.  7  F o r reporting purposes, the review has been divided into two main topics, alcohol exposure and tobacco exposure. Section 2.2 covers the literature on prenatal alcohol exposure and Section 2.3 covers the literature on prenatal tobacco exposure, with each o f these sections being further categorized according to physiological effects, cognitive development or learning problems and behavioural outcomes. Section 2.4 covers literature on socioeconomic background, followed by Section 2.5, that briefly describes the literature on a few other potentially confounding variables. Section 2.6 summarizes the previous sections. Finally, this literature review includes two theoretical perspectives, described in Section 2.7, that are useful in describing the possible relationships between various variables that could influence the specified outcomes.  2.2 Prenatal effects o f alcohol Alcohol is a known teratogen, with the most devastating outcome being fetal alcohol syndrome ( F A S ) . F A S is manifested by a consistent pattern o f anomalies that includes (1) physical growth deficiencies either pre- and postnatally, (2) observable facial anomalies (such as flattened philtrum, thin upper lip, specific disproportional eye positioning and narrow eye openings), and (3) central nervous system involvement (such as learning disability or behavioural problems). Fetal alcohol effects ( F A E ) is the term used when fewer than three o f these components are present. F A E often goes undiagnosed. Children who do not have the obvious features o f F A S can have troublesome lives because the cause o f outcomes such as low academic performance or behavioural problems go undiagnosed or mis-diagnosed. Some children have alcohol-related defects, but their conditions do not warrant a diagnosis o f either F A S or F A E .  8  Since the first reporting of the anomalies associated with prenatal alcohol exposure were published in the scientific literature in France (Lemoine et al., 1968) and in the U.S.A. (Jones & Smith, 1973), there have been repeated observations of the teratogenic effects of alcohol (Abel, 1998b; Abel & Hannigan, 1995a; Coles, 1994; Michaelis & Michaelis, 1994; Ernhart, 1991; Streissguth, 1991; Streissguth et al., 1989c; Coles et al., 1987; Ernhart et al., 1987; Little et al., 1986; Sokol et al., 1986; Streissguth et al., 1984; Rosett et al., 1983; Clarren, 1981; Clarren & Smith, 1978). Longitudinal studies have been conducted in Germany, Sweden, Canada and in at leastfiveU.S.A. locations (Atlanta, Cleveland, Detroit, Pittsburgh and Seattle), with some of these following the children longitudinally for more than 2 decades. The studies can be divided into two categories, those attempting to describe the physical, behavioural, cognitive and social characteristics of people with FAS/FAE during various stages of life and those describing children who were exposed prenatally to moderate or lighter amounts of alcohol but rather than having FAS or FAE, the children were born with a range of outcomes.  For the most part, the results have been inconclusive, partially because the relatively short time in which research has been conducted (that is, approximately 30 years) has left little time for replication of work, particularly in the case of longitudinal studies. Methods and the inclusion of confounding variables have varied considerably, as have the measurements of both the independent variable of interest (alcohol consumption during pregnancy) and outcome variables.  Only one Canadian longitudinal study was found in the published literature. It described children whose mothers consumed light or moderate amounts of alcohol (Fried et al., 1992a, 1992b; Fried 9  & Watkinson, 1990; Fried & O'Connell, 1987; Fried & Makin, 1987). The two longest running studies have been conducted in Germany and in Seattle. In Seattle, the research team has conducted two studies, one describing children with FAS or FAE, and the other examining the effects of those prenatally exposed to alcohol but not diagnosed as having FAS or FAE. The first study has followed a cohort of 92 subjects with FAS referred to a clinic in Seattle, or to the clinic of Dr. D.W. Smith in Vancouver (Streissguth et al., 1989c). The other Seattle study began as a population-based study, comparing children exposed to distinctly varying levels of exposure, but not necessarily diagnosed with full FAS. The German longitudinal study has followed people with FAS or FAE in the former West Germany since 1977 (Steinhausen, 1996; Spohr, 1996; Spohr & Steinhausen, 1996, 1987, 1984; Spohr et a., 1993). Subjects (n=158) who were the children of alcoholics, entered the study at various agesfromnewborn to age 11.5 years.  There have been numerous studies attempting to explain mechanisms underlying the teratogenity of alcohol (for example, see Schenker et al., 1990). Except for a brief discussion here, most of these will not be included in this review of the research. Animal research examining alcohol as a teratogen has focused primarily on the five general areas of placental dysfunction, nutritional deficiency, acetaldehyde (the primary metabolite of alcohol) toxicity, the role of prostaglandins (Randall et al., 1989; U.S. Department of Health and Human Services, 1990) and fetal hypoxia. Each of these has been observed or suggested as an etiological factor, but inconsistencies in the research results make a single mechanism unlikely. For example, placental dysfunction may have a role in fetal outcome but is unlikely to be the sole cause of pathophysiology, since incubated embryo exposed to alcohol also have been found to be affected (Pennington et al., 1983). 10  Maternal nutritional deficiency, often observed amongst alcoholics, has been studied as a possible mechanism or competing cause. The research in this area has been inconclusive, since the administration of supplements has been found to improve pregnancy outcomes in some studies but not in others (Weinberg, 1985; Rush et al., 1980; U.S. Department of Health and Human Services. NIAAA, 1990). With respect to the role of prostaglandins in alcohol teratogenesis, studies have shown a reduction in alcohol-induced fetal malformations by using a prostaglandin synthesis inhibitor (such as aspirin), but the inhibitor has not completely blocked the effects. Each of these mechanisms has been observed, although none appears to be the sole causal mechanism, so it is possible that either collectively they play a role, or that each manifests different features or outcomes.  2.2.1 Physical and health effects As noted above, alcohol has been documented as a teratogen with characteristic physical anomalies of heavy exposure being growth retardation and facial dysmorphia and often major organ damage. Longitudinal studies have shown that the craniofacial dysmorphology may not be obvious in the newborn and the most severe of these characteristics are reduced as the child grows (Spohr, 1996). Most studies describing children with FAS are case studies of offspring of alcoholic mothers, including binge drinkers (Abel, 1998b; Armstrong, 1998).  Moderate use of alcohol during pregnancy has not been associated with the same physical outcomes. A meta-analysis (Polygenis et al., 1998) was completed in an attempt to determine whether first trimester moderate alcohol consumption was associated with fetal malformations.  11  Birth outcomes were compared between those exposed to moderate drinking levels (defined in that analysis as ^2 drinks a week) and the offspring of abstainers (defined as less than 2 drinks a week). Although 24 studies met the inclusion criteria, only 7 had data that were extractable. Moderate alcohol consumption during the first trimester was not found to increase the risk of fetal malformations (Polygenis et al., 1998).  Few systematic studies have attempted to examine the extent of physical damage associated with moderate levels of drinking, at least beyond neonatal stages, except for studies examining the long-term effects on physical size. Results of studies on size at birth have been inconsistent, with some studies finding a statistically significant association between prenatal alcohol consumption and growth deficiencies (Plant, 1985; Barr et al., 1984) and others not observing such an association (Forrest et al., 1991; Larsson et al., 1985). Abel and Hannigan (1995b) re-analyzed the results of a prospective study and found a J-shaped relationship between alcohol use and birth weight. Mothers who abstained had smaller babies than those who were light drinkers, but the mothers who drank most heavily had children with the lowest mean birth weights. Lundsberg, Bracken and Saftlas (1997) argued that birth weight, being a mix of preterm birth and retarded growth, has been an inadequate outcome measure because low to moderate alcohol use may increase the risk of preterm delivery in spite of its apparent lack of association with low birth weight.  Results have been inconclusive with respect to long-term physical growth of young children who had been prenatally exposed to alcohol. Heavy alcohol intake has been associated with slower 12  growth (height or birth weight) pre- and postnatally. Swedish researchers (Aronson & Hagberg, 1998; Aronson et al., 1985; Kyllerman et al., 1985, 1983) found continued growth deficiency among children whose mothers were alcoholic during pregnancy, with children with FAS at age 5 years being leaner and shorter with no catch-up growth indicated. Coles et al. (1991) found that when those subjects with dysmorphia were removedfromthe sample, the growth deficits in others prenatally exposed to alcohol were not statistically significantly different than those not prenatally exposed. The Seattle studies found that children diagnosed with FAS, and who were dysmorphic, continued to be growth deficient (Streissguth et al., 1985a) and that moderate alcohol exposure was related to growth deficiency initially but diminished with age during the infancy period (Barr et al., 1984). In the Seattle longitudinal study of children not diagnosed with FAS, physical size at age 7.5 years was not found to be associated with prenatal alcohol exposure (Sampson et al., 1994; Streissguth et al., 1990). This was consistent with thefindingof a longitudinal study in Ottawa, in which children prenatally exposed to two drinks per day caught up in size at 12 or 24 months, in spite of being small-for-age earlier (Fried & O'Connell, 1987). Both the Seattle study and the Ottawa study had samples of predominately middle class mothers. In a comprehensive study with respect to size or growth of young children, Greene et al. (1991b) found that weight and height at birth were inversely related to drinking during pregnancy. Self-reports of 2 or more drinks per week throughout pregnancy were found to decrease birth weight by 76 grams. A small but statistically significant association with length was found for the same consumption level only during thefirsttrimester. As the children aged, some catch-up in weight and height was observed. Because their results differed according to whether alcohol use was measured using a screening tool or by pre- or post-natal recall of alcohol use, the researchers summarized their  13  findings as being "inconclusive" (Greene et al., 1991b, p. 911). These findings were inconsistent with studies in Pittsburgh (Day et al., 1999, 1993, 1991), Buffalo (Russell et al., 1991) and France (Larroque & Kaminski, 1998) that found prenatal alcohol exposure to have significant negative effects on height at preschool or school ages.  Discriminating between dysmorphic and non-dysmorphic children may clarify the results. It appears that those studies that included children born dysmorphic showed that these children continued to be small for age throughout childhood, while the results for studies that included non-dysmorphic children did not show continued growth deficiency by school-age. In addition, differences in characteristics of the sample and in the timing of the alcohol use may explain at least some of the inconsistency (Day et al., 1999). In particular, Day et al. (1993) noted that persistent growth deficits have been found in low income populations, whereas those samples with more socially and economically advantaged subjects found that their cohorts had caught up in growth over time. Barrison and Wright (1984) found that the interactive effects of moderate prenatal alcohol use, particularly around the time of conception, and prenatal tobacco use and lower social class background, influenced low birth weight.  Besides physical growth retardation, other physical effects have been studied for their associations with prenatal alcohol exposure. Visual and auditory impairments have been so commonly observed in children with FAS or FAE that it has been suggested that ophthalmological and auditory examinations be routinely given for those who have been diagnosed with FAS (Abel, 1998b; Stromland, 1985). Children with FAS have often required corrective lenses, usually  14  resultingfromabnormal curvature of the lens (Abel, 1998b). At least three types of hearing loss have been associated with children with FAS (Abel, 1998b), including conductive (resulting from middle ear problems and occurring in more than 75% of children with FAS), sensorineural (resultingfromauditory nerve damage and found in «28% of children with FAS) and central hearing loss, foundfrequentlyin children with FAS (Church et al., 1997b). The incidence of central hearing dysfunction is unknown, but Church et al. (1997b) found in a study of 12 children with FAS, all showed clinically significant central hearing impairments.  No research literature was found regarding the general health status of elementary school-aged children who were prenatally exposed to alcohol. Many specific health problems have been described in the case study reports of children with FAS or FAE. For example, Johnson et al. (1981) observed an increased incidence of life-threatening bacterial infections as well as a propensity for minor infections in their review of 13 documented cases of FAS. Five of 13 patients had at least one episode of pneumonia, two had meningitis and one had sepsis. Following a comprehensive immunologic evaluation of the children with FAS and comparing the results with an age-matched control group of children with intrauterine growth retardation without FAS, it was suggested by the researchers that an impairment of the immune system may explain an increased susceptibility to infection.  Vorhees (1986) hypothesized an ordering of teratogenic outcomes related to alcohol, with dysmorphological anomalies being the most sensitive to exposure, followed by birth weight and lastly cognitive outcomes. It has been suggested that behaviour may be more sensitive than  15  physical features as an indicator of prenatal alcohol exposure (Mattson & Riley, 1998; Landesman-Dwyer et al, 1981; Shaywitz et al., 1980). Two of the longest longitudinal studies have been conducted in Seattle and after many years of research on children with FAS or FAE as well as those moderately affected by prenatal alcohol exposure, the researchers concluded that neither physical growth nor facial morphology were as sensitive markers as the broad range of neurobehavioural outcomes (Streissguth et al., 1994a). Timing and dose of alcohol intake also have been found to be important, with some critical periods important to the development of different organs.  2.2.2 Cognitive and behavioural effects of children prenatally exposed to alcohol Both cognitive development and behavioural problems of children prenatally exposed to alcohol have been well documented in the research literature over the past 30 years. Although many researchers have described learning problems and behavioural effects in the same studies, these two outcomes are treated separately in this literature review. The studies cited in this section are listed in Table 2.1, pages 44-48.  2.2.2.i Cognitive development, academic abilities or learning problems The research literature consistently describes associations between FAS and lower IQ. scores. Studies describing small cohorts of children with FAS have consistently included language acquisition and expressive and receptive language deficits among the cognitive developmental problems. As noted above, various types of hearing loss can resultfromprenatal alcohol exposure. Hearing loss in preschool and school-aged children could adversely affect language,  16  speech and intellectual development and cause distractibility and hyperactivity (Church et al., 1997b).  The Seattle longitudinal study that followed children with FAS found that the degree of intellectual impairment and the severity of the mother's alcoholism were correlated. At follow-up, home environment was not found as a mediating factor, as the intellectual impairment was sustained regardless of home conditions (Streissguth et al., 1985a). Testing at later ages and comparing with non-alcohol-exposed and minimally alcohol-exposed adolescents with similar IQ. scores showed that those with FAS had more difficulties in processing speed and accuracy, declarative learning and cognitive flexibility and planning (Carmichael Olson et al., 1998).  A German longitudinal study (Spohr, 1996; Steinhausen, 1996; Spohr and Steinhausen, 1996, 1987, 1984; Steinhausen et al., 1994, 1993; Spohr et al., 1993) tested children (n=68) at preschool ages, early school age (6-12 years) and at late school age (^ 13 years) using a battery of psychiatric, behavioural and psychological assessments and intelligence tests including the Wechsler Intelligence Scale for Children-Revised (WISC-R). They found that early school-aged children severely affected by prenatal alcohol exposure were all well below normal on intelligence tests with a large proportion of the sample classified as being mentally retarded. These effects persisted over time, in contrast to dysmorphia, that reduced over time. At the latest testing (ages 12-14 years), 17 of the children attended specialized schools and the majority of children had difficulties in mathematics, logical conclusions, visual perception, spatial relations and short-term memory.  17  A Swedish study followed 21 children born to alcoholic mothers,frombirth to age 14 years, comparing them to controls matched by age, sex, birth weight, gestational age and living area (Aronson & Hagberg, 1998; Aronson et al., 1985; Kyllerman et al., 1985; Olegard et al., 1979). Ten of the 21 children had FAS characteristics. Children with the characteristics of FAS had statistically significantly lower IQ. scores, as measured on the WISC-R, than either the controls or those in the study group without signs of FAS. There was no significant difference in IQ. scores between those reared in foster parents' homes and in biological parents' homes.  These three longitudinal studies in Seattle, Germany and Sweden were consistent in their observations of low I.Q. scores amongst children with FAS. In the following cross-sectional studies, specific variables, such as outcomes on memory, or the effects of timing of alcohol exposure according to trimester of exposure, were studied.  Coles et al. (1991) compared the 5-year old offspring of women who continued to drink throughout pregnancy with children whose mothers stopped drinking during the second trimester of pregnancy and with a control group whose mothers never drank during pregnancy. The sample was drawnfroma predominantly black urban population. The children of those who continued to drink showed significant deficits in areas of intellectual functioning such as sequential processing (short-term memory and encoding) and mental processing. Using the Kaufman Assessment Battery for Children (K-ABC) for assessment, both prenatally alcohol exposed groups showed deficiencies in pre-math and reading skills, compared to those whose mothers never drank.  18  Mattson et al. (1996) tested verbal learning and memory by comparing children diagnosed with FAS (n=20) to a control group matched for mental age and to another control group of the same physical age. Assessing the children with the California Verbal Learning Test-Children's Version (CELT-C) to test immediate and delayed recall and recognition memory on verbal tasks, they found that the children with FAS had profound verbal learning and memory deficits in comparison to either control group. Children with FAS learned fewer words indicating difficulty with immediate recall, had more difficulty with delayed recall and made more errors of various kinds in the tasks. They were less able to remember verbal information they had learned, indicating that memory may be more impaired in terms of encoding rather than retrieval. The researchers suggested two possible explanations, that the encoding deficit may be related to attentional or impulsivity factors or that this deficit in encoding, in the absence of deficit in retrieval, is unique to children with FAS.  In studies in San Diego, Mattson et al. (1999, 1998, 1997; Mattson & Riley, 1998) compared three groups of children aged 5 to 16 years, those who had been diagnosed with FAS, those whose mothers were alcoholic during pregnancy but who had not been diagnosed with FAS because of the absence of physical features and matched controls. Both the alcohol-exposed groups showed statistically significantly lower IQ. scores than the control group both on the overall scores and on the subscales although the non-dysmorphic group had slightly higher scores than those with FAS. It was also observed that the two alcohol-exposed groups were more impaired, relative to the controls, on tests of language, verbal learning and memory, academic skills,fine-motorspeed and visual-motor integration. Because of this observed consistent pattern 19  of these neurological deficits in the two alcohol-exposed groups, the authors concluded that the degree of deficit existed independent of the physical anomalies of FAS. Compared with the control group, the alcohol-exposed groups were impaired on some, but not all of the memory tasks, implying that they were not globally impaired. They showed impairment on verbal fluency but not on implicit memory tasks (tests of unconscious recall of previously performed tasks). These results on memory are consistent with two other studies that found a relationship between prenatal alcohol exposure and deficits in either or both verbal and non-verbal memory (Uecker & Nadel, 1996; Streissguth et al., 1994b, 1998).  In another cross-sectional study, the neuropsychological performance of 10 preschool-aged children with FAS in Saskatchewan was compared to that of a control group. The children with FAS displayed deficits in verbal, performance, language and motor skills, but no deficits were displayed in visual perception, quantitative or memory skills. The failure to find statistically significant differences between the control group and the group diagnosed with FAS on both quantitative skills and on memory skills differsfromother studies (Jansen et al., 1995).  In one of the few studies in which various intellectual and neuropsychological tests were administered to a cohort of British Columbian children, Conry (1990) compared a sample of 19 aboriginal school-aged children diagnosed with FAS or FAE with controls matched for age and gender. The children, aged 6.4 to 18.5 years, lived in an isolated, economically depressed community. Scores on intelligence tests were significantly lower for the children with FAS compared with FAE and with the controls. The children with FAS also had significantly lower  20  test results on most neuropsychological tests, including reaction time, ringer tapping, grip strength, finger localization and motor speed and precision, while those with FAE showed significantly lower scores only on grip strength.  These longitudinal and cross-sectional studies and others (for example, Smith & Eckardt, 1991; Iosub et al., 1981; Majewski, 1981) all observed considerably lower I.Q. scores for children with FAS and most observed problems with memory and with verbal learning. Some of the studies observed the same effects for children with FAE. Although all indicated lower scores on the tests of intelligence, there has been a lack of consistency on the level and areas of deficits and some of the exposed children tested within the normal range. For example, although the German cohort of children of alcoholic mothers scored lower than normal, in fact, 24 (34.3%) of the children scored in the normal range on the I.Q. test, another 34% scored in the borderline range and 31% scored with varying degrees of mental retardation (Steinhausen, 1996). In his comparison of various studies of I.Q. scores amongst children diagnosed with FAS, Steinhausen (1996) observed a wide range amongst the scores of the various cohorts, so that, for example the proportion with below 70 I.Q. scores ranged from 19% to 60% amongst 8 studies. Although mental retardation has been observed to be one of the most consistent characteristics of FAS, it is not an invariable outcome, one of the possible reasons being that "none of [the] studies was veryrigorousin design with respect to their criteria for identifying FAS or eliminating other causes" (Abel, 1998b, p. 128). Even the morerigorousstudies differ on psychometric tests used for assessment, the country of the children, living circumstances of children (for example, isolated B.C. community, foster homes, schools for the mentally retarded) and in the ages at which children were tested,  21  making general conclusions difficult.  Each of the studies cited above has contributed to the description of characteristics of schoolaged children diagnosed with FAS/FAE or who had alcoholic mothers. More recently, researchers have become interested in testing children who were exposed to more moderate levels of alcohol exposure. The results have been inconsistent. One study found improved performance related to pregnancy and pre-pregnancy maternal alcohol consumption (Forrest et al., 1991). Fried et al. (1992b) and Greene et al. (1990, 1991a) found no adverse outcomes on IQ. assessments when testing 5 to 6 year olds who had been exposed prenatally to moderate amounts of alcohol. Jacobson et al. (1998, 1993a, 1993b; Jacobson & Jacobson, 1994) found a small amount of the variance in infants' I.Q. scores could be explained by prenatal alcohol exposure. Larroque and Kaminski (1998) found lower scores on tests of cognitive development after exposure to at least three drinks per week. These studies are described in more detail below.  The only study that found a positive outcome of alcohol exposure during and prior to pregnancy was completed in Scotland and assessed 18-month old children (n=592). After controlling for confounding factors, it was observed that for full-term babies, those whose mothers drank during pregnancy had better performance scores on the Bayley scale (Forrest et al., 1991). Alcohol consumption did not exceed 100 grams absolute alcohol. This is of interest given the J-shaped relationship that Abel and Hannigan (1995b) found between low birth weight and prenatal alcohol exposure, as noted above.  22  The Seattle Longitudinal Prospective Study on Alcohol and Pregnancy (Sampson et al., 1989; Smith & Eckardt, 1991; Streissguth, 1994b; Streissguth et al., 1994a, 1994b, 1993, 1990, 1989a, 1989b, 1989c, 1989d, 1986a, 1981) has been one of the enduring studies with repeated follow-up of babies born during a one-year period, 1975-1976 . It began with prenatal interviews of pregnant women who were receiving prenatal care at two hospitals. The women (n=1,529), who were predominantly white, married, middle-class women, were interviewed about their use of alcohol, tobacco, caffeine and other drugs. Subsequently, a cohort of approximately 250 heavier drinkers and smokers and 250 non-smokers who abstained during pregnancy or drank infrequently were selected from the original sample for follow-up studies. The highriskgroup (mothers who drank more heavily and smoked) was over-sampled and the lowriskdrinking group was stratified by tobacco use. Follow-up assessments of the children were conducted at ages 1, 8 and 18 months and 4, 7 and 14 years. Most relevant to this thesis were the results of the Seattle data collection at age 7.5 years (Streissguth et al., 1994a; Streissguth et al., 1993; Streissguth et al., 1990). Data were collected using a battery of tests, including I.Q. and achievement tests, neuropsychological tests, memory, reaction and attention time, and physical measurement. In addition, both teachers and parents rated health, ability to socialize and other behaviours. Assessing with the Wechsler Intelligence Scale for Children-Revised (WISC-R), it was observed that children exposed to daily alcohol use during the middle trimester had mean I.Q. scores 7 points lower than the controls and that the effect of alcohol was exacerbated by lower father education and by a larger number of children in the family. Binge drinking prior to pregnancy had the greatest influence on academic achievement at age 7.5, compared to binge drinking during pregnancy or with frequent drinking prior to or during pregnancy. Children whose mothers binge drank (* 5  23  drinks on one occasion) in the month prior to pregnancy had deficiencies in arithmetic and reading, but not spelling, when tested using the WRAT-R. Decrements of scores of 1-3 months in the arithmetic and reading subscale were associated with binge drinking. No statistical evidence of threshold level of prenatal drinking was observed and the effects could not be explained by any of the 150 covariates included in the analyses, including parents' education, maternal smoking or nutrition during pregnancy (Streissguth et al., 1993). At age 11 years, arithmetic scores were the most affected by prenatal heavy alcohol exposure, and the low arithmetic scores did not improve with age as the testing at age 14 showed (Streissguth et al., 1994a, 1994b).  Another U.S.A. study (Jacobson et al., 1998, 1996, 1993a; Jacobson and Jacobson, 1994) examined dose-response effects of prenatal alcohol exposure, since the results of previous studies had been inconsistent. African-American pregnant women attending prenatal clinics in inner-city Detroit were recruited between 1986 and 1989. The researchers included an over-representation of those drinking at moderate to heavy levels. Prenatal exposure to binge drinking (*5 drinks per occasion on an average of at least once per week) more than doubled theriskof obtaining low scores on cognitive development assessments at ages 6.5, 12 and 13 months. The researchers observed a dose-response relationship for prenatal alcohol exposure when testing at 13 months on the Bayley MDI scale for prenatal alcohol exposure, but did not find a dose-response relationship with scores on the Psychomotor Development Index (Jacobson et al., 1993a). Below the threshold level of 4 drinks per day, little effect of prenatal alcohol exposure was observed. The researchers controlled for numerous demographic, and prenatal, neonatal and postnatal variables (including use of tobacco, marijuana and other drugs during pregnancy) and concluded that: "As  24  with previous studies, after control for potential confounders, only 1 to 2 % of the variance in infant outcome was uniquely attributable to prenatal alcohol exposure" (Jacobson et al., 1993a, p. 182). When the analyses were performed for a higherriskgroup (offspring of women over 30 years and higher alcohol exposure), prenatal exposure explained up to 7% of the variance in developmental outcome. Further analyses were conducted taking into account outliers and potential problems with summary measures (for example, one mother drank daily while the majority drank no more than two times per week). They observed functionally significant deficits in cognitive development in children diagnosed with FAS as well as in the offspring whose mothers drank intermittently but heavily.  The results of that Detroit study were consistent with those of a Pittsburgh study (Russell et al., 1991), although testing was completed at 6 years of age for the latter research. Children (n=175) whose mothers had more than one indication of problem drinking were compared with children whose mothers reported fewer indications of problem drinking on a questionnaire selfadministered during pregnancy. Statistically significant lower scores on tests of verbal and language scales (measured by Wechsler Preschool and Primary Scale of Intelligence and the Token Test) were observed amongst those whose mothers had more indications of problem drinking, although no relationship was observed between prenatal alcohol exposure and performance on a psychomotor test.  The only Canadian longitudinal study found in the literature is the Ottawa Prenatal Prospective Study (OPPS) that began in 1978 and includes women (n=250) who volunteered to participate 25  after media solicitation (Fried et al., 1999, 1997, 1992a, 1992b; McCartney et al., 1994; Makin et al., 1991; Fried, 1989a; Kristjansson et al., 1989; Fried & Makin, 1987; Fried & O'Connell, 1987). The researchers observed the children of predominately middle-class, lowriskwomen who were prenatally exposed to alcohol, cigarettes and marijuana. The cohort was assessed repeatedlyfromages 1 to 12 years. At ages 2 and 3 years of age, decreased language comprehension was observed, but at ages 5 to 6 years, there were no statistically significant differences in general cognitive development or language between the moderately exposed children and those not exposed. The OPPS researchers suggested that their failure to confirm a relationship between prenatal alcohol exposure and cognitive outcomes at the later ages may be because their sample had relatively lower levels of consumption compared to other studies. Women averaged less than one drink a day with only four of their subjects drinking more than 3 drinks per day.  A study of the children of Cleveland women who drank moderately during pregnancy had similar results to the OPPS study, although the backgrounds of the populations of the two studies differed considerably (Greene et al., 1991a, 1990; Sokol et al., 1981). The Cleveland women (n=359) were disadvantaged socioeconomically and were recruited over a 33-month period when presenting at a general Cleveland hospital. Their offspring were tested at various ages (6 months, 1, 2, 3 years and 4 years 10 months) using age-appropriate assessment tools to measure cognitive development, including the Bayley Scale, Stanford-Binet Intelligence Scale and the Wechsler WPPSI. The researchers used several indices to measure maternal alcohol use, assessed the children at five separate times and applied multiple analyses, but were unable to find an inverse 26  relationship between prenatal exposure and cognitive development tests when the only child with FAS was removedfromthe analyses. Prenatal exposure to moderate alcohol use was not found to have a statistically significant relationship with lower language scores or with general cognitive measures, and they found no evidence of adverse effects on cognitive development in the absence of fetal alcohol syndrome. Interesting, they did find a significant relationship between prenatal alcohol use and birth weight, supporting Vorhees' (1986) hypothesis, noted above, that the ordering of teratogenic responses to alcohol means that birth weight is a more sensitive outcome than cognitive development. They concluded that home environment, even within a predominantly disadvantaged cohort, was more predictive of lower cognitive development scores than either prenatal alcohol or tobacco exposure (Greene et al., 1991a).  Another attempt to study the effects of moderate prenatal alcohol exposure was conducted with 4V2 year olds in France. Prenatal exposure of three drinks per week (1.5 oz. of absolute alcohol) was significantly related to lower scores on the McCarthy scales of cognitive development, after controlling for potentially confounding variables (Larroque & Kaminski, 1998).  The French results were consistent with the Seattle study but inconsistent with the Ottawa and Cleveland longitudinal studies cited above where children were tested at preschool or early school age with respect to the relationship between exposure and cognitive development. The number of heavy drinkers in the Seattle and French studies and rates of attrition at follow-up may have contributed to inconsistency among the four studies. In summary, rates of consumption of 2:1.5 oz. or more of absolute alcohol per day have been found to be associated with general cognitive  27  function amongst infants and young children (Larroque & Kaminski, 1998; Jacobson et al., 1996, 1993a, 1993b; Jacobson & Jacobson, 1994), Below an exposure level of 1.5 oz. per day, two studies found no adverse outcomes with respect to cognitive development (Greene et al., 1991a; Fried & Watkinson, 1990) while the Seattle study observed adverse outcomes (Streissguth et al., 1989a, 1998b).  Goldschmidt et al. (1996) observed a dose-response relationship between prenatal alcohol exposure and academic abilities in one of the few prospective, longitudinal studies examining differences in children whose mothers drank at varying levels during pregnancy. There was a continuum of alcohol use among the mothers (n=595), although most were light or moderate drinkers. All were from lower-socioeconomic backgrounds and were high school educated. At the average age of 6.5 years, the academic abilities of children who were exposed to alcohol during the second trimester of pregnancy were tested, using the Wide Range Achievement TestRevised (WRAT-R) to assess arithmetic, reading and spelling skills. After controlling for I.Q. scores, significant deficits in achievement were predicted by prenatal alcohol exposure. The results showed a dose-related relationship for the arithmetic subtest, and in contrast, a threshold effect for the relationship between alcohol exposure at the second trimester and reading and spelling achievements. The threshold occurred at ~ 1 drink per day in the second trimester. This study brought new observations regarding the effects of prenatal alcohol exposure on academic abilities, perhaps explaining the inconsistency with previous observations in studies that either grouped varying levels of exposure or were not specific in differentiating the type of effect according to academic subject.  28  To summarize, a number of longitudinal studies in U.S.A. cities and in Sweden, West Germany and Ottawa, as well as cross-sectional studies have been conducted to examine associations between prenatal alcohol exposure and general cognitive development or specific learning deficits. Most of these studies (e.g., in Germany, Seattle, Sweden and San Diego) found that children diagnosed with FAS have statistically significantly lower I.Q. scores than others prenatally exposed to alcohol but without the FAS diagnosis and than those not exposed to alcohol. Studies have shown that even some of children diagnosed with FAS have I.Q. scores within the normal range.  Many studies have observed that children heavily exposed to alcohol have deficits in memory (Mattson et al. 1999, 1998, 1997; Mattson & Riley, 1998; Mattson 1996; Spohr & Steinhausen, 1996, 1987, 1984; Spohr et al. 1993; Smith & Eckardt, 1991; Coles et al., 1991; Iosub et al., 1981; Majewski, 1981). In contrast, one study (Jansen et al., 1995) found no deficit on memory skills. Children heavily exposed to alcohol have been observed to have deficits in specific abilities, including reading abilities (Goldschmidt et al., 1996; Coles et al., 1991), math or arithmetic (Goldschmidt et al., 1996; Kopera-Frye et al., 1996; Spohr & Steinhausen, 1996, 1987, 1984; Spohr et al. 1993; Coles et al., 1991; Streissguth et al., 1990), verbal learning (Mattson, et al. 1999, 1998, 1997; Mattson & Riley, 1998; Goldschmidt et al., 1996; Uecker & Nadel, 1996, Streissguth et al., 1994, 1998; Russell et al., 1991) and reading skills (Goldschmidt et al., 1996; Coles etal., 1991).  Two studies identified the impact of binge drinking on I.Q. scores or cognitive development 29  assessments (Jacobson et al., 1998, 1996, 1993a, Jacobson & Jacobson, 1994; Streissguth et al., 1990) including binge drinking that occurred just prior to pregnancy (Streissguth et al., 1990).  Studies examining children exposed to moderate alcohol use differ in that two of the three published studies found no adverse outcomes. The Cleveland longitudinal study (Greene et al., 1991a, 1990) drew its samplefroma socially disadvantaged U.S. urban setting and the Ottawa longitudinal study drew its samplefroma middle class Canadian sample (Fried et al., 1999, 1997, 1992a, 1992b; McCartney et al., 1994; Makin et al., 1991; Fried, 1989a; Kristjansson et al., 1989; Fried & Makin, 1987; Fried & O'Connell, 1987), but neither was able to find statistically significant differences between those exposed to moderate alcohol use those not exposed groups in general cognitive development or language skills. The results of the Seattle study, that had a sample similar to the Ottawa sample, found statistically significant associations between moderate alcohol exposure and test results of cognitive development.  Methodological issues, levels of alcohol exposure, or differences in populations that have not been controlled may contribute to the inconsistency of results. Children of alcoholics, heavy drinkers (that is, more than three drinks per day) and binge drinkers are more likely to have lower scores on assessments of academic abilities. Children prenatally exposed to light or moderate drinking are more likely than those prenatally exposed to heavy drinking to have test scores similar to those of children not exposed.  30  2.2.2.H Behavioural effects of prenatal alcohol exposure In spite of nearly 30 years passing since FAS was first described, we have had to rely, for the most part, on descriptive and case studies for our understanding of human behaviour related to FAS (Abel, 1998b; Armstrong, 1998). Many of the longitudinal studies cited above that followed childrenfromthe neonatal stage through to school age attempted to describe both learning difficulties and behavioural problems. As with the review above on learning abilities or cognitive development, the studies are divided for the purposes of this review according to those focussing on children with FAS or FAE and those focussing on behavioural problems of people prenatally exposed to any amounts of alcohol. The mostfrequentlyreported, and specifically studied, behavioural characteristics of young children with FAS and FAE are attention deficits, hyperactivity and impulsivity (Abel, 1998b), cited as the "hallmark features of prenatal alcohol syndrome" (Mattson & Riley, 1998, p. 287).  Results have been inconsistentfromstudies of infants tested for irritability, hypersensitivity or poor habituation, in spite of the use of the same neonatal assessment tool (Brazelton Neonatal Behavioral Assessment Scale) being used. Abel (1998b) reviewed the research literature on neonatal behavioural outcomes, finding eight studies, of which most found no statistically significant effect of prenatal alcohol exposure on any of these three outcomes. Of the eight studies, only one study observed an effect on habituation (Streissguth et al., 1983), two studies observed an effect on lower arousal (Jacobson et al., 1984; Streissguth et al., 1983) and two studies observed an effect on orientation (Richardson et al., 1989; Coles et al., 1985). Measurement problems may have been the basis of the inconsistencies. In particular, 31  measurement of behavioural outcomes have often relied on subjective observations and the measurement of timing, duration and quantity of alcohol consumption has relied on self-reports since no biological markers exist (Abel, 1998b). Only some of these measurement problems are overcome in testing older children for behavioural problems, since more sensitive testing procedures are available for older children and other measures of alcohol consumption have been used in a limited number of studies.  Assessments of children beyond neonatal age are more relevant to the present thesis. The literature found to be consistent with respect to observing attention deficits or hyperactivity for children diagnosed with FAS, but inconsistent with respect to those effects in children without the characteristics of FAS but who were prenatally exposed to alcohol. Longitudinal studies were conducted by researchers in Germany, Sweden, Ottawa and various U.S. cities, including Atlanta, Cleveland, Detroit, Seattle and Pittsburgh. Cross-sectional studies also were conducted comparing children with FAS to other groups to determine differences in attentional deficits.  In a Seattle longitudinal study, in which 92 people diagnosed with FAS were assessed, it was observed that by preschool age behavioural problems included hyperactivity, inattentiveness and impulsivity. Over half (58%) of the subjects were classified as having maladaptive behaviour. Using the Vineland Adaptive Behavior Scales, specific behaviours found to be prevalent amongst the subjects included poor concentration and attention (77%), social withdrawal (62%), impulsivity (57%), dependency (53%), teasing/bullying (53%), having periods of high anxiety (51%), and stubbornness or sullenness (50%) (Streissguth et al., 1989c).  32  The German longitudinal study described above (Steinhausen, 1996; Steinhausen et al., 1994, 1993; Spohr et al., 1993; Spohr & Steinhausen, 1987, 1984) described above has followed people with FAS or FAE (n=158) in the former West Germany since 1977, testing preschool, early school-aged (6-12 years) and late school-aged (*13 years) children using behavioural assessments (Child Behavior Checklist), psychological examinations (Columbia Mental Maturity Scale) and psychiatric assessments. Subjects entered the study at various agesfromnewborn to age 11.5 years. The early school-aged children diagnosed with FAS showed hyperkinetic disorders on psychiatric testing and attention deficit and social problems according to the parents' and teachers' ratings. Hyperkinetic disorder was the mostfrequentlydiagnosed problem, persisting throughout the various assessments at different ages, leading the researchers to conclude that "hyperactivity and distractibility seem to be the major handicap for a normal school career of these children" (Spohr & Steinhausen, 1987, p. 13).  The results of Swedish researchers who followed a cohort of 24 children of alcoholic mothers were consistent with the Seattle and German longitudinal studies. Young children, regardless of whether they had physical characteristics of FAS, were observed to have attention deficits and hyperactivity (Aronson et al., 1985; Kyllerman et al., 1985, 1983). Follow-up assessments at ages 12-14 years showed that the children continued to have difficulties with attention (Aronson & Hagberg, 1998).  The results of the Atlanta longitudinal study differedfromthe Swedish, Seattle and German studies, described above, in that attentional problems but not hyperactivity were observed. In the 33  Atlanta longitudinal study (Brown et al., 1991; Coles et al., 1991) three groups of offspring were compared; those whose mothers drank throughout pregnancy (n=25), those whose mothers stopped drinking during the second trimester (n=22) and those whose mothers did not drink alcohol during pregnancy (n=21). All women were African-American and of low socioeconomic status. No differences in impulsivity or hyperactivity were observed, but those whose mothers drank throughout pregnancy showed less ability in sustaining attention, and were more often described by their teachers (although not their parents) as showing behavioural problems and difficulties because of inattention.  In a later study by Coles et al. (1997), children (n=149) with attention deficit hyperactivity disorder (ADHD) were compared with three groups; children diagnosed with FAS, children who were prenatally exposed to alcohol but who did not have dysmorphic characteristics and a control group of children not exposed to alcohol prenatally. The mean age was 7.6 years and all were African-American and of low socioeconomic status. As well as testing intellectual abilities, the children's behaviour was assessed both with conventional instruments, such as the SNAP and the Child Behaviour Checklist (both parent and teacher versions), and a more specific measure of neurocognitive functioning reflecting a four-factor model (focus, sustain, encode and shift) of attention developed by A F . Mirsky. Coles et al. (1997) found differences between the two alcohol-exposed groups and children with ADHD. Children with ADHD performed more poorly on the conventional assessments of attentional problems and conduct disorder than the two alcohol-exposed groups, in spite of the three groups being equally impaired intellectually. The researchers were able to distinguish behavioural differences and, in particular, in the patterns of  34  inattention between the alcohol-exposed groups and the children with ADHD. Children with ADHD were best identified by their inability to focus and sustain attention, while those with FAS were identified by difficulties with visual/spatial skills and flexibility in problem-solving. The study's design could be distinguished from many others in this area since it was not based on case reports, nor was its sample drawnfrompeople referred for psychiatric assessment. Children in this longitudinal study were living with birth or adoptive mothers or other relatives, had all been selectedfromthe same teaching hospital, and had been selected for a study of prenatal alcohol exposure or when they were diagnosed as having ADHD.  The results of Coles et al. (1997) were inconsistent with those of Nanson and Hiscock (1990) who studied attentional problems, comparing three groups, children diagnosed with FAS/FAE (n=20), those diagnosed with attention deficit hyperactivity disorder (ADHD) and an equal number of controls (not ADHD and not exposed to alcohol). Most of the subjects were aboriginal and all lived in Saskatchewan. Parents' rating scales (Connors Abbreviated Parent Teacher Questionnaire, the SNAP and the Child Behaviour Checklist) and assessment instruments (WPPSI or WISC-R short form) were used. The results indicated that the children with FAS/FAE were similar with respect to attention deficits and social behaviour to those with ADHD, although those with FAS/FAE were considerably more impaired intellectually.  In an early study of 15 children, the school records of children aged 6.5 to 18.5 years who had been heavily exposed to alcohol during pregnancy were examined (Shaywitz et al., 1980). It was found that for all but one of the subjects, hyperactivity was noted and all but two had been 35  referred for special education by the time they were infirstgrade, with the other two being referred by third grade.  All the above studies focused on samples of children with FAS or FAE. Studies were also found that examined the behavioural outcomes resultingfroma range of prenatal alcohol exposure levels.  In the Seattle longitudinal study (Sampson et al., 1989; Streissguth et al., 1989a, 1998b), described above, it was found that classroom behaviours and abilities most negatively associated with prenatal alcohol exposure were cooperation, impulsivity, retention of information, paying attention,finishingtasks, relating experiences and formulating ideas. The covariate that attentuated the correlation but did not fully explain the variance was paternal education. Several other covariates were included in the model, such as maternal tobacco and caffeine use, nutrition, maternal education and birth order (Sampson et al., 1989; Streissguth 1989a, 1989b, 1989d). Compared to the control group, overall activity was not significantly associated with prenatal alcohol exposure, indicating that hyperactivity did not contribute to the attentional deficits. The results were similar at two assessment ages, 4 and 7.5 years (Streissguth et al, 1986a; Streissguth et al., 1984). At age 11 years, hyperactivity and attentional problems were observed amongst those who had been prenatally exposed to binge drinking (Carmichael Olson et al., 1992) and attentional problems were observed again at age 14 years. Two-thirds of the sample who had poor performance on vigilance tests at age 14 years had previously been rated below average on attention by their teachers, compared to 26% of the remainder of the sample, leading the authors  36  to conclude that neurobehavioural "alcohol effects are not attenuating with age" (Streissguth et al., 1994a, p. 96).  Using the same cohort, Landesman-Dwyer et al. (1981) attempted to distinguish between the effects of alcohol and tobacco exposure by categorizing the mothers into four groups according to whether they used tobacco/alcohol, used tobacco/not alcohol, used alcohol/not tobacco and used neither tobacco nor alcohol. They categorized alcohol use into three groups, moderate, occasional and non-drinking, and tobacco use dichotomously (smoked or did not smoke). None of the women was alcoholic or an alcohol abuser. They tested children for attention span, activity level, fidgetiness and social compliance during naturalistic home setting observations. Consistent with thefindingsof Streissguth et al. (1994b, 1986a, 1984), described above, researchers observed the offspring of moderate drinkers, defined as mean 0.45 oz. absolute alcohol per day, to be less attentive, less compliant to parents' commands and morefidgetyduring mealtimes than either the non-drinking group or the occasional drinking group.  The results of these Seattle studies are inconsistent with three longitudinal studies in Ottawa, Cleveland and Pittsburgh. In a sample reasonably similar to the Seattle study, that is middle-class predominately Caucasian women, the Ottawa longitudinal study found no relationship between prenatal alcohol use and attention in a study of children repeatedly assessed at ages newborn to 12 years (Fried et al., 1998, 1997, 1992; Fried & Watkinson, 1990; Fried & Makin, 1987). In a distinctly different sample of children of predominately black, disadvantaged women in Cleveland, the results were similar infindingno association between prenatal alcohol exposure and sustained  37  attention, as measured by a vigilance task, leading the researchers to conclude that "if there is an effect of prenatal alcohol exposure on vigilance task performance in young children, that effect is small" (Boyd et al., 1991, p.55). The Pittsburgh study sample (Leech et al., 1999) was similar to the Cleveland study sample, with half the population being African-American and all from socioeconomically disadvantaged backgrounds. At age 6 years, they found no statistically significant difference between the alcohol exposed and non-exposed groups with respect to attentional processes. Interesting, lower I.Q. scores (as measured by the Stanford-Binet scale), male gender and adult male in the household predicted more errors of commission, indicating impulsivity, while lower I.Q. scores, younger child age, maternal work or school status and higher maternal hostility predicted more omission errors indicating inattentiveness. Unfortunately for purposes of comparisons, these factors have not been measured or independently assessed in all other studies examining prenatal alcohol exposure. As described earlier, the Ottawa and Cleveland longitudinal studies also differed in resultsfromthe Seattle study with respect to cognitive development and academic achievement.  Apartfromthe longitudinal studies, or the studies particularly focused on attentional problems, hyperactivity or distractibility, there are two cross-sectional studies worth noting briefly because of their contribution to the study of psychosocial problems of children prenatally exposed to alcohol. Both studies focused on people with FAS or FAE or who had been exposed to heavy heavy alcohol use. Roebuck, Mattson and Riley (1999) compared 32 children heavily exposed to alcohol (including 19 diagnosed with FAS) with controls matched for age, gender and ethnicity in an effort to understand psychosocial functioning of children with and without the diagnostic 38  features of FAS. Both the children with FAS and those heavily exposed to alcohol but without the physical features of FAS were observed to be impaired in psychosocial functioning.  Finally, the first study to determine the prevalence of youth with FAS in the criminal justice system was completed recently in British Columbia (Fast et al., 1999). After assessing all youth who had been remanded for a forensic psychiatric or psychological assessment, it was observed that 23% had an alcohol-related diagnosis, with 1% diagnosed with FAS and 22% diagnosed with FAE. Since this represented a disproportionately large number within the British Columbia juvenile justice system, the researchers suggested that the learning and behavioural problems associated with FAS and FAE may increase youth's susceptibility to criminal behaviour. It could also be suggested that learning or behavioural problems may increase the susceptibility to being apprehended or charged with criminal offenses.  2.2.3 Summary and discussion of prenatal alcohol exposure research Research over the past 30 years has repeatedly used case studies to describe the physiological, behavioural and learning problems associated with FAS. The outcomes for children with FAS are devastating with respect to health, cognitive development and behavioral problems. Case studies, longitudinal research and animal studies have established alcohol as a teratogen. Many studies of children with FAS have very small samples. For studies not describing FAS specifically, but rather the long-term effects of any dose of prenatal alcohol exposure, the varied methods of assessing alcohol consumption have rendered it difficult to conduct a meta-analysis in which data could be combined (Forrest et al., 1992), although one meta-analysis has been attempted 39  regarding physical malformations (Polygenis et al., 1998).  The scientific knowledge regarding the outcomes of exposure to varying doses of alcohol, with different patterns of use and at different times during pregnancy, is developing but could not be regarded as conclusive. It appearsfromthe research on physical outcomes of moderate prenatal alcohol exposure, at least in studies of children who were not dysmorphicologically affected, that if the children were disadvantaged by growth retardation, catch-up occurred by school age.  Central nervous system damage is one of the elements of FAS, including sustained damage, as observed in longitudinal studies that have been conducted in Germany, Sweden, Seattle and Pittsburgh. Pertinent to the present study was the research literature on the behavioural and learning problems found in children, other than those diagnosed with FAS or FAE. Another Seattle longitudinal study, as well as longitudinal studies in other U.S. cities and in Ottawa, have provided datafromlarger samples on moderate alcohol exposure during pregnancy. These studies have also provided comparisons between exposed and non-exposed groups within the same cohort and comparisons between alcohol exposed groups and groups of children with other diagnoses such as Attention Deficit Hyperactivity Disorder.  Attention deficit, hyperactivity and impulsivity have been the most often described behavioural characteristics of children with FAS or FAE in the case study reports. This led two research teams (Coles et al., 1997, 1991; Nanson & Hiscock, 1990) to conduct comparative studies on children with ADHD and with FAS. Since the results were not entirely consistent, more work is 40  needed on this topic.  Most of the studies observed no relationship between behavioural problems and moderate prenatal alcohol use, including longitudinal studies in Ottawa (Fried et al, 1992a, 1992b; Fried & Watkinson, 1990; Fried & Makin, 1987), in Pittsburgh (Leech et al., 1999), in Cleveland (Greene et al., 1991a, 1990; Boyd et al., 1991) and in Detroit (Russell et al., 1991). Comparing these results with those of attentional problems, impulsivity or hyperactivity may suggest the existence of a threshold effect. Indeed, this was observed in a study of infants under 1 year of age in Detroit wherein scores on tests of psychomotor development were lower after exposure to 4 drinks per day. The results of these studies are inconsistent with those of two research teams (Streissguth et al., 1994b, 1993, 1989a, 1998b, 1986a, 1981; and Landesman-Dwyer et al., 1981) who studied children from the Seattle cohort, one using standardized assessment instruments and the other using observational methods. These Seattle studies both observed deficits in distractibility, restlessness, lack of persistence and attentional problems, with one of the studies (Streissguth et al., 1994a, 1994b) showing these characteristics persisting until at least age 14 years.  With respect to intellectual development or learning problems, the same pattern of research results holds. The case-study reports of children with FAS or FAE described low I.Q. scores, although the range was large, so that some children with FAS or FAE came within normal range on intelligence tests. For those not diagnosed with FAS or FAE, the results of studies of preschool and school-age children were inconsistent. The Cleveland study (Greene et al., 1990, 1991a)  41  showed no relationship between abilities and exposure, concluding that the confounder of home environment was more predictive of low test scores than was prenatal alcohol exposure. Deficits in abilities in mathematics or arithmetic were observed in studies in Atlanta (Coles et al., 1991), San Diego (Mattson et al., 1999; Mattson et al., 1997; Mattson et al., 1996), Seattle (Streissguth et al., 1989a, 1989b) and Pittsburgh (Goldschmidt et al., 1996). The latter study specifically studied the relationships between abilities and exposure, finding a linear dose-related relationship between low arithmetic scores and prenatal alcohol exposure. Similarly, deficits in language abilities (including spelling and reading) were found in studies in Atlanta, Ottawa (at age 2-3 years), San Diego, Pittsburgh and Seattle. Not all studies found these deficits in mathematical or language skills. Re-assessment or more precise testing is needed before conclusions can be drawn. For example, the Ottawa study found that there were deficits in arithmetic scores at ages 2-3 years, but that these deficits no longer existed at re-assessment at ages 5-6 years. Either the earlier assessment was insufficient or, as in the case of physical growth retardation, the children caught up after initial deficits. More precise examination of results led researchers in the Pittsburgh study tofindthreshold effects on spelling and reading scores, contrasting with the linear effect found on arithmetic scores (Goldschmidt et al., 1996).  The studies of prenatal exposure to alcohol, as reflected from following children with FAS, indentify adverse effects on cognitive abilities, although many children with normal I.Q. scores have been documented. As Forrest et al. (1992) noted after reviewing the literature, the evidence regarding any adverse effects of moderate levels of prenatal alcohol consumption, beyond the neonatal stage is unconvincing. They suggested that the lowest potential threshold level for 42  prenatal alcohol effects to be 165 grams per week (that is, about 3 drinks per week) in the studies reviewed at that time (Forrest et al., 1992). From the work of Goldschmidt et al. (1996), it may be more productive to examine different types of effects, or to examine additive effects controlling for covariates including socioeconomic status.  Of the longitudinal studies found in the literature, five (Sweden, Germany, Detroit, one of the French studies and one of the Seattle studies) followed children with FAS or FAE exclusively and did not describe specifically the socioeconomic background of these children. It appears that the subjects were included in the studies after presenting at a clinic or other medical care facility. In five other longitudinal studies (Roubaix in France, Atlanta, Cleveland, Detroit, Pittsburgh), researchers selected their samples from socially and economically disadvantaged urban areas, often inner city areas near the clinics in which the research took place. Two other longitudinal studies (Ottawa and Seattle) had samples of women who were predominantly white, middle-class women with higher education levels. Caution should be taken in drawing conclusions from studies with homogeneous samples, especially given the numbers of samples on this topic that have been described as low income or living in socioeconomically disadvantaged areas.  43  Table 2.1 Studies of cognitive and behavioural outcomes associated with prenatal alcohol exposure, listed in alphabetical order by first author. Authors, country, sample size and ages of testing  Study description  Results: behavioural outcomes  Aronson and Hagberg, 1998; Aronson, Hagberg & Gilbert, 1997; Aronson, Kyllerman, Sabal, Sandia & Olegard, 1985; Kyllerman, Aronson, Sabal, Karlberg, Sandia & Olegard, 1985; Kyllerman, Aronson and Olegard, 1983; Olegard, Sabal etal., 1979 Sweden (n=30) ages: neonate to 14 years  Longitudinal study of children of alcoholic mothers, comparing with matched controls.  Hyperactivity, distractibility, perception disorders and short attention span, that continued throughout childhood and to latest follow-up at age 14 years.  Results: cognitive outcomes or learning problems FAS group had lower I.Q. and perceptual delay in early years; with no significant I.Q. difference between those reared in foster parents' home and in biological parents' homes. Six of the 24 children attended schools for the mentally retarded and 11 had some type of special education.  Boyd etal., 1991 - see Greene et al., below. Coles, Platzman, RaskindHood, Brown, Falekand Smith, 1997; Coles, Brown, Smith, Platzman, Erickson, andFalek, 1991 U.S.A. (Atlanta) AfricanAmerican, low-income (n=122)  Study 1: Comparison of three groups: women who drank throughout pregnancy, women who . stopped during pregnancy and women who never drank during pregnancy.  On scores of pre-math and reading: offspring of women who never drank during pregnancy were better than groups whose mothers stopped during pregnancy, or whose mothers continued to drink through pregnancy.  ages: 5-8 years  Study 2: Comparison of children affected by prenatal alcohol exposure and children with attention deficit, hyperactivity disorder.  Attention and conduct disorder: ADHD children did worse  I.Q. testing: Children with FAS had same scores as ADHD compared to controls.  Corny, 1990  Cross-sectional study of children with and without diagnosis of FAS or FAE in isolated native community.  Neuropsychological testing: FAS group significantly lower on reaction time, finger tapping, grip strength and motor speed and precision. FAE group significantly lower only on grip strength.  I.Q. testing: FAS group significantly lower on all components of I.Q. tests, but FAE group were not, comparing to controls.  Canada (BC) (n=15 with FAS and n= 10 with alcoholic mothers) ages 6.4 to 18.5 years  Comparing three groups: children with FAS, children without FAS bom to alcoholic mothers; and controls.  44  Table 2.1 continued Fried, Watkinson & Watkinson, 1992b, Fried, Watkinson & Gray, 1992a; Fried & Watkinson, 1990; Fried & Makin, 1987; Fried AO'Connell, 1987; Canada (Ottawa) (n=250)  Ottawa Prenatal Prospective Study longitudinal study of children prenatally exposed to marijuana, cigarettes and alcohol: cognitive and language assessment. Middleclass, lowriskwomen.  No relationship between prenatal alcohol exposure and attention deficits.  Prenatal exposure to low doses of alcohol resulted in lower cognitive scores and lower language skills at ages 2 and 3 years, but this was reduced to no significant difference by age 5,6 years.  Longitudinal study of offspring of sociceconomically disadvantaged women, moderate alcohol use.  At ages 4 year 10 mos., prenatal alcohol exposure not found to be associated with attention deficit, as assessed on vigilance task.  No relationship between cognitive abilities and maternal drinking in absence of FAS, when tested at various ages. Home environment a better predictor of low scores on cognitive assessments.  Longitudinal study testing for relationship between prenatal alcohol exposure and academic achievement, in cohort of children prenatally exposed to moderate drinking levels, from lower socioeconomic backgrounds.  No significant effects observed of prenatal exposure on attention deficits or hyperactivity.  Linear (dose-related) relationship for arithmetic deficits; threshold effects for spelling and reading.  ages: each age 1-6, 8, 9, 12 years Greene, Emhart, Ager, Sokol, Martier and Boyd, 199 la; Green etal, 1991b; Boyd, Emhart, Greene, Sokol, Martier, 1991; Greene, Emhart, Martier, Sokol, Ager, 1990; Sokol, Miller, Debanne, Golden, Collins, Kaplan & Martier, 1981; U.S.A. (Cleveland) (n=359) ages: 6 months, 1,2, 3 years, 4 years 10 months. Goldschmidt, Richardson, Stoffer, Geva and Day, 1996; Leech, Richardson, Goldschmidt and Day, 1999 U.S.A. (Pittsburgh) (n=595- 763 depending on study) age: 6 years  45  Table 2.1 continued Jacobson, Jacobson, Sokol etal., 1998,1993a, 1993b U.S.A. (inner city Detroit) (n=480)  African-American infants, over-represented on prenatal alcohol exposure at moderate to heavy levels. Examining threshold effects.  ages: 6.5,12 and 13 months  Possible threshold effect on test of psychomotor development, since scores reduced only below higher levels of exposure (4 drinks/day).  Deficits in development in infants heavily exposed to alcohol (2 5 drinks per occasion at least once weekly), especially in children whose mothers were older than 30 years. Dose-response relationship on Bayley tests of development at 13 months.  Kyllerman, Aronson, Sabal, Larberg, Sandia and Olegard - see Aronson et al. above Landesman-Dwyer, Ragozin and Little, 1981  Children of moderate drinkers were found to be less attentive, more fidgety and less compliant with parents, compared to occasional maternal drinkers and nondrinkers.  U.S.A. (Seattle) (n=l28) age: 4 years  Larroque and Kaminski, 1998 France (n=160, with overrepresentation of heavy drinkers) age: 4.5 years  Aimed to investigate moderate prenatal alcohol exposure. Many potential confounders measured including tobacco use, parity, maternal age, family state, employment, family stimulation of child.  Lower scores on McCarthy scales, after controlling for confounders, for those exposed to *1.5 oz. absolute alcohol (3 drinks) per week. Verbal, quantitative and performance perceptive scales related to prenatal exposure.  Leech et al. - see Goldschmidt et al., above Nanson and Hiscock, 1990 Canada (Saskatchewan) (n=29 FAS, 20 ADHD, and 20 controls)  Comparing groups to understand relationship between FAS and Attention Deficit Hyperactivity Disorder (ADHD).  46  Children with FAS/FAE were similar in respect to attention deficits and social behaviour as those with ADHD.  Children with FAS/FAE were considerably more impaired intellectually.  Table 2.1 continued Mattson, Goodman, Caine, Delis & Riley, 1999; Mattson and Riley, 1999; 1998; Mattson, Riley, Grambling, Delis & Jones, 1998,1997; Mattson, Riley, Delis, Stem, Gramling and Jones, 1996; Roebuck, Mattson and Riley, 1999,1998; Roebuck, Simmons, Richardson et al., 1998; Thomas, Kelly, Mattson and Riley, 1998;  Various cross-sectional comparative studies examining differences between those exposed to heavy alcohol use prenatally (children with and without physical features of FAS) and controls. Studied memory, cognitive and neuropsychological development.  Children exposed to heavy use of alcohol had significant impairments in psychosocial functioning, regardless of whether they had facial anomalies characteristic of FAS.  Lower I.Q.'s found in children exposed to heavy alcohol use, including those without physical features required for diagnosis of FAS; both those with and without physical features showed impairment on tests of language, verbal learning and memory, academic skills, fine-motor speed and visual-motor integration. Children heavily exposed to alcohol had deficits on tests of memory (verbal fluency tasks) and on other cognitive development tests.  U.S.A. (San Diego) ages: 3-16 years  Roebuck et al. - see Mattson et al., above Russell, Czamecki, Cowan, McPherson and Mudar, 1991 U.S.A. (Detroit) (n=175) age: 6 years  Shaywitz, Cohen, and Shaywitz, 1980  Longitudinal study, comparing those whose mothers indicated more than one drinking problem with those whose mothers indicated less drinking problems, on a self-administered questionnaire.  No relationship between prenatal alcohol exposure and performance on a psychomotor test.  Significantly lower scores on language and verbal scales of the Wechsler Preschool and Primary Scale of Intelligence (WPPSI) and on Token Test.  Retrospective study of children heavily exposed to alcohol prenatally  Hyperactivity evident on school records for all but one of the subjects.  All but two of the subjects referred to special education by grade 1.  U.S.A. (Connecticut) (n=15) ages 6.5 to 18.5 years  47  Table 2.1 continued Spohr, 1996; Spohr & Steinhausen, 1996,1987, 1984; Spohr, Willms& Steinhausen, 1993; Steinhausen, 1996; Steinhausen, Willms & Spohr, 1994,1993  Longitudinal study of psycho-pathology, behavioural and cognitive outcomes of children with FAS or FAE.  Psychopathology, including hyperkinetic disorders, emotional and sleep disorders and other abnormal habits that remained over time. At age 34 years, attention deficit, hyperactivity, and distractibility were prevalent; over time affected children remained "abnormal" in psychiatric testing. Males had more severe psychiatric problems than girls. Older ages: social and attention problems persisted.  Germany (Berlin) 1977 to present (n=60-158, depending on age) ages: preschool to 14 years  Streissguth, 1996,1994a; Streissguth, Sampson and Barr etal, 1989; Streissguth, Clarren & . Jones, 1985a; Carmichael Olson etal., 1998  Children and adults with FAS described.  Seattle Longitudinal Prospective Study on Alcohol and Pregnancy, beginning with predominantly middle class women who gave birth during the one-year period 1975-1976.  U.S.A. (Seattle) (n=500) ages: neonatal to 14 years  48  At latest testing (ages 1214 years) 17 of the children attended specialized schooling; difficulties in mathematics, logical conclusions, visual perception, spatial relations and short-term memory.  Deficits in attention and memory tasks.  Intellectual impairments, physical dysmorphia correlated with level of mothers' alcoholism. Difficulties in processing speed and accuracy, declarative learning and cognitive flexibility and planning.  At age 7.5 years: deficits in distractibility, overpersistence, reassurance-seeking and poor organization At age 11 years: distractibility, restlessness and lack of persistence. Age 14 years: attention problems persisted.  At age 7.5 years: deficit on I.Q. scores, decrements of 1-3 months on arithmetic and reading scores. Binge drinking prior to pregnancy had the largest effect on academic performance. At ages 11 and 14 years, arithmetic scores remained low amongst the children of heavy drinkers.  1  U.S.A. (Seattle) (n=92 or less at follow-ups) ages: neonatal to adult Streissguth, Barr et al., 1994b, 1990, 1989d, 1989a; Streissguth, Bookstein, Sampson & Barr 1993, 1989b; Streissguth & Ladue, 1985b, Streissguth, Martin, Martin & Barr, 1981; Streissguth, Bar, Martin & Hermon, 1980  Large proportion of sample were considered mentally retarded and did not change considerably over time, although morphological damage decreased over time.  2.3 Prenatal and postnatal exposure to tobacco Prenatal tobacco exposure has well-documented, profound, adverse consequences on the health of the fetus and neonate. Data have been collected on this topic for at least 65 years (Sontag and Wallace, 1935) and over those years there have been ample research findings, reviews of the literature (U.S. Department of Health and Human Services, Office on Smoking and Health, 2001, 1970; Anderson, 1997; Edwards et al., 1996; Rush & Callahan, 1989; Kleinman & Kopstein, 1987; Landesman-Dwyer & Emanuel, 1979) and meta-analyses (Windham et al., 1999; Strachan & Cook, 1997; Kramer, 1987) giving credence to these associations. Some of the physical effects last long into childhood with increased morbidity and mortality being observed amongst children prenatally exposed to tobacco. These physical outcomes, especially in relationship to the fetus and neonate, have been widely accepted because of comparable animal research, the replication of the epidemiological data and plausible mechanisms. As well, at least 23 human studies have linked behavioural problems and cognitive deficits with prenatal tobacco exposure, some with very large samples. The results of these latter studies appear to be less well-established, as compared with the research of neonatal or childhood physical morbidity and mortality associated with prenatal tobacco exposure, or as compared with behavioural problems associated with teratogens such as alcohol. A review of the literature on this topic follows.  Various longitudinal studies, some with very large samples, have been conducted in a number of countries, including the United Kingdom, Finland, the Netherlands, New Zealand, Australia, Canada and the United States. For example, the British National Child Development Study (NCDS) began with the study of every baby born in England, Scotland and Wales (n=16,000) 49  during a one-week period in March 1958 (Davie et al., 1972) with a follow-up at age 7 years. The Finnish Cohort Study followed 1,763 children who were prenatally exposed to tobacco from birth to age 14 (Rantakallio & Koeiranen, 1987; Rantakallio, 1983). Three longitudinal studies have taken place in the United States: 30,000 children were followed up to age 7 years (Naeye, 1992b; Naeye & Peters, 1984; Hardy & Mellits, 1972), a cohort of Washington state children has been followed since the late 1970s (Streissguth et al., 1984) and the U.S. National Longitudinal Survey of Youth (NLSY) studied the offspring of their original cohort with 2,256 children ages 4 to 11 years who had been surveyed in 1986 (Weitzman et al, 1992). In New Zealand, 1,265 urban children have been followed at annual intervals since their birth in 1977 (Fergusson, 1999; Fergusson & Lloyd, 1991), while in Australia, a cohort of 5,342 children has been followed since their birth in the early 1990's (Williams et al., 1998). A cohort (n=l,377) is being studied in the Netherlands but reporting has only reached the age of 3 years (Orlebeke et al., 1999). In Canada, Fried and his colleagues have followed 250 children in the Ottawa area for many years, those children now being in their early teenage years (Fried et al., 1999, 1998, 1997, 1992a, 1992b; Fried, 1993; Fried & Watkinson, 1990; Fried & Makin, 1987). Also in Canada a national longitudinal study, completed biennially and now in its third cycle of data collection, has produced data that may prove useful, but to date has not been used to report on the effects of prenatal tobacco exposure (Canada. Statistics Canada, 1997).  Following is a review of the literature describing the effects of prenatal tobacco exposure, with special attention to longer term effects that are applicable to the current study of children who are 8 and 9 years old.  50  2.3.1 Physical effects of prenatal tobacco exposure Extensive research has been reported on the effects of prenatal exposure to tobacco on the physical health of the fetus and newborn.  At the prenatal stage, evidence was found of a causal link between tobacco exposure and fetal growth and survival of the fetus (U.S. Department of Health and Human Services, Office on Smoking and Health, 2001, 1979; Ness et al., 1999; Lieberman, 1994; Naeye, 1992b; Kline et al., 1987; Kramer, 1987; Shiono et al., 1986). Risks to the fetus that have been observed include abruptio placenta, preterm rupture of the membrane and ectopic pregnancy possibly resulting in premature birth and spontaneous abortion (Castles et al., 1999; DiFranza & Lew, 1995; Armstrong et al., 1992; Kramer, 1987). At the newborn stage, there has been evidence of low birth weight, lung and respiratory problems and Sudden Infant Death Syndrome after prenatal tobacco exposure (Strachan & Cook, 1997; ACOG Technical Bulletin, 1994; McGee & Stanton, 1994; McDonald et al., 1992a; Fingerhut et al., 1990; Kleinman et al., 1988; Kline et al., 1987; Lumley et al., 1985; Sexton & Hebel, 1984; Fergusson et al., 1981; Abel, 1980; Harlap & Davies, 1974; Colley et al., 1974).  Strong evidence links prenatal tobacco use and low birth weight (U.S. Department of Health and Human Services, Office on Smoking and Health, 1979; Brooke, 1989; Kramer, 1987; Sexton & Hebel, 1984) that may have consequences for mortality and morbidity. An intervention trial observed improved birth weight outcomes when smoking was reduced during pregnancy (Sexton & Hebel, 1994). Brooke et al. (1989) investigated the effects of various socioeconomic and 51  lifestylerisksand conditions concluding that the main cause of low birth weight was maternal smoking during pregnancy and that social and psychological factors had little or no direct effect on birth weight. A dose-response relationship was established (Hebel et al., 1988) although it has been observed that stopping smoking by the 8th month of pregnancy was as predictive of outcome as dose (Hebel et al., 1988). To gain precision in measuring prenatal tobacco exposure, British researchers categorized people according to the brand (low or high yield cigarettes) and the quantity smoked during pregnancy. They adjusted both carbon monoxide yield and number of cigarettes smoked in an attempt to find a threshold for birth weight. They found that both yield and quantity were important in determining low birth weight, but brand smoked (determining carbon monoxide yield) to have more effect on determining the threshold than quantity smoked. They observed a threshold for birth weight loss at 13 cigarettes/day and 15 mg/cigarette carbon monoxide (Peacock et al, 1991).  Several researchers, attempting to determine the long lasting effects of the neonate weight differences, examined whether children whose mothers smoked during pregnancy, and consequently were born low birth weight, continued to have retarded growth later in childhood and early adulthood (Vik et al., 1996; Eskenazi & Bergmann, 1995; McGee & Stanton, 1994; Day etal., 1992; Rona et al., 1985; Naeye & Peters, 1984; Rantakallio, 1983; Fogelman, 1980; Dunn et al., 1976; Butler & Goldstein, 1973; Davie et al., 1972; Hardy & Mellits, 1972; Goldstein, 1971). Results of many of these studies showed small but significant differences in favour of the offspring of nonsmokers with respect to weight and height, even when controlling for various confounding variables such as child's gender and length at birth, age of mother and socioeconomic  52  factors. The research has not been conclusive, however, since other researchers found that initial physical differences between exposed and non-exposed groups diminished by ages 8 months (Barr et al., 1984), age 12 months (Nafstad et al., 1999), age 18 months (Day et al., 1992), age 3 years (McGee & Stanton, 1994; Fox et al., 1990), age 9 years (McGee and Stanton, 1994) or age 14 years (Rantakallio, 1983) after adjusting for other maternal and paternal factors (such as paternal smoking) or duration of breastfeeding or initial birth weight differences. The differences between the two groups (non-exposed and exposed) appeared to be extensions of deficits in fetal growth rather than postnatal growth retardation. Many studies that involved re-analyses of longitudinal studies in which data were collected years ago did not have the advantage of testing specific dose responses. Children of heavy smokers were found to be significantly shorter than non-exposed children at age 5 years, but even this difference disappeared when birth weight and gestational age were controlled. Vik et al. (1996) found that by age 5 years, children of smokers had a complete catch-up in weight and a partial catch-up in height. It appeared that the reduction in birth weight took place in the second or third trimester of pregnancy (Vik et al., 1996; Butler & Goldstein, 1972).  The longer term effects of low birth weight may persist in the area of behavioural problems (McCormick et al., 1996) and cognitive development (Breslau et al., 1994). Breslau et al. (1994) compared randomly selected groups of low birth weight children and normal birth weight children, using two disparate populations, an urban disadvantaged one and a suburban middle-class population. At age 6 years, the I.Q. scores of the four groups were compared, with a gradient of mean scores found: the disadvantaged, urban, low birth weight group had the lowest I.Q. scores,  53  the disadvantaged urban normal birth weight group, the suburban low birth weight group, and the suburban normal birth weight group had relatively higher I.Q. scores ranked in the order presented Controlling for maternal education and socioeconomic status, those born with low birth weight had lower I.Q. scores than those born with normal birth weight. These researchers did not extend their study to suggest causal relationships for the low birth weight, such as prenatal tobacco or alcohol exposure or other lifestyle factors, but the importance of their research was the suggestion that adverse neonatal health conditions may persist in cognitive development at early school age. McCormick et al. (1996) found more behavioural problems amongst children born with low birth weight, although they also suggested that considerable portions of the adverse outcomes could be modified by environmental changes.  Lassen and Oei (1998) reviewed nine studies on the physical development of children prenatally exposed to tobacco and concluded that the evidence has been consistent on reduced height after prenatal tobacco exposure, in spite of the differing variables controlled by the researchers. They also concluded that: "it can be seen that the deficits, although they appear to be stable and relatively long term, are small and may be of little functional importance" (Lassen & Oei, 1998, p. 641).  Most of the studies included in Lassen and Oei's (1998) review and in the present review controlled for various confounding variables and were found to be precise in the level and patterns of tobacco use measured. Unfortunately, very few studies controlled for alcohol consumption.  54  It has been suggested that the mechanism relating to delayed fetal or childhood growth amongst those prenatally exposed to tobacco is foetal hypoxia that may result from increased carboxyhemoglobin levels, vasoconstriction of the blood supply to the placenta, or attenuated blood oxygen unloading (Fergusson, 1999; Orlebeke et al., 1999; Lassen & Oei, 1998; Milberger et al., 1996; Fergusson et al., 1993; Morrow et al., 1988; Naeye & Peters, 1984; Abel, 1980; Lehtovirta & Forss, 1978; Longo, 1976; Denson et al.; 1975; Colley al, 1974). Investigators have debated whether the appetite suppressing effects of nicotine influences birth weight, but it appears from the literature that other mechanisms are responsible for at least some of the deficits since the birth weight effects persisted even after caloric intakes were increased amongst smokers in some research studies (Kramer, 1987).  In addition to height and weight deficits for which there has been considerable research, some other health problems may persist well beyond the newborn stage, so that children whose mothers smoked during pregnancy continued to be at risk for ear infections, asthma, allergies and respiratory tract illnesses in general (Stathis et al., 1999; Fried & Watkinson, 1990; Moessinger, 1989; Taylor & Wadsworth, 1987; Rantakallio, 1983, 1978; Fogelman, 1980). On the other hand, high blood pressure found in newborns whose mothers smoked during pregnancy was not found to persist at ages 9 or 18 years (Williams & Poulton, 1999). Fogelman (1980) found that asthma or wheezy bronchitis persisted well into the teenage years. Medical histories of 16-year-olds revealed that the offspring of women who smoked heavily (>10 cigarettes daily) during pregnancy were significantly more likely to have these conditions. The relationship between asthma or wheezy bronchitis and prenatal tobacco exposure was confusing, as those who were exposed at  55  medium levels (10 cigarettes daily) were the least likely to have these conditions, compared to the heavily exposed group and the non-exposed group. Rantakallio (1983) found that the children at age 14 years were more prone to respiratory illnesses if they had been prenatally exposed to tobacco and that there were statistically significant differences between the exposed and nonexposed in rates of mortality up to 14 years.  In an earlier study by Rantakallio (1978) found statistically significantly higher rates of mortality from age 28 days to 5 years, more and longer hospitalizations and more visits to doctors were not noted in those prenatally exposed to tobacco, but the differences between the exposed and nonexposed groups were clearest in those under the age of one year. Also examining mortality rates amongst the offspring of smokers, a Swedish researcher examined the relative risk of post-neonatal death and early childhood death (ages 1 to 8 years) amongst children whose mothers smoked (Hofvendahl, 1995). After controlling for birth weight, there was no higherriskfor death during the late neonatal age, but that there was an increased relativerisk(1.35) for death during the postneonatal and post-infancy ages, that is, up to about 4 years of age, but that there was no increased risk between ages 5 to 8 years. Hofvendahl (1995) controlled his analysis for various factors, including socioeconomic status.  These apparent causal links have led researchers to speculate about other physiological or chemical alterations that may in turn be the underlying mechanisms for cognitive and behavioural effects of prenatal exposure to tobacco (Fergusson, 1999; Eskenazi & Trupin, 1995; Rantakallio et al., 1992). During critical periods, fetal oxygenation and placental circulation may affect central  56  nervous system development (Niemela & Jarvenpaa, 1996, Naeye, 1992b). Interest in behavioural problems also may have increased because estimates of a high number of hyperactive children in North American children coincided with increased tobacco use by women over recent decades (Denson et al., 1975).  2.3.2 Cognitive and behavioural effects of prenatal tobacco exposure Various researchers have examined the possible association between behavioural and cognitive outcomes and prenatal tobacco exposure. Maternal smoking during pregnancy has been implicated as a risk factor for increased behavioural and cognitive problems (Brennan et al., 1999; Leech et al.; 1999; Orlebeke et al., 1999; Trasti et al., 1999; Fried et al., 1998, 1997, 1992a, 1992b, Lassen & Oei, 1998; Wakschlag et al., 1997; Milberger et al., 1996; Fergusson et al., 1993; Weitzman et al., 1992; Makin etal., 1991; Fried & Watkinson, 1990; Sexton et al., 1990; Fried & Makin, 1987; Naeye & Peters, 1984; Streissguth et al., 1984; Nichols & Chen, 1981; Hardy & Mellits, 1972). In addition to evidence provided by longitudinal studies in various countries (mentioned above), researchers have examined specific behaviours such as attention deficit hyperactivity disorder, impulsivity or motor hyperactivity (Conners et al., 1996; Milberger et al., 1996; Fried et al., 1992a), mental development (Trasti et al., 1999), conduct disorder (Wakschlag et al., 1997) or increased numbers of behavioural problems (Fergusson et al., 1993; Weitzman et al., 1992). In Vancouver, Dunn et al. (1977, 1976) studied the intellectual and neurological maturation of 6 Vi year olds whose mothers smoked during pregnancy.  To follow is a review of the results of these longitudinal and cross-sectional studies,firstdescribing  57  the research with respect to behavioural outcomes and then with respect to cognitive outcomes. Table 2.2, on pages 96-103 at the end of this section, summarizes each of the studies cited.  2.3.2.i Behavioural effects of prenatal tobacco exposure Twenty-four studies were found that examined the relationship between children's or young adults' behaviour and prenatal tobacco exposure. There was an attempt to be comprehensive and no other studies were found on the relationship between behaviour and prenatal tobacco exposure. The studies found were both retrospective and prospective and examined general behaviour as children aged; cross-sectional studies examined specific behaviours such as attention problems or impulsivity. Eleven studies were consistent in demonstrating a link between prenatal tobacco exposure and development of externalizing behaviours, such as aggressive or antisocial behaviour, conduct disorder or criminal behaviour (Weissman et al., 1999; Brennan et al., 1999; Gibson & Tibbetts, 1998; Williams et al., 1998; Wakschlag, 1997; Orlebeke et al., 1999, 1997; Milberger et al., 1996; Fergusson et al., 1993; Rantakallio et al., 1992; Weitzman et al., 1992; Rantakallio 1983). There were two studies that compared those diagnosed with ADHD to controls, after collecting data retrospectively on prenatal tobacco exposure (Landesman-Dwyer et al., 1981; Denson et al., 1975). The studies have demonstrated consistently the link between prenatal tobacco exposure and behavioural problems, from early age to adulthood, involving inattention, impulsivity or externalizing behaviour, with conduct disorder in later years and delinquency and violent criminal behaviour in adulthood.  58  General behavioural measures A study investigating the relationship between prenatal tobacco exposure and behavioural problems of children was conducted by Weitzman et al. (1992) who followed 2,256 U.S. children ages 4 to 11 years using datafromthe National Longitudinal Survey of Youth (NLSY). A strength of this research was the investigators' differentiation of varying amounts and timing of tobacco exposure. Amount of smoking were divided dichotomously (>1 pack a day or * 1 pack a day) and timing was categorized into three exposure periods (prenatal, postnatal or both pre- and postnatal exposure). Behaviour was measured by the use of the Behavior Problem Index (BPI), an adaption of the subscales of the Child Behavior Checklist (CBCL) (Achenbach & Edelbrock, 1979; Achenbach, 1978). The results indicated a dose-response relationship with children whose mothers smoked heavily (z 1 pack a day) after pregnancy, or both during and after pregnancy, having higher scores indicating behavioural problems. The odds ratio for extreme behaviour problem associated with both pre- and post-natal exposure was 1.41 if the mother smoked less than one pack a day (p<0i) and 1.54 if she smoked 1 pack a day or more (p<05). This relationship was not shown for children who were exposed only prenatally, however, the sample was small (n=19) for mothers who reported smoking * 1 pack during pregnancy but not after pregnancy. The dose-response relationship was unclear for those women who smoked only during pregnancy, since children whose mothers smoked less (<1 pack a day) showed higher behavioural problem scores than those whose mothers smoked more.  Children exposed to smoking only postnatally, or both pre- and postnatally, showed significantly higher scores on many subscales of behaviour assessments, including antisocial behaviour,  59  anxiousness or depression, hyperactivity and being headstrong. These results indicated a doseresponse. A limitation to drawing conclusionsfromthis research was that there were very few people who smoked a pack or more of cigarettes a day only during pregnancy, without continuing the habit after pregnancy. The results suggested that these effects are more likely to follow from passive smoking after the child is born than after prenatal exposure alone.  This study is of interest to the present thesis because of the age group studied and important because it is one of the few studies that controlled for simultaneous prenatal alcohol exposure and various demographic and psychological variables. Prenatal alcohol exposure was treated as a dichotomous variable according to whether the mother had one or more drinks of alcohol per week. It is possible that this was insufficiently precise as a measure of prenatal alcohol exposure, insofar as there has been some evidence suggesting that the offspring of abstainers fare worse at least for the outcome of birth weight of neonates than those who drink occasionally (Abel & Hannigan, 1995b; Forrest et al., 1991; Lumley et al., 1985).  Fergusson et al. (1993) attempted to replicate the research of Weitzman et al. (1992), using different behavioural rating scales. They addressed some of the limitations of the earlier study, namely the measurement of maternal smoking by self-report and the need for clearer differentiation between pre- and postnatal exposure. They conducted a prospective, longitudinal study, collecting datafroma cohort of 1,265 New Zealand children at ages 8, 10 and 12 years. Data collection and analyses included statistical control of a long list of potentially confounding family, social and parental variables. Their results were consistent with those of Weitzman et al. (1992), suggesting  60  that there may be a dose-response relationship of pre- and postnatal exposure to tobacco for disruptive behaviour problems at pre-adolescent age. The duplication of the results of Weitzman et al. (1992) led Fergusson et al. (1993) to conclude that the results "are generally supportive of the hypothesis that smoking during pregnancy may lead to small but statistically detectable increases in rates of problem behaviour in childhood" (Fergusson et al., 1993, p. 821).  Fergusson et al. (1993) found small but significant differences between those not exposed and those whose mothers smoked ^ 1 pack daily during and after pregnancy on the three dependent variables of conduct disorder, attention deficit and disruptive behaviour. Unfortunately, alcohol use during pregnancy was not included in their analyses, although the researchers did collect data on history of alcohol problems.  A well-designed New Zealand study (McGee & Stanton, 1994) examined the relationship between prenatal smoking and height, behaviour and cognitive development of children (n=l,037) at ages 3 to 9 years. They included a wide range of potentially confounding factors, including health, perinatal and family adversity and child rearing practices. They tested attention deficit and hyperactivity, measured by parent and teacher rating scales. The only significant relationship they found with respect to behaviour was more behavioural problems at age 5 years, as rated by mothers.  A group of Vancouver researchers conducted a prospective study to examine the neurological and intellectual development at age 6 A years of children prenatally exposed to tobacco, comparing X  61  those who had been born prematurely and small for gestational age, those born full-term with low birth weight, and controls (Dunn et al., 1977, 1976). They observed statistically significant differences on behavioural scales in favour of those not prenatally exposed, but they were ambiguous about the variables controlled and the types of behaviour measured and there was no mention of data collection regarding alcohol use.  Externalizing behavioural problems: As mentioned above, study results were consistent in observations of an association between prenatal tobacco exposure and externalizing behaviours, such as aggressive or antisocial behaviour, conduct disorder or criminal behaviour (Weissman et al., 1999; Brennan et al., 1999; Williams et al., 1998; Orlebeke et al., 1999, 1997; Gibson & Tibbetts, 1998; Wakschlag, 1997; Milberger et al., 1996; Fergusson et al., 1993; Rantakallio et al., 1992; Weitzman et al., 1992; Rantakallio 1983).  The most recent study found on this topic examined three-year olds (n=672)froma low-income U.S. urban population (Day et al., 2000). Children were assessed by maternal ratings using three standard tests, the Toddler Behavior Checklist, SNAP Checklist and Routh Activity Scale. Statistically significant effects of prenatal tobacco exposure on children's behaviour were found, with increased scores on subscales for Oppositional Behavior, Immaturity, Emotional Instability, Physical Aggression and Activity. Of all the subscales of the Toddler Behavior Checklist, the largest effect of prenatal tobacco exposure was observed on Oppositional Behaviour. Impulsivity and peer problems were found to be associated with the combination of pre- and post-natal  62  exposure. The researchers noted that the effect observed was associated predominantly with third trimester smoking, a time other researchers (Slotkin, 1998) found nicotinic receptors to be formed. Day et al. (2000) also found current tobacco use by mothers predicted only attention problems. For preschoolers, no effects of prenatal alcohol exposure were observed on behaviour, except for a positive effect offirsttrimester exposure on immaturity. As well as being the most recent study, it is one of the few that accounted for prenatal alcohol exposure.  Orlebeke, Knol and Verhulst (1999, 1997) used a large sample (n=l,377) of 2- and 3-year-old twins in the Netherlands to examine the effects on behaviour of those prenatally exposed to tobacco. They divided their sample according to those who werefirst-bornand those who were second-born in each family, and adjusted their analyses for the possible confounding effects of birth weight, socioeconomic status, maternal age and type of feeding. Measuring behaviour using the CBCL behavioural rating scale (Achenbach & Edelbrock, 1979; Achenbach, 1978), they observed a statistically significant difference between prenatally exposed and non-exposed children on externalizing behavioural problems (t=2.56, p<01 forfirstborn children, t=3.69, p<001 for second born children) but no difference on internalizing behaviours (e.g., being withdrawn, depressed, anxious). The enhanced externalizing scores for those whose mothers smoked was primarily attributable to higher aggression (t=2.37, p<05 forfirstborn, t=3.43, p<001 for second born), oppositional (t=2.42, p<05 forfirstborn, t=3.56, p<001 for second born) and overactive (t=2.15, p<05 forfirstborn, t=2.73, p<005 for second born) scores. The effect of prenatal tobacco exposure was the same for boys and girls, even though boys had higher external behavioural scores. Orlebeke, Knol and Verhulst (1999, 1997) controlled specifically for method  63  of feeding during infancy because of research showing that breast-fed children have an advantage on neurological development and that mothers who smoke during pregnancy tend to bottle feed their infants (Lanting et al., 1994). As in the case of most other studies, Orlebeke, Knol and Verhulst (1999, 1997) did not control for prenatal alcohol exposure.  Another study,, important because of its more precise measurement of nicotine exposure, found differences between the maternal ratings of 5-year-old children according to whether the children were pre- and postnatally exposed to tobacco or not exposed. A statistically significant higher number of children who were both pre- and postnatally exposed were rated by their mothers as being more active, compared to those who were only prenatally or only postnatally exposed. There was a dose-response relationship of current smoking exposure, so that children of mothers who smoked a pack a day or more had nearly twice theriskof "active" behaviour, compared to the children of nonsmokers (Eskenazi & Trupin, 1995). The importance of this study was the researchers' attempt to overcome the lack of precise measurement of dosage used by other researchers. As well as maternal interviews, they measured serum cotinine, a metabolite of nicotine to assess levels of exposure. They also differentiated pre-and postnatal exposure. Unfortunately, they did not describe the mothers' behaviour rating scale. They did include prenatal alcohol exposure as a potentially confounding variable, but did not describe whether they measured timing, duration or level of alcohol exposure.  Williams et al. (1998) also observed a link between externalizing behaviours and prenatal tobacco exposure among their Australian cohort of 5,342 children whose age averaged 5.5 years. They  64  found a clear dose-response relationship between higher levels of prenatal tobacco exposure and higher rates of externalizing behaviour problems, using the CBCL (Achenbach & Edelbrock, 1979; Achenbach, 1978) to measure behaviour patterns. They concluded that at that age, postnatal tobacco exposure accounted for 16% of the variance, while prenatal tobacco exposure accounted for 25% of the variance, even after controlling for various confounding factors including socioeconomic status.  A well-controlled study conducted in the U.S.A. followed the children of 50 smokers and 97 nonsmokers for 10 years (Weissman et al., 1999). They found an increased risk that males whose mothers smoked during pregnancy had four times the risk (relative risk=4.10, p<.01) of being diagnosed with Conduct Disorder before adolescence and three times the lifetime risk (relative risk=3.17, p<05). They also found that females whose mothers smoked during pregnancy had five times the risk of drug abuse or dependency (relative risk=5.21, p<05). Although these researchers acknowledge the covariance of alcohol and tobacco use during pregnancy, they merely mention that alcohol was examined as a potential interactive variable, as was coffee consumption.  Inattention, impulsivity and attention deficit hyperactivity disorder: Consistency was also found in the results of seven separate studies linking the specific behavioural outcomes, attention problems and impulsivity, with prenatal exposure to tobacco. Five teams of U.S. researchers (Day et al., 2000; Leech et al., 1999; Milberger et al., 1996; Naeye & Peters, 1984; Streissguth et al., 1984), Canadian researchers (Denson et al., 1975) and British researchers (Nichols & Chen, 1981) have examined these relationships, some using cohortsfromnormal  65  populations or prenatal clinics and others examining the backgrounds of those with attention deficit problems. One study (Landesman-Dwyer et al., 1981) found results inconsistent with these studies.  Day et al. (2000) found impulsivity to be associated with the combination of both pre- and postnatal tobacco exposure, in their study of 3-year oldsfroma low-income, urban U.S. population (n=672). Postnatal exposure, but not prenatal exposure, predicted attention problems. Day et al. (2000) accounted for prenatal alcohol exposure in their analyses, as well as many other variables that could potentially affect behaviour.  Naeye and Peters (1984), using the Boston site (n=T2,150) of the U.S. Collaborative Perinatal Study, found hyperactivity and shortened attention span scores for 7-year old children whose mothers smoked during pregnancy. Lower attention span scores were attained by children prenatally exposed either moderately (1-10 cigarettes daily) or heavily (over 10 cigarettes daily). A dose-response was found for hyperactivity, with significantly higher levels of hyperactivity found in children heavily exposed. One strength of the study was its control of genetic and child-rearing practices through the pairing of children in the same family, only one of whom was subjected to prenatal tobacco exposure. Unfortunately, the study did not control for prenatal alcohol exposure.  Milberger et al. (1996) studied 140 boys, aged 6-17 years, with attention deficit hyperactivity disorder (ADHD), comparing them to a normal cohort. Their results suggested that maternal smoking during pregnancy is ariskfactor for ADHD, even when controlling for socioeconomic  66  status, parental I.Q. and parental ADHD status (Milberger et al., 1996). ADHD was measured according to DSM-IU-R criteria and mothers were considered to be smokers if they smoked at least a pack a day for at least 3 months during pregnancy. The limitation of this study is that smoking status was measured retrospectively by maternal recall and there was no control in the analyses of either prenatal alcohol use or postnatal tobacco use. Most other studies consider half a pack a day to be "heavy" smoking.  Leech et al. (1999) studied attention and impulsivity problems associated with prenatal exposure to various substances, including tobacco. Their cohort consisted of 783 childrenfroman urban (Pittsburgh) prenatal clinic, selected in 1983. They found that prenatal tobacco exposure predicted inattention when they tested the children at age 6 years. They controlled for various demographic and social environmental factors, noting that it was difficult to separate the effects of pre- and postnatal tobacco exposure because of the high correlation between the two stages of exposure.  Streissguth et al. (1984) also showed prenatal exposure to tobacco was statistically significantly associated with poor attention and poor orientation amongst a cohort of 452 four-year-olds who were part of a longitudinal prospective study in Seattle, described previously. Although the main interest of the researchers was the effects of prenatal alcohol exposure, they also collected data on the use of tobacco and other drugs during pregnancy and reported on the effect of prenatal tobacco exposure at least in theirfirstfew years of reporting. They tested vigilance, reaction time and inattentiveness. Children of mothers who were heavy smokers scored statistically significantly . lower on a test of vigilance measuring both errors of commission (indicating impulsivity) and  67  errors of omission (indicating inattention) and for poorer orientation (that is, not on task). The exposed children were not affected in their reaction time. These researchers used the same tests for children prenatally exposed to alcohol and the findings were statistically significant when the tests for alcohol exposed children were adjusted for tobacco exposure, or visa versa, the tests for tobacco exposed children were adjusted for alcohol exposure. Thefindingsfor all tests were controlled for maternal education, prenatal nutrition and caffeine exposure and child's birth order. This study was one of the few found in which prenatal alcohol use was controlled. In addition, they measured nicotine exposure more accurately than many of the retrospective studies by computing the nicotine usefromthe number of cigarettes smoked daily multiplied by nicotine in the brand. Another important aspect of the study design was the selection of subjectsfroma population of children considered normal, that is, they were not selected because of abnormalities.  In contrast to the above studies, Landesman-Dwyer et al. (1981) did notfindthe same results as Streissguth et al. (1984), described above, even though they tested the offspring of the same middle-class cohortfromSeattle. The measures to test behaviour differed between the two studies. The smoking variable was treated dichotomously, according to whether the mother smoked during the fourth month of pregnancy. The methods of measuring behaviour at age 4 years were naturalistic observations in the home by the researchers and mothers' perceptions of behaviour. Results showed no difference between the exposed and non-exposed groups in the observations of behaviour, such asfidgeting,focused attention at various times such as during story time or at mealtime. Maternal smoking was related to only a few temperament measures (such as persistence or intensity). The authors suggested that their results may have been different  68  from the studies that did not measure or control for alcohol (such as Dunn et al., 1977; Butler & Goldstein, 1973; and Nichols & Chen, 1981). On the other hand, their results also differed from those of Streissguth et al. (1984) (see above) who were especially interested in measuring alcohol consumption during pregnancy.  The poorer orientation and attention that Streissguth et al. (1984) found was consistent with the results of Canadian researchers Denson et al. (1975). They compared 20 (18 boys, 2 girls) children diagnosed with "hyperkinetic reaction" with two matched control group, one described as normal and one with dyslexia. Subjects were ages 5 to 15 years. Both current smoking history and retrospective data on smoking during pregnancy were collected during interviews with mothers. The results showed statistically significant differences between the groups. The hyperkinetic group were exposed to smoking both prenatally (at twice the rate as the control groups) and postnatally (at three times the rate as the control groups). Unfortunately, no data were reported on prenatal alcohol use.  Nichols and Chen (1981) also found poor orientation and attention deficits amongst the children of smokers They reported the extensive work of the U.S. National Collaborative Perinatal Project (NCPP) in which 300 prenatal and postnatal variables were examined in a cohort of 29,887 children at ages 4 and 7 years. Maternal smoking during pregnancy was among the 10 most important variables found to reflect later hyperactivity and impulsivity. After controlling for socioeconomic and other factors, their results showed that those who were prenatally exposed to tobacco had statistically significantly higher results on tests of poor orientation and attention.  69  In the Ottawa Prenatal Prospective Study, predominately middle-class subjects volunteered for a study in which they were interviewed many times during and after pregnancy, their medical records were examined and their offspring (n*250) were tested at various ages for behavioural outcomes of prenatal exposure to tobacco, marijuana or alcohol. In infancy, prenatal cigarette exposure was found to be associated with increased tremors and poorer auditory habituation (Fried & Makin, 1987). At ages 4 to 7 years, prenatal cigarette smoking was found to be associated with increased activity level and increased errors of commission (that measure impulsivity) in auditory tasks, and to a lesser degree, in visual tasks (Kristjansson et al., 1989). Because of the correlation between pre- and postnatal tobacco exposure, prenatal marijuana exposure, family income and maternal education, the analyses controlled for these variables. Kristjansson et al. (1989) found only a relationship with errors of commission, while in the Seattle study that had a cohort with similar socioeconomic background, both errors of commission and errors of omission (measuring attention) were associated with daily tobacco use during pregnancy (Streissguth et al., 1984). While not identical in their results, the two studies indicate problems with vigilance and impulsivity in children prenatally exposed to tobacco. The Ottawa researchers noted that attention and impulsivity problems appeared to be more pronounced amongst children prenatally exposed to tobacco than to either moderate alcohol exposure or marijuana exposure.  One further report of interest from the Ottawa research compared children pre- and postnatally exposed to tobacco on learning deficiencies (reported below) and also on mothers' reports of their child's behaviour (Makin et al., 1991). Using a parents' behavioural scale (Conners Parent Questionnaire), children who were exposed both pre- and postnatally to tobacco rated higher on all  70  scales than those of the nonsmoking group, indicating more behavioural problems.  Conduct disorder; Conduct disorder has been described as a "devastating condition" both because of the harm youth inflict on others and on property and because of theriskof other forms of psychopathology (Loeber et al., 1995). The results of a recent study suggested that maternal smoking during pregnancy was strongly linked with conduct disorder in male offspring (Wakschlag et al., 1997). The researchers controlled many confounders considered to be potentialriskfactors for conduct disorder, including socioeconomic status, low birth weight, pregnancy complications, family factors (such as marital status), psychopathological characteristics of the biological parents, parenting styles and demographic variables. The sample was a cohort of 177 U.S. boys, ages 7 to 12 years, referred to a clinic. Prenatal alcohol and use of one or more illicit drugs were treated dichotomously according to whether they were used. Results of a bivariate analysis showed that mothers who smoked more than half a pack of cigarettes daily during pregnancy were significantly more likely to have a child diagnosed with conduct disorder, as measured by DSM-III-R criteria. Furthermore, when a logistic regression was conducted, smoking more than half a pack daily, socioeconomic status and maternal antisocial personality were independent predictors of conduct disorder. Regression analyses adding all the significant independent predictors indicated that smoking more than half a pack daily during pregnancy, maternal age, little parental supervision and harsh discipline were each significant independent predictors. Interestingly, use of alcohol or illicit drugs, or substance abuse were not shown to be independent predictors.  71  Loeber et al. (1995) concluded that A D H D predicted an early onset of conduct disorder and that low socioeconomic status, oppositional behaviour and parental substance abuse were key factors in contributing to the progression to conduct disorder, when controlling for various other factors. Unfortunately, only current history of paternal and maternal alcohol and drug use were measured.  Delinquency and criminal behaviour: Related to the research on conduct disorder are studies linking maternal smoking during pregnancy and juvenile delinquency and criminal behaviour among adult male offspring. Four studies were found, three from Scandinavia and the other from the U.S.A.  Brennan et al. (1999) studied 4,169 males born 1959-1961 in Copenhagen in a prospective longitudinal study. Mothers self-reported on the quantity of tobacco used during their third trimester of pregnancy and on the use of various prescription drugs during pregnancy, such as antihistamines, diuretics, antiepileptics, psychopharmaceuticals, antibiotics, analgesics, hormone treatment. History of alcohol use was not included in the data collection. The criminal records were checked when the male offspring of these women were 34 years old. Results indicated a dose-response relationship between arrests for nonviolent and violent crimes and amount of smoking by the mother during pregnancy. The relationship persisted when related variables, such as paternal criminal history, were controlled during the analysis.  A population-based cohort (n= 12,058) born in northern Finland in 1966 was the study population for two studies. Both studies excluded female children because of their low level of arrest or  72  recorded criminality. Data on alcohol consumption were not collected, because "here it was a fairly minor problem among Finland women in 1966" with only 2.7 litres per capita sold annually and a ratio of 11:1 or 12:1 male:female consumption for many years (Rantakallio et al., 1992, p. 1112). Smoking behaviour was categorized into three groups: none, stopped before pregnancy and smoked daily throughout entire pregnancy. Rantakallio et al (1992) examined delinquency at age 22 years, showing the incidence of delinquency amongst males (N=5,966) in the cohort was 10.3% amongst those prenatally exposed to tobacco and less than half that amount (4.6%) for their non-exposed counterparts. A weak relationship with respect to timing was observed, so that those whose mothers stopped smoking in the first trimester had a slightly lower rate of delinquency. The association between prenatal exposure and delinquency remained even after controlling for various possible social, family and demographic confounders. The researchers concluded that a clear association between prenatal tobacco exposure and juvenile crime existed, and could not be explained by other social or demographic variables.  Using the same cohort, Rasanen et al. (1999) studied the relationship between criminal behaviour and prenatal tobacco exposure. Comparing sons of mothers who did not smoke during pregnancy with those whose mothers smoked throughout the pregnancy, the latter group had more than double the risk of having committed a violent crime or having repeatedly committed crimes by age 28 years. The same relationship was not shown for nonviolent crimes.  A further contribution was made by Gibson and Tibbetts (1998) who used the cohortfromthe NCPP, the longitudinal U.S. study mentioned above, to examine the behaviours of a cohort of 832  73  inner-city Philadelphian Afro-American youths born between 1959 and 1966. They found that the combined effect of prenatal exposure to tobacco and low Apgar scores (1 minute) had a significant influence in predicting offending behaviour, measured according to whether they had criminal history in police records between the ages of 10 and 22 years. There was an additive effect, the independent effect of either variable alone was not found to be predictive of offending behaviour. Unlike the Scandinavian studies, youth of both genders were included. Prenatal alcohol use was omittedfromthe study report.  These studies of older children may be less relevant to the present thesis, but the results were included in this review because of the emerging pattern of behavioursfrompreschool to adulthood. The results may suggest that possible effects may persist throughout childhood nd continuing into adulthood, or they may suggest that similar behaviours are found at various ages amongst those prenatally exposed to tobacco. They could suggest that effects at later ages are the cumulative product of other effects at earlier ages, or that the same effects are expressed in different ways at different ages. Wakschlag et al. (1997) found a relationship with higher levels of prenatal exposure to tobacco and conduct disorder amongst boys in adolescent years (ages 12 to 17). Rantakallio et al. (1992) found an association between prenatal tobacco exposure and juvenile delinquency. Rasanen et al. (1999) found a relationship between prenatal exposure to tobacco and criminal behaviour up to age 28 years. Brennan et al. (1999) found the same relationship to criminal behaviour amongst subjects at age 34 years. The results of these studies were consistent, even controlling for a variety of potentially confounding personal, family and social variables. One of the studies (Rasanen et al., 1999) did not discriminate the smoking behaviour according to dose,  74  another did not measure a dose-response gradient (Rantakallio et al., 1992). The other two studies were consistent in their findings that increased dosage was related to the behaviours under study. Wakschlag et al. (1997) found that prenatal exposure to more than half a pack daily was associated with increased risk for conduct disorder in boys. Brennan etal. (1999) found that women who smoked more than 20 cigarettes a day were more likely to have male offspring who were arrested during adulthood for nonviolent or violent crimes.  One further study was included in this review, but not mentioned above because of the unique outcome variables examined, namely, affective and personal functioning. The researcher found no significant differences in exposed and non-exposed children at age 10 years on teachers' and parents' ratings of depression, happiness, school social behaviour, self-esteem and peer-rated popularity (Lefkowitz, 1981).  To summarize, the 24 human studies reviewed showed considerable, although not complete consistency, both in their results and in the types of behavioural problems found to be associated with prenatal exposure to tobacco. The results were consistent in Australia, New Zealand, and in North American and European countries, although many of the studies that measured inattention or impulsivity took place in North America, while those that examined delinquency and crime took place in Scandinavia and the U.S.A. The consistency between the research of two studies in particular (Fergusson et al., 1993 and Brennan et al., 1999) led one of the researchers to write a commentary on the relationship between prenatal tobacco exposure and antisocial behaviour (Fergusson, 1999). Observing the strength of the studies in the large representative samples in  75  both Denmark and New Zealand, the commentary noted the consistency in showing increased risk for externalizing behaviours that extend over a hfe-time and the resiliency in the relationships in spite of controlling for many social, family and individual factors. Behaviours appeared to be specific antisocial behaviours and not reoccurrence of the same behaviour over different times. He suggested further that the evidence meets the criteria used as evidence of causal relationships (Susser, 1973), including replicability of results. Fergusson (1999) also stated that it would be unwise to draw strong conclusions at this point given that underlying mechanisms have not been fully established, that genetic confounding could be possible, and these need further examination.  Not mentioned by Fergusson was the fact that prenatal exposure to alcohol has been found to be associated with inattentivity, impulsivity and other outcomes, including secondary consequences such as incarceration for delinquency, to which prenatal tobacco has been associated in the above studies. Maternal alcohol use and maternal tobacco use during pregnancy have been found to covary, as do maternal education, socioeconomic status and other variables (see below). Unfortunately less than half the reviewed studies attempted to control for prenatal alcohol use and many of these were brief in their reporting of measurement and treatment of this confounding variable. Rantakallio et al. (1992) point out that this may not be a problem since their large cohort was unlikely to drink alcohol, given the prevalence of alcohol use among women in Finland in the 1960's. It may be that this tradition of non-use of alcohol at the time of data collection existed in other countries as well, although this was not the case in Canada and the U.S.A. where some of the studies did take alcohol use into consideration. Even without direct measures and analytic control of alcohol, the consistency of results may provide opportunities to draw conclusions about the  76  relationship between prenatal exposure to tobacco and behaviour, especially respecting attention, impulsivity and externalizing behaviours in general, leading to more aggressive behaviours upon approaching adulthood. There was only one exception to the consistency of results, that being one of the reports of the Seattle study of 4-year olds (Landesman-Dwyer et al., 1981) that included alcohol use in its data collection. In that analyses, no association was observed between prenatal tobacco use and attention problems.  2.3.2.H Cognitive development, academic abilities or learning problems Studies that tested cognitive development and academic abilities associated with prenatal tobacco exposure found diverging results. Many studies failed to find a relationship between exposure and mental development (Trasti et al., 1999; Niemela & Jarvenpaa, 1996; Eskenazi & Trupin, 1995; McGee & Stanton, 1994; Baghurst et al., 1992; Roeleveld et al., 1992; Fergusson & Lloyd, 1991; Lefkowitz, 1981; Hardy & Mellits, 1972). Others found statistical significance on at least some tests when measuring the differences between exposed and unexposed groups when children were tested many years postnatally (Fried et al., 1998, 1997, 1998b; Olds et al., 1994a; Sexton et al., 1990; Naeye and Peters, 1984; Rantakallio, 1983; Nichols & Chen, 1981; Fogelman, 1980; Butler & Goldstein, 1973). The studies are reviewed below.  Preschool age Trasti et al. (1999) used various age-appropriate tests to assess the motor and mental abilities of randomly selected Scandinavian children (n=376), half of whom had been prenatally exposed daily to tobacco. No measure of dose was recorded. At age 13 months, both groups performed equally  77  well on the Bayley Scales of Infant Development. At age 5 years, cognitive development was not found to have an association, when measured by the Wechsler Preschool and Primary Scale of Intelligence (WPPSI-R) after controlling for maternal education. A small adverse effect on balance at agefiveyears was observed. It is possible that the lack of measurement of dose could have influenced the results. Alcohol use during pregnancy was not measured.  Similarly in Australia, Baghurst, Tong, Woodward and McMichael (1992) failed tofindan association between exposed and non-exposed children (n=548), followedfromages 1 to 4 years, after adjusting the analyses for socioeconomic status, home environment and mother's intelligence. The children were tested using the Bayley Scales of Infant Development and McCarthy Scales of Children's Abilities for various indicators of cognitive development, including memory, problemsolving abilities, language and speech development and motor performance. Unfortunately, dosage of exposure was not reported and women were considered non-smokers if they had ^5 cigarettes during their pregnancy. Alcohol use during pregnancy was recorded and treated as a confounding variable.  A Finish study (Niemela & Jarvenpaa, 1996) also found no statistically significant difference between tobacco exposed (prenatally) and not exposed groups of 4 and 5 year-old children on test scores of cognitive development, after adjustment for confounding variables. They reiterated the importance of lifestyle and social background, indicated by maternal education and parental factors, as determinants of children's cognitive development. Unfortunately, neither dosage of tobacco nor prenatal alcohol use was mentioned in reporting the study.  78  A U.S. study that had the most precise and valid measure of nicotine exposure (measurement of maternal serum cotinine levels during pregnancy) compared 5-year old children of women who smoked only prenatally or only postnatally or both pre- and postnatally. Assessments used were the Raven Progressive Coloured Matrices (measuring nonverbal reasoning and the ability to form comparisons and organize spatial perceptions) and the Peabody Picture Vocabulary Test (measuring receptive vocabulary). Scores for the three groups were not statistically significantly different when the analyses included confounding variables such as prenatal alcohol exposure, preschool attendance, parental education and various other factors (Eskenazi and Trupin, 1995).  The four studies cited above all failed tofinda significant relationship between prenatal tobacco exposure and deficits in learning or cognitive development. These studies were inconsistent with the following three studies that found associations.  Olds, Henderson and Tatelbaum (1994b) found that the 3- and 4-year old children of mothers who smoked had statistically significantly lower scores on the Stanford-Binet test of intelligence than those not prenatally exposed to tobacco. Their sample (n=400) was considerably different than studies described above in that it was comprised of womenfroma semi-rural area of New York state who entered the study at the 30* week of pregnancy and who were predominantly poor, unmarried or teenagers. In addition, the researchers defined prenatal smoking exposure as *10 cigarettes daily throughout pregnancy, differing from the two studies above that defined smokers with considerably less precision. Olds et al. (1994a) further developed their work by evaluating the effects of an intervention in which there were nurse home visitations during pregnancy or  79  during thefirsttwo years of the child's life. These interventions during or after pregnancy had a positive influence in reducing the adverse effects of prenatal exposure to tobacco, a somewhat similar conclusion to that of Baghurst et al. (1992) who found that adjusting their analyses for variables that were related to stimulating home environments was found to alter the association between prenatal exposure and cognitive development. Another important aspect of this study (Olds et al., 1994a, 1994b) was the examination of timing of tobacco exposure. Separate analyses of covariance were conducted for exposure beginning at gestational mean age of 17 weeks, at 34* week of gestation, and at postnatal exposure at age 4 years. The children's scores were measured at between ages 3 and 4 years. All three exposed groups had lower scores than the non-exposed group, with the largest difference found between the non-exposed and those exposed starting at the 34* week of gestation and the smallest difference noted between the non-exposed children and exposed only postnatally.  Sexton, Fox and Hebel (1990) also found a relationship between prenatal tobacco use and cognitive development, as tested at age 3 years. They used one of the same scales as Baghurst et al. (1992), the McCarthy Scales of Children's Abilities, as well as the Minnesota Child Development Inventory. They also defined smoking as £ 10 cigarettes daily. Interesting, theirs was not a comparison of those exposed and those not exposed, rather, it compared the offspring of those who stopped smoking at, or prior to, the 18* week of gestation. On the general scale and all subscales, the children of those who continued to smoke had significantly lower test scores. The researchers carefully considered the social, educational and other background variables.  80  In the only Canadian longitudinal study, children who were born between 1980 and 1983 to predominantly low-risk middle-class women in the Ottawa area were studied (n=698), examining prenatal exposure to alcohol, tobacco and marijuana. Results have been reported for cognitive effects at various ages,from12 months to 12 years. Prenatal tobacco exposure was observed to be associated with lower mental scores and altered responses to auditory items at ages 1 to 2 years (Gusella & Fried, 1984), with poorer language development and lower cognitive scores at ages 3 to 4 years (n=190) (Fried & Watkinson, 1990) and with lower cognitive and receptive language scores at ages 5 to 6 years (n« 136) (Fried et al., 1992b), after controlling for confounding factors including parental education, prenatal marijuana exposure and to moderate amounts of alcohol. _ Children were administered a battery of tests, including at ages 3 to 4 years the Reynell Development Language Scales, the Pegboard Test and Tactile Form Recognition Task and at ages 3, 4, 5 and 6 years, the McCarthy Scales (Children's Abilities, General Cognitive and Verbal Subscale) and the Peabody Picture Vocabulary Test. Smoking was categorized according to none, light or heavy (>15 mg nicotine a day, or approximately one package per day). Home environment was controlled as a variable in the analyses, since it was considered to be related to cognitive development. The verbal subscale of the McCarthy Scales was primarily responsible for discriminating the exposed and non-exposed groups for both 3 and 4 year olds. A negative doseresponse performance was observed, with the heavily exposed group having poorer verbal skills than either of the other groups, both with the 3 and 4 year olds and with the 5 and 6 years olds. The researches noted that, at least for the 5 and 6 year olds, the smoking exposed and the nonexposed groups were significantly different, but the former group means were still above the norm for their respective ages.  81  In summary, the three studies (Olds et al., 1994a, 1994b; Sexton et al., 1990; Fried & Watkinson, 1990; Gusella & Fried, 1984) that defined smoking as ^ 10 cigarettes daily found a relationship between cognitive development at preschool age and prenatal tobacco exposure, while four other studies (Trasti et al., 1999; Niemela and Jarvenpaa, 1996; Baghurst et al., 1992) that did not measure dosage or were vague about their measurement of prenatal tobacco exposure did not find a relationship between exposed and non-exposed groups. In contrast, the study (Eskenazi and Trupin, 1995) that had the most precise measurement of nicotine exposure failed tofinda doseresponse relation between prenatal exposure and cognitive development when confounding variables were accounted for in the analyses.  School age and beyond (ages 6 to 23 years) Studies of children between the ages of 6 and 12 years also showed contradictory results, with three studies (Fried et al., 1998, 1997, 1992b; Naeye & Peters, 1984; Nichols & Chen, 1981) showing some lower intellectual functioning, while other studies (McGee & Stanton, 1994; Roeleveld et al., 1992; Fergusson & Lloyd, 1991; Rantakallio, 1983; Lefkowitz, 1981; Hardy & Mellits, 1972) observed no differences between exposed and non-exposed children in general intellectual functioning after confounding factors were included in the analyses. Although the general test showed no differences, some areas of test scores discriminated the exposedfromthe non-exposed. Those who used assessments with subscales testing different components of academic abilities had some consistency in their results, as children exposed prenatally appear to be effected in their language abilities, but only one study showed deficits in arithmetic or mathematics test scores.  82  Three groups of researchers (Naeye & Peters, 1984; Nichols & Chen, 1981; Hardy & Mellits, 1972) each analysed and reported on the intellectual functioning of children at age 7 years who were part of the 12-site U.S. Longitudinal National Collaborative Perinatal Project (NCPP) in which 58,000 pregnancies were followed prospectively beginning in 1959. Naeye and Peters (1984) tested the children from the Boston site (n=12,150), administering the Wide Range Achievement Test (WRAT) assessment of spelling, reading and arithmetic. In an attempt to overcome the criticism that child-rearing practices and genetic factors might be potential confounders, childrenfromthe same family and gender were matched, with only one sibling from each family having been prenatally exposed to tobacco. Children whose mothers smoked throughout the pregnancy had significantly lower scores on the reading and spelling subtests, although the differences were slight with only 3% lower spelling scores and 4% lower reading scores than those of the children of non-smokers. There was no difference between the children of smokers and non-smokers on the arithmetic scores. As noted above when discussing research with the same cohort with respect to behavioural outcomes of prenatal tobacco exposure, Naeye and Peters (1984) were unable to control for alcohol exposure. Their analyses showed that reading and spelling were associated with a wide range of other variables including maternal age, family income, maternal and paternal education, birth order and gender.  These results contradict two other studies using different sites of the NCPP. Hardy and Mellits (1972) found no significant difference between the group exposed to £ 10 cigarettes daily and the non-exposed group (n=88) in general intellectual functioning when tested at ages 4 and 7 years. They used the WRAT as an assessment tool, and didfindscores on the spelling and reading  83  subscales to be significantly different between the two groups Nichols and Chen (1981) initially found significant differences in results on all components of "minimal brain damage," that is, learning difficulties, hyperkinetic impulsive behaviour (hyperactivity, impulsivity, short attention span and emotional lability) and neurological signs (e.g., coordination, gait, reflexes) in children whose mothers smoked 20 cigarettes daily or who smoked heavily (^40 cigarettes daily). When multivariate analyses were conducted controlling for socio-environmental factors, only hyperkinetic impulsive behaviour was found to be ariskfor those who were prenatally exposed to tobacco (Nichols & Chen, 1981). Women who smoked more than 20 cigarettes daily were 28% more likely to have children with hyperactivity-impulsivity.  Similar to the American study, a large retrospective British study, the National Child Development Study, followed all 16,000 children born during a 3-week period in 1958 in Scotland, England and Wales (Fogelman & Manor, 1988; Fogelman, 1980; Butler & Goldstein, 1973; Davie et al., 1972). Doses were categorized as none, "medium" (10 cigarettes per day) or "heavy" (>10 cigarettes per day). Reading ability scores were significantly different between the exposed and non-exposed groups at age 7 years (Davie et al., 1972), and again at age 11 years (Butler and Goldstein, 1973). At age 7 years, differences equivalent to 4 months of reading age were found between the heavily exposed and the non-exposed groups (Davie et al., 1972). At age 11, there was no dose-response shown in reading scores, but statistically significant differences were found between the exposed and non-exposed groups. Mathematics scores did not differ between the two groups at ages 7 and 11 years. At age 16 years, an association was observed between tobacco exposure and both reading and mathematics (Fogelman, 1980). At the older age, a small dose-response effect was  84  observed, with those exposed to 10 or fewer cigarettes scoring lower than those not exposed and those whose mothers smoked "heavy" doses scoring lower than those whose mothers smoked at "medium" levels. Analyses at each of these ages controlled for various confounding factors, including gender, birth weight, gestational age, social class (based on father's occupation) at birth and at age 16 years and number of siblings. Prenatal alcohol exposure was not included in the list of potentially confounding variables.  The same cohort was followed again to find that at age 23 years the prenatally exposed group did not achieve the same educational levels as their non-exposed counterparts, even after controlling various confounding factors such as social class (Fogelman and Manor, 1988). The researchers concluded that the relationship was observed to be even stronger at age 23 years than earlier. This may not be surprising, given that the exposed group from the same cohort had been doing relatively poorly on at least some academic subjects at ages 7, 11 and 16 years (Fogelman, 1980; Butler and Goldstein, 1972, 1973). This is the only study found to date to indicate that depressed intellectual development has persisted over a very long term.  In the Canadian longitudinal study, mentioned above when discussing preschool children, testing for cognitive effects also was completed at ages 9 to 12 years (n=131) (Fried et al., 1998, 1997). Results showed a dose-dependent association between prenatal tobacco exposure and lower language and reading scores, after analyses controlled for confounding variables. This is similar to the researchers' observations at earlier ages. The interesting more recent finding was that lower reading scores were particularly auditory-related aspects of the 7 assessment tools used, so that  85  exposure appeared to have less influence on nonverbal auditory performance than on higher language functioning. This led the researchers to conclude that "for reading, the phonological rather than the orthographic sphere was more affected" and that prenatal tobacco exposure alters a range of behaviours dependent on auditory functioning (Fried et al., 1997, p. 182). Significant differences between the exposed and non-exposed groups were not found on the reading subscale of the WRAT assessment. The researchers assessed only language and reading abilities and omitted arithmetic or mathematics in their testing. Their methods used only some of the subtests of the assessment tools, for example, only the reading subtest of the WRAT test.  A retrospective study (McGee & Stanton, 1994) of New Zealand women (n=1037) self-reporting their tobacco use was unable to establish a relationship between prenatal exposure and children's language skills or general cognitive development. Smoking was categorized according to doserelated levels. Any initial effect found in the relationship was lost when many confounding factors were added to the analyses, with the exception of a weak relationship between prenatal exposure and I.Q. scores at age 5 and reading at age 9, both only for girls.  Another New Zealand longitudinal study (Fergusson & Lloyd, 1991) also showed no relationship between daily prenatal tobacco exposure and general tests of intelligence, reading, writing and mathematical abilities for children tested at ages 8 to 12 years, after the analyses were adjusted for confounding variables. Children exposed to at least 20 cigarettes daily scored significantly lower on the tests. Analyses also showed that relatively disadvantaged home environments were more likely to influence score results than prenatal exposure to tobacco. This was one of the few studies  86  found in which considerable effort went towards recording dosage, timing (exposure at each trimester) and frequency of exposure. Alcohol use was not included in the data collection.  U.S. researcher Lefkowitz (1981) tested children at age 10 years and found no significant difference between the offspring of smokers and non-smokers, leading her to conclude that if these children survive the neonatal period, they will not suffer long-term effects with respect to intellectual functioning.  Linking data collected during a retrospective longitudinal study of a cohort of children born in two northern states in Finland with a cross-sectional study of children with mental retardation, cerebral palsy or epilepsy, it was observed that neither the number of handicapped children nor the numbers of handicaps differed between the exposed and non-exposed groups (Rantakallio & Koiranen, 1987). The same researcher examined the total cohort (n=1819) at age 14 years, finding that those exposed prenatally to tobacco had on average poorer school performance than those not exposed (Rantakallio, 1983). Controlling for various confounding factors led to the conclusion that although prenatal smoking exposure did have an effect on mental development, various other socio-biological factors were also considered to beriskfactors, including indirect exposure through paternal smoking, and that many of these otherriskfactors are co-linear with maternal smoking during and after pregnancy.  In another European cross-sectional study of 322 children with mental retardation who were admitted for care anytime up to age 15 years, prenatal tobacco exposure was not observed to be a  87  risk factor (Roeleveld et al., 1992).  As mentioned, in the discussion on studies of younger children, there is also inconsistency in the results regarding overall intellectual functioning amongst older children and adolescents who were exposed to tobacco prenatally. The inconsistency could be a function of measurement since some of the studies that observed no differences between the exposed and non-exposed may have included very low levels of exposure in their "exposed groups." Some consistency, although tentative, may be emergingfromthese studies with respect to types of deficits in intellectual functioning. In six of the studies, deficits in reading or language were observed, in some even when overall intellectual functioning did not discriminate the exposed and non-exposed groups (Fried et al., 1998, 1997, 1992b; McGee & Stanton, 1994; Naeye & Peters, 1984; Fogelman, 1980; Davie et al., 1972; Hardy & Mellits, 1972; Goldstein, 1971). For example, McGee and Stanton (1994) found that confounding variables accounted for any initial relationships they observed between intellectual functioning and prenatal tobacco exposure, except for a relationship, albeit weak, between prenatal exposure and I.Q. scores at age 5 and reading at age 9, both only for girls. Both Naeye and Peters (1984) and Hardy and Mellits (1972) found that spelling and reading scores, but not arithmetic scores, discriminated the exposed and non-exposed groups when using the WRAT to assess the children's abilities at age 4 or 7 years. Fried et al. (1998, 1997) only tested reading or language abilities during later testing of the Ottawa cohort, that is, ages 9 to 12 years. They observed lower test scores in these areas associated with prenatal tobacco exposure, especially on auditory-related aspects of the reading tests. Fried et al. (1998, 1997) attempted to be more precise in their testing of various components of language and reading abilities, finding,  88  for example, that more complex tasks such as reading comprehension caused more difficulty for exposed groups compared to non-exposed groups. They also noted that altered auditorydependent aspects of reading caused more problems for the exposed groups, leading them to conclude that altered auditory-based behaviours also discriminated the exposed and non-exposed in testing they had completed during earlier ages of the same cohort (Fried et al., 1998, 1992b, 1990). Children 6 to 11 years old were tested specifically using a central auditory processing task, and it was observed that prenatal tobacco exposure was associated linearly with poorer performance, leading to the conclusion that "the consequences to the child with impaired auditory processing capabilities may be reflected in difficulties in learning and behaviour that have been linked to in utero exposure to cigarettes" (McCartney et al., 1994, p. 275).  Only one study (Fogelman, 1980) observed deficits in mathematics scores when testing 16 year olds amongst a tobacco-exposed group although there was no reporting of deficits in arithmetic amongst younger groups of the same cohort (Butler and Goldstein, 1973; Davie et al., 1972).  2.3.3 Summary and discussion of prenatal tobacco exposure research Consistent evidence was found for an association between prenatal tobacco exposure and fetal and neonatal growth. In addition to retarded growth, the health issue that was most consistently found to be associated with prenatal tobacco exposure was respiratory problems, persisting into childhood.  With respect to behavioural outcomes, studies consistently found an association between prenatal  89  exposure and attentional problems, hyperactivity or impulsivity. These same behavioural characteristics were found well into childhood and adolescence. At later ages, conduct disorder and criminality were found to be associated with prenatal tobacco exposure.  With respect to academic abilities, there was also consistency in the research literature regarding outcomes, in that language, reading or spelling scores in various assessments were found to be lower for those who had been prenatally exposed to tobacco, while mathematics or arithmetic scores, if tested at all, were rarely found to be lower.  The inconsistency in the results of these studies might be, for the most part, caused by weaknesses in the research designs. For example, an early study that did not find a relationship between exposure and inattention in 4-year olds had insufficient numbers of heavy smokers (LandesmanDwyer et al., 1981). Another barrier to gaining better understanding of the effects of prenatal exposure has been the difficulty in discriminating pre- and postnatal exposure since very few women smoke during pregnancy and then stop at the birth of their child.  Few studies were found in which prenatal alcohol exposure and prenatal tobacco exposure were treated as interactive variables, rather, most analyses adjusted for the effects of the other variable. Some longitudinal studies have included prenatal alcohol exposure, prenatal tobacco exposure and other drugs used during pregnancy such as cocaine or marijuana, in examining outcomes on children in Pittsburgh (Leech et al., 1999), in Seattle (Streissguth et al., 1984) and in Ottawa (Fried etal., 1999, 1998, 1997, 1992a, 1992b; Fried, 1993; Makin et al., 1991; Fried & Watkinson,  90  1990). The importance of examining the prenatal use of these drugs in the same longitudinal studies is that alcohol, at least, has also been found to be related to attentional problems, hyperactivity and impulsivity, as well as to lower scores on assessments of academic abilities. Any research examining the outcomes of prenatal tobacco exposure that does not take into consideration the effects of alcohol (or visa versa) may be suspect in its conclusions since the use of alcohol and tobacco covary. Researchers of the Pittsburgh and Ottawa longitudinal studies appeared to fully understand the potential risks of the alternative drugs under study. This may not have been the case in data collections in earlier studies in which data were collected regarding prenatal tobacco use well before there was any understanding of the potential risks of prenatal alcohol exposure; for example, the British National Child Development Study, 1958 Cohort (Butler & Goldstein, 1973; Davie et al., 1972; Goldstein, 1971).  Rush and Callahan (1989), in their review of prenatal tobacco exposure, criticized the majority of researchers for not adequately controlling for socioeconomic variables, that are related to both the dependent (developmental outcomes) and independent variable (prenatal tobacco exposure). Inconsistencies were found with respect to analytic control of additional variables, some of which have been observed to be correlated with prenatal tobacco use (e.g., as seen below, low maternal education or alcohol use during pregnancy). Thus, there is the possibility that some results were spurious. Reviewers of the research literature (Rush & Callahan, 1989; Landesman-Dwyer & Emanuel, 1979) conclude that the observed abnormalities cannot be assumed to be caused by parental cigarette smoking given that so little attention has been given to the differences in sociodemographic and psychological backgrounds between those who smoke during pregnancy  91  and those who stopped during pregnancy or who have never smoked. Landesman-Dwyer and Emanuel (1979) suggested that researchers should have controlled for the variables of parental intelligence and psychological status. No more recent reviews have been found and it appears from the literature review for this thesis that this criticism still holds, likely because of the major methodological problems with the analytical comprehensiveness of any study including assessments of many demographic and environmental variables in a single study.  In spite of the criticisms, patterns in the research were found to emerge, especially with respect to behaviour. The research on cognitive development was less consistent, except in the case of language development or reading skills that may be related in turn to auditory development. In a number of studies examining cognitive outcomes, any initial relationship found between prenatal tobacco exposure and deficits in areas of cognitive development were no longer found to be significant when confounding factors, particularly mother's intelligence or mother's education or other socioeconomic status indicators were included in the analyses (Trasti et al., 1999; Niemela and Jarvenpaa, 1996; Eskenazi and Trupin, 1995; Eskenazi and Bergman, 1995; McGee and Stanton, 1994; Fergusson et al., 1993; Baghurst et al., 1992).  To further underline the argument about the importance of other variables, interventions have been found to overcome the relationship between poor cognitive development scores and prenatal exposure to tobacco. Olds et al. (1994a) found during a randomized clinical trial that intellectual impairments observed in children prenatally exposed to tobacco were offset by a comprehensive prenatal home-visitation program that concentrated on improving lifestyle behaviours (such as diet  92  and reducing smoking during pregnancy) of women who smoked. Lifestyle behaviours have been found to be correlated with maternal education (see discussion below).  Therefore, any conclusions regarding prenatal tobacco exposure and cognitive effects would have to be made with caution, considering whether the outcomes are influenced by maternal education and related factors such as early childhood stimuli at home. Any conclusions regarding prenatal tobacco exposure and behavioural effects should also be considered cautiously, given that all researchers have not controlled for prenatal alcohol exposure, that is associated with attentional problems and hyperactivity. Longitudinal research in which the histories of both alcohol and tobacco use during pregnancy is part of the data collection and analyses (e.g., Fried in Ottawa and Streissguth in Seattle, along with their colleagues) should be given more weight when making conclusions about long-term effects of prenatal tobacco exposure.  2.3.4 Postnatal exposure to tobacco Two reviews of the literature (Windham et al., 1999; Rush & Callahan, 1989) and a meta-analysis (Windham et al., 1999) on the effects of passive smoke (that is, not by the mother during pregnancy) on children's physical development conclude that postnatal exposure to environmental tobacco smoke has an effect equivalent to prenatal exposure in reducing birth weight. Secondhand smoke by exposure to paternal smoking throughout the life of the child was found to be a risk factor in a Finnish study (Roeleveld et al., 1992; Rantakallio, 1983). A group of 14-year olds (n= 12,068), all of whom had been prenatally exposed to tobacco, was matched with a non-exposed group according to age, marital status of mother, parity and place of residence. Exposure to  93  paternal smoking carried nearly as strong an association with retardation in mental and physical development as did exposure to prenatal maternal smoking (Rantakallio, 1983). Children who had been exposed prenatally to tobacco smoke had on average poorer school performance than those not exposed, but other socio-biological factors, including paternal smoking, were also risks for slower mental development (Rantakallio, 1983).  In another study, groups were categorized according to whether their mothers smoked at 17 weeks gestational age, 34 weeks gestational age and postnatally at age 4 years. Compared to a nonexposed control group, the children's I.Q. scores (at average age 3 and 4 years) were lower for all three groups, with the largest difference between exposed and non-exposed being the smoking exposure at the 34th week of gestation and the smallest being postnatal exposure (Olds et al., 1994a, 1994b).  Several other researchers have included postnatal exposure in their studies of the effects of prenatal tobacco exposure with inconsistent results. Causal relationships have been established which show that second-hand smoke has influenced the health and well-being of exposed children (Brondum et al., 1999; Gergen et al., 1998, Strachan & Cook, 1998; Brooke et al., 1989). Postnatal second-hand smoke was found to be an important risk factor for growth (Rona et al., 1985) and for respiratory problems (Stein, 1999).  With respect to behaviour and postnatal tobacco exposure, Williams et al. (1998) followed a large cohort of Australian children (n=5,342), finding that at 5 years of age, postnatal tobacco exposure  94  accounted for 16% of the variance in externalizing behavioural problems, while prenatal tobacco exposure accounted for 25% of the variance, even after controlling for various confounding factors including socioeconomic status.  The results of other research on this issue have been inconsistent, as indicated during the review above on prenatal tobacco exposure. This is not surprising given the varied methods of defining levels of tobacco exposure, measurement of outcomes and analytic control of confounding variables.  95  Table 2.2 Studies of cognitive and behavioural outcomes associated with prenatal tobacco exposure, listed in alphabetical order by first author. Authors, country and age of study sample  Controlled variables  Baghurst, Tong, Woodward and McMichael, 1992  Socioeconomic, mother's intelligence (tested), home environment, prepregnancy smoking, clinical characteristics, prenatal alcohol use, diet  none  Maternal rejection, socioeconomic status, maternal age, pregnancy and delivery complications, use of prescription drugs during pregnancy, paternal criminal history, parental psychiatric hospitalization  none  Gender, birth weight, gestational period, mother's height, social class at birth and at age 16 (indicated by father's occupation with children with no fathers being excludedfromstudy), birth order, number of siblings.  none  Australia (n=548)  Prenatal alcohol exposure measurement  Results: behavioural outcomes  No relationship between exposed and non-exposed groups on Bayley or McCarthy Scales, after adjustment for socioeconomic background and mother's intelligence  ages: 1-4 years  Brennan, Grekin, Mednick, 1999 Denmark ages: adults up to 34 years old  Davie, Butler and Goldstein, 1972; Butler and Goldstein, 1973; Fogelman, 1980; Fogelman and Manor, 1988; Goldstein, 1971  Results: cognitive outcomes or learning problems  Dose-response relationship between amount of maternal smoking during pregnancy and arrests for nonviolent and violent crimes and in particular persistent criminal behaviour (rather than adolescent arrests only). At ages 7,11 and 16 years, reading scores lower for exposed group; at age 16 mathematics scores also lower. At age 23 years, educational achievement higher amongst the nonexposed.  Britain 1958 cohort (n=8200) ages: 7,11,16,23  96  Table 2.2 continued Day, Richardson, Goldschmidt and Cornelius, 2000 U.S.A. (Pittsburgh) n=672 pairs age: 3 years  Alcohol, marijuana, cocaine, family characteristics (siblings, parents' marriage), home environment, life events. Sample drawn from a lowincome urban hospitalbased clinic.  Alcohol was measured (half sample drink 3 or more per week during 3"1 trimester), included in analyses.  Prenatal exposure associated with Oppositional Behavior, Immaturity, Emotional Instability, Physical Aggression and Activity subscales of 3 standard behaviour scales. Largest effect on Oppositional Behaviour. 3 trimester smoking, important. rd  Denson, Nanson, and M c Waiters, 1975 Canada (Sask.) (n=20) ages: 5-15  Dunn, McBumey, Ingram and Hunter, 1977, 1976 Canada (Vancouver) (n=318)  Case control study. Subjects who were diagnosed with hyperkinetic syndrome, plus two groups of matched controls (normal and dyslexic); excluded those with low I.Q. and adopted children. Controlled for pregnancy complications; father's smoking.  none  Children in hyperkinetic group exposed by mother both prenatally (twice the rate as controls) and postnatally (3 times average for both control groups). Fathers' smoking had no effect.  Prospective study  none  Behavioural ratings in favour of those not prenatally exposed to tobacco.  Asked about mother's alcohol use during pregnancy, but gave no further details.  No significant difference between exposed and nonexposed on mother's behavioural rating of level of activity. Trend towards rating of higher activity for those postnatally exposed.  Compared 4 birth groups: small for gestational age, prematures, premature low birth weight and lull term control group. N o other variables controlled.  age: 6.5 years Eskenazi and Trupin, 1995; Eskenazi and Bergman, 1995 U.S.A. (California) (n=2000+)  Parental education; family income; mother's age, employment status; father's social class; birth order, sex, gestational age at prenatal class.  age: 5 years  97  No significant difference between prenatally, postnatally and both pre- and postnatally exposed groups on cognitive development tests when controlling for other variables.  Table 2.2 continued Fergusson and Lloyd, 1991; Fergusson, Horwood, and Lynskey, 1993  ages: 8,10, 12.  Child's race, age, gender, family size; mother's age, education, socioeconomic status, average standard of living, emotional responsiveness, life events, changes of parents, punishment of child, parental discord; parental history of offending.  Parental history of drug use: mothers selfreport regarding drug and alcohol use by either biological or adoptive parents.  Significant increases in rates of childhood problem behaviours amongst those exposed either during and after pregnancy. Smoking during pregnancy associated with small but significant increases in problem behaviours.  After confounding factors included in analyses, no relationship found between exposure and scores on various scales of general intelligence, word recognition, math ability and reading comprehension.  Fried, 1989a; Fried and Makin, 1987; Fried and O'Connell, 1987; Kristjansson et al., 1989; Makin et al., 1991; Fried et al., 1992,1997, 1998; McCartney etal., 1994  Maternal age, education, socioeconomic status, including father's and mother's education, family income, home environment, mother's age. Also compared post and prenatal exposure to tobacco.  Impulsivity and attentional problems shown consistently at each of the ages for which the children were tested on tests of vigilance or behavioural ratings were given.  At all ages, language or reading scores were lower for children prenatally exposed to tobacco, particularly on auditory-related aspects of reading assessments.  Ottawa, Canada (n=250) ages: 1,2,3,4,5,6, 8, 9, 12 years.  Sample was healthy, middle-class, welleducated, white cohort from urban Ottawa, who gave birth in early 1980's.  Independent variables were prenatal exposure to low levels of alcohol, marijuana and tobacco. Analyses included and compared, exposed to each of these drugs.  Gibson and Tibbetts, 1998  None mentioned  none  Combination of two variables, prenatal tobacco use and low Apgar scores, predicted later police or criminal records.  Socioeconomic factors  none  Hyperactivity and impulsivity observed in the exposed group.  New Zealand (n=1020)  U.S.A. (N=832) ages: 10-22 Hardy and Mellitis, 1972 U.S.A. ages: 4, 7 years  98  Parents behavioural scale at age 6-9 years showed more behavioural problems in both the pre- and postnatally exposed groups, comparing with nonsmoking group.  Spelling and reading scores differed between the exposed and non-exposed groups.  Table 2.2 continued LandesmanDwyer, Ragozin and Little, 1981 U.S.A. (Seattle) age: 4 years Leech, Richardson, Goldschmidt, and Day, 1999 U.S.A. (Pittsburgh) (n=595) age: 6 years Lefkowitz, 1981 U.S.A. (N.Y.) (n=233) age: 10 years McGee and Stanton, 1994 New Zealand (n=1037) ages: 3-9 years  Prospective study. Healthy, middle-class, married, well-educated, white cohort  Controlled for alcohol use in analyses.  Longitudinal study of women, all of whom were from lower socioeconomic backgrounds; child's age and gender, presence of adult male; maternal work; maternal hostility.  Retrospective study; matched exposed and nonexposed cohorts on age, parity, income, education, occupation, father's occupation, family size.  Did not find a significant relationship between maternal smoking and attention. Number of heavy smokers was small. Prenatal tobacco exposure is predictive of inattention. 2" and 3 trimester tobacco use predicted increased omission errors (measures of inattention). d  rd  none  Family background (socioeconomic status, marital status, family adversity), perinatal adversity, health, child rearing (maternal attitude to child, maternal overprotectiveness, attendance at preschool, range of experiences given child, family adversity, child separation from mother).  99  No significant difference on depression, happiness, social behaviour at school, self-esteem, peer-rated popularity.  No significant difference on class achievement, figuredrawing I.Q., workstudy habits  Relationship between maternal reports of behaviour problems only during the age 5 testing. Behaviour measured: attention deficits-hyperactivity.  Smoking during pregnancy and cognitive development not found to be statistically significant, after confounding variables were included in analyses, except for weak relationship for girls with lower I.Q. at age 5 and reading at age 9.  Table 2.2 continued Milberger, Biederman, Faraone, Chen and Jones, 1993  Socioeconomic factors, parental I.Q. levels, parental ADHD status  None  Heavy smoking (at least a pack a day for at least 3 months) observed to be a risk factor for ADHD.  Demographic and pregravid factors (e.g., age, income, no. of abortions, length of time to become pregnant), numerous pregnancy factors, maternal behaviour (e.g., breast feeding, antenatal clinic visits, stopped work by 3 trimester), newborn characteristics (e.g., birth order, gestational age at birth, 5-min. Apgar score.  No mention of alcohol reporting, except: "Quantitative data on alcohol intake were not collected... so alcohol is a variable that is only partially controlled" (p.603).  Hyperactivity and short attention span more frequent in children whose mothers smoked throughout pregnancy (measures for attention span and motor activity tests not well described).  U.S.A. (n= 140 ADHD, n=120 controls) ages: 6-17 years Naeye and Peters, 1984 U.S.A. (Boston) (n=12,150) age:7 years  rd  Nichols and Chen, 1981  Poor orientation and poor attention found in exposed group. Prenatal tobacco exposure found to be amongst the 10 highest predictor variables for hyperactivity and impulsivity.  Britain (n=29,887) age: 7 years  Niemela and Jarvenpaa, 1996 Finland (n=426) age: 56 months  Lower reading and spelling test scores (using WRAT for reading, spelling and arithmetic).  Maternal education and age, duration of breast feeding, marital status, number of siblings, Apgar scores, type of delivery, paternal smoking during pregnancy  none  100  No significant difference in test scores after analyses adjusted for confounding variables.  Table 2.2 continued Olds, Henderson and Tatelaum, 1994 U.S.A. (Rural N.Y.) (n=400) ages: 1-4 years  Orlebeke, Knol, and Verhulst, 1999, 1997  Maternal age, partner support, marital status, diet, alcohol and first trimester drug use, maternal depression, gestational age at initial prenatal care visit, father's education, duration of breast feeding, number of prenatal visits  Analyses adjusted for alcohol use, but measurement or doses not reported  Birth weight; maternal age; socioeconomic factors, type of feeding (bottle or breast)  none  The Netherlands (n=l,377)  Rasanen, Hakko, Isohanni, Hodgins, Jarvelin, and Tiihonen, 1999 Finland (n=5636 males) age: 28 years.  Rantakallio, Laara, Isohanni and Moilanen, 1992; Rantakallio and Koiranen, 1987; Rantakallio, 1985,1983  Socioeconomic factors (father's socioeconomic, mother's education, urban/rural residence); pregnancy and birth factors (e.g., complications, mood and attitude to pregnancy); obstetrical complications; family and parenting risk factors (e.g., marital status, age, family size).  none  Various  Children whose mothers smoked *10 cigarettes daily had lower I.Q. scores as tested by StanfordBinet.  Prenatally exposed children showed more externalizing behavioural problems, but not more internalizing behaviours than controls. Same for boys and girls. Aggression subscale scores increased the overall externalizing score. Maternal smoking associated with violent offenses and persistent offenses, not with nonviolent offenses.  Association found between prenatal exposure and juvenile delinquency. Failed to find a dose-response relationship.  Finland (n=5757 males)  101  -  Table 2.2 continued Sexton, Fox and Hebel, 1990 U.S.A. (Maryland) (n=101who stopped smoking, 263 who continued smoking) age: 3 years Streissguth, Martin, Barr and Beth, 1984 U.S.A.  Socioeconomic factors (maternal education, income); maternal age, prenatal visits, race, alcohol and coffee use during pregnancy, number of siblings, mother employed, birth order, number of child illnesses and visits to the doctor and hospitalization days.  Alcohol use controlled, but measurement not reported  Child's birth order; race; maternal education; nutrition; prenatal caffeine exposure and prenatal alcohol and other drug exposure.  Specific measures (quantity and frequency) of alcohol at 2 time periods (prepregnancy and at 5th month) during an interview.  Maternal age, paternal and maternal income, breastfeeding, maternal education.  none  Maternal age, birth complications, prematurity, low birth weight; family factors (e.g., marital status); parental antisocial personality, maladaptive parenting, SES; substance abuse during pregnancy  Controlled; treated as dichotomous variable (yes or no)  ages: 5-15  Trasti, Vik, Jacob sen and Bakketei, 1999  Higher I.Q. scores (on McCarthy Scales and Minnesota Chid Development Inventory) for those who stopped during pregnancy, comparing to those who smoked throughout pregnancy.  Poor attention and poor orientation to the task (no significant different in reaction time).  No difference at age 13 months on Bayley Scales of Infant Development; at age 5 years, no difference on scales of intelligence (WPPSIR) when adjusted for maternal education. At age 5 years, differences on balance.  Scandinavia (n=376) ages: 13 mo., 5 years  Wakschlag, Lahey, Loeber, Green, Gordon and Leventhal, 1997 U.S.A. (n=177) ages: 6-12 years  age: 12-17 years olds referred to clinic (assessed over 5-year period) Results: Maternal smoking is independentriskfactor for conduct disorder in males  102  Table 2.2 continued Weissman, Warner, Wickramaratne and Kandel, 1999 U.S.A. (n=50) ages 16-33 years Weitzman, Gortmaker, and Sobol, 1992 U.S.A.  Williams, O'Callaghan, Najman, Bor, Andersen, Richards, and Chinlyn, 1998; O'Callaghan, Williams, Anderson, Bor and Najman, 1997  Demographic factors, parental psychopathological conditions, pregnancy risk factors and family risk factors.  Mentioned that smokers had higher rates of alcohol use than controls, but analytic treatment not discussed.  4-fold increased risk in prepurbertal-onset Conduct Disorder in boys, and a 5-fold increasedriskof adolescent drug use in girls.  Child's race, age, gender, birth weight, chronic asthma; family structure, income, divorce or separation of parents, mother's education, intelligence, self-esteem, employment status, chronic disabling health conditions. Compared pre- and/or postnatal tobacco exposure.  Controlled: dichotomous variable of regular use, defined as one or more alcoholic drinks per week during pregnancy  Most behaviour problems amongst those whose mothers smoked both during and after pregnancy, or only after birth of child. Dose-related with strongest link with behavioural problems at exposure of 2 pack a day.  maternal age, education, social class, marital status, mental health; gestational stage at first clinic visit, complications during pregnancy; child's gender, gestational age at birth; family structure and maternal mental health.  none  Results: doseresponse; higher levels of smoking associated with higher rates of externalizing (aggressive) behaviour problems; weaker relationships with internalizing behaviour.  Australia (n=5342) age: 4-6 (mean 5.5 years).  103  2.4 Socioeconomic background Socioeconomic status (SES) as a variable is complex in that researchers have often controlled for it, rather than examined its main effect (Gazmararian et al., 1996; Marmot et al., 1995; Adler et al., 1994; Green, 1970a) and conceptualized and defined it differently by various researchers (Williamson, 1999). Its indicators for measurement have not been well established (some of the indicators have been found to be interrelated but overlapping) and some of the behavioural outcomes for which SES has been an independent variable in themselves covary. Marmot, Bobak and Davey Smith (1995) suggested that it may be better to attempt to understand the links between social position and health, rather than treating socioeconomic background "as a nuisance variable that messes up statistical analysis" (p. 172).  Associations between socioeconomic status and health, or the absence of health as measured by mortality and morbidity, have been consistently found over time and across cultures (Susser et al., 1985). In the research literature, this variable has been found to be more often considered a confounder than ariskfactor (Liberatos et al., 1988).  At least four theoretical viewpoints have been proposed in attempts to explain the consistency of results in studies of the relationship between social inequality and health, regardless of the outcome variable (e.g., infant mortality, accidents, heart disease) (Cubbin et al., 2000; Reutter et al., 1999; Marmot et al., 1995; Kennedy et al., 1998). The possible explanations are that: (1) the relationship is an artefact or not really present, (2) there is natural selection with those who are more often ill having a tendency to drift down economically and socially ("drift theory"), (3) there are  104  behavioural or lifestyle differences such that those who choose unhealthful lifestyles are those who comefromlower levels of the socioeconomic gradient and (4) there are structural differences between the wealthy and the poor (such as housing, access to health care, employment circumstances) that contribute to poor health. Since socioeconomic inequalities may influence, or covary with, the independent variables that are behavioural in nature (i.e., alcohol and tobacco use during pregnancy) (see, for example, Olkinuora, 1984) and may influence outcome variables (health, academic abilities, behaviour) (see, for example, Kennedy et al., 1998), socioeconomic inequalities are considered an important variable to further explore.  Of concern to this thesis, and indeed to most health studies that include socioeconomic inequalities, is that measurement of socioeconomic status and inequalities has been challenging (Murray et al., 1999; Blane et al, 1998; Wilkinson, 1996; Liberatos et al., 1988; Mackenbach & Kunst, 1997; Russell etal., 1990). Even when individual indicators are used, as opposed to social conditions in the aggregate, there is inconsistency in the use of those indicators (Wilkinson, 1996) with occupation being more often used by British researchers, educational attainment by European researchers and racial background being included with other indicators by researchersfromthe United States.  Also of concern to the present thesis is that socioeconomic inequalities may influence both independent and dependent variables, possibly in different ways. Canadian data indicated that tobacco use has an inverse relationship with education level so that smoking prevalence decreases with higher educational attainment in both the general population and in the subcategory of  105  pregnant women (Canada. Health Canada and Statistics Canada, 1999). An extensive examination of this inverse relationship found a threshold effect, with a sharp change between those with less than 9 years of education and those with 10 or more years of education (Zhu et al., 1996). Various categories of smokers (e.g., currently smoking, heavy smoker, those who have not stopped) were associated with numerous indicators of socioeconomic status, including educational status (less than 9 years of education), marital status (divorced or separated), employment status (not being employed) and poverty status (income below poverty level) (Zhu et al., 1996). Likely similar to other western countries, Canadian women with the highest education practice more healthful behaviours, including smoking less (Pomerleau et al., 1997; Hofer & Katz, 1996, Millar and Stephens, 1993). Studies of pregnant Canadian women were consistent in their findings in that women who are younger (less than 24 years), unmarried and with less education have the highest rates of tobacco use both before and during pregnancy (Dodds, 1995; Stewart et al., 1995). Stewart et al. (1995) also found that, at least in the Ottawa area, there was a significant difference in smoking prevalence during pregnancy between those living above and below the poverty level, based on Statistics Canada's tables of annual income, family size and geographic region of residency. U.S. research is consistent with the Canadian studies, with more tobacco use during pregnancy amongst less educated, low-income women (Leech et al., 1999; MMWR, 1999a; NLDA, 1996; Kleinman & Kopstein, 1987) and unmarried women (Williamson et al., 1989), although the subgroup whose smoking prevalence rates are declining the fastest are black pregnant teenagers who are typically unmarried and have low education (Land & Stockbauer, 1993). A British study found the same association (Haslam et al., 1997), while a French study found no significant associations with education or paid work (Kaminski et al., 1995).  106  Regarding alcohol use in Canada's general population studies, people in the highest income and highest education categories have been observed to have the lowest number of occasions of heavy drinking (Pomerleau et al., 1997). No population-based studies on use by pregnant Canadian women were found, either with respect to use of alcohol or of other drugs. Population studies of women in the U.S.A. and in France were found. The sociodemographic background of women who continued to drink alcohol during pregnancy (rather than stopping when they became pregnant) has also been studied, with the results being inconsistent. In two U.S. studies, women with the highest incomes were more likely to drink alcohol during pregnancy (Ebrahim et al., 1998; NTDA, 1996), and variables associated with binge drinking during pregnancy were being unmarried, employed and a smoker (Ebrahim et al., 1999). In contrast, another U.S. study (Leech et al., 1999) found that pregnant women who used alcohol heavily did not have any distinguishing sociodemographic characteristics. A 1988 U.S. study using national data found that frequent prenatal drinking was more prevalent amongst older women (^35 years), all racial groups other than Caucasian, those with low incomes and those who are smokers (MMWR, 1995a). A more recent study in Pittsburgh found that the heaviest smokers, at least during the first trimester of pregnancy, were more likely to be Caucasian and less educated (Leech et al., 1999). A study that was conducted over two time periods (before and after medical and media attention regarding FAS) showed that there was a decrease of alcohol use during pregnancy in the 6-year period, but that the decrease was most marked in the highly educated groups (Streissguth et al., 1983). In a study of socially disadvantaged cohort of women in France, alcohol consumption was highest among those with the least education. Abel (1995) has argued that even if some studies have shown prevalence ofriskydrinking during pregnancy to be at the higher ends of the socioeconomic  107  gradient, it may not be their offspring who are those most affected by alcohol-related birth problems since other factors (for example, nutrition) may interact with alcohol exposure.  The only Canadian longitudinal study of pregnant women found that lower socioeconomic status, as defined by family income and mother's education were associated with both tobacco and marijuana use during pregnancy but not with alcohol use (Fried & Watkinson, 1990). It should be noted that the study drew its sample from a homogeneous predominantly middle-class low-risk Caucasian population.  In summary, there appears to be an association, at least correlational, between the independent variables, socioeconomic status and alcohol and tobacco use during pregnancy. In addition, there may be a relationship between socioeconomic status and the outcome variables of this thesis. The relationship between social conditions and health has been well-documented (Mackenbach & Kunst, 1997; Marmot et al., 1997, 1995, 1991, 1987, 1986; Wilkinson, 1996; Adler et al., 1994, 1993; Evans et al., 1994; Feinstein, 1993; Liberatos et al., 1988; Green, 1970a). Socioeconomic disadvantages during childhood have been shown to influence health well into adulthood (Hertzman & Wiens, 1996; Schwartz et al., 1995; Peck, 1994). Academic abilities have been found to be associated with socioeconomic background. Amongst 4-year old children, those with multiple social-environmentalriskfactors were observed to be more than 24 times as likely to be have low verbal I.Q. scores (Sameroff et al., 1987) and amongst school-aged boys socioeconomic gradient correlated with both slow physical growth and lower reading achievement (Weinberg et al., 1974).  108  Independent associations were found between lower income, minority status and lower parental education and poorer performance on cognitive assessments amongst children aged 6 to 16 years from a large U.S. national cohort (n=2531) (Kramer et al., 1995). The same study found that general health status was a predictor of performance on cognitive assessments (including the WRAT-R) (Kramer et al., 1995).  It has been observed that infants born with low birth weight more often comefromdisadvantaged backgrounds, even controlling for behavioural factors such as alcohol and tobacco use during pregnancy. This observation has led to suggestions that women in the socially-deprived "group has associated with it a few significant and common unidentified pathological factors or, more likely, that numerous factors in the social environment are acting to reduce birth weight" (Wilcox et al., 1995, p.924). Evidence on this issue has been inconsistent. Brooke et al. (1989) examined several risks (including the use of tobacco, alcohol and caffeine, psychosocial stress and socioeconomic background) and concluded that smoking during pregnancy was the main contributor to low birth weight, while social and psychological factors had little or no direct effect. Orr et al. (1996) found that psychosocial stressors had different effects on thosefromdifferent racial backgrounds and that for Caucasian women, significant predictors of low birth weight were smoking, drug use and other factors and that psychosocial stressors may be only indirectly associated. These latter studies suggested that smoking may have a more direct association with low birth weight, but that tobacco use during pregnancy and socioeconomic background covary.  Measurement may be complicated by the socioeconomic background of mothers. It has been  109  found that some groups of women are more likely to assess their children as having poor health; for example, single mothers were observed to be significantly less likely to report their child as having "excellent" health on a 5-point adjectival scale (Angel & Worobey, 1988), mothers under stress reported more illness for themselves and for their children than those not under stress (Mechanic, 1964), and different racial groups assess the state of their children's health differently (Angel & Worobey, 1988). Of particular interest to the present study were the findings that women who used alcohol or drugs during pregnancy were observed to under- or over-rate considerably their infant child's development, with heavy alcohol users being 15-fold more likely to overestimate their infants mental development and illicit drug users being 4-fold more likely to underestimate their infant's physical development, comparing with test results of the Bayley Scales of Infant Development (Seagull et al., 1996).  The results of at least one study suggested that "developmental outcomes are more strongly related to environmental influences than to prenatal insult" (Greene et al., 1991a, p.66). Home environment was more predictive of lower scores on cognitive development than either prenatal exposure to tobacco or to moderate levels of alcohol exposure, even within a predominantly socioeconomically disadvantaged cohort of mother-child pairsfromCleveland.  2.5 Other potentially confounding factors There are, of course, several other potentially confounding variables that could mediate or moderate the association between prenatal risk factors and outcomes regarding children's health, behaviour or academic abilities.  110  Prenatal and postnatal nutritional deficiencies can adversely affect the newborn and child both physically and cognitively. Some researchers have suggested confounding factors that could explain the link between poor nutrition and learning problems as being indirect (Karp, 1993; Karp et al., 1992). The correlational relationships between measures of nutrition and aptitude or school achievement among children from disadvantaged neighbourhoods have not been found among white, upper middle class children (Richards et al., 1984; Karp et al, 1992). Social environments that adversely affect both nutrition and learning abilities appear to exist for children living in poverty (Karp, 1993; Hallberg, 1989).  Poverty and culture have been linked to incidence rates of FAS. The relationships between FAS and poverty or culture are unclear, however, because of the strong relationship between these two factors (some cultural groups and poverty) and between these two factors and other risk factors including nutrition and smoking. Several studies have indicated a positive relationship between specific cultural groups, especially aboriginal groups in Canada and black and aboriginal groups in the United States, so that early researchers hypothesized a genetic link between FAS and those cultural groups. Although epidemiological data have suggested higher incidence rates of FAS among aboriginal than among non-aboriginal populations in Canada (Square, 1997; Robinson et al, 1987), there is no clear evidence of genetic predisposition, rather that poverty may be an intervening factor. Bingol et al. (1987) found a highly statistically significant difference in the prevalence of FAS or FAE between low and high socioeconomic status women. They found FAS to be rare in the high socioeconomic group of chronic alcoholic women, although there were occurrences of lower educational achievement and attention deficit disorder amongst their  111  children. Even those rates were statistically different, for example 21% of offspring of alcoholic women in the higher socioeconomic group had attention deficit disorder, compared with 71% of the children in the lower socioeconomic group.  2.5.1 Other maternal and child factors Other factors have been shown to influence the outcomes under study in the present thesis, including maternal age at delivery of the child under study and birth order.  Maternal age and parity have been observed as increased risk for fetal alcohol syndrome both in case reports and epidemiological studies (Abel, 1998b; Jacobson et al., 1996; Sokol et al., 1986). Abel (1998b, 1988, 1984; Abel & Dintcheff, 1985) calculated that there is a 79 fold increased risk of having a second child with FAS if a mother has already had one child with FAS. Whether parity or age increases theriskhas not yet been determined in humans, since maternal age, parity, duration of alcoholism and maternal health have been found to covary. Some animal studies suggested that advanced age increases fetal vulnerability more than does parity (Jacobson et al., 1996; Abel & Dintcheff, 1985, 1984). Speculated explanations found in the literature are that there are increasingly greater blood alcohol levels as age advances even when the same amounts of alcohol are consumed, reducing the ability to metabolize alcohol as age increases (Jacobson et al., 1996) and that increased age correlates with an increased stage of alcoholism and the general unhealthiness of pregnant alcoholics (Majewski, 1981). A positive aspect of this relationship is that those generally considered most atrisk,that is teenagers or younger mothers, because of the higher rates of alcohol consumed in that age category, may have decreased risk for having  112  offspring with FAS (Abel, 1998b).  The research on maternal age and parity has been conducted in animal research with high doses of alcohol. In humans, only case reports and other studies involving people with FAS have shown the associations with maternal age and parity, and there have been no studies where more moderate levels of alcohol were consumed.  Only one study was found on the relationship between the effects of prenatal exposure to tobacco and maternal age. Ahluwalia (1997) observed a slightly elevatedriskof low birth weight for the offspring of women over age 30 years who smoked during pregnancy.  2.5.2 Other drugs Outcome variables of this study may be modified by the mother's use of drugs other than alcohol or tobacco.  Marijuana More prevalent than other illegal drugs amongst the sample in the Vancouver Island Pregnancy Study was the use of marijuana during pregnancy, with approximately 10 percent (n=40) of the women admitting to using marijuana prior to pregnancy and 5 per cent (n=20) reported using marijuana during at least one of the trimesters of pregnancy, with about half that number using marijuana throughout their pregnancy (Armstrong et al., 2001a). Hashish was also used by 12 (3%) of the women before pregnancy, but half of these women stopped during the first trimester  113  and only two women used hashish on an infrequent basis in the third trimester.  The rates of reported marijuana use were lower than those found in other studies; for example the Ottawa Prenatal Prospective Study found in 1978 that 20% of women smoked marijuana at some time during pregnancy (Fried et al., 1995a, 1998b) and in a cohort of women of predominately lower socioeconomic backgrounds in Pittsburgh, U.S.A., the rates for each trimester were about twice that of the Vancouver Island Pregnancy Study (Leech et al., 1999; Day et al., 1992, 1991). Hashish use in the latter study was rare and cocaine use was low (3% duringfirsttrimester, 1% in second trimester and 0.3% during third trimester) (Leech et al., 1999).  A limited number of longitudinal studies of the effects of prenatal marijuana exposure on schoolaged children were found. Leech et al. (1999) tested the offspring of the cohort of Pittsburgh women for attention and impulsivity problems at age 6 years,findingthat second trimester marijuana use was predictive of more errors of commission (reflecting impulsivity) and fewer errors of omission (reflecting inattentiveness). In contrast, Fried, O'Connell and Watkinson (1992a) found that the 6-year old offspring of middle-income mothers who used marijuana during pregnancy had an increased number of errors of omission in a task of vigilance and that there was a dose-response relationship between prenatal marijuana exposure on two indicators of impulsivity/hyperactivity. Prenatal marijuana exposure was not associated with deficits in cognitive and language skills (Fried et al., 1992b). Fried et al. (1998, 1997) found that prenatal marijuana exposure was not significantly related to reading or language skills, nor to global intelligence scores, by ages 9 to 12 years. On the other hand, they found that two of their many  114  subtests were consistent at testing during the neonatal stage, and at ages 3 and 9-12 years. Their results showed an association between prenatal marijuana use and poorer habituation to visual but not auditory stimuli at neonatal stage and poorer abstract/visual reasoning in 3-year olds. For the 9-12 year olds, overall intelligence was not significantly different between the exposed and nonexposed groups, but the two groups were differentiated on the subtests of Block Design and Picture Completion on the Wechsler Intelligence Scale for Children. The results showed a linear trend. In summary, Fried and his colleagues in Ottawa found that prenatal exposure was negatively associated with integrative abilities based on visual analysis, visual hypothesis testing and impulse control, but not with verbal skills or overall intelligence.  Caffeine The research on the effects of prenatal exposure to coffee and other caffeine products is inconclusive. Most women have reported use of some source of caffeine during their pregnancy (Watkinson & Fried, 1985) although many decrease their use of caffeinated beverages partly because of nausea, possibly causing bias in the research (Fenster et al., 1991a, 1991b). Animal studies have shown that caffeine does have teratogenic effects but human studies have been sparse, especially those that controlled for the use of other substances, such as tobacco and alcohol (Nehlig and Derby, 1994; Sobotka, 1989). A weak (Armstrong et al., 1992) or no (Fenster et al., 1997, 1991b; Mills et al., 1993) association has been reported between caffeine and spontaneous abortion. In contrast, Fenster et al. (1997, 1991b) observed that consumption during the first trimester of decaffeinated beverages was associated with spontaneous abortion. Only weak associations have been found between caffeine consumption and prematurity or congenital defects  115  (McDonald et al., 1992a; 1992b). Although birth size may be somewhat lower (Fenster et al., 1991a; Fried & O'Connell, 1987), this effect has not been shown to last, as the children were , observed to catch up in size within their first few months (Fried & O'Connell, 1987; Barr et al., 1984). Mills et al. (1993) found that any initial observation of a relationship between low birth weight and caffeine use was no longer statistically significant when otherriskfactors, especially smoking, were controlled in the analyses. The only well-controlled longitudinal study found no effects of prenatal caffeine exposure on the cognitive and neurobehavioural variables measured during thefirsttwo days of life, including reactivity, arousal and suckling and in the longer-term, no effects on motor ability or intelligence at age 4 years, nor on alertness or intelligence at age 7 years (Barr & Streissguth, 1991).  It is possible that caffeine has the potential of mediating the teratogenic effects of other substances, particularly tobacco and alcohol (Nehlig & Derby, 1994).  2.6 Summary Both prenatal exposure to alcohol and tobacco have been shown in the research literature to be associated with outcomes affecting physical growth, behaviour and academic performance. The research has been inconsistent with respect to physical growth, in association with either tobacco exposure or moderate alcohol exposure and if physical growth deficiencies are influenced by either exposure, this association may have little functional consequence, if the children catch up to the normal in later years, as has been shown in many studies.  116  As an effect of both prenatal tobacco exposure and prenatal exposure to alcohol, attention deficits dominate the research literature. Independently each has been observed to have a relationship with specific behavioural problems relating to attention deficits or related behaviours including impulsivity and hyperactivity.  Mechanisms have been suggested for the adverse effects of both tobacco and alcohol exposure. Five different mechanisms have been most often suggested, however, fetal hypoxia and malnutrition are commonly linked with both tobacco and alcohol exposure. The evidence is unclear and probably no single mechanism acts alone as a causal factor.  Up until 1941, when rubella was observed to be a teratogen, environmental factors were given little attention as possible risks to the fetus (Naeye, 1992b). In the 1970's, data were collected on large populations in many countries on numerous conditions of newborns, including maternal tobacco use during pregnancy. At the time these studies were being designed, it was still believed that a barrier existed so that ethanol did not reach the fetus. This is most likely the reason for not collecting data on alcohol consumption during pregnancy. Since the late 1960's, thousands of research studies have reported on the effects of prenatal alcohol exposure, many being case reports of people with FAS and only recently have there been studies attempting to study threshold or linear outcomes on specific variables.  Refined methods and measures continue to challenge researchers in thisfield,since no biological markers have been found to reliably measure amount of alcohol or tobacco used. Self-reporting  117  has been the measure most often used. Dose and timing of either or both alcohol and tobacco pose measurement problems given that many women use both tobacco and alcohol at varying levels and times during pregnancy. Only one study was found in which tobacco exposure, alcohol exposure and social background were examined for interactive effects. Barrison and Wright (1984) observed that social background and prenatal alcohol and tobacco exposures had an interactive effect on birth weight. In their prospective study of 900 Caucasian women in England, they found that alcohol consumption of more than 100 g of alcohol per week was associated with more than double the risk of low (below the 10* percentile) birth weight and that the risk was significantly increased in offspring of women who smoked cigarettes during pregnancy and who were classified in lower social classes, while controlling for other potential confounding variables such as parity and maternal age. Theriskof drinking and smoking was found to be additive (Wright et al., 1984). Moreover, timing of alcohol exposure was included in the analyses, with the observation that preconception and early pregnancy drinking appeared to have the most potential forrisk,at least in terms of birth weight outcome (Barrison & Wright, 1984; Wright et al., 1984).  One other study attempted to examine the effects of the combinedrisksof prenatal alcohol and tobacco exposure to behavioural outcomes such as attention span, activity level, fidgetiness and social compliance during a naturalistic home observation (Landesman-Dwyer et al., 1981). This study was described previously with respect to behavioural outcomes and prenatal tobacco exposure. Caucasian women (n=128)fromthe Seattle longitudinal study were categorized accordingly: (1) 29 moderate drinkers/smokers, (2) 38 moderate drinkers/non-smokers, (3) 26 occasional or non-drinkers/smokers, (4) 35 occasional or non-drinkers/non-smokers, with the four  118  groups being similar in terms of other demographic factors such as maternal age, parity and educational levels. Alcoholic women or alcohol abusing women were excludedfromthe sample. This seemed a useful contribution to the literature in terms of methodology, with the potential of discriminating the results ofrisksfromeither tobacco or alcohol, while holding constant the socioeconomic levels, since the women camefromrelatively similar backgrounds. Unfortunately, tobacco was treated dichotomously, rather than categorically by dose.  The importance of these two studies (Barrison & Wright, 1984 and Landesman-Dwyer et al., 1981) was that interactive effects were considered in their statistics models. Very few studies have replicated this approach because of the complexity of the research and the many factors influencing both the independent and dependent variables. With respect to low birth weight and stillbirth as outcomes, it was noted that "the literature on moderate drinking is contradictory and suffers from poor control of factors known to confound pregnancy outcomes such as social class, parity and smoking habit" (Barrison & Wright, 1984, p. 167). From the present review of the literature, the same appears to hold today. As the research accumulates in this relatively newfield,patterns are emerging and our understanding of the influence on health of factors such as socioeconomic background are becoming clearer, in an albeit complex topic both to study and to interpret.  Each of the longitudinal studies to date have had considerably different populations, ranging from European populations whose mothers smoked but consumed little or no alcohol to disadvantaged black inner city American populations. Replication of the current studies in more geographic settings may present more data on which to identify patterns.  119  2.7 Theoretical guides The literature review above on prenatal alcohol and tobacco exposure summarized the field of human research particularly with respect to academic abilities, health or behavioural problems. To follow is a description of two theoretical perspectives presenting ecological approaches to the relationships between many of the variables under study in the Vancouver Island Pregnancy Follow-up Study. These are not meant to indicate comprehensiveness in all theoretical models on this topic; rather they are two perspectives found useful in formulating the relationships between and realizing the complexity of the interaction of variables.  These two ecological models serve as the organisational frameworks for examining the relationship between the variables and as heuristic devices in designing this research. Both Sewell, Price and Karp (1985) and Abel and Hannigan (1995a) described relationships between various risk factors and conditions, including prenatal alcohol exposure, poverty, poor nutrition and cognitive development. Conceptualizations of each of these models is shown in Figures 2.1 and 2.2 below.  120  Figure 2.1 Conceptualization of the interacting variables influencing birth outcomes (Sewell Price and Karp, 1985).  £ POVERTY  lack or s o c i a l support  lack of education in-utero exposure to ETOH ability to De e f f e c t i v e parents  ETOH use in the f a m i l y  POOR DIET  excess fiber low vitamin C l i t t l e available dietary iron  LOW ABSORPTION OF DIETARY IRON  IRON DEFICIENCY  LEARNING FAILURE  I  Absorption of lead  I  EXPOSURE TO LEAD  Other nutritional factors ( e g hunger, and zinc or protein d e f i c i e n c y )  121  Figure 2.2 Schematic representation of summary of relationships among permissive and provocative maternal risk factors underlying FAS/ARBDs (alcohol-related birth defects) (Abel and Hannigan, 1995a). BIOLOGICAL RESPONSES (E.G., GLUCOCORTICOIDS)  ~  ENVIRONMENTAL POLL LUOTTAKNTTSS * 1  A  \  '  UNDERNUTRITION  ALCOHOL ABUSE -  \  /  \  /  S T R E S S ^ - "  ANTIOXIDANT  h^SlflSiSSfy (VITAMINS A  PROTEIN DEFICIENCY  INTRAUTERINE GROWTH RETARDATION  /  H  O  L  I  N  T  * ^  >  X  ) - < - - - > - o f H E R O « U G ^  PATTERN /  USE/ABUSE /  A  I  |- ' .  MINERALS)-*-  DECREASED ENERGY  C ^ ^  ALCOHOLISM » ALCOHOLISM-. N  I  FREE r* I JIC^T. RADICALS J  _  +7  CELL " • DAMAGE '  -> FAS/ARBDs  Notes: Sociobehavioural permissive factors are shown inside the circles, key biological provocative factors are shown in the rectangles, dotted-line arrows indicate pathways by which permissive and provocative risk factors act on the fetal unit, and solid-line arrows indicate biological relationships and physiological pathways.  122  Sewell et al. (198S) argued that assessments of cognitive abilities have been insufficient, resulting in over-emphasis on labelling and classification. Instead, they suggested, environmental factors should be included in any assessment of cognitive abilities and the consequent focus should be on the educational needs of children rather than on apparent intelligence classifications. As a basis for their argument, they noted the concomitant occurrence of poverty, nutritional factors (including hunger, iron and zinc deficiencies, absorption of lead), prenatal exposure to alcohol and alcohol use in the family, all of which have interrelated effects on failure to learn. Their model, illustrated in Figure 2.1, showed an indirect relationship between poverty and learning failure, influenced by synergistic environmental conditions and risk factors. Lack of social support and poor education could result in poverty and poor parenting. When these two conditions co-exist, there could be increased levels of social problems (including alcohol use of the parents) and poor nutrition, with consequences to the child's ability to learn.  Sewell et al. (1985) focused on the outcome variable of learning abilities. The other two variables of the Vancouver Island Pregnancy Follow-up Study, behaviour and health status, may be influenced by the same interrelationship of risk factors and conditions.  Like Sewell et al. (1985), Abel and Hannigan (1995a) placed importance on socio-behavioural factors. Abel and Hannigan's model is shown in Figure 2.2. They were more explicit than Sewell regarding the relationships between the variables and explanations of biological mechanisms influenced by social conditions, such as low socioeconomic status and specific risk factors, such as under-nutrition. In addition, their model has been developed with respect to the outcomes of FAS  123  or alcohol-related birth defects and therefore seems more specific and relevant to the present study.  Abel and Hannigan (1995a) noted that not all women who drink heavily produce offspring with FAS so that alcohol is a necessary, but not sufficient, factor resulting in the occurrence of FAS. Rather than alcohol being a single causal agent (with expectations of a response), there are factors that increase the susceptibility to alcohol or weaken the defence mechanism. They identified two groups of conditions, permissive and provocative, as possible explanations for the occurrence or nonoccurrence of FAS in spite of the mother drinking alcohol. Permissive conditions are "predisposing behavior, social or environmental factors" (Abel & Hannigan, 1995a, p. 445) that differentiate the fetal response to alcohol. The permissive conditions influence "physiological changes in the internal milieu, called provocative conditions, that increase the vulnerability to alcohol's toxic effects" (Abel & Hannigan, 1995a, p. 445).  In their model shown in Figure 2.2, "permissive conditions" are shown inside circles and "provocative conditions" are shown in rectangles. The pathways between these variables can be causal or single-directional, indicated by solid line arrows in their schema, or bidirectional that have been shown by dotted line arrows. The "provocative conditions" produce the mechanisms that cause cell damage, the proximal cause of FAS.  Abel and Hannigan (1995a) included in their realm of "permissive conditions" both social conditions, such as lower socioeconomic status and aspects of culture andriskfactors, such as  124  alcohol intake patterns and smoking, although they did not make the categorical distinction of environmental conditions and particular risks. "Provocative conditions" include blood alcohol levels (that are also affected by body weight and other factors), under nutrition and tobacco smoke components.  Their theoretical model offered an opportunity to link animal and behavioural research regarding the teratogenicity of alcohol and provided a more advanced basis for understanding susceptibility to the effects of alcohol, since the specific mechanisms are not yet entirely understood. Rather than focussing on specific dose-responses as the factor determining whether FAS or any alcoholrelated birth defect occurs, they proposed these other conditions as determinants.  Abel and Hannigan (1995a) have used various examples to give credence to their theoretical model, one of which is maternal age and parity. The probability of prenatal alcohol-related birth defects increases with parity and maternal age (Abel, 1998b; Jacobson et al., 1996; Sokol et al., 1986; Abel & Dintcheff 1984, 1985). Abel and Hannigan (1995a) noted the relatively high rates of drinking amongst teenagers, but that if alcohol consumption were the single causal factor for fetal alcohol syndrome, then the number of years of alcohol consumption would be irrelevant. In fact, the opposite has appeared to be the case. Abel and Hannigan (1995a) suggested a number of possible explanations, including increased age allowing for a longer history of alcoholism that possibly comprises maternal health, higher blood alcohol levels associated with consistent levels of consumption, or development of physiological problems such as cirrhosis.  125  These models (Sewell et al., 1985 and Abel & Hannigan, 1995a) guided the Vancouver Island Pregnancy Follow-up Study in their suggestions that prenatal alcohol exposure is multidimensional, and that as a variable it should not be studied independentlyfromother risks to which the child might be subjected. Sewell et al. (1985) suggested the complexity of the relationship between variables and the outcome variable of academic abilities or failure to learn. Abel and Hannigan (1995a; Abel, 1998) focused on alcohol-related birth defects and FAS. Although there is direct concern with alcohol-related effects, there was no attempt to diagnose FAS in the Vancouver Island Pregnancy Follow-up Study. Rather, the study has been guided by their model in the complex risk factors and environmental conditions that they suggested contribute to adverse effects on the child.  126  Chapter Ul  Research Questions and Methods  3.1 Research Questions This research examined prenatal exposure to alcohol and prenatal exposure to tobacco in the attempt to determine the impact of these risk factors to children's health, academic abilities and behaviour. The research also considered the interaction of three variables, prenatal alcohol and prenatal tobacco exposure and socioeconomic status, as potentially posing higher risk than would each of these variables alone.  The general aims of the research were to examine behavioural characteristics, academic abilities and health of elementary school age children who were prenatally exposed to alcohol, or to tobacco, or to both, and to study the effects of the combined risks to the child of prenatal alcohol exposure, prenatal tobacco exposure and lower socioeconomic status. A conceptual model of the relationships among the variables is shown in Figure 3.1, page 129.  The research questions were. 1.  Are there interactive effects of (i) prenatal alcohol exposure, (ii) prenatal tobacco exposure, and (iii) socioeconomic status on the health, academic abilities or behaviour of children?  2.  Is there a relationship between alcohol exposure during pregnancy and childhood behaviours?  3.  Are there more health problems, both in quantity and severity, amongst children who were prenatally exposed to alcohol?  4.  Do children prenatally exposed to alcohol have lower scores on the WRAT-3 and lower  127  ratings regarding school performance compared to those who were not prenatally exposed? Is there a relationship between tobacco exposure during pregnancy and childhood behaviours? Are there more health problems, both in quantity and severity, amongst children who were prenatally exposed to tobacco? Do children prenatally exposed to tobacco have lower scores on the WRAT-3 and lower ratings regarding school performance compared to those who were not prenatally exposed? What are the relationships between socioeconomic background and health, academic abilities and behaviour in this cohort of children? What are the relationships between current weight and prenatal exposure to alcohol or to tobacco or to lower socioeconomic background?  128  Figure 3.1. Conceptual model of relationships between independent variables (prenatal alcohol exposure, prenatal tobacco exposure and socioeconomic status) and outcome variables (academic abilities, health and behaviour).  129  3.2 Methods This was a population-based study in which data were collected at three stages. The first data collection took place during a one-year period, July 1, 1990 to June 30, 1991, when a cohort of pregnant women living on central and northern Vancouver Island visited physicians who had been asked to completed an adapted medical history form which included the T-ACE screening assessment. These women became subjects of the study when they delivered babies in any of prespecified hospital regions on Vancouver Island. They were identified through medical records at the time of delivery. Physicians had been asked to participate in the study for the one-year period by including questions in the regular medical history, but the subjects did not become known to the researchers until the time of delivery. The women became subjects when they granted permission to participate in the study. Approximately 10% of the women delivering within the region during the week were selected randomly for an indepth interview in the home which took place within eight weeks of giving birth. These interviews lasted approximately 1 A hours and were conducted X  by research nurses. These two stages comprised the Vancouver Island Pregnancy Study. The present study design extended that earlier research into a longitudinal study, known as the Vancouver Island Pregnancy Follow-up Study, by collecting data on the dependent variables from the children, their parents and their teachers. Data were collected through interviews, selfadministered questionnaires, rating assessments and standardized test instruments.  3.2.1 Ethics review Review by The University of British Columbia, Office of Research Services and Administration, Behavioural Research Ethics Board, was completed and approved on February 27, 1998 and re-  130  approved March 31, 1999.  3.2.2 Stages of data collection The three stages of data collection, along with time periods, subject selection and number of subjects and data sources, were as follows:  Stage 1 Time period: July 1, 1990 - June 30, 1991 Subjects: all birth deliveries in pre-designated hospital districts of central and northern Vancouver Island Data source: adapted medical records forms n=3,659  Stage 2 Time period: June 1990 - mid-1991 Subjects: random sample selectedfromStage 1 population Data source: home interviews conducted within 8 weeks of birth n=403 (11.0% of population)  Stage 3 Time period: May 1999 - January 2000 Subjects: children of Stage 2 sample Data sources: written questionnaires for moters and teachers and testing of child n=265 (65.7% of Stage 2 sample; 7.0% of population)  131  This Chapter reports the methods of Stage 3. Methods and results of the research of earlier phases have been reported elsewhere (Armstrong et al., 2001a; Armstrong et al., 1994), however, because of the associations between data collected during the first two stages and the final stage, some details of the methods used in previous stages have been repeated here, as required for clarity or for interpretation of results.  3.2.3 Population and sample 3.2.3.i Study population and procedures for locating subjects The population consisted of all women who delivered babies withinfiveregional hospital districts of central and northern Vancouver Island during the 12-month period, July 1, 1990 to June 30, 1991. The population size was 3,659 women, according to hospital medical records of births in the region during the time period. Women were identified at the time of delivery. The expected number and actual number of deliveries was ascertained by noting all deliveries in the study area, that is, in all hospitals, each week. From the list of actual deliveries, 10% of the subjects were selected randomly for in-depth follow-up interviews. If a subject refused, a replacement random selection was made.  The birth records included demographic, pregnancy, delivery and newborn outcome information, and were adapted for the purposes of this study, to include alcoholrisklevels. All physicians providing antenatal and obstetrical care within the catchment area had been asked to participate in the study by administering an alcohol screening tool to all pregnant women under their care and physicians were trained in its use. The screening tool used was the T-ACE questionnaire on risk  132  levels of alcohol use, designed for use with pregnant populations (Sokol et al., 1989). Physician records were linked with the hospital birth records.  Demographic characteristics Demographic characteristics of the study population are shown in Table 3.1 below. Women who delivered during the study period ranged in age from 14 years to 47 years of age. Over two-thirds (66.8%) were married, 16.3% lived common-law with a partner and 16.9% were single (including separated, divorced, widowed, or never married). Of those whose employment status was known, most (68.2%) were employed either outside the home or as a homemaker at home, or were students, and the remainder (31.8%) reported being unemployed. Only 40% of the study population's ethnicity was known, and of these, 76% reported Caucasian, 19.5% aboriginal and 5.5% other ethnic backgrounds.  133  Table 3.1 Selected maternal demographic variables of study population of women with live births in central and northern Vancouver Island, 1990-91 data collection. Variable  n=3,659  Age  mean range undetermined  Marital status  married common law single separated, divorced, widowed  Employment status*  undetermined Unemployed Employed Homemaker Student undetermined  Ethnic background*  Caucasian Aboriginal Asian or South Asian Other undetermined  26.95 years 14-47 years 4 2,436 (66.8%) 594 (16.3%) 534 (14.6%) 83 (2.3%). 3,647 (100.0%) 12 853 (31.6%) 1,269 (47.0%) 482 (17.8%) 93 (3,4%) 2,697 (100.0%) 962 1,445 (76.0%) 351 (18.5%) 96 (5.0%) 10 (0W^ 1,902 (100.0%) 1,657  *Data for these variables collected only if indicated on hospital record, or for those women who had a home visit interview, explaining the high ' number in the "undetermined" category.  Maternalriskbehaviours during pregnancy: alcohol and tobacco Risk levels of alcohol use were measured with the T-ACE screening tool (Sokol et al., 1989). Sokol (1989) dichotomizedrisklevels (risk or norisk),with any score above 2 rated as risk. Risk levels for the Vancouver Island cohort had been categorized into three levels, that is, low, medium or highrisk(see below, section 3.2.4.i (a)). At the Stage 1 T-ACE assessment, nearly half the  women (49.4%) were considered to be lowriskwith respect to alcohol, 36.4% had scores considered to be mediumriskand 14.7% had scores considered to be highrisk,as shown in Table 3.3 below (Armstrong et al., 2001a, 1994). Nearly one-third (32.8%) of these women smoked at some time during their pregnancies.  3.2.3.ii Study sample and procedures for locating subjects After the birth of their newborns, a sample of 403 women was randomly selectedfromthe original population of 3,659 women (11.0% of the population). All 403 randomly selected women who were interviewed during Stage 2 in 1990-1992 were considered part of the Stage 3 sample and were sought for inclusion in the 1999-2000 data collection.  Extensive efforts were used to locate each of the subjectsfromthe earlier stages of the research since sample loss can be a serious problem in longitudinal studies. As shown in Table 3.2, a total of 299 (74.2%) of the original randomly selected sample was located, of whom 65.7% (n=265) participated in Stage 3 and another 3% (n=12) were found but did not participate for various reasons (refusal, death, difficult to contact). Given the long period with no contact, the low rate of loss to follow-up compared favourably with other Canadian longitudinal studies. For example, an Ontario Child Health Study located and enlisted 75.5% of its sample four years after original follow-up (Boyle, Offord, Racine & Catlin, 1991). The Collaborative Perinatal Study began in 1959 at 12 sites at hospitals throughout the U.S.A. At one of those sites, the Boston Lying-in Hospital, less than 20% of their sample was lost at testing 7 years later mostly because of the researchers' efforts in contacting the subjects annually between initial contact and re-test (Naeye &  135  Peters, 1984).  Procedures for locating subjects A multi-step process was used to locate the subjects for Stage 3 data collection. Three directories were used to find current addresses: local telephone books and two electronic databases (CanadaPhone available on CD-ROM and Canada 411 available on the internet). The two electronic databases were searched for subjects and if either the address or the telephone number matched, the subject was sent a letter explaining the study. This yielded nearly 23% of the sample (n=92). These people lived at the same address as they had lived approximately nine years earlier, or at least near enough to have retained the same telephone number. Next, contact was made with people with the same surnamefromtowns in which the subject had been living, andfrequentlythe subject was located through contact information given by relatives. Through this searching, some women were found who had moved out of the province, including to the Yukon, Ontario, Saskatchewan and Alberta. In two cases, ex-husbands would not reveal the whereabouts of their ex-spouses, but in all other cases, family members helped to locate the subjects. Alternatively, if the surname was unusual we searched the electronic directories for any names throughout British Columbia or Canada. The searching through relatives and directories yielded another 48% (n=128) of the subjects who were eventually interviewed. In addition, public health nurses working in the area assisted with locating individuals who were known to them. Public health nurses generally knew subjects if they had new babies or because they lived in the same neighbourhood. The public health nurses asked subjects known to them if they were willing to have us contact them by letter. Four subjects (1.5%) were found through public health nurses.  136  In summary, 83.5% (253) of the sample was located through publicly available directories, either directly or through someone who knew the whereabouts of the subject. The women were sent two-page letters explaining the study, followed by phone calls asking for their participation.  For the remainder of the sample who had not yet to be traced, information was matched with the British Columbia's Ministry of Health Linked database. Personal Health Numbers or mother's date of birth, either of which were obtained during the 1990-1991 study, were matched with present addresses and telephone numbers. Permission for this matching procedure was granted after application to the provincial Ministry of Health, who restricted the methods of contact. Subjects were sent a letter requesting their participation and asking them to return a card authorizing further contactfromthe researcher. No further contact could be made with the subject, prior to the card being returned. Of the 150 names sent to the Ministry of Health, 110 addresses were supplied, with 40 names having no match in the database. It appeared from reviewing the original (1990-91) medical data forms that in two cases, the women had no medical numbers nor any relatives in British Columbia and may not have been registered in any way within the medical system. Other women may have moved out-of-province, changed names or were otherwise unknown to the health care system's database. Of the 110 addresses supplied by the Ministry of Health database, 25 (22.7%) were returned by the post office as "address unknown", 39 yielded no response (35.5%) and 41 (37.3%) resulted in the subject being interviewed, three sent refusal responses and one could not be found because she did not supply telephone contact information. A summary of these methods of locating subjects, the contact procedures and the numbers of subjects found and who participated or refused is found in Table 3.2 below.  137  Table 3.2 Methods of attempting to locate subjects, with refusal and participation rates. Subjects located Names linked through publicly with Ministry ofHealth available database sources  Total  Subjects found: Interviewed Refused  224  41  265 (65.7%)  18  4  22 (5.5%)  Total  287 (71.2%)  Subjects or'their relatives found, but interview not possible: i. Moved to address unknown to relatives  3  ii. Deceased (one child; one mother)  2  iii. Subjects found but contact to make arrangements for interview not possible (e.g., no phone)  5  iv. English too limited to interview  1  Total  253  1  12 (3.0%) 1  46  299 (74.2%)  Subjects never found: i. Not matched with Ministry database  not applicable  40  ii. Letters returned "address unknown"  not applicable  25  iii. No response to our correspondence  not applicable  39 Total  104 (25.8%)  403 (100.0%)  Of the original sample of 403 women, 287 women were found, another 104 were never found and 12 were unavailable because of death (n=2), relocation to unknown locations (n=3), limited English (n=l), or could not be contacted to arrange time and venue of interview (n=5). Of the 138  287 women who were located, there was a 92.3% participation rate.  3.2.3.U (a) Refusal rate The refusal rate was small. Only 22 (7.7%) located women refused to participate, that is 5.4% of the original sample. The most common reasons for refusal were that their husbands did not want them to participate, that they were too busy, or that they did not want to divulge private information or thought anonymity was not assured. Although only 22 women formally refused, another 39 people (noted in the "subjects never found" category in Table 3.2) did not respond to correspondence sent to them regarding the study after their names and addresses were found in the Ministry of Health Linked database. It is not possible to know whether they did not receive the correspondence or whether their non-response was indicative of a passive mode of refusal.  3.2.3.ii (b) Teachers'participation Parents did not only consent to their own participation, but also that of their child and of their child's teacher since a further data source was teachers' responses to a brief written questionnaire. The process for soliciting participationfromthe teacher was multi-staged, requiring three levels of permission. First, 10 school boards were asked for permission to have teachers in their jurisdiction participate. Nine of the school boards granted permission and one requested additional permission from the local Teachers' Association, who agreed that their membership could participate. One school board sent our request to their Education Committee who did not approve their teachers' participation, citing that their teachers had been surveyed too often and were too busy to be surveyed again. Not all teachers' responses were lost in that district, however, because parents,  139  teachers and principals agreed that the individual teachers could participate since both the teachers and parents wanted to contribute to the study.  Secondly, each parent who participated was asked to give written consent for the child's teacher to participate, after being shown a copy of the teacher's questionnaire. The teacher was given a copy of the parent's written consent form and asked for written consent for his or her own participation. Finally, written permission was requiredfromeach teacher. The interviewer took the questionnaire and consent forms to the school and either waited while the teacher completed them, asked to have them mailed back, or picked them up on another occasion.  The teachers' participation rate was high. Completed questionnaires were receivedfrom242 teachers (91.3%). Four teachers (1.5%) refused to participate and 10 parents (3.8%) denied permission for their child's teacher to participate. The most common reason for the parent refusing the teacher's participation was that they thought that teachers were busy and should not be given further work.  There were nine other children for whom there were no teachers' questionnaires. Seven (2.6%) children were home-schooled and two (0.8%) moved within days of the interview (one was not found again and for the other it was not possible to find a teacher who knew her well enough to complete the questionnaire).  140  3.2.4 Variables and measures The dependent variables of interest were health, behaviour and academic abilities. Data were collected for each of these variables from three sources, parent, teacher and child. They were collected when the children were on average 8 years old. All data were collected during a 9-month period, May 1999 to February 2000.  Data for the independent variables were collected in the earlier stages of the study, that is, during a one-year period, 1990 to 1991, when the mothers were pregnant and during thefirstweeks of life of their children. Data for potentially confounding variables were collected during all stages of data collection.  The variables under study: Independent variables: • prenatal alcohol exposure, treated both as a dichotomous and a continuous variable • prenatal tobacco exposure, treated both as a dichotomous and a continuous variable • socioeconomic status, defined by parental education, parental occupations and income Dependent variables: • health of child at approximately age 8 • behavioural problems of child at approximately age 8 • academic abilities of child at approximately age 8 Each of these independent and dependent variables, their definitions, measurements, methods of data collection and data sources are described below. Table 3.6, found at the end of this chapter on pages 171-2, lists the variables, according to category, and the stage at which the data were collected.  141  3.2.4.i Independent Variables The independent variables are prenatal alcohol exposure, prenatal tobacco exposure and socioeconomic background. Data were collected for these variables during the 1990/91 data collection stages through self-reports by the mother.  3.2.4.i (a) Prenatal alcohol exposure There were four measures related to maternal alcohol use, two assessments ofriskand a selfreporting of actual alcohol intake at two time periods, before pregnancy and during pregnancy. The T-ACE test measures risk (Sokol et al, 1989) and it was administered twice, once during pregnancy in the physician's office (by the woman's physician or office staff), and for a second time during an in-depth home interview that took place within 8 weeks of delivery. All data regarding the measurement of prenatal alcohol exposure were collected through maternal selfreports. During the home interview, the interviewer administered the T-ACE test, that referred to tolerance level before pregnancy, and asked specific questions about alcohol consumption frequency and quantity prior to and during each trimester of pregnancy. Therefore, any recall data were reported within 10 months or less of the actual behaviour.  Measurement of alcohol risk level The T-ACE questionnaire (Sokol et al., 1989), administered during a office visit to the physician and repeated during the home interview, is a four-question screening tool with a summative score rangingfrom0 to 5.  142  All physicians in the geographic area were asked to participate in the 1990-1991 Vancouver Island research project. They were provided with training at local hospital sites in the use of the study instrument. Each physician was encouraged to participate by completing a T-ACE questionnaire as part of the routine pre-natal visit for each patient who was likely to give birth during the study period. The standard Provincial Prenatal Record was amended to include 10 questions about alcohol use, beginning with a question about whether the patient had ever had a drink of alcohol. If the response was positive, the patient was asked age of first drink, age she first got drunk, her preference for type of drink (beer, wine, liquor) and the four questions comprising the T-ACE questionnaire (Sokol et al., 1989) with reference to drinking prior to pregnancy. Those questions are shown in Appendix A.  The T-ACE does not ask about quantity and frequency of drinking and therefore does not measure consumption. Rather, it measures risk level, with a score of 0 indicating no risk and at the upper extreme a score of 5 indicating high risk level. The T-ACE screening tool was adaptedfromthe CAGE (Ewing, 1984) screening tool that has four questions about attempts to cut down on alcohol use (C), being annoyed by others suggesting the need to cut down (A), feeling guilty about one's consumption (G) and using alcohol in the morning as an eye-opener (E). Sokol et al. (1989) found that the CAGE did have good predictive properties except for the item about feeling guilty. This question was replaced with a question about tolerance (T) with scores for that question rangingfrom0 to 2 (0 if she does not get high, 1 if she needs only one drink to get high and 2 if she needs two or more drinks to get high). The other three questions score one point each, that is, the questions about being annoyed (A), cutting down (C) and using alcohol as an eye-opener (E).  143  An external reviewer found a sensitivity of 76 percent, 79 percent for specificity, 14 percent for positive predictive value and 79 percent for efficiency (Russell, 1994), using the original scoring criteria of Sokol, Martier and Ager (1989) that classifies women with a score of two or great as being "at risk".  T-ACE scores rangefrom0 to 5 were dichotomized for level or risk (risk or no risk), according to Sokol, Martier and Ager (1989). For the purposes of the Vancouver Island Pregnancy Study, the scores were categorized into three levels, low risk (scores of 0 - 1), medium risk (score of 2) and high risk level (scores of 3, 4 or 5). This categorization was found to be more appropriate, based on an examination of the Vancouver Island cohort data and an examination of associated risk variables (Armstrong, 2001; Armstrong et al., 2001b).  Numbers of women scoring in each of these categories have been reported previously (Armstrong et al., 2001a, 2001b), and are shown in Table 3.3 below, for both the total population with T-ACE scores and repeated T-ACE for the 403 randomly selected original sample, and for the sample included in the Stage 3 data collection. The correlation between the two sets of scoresfromthe different administrations of the T-ACE (that is, at Stage 1 and Stage 2 with n=403) was .53 (p=.001).  144  Table 3.3 Distribution o f T - A C E risk levels for population (Stage 1), random sample (Stage 2) and sample found at Stage 3. Risk level (and scores)  total population frequency (%) Stage 1  L o w risk level (scores 0 or 1)  1,511  (49.4%)  237  60.0  164  (61.9%)  Medium risk (score o f 2)  1,114  (36.4%)  128  32.4  85  (32.1%)  434  (14.2%)  30  7.6  16  (6.0%)  3,059*  (100.0%)  395  100.0  265  (100.0%)  H i g h risk level (scores o f 3, 4, or 5) Total  random sample frequency (%) Stage 2  random sample frequency (%) Stage 3***  •Data missing from 600 (16.4%) women from total population (n=3,65). **Data missing from 8 (2.0%) of the sample (n=403). ***Sample (n=265) refers to those participating at Stage 3, but data were collected for T A C E assessments at Stage 2.  Alcohol consumption The postnatal data collection involved interviews at the women's homes with 403 women randomly selected from the total population of women who delivered during the study period. Home interviews were one to two hours in duration and conducted by trained community health nurses. During the home interview, respondents were asked about their alcohol consumption. They were asked 15 questions about alcohol consumption, beginning with whether they had ever had a drink. The T - A C E questions were repeated, again referring to pre-pregnancy behaviour. In addition, during the home interview, they were asked about their preference for alcohol beverages during pregnancy and the amount and frequency o f drinking each type o f alcohol (beer, wine or liquor). They were questioned on drinking five or more drinks on any one occasion. Although respondents were not questioned on the period o f time they took to drink those five or more drinks, that quantity o f alcohol consumed over a 2-hour period has been considered to be binging  145  (Abel, 1998b; Abel & Kruger, 1995; Streissguth et al., 1994a; Jacobson et al., 1993a). Respondents were asked about quantity, frequency and type of alcohol beverage for each of four time periods, pre-pregnancy and each of the three trimesters of pregnancy.  These data regarding frequency and amount of alcohol consumed and period in which the alcohol was consumed were treated both as dichotomous data and as continuous data. As a dichotomous variable, analyses were completed for whether the women drank at all during pregnancy. As a continuous variable, total amounts of alcohol consumed for each trimester and over the length of the pregnancy were calculated. Data collected regarding alcohol use during pregnancy are summarized in Table 3.4 below.  Table 3.4 Data collected during home interview regarding alcohol consumption. 1. Type of alcohol used: beer, wine, liquor 2. Quantity (a) On a day when you drank beer (or wine, or liquor), how many cans/bottles would you usually have? (b) What was the greatest number of drinks you had in a single day in these periods (that is, before pregnancy, first, second and third trimester.)? (c) Did you ever drink more than that described on a usual occasion, such as on weekends or at parties? (d) What was the greatest number of drinks you had in a single day? 3. Frequency How often did you drink (categorized by each period relating to pregnancy (before pregnancy, first trimester, second trimester and third trimester): 1-2 times a week; 2-3 times a month; about once a month, or less than once a month? 4. Binging How often did you drink at least 5 drinks per day (categorized by each period relating to pregnancy): 2-3 times a month, about once a month, or less than once a month?  146  3.2.4.i (b) Prenatal and postnatal tobacco exposure Women were questioned about their tobacco use during Stage 2, that is, at the postnatal home interview. All data on prenatal tobacco exposure were collected through self-report of the mother about her own use or use by others in the household. As part of the structured interview, women were asked five questions about tobacco use: (1) whether anyone in the household smoked during the pregnancy, (2) whether the respondent smoked during her pregnancy, (3) how many cigarettes she smoked daily and what brands, (4) during which trimesters, if any, she had an aversion to tobacco, and (5) during which trimesters, if any, she had a craving for tobacco during her pregnancy.  For the purposes of statistical analyses, this exposure was treated both as a dichotomous variable (whether the women smoked and whether the fetus was exposed through others in the household smoking), as a continuous variable (calculated by multiplying thefrequencyand amount of use for the duration of the pregnancy) and as a categorical variable according to the number of cigarettes smoked by the mother daily (<10 per day, 10-20 per day or >20 per day).  Postnatal tobacco exposure data were collected during Stage 3. Respondents were asked about the status of their smoking, that is, whether they were currently smoking, had stopped or had never smoked. If they had stopped, they were questioned about the amount andfrequencyof their past smoking and for how long they had stopped. If they continued to be smokers, they were questioned about the amount andfrequencyof their current smoking behaviour and if they smoked  147  in the house. Two additional questions asked about smoke exposurefromanyone else in the household and the number of times per day the child was exposed to second-hand smoke.  3.2.4.i (c) Socioeconomic background From the literature review reported in the previous chapter, it was noted that the indicators generally used to place people along a socioeconomic gradient are income, parental education and occupation with other factors such as ethnicity sometimes being included (Murray et al., 1999; Green, 1970a). These three indicators were used for the current study:  Income. Respondents were asked to report their income level,froma list of 10 levels ranging from less than $10,000 to over $80,000. They were also asked about all sources of household income,froma list of 17 items, as well as their main source of income, under the assumption that some sources of income (such as welfare payments) were further indicative of income levels.  Education. Respondents were asked to report the highest level of schooling completed for each of four possible parents, birth mother and birth father, adoptive mother and adoptive father. In all but five cases, the mother was the respondent so she self-reported as well as reported for the child's father or step-father. In the five cases when the father responded, he self-reported and reported for the mother and step-mother. There were no adoptive mothers in this cohort. The question about parental education was asked at two times, Stage 2 (home interview after birth of child) and Stage 3 (during the 1999-2000 interview). For the purposes of internal checking, respondents were asked at Stage 3 if they had returned to school since the earlier interview. The  148  categories for levels of education were adaptedfromthe (National Longitudinal Study on Children and Youth (NLSCY) (Canada. Statistics Canada, 1997), with 14 levels of education ranging from grade 9 or less to graduate degree.  Occupation. Data on parental occupation were collected at Stage 3. Using the NLSCY (Canada. Statistics Canada, 1997) survey, the occupational categories had 21 groups according to the 1980 Standard Occupational Classification (Canada. Statistics Canada, 1981). These categories proved to be too specific for the needs of this study and for the sample size, so these categories were collapsed intofivegroups, according to the 1991 Standard Occupational Classification (Canada. Statistics Canada, 1993), rangingfromprofessional or administrative occupations to unskilled occupations.  To summarize, there are six indicators of socioeconomic background: amount of income, source of income, mother's education, father's education (or education of step-father), mother's occupation and father's occupation.  Each of these indicators was entered separately into the statistical analysis to examine their influence on the outcomes. Although there was an attempt to calculate one composite score for socioeconomic background by giving weights to income, education and occupation, according to the model suggested by Green (1970a), this became impractical, mostly because of the extensive list of occupations that would require more precise ranking. There was also an assumption that each of these indicators could have different effects on outcome variables because they operate  149  through different causal pathways.  3.2.4.U Dependent Variables The dependent variables are health, academic abilities and behaviour. Data were collected from three sources for these variables: parent, child and teacher. Data on the health and behaviour variables were gathered through second-hand reportingfroma parent andfromthe child's teacher. Data on the academic abilities were collected by testing the child and by second-hand reporting from the child's teacher. The specific instruments developed for the data collection were discussed in more detail below, and sources of data collection for measurement of these dependent variables are shown in Table 3.5. The table shows that each dependent variable had more than one data source; for example, indicators of academic abilities were measured both by testing the child and by asking the teacher to rate the child's performance compared to others in the class. Each of these variables, shown in Table 3.5, are described in detail below.  Table 3.5 Sources of data for dependent variables at Stage 3 data collection. Data source Variable  Mother's report of child  Teacher's report of child  Child  Health  yes  yes (1 question)  no  Behaviour  yes  yes  no  Academic abilities  yes  yes  yes  3.2.4.H (a) Health Health was measured by three indicators: general health rating, a composite of various measures of  150  activity and morbidity, and absenteeism from school. In addition, health was measured in both general terms (that is, general state of health) and according to illness in specific terms (for example, number of illnesses or hospitalizations).  Rating of health, compared to others the same age, was measured through second-hand reporting from the mother (except in the five cases where the father was the respondent), who was asked the question: "In general, how would you rate your child's health compared to others of the same age?" Although there have been various attempts to measure health using questionnaires or instruments (for example, Donovan et al., 1993; Chambers et al., 1987), one of the most widely used measures is the single-question self-reported health status question asking respondents to rate health status as excellent, very good, good, fair or poor (Ratner et al., 1998). Ratner et al. (1998) refuted earlier studies in their reporting that this question is useful only in relation to physical health status, with respondents interpreting the word "health" to mean physical health. The number of items on the scale varies (usually three or four), but to be consistent with the Canadian NLSCY (Canada. Statistics Canada, 1997),fiveoptions were used in this research.  This measure allowed comparison of general health with that of others within the same age group. In addition, the mother was asked to rate her own health compared to others of similar age and to rate the health of other children in the family compared to others the same age. This allowed for comparisons between the general health rating of the subject and others in the family. For each of these health rating questions, an open-ended question followed, asking the reasons for giving such a response.  151  In addition, data were collected on indicators of illness or health through a set of questions specific to activity level and generally regarding health. Most of these questions were adapted from the NLSCY (Canada. Statistics Canada, 1997). These indicators and their types are listed below: Health problems: • description of health problems, both current and past history (qualitative data collection, categorized), • description of major illnesses during lifetime (categorized from qualitative data), • number of hospitalizations both during past 12 months and since birth, length of time in hospital and reasons (ratio data), • frequency during past year of contact with various health care professionals, including physicians, pediatricians or other doctors, psychologists or psychiatrists, public health nurses or nurse practitioners, social workers, child welfare or children's aid workers and others (ordinal data).  Activity levels questions: • level of physical activity compared to other children of the same age, measured on a 5-point scale (interval data), • sleep requirements compared to other children, measured on a 5-point adjectival scale (interval data). Illness or disability-related questions: • use of specific medications (ventolin or inhalers for asthma, Ritalin, tranquillizers, anticonvulsants or anti-epileptic pills or other medications) (nominal data), • causes for worry or unhappiness (qualitative data), • abilities (in sight, hearing, speech, mobility, dexterity), feelings, memory, thinking and problem-solving (nominal data), • pain and discomfort (nominal data), • serious injuries:frequency,type and cause (various types of data, including nominal), • asthma and other long-term conditions (nominal data).  Parents were asked to report the height and weight of the child and that of the parents. Data were treated as dichotomous, categorical, or continuous, as appropriate according to the variable. For example, whether the child had problems with vision or hearing was treated as a dichotomous variable, while those data for which scales were used (such as the extent to which the child is 152  worried) were interval data.  The third measurement used to indicate health status was the frequency of absenteeism from school, data that were collectedfromthe teachers. This may not be a valid measure of state of health if used as the only measure to indicate health status, since there are various reasons other than the student's health that could cause absenteeism, however, it was used as an indicator contributing to the other measures, rather than as a stand-alone measure. This question also offered the possibility of another data source, namely the child's teacher, to contribute to data regarding health. Having an additional data source was deemed important because mothers' evaluations of their children's health can be influenced by many factors beyond their actual health status, including culture, socioeconomic status and marital position (Angel & Worobey, 1988; Zola, 1966).  These data regarding absenteeism were categorical: 0 to 3 days absent, 4-6 days absent, 7-10 days and 11 or more. The number of days absent per month was calculated according to the teachers' response and when the questionnaire was completed. All questionnaires were completed after at least 3 months of the school being in session (that is, none were completed before the end of November).  3.2.4.ii (b) Academic abilities Academic abilities were defined as the abilities of the child compared to normative age-specific standards in reading, spelling and arithmetic. Three measures were used: assessment of children  153  using the revised Wide-Range Achievement Test (WRAT-3) (Wilkinson, 1993; Jastak and Wilkinson, 1984; Jastak and Jastak, 1978), the teacher's rating of the child compared with others in his or her classroom, and the parent's assessment of the child's difficulty with core subjects.  The WRAT (Wilkinson, 1993; Jastak and Wilkinson, 1984, 1979) is a test of academic abilities for childrenfrompre-school ages to grade 12. It can be used for either individual testing or for groups of children. The third revision (WRAT-3) was used in this study and each child was tested individually. Its purpose was to test abilities in three standard subjects, spelling, reading and arithmetic, comparing individualized performances with others of the same age of a normative, albeit non-Canadian, standard.  The WRAT-3 has three subtests: (a) decoding skills of reading (that is, pronouncing printed letters and words), (b) spelling (measuring the ability to write one's own name and write words from dictation) and (c) arithmetic (counting, reading number symbols, solving problems given orally and computing written problems). The three subtests are intercorrelated, indicating these measures may have a developmental component (Wilkinson, 1993).  The WRAT-3 has acceptable reliability and content and construct validity (Wilkinson, 1993). Test-retest reliability correlations rangefrom.91 to.98 on tests of the WRAT-3 given to a sample of 142 subjects, ages 6 to 16 years (Wilkinson, 1993), indicating good stability, although these reliability tests are not well-described by the test's authors. Reliability testing (test-retest) has shown to be high (r=94) for the 7-10 year age group (Sattler, 1992) on the WRAT-R, the  154  previous edition on which the WRAT-3 was based. In addition, the test appears to be reasonably consistent with other tests, although it does appear to yield lower scores than some tests with similar purposes, such as the widely-used Kaufman Test of Educational Achievement (Prewett et al., 1991a, 1991b; Kaufman & Kaufman, 1985).  For the purposes of this study, the advantages of using the WRAT-3 were that it could be administered by interviewers who were not specially trained in psychology or education and it could be administered individually in a reasonably short time period (less than A hour). It also had l  reasonable reliability for the age cohort under study as well as for older children so research could use the same assessment tool if follow-up of these children were to be done. Furthermore, WRAT-3 and its predecessors have been widely used and scores are standardized on U.S. national samples according to age, gender and urban/non-urban stratifications.  A disadvantage of this test is its limitation in two areas of measurement; neither reading comprehension nor complex arithmetic computations are tested. A further limitation is that no standard scores are provided for varying socioeconomic stratifications nor according to cultural backgrounds, such as Canadian children or children of aboriginal heritage.  A further data source for the child's academic abilities was the teacher who had known the child for a minimum of two months. Data were collectedfromthe teacher through the use of a short questionnaire designed specifically for this study (shown in Appendix B). Some of the questions were adapted for usefromthe work of Conry (1999). The teacher rated the child, compared to  155  other children of the same age, on overall abilities, pre-reading or reading or language skills, numbers or arithmetic skills, movement, physical education or sports abilities, citizenship or social interaction, and school or academic motivation. Teachers reported on their direct knowledge or opinions of the children. The data from all of these questions were interval.  A history of repeating or skipping grades was collected. Although this may have been a useful indicator of academic history, current trends suggest that few children repeat or skip grades in elementary school in British Columbia. These two questions were adaptedfromthe NLSCY (Canada. Statistics Canada, 1997) and the data source was the mother's questionnaire. It was expected that the mother or father would be able to answer questions about skipping or repeating grades more readily than the teacher who may be inconvenienced in seeking that datafromschool records.  3.2.4.ii (c) Behaviour Behaviour was defined in terms of the prevalence of both social competencies and antisocial problems. An assessment tool with 8 subscales was used to measure Hyperactivity, Inattention, Anxiety, Emotional Disorder, Conduct Disorder, Physical Aggression, Indirect Aggression and Prosocial Behaviour. It is referred to as the Behaviour Rating Scale in this thesis. It is shown in the parents' questionnaire and the teachers' questionnaire in Appendix B.  The assessment instrument had been adapted previouslyfromthe Child Behavior Checklist (CBCL) (Achenbach et al., 1987; Achenbach and Edelbrock, 1981, 1978; Achenbach, 1978) for  156  use by Canadians for the NLSCY (Canada. Statistics Canada 1997) and other research (Offord & Lipman, 1996; Boyle et al., 1987; Offord et al., 1987). Originally, the CBCL was devised to assess and categorize clinical behavioural problems of children (Achenbach et al., 1987; Achenbach & Edelbrock, 1981, 1978; Achenbach, 1978). It was used with normal children to test validity, allowing standardization of normal populations of English speaking 4-16 year-old American children (Achenbach and Edelbrock, 1981) and since then has been widely used for clinical purposes and for research. The Canadian adaptation (Offord and Lipman, 1996; Boyle et al., 1987; Offord et al., 1987) has contributed a more positive slant through the inclusion of a 12-item subscale on Prosocial Behaviour and has changed some of the vocabulary. This adapted version has parents' and teachers' versions and both were used in the NLSCY (Canada. Statistics Canada, 1997).  Both versions of the behaviour scale were used for the current study. Teachers and parents are asked to report on the behaviour of a child in self-administered questionnaires. The parents' version of the instrument consists of 47 items describing the child's behaviour, emotions or attitudes and social relations. For example, respondents were asked whether the child "tells lies or cheats", "seems to be unhappy, sad or depressed", or "shows sympathy to someone who has made a mistake." Respondents rated each child on a three-point scale, "not true", "sometimes or somewhat true", or "often or very true." The teachers' version is the same in all questions except that it drops one item ("steals at home") about which the teacher would have little or no information.  157  There were three advantages to using this instrument. First, it had been tested and used on a large national Canadian cohort for the NLSCY (Canada. Statistics Canada, 1997). Second, its subscales measure attributes (such as inattentivity) that are important to this study at least on a theoretical basis. Third, it was practical to use in that it is self-administered, could be easily explained by interviewers who were not psychologists, and it has both parents' and teachers' versions. The use of both parents' and teachers' rating scales was deemed important because assessments have differed according to rater and according to the contexts in which the rater has known the child (Kohen et al., 1997; Verhulst & Ende, 1991; Verhulst & Akkerhuis, 1989; Achenbach et al., 1987).  Scores for the behavioral scale were calculated in two ways, first with a total summative score and second calculating scores for each of the subscales, with all scores being treated as interval-level data. Correlational analyses were conducted to compare scoresfromboth the parent and the teacher.  3.2.4.iii Other variables Several other variables were measured as potential confounders. Data for each of these were collected as part of the self-administered parents' questionnaire. They include: a. physical height and weight of both parents; b. mother's access of educational, social or health services when the child was an infant; c. mother's well-being within year of giving birth (e.g., duration of breast-feeding, post-partum depression);  158  d. CAGE questionnaire (Ewing, 1984) regarding current alcohol risk level of mother; e. changes in home, school or parenting (that is, adoption) of child; f.  family factors (age of mother, birth order of child, number of siblings, marital status of mother;  g. child's activities and hours spent at each (e.g., television and video games, reading, sports); h. social support (of mother and of other adults in child's life); i.  socialization of child (time spent playing with other children, ability to get along with other children and ability to get along with siblings);  j.  nutrition (mother listed the child's food intakefromthe day previous to the interview, with the food intake later being categorized according to type, such as bread or dairy products).  Data on each of these variables were collected by self-reporting of parents or by second-hand reporting by the parents about the child. In most cases, variables were dichotomous except in a few cases when variable responses were rated on a scale.  3.2.5 Study instruments and means of administration Study instruments were designed to collect datafromthree sources; mother (or other parent or guardian), teacher and child. Instruments used to collect datafromparents andfromteachers are described below and these instruments are included in Appendix B.  3.2.5.i Parents' questionnaire Data were collectedfromthe mother using a self-administrated written questionnaire devised for the project. As noted above, there werefiveexceptions to the mother self-reporting or reporting  159  on her child, these cases being when the father was the primary caregiver and he completed the questionnaire.  Although it was self-administered, an interviewer was present during the time of administration to answer questions readily. The interviewer was trained to encourage questionsfromthe mother for clarification of questions. In many cases, the mother had no difficulty completing the questionnaire, in other cases individual questions were clarified and in a few cases, the interviewer read through the total questionnaire for the respondent.  There were 18 exceptions to the interviewer being present, all because the questionnaire was sent by mail. In these cases, the child had moved to distant locations (13 in British Columbia, 2 in Alberta, 1 in Saskatchewan, 1 in Ontario and 1 in the Northwest Territories) where interviewers were not sent for practical reasons. Although it is not ideal that some of the respondents did not have an interviewer present as planned in designing the study and the questionnaire, it appeared more important to include these subjects in spite of this deviationfromthe original research design. For those completing the questionnaire without the presence of an interviewer, follow-up telephone calls were made before and after completion to clarify the meaning of any questions and to encourage completion of the total questionnaire.  Development and pre-testing Many of the items for the questionnaire were used in the NLSCY (Canada. Statistics Canada, 1997) and were selected so that comparisons could be made with national data. Other questions  160  were devised specifically for this research. Thefinalproduct is a 97-item questionnaire divided in the following sections: child's background (26 questions), child's health and family health (27 questions), environment (22 questions), child's education (13 questions), child's activities (8 questions) and a 46-item Behaviour Rating Scale. Most mothers took 3/4 to 1 hour to complete the questionnaire, and the range was Vi to 2 hours.  The NLSCY survey questions were designed to be asked verbally, either in person or by telephone. The current design required two data sources (mother and child), thus to conserve scarce time and financial resources, it was decided to adapt the NLSCY questions into written questions. The mother responded to the written questionnaire while the child was being tested, thereby limiting the amount of time the interviewers were with the family.  The questionnaire was pre-tested with 13 women (7 at first pretest and 6 at subsequent pretest), all with early elementary school age children, to estimate the time required for completion of the questionnaire, to test for discriminating responses and to get feedback on the comprehensibility of the questions and the instructions. The questionnaire was revised after the pretests according to the input to clarify instructions or wording of some questions, not including thosefromthe NLSCY orfromestablished instruments.  3.2.5.U Teachers' questionnaire The teachers' questionnaire was a short, three-page, eight-question, self-administered, written questionnaire. The usual length of time for completion was eight minutes.  161  There was a temporal process for administering the questionnaire. First, the interviewer interviewed the mother, obtained signed written consent forms including permission to approach the teacher, and to ascertain the child's grade and name of the teacher and the school. This was followed by the interviewer delivering a package to the school that included a teachers' questionnaire, written consentsfromthe mother and the relevant school board and a letter describing the study. The interviewer either waited for it to be completed, picked it up on another day, or left a self-addressed stamped envelope for it to be returned to the interviewer. The interviewer was not always present during the completion of the questionnaire because of the varying time pressures on teachers.  The questionnaire included two questions establishing the relationship with the child (amount of time teacher saw the child each week, length of time teacher knew the child), one question about absenteeism, and four questions asking the teacher to rate the child on abilities and activities compared to other children of the same age and gender (Conry, 1999). Thefinalquestion was the 46-item Behaviour Rating Scale described above. A copy of this instrument is in Appendix B.  Development and pre-testing Fifteen questionnaires were completed by six different teachers to pre-test the instrument. The pre-testing assessed: (a) testing procedures such as amount of time required to complete the questionnaire and its instructions; (b) intra-rater reliability; and (c) inter-rater reliability. Since the majority of the instrument comprised the Behaviour Rating Scale, a standard tool, limited pretesting was required. Because of the inputfromthese teachers, a small number of changes  162  were made to the wording of new questions and to clarify the instructions, but the Behavior Rating Scale was not altered.  3.2.5.iii Child's WRAT-3 test The child was given one test, the Wide Range Achievement Test 3 (WRAT-3), described above. Length of the assessment and completion time of this test depends on the level of achievement with those at higher levels being given more questions. This group took an average of approximately 20 minutes and none took more than 30 minutes for completion.  The assessment was administered at the same time the mother was completing her questionnaire. Prior to testing, mothers gave written consent for their child to participate. The interviewer explained the test procedures and administered the test. Interviewers reported that the children were generally pleased to do the test, willing to participate and felt important by being included in the study. There werefiveexceptions to this procedure, all because the families lived in places other than Vancouver Island so no interviewer was present. For these cases (four individuals plus one set of twins), three school teachers, one school psychologist and one school principal administered the assessment and all were experienced in the administration of the WRAT-3.  3.2.6 Analytic procedures All data were examined for their distributions. Chi-square analyses were completed to compare the samplesfromthe randomly selected sample of Stage 2 (n=403) and follow-up sample of Stage 3 (n=265) of the study, examining differences with respect to maternal age, socioeconomic factors,  163  drinking patterns, birth weight or other factors that could influence further analyses. Chi-square analyses had been completed previously to compare the population (Stage 1) and the randomly selected sample of Stage 2 (Armstrong et al., 2001a). From the sample at Stage 2, the subjects were classified according to participation, that is, participated at Stage 3, lost for the follow-up study, or refused. These three groups were compared according to the same demographic, maternal behavioural and social factors, and infant variables.  Correlations were interpreted as weak if they were below .3, even if statistically significant, since there is very low predictive value below that level. They were interpreted as moderate if they fell between .4 and .6, and as high if above .7. Analyses were considered statistically significant at p=.05, and were reported as .05, .01 or .001.  Bivariate analyses were completed to determine the independent relationships between each of the risk conditions and each of the dependent variables. Cross-tabulations, univariate and multivariate analysis and logistic regression analysis were conducted to estimate the relationships between the independent and dependent variables, controlling for the covariates.  Finally, a series of tests were conducted in which both the main effects and interaction effects of variables were compared with the scores of children who were exposed to none of the measured risks. These were one sample t-tests with the null hypothesis that there was no difference between the scores of those exposed to norisks(of the various variables) and either the mediumrisklevels or highrisklevels. The variables included in these statistical models were smoking during 3  164  rd  trimester (dichotomized variable according to whether exposed or not to smoking during 3  rd  trimester that is essentially throughout pregnancy), current smoking status of the primary caregiver (dichotomized variable as to current smoker or not), T-ACErisk(Stage 2 adrninistration), alcohol consumption patterns (three levels, abstainer, social drinker and heavy drinker), each of the five indicators of socioeconomic status (each of which had three levels ofrisk),and marital status (dichotomized according to married or not currently married). Separate analyses were required when adding each of the five indicators of socioeconomic status to the models, since the analyses became too restrictive in cell size if all were added into one model. Scores were compared between those with the lowestrisklevel and other levels of risk. The Beta, or mean difference in scores, of lowestrisklevel for these variables (for example, no smoking or alcohol use during pregnancy) were set as the baseline for comparison. Regressions were calculated for the main effects of each of the variables and for the effects of the combination of variables on the three academic scores, with the health ratings and with the behavioural outcomes.  A series of Independent sample t-tests for equality of means were calculated to compare scores in reading, spelling and arithmetic of those in the high-risk categories and low-risk categories. This final step in the analysis was to include all three independent variables, alcohol exposure, tobacco exposure and socioeconomic background into the model. Extreme levels of the combined variables (for example, lowest alcoholrisklevel and lowest tobacco exposure and lowest socioeconomic indicator are combined and compared with highest levels these variables. At one extreme corner is a cell that combines the highestriskof these three dimensions and at the other extreme corner is a cell that combines that constitutes the least risk. A comparison was made  165  between the combined high risks and the combination of no or lesser risks, and subsequently it was planned to move in a level on each variable to compare lesser extremes..  The highriskgroup consisted of women who had a high T-ACE risk level, drank alcohol during their pregnancy and smoked cigarettes throughout pregnancy and had the lowest indicators of socioeconomic status. Initially, tobacco exposure during pregnancy was included in the model but when no effect was observed on any of the academic abilities, consistent with all previous analyses, it was excludedfromthe model that included multiple independent variables! This permitted less restriction in the analyses when including multiple variables. Socioeconomic indicator variables needed to be added to the model separately since including all at once was restrictive on the analyses, resulting in the degrees offreedombeing exhausted. By deleting tobacco exposure from the model, it was possible to keep three categories for each variable, that is three levels of socioeconomic status and threerisklevels of T-ACE and alcohol consumption (abstainers, social drinkers and heavy drinkers) and to conduct a series of contrasts between the various levels for each of the academic subjects. The remaining variables combined to determinerisklevel were each of the socioeconomic indicators and T-ACE risk and alcohol consumption patterns.  In each case, the results of Levene's test for equality of variances guided whether the t-test for equality of means assumed equal variances or equal variances not assumed.  Three aspects of the analyses require further detail, that is, how twins, missing data and nonnormally distributed data were treated.  166  Twins There were two sets of twins. The results might have been skewed by including two children's outcomes but only one mother's behaviour, therefore only one childfromeach