CANCER CLUSTER DETECTION INBRITISH COLUMBIA SCHOOL DISTRICTS1983—1989byRHONDA JEAN ROSYCHUKB.Sc., The University of Alberta, 1992A THESIS SUBMITTED IN PARTIAL FULFILLMENT OFTHE REQUIREMENTS FOR THE DEGREE OFMASTER OF SCIENCEinTHE FACULTY OF GRADUATE STUDIESDepartment of StatisticsWe accept this thesis as conformingto the required standardTHE UNIVERSITY OF BRITISH COLUMBIAMarch 1994©Rhonda Jean Rosychuk, 1994In presenting this thesis in partial fulfilment of the requirements for an advanceddegree at the University of British Columbia, I agree that the Library shall make itfreely available for reference and study. I further agree that permission for extensivecopying of this thesis for scholarly purposes may be granted by the head of mydepartment or by his or her representatives. It is understood that copying orpublication of this thesis for financial gain shall not be allowed without my writtenpermission.(Signature)Department ofThe University of British ColumbiaVancouver, CanadaDate________________DE6 (2188)AbstractA disease cluster is an aggregation of occurrences of a disease. The observation of a perceived excess number of similar illnesses is termed disease clustering.Statistical tests for disease clustering investigate if the observed pattern of casesin at least one geographical area could possibly have happened by chance alone.This pattern may be spatial, temporal, or both. Investigating possible cancerclusters in British Columbia for the period of 1983—1989 inclusive is the objectiveof this thesis. Whether or not cancer clustering appears near pulp and paper millswithin the province is of specific interest. The geographical units upon which ourinvestigation will be based are the B.C. school districts. The variation in sizeand population demographics among districts requires a cluster detection methodwhich is considerate of the underlying population distribution within the studyregion. School district population size diversity requires a modification to theBesag and Newell (1991) method. This modification is implemented with B.C.school district data and several possible clusters are detected for various types ofcancer.11ContentsAbstract iiTable of Contents iiiList of Tables vList of Figures viAcknowledgements vii1 Introduction 12 Literature Review 42.1 General Tests 52.1.1 The Whittemore et al. Test 62.1.2 The Geographical Analysis Machine 92.1.3 Cluster Evaluation Permutation Procedure 102.1.4 Besag and Newell Statistic 122.2 Focused Tests 152.2.1 Besag and Newell Focused Test 152.2.2 Stone Statistic 162.2.3 Wailer et al. Statistic 173 Data Description 203.1 Cancer Data 203.2 Population Data 213.3 Centroid Determination 234 Analysis 314.1 Combined Year Analysis 324.2 Modified Besag and Newell Method 371114.3 Yearly Analysis.425 Discussion of Results 455.1 Combined Year Analysis 455.1.1 General Male Findings 455.1.2 General Female Findings 565.1.3 Focused Findings 635.2 Yearly Analysis 665.2.1 General Male Findings 665.2.2 General Female Findings 805.2.3 Focused Findings 926 Conclusion 94References 99Appendix: Tables 101ivList of Tables1 Cancer Sites Investigated with Patient Counts2 British Columbia Pulpmills3 Nearest School District Centroids4 Nearest School District Centroids to Foci5 Combined Year Male Results for Site 1 Lip, k1,. . . ,6 Combined Year Male General Results by Site 1983—89 .7 Combined Year Female General Results by Site 1983—898 Combined Year Male Focused Results by Site 1983—899 Combined Year Female Focused Results by Site 1983—8910 Yearly Male General Results by Site 1983—8911 Yearly Female General Results by Site 1983—8912 Yearly Male Focused Results by Site 1983—8913 Yearly Female Focused Results by Site 1983—89101102103104105107125136• . . 137• . . • 138• • . . 154• . . 165169VList of Figures1 British Columbia School Districts 1983—89 222 Population Distribution by School District 1989 233 School District Centroids in British Columbia 1983—89 284 Pulpmill Locations in British Columbia 1983—89 305 Minimum Relative Risks for Significance at 5% 36viAcknowledgementsSeveral individuals were instrumental in this work. My deepest gratitude to Dr. Nhu Lefor his guidance and encouragement. Dr. John Petkau provided valuable assistance andcommentary. The Division of Epidemiology, Biometry, and Occupational Oncology atthe B.C. Cancer Agency provided financial support. Dr. Pierre Band provided valuablesuggestions. Mr. John Smith’s help in getting a B.C. map with school districts is appreciated. School district populations were provided to the Division by Mr. Dave O’Neilfrom the Population Section, Ministry of Finance and Corporate Affairs. In addition,the UBC Statistics Department provided financial support during my studies which wasgreatly appreciated.R.J.R.vii1 IntroductionHealth authorities are often alerted to suspected cancer clusters by members of thepublic. A cluster is an aggregation of cases of a disease. The aggregation may bein space or time or both. In space, cases may occur physically close to each otherwhile in time, cases may occur temporally close. These local reports may be made byobservant clinicians who notice a perceived increase in the number of people afflictedwith a specific disease or by individual citizens who have anecdotal observations aboutan illness afflicting their neighbourhood, family, or friends. Health authorities then needto respond to inquiries about areas of seemingly excess disease in a timely fashion, eitherto assure the public the numbers of cases observed do not indicate clustering or to informthe public and initiate further investigation if strong evidence of clustering is found.Generally speaking, the majority of these local reports are false alarms. Becausethorough investigations of these reports would require extensive resources and involvemuch time, detailed studies on all local reports of clustering are impractical. On theother hand, there probably are real cancer clusters which are caused by some specificrisk factors which may not be reflected by coincidental observations. This situation maypersist for several years and as a direct result many more individuals will be exposed tothe risk factors.To overcome these potential problems to a certain extent, health departments needobjective and standardized approaches to detect possible clusters with or without localreports. One such method is to use statistical techniques for disease cluster detectionin conjunction with available databases such as cancer registries. Statistical tests fordisease clustering investigate if the observed pattern of cases in administrative areascould possibly have happened by chance alone. The main advantages of this approachare that it is cost effective and it allows for timely cluster detection because thesedatabases for populations of the administrative areas and their associated cancer cases1are generally available.A surveillance system, where administrative regions are monitored for high diseaserates, identifies areas requiring further study and has the potential to target resourcesfor the most plausible clusters. Such a system allows for responses to lay reports ofexcess cases which are easier, timelier, and less expensive than individual thoroughinvestigations. The detection of clusters may also lead to etiological findings or suggestpossible causes.Identifying possible cancer clusters in British Columbia for the period 1983—89 is theobjective of this thesis. Whether or not cancer clustering appears near puip and papermills within the province is of specific interest.Particular information is required for a cancer cluster detection method to operate.Both disease data and population data must be known for the administrative zones underinvestigation. The administrative unit upon which our investigation will be based is theB.C. school district. Data about cancer cases in B.C. during 1983 to 1989 is providedby the British Columbia Cancer Agency (BCCA). For each patient, the year of cancerdiagnosis, sex, type of cancer, and age at time of diagnosis are contained in the database.The patient’s age is categorized into an age group. Information about the patient’sresidence at the time of diagnosis is also recorded, including address, postal code andschool district. During the study period 74 school districts existed and 22 pulpmillswere located within 17 of the school districts. Population data is known for each schooldistrict, age group, sex, and year.British Columbia school districts have a variety of shapes and sizes. The population within each district can range from below 10, 000 to above 100, 000. The clusterdetection method used should allow for the differences in population size among theschool districts. Of the choices available, a procedure proposed by Besag and Newell [1]appears to be appropriate to deal with our clustering questions.2The Besag and Newell method requires the identification of the nearest neighboursof each school district. The spatial relationship between school districts is based on theircell centroids. Cancer cases which have complete postal code information are used toestimate the location of each school district’s population-based centroid. An estimate ofa cancer patient’s location of residence is calculated using its postal code and a softwareconversion package called the Postal Code Conversion File [12].The Besag and Newell method was designed for very small administrative zones withsimilar population sizes. The dissimilarity of population sizes in B.C. school districts requires a modification to their method. This modification is used in yearly and combinedyear analyses for each sex and cancer type.The sections which follow give a detailed account of the analysis. Section 2 provides a summary of the statistical cluster detection methods which are found in theliterature. The data sources and centroid determination are detailed in Section 3. Section 4 describes the analysis based on the Besag and Newell method and the proposedmodifications to the approach. A discussion of the individual results obtained from themodified approach follow in Section 5 with concluding remarks in Section 6. Tables ofresults are found in the Appendix.32 Literature ReviewMany methods have been proposed in an attempt to provide detection plans whichallow for timely responses, with or without a local report of a suspected cluster. Todetect spatial clustering, general tests and focused tests are two varieties which emerge.General tests are designed to detect clusters within the overall pattern of disease inthe complete region. The null hypothesis is defined as complete randomness, i.e. eachindividual in the population is assumed to have an equal chance of developing thedisease. As the terms general and focused, introduced by Besag and Newell [1] suggest,the general test is used when one has no particular alternative in mind and the focusedtest is used when one has a particular alternative hypothesis that incidence rates willbe higher near sources of contamination. The general and focused tests to be describedin this section include those developed by Whittemore et al. [13], Openshaw et al. [6],Turnbull et al. [9], Besag and Newell [1], Stone [8], and Waller et al. [10], [11]. Generallyspeaking, these tests allow for the underlying populations of the administrative regionsto be different, which is necessary.Other cluster detection techniques, such as those proposed by Pinke and Nefzger [7],Knox [4], and Mantel [5], will not be discussed. Generally speaking, these methodologiesand many others have been proposed over the years where unrealistic assumptions aboutthe underlying population at risk within the study region are made. It is usually assumedthat the administrative areas within the region have approximately the same populationand age distribution. These techniques are unsuited for comparing disease rates betweendense and sparse regions. For example, a retirement community may be identified as acluster for a disease that affects only elderly people when age distribution is assumedto be uniform over the study region or a city may be identified as a cluster because ofits high numbers of cases due to dense population. Such methods are unsuitable for oursituation and will not be discussed.42.1 General TestsWe now describe several test procedures which could be classified as general tests, alongwith their advantages and disadvantages. The common null and alternative hypothesesin these test procedures are:H0: Every person is equally likely to contract the diseaseindependently of other cases, and the location oftheir residence,H1: Not H0.Suppose that the region of interest is divided into I cells. These cells may be administrative areas of interest such as health units, school districts, political ridings, censusdivisions/sub-divisions, or enumeration areas. The population residing in cell i is denoted by n, i = 1,. . . , I. For cell i, the number of incident cases of the disease isdenoted by a random variable C with observed value C:, = 1,. .. , I. The total numberof cases in the entire study region is C.= >j C. The total population and the totalnumber of observed cases are denoted by n.=n and c.=c, respectively.For a rare disease, the number of incident cases within a particular cell may be modeledas a Poisson random variable. The hypothesis above is equivalent toH0: The C, i = 1,. . . , I, are independent Poissonrandom variables with E(C:) nO,where 0 denotes the individual baseline risk of contracting the disease. With this formulation, the baseline risk of the disease multiplied by the population of a cell yieldsthe expected number of diseased individuals for that cell. This model assumes thatthe disease investigated is not hereditary or contagious. The model could accommodatesuch diseases if more features were added.The underlying population must be available for most of the methods detailed below.The number of cases per cell is also necessary. Several procedures require the centroid5of the cell, if not the actual location of each case.2.1.1 The Whittemore et a!. TestWhittemore et al. [13] developed a method for detecting spatial clusters for chronicdiseases. Their rationale for examining only spatial clustering is that chronic diseasescaused by a pollution source tend to be diagnosed long after exposure. Cases may beclose in space but not in time due to the fact that the disease development is long andwill vary from patient to patient. Their proposed test statistic is based on the meandistance between all pairs of cases. Cell centroids may be used if precise locations of thecases are not available. The degree of clustering within the entire study region is relatedto the mean distance between all pairs of cases. The smaller this mean, the closer thecases are together and the more clustering is indicated.Specifically, the population of an administrative area is partitioned into S subgroupscalled strata. For example, the strata could be defined by sex, age group, or race.Each stratum is assumed to have a different disease incidence rate, O. The numbersof observed cases and population in cell i for stratum .s are denoted by c and n,respectively, i = 1,. . . , I, s = 1,. . . , S. The random variable C8 represents the numberof cases in cell i and stratum .s. The G’s are assumed to be mutually independentPoisson random variables under the following null hypothesis,H0: E(C23) = O3n, i = 1,.. .,I, .s = 1,.. .,S.Under this hypothesis the vector of strata-specific total disease numbers is a sufficientstatistic for the unknown parameters Ox,. . ., 0s. Given its value c = (c.1,. . . , c.s), thecell frequencies (c13,.. . , Cl3) are mutually independent multinomials with size c.. andestimated probability vector= (n . . ,n13)/n.8 s = 1,. . . S. (1)6The mean distance between all pairs of cases, regardless of their strata membership, isconsidered as the test statistic. Let p(qi, q) be the distance between the q1th and theq2thl case. The test statistic is given by,w= ( )‘ p(qi,q). (2)ql<q2To calculate the distribution of W under the null hypothesis, the cases are assumed tobe located at the centroid of the cell. The distance between cases within a cell will bezero by this assumption. The total distance between all pairs of cases will be the sumover all cell pairs of the product of the number of cases contained in the pair of cellswith the distance between the pair’s cell centroids. This total distance can be written as1/2c..2rTDr, where r = c:1 Z= (c13,. . . , ci3) is the vector of relative cell frequenciesand D is an I x I matrix whose (k, l)th entry is the distance between cells k and 1.Equation (2) can then be written asW= ••rTDr. (3)c —1Since r is the mean of the S multinomial vectors (c13,. . . , Cl8), for s = 1,. . . ,S, itsexpected value is equal to E(R) = = c..’ c.8p. Let U3 = diag(p3)— 33T,= 1,... , S. Then U = c..1 c.8U3 is the mean of the multinomial covariancematrices c.3U and the covariance matrix of r is c..’U. Whittemore et al. show that Wis asymptotically normally distributed under H0. Assuming that 1im.,c.3/c. ) > 0for s = 1,...,S, then under H0, as c.. — 00,1 W_irTThr v-/ —1\r(o,i). (4)2 /TDUDWhittemore et al. used their statistic to examine clustering of 63 cases of anal andrectal squamous cell carcinoma in San Francisco during 1973 — 1981. There were threeage groups, two races and the two sexes yielding S = 3 x 2 x 2 = 12 strata, for thatstudy. With the stratification, the results indicated statistically significant evidence forclustering.7This Whittemore et al. method allows incorporation of the underlying facets of thepopulation. When administrative areas are not homogeneous, the population distribution within each area must be considered. Methods may produce clusters becauseof population variation. If the population within the stratifications at high risk variesfrom area to area, a method assuming homogeneous areas may produce spurious clusters. The method proposed by Whittemore et al. alleviates any possibility of clustersbeing detected because of nonuniformity of subgroups at different disease risk over thestudy region. Previous cluster detection tests, such as those proposed by Pinke andNefzger [7], Knox [4], and Mantel [5] were designed to detect cases clustered in spaceand time simultaneously, without taking the population into account. The Whittemoreet al. method addresses the problem of spatial clustering alone. Spatial clustering maybe more informative than space-time clustering when disease development varies greatlyfrom patient to patient.The procedure has some shortcomings. The location of significant clusters, if oneor more exist, is not indicated. The statistic only reflects the degree of clusteringin the overall study region. Since the test statistic is based on the pairwise distancesbetween cases, a small observed value of W could arise in two different situations. Wherepopulation density is high, such as in metropolitan areas, there may be an apparentcluster because the distance between people, and thus cases, is small. In rural areas,the same value of W could represent a true cluster since rural individuals generallylive farther apart than urban individuals do. The method cannot distinguish thesetwo types of close cases when only the distance between cases is considered and notthe population density. Clusters detected may arise either because of high populationdensity or a genuinely elevated disease incidence rate. If clusters occur in several cells,the test statistic has poor power. The power will also be poor when the cells with thehighest rates vary among the strata.82.1.2 The Geographical Analysis MachineThe Geographical Analysis Machine (GAM), proposed by Openshaw et al. [6], is agraphical attempt to identify disease clusters. Openshaw et al. originally carried outthe method on 1968—85 childhood leukemia data from northern England. The procedureanalyzes a large number of circles of fixed radius r to identify regions of particularlyhigh incidence rates. A square lattice is laid on a map of the study region with gridpoints evenly spaced at intervals r/5 apart. Each grid point is considered in turn. Forgrid point w, compute the number of cases in a circle of radius r centered at that gridpoint and call this number C. A circle is drawn at grid point w if the observed valuedof C is at least two and exceeds the 99.8 percentile of a Poisson distribution with mean= P,.c./n.. Here, F,. is the population contained in the circle centered at the wthgrid point. The procedure is repeated with different values of r. The procedure beginswith r equal to 1 km. After each iteration, r is incremented by 1 km and the procedureis terminated once r is greater than 25 km.The GAM method has some drawbacks. The procedure is computationally intensiveand does not offer a quantitative assessment of an observed pattern’s significance. Thepopulation at risk within these circles of varying radii is difficult to determine. Whenone circle is drawn, many neighbouring circles are drawn because of the large numberof circles considered and the amount of overlap. High correlation between overlappingcircles will produce some apparent clusters even under the null hypothesis. The originalversion of GAM was unable to handle stratification. Diseases, such as cancer, withdifferent forms afflicting different age groups and sex categories could not be analyzedproperly with this methodology.There seem to be multiple testing problems as well. Each individual significance testis correct, but an adjustment for multiple testing has not been offered. The problem ismanifested globally and locally. With tests on a sufficiently large region, chance alone9will yield some apparent clusters. The local problem is more severe. The use of differentradii and shifts of location are an attempt to identify an optimal circle for any set ofcases. The area contained in the optimal circle is considered the most likely cluster.However, the significance level calculations do not incorporate the varying radii andlocation shifts. Perhaps the low significance level (0.002) was meant to compensate forthese multiple testing problems.The procedure provides an excellent descriptive method for finding areas of highincidence rates that are free of geographical boundaries. Some of the above deficiencieshave been rectified. The GAM method is a good descriptive tool, however, any supposedareas of clustering may have happened by chance.2.1.3 Cluster Evaluation Permutation ProcedureWhere the GAM procedure analyzes circles of constant area, Turnbull et al. [9] considerregions of constant population, called “balls”. The distance between cells is again definedas the distance between their respective cell centroids. For each cell, a two-dimensionalball of adjacent cells is formed so that the population is equal to some fixed number ofpeople, R. Each cell’s population is examined in turn. For cell i, i = 1,. . . , I, if n isless than R, the cell with the closest centroid is examined. Suppose the nearest cell tocell i is cell j. If the combined population of cells i and j, n + nj, equals R, then thecells i and j form the ball for cell i. If the combined population exceeds R, the ball forcell i is formed with a portion of cell j. This portion is the population required from cellj to attain R people within the ball for cell i. If n + n3 <R, both cells become part ofthe ball for cell i and the next nearest neighbour to cell i is examined. To ensure eachball contains at least one cell, R is chosen to exceed each cell’s population. Altogether,there will be a collection of I balls.The number of cases within the ball for cell i is GR. When a proportion of a cell10is combined to a ball to attain the population size required, the same proportion of thecases in that cell will be added to the ball’s cases. For example, if the ball for cell i isformed by combining cell i and 1/8th of cell j, CiR is equal to c + c3. The CiR valuesare considered identically distributed random variables under the null hypothesis. Thesevariables are dependent though, due to the procedure’s combining scheme. Since theballs have equal populations at risk under the null hypothesis, the expected number ofcases per ball is directly proportional to the disease rate. Therefore, the ball with themost cases, and thus the highest disease rate, is of concern to health authorities. Themaximum rate statistic, MR = max(C1R,C2R,.. . , CJR)/R becomes a natural choicefor a test statistic. If MR is greater than some cutoff value determined by the nulldistribution of MR and the significance level, the null hypothesis is rejected. The cellwhose ball has the highest rate is noted as the most likely cluster. The second and thirdhighest rates may also be investigated in a analogous fashion. Null distributions forthese statistics can be found similarly.The randomization test ideas of Fisher [3], can be used to evaluate the null distribution of MR. The distribution is the group of MR values achieved by consideringall ways of assigning c. cases to n. individuals in the population. Each assignment isequally likely. The number of such permutations, n.!/(c.!(n.— c.)!), is typically too largeto compute the null distribution exactly. A Monte Carlo sample of the permutationscan be used to estimate the distribution under H0.The Turnbull et al. method does not have the multiplicity problems of the GAMmethod. This method also provides a quantitative assessment of the statistical significance of a supposed cluster, which the GAM procedure is unable to accomplish, and isless computationally intensive. Stratified analyses can also be accommodated with thisprocedure. Turnbull et al. have not addressed what effect the underlying populationhas on their test statistic. If population density within a cell is uniform, then taking11the proportion of cases corresponding to the proportion of population required may besuitable. However, if the population is not uniformly distributed within a cell, the amalgamation of fractions of other cell populations may not represent the true populationwithin the ball.Turnbull et al. also require that the cells have quite similar population sizes. If, forexample, one cell has much more population than the others, the population of the ball,R, has to be greater than the largest cell’s population. Many small cells will have to becombined to form a ball with population R and smaller clusters may be undetectable.2.1.4 Besag and Newell StatisticBesag and Newell [1] chose to examine regions with the same number of cases ratherthan areas of fixed population. The method requires a minimal amount of informationincluding the number of cases, the population at risk, and the centroid of each district.Their method is best used for an extensive area that is divided into relatively smalladministrative districts for which census information is known.Each cell is considered separately. All cases within a cell are considered to be locatedat the cell’s centroid. The distance between cells is defined as the distance between theirrespective centroids. For cell i, the remaining districts are ordered in increasing distancefrom cell i. An integer k is selected to represent cluster size. The number of cells, L,which must be added to cell i to include the nearest k cases is the test statistic. Toillustrate, cell i’s closest neighbours may be cells i1,i2• with cell i, being the p’farthest cell centroid from cell i. If c > k then the observed value of the test statistic,1, is equal to zero. If c < k but c + c1 k then £ is one and so on. Small observedvalues of L indicate that k cases are nearby and reflect the degree of clustering.Under the null hypothesis, the significance level of each suspected cluster can beevaluated. Let mL represent the total population in cell i and in its L nearest neigh12bours. Note that L > £ if and only if less than k persons among the me have the disease.For a rare disease, the hypergeometric probability that exactly x individuals amongpersons have the disease can be very closely approximated by a Poisson term under H0.This Poisson term is given byexp(—Ae)i\//x!, (5)where )e = mec./n.. Therefore, equation (6) below determines the significance levelfor cell i with respect to the k nearest cases,k—iF(L £) = 1— exp(—Aje)At/x!. (6)Cells which attain some nominal significance level, a = 0.05 say, are recorded. All cellswhich achieve significance at the 5% level can be plotted on a map of the region as auseful diagnostic tool. The cell with the lowest p-value shows the strongest evidence ofclustering for cluster size k. If there are S strata, the Poisson formulation is adaptedslightly. The method proceeds in the same manner as if stratification was not presentwith one exception: the Poisson distribution has a mean which incorporates the stratification. The total number of cases and individuals in stratum s are denoted by c.3 andn., respectively. The mean of the Poisson term in equation (5) would be replaced by=m8c.3/n.,where m8 is the number of cases in stratum s in cell i and its £closest neighbours.By chance, however, some cells are expected to be nominally significant at the 5%level. A rough approximation for the expected number of cells significant at lOOa%would be al. Let R be the total number of cells that are significant at the nominal levela. An overall test for R can be pursued. The true significance level for cell i is(7)where £* is the smallest value of £ such that P(L £) > a. The exact null expectation13of R is given byIE(R)=a. (8)If the observed values of R exceed E(R), a Monte Carlo simulation can be used to assessstatistical significance. Samples are generated by distributing the c. cases among then. individuals randomly. A particular case has a probability of being placed in cell iequal to ni/n., i = 1,. . . , I. If different strata are incorporated, each stratum is considered separately with cases allocated to cells with probabilities proportional to theapplicable numbers at risk. This simulation may provide some information about number of clusters detected, however, Besag and Newell stress that lack of overall statisticalsignificance should not preclude investigation of clusters that appear to exist.The Besag and Newell procedure also has multiple testing problems. Each cell isconsidered in turn with the obtained significance recorded. With the combination ofcells, each test is not independent. An overall test of clustering is offered as an attempt to assess the overall significance in the multiple testing environment when manycells appear significant. The Besag and Newell approach accommodates the underlyingpopulation distribution, and can easily incorporate different strata. Different population densities are assumed among the cells, however, the population within each cell isconsidered to be uniformly distributed. In sparsely populated regions, this uniformityassumption probably would not apply.This method looks specifically for clusters of size k. Besag and Newell chose 2, 4,6 and 8 as their values for k in an analysis of Northern England leukemia data. Thechoice of k is not an obvious one. Since the significance level depends k, the value of kto be employed (hereafter referred to as the “cluster size”) is an important issue. If thecluster size investigated is too small, larger clusters cannot be detected. With large k,a spurious cluster may be produced because the significance level is based on the sumof k terms. If k is large enough and the mean is appropriate, the sum of these k terms14will be close to 1 and the p-value will be close to zero.2.2 Focused TestsFocused tests are intended to determine if disease cases are clustered around suspectedsources of exposure. Such sources are called foci. The detection of a cluster near a focusneed not indicate that the source is the cause of the cluster. However, it would suggestthat further investigations be considered to establish any causality.For the validity of the focused test, several issues need to be carefully considered.When looking at foci, the sources should be chosen for reasons other than suspectedhigh incidence rates. A nonarbitrary choice of cell boundaries and risk categories isimportant. With a focused test, boundaries can be chosen deliberately to achieve themost extreme value of a test statistic. For example, investigators wishing to prove thata specific chemical causes a disease may alter their cell borders to give their desiredresult. Investigators must ensure knowledge of a potential source does not bias theirtest statistic. An additional problem may arise when results from a particular site arenot robust to small changes in the definition of the population at risk. The choice ofsubgroups and geographical borders are critical considerations for any inferences madeabout a possible cluster.The general test null hypothesis and notation apply to the focused tests as well.However, a different alternative is specified for these tests. The alternative hypothesisis that disease rates are elevated in the vicinity of sources of environmental exposure.That is, regions in the proximity of the source will have higher incidence rates.2.2.1 Besag and Newell Focused TestBesag and Newell [1] proposed a slight alteration of their general test to examine clustering about a focus. Instead of looking at each cell, only each focus is examined. The15location of a focus is considered a separate cell which does not contain population orcases. For focus f, the I cells are ordered in increasing distance from the focus. A clustersize k is chosen as it was in the general test. The test statistic, L is the number of cellswhich must be added to the focus to include the nearest k cases. The probability thatL is greater than £, as defined for the general test, determines the unadjusted p-value.The focused version of the Besag and Newell test has the same problems and advantages as its general parent. Multiple tests are still performed and the choice of clustersize to investigate is still an issue. This focused test does, however, allow an analysiswithout requiring exposure data. The test examines if a focus and some of its closeneighbours have statistically significant excess numbers of cases. Intuitively, if a sourceemits some contaminant which increases an individual’s risk for a disease, many casesshould be found in close proximity to that source. This intuition does not necessitatethe use of exposure data which is generally unavailable.2.2.2 Stone StatisticStone [8] developed a method to test clustering about a single focus. Cells are orderedaccording to increasing distance to the focus. The distance here is determined by thedistance between the cell’s centroid and the focus. The cell whose centroid is closest tothe focus is labeled cell 1, and the farthest cell is called cell I. The expected numberof cases in cell i under the null hypothesis is E: = nc./n.. The Stone test statistic isdefined asz’=1 cTtone (9)Intuitively, this statistic is the maximum value of the estimated relative risks amongsuccessively larger geographic areas about the source. These areas are composed of they cells closest over the values of y = 1,. . . , I. Let y be the value of y which maximizesTtone. The estimated relative risk is maximized in the closest cells to the focus.16This test statistic identifies both the cells which are most likely to be a cluster and thenumber of cases in the presumed cluster.Consideration of a two-dimensional random walk is required for the distribution ofTtone under H0. Let the plane be defined by the cumulative observed and cumulativeexpected numbers of cases and let A be the line through the origin with slope equal toTtone. The probability of a random walk beginning at the origin and traversing aboveline A is equal to the significance level for the test of H0. This significance level will beunadjusted for the multiplicity of tests.Stone ordered the cells by their distance to one source to examine clustering aboutthat focus. The cells could be ordered by their distance from any focus. Suppose thereare two foci in a study region. The cell which is closest to either of the sources than anyof the remaining cells, would be labeled cell 1. The process would continue as it did forthe investigation of a single focus. Suppose there were two foci and the distance betweencell i and focus f is denoted d2f, f = 1,2. The cell that satisfied min(di, d22) would belabeled cell 1. This adaptation can be used to see if clustering seems to occur around thefoci as a group. Some problems can arise from this change. The foci are treated as if theassociation between exposure and geographic distance is the same for each focus. Thefoci may interact to increase exposure and may release different levels of contamination.Detailed exposure data would be required to accommodate any heterogeneity betweenthe foci. Because the test statistic is based on the maximum value of a ratio of observedto expected cases, it may not effective for detecting mild clustering.2.2.3 Wailer et al. StatisticIf quantitative exposure data are available, or an exposure model is proposed, thesevalues can be used in the definition of a focused test. Otherwise, a surrogate for measuredexposure may be used. Waller et al. [10], [11] consider a model where exposure and17geographic distance are inversely related. All individuals within cell i are assumed tohave the same exposure to the foci, labeled g2, i = 1,. .. , I, which is taken to be thereciprocal of the distance from the focus to the centroid of cell i, i = 1,. . . , I. Thisconvention is reasonable since a cell which is 200 km from a focus may have 1/2 theexposure of a cell which is 100 km from the source.Let E1 be the expected number of cases in cell i under the null hypothesis defined earlier, where E1 = nc./n., i = 1, . . . , I. Wailer et al. consider a multiplicative alternativewhich isHi:E(C1)=riO(l+ge), i=l,...,I.and represents an increase in relative risk for individuals living in cells near the foci.Here, e is unspecified but positive and equals zero under the null hypothesis. With thisalternative, the expected number of cases in cell i is proportional to the exposure of theindividuals within that cell, i = 1,. . . , I. Under H1, the disease counts C1, i = 1,. . . , I,are still independent.Wailer et al. developed a focused test of the null hypothesis which would maximizelocal power against this alternative. Following the development of locally most powerful(LMP) tests described in Cox and Hinkley [14], the test statistic becomesIU =g1(C — E1). (10)Therefore, U is the sum of the deviations of the observed incidence for each cell fromits expectation under H0, weighted by the exposure. Under the null hypothesis, e = 0and 0 is the only parameter. The likelihood may be expressed as a single-parameterexponential family. This test will be uniformly most powerful (UMP) because a UMPtest exists for the one parameter exponential family, and the fact that UMP tests mustmaximize local power when the family of alternatives includes local values. Thus, Uwill be a UMP test as well.18Because the test is LMP, under the null hypothesis U has mean zero and varianceequal to the Fisher information, which is approximatelyIvar(U) =With standardization, U* = U/(var(U))”2and U has an asymptotic standard normaldistribution.The test proposed by Wailer et al. is based on a more specific alternative than Stone’stest. Whether or not exposure and distance are inversely related may be questionable.An individual residing downstream from a river polluting source may suffer more exposure than an individual living nearby but upstream from that source. In addition, ifcells are large the assumption that each individual within a cell has the same exposuremay not be justified.The methods described above are all considerate of the underlying population distribution. The discussion outlined some of advantages and disadvantages of each statisticalmethod. The Whittemore et al. statistic is unable to distinguish close cases due to clustering or population density. British Columbia has some areas of high population densityand it is as important to find clusters there as well as in rural, sparsely populated areas.Openshaw’s GAM cannot provide quantitative assessment of clustering. The methodsproposed by either Turnbull et al. or Besag and Newell would be suitable for our situation. The Besag and Newell procedure is preferred, because it could also be used fora focused test. Hence, the focused and general tests devised by Besag and Newell areused for our analysis.193 Data DescriptionThe Besag and Newell method and its modification detailed in Section 4.2 require cancer cases by administrative regions as well as the corresponding populations at risk.Section 3.1 and Section 3.2 specify the sources of data for cancer patients and population, respectively. The centroid determination for each school district is outlined inSection 3.3.3.1 Cancer DataThe provincial Department of Health recognized cancer as a disease in 1932. Privatephysicians began informing the B.C. Cancer Registry of cancer cases in 1935. By 1966,the register contained all known live cancer patients. Information was collected fromprivate physician submissions, larger hospital medical records and from data collectedby then the B.C. Cancer Institute.The Registry has data on every B.C. resident who has been diagnosed with a malignant neoplasm since 1970. Malignant neoplasm recorded includes in-situ and borderlinetumours. Many details are collected from each patient including the patient’s sex, age,year of diagnosis, address of residence at the time of diagnosis, including postal code,along with its school district. The type of cancer, or site, diagnosed is also recorded.The B.C. Cancer Registry is the primary source of cancer information in this study.During the years 1983 to 1989, there is a total of 80, 131 cases of cancer reported tothe Registry. Of these, 78, 378 (41,282 male and 37, 096 female) have complete schooldistrict information representing 98%. There are over thirty cancer sites of interestin this study. All skin cancers, except for melanoma, are excluded from the study.The term “all cancers” represents all non-skin cancers plus melanoma. A list of sitesinvestigated along with the corresponding number of patients afflicted is listed in Table 1of the Appendix. It is easily seen that not only does each site have a different incidence20rate but the cancers affect the sexes differently. The different eitologies require analysesto be conducted on each sex separately.3.2 Population DataThe school district is the smallest administrative region for which population figuresare available in the Division of Epidemiology, Biometry and Occupational Oncology atthe BCCA. The school district is our choice for a cell. British Columbia had 74 schooldistricts during the years 1983—89, which can be seen in Figure 1. The school districtshave various geographical sizes. The northern part of the province has much largerschool districts than the lower mainland area.The yearly estimated populations by school district were provided to the BCCA byMr. Dave O’Neil, Population Section, Ministry of Finance and Corporate Affairs. Thesepopulations were estimated based on the 1981 and 1986 census data. The sex and agegroup of an individual is recorded in the population data. The age groups are formedby combining ages into five year intervals. Thus, ages 0—4 define age group 1, 5—9 formage group 2, 10—14 form age group 3 and so on with 18 as the maximum age group.The population data gives the number of individuals in each sex and age group in eachschool district for every year of interest.The population size also varies between school districts. The lower mainland regionhas very high population density and the northern areas are sparsely populated. Thedisparity can be seen when Stikine is compared with Vancouver. In 1989, the total population of males residing in Stikine and Vancouver are 1, 319 and 222,037, respectively.The bar chart found in Figure 2 shows the total number of people in each school districtfor 1989. The bars are ordered by the school district number and we can see the vastdifferences in population between cells.21Figure 1: British Columbia School Districts 1983—891 Fernie2 Cranbrook3 Kimberley4 Windermere7 Nelson9 Castlegar10 Arrow Lakes11 Trail12 Grand Forks13 Kettle Valley14S Okanagan15 Penticton16 Keremeos17 Princeton18 Golden19 Reveistoke21 ArmstrongSpallumcheen22 Vernon23 Central Okanagan24 Kamloops26 N Thompson27 Cariboo-Chilcotin28 Quesnel29 Lillooet30 S Cariboo31 Merritt32 Hope33 Chilliwack34 Abbotsford35 Langley36 Surrey37 Delta38 Richmond39 Vancouver40 New Westminster41 Burnaby42 Maple Ridge43 Coquitlam44 N Vancouver45 W Vancouver46 Sunshine Coast47 Powell River48 Howe Sound49 Central Coast50 Queen Charlotte52 Prince Rupert54 Srmthers55 Burns Lake56 Nechako57 Prince George59 Peace River S60 Peace River N61 Greater Victoria62 Sooke63 Saamch64 Gulf Islands65 Cowichan66 Lake Cowichan68 Nanaimo69 Qualicuin70 Alberni71 Courteny72CampbeU River75 Mission76 Agassiz- Harrison77 Summerland80 Kitimat81 Fort Nelson84 Van Isl West85 Van Is! North86 Creston-Kaslo87 Stikine88 TerraceMap courtesy John Smfth & Maplrifo.22Figure 2: Population Distribution by School District 19898888-0.08-88-0-3.3 Centroid DeterminationTwo types of centroids can be considered as the centroid of a school district. A geographiccentroid is defined as the geographic center of a region, whereas a population centroidis the location which represents the center of the population. In small cells with highpopulation density, the geographic and population ceiltroids coincide. In our study, mostcells are quite large and not very densely populated. For large areas, the population andgeographical centroid can very different. Since the goal is to estimate the location ofcases within a cell, the geographical centroid is inappropriate. Where population doesnot exist, no cases will be found. If the geographic centroid of a cell is in an unpopulatedregion, the centroid is not at all an estimate of the location of the cases. Cases can onlybe located in areas of population. We expect more cases in a highly populated area thanISchool District23in a sparse one, so it makes sense to choose the population-based centroid as an estimateof the location of the cases within a cell. Only population centroids will be used in thisstudy. The exact locations of all people in the province is unknown. Under the nullhypothesis, the location of all cancer cases from all sites estimates the population-basedcentroid. This determination is a reasonable estimate of the real population centroidsince large numbers of cases are involved. Of the total cases reported to the Registryin all sites considered, about 91%, or 73,092, patients have both postal code and schooldistrict information complete. These patient’s postal codes are used to estimate theschool district centroids.The data is transformed by Geocodes/PCCF software [12] into latitude and longitudecoordinates. The Geocodes Postal Code Conversion File (PCCF) system is producedby the Geography Division of Statistics Canada. The current version contains over118,000 postal code records for British Columbia up to the year ending 1991. Thepackage enables its user to gain information about standard geographical areas by useof the six character postal code. Users can cross-reference geographic coordinates andcensus regions.The content of the PCCF has three key factors. These components are the postalcode, the 1991 Enumeration Area (EA) code, and the postal code coordinates. Thepostal code coordinate is given in Universal Transverse Mercator System and latitude/longitude coordinates. The coordinates are determined by representative pointsthat are “proxies” for postal code locations. They may be EA representative pointsor block-face representative points. An enumeration area representative point is basedon either automated or manual judgement of the visual center of an EA. This pointis usually located at or beside predominant building or street clusters. A block-face isgenerally one side of a city street between consecutive intersections with other streets inurban areas. The block-face representative point is the midpoint of the block-face, set24back a perpendicular distance of 22 metres from the street center line. The enumerationarea code contains many levels of census divisions such as Census Subdivision (CSD)and Federal Electoral District (FED).The program first links any postal codes which are unique on the PCCF. Uniquein this case means that the postal code is linked to a single FEDEA or block-face. Ifa postal code is not unique on the PCCF, other levels of geography such as CensusDivision/Census Subdivision can be used to clarify any ambiguity.With nonunique postal codes on the PCCF, observations with the same postal codewill be distributed across all applicable enumeration areas (or block-faces) which areidentified by that postal code. Whenever a postal code can be linked with more thanone block-face or FEDEA, observations are distributed uniformly, within a postal community’s service region, to each applicable FEDEA. Suppose the PCCF indicates aparticular rural postal code serves ten different FEDEA’s, with 7 in one CSD and therest in another. The FEDEA’s would be randomly mixed and each successive occurrenceof the postal code would be assigned to the next applicable FEDEA. The distribution ofevents would be 7/10 in the first CSD and 3/10 in the second CSD. This method givesan approximately population-weighted random distribution since FEDEA’s have aboutthe same population throughout the postal code’s service area.The first three characters in the postal code represent a set of well-defined and stableareas known as Forward Sortation Areas (FSA’s). The last three postal code charactersdesignate areas known as the Local Delivery Unit (LDU). A single postal code cancorrespond to several types of LDU’s in urban areas. One block-face, a CommunityMail Box, an apartment building, a business building, a large firm or organization, afederal government department, a mail delivery route, general delivery at a specific postoffice, one or more post office boxes are all examples of LDU’s. In rural areas, a postalcode refers to all services which originate from a post office or postal station.25When a postal code was not exactly matched on the PCCF an error, termed Error 0,is reported. When a postal code is not exactly matched to one contained in the programpartial geography can be assigned based on the consecutive characters which do matchon the PCCF. For example, if the LDU does not match, the geography may be assignedby the first five characters of the postal code which do match on the PC CF. If the postalcode was linked to a Post Office Geography (Error 1) rather than a place of residenceor was not linked to a residential address at all (Error 2) the user is also informed. Theprogram also warns if a postal code was linked to a business building, commercial orinstitutional building, or a retired postal code.The school districts of Langley, Surrey, Delta, Richmond, Vancouver, New Westminster, Burnaby, Coquitlam, North Vancouver and West Vancouver have particularly highError 0 rates. These regions contain 35, 128 cases of which 32, 163 are not matched onthe PCCF. The highly populated regions seem to have the most unmatched postal codes.Overall, about 59% of codes do not match exactly, 92% due to the lower mainland schooldistricts just mentioned. For these cases, some of the first characters of the postal codeare used to assign a latitude/longitude coordinate. These rates may sound alarming,however, the coordinates assigned are representative of the school district in question.More FSA’s exist in highly populated areas. If a postal code is not exactly matched,the FSA can be used to assign geographic coordinates. With the LDU unmatched,the geographic coordinates assigned by the FSA may differ from the real location bya block-face or apartment building or Community Mail Box. The coordinates will berepresentative of the area, even though they may not be as accurate as if the postalcode was exactly matched. One can easily tell if the geographic coordinates given byPCCF actually are plausible. Comparing the coordinates to the school districts on amap indicates any which are not contained within the school district’s boundary. Theselower mainland districts occupy a small geographic area with respect to the rest of the26province. High rates of nonmatching postal codes are not disconcerting since the exactlocation of each cancer case is not the objective. In the lower mainland, the populationdensity does not vary greatly within a school district. Even with inexact geographic coordinates, the location is likely still representative of the population which resides there.We also recognize that with rural postal codes, the latitude/longitude may correspondto a postal outlet and not to an actual place of residence. Postal outlets are located inareas of population, so the coordinates are accurate estimates of the location of a case.With the latitude and longitude coordinates that Geocodes/PCCF produced, thepopulation centroids can be approximated. The set of coordinates which correspondedto a specific school district are used to estimate the population centroid. The coordinatesof patients in Vancouver school district estimate the population centroid of Vancouver,for example. For each school district, the median latitude and median longitude aretaken to represent the population centroid. With this determination, the choice ofcentroid is based on the underlying population distribution. If a school district is square-shaped and all the population lives in the northwest corner, placing the centroid at thegeographic center would not be an accurate centroid. If the disease incidence is randomand independent of location, highly concentrated areas of cases should also have a highunderlying population at risk. Under the null hypothesis, distribution of cases shouldbe representative of the distribution of the population at risk. Of course, the populationcentroid chosen by this method will only approximate the actual cell population centroid.Figure 3 displays the estimated population centroid of each school district.British Columbia has twenty-two pulp and paper manufacturers within its borders.These mills are considered as foci for the focused tests. Table 2 found in the Appendix,lists the name, location and corresponding school district of each pulpmill. Postal codesare known for each mill and the mill coordinates are also given by the PCCF. Of the 22pulpmills, five postal codes are unmatched. The coordinates given are good estimates of27Figure 3: School District Centroids in British Columbia 1983—89&LJ(ZL)4SEThNcNMap courtesy John Smith & Mapinfo.28the actual locations. Mills are generally located near or within a town. The mills are notfound in deserted areas, but places where resources and the work force are nearby. Thecoordinates can be double checked by using a map. The Prince George mills coordinatesgiven by the PCCF can be cross-referenced with the town’s coordinates. The town ofPrince George can be seen on a map and the latitude and longitude can be recorded. ThePrince George mill coordinates should be fairly close to the town of Prince George to bereliable. For all mills considered, the coordinates determined by PCCF are very closeto the towns the mills actually are near. From this, we conclude that the coordinatesgiven by the PCCF program are excellent estimates for the geographic locations basedon their full addresses. The mill locations are plotted on the map given in Figure 4. Afew school districts have multiple mills located within their borders. The Prince Georgeschool district, in particular, has five mills within its boundary.29Figure 4: Pulpmill Locations in British Columbia 1983—8919,2 1‘rft_1_5 4 V ,¾4S 1,2,1016,22N•14çz •Map courtesy John Smith & Mapln(o.1 Prince George 9 Powell River 17 Skookumchuck2 Prince George 10 Prince George 18 Ocean Falls3 Castlegar 11 Prince Rupert 19 Mackenzie4 Kitimat 12 Port Alice 20 Campbell River5 Crofton 13 Squamish 21 Mackenzie6 Port Mellon 14 Kamloops 22 Quesnel7 Port Alberni 15 Gold River8 Nanaimo 16 Quesnel304 AnalysisThe application of the Besag and Newell method is now discussed. The data is firsttested under the general test. Problems encountered in the application of the Besag andNewell general test also arise in the focused test. Because the two tests are analogous,discussion is concentrated on the general test. A combined year analysis is discussed forthe general and focused tests followed by a yearly examination. A modification for theBesag and Newell method is motivated by the combined year analysis and is introducedin Section 4.2. The differences between the modification and the original method arealso examined.The nearest neighbours of a cell or foci are required for the method. The distancebetween cells is taken to be the distance between their centroids. In Table 3 the threenearest neighbours to each school district are listed.For the focused analyses, the location of a focus is considered its own cell. Within thiscell the number of cases and population are both equal to zero. Each focus is analyzed inturn by looking at the closest school districts to it. Table 4 gives the five nearest schooldistrict centroids for each mill. The Prince George school district has five pulpmillswithin its boundary. Two of these are in MacKenzie and three are in Prince George. Thelatitude and longitude coordinates, as calculated by the GEOCODE/PCCF software,are identical for the three mills located in Prince George. This means that the nearestneighbouring school districts are the same for all three mills. Thus, one analysis basedon only one focus of the three is required. From Table 4, the two mills in MacKenziehave quite similar nearest neighbours, with identical first two nearest neighbours, tothe mills in Prince George. The Skookumchuck mill is located in the Cranbrook schooldistrict, but it is closest to the centroid of the Kimberley school district. All other millsare closest to the centroid of the school district in which they are located.The cancer sites investigated have varying incidence rates in the different age groups31and sexes. Both general and focused tests are conducted on each site and sex separately.Once one analysis is conducted for any strata involved, it is repeated for every site andfor each sex. For example, male lip cases are tested then female lip cases are tested. Allanalyses use age group as a stratum.4.1 Combined Year AnalysisThe general test is applied first for the combined year analysis and the details belowdescribe the procedure.The choice of cluster size, k, is a serious consideration. Besag and Newell use thevalues 2, 4, 6 and 8 for k in their analysis of the incidence of childhood leukemia innorthern England. In their case, the counties, or cells, are relatively small and leukemiaincidences in each cell are probably either zero or one. A cluster size larger than 8 isprobably not very realistic in their situation. In our case, the different regions have verydissimilar numbers of incident cases. The cases may be lower than 10 or higher than100 in each administrative region. The variation among regions increases when morecommon cancers are examined. The lower mainland school districts may reflect incidentcases in the low thousands while a northern district has less than 100 cases. With thisvariability, the cluster size we should test is not clear.To follow their approach, several different cluster sizes are chosen to account for differences between administrative regions’ population sizes. The selection of each clustersize depends on the population of the school districts. Smaller school districts need tobe tested at smaller cluster sizes while large school districts need a large cluster size.Several different cluster sizes may be chosen to reflect the school district diversity. Thisdetermination hopefully ensures that the choice of k does not favour any particular population size. The detection of clusters in a densely populated area is just as importantas detecting a cluster in a sparsely populated region. Choosing multiple cluster sizes32may alleviate any bias towards a specific population size. To examine its feasibility,ten different cluster sizes, k1,k2,.. . , k10 are used for each sex and site. Let Cj represent the number of cases in school district i, year y, and age group s, i = 1,. . . , 74,y = 83, . . . ,89, s = 1,. . . , 18. The total number of cases, ci.., is calculated in each schooldistrict i over the years of 1983—89, where c.. 83 ZI c. The median value ofthe Cj.. ‘s is evaluated and labeled k1. Twice the maximum value of the c:.. ‘s is namedk10. The remaining eight choices for the cluster size are evenly spaced between k1 andk10 with any fractional parts dropped. Thus kr = k1 + Lx(kio —k1)/8] for x = 2, ... , 9with Li] the greatest integer less than or equal to t.The analysis proceeds for each cluster size. Let nj denote the population in schooldistrict i, year y, and age group s, i = 1,. . . , 74, y = 83,.. . , 89, s = 1,. .. , 18. Theaverage provincial population in age group s during the seven study years is labeled..s, ..3 = n28. The estimated mean number of cases expected over thecombined years for each school district is determined using age categories. Letrepresent the total population in cell i and its £ nearest neighbours for year y and agegroup s. The average population over the seven study years for cell i and its £ nearestneighbours is denoted The estimated mean number of cases expected during theyears 1983—89 for school district i, i = 1, ... , 74, is calculated as=(11)The result for male lip cancer, which is somewhat typical for this approach, is givenin Table 5. Results are not consistent from cluster size to cluster size. Most districts arepossible clusters for one or two values of k,, but fail to show any consistency throughoutthe cluster sizes considered. The observed value of the test statistic also fluctuatesquite noticeably. Many school districts are detected as possible clusters because of themultiple cluster sizes investigated. For rare cancers, such as lung adenosquamous cancer,the resulting cluster sizes detected are less than twenty. For more common cancers, like33cob-rectal, the cluster sizes range from 32 to 1, 643. A problem could arise if thecluster size is too large. The significance level of the Besag and Newell test is dependenton the cumulative Poisson distribution. If enough terms are summed the cumulativedistribution will be very close to one. Since k is the number of terms summed, withsufficiently large cluster sizes the Poisson cumulative distribution can be close to one.If this case occurs, by equation (6) the significance level will be very close to zero and acluster will be identified. When a cell attains a low significance level, an interpretationproblem emerges. Investigators must decide if this significant cell is a result of a realcluster or a consequence of the cluster size being too large.Just as Besag and Newell use four values in their analysis, we try to systematicallydetermine the cluster size based on the number of cases in different school districts.This situation makes the multiple testing problem even worse. The regular multiplicityquandaries are compounded by the assorted cluster sizes. How these results are interpreted is also a problem. Of course, most school districts for which a possible clusteris detected for one value of k, are no longer identified as possible clusters at higher orlower values of this cluster size. Since the significance level achieved is influenced by thechoice of k, the interpretation of any results must be based on the choice of k.These difficulties lead us to conclude that detecting cancer clusters with a fixedsize across British Columbia, where school districts have substantially different populations, is neither reasonable nor worthwhile. Instead, we try to detect cancer clusterswith different sizes for distinct areas in B.C. The sizes are chosen so that a commoninterpretation can be obtained for the analyses.As commonly seen in epidemiology literature, a relative risk of 1.5 (with at least 5cases) due to some specific exposure is considered to be sufficient evidence for furtherexamination. We incorporate this idea in choosing cluster sizes. That is, for eachschool district, a cluster size is chosen so that it is about 1.5 times the expected number34of cases if no clustering is present. For example, if cell 1 has 1000 residents and theprovincial incidence rate is 0.01, 15 would be the cluster size. When the age groups areincorporated, the cluster size investigated for cell i is k1.5, defined by,= [1.5Ethj.3Z!j, i = 1,... ,74. (12)With this cluster size valuation, the analysis proceeds. If the cluster size is less than five,5 is taken as the cluster size. The estimated mean number of cases expected for cell i isstill given in equation (11). Note that with this choice of the cluster size, a conclusionof cancer clusters in the region is interpreted as evidence of local areas having excessrelative risk.Equation (12) shows that the cluster size evaluated for each school district reflectsthe inherent variation in the population sizes. Large values of L are no longer requiredto achieve k cases when a cell with small population is considered, which helps ininterpretation. Certainly a cell which is a possible cluster, has a relative rate of 1.5, andat least five cases causes alarm. Some clusters are detected using this determinationof the cluster size, however, the method proves to be unable to detect larger clustersin small regions where expected numbers of cases are low. If, for example, cell i has 2cases expected ,o = 2), k1.5, equals 5. If that cell has 7 cases observed, £ = 0 and thep-value is insignificant. The cell is intuitively alarming because of the very high relativerate. With over 5 cases and a relative rate of 3.5, the district causes concern and thecluster detection method should support this conclusion. Besag and Newell also thinkthe relative rate is important but did not incorporate it in their method.In fact, the procedure performs relatively well when the expected number of caseswithin a region is greater than 15. With the mean less than 5, however, the mean needsto be multiplied by a much higher relative risk to detect unpredictably large clusters ina small area. Figure 5 shows what relative risk is required to detect a cluster at 5% foreach value of the mean, smoothed for display.35Figure 5: Minimum Relative Risks for Significance at 5%LOC’)qc)LqC%JqC”The y-axis in this figure is the 95 percentile of the Poisson distribution with mean.A divided by )... Any observed relative risks which lie above the curve yield a cell withthe corresponding expected number of cases to be significant at 5%. For .A > 15, thisprocedure is able to perform well because 1.5 lies above the curve in that region. Withsmall values of A, 1.5 lies well below the curve and no cells are significant for those valuesof ).. We also notice that lower than 1.5 relative rates will return a significant resultwhen the expected number of cases is large. From this figure it is easy to see that withsmaller means, the relative risk required to detect a possible cluster increases. A fixedrelative risk across the region may be reasonable for common cancers where expectednumber of cases are relatively large, but it may not be desirable for rare cancers wherethe expected numbers may be small and a potential cluster may be missed. This led us0 5 10 15 20 25 30Expected Cases36to try an alternate function of the expected number of cases within each cell.4.2 Modified Besag and Newell MethodThe previous discussion displayed some of the problems we encounter with the application of the Besag and Newell statistic. Specifically we have problems with the appropriate choice of cluster size. One cluster size cannot detect all possible clusters within theprovince. A different cluster size for each school district based on the population withinthat district seems desirable. We now propose a method to determine which clustersizes to test for each school district.The significance level calculations with the Besag and Newell approach are based onthe cluster size investigated. The p-value may not represent the true degree of clusteringif the observed number of cases is much larger than the cluster size being tested. If thecluster size is 10, for example, the same p-value will be achieved for cell i if it has 10 or50 cases within its boundary. The test statistic is zero in both cases and the p-valuesare identical. Surely the latter situation represents more clustering than the former.The Besag and Newell method would benefit from a systematic way of determining thecluster size so that small changes in the choice of cluster size do not alter the significancelevel dramatically. Certainly, it is undesirable to have a particular cell insignificant for kbut significant for k + 1, when in both situations £ = 0. Testing for a specific cluster sizemay not be of interest. Instead, areas which have statistically significant excess casesmay be of concern, whatever that excess size may be.These observations motivate some alterations based on the idea of allowing the number of cases representing a cluster size to vary as cell i is combined with its nearestneighbours. If the value of the cluster size is not important to the investigator, thecluster size shall be chosen to represent the minimum size for which the area examinedis a statistically significant cluster. This size corresponds to a specific percentile of the37Poisson distribution with mean determined by the area under scrutiny.Consider cell i, where 1 < I. Recall that the estimated total number of casesexpected in cell i and its w nearest neighbours is denoted by =0 w I — 1. Also recall that m2 is the total population in cell i and its w nearestneighbours. Define ki._a,j,w to be the 100(1 — a) percentile of the Poisson distributionwith mean Thus, ki_a,t,w is the smallest integer such thatki 11 — <. (13)A cluster size equal to is interpreted as the minimum number of cases whichwould have to be observed to cause cell i and its nearest w neighbours to be significantat level a.Each cell has some expected number of cases no matter how minute, so for any cell i,the sequence of Aj’s is strictly increasing, ),o < > •.. < )jj. It follows that forany i, the sequence of ki_a,j,w’s is non-decreasing, ki_a,j,...k1_,,,j_;i.e. the mean of the Poisson distribution increases so the 100(1— a) percentile is non-decreasing. However, the Poisson distribution is discrete, so very small increments inthe mean may not change the 100(1— a) percentile.Cell i may be tested at cluster size ki_a,j,. If the actual number of cases in cell iexceeds this cluster size, cell i is significant at lOOa%. The determination of the clustersize guarantees this result. With a cluster size ofk1,0 only cells which are possibleclusters on their own are guaranteed be significant. This cluster size may only identifysignificant cells for which the test statistic is zero, £ = 0. If, on the other hand, theactual number of cases in cell i is less than ki_a,j,ü, the test statistic increases by oneand the expected number of cases in cell i and its nearest neighbour equals . The100(1 — a) percentile is re-evaluated and these two cells are probably insignificant evenif a true cluster exists.In general, testing cell i at cluster size ki_a,j, determines if cell i alone is a possible38cluster. Testing the same cell at ki_a,j,i indicates if cell i and its nearest neighbourare significant when combined together. Two cells can be significant together withouteither being a possible cluster individually. This circumstance requires that not one,but several cluster sizes be tested.The following testing algorithm is proposed. Consider the sequence of cluster sizeski_a,j,o, ki_cr,i,i,. . . , ki_a,j,w for some integer w {O, 1,. . . , I — 1} for cell i. Test cell iat cluster size ki_a,j,0. If it is significant, no further testing is done on cell i. If cell i isnot significant, it is combined with its nearest neighbour and tested at the new clustersize ki_a,j,i. Again, if cell i is significant for this cluster size the algorithm terminates.Otherwise, cell i is tested at the next cluster size in the sequence. In general, test eachcell until either it is significant or it has been tested at every cluster size in the sequence.For cell i, the results from the last cluster size tested are recorded. The p-value fromthis last test will be referred to as the p-value for the cell.Generally speaking, w should not be very large. Large values of w correspond totesting if huge portions of the study area are significant. These tests probably do notyield any new information on clustering: either smaller groups of the huge portion aresignificant at smaller cluster sizes or the combination of many cells yields normal diseaserates. If cells with low disease counts are combined with cells of high disease counts, theresulting combination probably has about normal disease counts and is not a cluster.The multiple testing problem becomes more serious as w becomes large. The upperbound for the expected number of cells which may be significant due to chance is equalto oI for each cluster size investigated. If the number of cluster sizes investigated islarge, many spurious clusters may be identified. Consequently, keeping w small aids ininterpretation of results.The interpretation of the p-value for a cell is slightly different with the values ofk chosen. The school district with the lowest p-value is the most likely cluster of the39particular size investigated, if all cells are tested at the same cluster size. With themodified method, most of the p-values are very close to 0.05. Hence, the individual pvalue is no longer the best measure of the most plausible cluster of size k. The relativerate is a way of assessing which of the possible clusters are actual clusters. Those cellswith the highest relative rates should be of most concern to health authorities.The overall test of clustering introduced by Besag and Newell is important to theinterpretation of the results produced by our modified method. When several clustersizes for each cell are tested, many cells may be significant by chance. If only one clustersize is tested for each cell, cI cells are expected to be significant. If each cell is testedat w or less cluster sizes, the expected number of cells found significant by chance isat most wcd. If the investigator tests each cell at many cluster sizes, the Monte Carlosimulation can help determine if any apparent clustering can be attributed to chance.For our school district application we consider three cluster sizes (w = 2). Most ofthe school districts are geographically large. Choosing to combine 4, 5 or more of thecells may result in testing a large fraction of the province. Testing more cluster sizesallows more significant results by chance. Spurious clusters are unappealing. With theseconsiderations, testing at most three cluster sizes seems reasonable in our situation.The cluster sizes k.95,10k.95,1 and k.95,12 for cell i are the 95’ percentiles of thePoisson distributions with means )qo, )i1, and )z2, respectively. These means correspondto the expected number of cases in cell i alone, cell i and its nearest neighbour, andcell i and its two closest neighbours. Specifically, Xje=thj.jc..8/n.3and fii1. is theaverage population over the seven study years in age group s for the £ nearest neighboursof cell i. If any of the three cluster sizes are less than 5 we set them equal to five. Eachcell is tested at k.95,10then at k.95,1 and at k95,12if necessary. The algorithm proceedsfor each cell and the significant cells with relative rates at least 1.5 are recorded. Theindividual p-values will all be near 0.05 because of the way each k.95,1 is determined.40Those cells with the highest rates should be of most concern to health authorities.A Monte Carlo simulation is performed to help determine the overall significance ofthe number of clusters detected. The total number of cases c... are distributed randomlyto each of the strata based on the appropriate probability. The probability that a casewould fall in cell i is the proportion of the provincial population which that cell contains.The above method of analysis is then applied to the simulated data. The simulation isiterated 1000 times and the number of significant cells, R, for each iteration is recorded.The overall p-value is then the proportion of the simulations which attain at least theobserved value of R. Sites for which many school districts appear significant may bemore alarming than sites which have few possible clusters. With three values of k usedon each cell at most 3 x 0.05 x 74 = 11.1 cells, not necessarily distinct, are expected to besignificant by chance. This expectation is much lower when considering only significantcells with relative rates of at least 1.5 and testing each cell at a maximum of threedifferent cluster sizes. The Monte Carlo simulation may help to evaluate the significanceof overall clustering when more than the expected number of clusters appear significant.Note that the lack of statistical overall significance should not deter investigation ofindividual clusters which appear to exist. Overall tests are of secondary consequence andonly help assess the situation when many possible clusters are detected. In our analysis,only a few sites seem to have many possible clusters. The Monte Carlo simulation,nevertheless, is done on all sites.The focused test for combined years is conducted analogously. For each mill i,i = i,. . ., 18, three cluster sizes are determined. The mill itself has exactly zero casesexpected, so the cluster sizes are evaluated based on the nearest neighbours. The 95thpercentiles k95,10k.95,21 and k95,,2 now correspond to slightly different Poisson distributions. The first cluster is the 95th percentile based on the number of cases expectedin the nearest neighbour to mill i. The other two percentiles are based on the number of41cases expected in the two and three closest neighbours to mill i, respectively. A MonteCarlo simulation is also performed in the same manner described for the general test toobtain an overall p-value.Our modification yields some improvements to the Besag and Newell method. Mostimportantly, a systematic determination of the cluster size investigated is offered. Thiscluster size is guaranteed to detect possible clusters at a certain significance level basedon the size’s determination. The cluster size is determined by a percentile of a Poissondistribution. If the observed number of cases is as large as or greater than the percentile,the cell is significant and is detected as a possible cluster. We have the added advantagethat the value of k is chosen according to the underlying population or equivalently, itsexpected number of cases. The choice of cluster size is less ad hoc with this modification.A Monte Carlo simulation can still be applied to provide some overall significance tothe number of apparent clusters which are detected in the study region.A discussion of the general and focused results obtained follows in Section 5. Thenext section provides the yearly analysis procedure.4.3 Yearly AnalysisThe combined year analysis identified school districts which have enough cases over thestudy period to be considered possible clusters. Investigators may wish to know if theexcess of cases occurs in a specific year or is consistent across the years. To see if excessesof cases are consistent through time, we look at each year separately to see if any patternemerges. A school district may have a significant excess number of cases in particularyears within the study period. Although a cell is not significant for the combined yearanalysis, it still may be significant for a particular year. We specifically want to seeif the relative rates are consistent through time. If a school district is significant forseveral years we may be more concerned with this area than a cell which is significant42for an isolated year.To try to compare results from seven years of data, it seems appropriate, for aparticular site and sex, to choose one cluster size for each school district and test eachyear at that cluster size. With the cluster size fixed across all years, the statistic Lwould show different values for different years and the p-value calculated would indicatethe degree of clustering. We would be able to easily compare years if the same clustersize is used for each year. This would be feasible because the population in each schooldistrict does not fluctuate too much from year to year, and hence the expected numberof cases are about the same for each year. We use the combined year cluster size chosenfor a cell and divide by seven to get the yearly cluster size. The minimum yearly clustersize investigated is two since a cluster of size one is unreasonable.The same sort of problem arises when trying to look at one cluster size across all yearsas we had for one cluster size across all school districts. An area may be not significantfor any year because the cluster size chosen is too low. Consider the situation where acell has disease counts equal to zero for all years but one, 1988 say. In 1988, suppose thiscell has 13 cases observed. If the expected number of cases over all years is seven, thiscell is significant at 5% when k equals thirteen. When the yearly analysis is conducted,we use a cluster size equal to 2 since 13/7 is less than two. For each year, the expectednumber of cases are about one. None of the years are significant even though the numberof cases in 1988 seems to be a cluster. With this type of situation arising in our analysiswe conclude that the use of one cluster size for each year is inappropriate.If the cluster size must differ across the study years, we want to choose the best kfor each year, so that cells with seemingly excess numbers of cases are significant. Themodified method developed for the general test could be applied to the yearly analysis.Each year is treated separately, as if it was the only data, and the general test with threepossible cluster sizes is performed. For cell i and year y the sequence of cluster sizes43would be and The 95 percentile, k.95,j,e,y, would be basedon the Poisson distribution with mean=The term m8e is thepopulation in cell i and its £ nearest neighbours for age group s and year y, i = 1,.. . , 74,s = 1,. . ., 18, y 83,... , 89. Here, as in the combined year case, the minimum clustersize for which a cell is significant is recorded. The p-value obtained at the minimumcluster size is called the p-value for the cell.Seven p-values are recorded for each school district under the yearly analysis. Acombination of these p-values may indicate if a cell is part of a possible cluster consistently throughout the study period. Let Piy be the p-value obtained for cell i in year y,j = 1,... , 74 and y = 83,••• ,89. Assuming that the yearly analyses are independent,the statistic,T = —2 1np (14)can be approximated by a x2 distribution with 2 x 7 degrees of freedom. Consistentlysmall p-values for a school district produce a large value of T. Large p-values would yielda small value of the statistic. If T is small, we accept the hypothesis that no consistencyof apparent clustering is seen in the yearly analyses. For large values of T we reject thehypothesis. The p-value from each year can be transformed in this way and an overallp-value for consistency can be calculated from the X214 distribution. We call this p—valuethe consistency p-value in order to distinguish it from the p-value calculated for eachyear.The focused yearly analysis is carried out in the same manner as the general yearlysituation just described. The years are considered individually. Three cluster sizes aretested for each mill and a combination of the yearly p-values is made. Again, the clustersizes are based on the population in the mill’s nearest neighbours.445 Discussion of Results5.1 Combined Year AnalysisThe male and female findings by cancer site, in Tables 6 and 7, are listed in the Appendix.Only the sites for which at least one school district or mill, with a relative rate of atleast 1.5, was significant at 5% are displayed. The observed value £ of the statistic L islisted along with the cluster size, k, and the expected number of cases within the schooldistrict and its L nearest neighbours, A. The actual number of cases observed withinthe school district or mill and its nearest L neighbours is denoted 0. The relative rate islabeled 0/A and individual p-value is called p. The p-value obtained from Monte Carlosimulation for the overall test is called P. Sites for which this P is less than 0.1 havethe significant school districts shaded on a map of the province in the general analysis.The term “NSD” for the focused test refers to the mill’s nearest school district.5.1.1 General Male FindingsNo significant school districts with relative rates at least 1.5 are identified for the sitesof colon, lung adenocarcinoma, all cancers except lung, and all cancers for males.LipLip cancer proves to be a site with many significant cells. From Table 6.1 andshaded regions of the map, the school districts around Grand Forks have some of thehighest relative rates for lip cancer. Within Grand Forks’ borders, 7 cases are observedwhen only 1.17 are expected. The relative rate here is a high 5.98 and influences theneighbouring school districts. We note that this result would not be detected if k15was the cluster size investigated. The p-value for Grand Forks is equal to 0.01 and itsneighbours have individual p-values which are between 0.02 and 0.04. Peace River Southis also startling with a relative rate of 4.08. The high rate here heavily influences the45results for Peace River North. Cranbrook comes next with 3.37 times more cases thanexpected. Neighbours Fernie, Kimberly and Creston-Kaslo also form a possible cluster.Chilliwack, Creston-Kaslo, and Abbotsford each have high relative rates which affectsthe cells neighbouring them. Cariboo-Chilcotin and Quesnel have a relative rate of 1.96when considered together. Lip cancer is one of the rarest forms of cancer considered inthis study. It is surprising to find so many cells which have very high rates. The zeroMonte Carlo p-value supports the extreme overall clustering we appear to see.Oral CavityCastlegar and New Westminster are listed for oral cavity cancer. Some clusteringis seen in the Vancouver, New Westminster, Castlegar and Trail areas in Table 6.2.Castlegar has the highest relative rate and lowest individual p-value of the group equaling2.93 and 0.02, respectively. These values influence the neighbouring district of Trail.Trail’s relative rate is 1.56 which is slightly above 1.5. Vancouver has a relative rate of1.63 with 314 observed cases. New Westminster has just over twice the observed casesthan expected. Both Vancouver and New Westminster are considered clusters alone.Both of these two have observed test statistics equal to zero. The Monte Carlo p-valueis evaluated as 0.402. Overall clustering is not supported with this site although someindividual relative rates are quite high.EsophagusThe same can be said for esophagus cancer. Although, overall clustering is notsupported by the Monte Carlo p-value of 0.8 13, Cariboo-Chilcotin and Quesnel have highrelative rates which are listed in Table 6.3. Cariboo-Chilcotin has nine cases observedwhen only 4.65 were expected, yielding a relative risk of 1.93. Quesnel’s results areeffected by neighbour Cariboo-Chilcotin. Consequently, the relative rate for Quesnel islisted as 1.96.46StomachTable 6.4 lists seven significant districts for stomach cancer. The Monte Carlo pvalue is equal to 0.052 which suggests some overall clustering. Clustering appears to belocated in some northern regions of Vancouver Island, Central Coast, Queen Charlotte,and Prince Rupert. Princeton has the highest relative rate of 2.59 with an individualp-value of 0.03. Vancouver Island North has a relative rate of 2.37 which influencesneighbour Central Coast. Alberni has a relative rate of 1.78 and does not require anycells to be combined to achieve at least 19 cases. Campbell River also has the teststatistic equal to 0 and a relative rate of 1.64. Prince Rupert and neighbour Kitimatgive Prince Rupert a significant result. The relative rate is 1.78 with 14 cases observedin the two cells. Queen Charlotte’s test statistic is 2. It along with Prince Rupertand Kitimat have 16 cases observed and a relative rate of 1.75. Both Prince Rupertand Queen Charlotte appear to be influenced by Kitimat even though Kitimat does notappear in this table.RectumRectal cancer has one area of possible clustering. The Hope and Agassiz-Harrisonregion have high relative rates which can be seen in Table 6.6. Agassiz-Harrison has astartling relative rate of 2.29 which influences its neighbour Hope. Hope together withAgassiz-Harrison have a relative rate of 1.64. The overall p-value is equal to 0.565. Thisvalue suggests that overall clustering for this site is not seen in the province.LiverParts of the lower mainland region are significant for liver cancer as seen in Table 6.7. Vancouver and New Westminster have test statistics equal to zero and relativerates 1.84 and 2.19, respectively. Richmond is a possible cluster when considered withneighbours Delta and Vancouver. New Westminster’s high relative rate influences neighbour Burnaby. Burnaby has a relative rate of 1.58 when considered with its neighbour.47A relative rate of 1.71 is seen for North Vancouver when combined with Vancouver.West Vancouver’s relative rate is equal to 1.64 when considered along with neighboursNorth Vancouver and Vancouver. Although several school districts are listed the MonteCarlo p-value of 0.266 suggests that overall clustering is not apparent.PancreasPancreatic cancer has seven districts which have relative rates above 1.5 and aresignificant at 5%. The Monte Carlo p-value found in Table 6.8 is calculated as 0.115.Arrow Lakes has the highest relative rate of this group equal to 3.26. only 1.84 cases areexpected in this school district and five are actually observed. Prince George, Cowichan,and Shuswap each have test statistics of zero. The relative rates for these three are 1.58,1.90, and 1.73. Reveistoke and Lake Cowichan both have require one neighbour to becombined with them. Shuswap is instrumental in Revelstoke’s significance and relativerate of 1.51. Cowichan does the same for Lake Cowichan which results in a relativerate of 1.78. Nechako is influenced by neighbours Prince George and Burns Lake. TheNechako and neighbours expect 17.75 cases but observe 26 cases which gives a relativerate of 1.52.LarynxNechako and Prince George also have high relative rates for cancer of the larynx.As seen in Table 6.9, Prince George has a relative rate of 1.56 and 18 observed cases.Neighbour Nechako, influenced by Prince George, has a relative rate equal to 1.52.Princeton and Vancouver Island West also have large relative rates of 2.5 and 1.58,respectively. Princeton is combined with neighbour Keremeos to achieve that result.Vancouver Island West must be combined with Campbell River and Courtenay to givea relative rate of 1.58. With four school districts listed the Monte Carlo p-value is ahigh 0.479.48Lung SquamousThe lung squamous results are given in Table 6.10. Howe Sound, Smithers, Sooke,and Campbell River have test statistics equal to zero. Of these, Smithers has the highestrelative rate equal to 1.87. Howe Sound, Campbell River, and Sooke follow with relativerates of 1.71, 1.65, and 1.58, respectively. Vancouver Island West along with neighbourCampbell River have a relative rate of 1.61. Stikine is influenced by Fort Nelson andSmithers with relative rate 1.74 and 14 cases observed. This site has an overall p-valueof 0.030. This value implies that overall clustering within the province exists.Lung AdenosquamousLung adenosquamous has four significant school districts which are found in Table 6.11. These four do not suggest overall clustering since the Monte Carlo p-value is0.306. Nanaimo has zero as the observed test statistic. Only 1.92 cases are expectedin the school district and six are actually observed giving Nanaimo a relative rate of3.13. Qualicum is influenced by Nanaimo’s results and has a similarly high relative rate.Sunshine Coast when considered with Nanaimo is also significant and has a relative rateof 2.77. Lake Cowichan along with Cowichan and Nanaimo are also significant. It isapparent that the high relative rate in Nanaimo causes its neighbours also to appear aspossible clusters.Lung Small CellNanaimo is again listed as significant for lung small cell cancer. In Table 6.13 we cansee that Nanaimo has 44 cases observed when only about 27 were expected. Revelstokehas the highest relative rate of those listed. With 2.67 times more cases observedthan expected, it has an individual p-value of 0.03. Golden is influenced by neighbourReveistoke, and thus the relative rate for the combined area is 2.35. Lillooet is alsosignificant when neighbours South Cariboo and Merritt are combined with it. Lillooetand neighbours have 12 cases observed and a relative rate of 1.80. The overall p-value49for this site is 0.413. This number implies that overall clustering is not supported.Lung OthersOverall clustering is supported in the lung others sites, however. In Table 6.14, theMonte Carlo p-value calculated to be zero. The school districts which have zero astheir test statistic are Fernie, Revelstoke, Merritt, New Westminster, Queen Charlotte,Prince George, Mission and Fort Nelson. Fort Nelson tops the list when relative rate isconsidered. Fort Nelson has about one case expected but has five observed within itsborders. These values give a high relative rate of 4.64. These cases obviously influenceStikine whose relative rate is above 3. Queen Charlotte expects 1.9 cases but observesfive. New Westminster has about 41 cases expected and 63 are actually observed.Reveistoke has a relative rate of 2.25 and influences Golden. Merritt has a relative rateof 2.16 with 12 cases observed. Lillooet together with South Cariboo and Merritt has 21observed cases while about 13 were expected resulting in a relative rate equal to 1.59.Prince George has a relative rate of 1.71 and effects the results found for neighbourNechako. Mission has a relative rate of 1.69 as 31 cases were observed when only 18.39were expected.All Lungs and Non-small Cell LungBoth all lungs and non-small cell lungs have the same significant school districts.These sites are found in Table 6.15 and Table 6.16 listing Reveistoke, Merritt and FortNelson. Each school district has test statistic equal to zero for both sites. Fort Nelsonhas the highest relative rate followed by Merritt and Reveistoke. For all lungs the relativerates are 2.40, 1.85, and 2.70, respectively. When non-small cell lungs are considered,the values are 2.57, 1.81, and 1.53. While both sites have the same significant schooldistricts, they have dissimilar Monte Carlo p-values. The all lungs site is significantfor overall clustering with a Monte Carlo p-value equal to 0.036. On the other hand,non-small cell lungs had 0.069 as the Monte Carlo p-value and mildly suggests an overall50pattern of clustering.Soft Tissue SarcomaSoft tissue sarcoma findings are displayed in Table 6.17. No overall pattern of clustering is suggested by the Monte Carlo of 0.071. Grand Forks and Kettle Valley haveextreme relative rates for soft tissue sarcoma, 2.93 and 1.89, respectively. Kettle Valley’sresult is influenced by the five cases seen in Grand Forks. Vancouver has a relative rateof 2.01 and test statistic equal to 0. North Vancouver is affected by Vancouver and givesa relative rate of 1.85 when both regions are considered together. Both school districtscause West Vancouver to be significant with relative rate 1.79. Vancouver apparentlyalso influences Burnaby and Richmond. Both areas list the same relative rate of 1.69and test statistic equal to two. Mission is significant itself with a relative rate of 1.89and ten observed cases.MelanomaThree districts are noted for melanoma in Table 6.18. All three have test statisticsequal to zero and the Monte Carlo p-value is evaluated as 0.544. Merritt has the highestrelative rate of the bunch, 2.43, and nine cases are observed in its school district. Saanichhas 34 cases observed to give a relative rate of 1.72. Thirty-one cases are seen in WestVancouver which is just over 1.5 times what is expected.ProstateTable 6.20 lists three districts for prostate cancer. The overall p-value is equal to0.006 here and strongly indicates overall clustering. Burns Lake, Nechako, and PrinceGeorge are each significant individually. The relative rates are 1.69, 1.55, and 1.55,respectively. Burns Lake has 24 cases observed and Nechako has 42 observed. Onehundred seventy-eight cases are seen in Prince George when only about 115 are expected.TestisTesticular cancer results show some clustering in the Gulf Islands, Courtenay, and51Vancouver Island North areas. Table 6.21 shows Gulf Islands with the highest relativerate of 4.16 and individual p-value 0.01. Only 1.2 cases are expected in that area and fiveare actually found. Vancouver Island North has 2.95 cases expected and eight observedgiving a relative rate of 2.71. This result affects neighbour Central Coast producinga relative rate of 2.55. Courtenay is significant with a relative rate 1.92 which allowsneighbour Powell River to also be listed with a relative rate of 1.85. Trail is seen withabout three cases expected and seven actually observed. With six school districts listed,the Monte Carlo p-value is calculated to be 0.254.BladderSeven districts give a Monte Carlo p-value of 0.019 indicating overall clustering forbladder cancer. In Table 6.22, South Cariboo has the highest relative rate with 2.45and lowest individual p-value equal to 0.02. These values seem to induce a significantresult for Lillooet as well. Lillooet’s test statistic is equal to two and has a relative rateof 1.59. Campbell River follows next with a relative rate of 2.36 and thirty-two casesobserved. It apparently affects neighbours Vancouver Island North and West for theytoo have relative rates above two. Specifically, Vancouver Island West has a relativerate of 2.36 while Vancouver Island North has Relative rate 2.20. Mission and Terraceare also listed with relative rates 2.01 and 1.70, respectively.KidneyA few patches throughout the province are seen to be significant for kidney cancer.The Monte Carlo p-value of 0.069 suggests some overall clustering. The largest relativerate is visible at Lake Cowichan in Table 6.23. Five cases are observed when less thantwo are expected. The relative rate for Lake Cowichan is 2.70. Kitimat has a relativerate just below Lake Cowichan at 2.68. Seven cases are observed in this district whenabout three would be expected. These results produce significance for Queen Charlotteand Prince Rupert. The former has a relative rate of 1.71 while the latter has 1.84.52Mission, Cranbrook, and Maple Ridge each have zero as their observed test statistic.Cranbrook has 12 cases observed when oniy 6.10 are expected. Maple Ridge and Missionhave relative rates of 1.74 and 1.76, respectively.BrainMaple Ridge is the only possible cluster for brain cancer. Table 6.24 shows that thisschool district has a relative rate of 1.76 with nineteen cases observed. With only onepossible cluster the intuition is that overall clustering is not present. This is supportedby the overall p-value which is 0.923.Hodgkins DiseaseSmithers and Burns Lake are possible clusters for Hodgkins Disease. Smithers hasan observed test statistic equal to 0 as seen in Table 6.25. Smithers has a relative rate of3.52 with five observed cases and causes Burns Lake also to appear on the table. BurnsLake, when combined with Nechako and Smithers give a relative rate of 2.43, mostlydue to the five cases in Smithers. The Monte Carlo p-value is a high 0.813 which is notsuggestive of overall clustering.Non-Ho dgkins LymphomaNon-Hodgkins lymphoma comes next in Table 6.26. The Monte Carlo p-value of0.298 is also not suggestive of overall clustering for this site. Maple Ridge and Vancouverare significant on their own. The former has 33 observed cases and relative rate equalto 1.61 while the latter has 42 observed cases and a relative rate of 1.52. Cranbrookand Kimberly are significant together, have a relative rate of 1.55 and twenty-one casesobserved.Multiple MyelomaMultiple myeloma has a Monte Carlo p-value of 0.002, found in Table 6.27. A fewpatches of high relative rates are seen. The Nelson, Castlegar, Arrow Lakes, and Trailregions are a possible cluster. Arrow Lakes has a relative rate of 2.00 and the remaining53school districts of this area have 1.74 as a relative risk. The South Okanagan andKeremeos need to be combined with Penticton to have significant results. Twenty-onecases are observed in the three school districts giving a relative rate of 1.5. Merritt hassix cases observed, a relative rate of 4.66 and individual p-value equal to 0.01. Merrittinduces Lillooet also to be significant for this site with a relative of 2.30. Nechako andPrince George together have a relative rate of 1.73 and fifteen observed cases. CampbellRiver’s eight cases influence Vancouver Island West and North. Campbell River has arelative rate of 2.22 while both Vancouver Island North and West have relative rates of2.09.LeukemiaFor leukemia, listed in Table 6.28, overall clustering is not supported by the MonteCarlo p-value of 0.136. Vancouver Island North has the highest relative rate of 2.48with eight cases observed. Its result influences Central Coast giving 1.92 as the relativerate. Qualicum comes next with twenty-five cases observed and a relative rate equal to1.83. Cowichan has a relative rate of 1.72 which affects neighbouring Lake Cowichan.The resulting relative rate for Lake Cowichan is thus 1.65. Cowichan also affects itsneighbour Gulf Islands which subsequently has a relative rate of 1.6 and forty observedcases.Acute LeukemiaQualicum also appears under the acute leukemia site appearing in Table 6.29. Ithas nine observed cases and a relative rate of 3.78. Nanaimo is influenced be Qualicumand together they have a relative rate of 1.79 and 21 cases observed. Lillooet, whencombined with South Cariboo and Merritt, has six observed cases and a relative rate of2.32. Neither South Cariboo nor Merritt appear significant when combined with any oftheir neighbours. The overall p-value is evaluated as 0.681 and overall clustering is notsuggested.54Chronic LeukemiaCowichan appears to be the center of clustering for chronic leukemia. In Table 6.30,Cowichan has twenty-seven observed cases when only 13.39. The relative rate is calculated to be 2.02. Gulf Islands and Lake Cowichan are effected by this and are also listedas significant. Gulf Islands has a relative rate of 1.85 while Lake Cowichan’s relative rateis evaluated as 1.96. Vancouver Island itself is significant with six cases and a relativerate of 2.82. The overall p-value for this site is 0.465 and is not suggestive of overallclustering.Other SitesNechako and Powell River appear for other sites in Table 6.31. Nechako has elevencases when 5.20 are expected, producing a relative rate of 2.12. Powell River has arelative rate of 1.64 with fifteen observed cases. With only two districts listed, overallclustering is not expected and the Monte Carlo p-value of 0.733 supports that expectation.Primary UnknownPrimary unknown possible clusters are noted in Table 6.32. Prince Rupert has17 observed cases and a relative rate of 1.75. These results influence Queen Charlotte.When Queen Charlotte is combined with Prince Rupert the relative rate is 1.64. Overallcluster is not apparent with an overall p-value of 0.546.Cob-RectalCob-rectal results appear in Table 6.36. Only Agassiz-Harrison is noted with nineteen cases found within its borders. The relative rate is 1.85 and the individual p-valueis 0.03. With only one significant school district, overall clustering is not expected. TheMonte Carlo p-value is calculated as 0.632 and supports the expectation.555.1.2 General Female FindingsThe female findings for the general test are tabled in the Appendix starting at page 125.There is insufficient evidence to suggest that the sites of lip, liver, lungs:non-small cells,all lungs, soft tissue sarcoma, breast, all cancers except lung, and all cancers have anypossible clusters in our analysis.Oral CavityAlberni is the only cell listed as significant for oral cavity cancer found in Table 7.2.Its relative rate is 2.28 with twelve cases observed. Overall clustering is not seen asindicated by a Monte Carlo p-value equal to 0.922.EsophagusTable 7.3 lists four significant school districts for esophagael cancer. Prince Georgeis the only one listed with a test statistic equal to zero. The relative rate is 2.46 andseven cases are observed. When Quesnel is combined with Cariboo-Chilcotin and PrinceGeorge it is significant with relative rate equal to 1.84. Gulf Islands and Cowichan arecombined with Saanich both cells have a relative rate of 1.66 and sixteen observedcases. The overall p-value is calculated as 0.501 for this case. Thus, no overall patternof clustering is implied.StomachFour school districts are listed again for stomach cancer noted in Table 7.4. CampbellRiver has a high relative rate of 1.92. Nine cases are observed for that school districtwhen 4.69 are expected. When Vancouver Island North is considered with neighboursVancouver Island West and Campbell River, it too is significant. Those three districtscombined have a relative rate calculated as two. Thirty-seven cases are observed inRichmond when only about 22 are expected. This situation produces a relative rate of1.68. Richmond also influences Delta’s results. Together the two have a relative rate of1.54. Overall clustering is also not suggested for this site with a Monte Carlo p-value of560.497.ColonTable 7.5 displays the significant school districts for colon cancer. Kettle Valley andSunshine Coast both have zero as their observed test statistics. Kettle Valley has sevencases producing a relative rate of 2.37. Sunshine Coast has a lower relative rate evaluatedto be 1.69. Again, the overall p-value (0.454) does not indicate overall clustering.RectumRectal cancer is next in Table 7.6. Only two school districts appear significant. TheMonte Carlo p-value was calculated to be 0.694. Nine cases are observed in Princetongiving an extreme relative rate of 3.48. Powell River has a lower relative rate which isequal to 1.65 and eighteen cases are seen within its border.PancreasTrail appears to be the center of a possible cluster containing Nelson, Castlegar, andGrand Forks for the pancreas site. Table 7.8 shows Trail with a relative rate equal to2.49 with seventeen cases observed. Castlegar’s rate is 2.01 while Nelson and GrandForks have relative rates just around 1.75. Langley and Cowichan also are significantwith relative rates 1.53 and 1.63, respectively. Both are possible clusters on their ownand have around twenty cases observed. The Monte Carlo p-value for this site is equalto 0.186. With such a value, overall clustering in the province is not evident.LarynxThe larynx site is given in Table 7.9. Greater Victoria is the only cell listed with zeroas the observed test statistic. The relative rate is 1.81 with seventeen actual cases withinits boundary. Sooke’s nearest neighbour is Greater Vancouver. Hence, Sooke’s relativerate of 1.70 is influenced by Greater Victoria. Lake Cowichan along with neighbourCowichan have six observed cases when only 1.77 were expected. These numbers resultin a relative rate of 3.39. This site does not appear to have overall clustering as conveyed57by the Monte Carlo p-value, 0.538.Lung SquamousTable 7.10 lists the significant school districts which are found for lung squamouscancer. With nine possible clusters, the Monte Carlo p-value of 0.044 suggests overallclustering within the province. Quesnel, Nanaimo, and Courtenay have the test statisticequal to zero. The relative rates for these three are calculated as 2.16, 2.02, and 1.99, respectively. Nanaimo’s results influence Sunshine Coast, Lake Cowichan, and Qualicum.Sunshine Coast has a relative rate of 1.92 while Lake Cowichan and Qualicum haveabout 1.6 times as many cases observed than expected. Courtenay affects Powell Riverand Campbell River producing relative rates of 1.84 and 1.93, respectively. When Vancouver Island West is combined with both Courtenay and Campbell River, the cell issignificant with a relative rate of 1.90 and an individual p-value of 0.03.Lung AdenosquamousFourteen cells are significant for the site lung adenosquamous. Table 7.11 indicatesthat the relative rates are very high for the cells listed. The value of L is very interestingfor these districts. The lowest value for the statistic occurs for Nanaimo. Nanaimo andits two neighbours, Qualicum and Sunshine Coast, have five cases observed producing arelative rate of 3.13. The remaining cells are influenced by these three and also appearsignificant even with large values of L. With only 30 cases in the entire province, itis quite startling to have 1/6 of the cases in this area of the province which containsroughly 4% of the province’s population of women. The Monte Carlo p-value is givenas 0.003 which is not surprising under the circumstances.Lung AdenocarcinomaCariboo-Chilcotin and Kamloops have the highest relative rates for lung adenocarcinoma found in Table 7.12. The relative rates are, respectively, 1.76 and 1.55. Kamloops’values influence neighbouring North Thompson. Its relative rate is 1.52 with 43 cases58observed. With only three districts significant, overall clustering is not expected. TheMonte Carlo p-value is 0.521 which does not connote overall clustering.Lung Small CellLung small cell cancer has many districts listed as significant with a Monte Carlop-value (0.014) which suggests overall clustering. Chilliwack and Courtenay appear tobe the regions of most concern in Table 7.13. Chilliwack has nineteen observed casesresulting in a relative rate of 1.61 while Courtenay has three fewer observed cases and arelative rate of 1.81. Their neighbours appear to be influenced by the number of casesobserved in these two districts. Hope and Agassiz-Harrison are influenced by each otherand Chilliwack. Together, the three have a relative rate of 1.53. Campbell River andPowell River significant results are based on Courtenay’s. Alberni and Vancouver IslandWest are also influenced by Courtenay and have relative rates 1.53 and 1.68, respectively.Lake Cowichan along with Cowichan have a relative rate of 1.62. Kamloops and Merritttogether have a relative rate of 1.54.Lung OthersTable 7.14 has lung others with an overall p-value calculated as 0.162. Overallclustering is not implied by this result although Quesnel has a relative rate of 2.33.With eleven cases observed and 4.73 expected the individual p-value for Quesnel is 0.02.New Westminster has a relative rate of 1.5 with 36 cases observed within its border.Sunshine Coast has fourteen cases observed with a relative rate of 1.68. Peace RiverSouth has 2.06 times more cases observed than expected. This value causes Peace RiverNorth and Fort Nelson to also appear significant.MelanomaMelanoma has six school districts with an overall p-value for 0.113. In Table 7.18Kimberley has the highest relative rate which equals 2.13. West Vancouver has a relativerate of 1.82 with 41 observed cases. Howe Sound is influenced by neighboui West59Vancouver and also appears in this table. Cowichan’s relative rate is 1.52 with anindividual p-value of 0.04. Cowichan induces Gulf Islands to be significant as well. Therelative rate for Langley is 1.54 with 44 cases found in its district.BladderBladder cancer comes next in Table 7.22. With a Monte Carlo p-value of 0.673overall clustering is not suggested. West Vancouver has a relative rate of 1.68 whichinduces neigbhouring Howe Sound to appear significant. When Princeton is combinedwith Keremeos, six cases are observed producing a relative rate of 3.11.KidneyWhen Golden is amalgamated with Revelstoke, five cases of kidney cancer are observed in Table 7.23. The relative rate is given as 2.55 and 1.96 cases were expected.Queen Charlotte in combination with Prince Rupert yields a relative rate of 2.73 andseven observed cases. Chilliwack must be combined with Mission and Agassiz-Harrisonto produce a 1.52 times more cases than expected. The Monte Carlo p-value for thissite is 0.649 and overall clustering is not supported.BrainBrain results are found in Table 7.24. Castlegar has a relative rate of 2.56 whichcauses Trail to also be significant. North Vancouver, West Vancouver, and Cowichanhave relative rates 1.76, 1.64, and 1.78 respectively, with zero as the test statistic.Sunshine Coast is influenced by Nanaimo and North Vancouver for its significance. HoweSound has a p-value equal to 0.04 when combined with North and West Vancouvers.Lake Cowichan with Nanaimo and Cowichan gives a relative rate of 1.51. The MonteCarlo p-value is 0.103 and implies some overall clustering.Hodgkins DiseaseWhen Delta and Richmond are combined, twenty-two cases of Hodgkins Disease areobserved. The relative rate for these two is 1.57 as seen in Table 7.25. Coqiiitlam has60fifteen cases observed when only 8.87 are expected yielding a relative rate of 1.69. TheMonte Carlo p-value in this case is 0.652 and again clustering is not suggested.Non-Hodgkins LymphomaThe Monte Carlo p-value for non-Hodgkins lymphoma is 0.258 as seen in Table 7.26.Reveistoke and Hope have relative rates 2.47 and 2.73, respectively. Terrace has ten casesobserved yielding a relative rate of 2.02. The high rate in Terrace produces Smithersas a significant school district with relative rate 1.92. Lake Cowichan has test statisticzero with five observed cases and a relative rate of 2.98.Multiple MyelomaMultiple myeloma produces eight possible clusters in Table 7.27. Penticton has arelative rate of 2.28 and fifteen observed cases. Keremeos and South Okanagan areinfluenced by the high relative rate and are also significant with 1.90 as their relativerates. Summerland also is significant due to its neighbour Penticton. Langley has arelative rate of 1.83 and causes Abbotsford to also be a possible cluster. Nechako andPrince George together have relative rate 2.11. With an overall p-value of 0.104 this sitemay suggest some overall clustering.LeukemiaPenticton influences South Okanagan and Keremeos for leukemia. In Table 7.28Penticton has test statistic zero and relative rate 1.76. The Gulf Islands and Cowichanform an area with a relative rate of 1.57 and twenty-five observed cases. The MonteCarlo p-value of 0.283 suggests that overall clustering is not apparent.Acute LeukemiaTable 7.29 shows Campbell River as the primary location of clustering of acuteleukemia. With a relative rate of 2.95 and seven observed cases, it causes neighbouringVancouver Island West to also appear on the table. With only two districts listed theoverall p-value is 0.799.61Chronic LeukemiaChronic leukemia findings are noted in Table 7.30. Here again, as it was in leukemia,Penticton has a high relative rate which influences South Okanagan and Keremeos. Penticton has a relative rate of 1.86 while the other two have relative rate 1.68. Abbotsfordhas 23 cases found within its borders and relative rate 1.71. Mission’s nearest neighbouris Abbotsford and is significant as well. Cowichan has relative rate 1.83 and inducessignificance in Gulf Islands and Lake Cowichan. With eight school districts significant,the overall p-value is calculated as 0.073.Other SitesIn Table 7.31, Grand Forks has the highest relative rate for other sites. With a teststatistic equal to zero, Grand Forks has a relative rate of 2.35. Neighbour Kettle Valleyis affected with fifteen observed cases and relative rate 2.23. The overall p-value was0.655 and does not suggest any overall clustering.Primary UnknownOverall clustering is suggested for primary unknown cancer. Five school districts arelisted in Table 7.32 with a Monte Carlo p-value calculated as 0.088. Cariboo-Chilcotinhas relative rate 1.56 with twenty-two observed cases. Merritt’s relative rate is 2.41and causes neighbouring Lillooet also to be significant. When Lillooet is combined withMerritt and South Cariboo the resulting relative rate is 1.65. Prince Rupert and Missionhave zero as their test statistics and relative rates 1.78 and 1.55, respectively.OvaryOnly Creston-Kaslo appears for ovarian cancer in Table 7.34. With seventeen observed cases the relative rate is 1.72. No overall clustering is apparent with a MonteCarlo p-value of 0.895.CervixTable 7.35 has three significant school districts for cervix. Princeton, with five ob62served and 1.65 expected cases, has an individual p-value of 0.03. Quesnel has a relativerate equal to 2.23. Central Coast is significant with relative rate 1.78 when combinedwith Vancouver Island North and Burns Lake. With an overall p-value of 0.535 overallclustering is not indicated.Cob-RectalPrinceton and Sunshine Coast appear in Table 7.36 for cob-rectal. Princeton hassixteen observed cases giving a relative rate 2.19. Sunshine Coast has a relative rateof 1.53 and 56 actual cases found within its borders. Again, no overall clustering issuggested with a Monte Carlo p-value of 0.314.EndometriumEndometrial results are displayed in Table 7.39. The Monte Carlo p-value is 0.221with Gulf Island producing the highest relative rate. Gulf Islands has a relative rate of1.87 with test statistic zero. West Vancouver’s rate of 1.68 causes neighbouring HoweSound to appear significant as well. Summerland expects about ten cases but observessixteen giving it a relative rate of 1.61.5.1.3 Focused FindingsThe combined year results for the focused test start at Table 8 found in the Appendix.The notation used is the same as was found in the tables for the general test. The termNSD means the nearest school district centroid to the focus. We call the school districtwhere the focus is located, the home school district. The results mostly mimic thosefound in the general test for school districts which contain pulp and paper mills. Thenearest cell centroids to the foci are generally the nearest centroids to the foci’s homeschool district. When we consider the first few nearest neighbours results this is certainlytrue. The Ocean Falls mill is an exception. Ocean Falls home school district is CentralCoast. Ocean Falls’ three closest neighbours are Central Coast, Kitimat, and Terrace,63respectively. However, the Central Coast’s first two closest neighbours are VancouverIsland North and Burns Lake. Therefore, the focused results for Ocean Falls will notcorrespond to the general results for its home school district. Skookumchuck is anotherexception. Although the Skookumchuck mill is located in the Cranbrook school district,its nearest neighbour is the Kimberley school district. Any result which Skookumchuckhas will be represented in the general results for Kimberley.MacKenzie and Prince George have the same first three nearest neighbours. Thus,their results are identical. Although both will appear significant for some site, theyreally only represent one region.Individual results for the foci are almost identical to the results found in their homeschool district. Only the observed value of the test statistic is one larger because thefoci itself does not contain cases or population. When only testing foci and not all cellswithin the province, the Monte Carlo p-value will not be the same as it was for thegeneral test.In the general test for males, the sites of lip, stomach, lung squamous, lung others, alllungs, lungs:non-small cells, soft tissue sarcoma, prostate, bladder, kidney, and multiplemyeloma all had Monte Carlo p-values at most 0.1. The focused test also shows evidenceof overall clustering for stomach, lung squamous, lung others, prostate, bladder, kidney,and multiple myeloma. The Monte Carlo p-value for pancreatic cancer suggests overallclustering in the focused test with a p-value of 0.070. In the general framework, thepancreas site had an overall p-value just above 0.11.A few mills are possible clusters for a few different sites while others appear significant. Quesnel is significant for lip and esophagus. Skookumchuck is listed for lipand only Castlegar is a possible cluster for oral cavity. Port Alberni only appearssignificant for stomach cancer. Prince Rupert is significant for stomach, lung others,kidney, and primary unknown. Port Alice is also a possible cluster for stomach and the64sites of testis, leukemia, and chronic leukemia. Ocean Falls is listed for stomach andkidney while Campbell River is significant for stomach, lung squamous, bladder, andmultiple myeloma. Prince George and MacKenzie possible clusters for pancreas, larynx,lung others, prostate, and multiple myeloma. Crofton appears significant for pancreas,leukemia, and chronic leukemia. Squamish is only listed for lung squamous. Gold Riveralso appears for that site and the sites of bladder, multiple myeloma. Nanaimo maybe a cluster for lung adenosquamous and lung small cell. Powell River is significant fortestis and other sites. Kitimat only appears significant for kidney cancer.The female lung squamous, lung adenosquamous, lung small cell, chronic leukemia,and primary unknown sites has overall p-values at most 0.1 for the general test. In thefocused situation listed in Table 9, lung squamous, lung adenosquamous, and lung smallcell sites also have Monte Carlo p-values which suggest overall clustering. Melanomaand endometrium have focused Monte Carlo p-values 0.052 and 0.099 which also suggestoverall clustering.Port Alberni is a possible cluster for oral cavity. Prince George and MacKenzie havesignificant results for esophagus, lung adenosquamous, and multiple myeloma. CampbellRiver may be a cluster for stomach, lung squamous, lung small cell, and acute leukemia.Port Mellon is listed for colon, lung others, melanoma, bladder, brain, cob-rectal, andendometrium. Powell River is a possible cluster for rectum, lung squamous, and lungsmall cell. Castlegar is significant for pancreas and brain while Crofton is significantfor those sites as well as melanoma, leukemia, and chronic leukemia. Lung squamousand lung adenosquamous have Nanaimo as a possible cluster. Quesnel is listed for lungsquamous, lung others, and cervix. Kitimat only appears significant for lung adenosquamous. Prince Rupert may be a cluster for lung adenosquamous and primary unknown.Lung adenocarcinoma and lung small cell both have Kamboops as a possible cluster.Squamish appears significant for melanoma, bladder, and endometrium. Gold River is65only listed for acute leukemia.Generally speaking, the results are quite consistent between the general and focusedtests. This consistency is expected because of the choice of cluster size each mill orschool district is tested at and the fact that the mill neighbours are similar, if not exact,to the home school district.5.2 Yearly AnalysisThe results for the yearly general analysis are listed starting on page 138 of the Appendix.The school districts or mills which satisfied at least one of the following conditions arelisted for each site. Only cells which had either a consistency p-value at most 0.05, or twoconsecutive significant years, or at least 3 significant years are listed. The consistencyp-value calculated over the seven years is denoted Pc. The cluster size, observed valueof the statistic, and the relative rate, labeled k, £, and Of.\, respectively, are listed foreach year, 1983 through 1989. When a school district or mill is significant at 5% for aparticular year, that year’s relative rate has an asterisk (*) beside it. When k is smallthe number of expected cases is also very small, usually less than one. The relativerates may seem very extreme since the expected cancer counts are real numbers and theobserved are integer values. When k is small, we note that the relative rate is inflatedbecause a fraction of a person cannot have cancer even though that is what is expected.5.2.1 General Male FindingsLipGrand Forks and Kettle Valley appear significant in the years 1985 and 1986 forlip cancer in Table 10.1. Kamloops and South Cariboo have two significant years inthe beginning of the study period. The same can be said for the Peace River districtsand Fort Nelson. The years 1986—88 are significant for Hope, Chilliwack and Agassiz66harrison. The cluster sizes are very small here and the relative rates are inflated. Withsuch small cluster sizes, the expected number of cases is very small, close to zero forsparsely populated cells. The relative rates, thus, look more startling that they actuallyare.Oral CavityOral cavity cancer displays some school districts with consistently significant years.In Table 10.2, we see that the lower mainland cells figure prominently; Vancouver andBurnaby have six significant years each. Vancouver has relative rates ranging from1.36 to 2.06 during the significant years. Except for 1985 Burnaby has relative ratesranging from 1.30 to 1.90 with a consistency p-value equal to zero. New Westminsterhas high relative rates for the last four study years. The relative rates are between 1.62and 3.88 for that period. Richmond is significant for all years except 1986 and 1988.The maximum relative rate occurs during 1983. Castlegar is an apparent cluster for1983, 1988 and 1989. In those years, the test statistic is zero while in the remainingyears, the test statistic is ten or eleven. The relative rates are very high due to thevery small number of cases expected, usually less than one. West Vancouver’s resultsare closely related to neighbour North Vancouver. Both are significant for 1983—85 and1989. Both have consistency p-values less than 0.05. Prince Rupert only has low valuesof the test statistic for 1985—86. During these two years, the relative rates are 3.15 and4.41, respectively. Both years have fewer than five cases. Castlegar, Vancouver and NewWestminster are also significant for the combined year analysis.EsophagusVancouver also shows a consistency for esophagael cancer, which is listed in Table 10.3. Vancouver has relative rates of 1.83 and 1.47 in 1984 and 1986, respectively.The remaining years have large values of the test statistics. North Thompson andCariboo-Chilcotin have consistency p-values equal to zero. Cariboo-Chilcotin was sig67nificant in the combined year analysis and has significant results for the years 1984,1988, and 1989. The relative rates exceed 3.5 for these years but less than five cases areobserved. North Thompson has a very extreme relative rate of 23.65 in 1984. only avery small fraction of one case is expected when the cluster size is 2. In 1989 the sameis seen where the relative rate is 28.10 and a minimum of two cases are observed.StomachQueen Charlotte, Prince Rupert, and Vancouver Island North are significant for 1986and 1987 for stomach cancer. Table 10.4 shows that the lowest consistency p-value isequal to 0.04 for Queen Charlotte. Queen Charlotte is influenced by Prince Rupertwhich has relative rates equal to 3.59 and 4.38 for 1988—89, respectively. VancouverIsland North, when combined with Vancouver Island West and Campbell River, hasrelative rates of 2.61 and 2.64 for the same two years. All three cells listed are alsosignificant for this site when years are combined.RectumMost of the districts listed for rectal cancer in Table 10.6 have a couple of significantyears without any overall consistency shown. Hope and Agassiz-Harrison were significantfor the combined year analysis although neither appears here. Sunshine Coast is onlysignificant for 1988—89 with relative rates about 1.66 in each of those years. Peace RiverSouth has a relative rate of 2.62 and five observed cases in 1984. In 1985 the districtis also significant with a relative rate of 1.84 when considered with Peace River Northand Prince George. Nanaimo has three significant years. In 1986, 1988, and 1989 thecluster sizes are all equal to fifteen with observed test statistic zero. The relative ratesare 1.66, 1.97 and 1.75 for these years respectively. These values influence neighbourLake Cowichan which is significant for the last two study years. Campbell River is alsosignificant for those two years with relative rates 2.74 and 1.78. Vancouver Island Westis influenced by these rates and is only significant for those two years as well with similar68relative rates. Creston-Kaslo is a possible cluster for the first two years when the teststatistic is zero. Both years have relative rates of about 2.5 with five or six observedcases.LiverThe liver cancer results in Table 10.7 have the same cells listed which were seen inthe combined year results. Vancouver seems to affect its neighbours. Vancouver is nota possible cluster for 1986 where the test statistic is four, but is for every other year. In1985 and 1989 three relative rates are 2.21 and the lowest relative rate for a significantyear is 1.58 in 1988. The significant years all have test statistic equal to zero and theconsistency p-value is found to be the same. Burnaby is significant for each study year.For most years the test statistic is equal to two meaning Burnaby was combined withNew Westminster and Vancouver. The relative rates range from 1.53 to 2.08. NewWestminster is a possible cluster for the last two years of study. In 1988 the relativerate was 1.95 while in 1989 the relative rate was 3.53. North Vancouver is significantfor the years when it is combined with Vancouver. West Vancouver is a possible clusterwhen considered along with the two neighbours North Vancouver and Vancouver. OnlyNew Westminster has a consistency p-value which is different from zero and is equal to0.07. The other school districts significance is based on the high relative rates found inVancouver.PancreasNo school districts satisfied our requirements to be tabled for pancreatic cancer. OnlyCoquitlam is listed for larynx cancer in Table 10.9. Although it was not significant whenyears were combined, it is significant for 1986 and 1987. For both years, the statisticequals one. When Coquitlam is combined with New Westminster the relative rates are2.19 and 2.15 in 1986 and 1987, respectively. The consistency p-value is calculated as0.26 indicating no overall consistency.69Lung SquamousFor lung squamous, all districts tabled have two consecutive significant years. InTable 10.10 every cell is significant for 1987. Vancouver Island West and Campbell Riverare also possible clusters for the combined year analysis. Courtenay appears to be theschool district which influences its neighbours. In 1986 and 1987 the test statistic is zeroand the relative rates are 2.43 and 2.01, respectively. The resulting consistency p-valueis equal to 0.09. Campbell River is significant for those two years as well when combinedwith Courtenay. Vancouver Island West shows the same influence when added to bothCampbell River and Courtenay. The relative rates are 2.24 and 1.83 for 1986 and 1987,respectively for Vancouver Island West. Powell River has relative rates around 1.9 forthose years when Courtenay is combined with it. For 1987—88 Chilliwack and AgassizHarrison are significant with relative rates of about 1.75 in each of those years. WhenHope is combined with these two it also has similar relative rates for those significantyears.Lung AdenosquamousNanaimo has three consecutive significant years in the site lung adenosquamous. Itwas significant for combined years and in Table 10.11 the cluster sizes involved are verysmall. Even with a test statistic of four in 1987 the result is significant. During 1986—88the relative rates are 6.69, 4.10, and 9.51, respectively, because less than one case isexpected. Nanaimo’s 1988 value makes neighbour Qualicum significant with a relativerate of 6.24 and cluster size 2. Qualicum is also a possible cluster of size two for 1983with a relative rate of 11.47. Alberni also has a high relative rate in this year due toits neighbour Qualicum. Lake Cowichan has relative rates of 4.24 and 8.70 for 1987and 1988, respectively. The cluster sizes are three or less and are affected by Nanaimoand Qualicum. Lake Cowichan, Nanaimo, and Qualicum were each seen listed in thecombined year results for this site.70Lung AdenocarcinomaLung adenocarcinoma findings are displayed in Table 10.12. Merritt is a possiblecluster for 1984—85. Merritt alone has a relative rate of 4.35 for cluster size three in1984. When Merritt is combined with Kamloops in 1985 the relative rate is 2.05 for acluster size equal to eleven. Richmond is a possible cluster for the first two study years.In 1983 the relative rate is 1.75 and equals 2.46 in 1984. When Coquitlam is combinedwith New Westminster it is significant for the same two years with relative rates 1.61and 1.69. Campbell River is significant for 1986 and 1987 with relative rates 1.74 and1.86, respectively.Lung Small CellCampbell River also has two significant years for lung small cell cancer seen inTable 10.13. In 1988 it has a relative rate of 2.97 for cluster size four. Together withCourtenay, Campbell River has a relative rate of 2.49 for 1989. The consistency p-valueis far above 0.05 for this school district and is evaluated as 0.24. The years between1986 and 1988 inclusive are significant for Golden. During this period the cluster sizeis equal to three with relative rates 5.65, 3.67, and 6.66. In 1987 Golden must becombined with Reveistoke and Windermere while the other two significailt years requireonly Reveistoke be added to Golden. Golden is the only one listed here which was alsofound in the combined year analysis. Burns Lake also has two significant years with aconsistency p-value equal to 0.06. In 1986 it is significant alone and significant for 1985when amalgamated with Nechako.Lung OthersFor lung others, fourteen school districts are tabled. Table 10.14 has Golden significant for the first two study years. The relative rates are equal to 5.18 and 6.61 withcluster sizes 4 and 3 for those two years respectively. Kamloops is a possible cluster ofsize twelve for both 1987 and 1988. The relative rates are equal to 2.00 and 1.74 for71those years. Quesnel is significant for 1987 and 1989 when combined with neighboursCariboo-Chilcotin and Prince George. The relative rates are 2.07 and 1.71 for clustersize fifteen. Prince George has zero as its observed test statistic for 1986, 1987, and 1989.The relative rates range from 1.96 to 2.80 in those years while the cluster size tested iseither eight or nine. The consistency p-value for this site is equal to 0.01. Nechako hasvery similar results based on its close proximity to Prince George. Peace River Southand Peace River North combined with Prince George are significant for 1986 and 1989with relative rates 2.21 and 1.79 in those years. Peace River South is also significantwith a relative rate of 2.68 in 1984 which also causes Peace River North to be significantfor the same year. Merritt is a possible cluster for four of the study years. In 1985 witha cluster size of three the relative rate is calculated to be 3.92. The test statistic is alsoequal to zero for 1989 where the relative rate is 4.75 for a cluster size of four. Merritttogether with Kamloops is significant for 1987—88. South Cariboo has a consistencyp-value equal to 0.02 with significant years 1986 and 1987. The 1986 finding is due tothe high relative rate for that year in neighbouring Lillooet while the 1987 relative rateof 1.83 occurs when Lillooet and Kamloops are combined with South Cariboo. Thesignificance obtained by Lillooet in 1985 with a relative rate of 3.30 occurs when it iscombined with South Cariboo and Merritt. Surrey is significant for 1985, 1987, and1989 with relative rates less than 1.5. New Westminster has relative rates for 1985,1986, and 1989 with relative rates 2.75, 2.04 and 1.76, respectively. Queen Charlottehas high relative rates for 1986—86 when two is the cluster size. Kitimat is significantfor the two years 1987 and 1988. In the latter year, the test statistic is zero for clustersize three. In the former year, the relative rate is 2.22 when Kitimat is combined withTerrace and Prince Rupert.All LungsFor the site of all lungs, South Cariboo, Chilliwack, Surrey, Richmond, Vancouver,72Burnaby, Coquitlam, North Vancouver, and Powell River fail to have significant yearsin which the relative rate is at least 1.5. In Table 10.15, Kamloops is significant onits own for 1988 with a relative rate of 1.55. This rate induces neighbouring NorthThompson to be significant with a relative rate of 1.52 and Merritt as well. Merritt hasthree significant years, a p-value equal to zero, and was significant in the combined yearframework. In 1985 and 1989 the relative rates are 2.42 and 3.49. When a cluster sizeof twenty is tested for the years 1985 and 1986 Maple Ridge has relative rates 1.52 and1.65. Courtenay’s relative rates are around 1.55 for 1987—88 and produce significancein Campbell River as well for those years. Vancouver Island West’s high relative rateshappen when it is combined with Campbell River.Non-Small Cell LungTable 10.16 has the non-small cell lung cancer results. Again, several school districtslisted have relative rates below 1.5. Merritt appears again as it did in the combinedyear situation. In 1984 and 1989 the relative rates are 2.37 and 3.66 with a consistencyp-value of zero. Kamloops has two consecutive years in which the relative rate was justhigher than 1.5. Chilliwack has a relative rate for 1988 equal to 1.58. New Westminster’srelative rates for 1985 and 1989 are 1.67 and 1.54, respectively. Courtenay has relativerates of 1.69 in 1986 and 1988 with nineteen as the cluster size. These values inducesignificance in Powell River and Campbell River for those years and Vancouver IslandWest in 1986. Campbell River also has a relative rate of 2.09 in 1984 when the teststatistic is zero.Soft Tissue SarcomaParts of the lower mainland area show signs of consistency for soft tissue sarcoma.Burnaby and Vancouver are significant for every year except 1983 as noted in Table 10.17. All cells found here were also obtained significance for the combined yeartest. Vancouver seems to be the main school districts which influences the lower main73land. The relative rates range from 1.6 in 1984 to 2.56 in 1987 with zero as both thetest statistic and consistency p-value. When Richmond is combined with Vancouverand Delta it is significant for 1985 and 1987—89 with a consistency p-value of zero aswell. Burnaby requires no combination of neighbours to be significant for 1984 with arelative rate of 2.10. For the years past 1984, it is significant when combined with NewWestminster and Vancouver. North Vancouver is also significant, in combination withVancouver, from 1985 with a consistency p-value calculated as zero. West Vancouvershows the same results when combined with Vancouver and North Vancouver. GrandForks also shows some clustering for 1987 and 1989 with relative rates 9.77 and 6.42,respectively. The cluster size is only two here and expected number of cases are lessthan one.MelanomaIn 1988 and 1989, the Sunshine Coast is a possible cluster for melanoma. In Table 10.18 Saanich also appears significant for 1984—86 with a consistency p-value of0.07. Saanich is significant by itself for the years 1984 and 1986 where the relativerates are 2.84 and 2.97, respectively. However, the relative rate of 2.15 in 1985 occurswhen Saanich is combined with Gulf Islands. Saanich also was significant when yearsare combined while Sunshine Coast was not. In 1988 Sunshine Coast and Nanaimoare a significant combination with relative rate 1.97. In 1987 the two along with WestVancouver have a relative rate of 1.88.ProstateThe Okanagan area figures prominently for cancer of the prostate. South Okanagan,Penticton, and Keremeos are significant for all years but 1984 as seen in Tables 10.20aand 10.20b. Most of the school districts listed have relative rates less than 1.5 for thesignificant years and will not be discussed. In 1985 and 1986 the test statistic is zero forSouth Okanagan and its relative rates are 1.55 and 1.86 for those years. In 1988 with the74influence of Keremeos the relative rate is 1.56. Penticton’s relative rates which exceed1.5 occur for 1986 and 1987 and are both close to 1.7. Keremeos is also significant forfor 1986 when South Okanagan’s cases are added to give a relative rate of 1.59. Thesethree school districts each have consistency p-values equal to zero. North Thompson hasa relative rate of 3.14 for the test statistic equal to zero. Cariboo-Chilcotin has relativerates 1.81, 1.59, and 1.60 for the years 1984, 1988, and 1989. Only in 1989 is the additionof Quesnel’s cases required for a significant result. Quesnel is also significant for 1984when combined with Cariboo-Chilcotin and Prince George yielding a relative rate closeto two. South Cariboo’s significant relative rate of 2.92 in 1985 causes Lillooet to also besignificant for that year. The two combined have a relative rate of 1.98 in 1986. Merritt’srelative rate of 1.5 in 1986 arises from combination with Kamloops. Abbotsford, NorthVancouver, and West Vancouver mostly have years with relative rates below 1.5. NorthVancouver has a relative rate slightly higher in 1988 while West Vancouver’s relative rateis 1.58 in 1986. Powell River’s three significant years occur by itself or with neighbourCourtenay. Central Coast has two years where the relative rates are 3.72 and 1.86. Thelatter year’s result occurs due to combination with Vancouver Island North and BurnsLake. Similar findings are shown for those years, 1987 and 1988, under Burns Lake.Nechako has several significant years with a consistency p-value equal to zero. The 1985and 1989 relative rates of 2.18 and 2.16 are when the test statistic is zero. The othersignificant years, 1984, 1987, and 1988, are significant because of the high relative ratein Prince George for those years. Similarly, Prince George is significant for 1989 whenNechako has a high relative rate. For the years 1984, 1987 and 1988 the relative ratesfor Prince George are 2.61, 2.01, and 1.43, respectively. Peace River South has a relativerate of 1.74 in 1985 and when combined with Peace River North and Prince George therelative rate is 2 in 1984 and 1.63 in 1987. The same type of results are seen for PeaceRiver North because of its closeness to Peace River South. Mission is significant by75itself with a relative rate of 1.79 in 1984. Summerland is significant by itself or whencombined with Penticton. Burns Lake, Nechako, and Prince George all appeared to bepossible clusters in the combined year analysis as well.TestisLangley was a possible cluster for the first three years of the study period for testicular cancer in Table 10.21. With a cluster size equal to five, the first two years giverelative rates 3.36 and 3.26. In 1985 Langley is combined with Surrey and Maple Ridgeyields a relative rate of 1.80. Maple Ridge’s relative rates above two are caused by thesignificance obtained in Langley for 1984—85. Powell River and Courtenay together havea relative rate for 1986 which equals 3.23. Four was the cluster size investigated in bothregions for that year. Powell River has a relative rate of 6.98 in 1984 when three is thecluster size. This result causes Courtenay also to be significant for that year. Campbell River’s results are similar to Courtenay’s and has a consistency p-value of 0.03.Vancouver Island North has a high relative rate in 1988 when the cluster size tested isthree. Courtenay, Powell River, and Vancouver Island North were each significant inthe combined analysis as well and have consistency p-values calculated to be 0.04, 0.02,and 0.04, respectively.BladderBladder cancer has shows some consistency in the Campbell River, North and WestVancouver Island as found in Table 10.22. For four of the study years Campbell Riveris significant with test statistic zero and cluster size five. The relative rates for thoseyears are as low as 2.72 and as high as 3.58. Vancouver Island West is significant for thesame years as neighbour Campbell River as well as in 1985. Together, with a cluster sizeequal to six, the two have a relative rate of 2.54. Vancouver Island North also showssimilar results due to neighbours Vancouver Island West and Campbell River with aconsistency p-value of zero. Lake Cowichan has two significant years when combined76with Cowichan or both Cowichan and Nanaimo. North Vancouver has two consecutivesignificant years with cluster size fourteen. South Cariboo is significant by itself in 1983and significant in 1988 when combined with Lillooet and Kamloops. Lake Cowichan andNorth Vancouver were not significant in the combined year tests while the remainingschool districts were.KidneyOnly Queen Charlotte and Kitimat are tabled for kidney cancer. In Table 10.23both are significant for small cluster sizes during the last two study years. When QueenCharlotte is combined with Prince Rupert and Kitimat it is significant for a cluster sizeof four. Kitimat has a relative rate of 7.16 when the cluster size is three in 1989. In1988 the combination of Kitimat, Terrace, and Prince Rupert have relative rate 2.80.Both of these districts also appeared as possible clusters in the combined year analysis.BrainNorth Thompson and Maple Ridge have two consecutive years for brain cancer.Maple Ridge was significant for this site in the combined year analysis and also appearsin Table 10.24. In 1988 the test statistic is zero and the relative rate is 3.34 for thecluster size of five. For 1985 and 1989 Maple Ridge is combined with Langley to besignificant with relative rates 2.17 and 2.62, respectively. North Thompson has a veryhigh relative rate for 1987. The cluster size investigated is only two and the expectednumber of cases are less than one. In 1986 it is also significant when Kamloops is added.Both cells listed here have consistency p-values equal to 0.06.Non-Hodgkins LymphomaFor Non-Hodgkins lymphoma, Courtenay has the most significant years, which canbe seen in Table 10.26. In 1987 and 1989 the test statistic is zero for this cell and therelative rates are 3.20 and 2.33. Courtenay is also significant for the first two studywhere Powell River or Powell River and Campbell River are added. Powell River is a77possible cluster itself for 1983 with a relative rate of 3.13 and 0.04 as the consistencyp-value. Powell River is also significant in 1987 when combined with Courtenay. Albernihas relative rates 2.88 and 1.80 for 1986 and 1987. In 1987 Alberni is combined withCourtenay and Qualicum to achieve its result. Richmond is significant for 1985-87 witha relative rate exceeding 1.5 only in 1987. Langley has two significant years, however,only 1989 has a relative rate (2.03) which is above 1.5. Surrey has two significant yearswhich have relative rates 1.55. In 1984 combination with New Westminster is required,while in 1988 Surrey is significant on its own.Multiple MyelomaMultiple myeloma testing only lists Prince George in Table 10.27. In both 1984 and1985 the cluster size investigated is four. In the former year the relative rate is 3.48and the test statistic is zero. The latter year has the combination of Prince Georgeand Nechako yielding a relative rate of 3.47. The consistency p-value is 0.07 and PrinceGeorge also appeared significant for the combined year situation.LeukemiaCowichan has two significant years for leukemia. In Table 10.28 for the clustersize of six, Cowichan is significant for 1983 and 1986 with relative rates 2.42 and 3.08,respectively. Lake Cowichan is affected by these results and shows similar numbers.However, Lake Cowichan had a consistency p-value of zero while Cowichan’s consistencyp-value is 0.04. Nanaimo also appears significant for 1987 and 1989. In the last yearthe test statistic is zero with relative rate 2.08. In 1987 the neighbours Qualicum andSunshine Coast must be added to Nanaimo to have at least fourteen cases. Nanaimowas not significant during the combined year tests although both Cowichan and LakeCowichan were.Acute LeukemiaTable 10.29 contains the acute leukemia results. The last three years are signifi78cant for Qualicum with relative rates 6.04, 2.38, and 3.32 for cluster sizes 3, 6, and5, respectively. Qualicum was significant for the case where years were combined aswell. Vancouver for 1987 and 1988 has relative rates about 1.76. These values effectNorth Vancouver’s results which are also significant for those two years. The CranbrookKimberley area is significant for the last two study years for chronic leukemia. Togetherthe two have a relative rate of 3.93 for 1988 in Table 10.30. The two also are significantfor 1989 where the cluster size investigated is still quite small. Keremeos has a veryhigh relative rate for a cluster size equal to two in 1983 In 1984 it, along with SouthOkanagan and Penticton, has a relative rate of 2.21. South Okanagan’s results are similar to that of Keremeos because of their close proximity. Lake Cowichan has only oneyear significant and the relative rate is 3.27. The school district was significant also forthe combined year tests. Greater Victoria has a relative rate of 1.64 in 1984. In 1985the relative rate is 1.55 when Sooke and Saanich are combined with Greater Vancouver.Sooke similarly shows that result and is also significant for 1984.Primary UnknownOnly Queen Charlotte and Prince Rupert were significant when years were combinedbut are not listed for primary unknown in Table 10.32. Powell River has the first twoyears significant with relative rates 2.76 and 2.08 respectively. The 1984 results occurwhen Powell River is combined with Courtenay. Quesnel in 1988 has a relative rateof 2.69 with five as the cluster size. The relative rate of 1.96 in 1987 happens whenCariboo-Chilcotin is added. The relative rate in 1985 occurs when Cariboo-Chilcotinand Prince George are considered together. Vancouver has a consistency p-value of zeroand is significant by itself for the years starting after 1984. The relative rates, however,are less than 1.5. Vancouver results affect Richmond, Burnaby, and North Vancouver.In 1988 a relative rate of 1.63 is seen for Richmond. Burnaby for the same year has itsonly relative rate above 1.5.79Cob-RectalCampbell River has a relative rate above 1.5 for cob-rectal cancer. Table 10.36shows that the relative rate is 1.92 in 1988. This result also causes Vancouver IslandWest to have a relative rate of 1.99 for the same year. Mission is significant with arelative rate above 1.5 for 1988 only. The relative rate is 1.89 for that year and theconsistency p-value is 0.47. Nanaimo is significant for three years with one year havinga relativer rate less than 1.5. In each of these years the test statistic is equal to zero.These results influence Qualicum which shows a similar pattern. Surrey’s significantyears have relative rates less than 1.3. Vernon has relative rates just slightly above 1.5for 1986—87.For all cancers except lung, the relative rates do not exceed 1.5. Vancouver, Northand West Vancouver, Keremeos, and South Okanagan are significant for several of theyears studied. Table 10.38 shows the results for all cancers except lung while Table 10.46chronicles the site all cancers. The same is basically seen in the all cancers site. OnlyMerritt has a year in which the relative rate exceeds 1.5. Most of the school districtshave relative rates which are not alarming.5.2.2 General Female FindingsLipLangley has two significant years for lip cancer. Table 11.1 shows that the years 1985and 1986 are significant with a consistency p-value equal to 0.08. In both years the teststatistic is zero requiring Langley to be combined with neighbours Surrey and MapleRidge. The cluster sizes are very small and produce the high relative rates 3.10 and4.52. For the other years this school district has fairly large values of the test statistic.In the combined year examination no school districts were found to be significant withrelative rates above 1.5.80Oral CavityAlberni was the only school district found significant for the combined year analysisof oral cavity cancer. It does not appear in the yearly examination given in Table 11.2.Surrey has three possible years of clustering. In 1983 and 1989 a zero test statisticis found with relative rates 12.17 and 1.80, respectively. In 1988 Surrey amalgamatedwith New Westminster and Langley produces a relative rate of 1.94 for cluster size13. Richmond, on the other hand, must be combined with Delta and Vancouver to besignificant for 1986—87. Only in 1986 though, does the relative rate exceed 1.5. Bothdistricts show a consistency p-value equal to 0.08.EsophagusLower mainland areas are listed in Table 11.3 for cancer of the esophagus. Whilenone were featured as significant for the combined year situation, the consistency pvalues are all at or below 0.04. Vancouver is significant by itself for 1985 and 1988with relative rates equal to 1.87 in both years. Vancouver’s other significant year is dueto North Vancouver’s high relative rate of 3.86 in 1989. Similarly, North Vancouver’ssignificance in 1988 is because of neighbour Vancouver. Vancouver’s value for thatyear also influences Burnaby which also needs to be considered with New Westminster.The relative rate is listed as 1.78. Richmond is also significant in 1985 and 1988 whencombined with Delta and Vancouver. Vancouver seems to be the cell which influencesthe others to also be significant.StomachStomach cancer is exhibited in Table 11.4. Richmond has a relative rate of 3.38 in1984 which causes neighbour Delta also to be significant for that year. Richmond andDelta together have a relative rate of 2.21 for the year 1989. Delta’s relative rate of 2.59in 1987 occurs when the cluster size is five. Both of these districts have consistency pvalues which are below 0.04 and were possible clusters in the combined year examination.81Cowichan is listed with a relative rate of 3.26 for 1987 and when considered with closecells Gulf Islands and Saanich has a relative rate of 2.18 for 1986.ColonOnly the Sunshine Coast is listed for colon cancer in Table 11.5. In the first twostudy years the relative rates are 2.54 and 2.40, respectively, with test statistic zero andcluster sizes at or near seven. The remaining years have a test statistic equal to four.Sunshine Coast was also tabled in the combined year situation and has a consistencyp-value of 0.36 in this yearly analysis.RectumTable 11.6 has Princeton and New Westminster listed for rectal cancer. Princetonwas significant in the combined year analysis and has two significant individual years.In 1985 and 1989 the relative rates are 7.27 and 8.07, respectively, when the cluster sizeinvestigated is three. New Westminster is a possible cluster for 1983, 1984, and 1988.The last two of these three years have zero as a test statistic and relative rates of two orabove. The first study year requires the combination of New Westminster and Burnabyto be significant with a relative rate of 1.57.LiverAlthough no cells were significant for liver cancer in the combined year investigation,Qualicum appears in Table 11.7. The two consecutive years 1986 and 1987 caused thissite to be listed. In 1987 the cluster size is very small and the resultant relative rateis 5.97. In the previous year, Nanaimo and Qualicum together have significance with arelative rate of 3.36 when the cluster size is four. The consistency p-value for this schooldistrict was evaluated to be 0.46.PancreasThe Castlegar-Trail area shows some consistency for pancreatic cancer and was alsosignificant in the combined year analysis. Trail 1986 and 1987 as significant years in82Table 11.8. In both periods the test statistic is zero and the relative rates are abovethree with very small cluster sizes. These results cause Castlegar also to be significantfor the same years. In addition, the two together have a relative rate of 3.12 in 1984when the cluster size examined is four. The consistency p-values are 0.02 for Castlegarand 0.03 for Trail.LarynxLarynx cancer has five districts listed in Table 11.9. In the combined year framework,Greater Victoria was also found to be significant. The yearly investigation shows that1983 and 1989 have relative rates 3.73 and 2.96, respectively. In the first year, the clustersize is four while in the last year the cluster size is a bit larger at six. Cariboo-Chilcotinhas a high relative rate of 11.79 for 1984. Here, however, the cluster size is only twoand less than one case is expected. Similarly high relative rates are seen in Kamloopsfor 1985—86 and 1988. In all these years the cluster size investigated is two. The highrates in Kamloops cause neighbours North Thompson and South Cariboo also to besignificant for the same years. In 1986 Kamloops must be combine with South Cariboo,Merritt, and Shuswap to be significant. The consistency p-values are 0.03 for all cellsexcept South Cariboo. South Cariboo has a consistency p-value equal to 0.04.Lung SquamousAll districts listed for lung squamous in Table 11.10 have consistency p-values whichare zero or 0.01. Courtenay, Campbell River, Sunshine Coast all have four significantyears. Sunshine Coast is significant itself in 1985 for the cluster size three. The years1984, 1987, and 1988 have relative rates 2.78, 4.86, and 2.46, respectively, when SunshineCoast is combined with Nanaimo. Nanaimo’s test statistics for all of those years is zerowith relative rates exceeding 2.7. Qualicum is influenced by Nanaimo and is significantfor the years 1987—88. The relative rates are 2.27 and 2.40 for those years, respectively.Courtenay is significant itself for 1983 and 1987 where the cluster size is four. The high83relative rates cause Powell River to obtain significance for the same years. Courtenayand Powell River together have a relative rate of 3.13 in 1985 which causes CampbellRiver to be significant with a relative rate of 2.75 in that year as well. Powell River’srelative rate of 5.31 for cluster size three in 1984 also produces a significant year forCourtenay and Campbell River. Campbell River’s relative rate of 4.68 when three isthe cluster size influences Vancouver Island West and Powell River to also be significantin 1983. Vancouver Island West also has a significant year in 1987 when considered alongwith Courtenay and Campbell River. All of these districts were also possible clustersfor the combined year tests.Lung AdenocarcinomaTable 11.12 has the significant cells for lung adenocarcinoma cancer. Surrey hasthe lowest consistency p-value which is 0.03. In 1985 it is significant with relative rate1.96 and test statistic zero. The other two significant years found have relative rateslower than 1.5. New Westminster has a relative rate equal to 2.36 in 1987 when thetest statistic is zero and the cluster size is eight. This induces Coquitlam also to besignificant for that year. In 1984 the combination of New Westminster and neighboursBurnaby and Coquitlam produces a relative rate of 1.58. Burnaby’s only significantyear where the relative rate is above 1.5 is 1988. The cluster size investigated is 18and the corresponding relative rate is 1.56. This result influences Coquitlam and NewWestminster which produces relative rates smaller than 1.5. Coquitlam is also significantitself for 1984 with relative rate 1.90.Lung Small CellLake Cowichan has a very high relative rate for lung small cell cancer in 1986.Table 11.13 shows this rate to be 16.50 for a cluster size of two. In this situation, theexpected number of cases is extremely small. This school district is also significantfor the following year when considered with Cowichan. The cluster size there is a bit84larger at five and the relative rate is 3.46. Powell River is significant on its own forthe last study year with cluster size three. In 1987 the addition of Courtenay causes arelative rate of 2.93. When Campbell River is combined with Courtenay in the sameyear the relative rate is 2.77. These two school districts and Vancouver Island Westare also a possible cluster for that year with a relative rate of 2.73. Powell River hasthe lowest consistency p-value of 0.01. Lake Cowichan and Vancouver Island West bothhave consistency p-values equal to 0.03. All of these districts were also significant whenyears were combined.Lung OthersLung others is featured in Table 11.14. Peace River North has a consistency p-valueof 0.04 without one individual significant year. Peace River South itself is significant for1986 when the cluster size is three. Prince George has relative rate 2.91 with at leastfive observed cases for 1987 with a consistency p-value equal to 0.04. Sunshine Coasthas two consecutive years with cluster sizes less than five. In 1984 the relative rate is3.96 while in 1985 the relative rate is 3.72. All cells except for Prince George were alsofound in the combined year results.All LungsPowell River has the most significant years for all lungs. In Table 11.15 Powell Riveris significant by itself in 1984 and 1989 with relative rates 2.70 and 2.20, respectively.For the years 1983 and 1987 the relative rates are slightly above 1.5 when it is combinedwith Courtenay and Campbell River. For those years, Courtenay and Campbell Riverare each significant. Courtenay is significant by itself in 1984 and 1987 with relativerates 1.86 and 2.06, respectively. These values influence Campbell River which is alsosignificant for that time period. Vancouver Island West is significant for the years whereCourtenay and Campbell River are combined with it (1983,1984 and 1987). SunshineCoast has two consecutive significant years. In 1984 the relative rate is 2.34 while the85relative rate in 1985 is 2.63 where the cluster sizes are both eight. Qualicum is significantfor a couple of years in which the relative rates do not surpass 1.5. Cowichan has thelast two study years as possible clusters and a consistency p-value equal to 0.40. Therelative rate for both of these years is around 1.63 and the cluster sizes investigated arefifteen or sixteen.Non-small Cell LungNon-small cell lung cancer is featured in Table 11.16. Alberni has two years inwhich its test statistic was zero and the relative rates exceed two. Cowichan has thesame situation. In every year prior to 1988 the test statistic is five and Cowichan isnot significant. With zero as the test statistic in the last two years the relative ratesare around 1.7. Sunshine Coast also shows the same feature when the test statisticis zero for 1984—85. In those years the relative rates are 2.52 and 3.24 with at leastseven cases observed. Cariboo-Chilcotin is significant by itself in 1985, in 1986 whencombined with Quesnel and North Thompson, and in 1989 when amalgamated withQuesnel. Courtenay has a relative rate of 2.12 in 1984 and 2.22 in 1984. CampbellRiver is influenced by Courtenay and has significant years four 1983—84 and 1987. Therelative rate range from 1.65 to 1.90. Vancouver Island West also shows the same patternwhen combined with both Courtenay and Campbell River.MelanomaWest Vancouver and Howe Sound have the most significant years for melanoma. WestVancouver is significant for every year except 1984 and 1989 as depicted in Table 11.18.For 1983, 1985, and 1986 the relative rates are 2.90, 2.84, and 2.73, respectively. In 1988—89, it must be combined with North Vancouver to obtain significance. The consistency pvalue is zero for this school district. Howe Sound is significant when combined with WestVancouver for those same years. It is also a possible cluster in 1987 when consideredwith West Vancouver and North Vancouver with a relative rate of 1.56. Burnaby is86significant by itself for 1984 with a relative rate equal to 1.72 and cluster size 13. Whencombined with New Westminster in 1983, it is also significant with relative rate 1.67.Langley is significant for 1984—86 when the test statistic is zero. The relative rates thererange from 2.02 to 2.64 with cluster sizes just under ten. The consistency p-value forLangley is calculated to be 0.06 while the Sunshine Coast has a consistency p-valueof 0.44. Sunshine Coast is significant when the cluster size is four in 1986 and in theprevious year when influenced by West Vancouver and Nanaimo. Sunshine Coast andBurnaby were not featured in the combined year examination, however, the remainingdistricts were also significant for that situation.BladderOnly West Vancouver is tabled for bladder cancer, with three significant years. InTable 11.22 the consistency p-value for West Vancouver is calculated to be 0.03. WestVancouver was also found to be significant when years were considered together. In 1985the relative rate is 3.64 with cluster size five. For 1987 and 1989 West Vancouver mustbe combined with North Vancouver to give relative rates of 2.06 and 1.92, respectively.KidneyQueen Charlotte, Creston-Kaslo, and Stikine are listed for kidney in Table 11.23.In 1987, Queen Charlotte has a relative rate of 21.04. Since £ 0 and the cluster sizeinvestigated is only two, a minimum of two cases are found in Queen Charlotte. Whenthe expected number of cases are less than one the relative rate becomes quite extreme.In 1984 the same school district is also significant at the same small cluster size whencombined with Prince Rupert. Creston-Kaslo for the last study year has a relative rateof 5.24 and cluster size two. In 1986, Creston-Kaslo along with Nelson, Cranbrook, andFernie, has a relative rate of 2.54 and is significant. Stikine is significant for two yearswhen combined with Fort Nelson and Smithers. Again, the cluster size is two and veryfew cases are actually observed. The first two districts listed have consistency p-values87equal to 0.03 while Stikine’s is 0.05.Non-Hodgkins LymphomaNon-Hodgkins lymphoma findings are seen in Table 11.26. Langley is significant forthe first two years with relative rates 2.61 and 2.40. Delta is a possible cluster itselffor 1984 and in 1983 and 1989 when merged with Richmond and New Westminster.Nechako has a relative rate of 4.70 in 1989 with small cluster size. In 1985 it is alsosignificant when considered with Prince George and Quesnel.Multiple MyelomaFor multiple myeloma in Table 11.27, Burns Lake, Nechako, and the Peace Riverdistricts are significant in the first two study years. Nechako has a very high relativerate and small cluster size for 1984. This influences Burns Lake which is also significantfor that year. Nechako also has a high relative rate for the first study year when combinedwith Prince George. The extreme rates in Burns Lake and Nechako also cause PeaceRiver South and North to be significant with high relative rates. The cluster sizes arevery small here and very few cases are actually observed. South Okanagan and Keremeosare significant for 1988 with a relative rate of 3.84 when combined along with Penticton.Summerland has a couple of years where the relative rates are high when combined withPenticton. South Okanagan, Keremeos, and Summerland all have consistency p-valueswhich are equal to 0.05. Summerland, Nechako, South Okanagan, and Keremeos wereeach significant in the combined year investigation and the cluster sizes tested in theyearly analysis are very small.LeukemiaVancouver Island North has 1987 and 1988 as significant years for leukemia. It isthe only school district listed in Table 11.28. In 1987 it is significant by itself and in1988 when combined with Vancouver Island West and Campbell River. With less thanone case expected in 1987 the relative rate is 7.21.88Acute LeukemiaSunshine Coast and Nanaimo are both listed for acute leukemia. In Table 11.29 wesee Nanaimo is significant for the first two years with test statistic equal to zero andcluster size four. The high rates found for that school district also cause its neighbour tobe significant for the same year. Neither Sunshine Coast nor Nanaimo have consistencyp-values less than 0.05.Other SitesNorth Vancouver has the most significant years for other sites with 0.00 as its consistency p-value. Five years are significant for this district found in Table 11.31. Forthe years in which the test statistic is zero, the relative rates exceed 1.5. West Vancouver shows similar results when combined with North Vancouver. In 1985, 1986, and1988 Grand Forks has relative rates which exceed 3.5. In 1985, Grand Forks must beconsidered with Kettle Valley while in 1989, the relative rate is 2.15 when Kettle Valleyand Trail are amalgamated with it. Nechako is significant for 1984—85 when consideredby itself or combined with Prince George. The relative rates during these years are4.20 and 2.39 with fairly small cluster sizes. Nanaimo is significant with at least elevencases observed in 1986. This cell also has a relative rate of 1.70 when merged withneighbouring Qualicum and Sunshine Coast.Primary UnknownVancouver has three years of possible clustering for the primary unknown site displayed in Table 11.32. All years, however, have relative rates less than 1.5. Maple Ridgeis significant by itself with relative rate 2.52 in 1984. When it is combined in 1985 withLangley the relative rate is 1.81. Mission is only significant when the test statistic is zerofor the years 1987—88. The relative rates are 2.83 and 2.67 for those years, respectively.BreastIn Table 11.33 the significant school districts for breast cancer are displayed. Only89West Vancouver has a year which is significant and has a relative rate above 1.5. In 1984with a cluster size of 37 the relative rate is 1.53. None of the cells have a consistencyp-value which is less than 0.10.CervixFor cervical cancer in Table 11.35, Vancouver has rates which seem to influence itsneighbours. Five years are significant for Vancouver although oniy two have rates above1.5. The 1989 result causes Richmond to be significant with a relative rate of 1.70. Theother years show small relative rates. The same can be said for Burnaby and NorthVancouver which have relative rates for 1989 equal to 1.73 and 1.67, respectively. Theconsistency p-values in these school districts are close to zero. Surrey has a relative rateof 1.87 in 1988 and 1.51 in 1987. The last value occurs when Surrey is combined withNew Westminster. The Central Coast has excessive relative rates with small cluster sizesfor 1985—86. Alberni in combination with Qualiciim is significant for 1987 and 1988.The Richmond, Vancouver, Burnaby, and North Vancouver regions all have consistencyp-values close to zero. Central Coast is significant for 1985 and 1986. With test statisticzero in 1986 and cluster size 2, the relative rate is 22.71. only a fraction of one case isexpected here, so observing at least two cases gives the extremely high relative rate. In1985, Central Coast must be combined with Vancouver Island North and have relativerate 5.31 when combined.Colo-.RectalNew Westminster, Sunshine Coast and Kitimat have three significant years each forcob-rectal cancer. Sunshine Coast was present in the combined year investigation andis also listed in Table 11.36. Sunshine Coast is significant for years in which the teststatistic equals zero. The relative rates seen are close to two. Cariboo-Chilcotin hasrelative rates equal to 1.99 for two consecutive years with a consistency p-value of 0.22.In 1985 Hope is significant with relative rate 2.93 when Agassiz-Harrison is added to it.90The addition of Agassiz-Harrison and Chilliwack yield a relative rate of 1.71 for Hopein 1987. New Westminster’s significance in 1984 gives a relative rate of 1.73 while theother years are less than 1.5. Kitimat is significant for three years when considered withTerrace or both Terrace and Prince Rupert. The consistency p-value for Kitimat is 0.09.EndometriumFor the endometrium site found in Table 11.39, West Vancouver has the most significant years. The years 1984 and 1987 have relative rates above 1.5 for Richmond with0.06 as the consistency p-value. West Vancouver has a consistency p-value of 0.01 andhas relative rates above two when the test statistic equals zero. Howe Sound is onlysignificant if combined with neighbour West Vancouver. Prince George is a possiblecluster for the years 1983, 1985, and 1987 with relative rates 2.15, 1.97, and 2.22. GulfIslands is only significant if considered alone or with neighbour Cowichari for 1985 and1986.When all cancers except lung are considered, Kettle Valley has a high relative ratefor 1983 and 1984. In Table 11.38, Kettle Valley has relative rates 2.73 and 2.43 forthose two years. With a cluster size of 9 investigated in each of these two years, therelative rate is large. The remaining districts displayed have relative rates below 1.5.The same can be said for the site all cancers. Kettle Valley has large relative rates forthe first two study years. In Table 11.46 Hope has a relative rate of 1.6 for 1984 and isalso significant for 1983.Overall the results given by the yearly analysis are fairly consistent with those seenin the combined year analysis. If a school district is very significant over all years, it islikely to be significant for at least one year in particular. With school districts whichare not as significant, they may not appear significant in any one year. This fact ismainly because of the discreteness of the Poisson distribution. If the expected numberof cases is small, the cluster size may be more than the 95 percentile. For example, if91= 2 then the approximate 95’ percentile of the Poisson distribution is 6. However, 6is almost exactly the 98’ percentile of that Poisson distribution. With low numbers ofcases expected, the yearly results may not be consistent with the combined year analysisfor a particular cell. Another reason for lack of consistency can be based on the numberof cells combined. Suppose a cell is significant in the combined year analysis with anobserved test statistic of 1. Perhaps in the yearly framework the observed test statisticsfor that cell are all 2. The yearly and combined year analyses will not be examining thesame thing because the populations involved are different.It should also be noted that some of these results showed very high relative rates.The relative rates should be considered along with the cluster size. Small cluster sizesmean that few cases are required to have significance achieved. Small values of \ needsmaller values of the cluster size to be able to detect possible clusters at a specifiedsignificance. A relative rate of ten would mean very different things depending on thecluster size and underlying population at risk. For example, if one cell has five casesexpected and another cell has only 0.1 case expected the relative rate of ten would havea different interpretations. In the first case it would be extremely alarming to have 50cases observed when oniy five are expected. In the other cell, expecting 0.1 cases andobserving one case would give a relative rate of ten but not be so alarming.5.2.3 Focused FindingsThe focused yearly results for males start on page 165. These findings display someagreement with the with the general yearly results when £ is small. When £ is larger,the nearest neighbours of the mill are not exactly the same as the nearest neighbours ofthe school district where the mill is located.From the male results we see that Port Mellon has several years significant for thesites oral cavity, liver, soft tissue sarcoma, and prostate. Port Mellon also shows this92pattern in some other sites, although the relative rate is not above 1.5 in those years.Campbell River is listed for several sites including all lungs, krngs:non-small cells, testis,and bladder. For the sites of all lungs, lungs:non-small cells, bladder, Gold River shows apattern of high relative rates. Powell River figures prominently for all lungs, lungs:nonsmall cells, testis, and non-Hodgkins lymphoma. Prince George has three significantyears for lung others. MacKenzie shows similar results but not exactly the same. Withlarger values of the statistic, the neighbours are not identical for these two mills. Similarresults are also seen for prostate and multiple myeloma. Those mills listed under rectumcancer are all significant for 1988 and 1989.When the female results are considered we see that Port Mellon also appears severaltimes. For oral cavity, colon, melanoma, cob-rectal, and endometrium it has severalsignificant years, and two consecutive significant years for some sites. Powell River,Gold River, and Campbell River are displayed for lung squamous, lung small cell, andall lungs. The last two also appear together for lung:non-small cells. Squamish has foursignificant years for melanoma.936 ConclusionThe objective of a large scale surveillance scheme for potential disease clusters is toprovide information which can be used to prioritize areas or foci which require furtherstudy. Surveillance studies are intended to indicate where local studies should be conducted. The findings from these studies can potentially provide guidelines for allocatinglimited investigation resources to areas most likely to be actual clusters.Determining which school districts and mills are possible clusters in British Columbiawas the goal of this surveillance study. The diversity of school district population sizesforced the cluster detection technique employed to incorporate the population distribution within each cell. After examination of both general and focused tests considerateof the population distribution, the method proposed by Besag and Newell was deemedmost appropriate. Concerns about the choice of cluster size in our analysis motivateda modification to the Besag and Newell method. The modified method was used forboth general and focused tests, and for both combined year and yearly analyses. Thecombined year examination identified mills or school districts with statistically significant excess numbers of cases. To investigate if cells were consistently part of a possiblecluster throughout the study years, a yearly analysis was conducted as well.The Besag and Newell method seemed the most suitable choice due to several factors.Their procedure did not require the amalgamation of partial cells. Trying to split theschool districts into regions of equal population was unacceptable with large, sparselypopulated regions. Certainly in school districts such as Fort Nelson, a third of thepopulation did not reside in a third of the geographical area. Neither was exposure datarequired to determine the clustering around a pollutant source, nor did the existing cellsneed to be redefined to incorporate the focused test. Only the disease counts, populationdata, and nearest neighbours for each administrative zone were required for the analysis.These aspects made their method suitable for our situation.94The cluster size which should be used was not entirely clear. When cancer countsvary dramatically from cell to cell, the cluster size chosen becomes critical. When zonesare very small and disease counts are also small, this choice is not as paramount. Forexample, when disease counts within a cell are either zero or one, to achieve a certaincluster size k, the test statistic will be at least k. However, when combining regions withhigher disease counts, adding one more cell can drastically increase the number of casesobserved. With the disparity of population among B.C. school districts, the cluster sizechosen had to reflect the possibility of large jumps in the cases observed when the teststatistic increased slightly. It was readily apparent that the diverse populations requireddiverse cluster sizes. Using one cluster size for all cells was not reasonable. Each cellneeded its own cluster size dependent on its population. The cluster size for a cell was afunction of its population, or equivalently, a function of its estimated expected numberof cases.Our proposed modification used percentiles of the Poisson distribution based onestimated expected number of cases. This modification guaranteed all statistically significant cells or foci will be identified for some prespecified level of significance. Withoutour alteration, a cell could be tested at two different cluster sizes and produce conflictingresults even if the observed test statistic was exactly the same in both situations. Thistype of incongruency motivated us to choose a cluster size evaluated as the minimumnumber of cases necessary to be statistically significant.Each cell was tested at a maximum of three different cluster sizes. Each of these cluster sizes was the 95 percentile of a Poisson distribution with a specific mean adjustedfor age distribution in the underlying population. The three means were the estimatedexpected number of cases for a cell individually, the cell and its nearest neighbour, andthe cell and its two closest neighbours. Our cluster sizes represent the minimum numberof cases required to obtain significant results based on the estimated expected number95of cases for the corresponding region.The modified procedure was implemented on each sex and cancer type separatelywith age group as a stratum. Combined year analyses identified possible clusters for theentire study period. When the modified method was applied to each year separately,the yearly analyses showed if a cell was consistently part of a cluster. A Monte Carlosimulation was conducted to ascertain the degree of overall clustering for the combinedyear analyses, while in the yearly examination, yearly p-values were transformed toreflect the degree of consistency of yearly results.The findings of this study were discussed in detail in Section 5. With various sites,sexes, and analyses, the results are difficult to summarize briefly and listing each significant cell would be repeating that discussion. We offer only broad comments aboutsome sites.Several sites did not have any significant areas with at least a relative rate of 1.5.In males, the sites for which clustering evidence was not supported by the general testare colon, lung adenocarcinoma, all cancers except lung and all cancers. No apparentclustering around mills was indicated by the male focused test in those sites and thesites of rectum, liver, all lungs, lungs:non-small cells, soft tissue sarcoma, melanoma,brain, Hodgkins Disease, non-Hodgkins lymphoma, and cob-rectal. Lip, liver, all lungs,lungs: non-small cells, soft tissue sarcoma, breast, all cancers except lung, and allcancers sites did not have any possible clusters listed for the combined year generalfemale analysis. The focused combined year female results did not suggest clusteringaround the foci for those same sites, with the exception of lungs:non-small cells, and thesites of esophagus, rectum, lung adenosquamous, lung adenocarcinoma, bladder, kidney,brain, non-Hodgkins lymphoma, chronic leukemia, primary unknown, and ovary. Yearlyresults showed some school districts and mills were consistently part of possible clustersfor a few sites.96Several sites showed overall significance in the Monte Carlo simulation. In males,the sites of lip, stomach, lung squamous, lung others, all lungs, lungs: non-small cells,soft tissue sarcoma, prostate, bladder, kidney, and multiple myeloma suggested overallclustering for the general test. Overall clustering around mills for stomach, pancreas,lung squamous, lung others, prostate, bladder, kidney, and multiple myeloma cancerswas supported by the focused test. Overall clustering was not as apparent when womenare considered. The general test displayed lung squamous, lung adenosquamous, lungsmall cell, chronic leukemia, and primary unknown sites with low overall p-values. Thosesame lung sites as well as melanoma and endometrial cancer also were indicative ofoverall clustering for the focused test.This surveillance study involved over thirty cancer sites, two sexes, and seven yearsof data. Utilization of the modified method identified all cells which were statisticallysignificant. Not all school districts or mills identified can be interpreted as actual clustersdue to multiple testing problem. Any apparent excess of cases may be due to factorswhich were not known to us. Although the age structure was considered, people withinan age group were assumed to be homogeneous within that group. We assume, forexample, that the women between the ages of 20—24 in Surrey smoke as much or aslittle as the women in that age group living in Stikine. This may not necessarily be thecase. Covariates such as smoking, drinking, diet, occupation, or family history may befactors which contribute to a cluster being detected in a particular school district. Thepossible clusters detected would be actual clusters if everyone within a stratum, wassimilar from cell to cell. People with similar ages and of the same sex were consideredtogether irrespective of factors which may make them more susceptible to cancer thansomeone else in their strata.Causal relationships between cancer incidences and pulpmills should not be established by our findings. Nor can we say that someone is more likely to get a specific cancer97due to their locations of residence. Even though strong associations between locationsof residence and elevated cancer risks may be established, cancer has a long latencyand people frequently change their locations of residence, making such statements inappropriate. Every significant cell should not be interpreted as an actual cluster. All wecan conclude is that some areas have statistically significant excessive numbers of casesduring the study period which cannot be attributed to age group or sex distributions.The health authority must assess, given its current knowledge on the etiologies ofthese diseases, which school districts or mills and sites require further study and whatform this further study will take. Smaller administrative zones, such as enumerationareas, may also be tested for clustering. The focused test may be more informativeif some environmental factors are incorporated in the nearest neighbour relationship.Using wind patterns to determine the nearest neighbours to a pulpmill may be morereasonable than distance alone, for example. Mortality studies could be performed onthe school districts which were significant followed by in-depth epidemiological studies,if necessary. The cluster detection analysis conducted here is a preliminary step towardsidentifying cancer clusters.98References[1] Besag, J. and Newell, J. (1991). The detection of clusters in rare diseases. Journalof the Royal Statistical Society A 154, 143-155.[2] Cox, D.R. and Hinkley, D.V. (1974). Theoretical Statistics. London: Chapman andHall.[3] Fisher, R.A. (1935). The Design of Experiments. Edinburgh, Scotland: Oliver &Boyd.[4] Knox, G. (1964). Epidemiology of childhood leukemia in Northumberland andDurham. British Journal of Preventive and Social Medicine, 18, 17-24.[5} Mantel, N. (1967). The detection of disease clustering and a generalized regressionapproach. Cancer Res. 27, 209 — 20.[6] Openshaw, S., Craft, A.W., Chariton, M., and Birch, J.M. (1988). Investigation ofleukemia clusters by use of a geographical analysis machine. Lancet, 272-273.[7] Pinkel, D. and Nefzger, D. (1959). Some epidemiological features of childhoodleukemia in the Buffalo, N.Y., area. Cancer 12, 241-8.[8] Stone, R.A. (1988). Investigations of excess environmental risks around putativesources: statistical problems and a proposed test. Statistics in Medicine 7, 649-660.[9] Turnbull, B.W., Iwano, E.J., Burnett, W.S., Howe, H.L., and Clark, L.C. (1990).Monitoring for clustering of disease: application to leukemia incidence in UpstateNew York. American Journal of Epidemiology 132, supplement, S136-S143.[10] Wailer, L.A., Turnbull, B.W., Clark, L.C., and Nasca, P. (1992). Chronic disease surveillance and testing of clustering of disease and exposure: application to99leukemia incidence and TCE-contaminated dumpsites in Upstate New York. Environmetrics 3, 281-300.[11] Wailer, L.A., Turnbull, B.W., Clark, L.C., and Nasca, P. (1993). Examining SpatialPatterns of Incidence Data to Detect Clusters in a Rare Disease. In Case Studiesin Biometry. N. Lange and L. Ryan (eds)., New York: John Wiley & Sons.[12] Wilkins, R. (1993) Geocodes/FCCF User’s Guide. Automated Geographic CodingBased on the Statistics Canada Postal Code Conversion File. Ottawa: CanadianCentre for Health Information, Statistics Canada.[13] Whittemore, A.S., Friend, N., Brown, B.W., Holly, E.A. (1987). A test to detectclusters of disease. Biometrika 74, 631-635.[14] Cox, D.R. and Hinkley, D.V. (1974). Theoretical Statistics. London: Chapman andHall.100Appendix: TablesTable 1: Cancer Sites Investigated with Patient CountsSite Number Cancer Site Males FemalesI Lip 14U2 Oral Cavity 141 — 9 1167 6043 Esophagus 150 550 2614 Stomach 151 1419 7535 Colon 153 3123 32286 Rectum 154 2283 17387 Liver 155 396 2178 Pancreas 157 952 8559 Larynx 161 653 10710 Lung Squamous 2368 71311 Lung Adenosquamous 73 3012 Lung Adenocarcinoma 1647 129813 Lung Small Cell 1038 63414 Lung Others 2219 112615 All Lungs 7345 380116 Lungs: non-small cells 6307 316717 Soft Tissue Sarcoma 171 605 33618 Melanoma 172 1210 128320 Prostate 185 9613 —21 Testis 186 494 —22 Bladder 188 1860 60623 Kidney 189 1085 57824 Brain 191 747 50125 Hodgkins Disease 201 289 22026 Non-Hodgkins Lymphoma 202 1506 117827 Multiple Myeloma 203 503 41128 Leukemia 204,5,7, 10, 11 1231 84929 Acute Leukemia 204,5 350 26730 Chronic Leukemia 207, 10, 11 881 58231 Other Sites 1371 177332 Primary Unknown 199 2552 206833 Breast (female) — 1081034 Ovary — 164235 Cervix — 108536 Cob-Rectal (4,5) 5406 496638 All cancers except lung 33937 3329539 Endometrium — 212746 All cancers (ex. non-melanoma skin) 41282 37096101Table 2: British Columbia PulpmillsCode Company & Division Location School District1 Canadian Forest Products Ltd.Prince George Pulp & Paper Mill Prince George 572 Canadian Forest Products Ltd.Intercontinental Pulp & Paper Prince George 573 Celgar Pulp Co. Castlegar 94 Eurocan Pulp & Paper Co. Kitimat 805 Fletcher Challenge Canada Ltd.Crofton Division Crofton 656 Howe Sound Pulp & Paper Ltd. Port Mellon 467 MacMillan Bloedel Ltd.Alberni Pulp & Paper Division Port Alberni 708 MacMillan Bloedel Ltd.Harmac Division Nanaimo 689 MacMillan Bloedel Ltd.Powell River Division Powell River 4710 Northwood Pulp & Timber Ltd. Prince George 5711 Skeena Cellulose Inc.Skeena Pulp Operations Prince Rupert 5212 Wester Pulp Limited PartnershipPort Alice Operation Port Alice 8513 Wester Pulp Limited PartnershipSquamish Operation Squamish 4814 Wyerhauser Canada Ltd. Kamloops 2415 Canadian Pacific Forest Prod. LtdGold River Mill Gold River 8416 Cariboo Pulp & Paper Co. Quesnel 2817 Crestbook Forest Industries Ltd. Skookumchuck 218 Crown Zellerback Ltd.Extinct 78 Ocean Falls 4919 Finlay Forest Industries Ltd. Mackenzie 5720 Fletcher Challenge Canada Ltd.Elk Falls Mill Campbell River 7221 Fletcher Challenge Canada Ltd.Mackenzie Pulp Division Mackenzie 5722 Quesnel River Pulp Co. Quesnel 28102Table 3: Nearest School District CentroidsNearest SD Nearest SDCode Name 1st 2nd 3rd Code Name 1st 2nd 3rd1 Fernie 2 3 86 45 W Vancouver 44 39 412 Cranbrook 3 1 86 46 Sunshine Coast 68 45 693 Kimberley 2 1 86 47 Powell River 71 72 694 Windermere 3 18 2 48 Howe Sound 45 44 467 Nelson 9 11 10 49 Central Coast 85 55 809 Castlegar 11 7 12 50 Queen Charlotte 52 80 8810 Arrow Lakes 7 9 11 52 Prince Rupert 80 88 5011 Trail 9 7 12 54 Smithers 88 55 8012 Grand Forks 13 11 9 55 Burns Lake 56 54 8813 Kettle Valley 12 14 11 56 Nechako 57 55 2814 S Okanagan 16 15 77 57 Prince George 56 28 5515 Penticton 77 14 16 59 Peace River 5 60 57 5616 Keremeos 14 15 77 60 Peace River N 59 57 8117 Princeton 16 77 15 61 Greater Victoria 62 63 6418 Golden 19 4 10 62 Sooke 61 63 6519 Revelstoke 89 18 21 63 Saanich 64 62 6121 Armstrong- 22 89 23 64 Gulf Islands 65 63 37Spallumcheen22 Vernon 21 23 89 65 Cowichan 64 63 6623 Central Okanagan 77 15 22 66 Lake Cowichan 65 68 6424 Kamloops 31 30 89 68 Nanaimo 69 46 6626 N Thompson 24 89 19 69 Qualicum 68 70 4627 Cariboo-Chilcotin 28 26 30 70 Alberni 69 71 6828 Quesnel 27 57 56 71 Courtenay 47 72 7029 Lillooet 30 31 24 72 Campbell River 71 47 8430 S Cariboo 29 24 31 75 Mission 34 42 3531 Merritt 24 17 30 76 Agassiz-Harrison 33 32 7532 Hope 76 33 17 77 Summerland 15 23 1633 Chilliwack 76 75 34 80 Kitimat 88 52 5434 Abbotsford 75 35 33 81 Fort Nelson 60 59 8735 Langley 36 42 40 84 Van Isl West 72 71 7036 Surrey 40 35 43 85 Van Isi North 84 72 7137 Delta 38 40 36 86 Creston-Kaslo 7 2 338 Richmond 37 39 41 87 Stikine 81 54 8839 Vancouver 44 41 45 88 Terrace 80 54 5240 New Westminster 41 43 36 89 Shuswap 21 22 2441 Burnaby 40 39 4342 Maple Ridge 35 43 3643 Coquitlam 40 41 3644 N Vancouver 39 45 41103Table 4: Nearest School District Centroids to FociNearest SD’sCode Mill Location 1st 2nd 3rd 4th 5th1,2,10 Prince George 57 56 28 55 273 Castlegar 9 11 7 12 134 Kitimat 80 88 52 54 555 Crofton 65 64 63 66 686 Port Mellon 46 45 39 44 387 Port Alberni 70 69 71 68 668 Nanaimo 68 69 46 66 659 Powell River 47 71 72 69 7011 Prince Rupert 52 80 88 50 5412 Port Alice 85 84 72 71 4913 Squamish 48 45 44 46 3914 Kamloops 24 31 89 30 2115 Gold River 84 72 71 85 4716,22 Quesnel 28 27 57 56 2617 Skookumchuck 3 2 1 4 8618 Ocean Falls 49 80 88 52 8519,21 MacKenzie 57 56 60 59 5520 Campbell River 72 71 47 84 70104Table 5a: Combined Year Male Results for Site 1 Lip, k1,.. . ,Cluster Size (k)SD 2 10 18 26 34 42 50 58 66 74T 0.79 0.12 0.01 0.01 0.02 0.24 0.14 0.06 0.01 0.04£ 1 3 5 10 13 15 16 17 18 22A 2.95 6.47 9.70 16.08 22.83 37.22 42.38 46.79 49.53 59.300.53 0.12 0.01 0.01 0.02 0.24 0.14 0.04 0.02 0.04£ 0 3 4 9 13 15 16 17 18 22A 1.78 6.47 9.06 15.58 22.83 37.22 42.38 45.12 49.53 59.30W 0.29 0.12 0.01 DM1 0.02 0.24 0.11 0.06 0.02 0.04£ 0 3 5 9 13 15 16 17 18 22A 1.08 6.47 9.70 15.58 22.83 37.22 41.63 46.79 49.53 59.30T 0.51 0.01 0.03 0.16 0.10 0.01 0.11 0.06 DM3 0.03£ 1 3 7 12 13 14 16 17 19 21A 1.72 4.01 10.82 20.94 26.53 27.70 41.63 46.77 51.20 58.41T T 0.39 0.27 0.04 DM1 0.00 0.00 0.01 0.02 0.03 0.0466£ 0 3 6 8 9 10 11 13 19 23A 1.34 7.85 11.35 15.17 16.95 22.11 35.61 42.86 51.42 60.19TT 0.92 0.27 0.03 0.01 0.00 0.00 0.03 0.02 0.03 DM5£ 1 3 5 8 9 10 12 14 19 23A 4.09 7.89 10.73 15.17 20.32 22.11 37.27 43.58 51.23 60.1912 T 0.33 0.06 0.14 DM1 0.39 0.14 0.09 0.OS 0.01 0.03£ 0 3 5 6 9 11 12 16 17 21A 1.17 5.69 13.59 16.18 32.07 35.13 40.72 47.65 49.43 58.72T3 0.4 0.01 0.14 0.01 0.23 0.29 0.13 0.06 0.01 DM1£ 1 2 5 7 8 11 14 15 18 19A 1.60 4.35 13.59 15.98 29.48 38.28 42.15 46.56 49.37 56.22rw 0.96 0.04 0.91 0.35 0.35 0.0S 0.04 0.13 DM5 0.04£ 0 0 4 4 8 10 12 15 18 19A 5.15 5.15 23.78 23.78 31.52 33.30 38.55 49.48 53.32 59.3517 0.19 0.29 0.09 0.53 0.43 DM7 0.04 0.04 0.10 0.01£ 2 3 6 7 9 9 10 11 13 13A 2.93 8.08 12.56 26.05 32.71 32.71 38.30 45.14 55.58 55.582 T 0.99 0.19 0.12 0.02 DM3 0.23 0.12 0.07 DM3 0.03£ 1 1 4 6 9 10 14 15 19 25A 7.22 7.22 13.25 16.77 23.67 37.17 41.77 46.93 51.78 58.5877 0.77 DM3 0.11 0.34 0.33 0.76 0.32 0.74 0.76 0.79t 0 2 5 8 11 14 14 20 22 23A 2.78 4.97 12.99 23.61 31.17 46.49 46.49 62.62 71.60 80.732 0.33 0.40 0.15 0.15 DM1 0.15 0.22 DM3 0.74 0.37£ 1 3 6 9 9 11 12 12 15 15A 1.17 8.90 13.62 20.82 20.82 35.32 44.45 44.45 70.95 70.95W. 0.17 0.29 0.15 DM4 DM4 0.27 0.36 0.07 0.02 0.01£ 0 2 5 8 10 11 15 15 16 17A 0.74 8.02 13.69 17.85 24.33 37.84 47.26 47.26 49.98 55.133T 1.00 0.22 0.03 0.06 DM0 0.02 0.06 0.61 0.23 0.04£ 2 2 4 5 5 8 9 10 10 10A 7.50 7.50 10.77 18.49 18.49 30.26 39.39 59.82 59.82 59.8233 D.9 0.09 0.01 0.03 0.00 0.87 0.48 0.12 0.01 0.03£ 0 0 2 3 3 7 7 7 7 8A 6.03 6.03 9.38 17.10 17.10 49.26 49.26 49.26 49.26 58.39105Table 5b: Combined Year Male Results for Site 1 Lip, k1, . . . ,Cluster Size (k)SD 2 10 18 26 34 4 50 58 66 743T 7 1.00 0.25 0.02 0.30 0.02 0.82 0.40 0.09 0.01 0.02£ 0 0 1 3 3 5 5 5 5 7A 7.72 7.72 10.44 23.10 23.10 47.79 47.79 47.79 47.79 57.545W 7 0.58 0.00 0.28 0.55 0.09 0.04 0.06 0.33 0.08 0.12£ 0 1 7 13 13 15 20 23 24 28A 1.96 3.56 15.34 26.36 26.36 31.51 39.39 54.41 55.03 64.20&f 7 0.48 0.00 0.36 0.08 0.15 0.07 0.09 0.37 0.09 0.15£ 0 1 8 12 16 18 22 24 24 32A 1.60 3.56 16.23 19.25 28.05 32.90 40.79 55.17 55.17 65.2175 7 0.76 0.60 0.02 063 0.12 0.82 0.39 0.08 0.01 0.02£ 0 1 1 4 4 5 5 5 5 6A 2.73 10.44 10.44 27.36 27.36 47.79 47.79 47.79 47.79 56.927W 7 0.99 0.14 0.02 0.04 0.00 0.28 0.88 0.54 0.18 0.03£ 1 1 3 4 4 7 8 8 8 8A 6.66 6.66 10.23 17.94 17.94 37.96 58.39 58.39 58.39 58.398T 7 0.54 0.01 0.34 0.12 0.27 0.12 0.02 0.04 0.33 0.21£ 1 2 10 12 20 21 23 26 29 31A 1.81 3.77 15.98 20.02 30.22 34.63 36.29 45.27 62.07 67.03W jT 0.70 0.15 0.00 0.01 0.00 0.00 0.03 0.03 0.03 0.04£ 0 2 3 8 11 12 13 15 20 22A 2.44 6.82 7.90 14.94 18.76 23.91 37.41 44.66 51.92 59.30106Table 6.1: Male Combined Year General Results 1983—89Site School District k £ 0 A 0/A p PI Lip 1 Fernie 9 2 9 4.03 2.24 0.02 0.0002 Cranbrook 5 0 6 1.78 3.37 0.04. 3 Kimberley 7 1 9 2.86 3.14 0.0312 Grand Forks 5 0 7 1.17 5.98 0.0113 Kettle Valley 5 1 7 1.60 4.36 0.0214 S Okanagan 15 2 15 8.62 1.74 0.0315 Penticton 10 0 10 5.15 1.94 0.0416 Keremeos 15 2 15 8.62 1.74 0.0324 Karnlcops 14 2 14 8.46 1.65 0.0527 Cariboo-Chilcotin 9 1 9 4.59 1.96 0.0428 Quesnel 9 1 9 4.59 1.96 0.0432 Hope 13 2 16 7.50 2.13 0.0433 Chilliwack 11 0 15 6.03 2.49 0.0434 Abbotsford 14 0 14 7.72 1.81 0.0336 Surrey 29 0 32 20.43 1.57 0.0459 Peace River S 5 0 8 1.96 4.08 0.0560 Peace River N 8 1 10 3.56 2.81 0.0375 Mission 17 1 18 10.44 1.72 0.0476 Agassiz-Harrison 12 1 16 6.66 2.40 0.0481 Fort Nelson 8 2 10 3.77 2.65 0.0486 Crestori-Kaslo 6 0 6 2.44 2.46 0.04///q , IMp ty 3&n Sn,th & Mpnfo.107CdIHIcc —(--,-IrI I_-)II—’÷ICDCDC.I.ICDCD©I ICDCD—-ciIC)Cl)Cl)Cl)I cccc c.zII0000ccèbcèccCc.-C.,’C’ &0 0 C L.0 D (n 3 t 0C4c+,- 0 S p,Co...—c.’c’,,cS.o-:-cp-.’,—CCCCCCoCc-c—.-ci—‘‘C-•CoCoC1CC CC.ZCJIC C ICD CD CD CD CD Co Co cc 00 00 ccCiD 0 0 0 nj 0I--1CC1CCC?c, I_ , CD(4 CD (4 0 0 0 rj 00 CD—0CD÷CDI.I.C, CD C, 0 0 0 0 CDC-.: CCIU) 0CD CD CD CD IC, CD 0 0 U) C, C CCD oo CD CD CD IL:.CO0CDU) CDCOC.00C0)0000DCI—I—I—’IC..)I•.Di0)CDC)1C0)0COI—’—.0)CD0C.)C.70i—ii—I.D.0)—1C)’-000I-..00CCC C.C)’C C CCD -4 CD CD CD000)0)C)1C)1I•—I-(4CDC)C)’—1C)CDCCDD9)0•CDCD‘-0 —,mCD-U)g•)0-;—.0C.)-CD+U)CDC.)c)?D i’C..)CD00I—-C.ZCC--—C.)C).C.)-0)1-Is.)Is)-C.)-..: Is:)00)0)Is:)C.)u=- CC..)0000L’,:)I-0-CeeeeC.:).0)1C.).C 0)CD CD CD CDC.300)—CC-C)’C7000)00-.10)1CC00CC-C.CT’CICD-—).--C..C—E0t’:)CCOi CC..)C C. C CI.I•0 0 C a L a 3 0 0C aq C,—icCCr2(Cl)Cl)e,C-lC—.—.(_-c-Cc11occ)1.C-—‘i-.’—..cyCo-.-)- C eC21C C CCD CD CD CD ICID C C 0I—I..0C, IirJq0 0HHI o00CCiDç1I1(x)COCDrCD00OrIIOoIIo0II.O_CD•IIIiIICDIII—IiCD.CDIICDCDIICDCDI—II I!ICDCD•_II——CtJCl)C))Cl) JCOcc000000àooècXcc00CDCICC03JC-c-h3->-C C CTable 6.14: Male General Results 1983—89Site School District k £ 0 ) O/.) p 1’14 Lung Others 1 Fernie 13 0 13 7.16 1.81 0.03 0.00018 Golden 13 1 - 15 7.50 2.00 0.0419 Revelstoke 9 0 10 4.44 2.25 0.0429 Lillooet 20 2 21 13.21 1.59 0.0531 Merritt 11 0 12 5.55 2.16 0.0340 New Westminster 53 0 63 40.87 1.54 0.0450 Queen Charlotte 5 0 5 1.90 2.63 0.0452 Prince Rupert 18 1 19 11.23 1.69 0.0456 Nechako 48 1 56 36.48 1.53 0.0457 Prince George 40 0 51 29.86 1.71 0.0475 Mission 27 0 31 18.39 1.69 0.0481 Fort Nelson 5 0 5 1.08 4.64 0.0187 Stikine 5 1 6 1.82 3.29 0.04/16,22.Ji. /“4.14Map courtasy John Smith & Mapinfo.113t C) 0 C C)-C C 0 (0 3 00 C)•0 0I V 1:- C Gq 0C3-‘ccoe-t-Cl-—00Cl:0 Cr 0 0 0 Cl I-I n 0+0 a P C) a C) a a a Co 0+’ 0± Co cc ccC C-a P Ci C- a C) a a ;kCOC?tCCC N00)0CI-’II—al0)NNhC)hQOCcnC>-CCCoaeCe3C C CC 0Table 6.17: Male General Results 1983—89Site School District k £ 0 ,\ O/.X p 1’17 Soft Tissue 12 Grand Forks 5 0 5 1.71 2.93 0.03 0.071Sarcoma 13 Kettle Valley 6 1 6 2.42 2.48 0.0438 Richmond 156 2 229 135.23 1.69 0.0439 Vancouver 117 0 200 99.44 2.01 0.0541 Burnaby 161 2 236 140.00 1.69 0.0444 N Vancouver 140 1 223 120.43 1.85 0.0445 W Vancouver 150 2 233 129.85 1.79 0.0475 Mission 10 0 10 5.28 1.89 0.0419,21y 1,2.10\ >1 1 6,22—?’A14• -_______________)Map courtesy John Smith & Mapinfo.1150HCCID‘CC7c.c-::: C oae C,,’ CI,I.C) 0 C I— 0 3 t 0C?CC.I-.ts—.,.CI—DICD CD CD CD IC 0 Ct’C.”C”C-..0C.CD& Ct’ C, 0 0-_I:_-CC)1C.’1CCTICT.C C C C C C C(ID (C)CC?DCCl)IIc1C-cD)-CCCCCCCCCCC C” C is: C”•0 0 C “4 0 U) 3 :7 (4 t 0cjL’ct,c “CoCoCo—1-C.COC1C)-I,.3L:3Is.DI-4I.C4.ZI—i..C—I—t—I—ICoCC)CC41COC7C3CDC C 0CD CD CD CD ;0 0 4+ 2+ CCC,’1.Di4Table 6.23: Male General Results 1983—89Site School District k £ 0 A 0/A p 7’23 Kidney 2 Cranbrook 11 0 12 6.10 1.97 005 0.06942 Maple Ridge 22 0 25 14.36 1.74 0.0450 Queen Charlotte 14 2 14 8.21 1.71 0.0452 Prince Rupert 13 1 13 7.06 1.84 0.0366 Lake Cowichan 5 0 5 1.85 2.70 0.0475 Mission 15 0 16 9.09 1.76 0.0480 Kitimat 7 0 7 2.62 2.68 0.02‘S19,211,2.101622‘L14Map courtesy John Smith & Mapinfo.120is:CD0aq (i)CD Ct’HHHCzJ)“c1rcr’-—_—•oCD0CD0CD .o.‘CDCDCDCDCDCDCDCDCDICDCD0000IIIIIIIIis:)Ca:)I-Ci—iI.ccccIccIIcccccc00Ccc_51C e Cl C is:Table 6.27: Male General Results 1983—89Site School District k £ 0 A 0/A p P27 Multiple Myeloma 7 Nelson 17 2 18 10.37 1.74 0.04 0.0029 Castlegar 17 2 18 10.37 1.74 0.0410 Arrow Lakes 10 1 10 5.00 2.00 0.0311 Trail 17 2 18 10.37 1.74 0.0414 S Okanagan 21 2 21 13.99 1.50 0.0516 Keremeos 21 2 21 13.99 1.50 0.0529 Lillooet 7 2 7 3.04 2.30 0.0431 Merritt 5 0 6 1.29 4.66 0.0156 Nechako 15 1 15 8.69 1.73 0.0357 Prince George 15 1 15 8.69 1.73 0.0372 Campbell River 8 0 8 3.60 2.22 0.0384 Van IsI West 8 1 8 3.82 2.09 0.0485 Van IsI North 10 2 10 4.79 2.09 0.02Map courtesy John Smith & Mapinfo.122CDCDCDoCDCDCDCDCDCDHHH00COCl)00COCCCCC00<CCCC.1.C<:-1C)0LC0o—0U)U)c.)CDCDCDCZCDCDCDI—00ICC4——U)U)czcCC-4CC1-CCC)1I-D.:-IICOCO00CDCO00-4CD0---‘-.000CC--4000)-0)0)00CCcoczcV1COCt,o-.oSCD CD Cl)0HHHC!)“C.’lC)CCcI—0CD0CD0CDCDOQ0roC)CDCOI—0-CDCDCDCD CDCDCDciCDCDCiCoCoCoCOCoCl)000000CCI000000-4CCCC-4Ic7C,o(0o0H CDC)C)Ci)C)C:•.CDCl)00-1C?3C)C)C?Ci)C)C)C?’C?’ZC?’00C)C?’4C)CDCDCDoCD<coCDCD-0_-_0—.—1—CD“‘..iCDCDrjII.iLIiCDF-j’‘)ICDCDCDCDCDCDCDCDCDICDICDCDCDI—I—C?’-CDCZ—C)00I-‘CD—11C)C)_IV•Is.,ICDCOCOCOICOICOCOIICDI—I—Ip—’V.CO•...CO_COCl)COCOC)001_00-.000CD-)CY0C)C)-CCCCCCCCICCCCH0000CO00000000czI-’tZC?’C)CDI00CCD0000-d00CDC)cd0000CO-CCC?’00.C?) C?’CCCCCCCDCDCDC9cc?‘CCCC0.0C)C?’-C)C)C)CDC?)C?)C?’CD-‘C?’C)C)C)00 CCC?)C?’C?) C?’C?’C?)C?’jC?)CV a 0 C C 0 5.S 3 :1-•0 S 0‘—c,c .ct,0’4 J I00-.—1O00?-.1CQ<C)(DOZ—0b—I-’‘.ZI-C?5CO0-I00COCOC?COC?’—CCCCCCCCCCCCCCOC?1COCTC?’C?CD I, CD CD CD CD CD I, cc 00 CO 00 cc0 0 0 000—i00C?1C-CTable 7.11: Female General Results 1983—89Site School District k £ 0 A 0/A p p11 Lung 46 Sunshine Coast 5 3 5 1.96 2.55 0.05 0.003Adenosquamous 52 Prince Rupert 5 14 5 1.72 2.90 0.0354 Smithers 5 10 5 1.30 3.84 0.0155 Burns Lake 5 9 5 1.39 3.59 0.0156 Nechako 5 8 5 1.28 3.90 0.0157 Prince George 5 6 5 1.28 3.89 0.0159 Peace River S 5 7 5 1.30 3.84 0.0160 Peace River N 5 8 5 1.38 3.63 0.0168 Nanaimo 5 2 5 1.33 3.77 0.0169 Qualicum 5 3 5 1.60 3.13 0.0280 Kitimat 5 11 5 1.47 3.39 0.0281 Fort Nelson 5 9 5 1.19 4.21 0.0187 Stikine 5 11 5 1.32 3.79 0.0188 Terrace 5 11 5 1.47 3.39 0.02‘4__1Map courtesy John Smith & Mapirtfo.127Djoq0 0Cl)c’—I—II000c.)‘?I___0— CD CD CD-QoQoCcc cc0L..zTable 7.13: Female General Results 1983—89Site School District k £ 0 .\ O/) p p13 Lung Small Cell 24 Kamloops 22 1 22 14.28 1.54 0.03 0.01431 Merritt 22 1 22 14.28 1.54 0.0332 Hope 22 2 22 14.37 1.53 0.0433 Chilliwack 19 0 19 11.79 1.61 0.0347 Powell River 19 1 23 12.32 1.87 0.0566 Lake Cowichan 17 1 17 10.52 1.62 0.0470 Alberni 30 2 33 21.54 1.53 0.0571 Courtenay 14 0 15 8.28 1.81 0.0472 Campbell River 20 1 21 12.90 1.63 0.0476 Agassiz-Harrison 22 2 22 14.37 1.53 0.0484 Van Isi West 20 2 22 13.10 1.68 0.0519,21.s c\7 &vr16,22Map courtBsy John Smith & Maplnfo.129& DCD. 9’I4.C00C’DZ09’• -Cl CD Cl 0 0 0 Co 00 0 0 Co 0©C CD CD 0 CD CD 9, CD Co Co cc 00 00 cc0OC.1—o Cl)0°0CDC;I—C;‘CtCDC;1Ci-Cl)0Q0 —CD09)0 ÷0 ...c,CC7IC,’--CCC—IC)1C;C;C; CC)’C.0C; C C CCC CD CD 0 CD CD 9, CD Co Co cc 00 ccC CD 9, CD 0 CD CD 9, FECto CDCD 9, 0 9’C3CCl“00C1C,100.0C00cD0?-.oCSC?I—00C))-—.-‘C -.C eCC C;I.zI-C;C;IttsZCI—Ccz:C.CC—i00CTCO—C.”-I-I-C.”C;CC;.’C.’C00CC-CCCCCCCC;.’C;C CD 00 9, CD 0 CD CD 9,1C4C CD Cl)00CC).t’.Ci 5OCDCDCD0‘OC)CiDr.CDCDI-.’C;ZIC.II—I—iI_C;.’ICc-.CI-CcccccDoC;).D-C;- CC.”C)00CC)CCCCCCCCbööè.C..’ C.”.CC CC CD I. 9, CD 0 CD CD IC E”C;C.ZC.,CCI-.C,’..’C.’0—4C.——.C.CyC.’I-—CCC C C C C.Is 0 CD IC.)’CCYC_•“CD(I—4(•)-CDcoGq0- COIsCt±-CD+‘CD CDC, CD C, 0 0 CO C)Is = CD Ct,00CD-.C CC, CD C, C) 0 0 CO C) CI.Is C 0 C cjq CO CO CD CD,CCeoo0“CD—1CY1C)’C.3C:.‘-•-IC)C)’C)C)’=‘.CDCD--) CDC, C) 0 0 CO C)ci CD C, C) 0 0 CO C)CD cc CD CD CD CD CI) CO I. cc 00 00 ccCD 00 CD CD CD CD CO C’) I cc 00 00 ccCD CD 0 CD Ct’NCj CD0CO00O)C)’C)I-Ct00IsCCC)CJ)i-’I-CC”,.C)O?CCCC-,t-C-CC.-.4Is.’IS.00C.)-4ISIs:Is:’I—Is.DISCISD0OIsC))—-CCCCCCCCC))C))C))C)’, C Is C) 0CD -4 CD 0 CD CD ;IS) IS)IS)IsIS)IS)C)) IS)C.Cl’Cl’CI—CC—.Cr1CT’00I—i 0CCCC00CC0C) I—C——-CC)’C)’C)’—1C-.c’C)o—CCCCCCCCCC,C..C)’.C C N 0(I—II—IS)I—I-I3CC.4IS)-?t,—I—I—’CIS)Is)CtsI—’I—’•C))Is)IsC)Is) IS)IS)CC)’CCIS)I—ICXCTl Cl’-.4CCC)CC)C)C)C)C))C)’C)’C0000.CC—4C—.)IS:H—’—‘IS)I-C)I—’’,—’C)300CCXCCCCCCCCCC))C))C)) C)•’IS)kCC31C)PCCCIS)C) C CC CCIS)NI—Is)NC)’Is)NCI-.CX 00CC—C)’C)-‘r-CC)C)’C)CCCCCCC)’C) C C C) ISC ‘c‘0 C 0 0 3 0cz Ic,(-;cD0 CD CD CD S ,-I-rj0C1CDCCDCD‘-I—I—.•ts.ISocic,i00CCCt—)—‘I—I—-I0000I—CC.00CCCQ0CD0I C C cCDCD CDICiDI— CD CD CDI.CD CDC0 C 0 L 0 3 0 0—1C)-’C—.(.)oCDCI—..I—”-fl,CII-.C)’C CC,—cCD ;Cr CD 0 0 to 0CDCC,-C?’.C)’ C CCC) C C 0 0C—c(•DCO(•D--.,C0.-0.-IC)Ct)C)CD0oCD00CDCDCDCD0C)CDCDCDCDCDCDICDCDI-ICOIIIC0.CI.I.cccccccc000000 czCIIICO00000000cccccccc‘I-I:QQCOCCCCCeCCCC1cv1CI00 C,’Table 8: Combined Year Male Focused Results by Site 1983—89Site Mill Location NSD k £ 0 A 0/A p V1 Lip Quesnel 28 9 2 9 4.59 1.96 0.04 0.303Skookumchuck 3 8 2 9 3.56 2.53 0.032 Oral Cavity (Jastlegar 9 10 1 14 4.78 2.93 0.02 0.4463 Esophagus Quesnel 28 13 2 15 7.65 1.96 0.05 0.6054 Stomach Port Alberni 70 19 1 22 12.37 1.78 0.05 0.001Prince Rupert 52 14 2 14 7.86 1.78 0.03Port Alice 85 7 1 7 2.96 2.37 0.03Ocean Falls 49 8 2 8 3.76 2.13 0.04Campbell River 72 17 1 17 10.37 1.64 0.04S Pancreas Prince George 57 21 1 21 13.32 1.55 0.03 0.070Crofton 65 22 1 28 14.76 1.90 0.05MacKenzie 57 21 1 21 13.32 1.58 0.039 Larynx Prince George 57 18 1 18 11.57 1.56 0.05 0.234MacKenzie 57 18 1 18 11.57 1.56 0.0510 Lung Squamous Squamish 48 13 1 13 7.61 1.71 0.05 0.017Gold River 84 27 2 30 18.62 1.61 0.04Campbell River 72 26 1 29 17.59 1.65 0.0411 Lung Nanaimo 65 5 1 6 1.92 3.13 0.05 0.386Adenosquamous13 Lung Small Cell Nanaimo 68 37 1 44 27.03 1.63 0.04 0.46614 Lung Others Prince George 57 40 1 51 29.86 1.71 0.04 0.011Prince Rupert 52 18 2 19 11.23 1.69 0.04MacKenzie 57 40 1 51 29.86 1.71 0.0420 Prostate Prince George 57 134 1 178 115.13 1.55 0.05 0.001MacKenzie 57 134 1 178 115.13 1.55 0.0521 Testis Powell River 47 15 2 17 9.19 1.85 0.05 0.242Port Alice 85 7 1 8 2.95 2.71 0.0322 Bladder Gold River 84 22 2 34 14.38 2.36 0.04 0.096Campbell River 72 21 1 32 13.53 2.36 0.0423 Kidney Kitimat 80 7 1 7 2.62 2.68 0.02 0.048Prince Rupert 52 13 2 13 7.06 1.84 0.03Ocean Falls 49 8 2 9 3.51 2.56 0.0327 Multiple Myeloma Prince George 57 15 2 15 8.69 1.73 0.03 0.044Gold River 84 8 2 8 3.82 2.09 0.04MacKenzie 57 15 2 15 8.69 1.73 0.03Campbell River 72 8 1 8 3.60 2.22 0.0328 Leukemia Crofton 65 27 1 32 18.56 1.72 0.04 0.143Port Alice 85 8 1 9 3.63 2.48 0.0329 Acute Leukemia Nanaimo 68 19 2 21 11.73 1.79 0.03 0.54130 Chronic Leukemia Crofton 65 21 1 27 13.39 2.02 0.03 0.183Port Alice 85 6 6 2.12 2.82 0.0231 Other Sites Powell River 47 15 15 9.12 1.64 0.05 0.41332 Primary Unknown Prince Rupert 52 16 17 9.73 1.75 0.04 0.318136Table 9: Combined Year Female Focused Results by Site 1983—89Site Mill Location NSD k £ 0 A 0/A p 1’2 Oral Cavity Port Alberni 70 10 1 12 5.26 2.28 0.04 0.5753 Esophagus Prince George 57 7 1 7 2.85 2.46 0.03 0.266MacKenzie 57 7 1 7 2.85 2.46 0.034 Stomach Campbell River 72 9 1 9 4.69 1.92 0.05 0.5855 Colon Port Mellon 46 33 1 40 23.70 1.69 0.04 0.2266 Rectum Powell River 47 18 1 18 10.93 1.65 0.03 0.3618 Pancreas Castlegar 9 16 2 20 9.95 2.01 0.05 0.201Crofton 65 19 1 20 12.25 1.63 0.0410 Lung Squamous INanaimo 68 27 1 37 18.32 2.02 0.03 0.027Powell River 47 21 2 25 13.60 1.84 0.04Quesnel 28 7 1 7 3.24 2.16 0.05Campbell River 72 21 2 27 13.99 1.93 0.0511 Lung Prince George 57 5 7 5 1.28 3.89 0.01 0.019Adenosquamous Kitimat 80 5 12 5 1.47 3.39 0.02Nanaimo 68 5 3 5 1.33 3.77 0.01Prince Rupert 52 5 15 5 1.72 2.90 0.03MacKenzie 57 5 9 5 1.49 3.35 0.0212 Lung Kamloops 24 37 1 42 27.04 1.55 0.04 0.433Adenocarcinoma13 Lung Small Cell Powell River 47 19 2 23 12.32 1.87 0.05 0.100Kamloops 24 22 2 22 14.28 1.54 0.03Campbell River 72 20 2 21 12.90 1.63 0.0414 Lung Others Port Mellon 46 14 1 14 8.32 1.68 0.04 0.148Quesnel 28 10 1 11 4.73 2.33 0.0218 Melanoma Crofton 65 26 1 27 17.82 1.52 0.04 0.052Port Mellon 46 41 2 51 30.60 1.67 0.04Squamish 48 38 2 45 27.91 1.61 0.0422 Bladder Port Mellon 46 26 2 29 18.09 1.60 0.05 0.262Squamish 48 23 2 23 15.14 1.52 0.0424 Brain Castlegar 9 5 1 5 1.95 2.56 0.05 0.125Crofton 65 13 1 13 7.31 1.78 0.04Port Mellon 46 20 2 20 12.64 1.58 0.0327 Multiple Myeloma Prince George 57 11 2 12 5.68 2.11 0.03 0.291MacKenzie 57 11 2 12 5.68 2.11 0.0328 Leukemia Crofton 65 24 2 25 15.94 1.57 0.04 0.49729 Acute Leukemia Gold River 84 7 2 7 2.64 2.65 0.02 0.256Campbell River 72 6 7 2.37 2.95 0.0330 Chronic Leukemia Crofton 65 14 15 8.22 1.83 0.04 0.50332 Primary Unknown Prince Rupert 52 12 12 6.73 1.78 0.04 0.35535 Cervix Quesnel 28 13 16 7.17 2.23 0.03 0.44836 Cob-Rectal Port Mellon 46 48 1 56 36.50 1.53 0.04 0.13739 Endometrium Port Mellon 46 75 2 94 61.02 1.54 0.05 0.099Squamish 48 64 2 80 51.20 1.56 0.05137I.cr Cl; I. CD CD CD I0-CC;C;C;C;II--C4CCCC;ICCCC=rj77aq:0-,CDL—e+rj-—E’CDCD—CDCD.CD7C07,0-— 0C,02222222222------)----.-CC;D0oI.X’C;c’1.C;.C;Cào‘bo‘cccc0c00:C;CCC00*..—ccc.00—C;CccC-C?C;3CC;1C;Ccc:CC•C;ccC’cc.-I—ICC;C;:•9cc*CC;C..C’C;C;1C;CDI—‘---‘C—C;C?1C;C?’C;*—CIC;C;C;QI—’I—I—:i—i——D.,-,.‘.I—.01:’C;.C;001C;C;ccCCCCCCCCCCCCCCCCCCCCI-C-CI-CCCIsC;-ccCD I. CD CD CDCD I CD CD CD ICD I. CD CD CD0C)Cc ,(•)U)cCl)-:— rI—.—i..c,icct—c)—Ir I—.’.I;—.CCCCCCI..,C.ZCCC)—***C:—.I—aoCcCCTCC:-‘—-CCCcCIs.DCCC,- CCIc—,C.o,,000.PCl) ±ct,- L E 0 ÷ E2?---I—CI C0Il4C)C)’C3C)iC)I-CCCCc)CIS0**C)’i*C—‘I.jC)C’ 0—iccy C C0C)000,——CC)i-C)CC)1CC0*I—’I—CC?3CCCC.-.I-.CC 0C0C00C- CC00C0*CCC*CCCèèCCCCCCC).C)C-C00.C-.D .0C•0-.-U) ‘-------->-C CCg•CCC CCC******C)):‘‘-CcC CC******C.I•‘CCC-C’CCtDC)Ci—••..C))CI-C’C)CC3CCL-CCC-CCcCCCCCC’CCCCCI—CCCCC’.CC.C)1CCCCC..C0CcCCCCCC’CCC’C,C-CCcc*******C’CCCCCCCCèèèè-CCI-CCC)CCCD cc CD CD CD CD 0) Ci) C,)CD cc > cc 00 00 ccCD p CD CD CDCD I. p CD CD CD CD Ci) Cl) CD C) CD cc 00 00 ccC) C C SC’ C C 0 I- 0 0 C)C,I-C000CD‘)...,,p’.C0C•CDS S C) + CD222222)->-----I—I—0 0*C?I—I—’:C):0:.,I—:)—‘I—•C):C)c...tI—:••-....IDi—0C)1C7C5’****0I—C?3ts— ccC.3C******0CCCCCeeeeeeCC—CCI C) I 0 0*CI—I 000-CCCC4CI0C0cCD,,,p;c’-CD0.CDç0ct -SI0—C2222222-)- I—’IIèC?CTIào-4C)ào’C)Z0000D__jICC1C))—I.L1C)-C)’CI—I—I.-I—I—’I00C?C)CI—’II—ILC)DCC)1C)1&cI-00CDCCI——‘IcCcc-0—a00*0I—•I——‘pI—a004C0000****CCCCCCCCee ccCC?C CI- CC 0 CC 0CD CD CD CD CD CO CO CD I. I. CD C CO 0 Co cC 00 00 CCCD I’ CD cD CD CD I. I-CCCiCCC C0—,?aCaa?a2a-----c1LIs0030*.-0**I—I.I—1s301I—:•:•I..ao:;Is153Is03c1535315c1g*CCc153151153-àø—4000CIsI—153**C05C003..IsC)1N1531505IsI-***03oI—’I—CC’‘I—01ISCCCCCCCCCccocc—-c0(4CNI-—C..CCCo‘aqQ’0000-CDL:q0CD-I—CD(ICOCD÷.COCD0,-t2a2a2a2a2a2a2a-->->----I—I—’:•I—p—‘iCCC155CI—CCI0.4-.4.•N-.4i—CCI—’‘I—ppIoz00C100.515503CCCC.I1I—iI..pp,C3103CyI0155-.4152155c.2p‘-I015201155CCCC—01:—:l05(X>03155**CC-.4C.I—I—_0—1i—00.‘CCCC-1—0503030.15150*******—.I—I..I.. CIS.5155C155010505I—I—’II—’p1e,01;:cjCCCC0505005153.CC152CCCCCCCeeee00CC00.I-CD I. I’ cz CD CD CD CD cf: CD cz oq CJD C-) CD cc 00 ccCD I. CD CD CD—C0tCCC5, —- L. c -2?---I ‘ 0II—I ‘bC1b—000C”pI- C’ 0 Cp•“I-.-Ec000C”0**C—ICC00000-*pP01I—’0C’ 0*0 I C-‘cC’0a*c,1CCC.C-.—C’CJE0. c .?—i:r’:bC00***I—•.CCI—’I-.c.1tISC—III—?-EI—I——101C’CCCCCC’C CCCC’CD I. I. CD c-)CD CD o.I. CzCCCCV.CC—CCCCI-CC00—CDCDCDCDc—‘.,,-CDCD-CD,0,CD0.t-CC0CCD0CD0CDCD0...Gq CDC a.,-++CDCD22?222222222222-,_->------>-.---- CTIoo-cViii—VCccnxci,—‘I—’0o-àoooL1ercc0o00C0D.hC.-4..—4I—C.jQo—4-V1CCt..,oco99C*.-4i—.I—i00c,10DCCCCC0CC’—1Cc******I-..,—I--‘_i_Q0C—‘I—**.00****Vi-ViC30000000ViI—I—I—0CVI*C**CCCCCCCCCCCCCC-èèèè-èè-è..C-sC--CCTable 10.15: Male General Yearly Results Site 15 All Lungs 1983—89School District 1983 1984 1985 1986 1987 1988 198924 Kamloops F 35 37 34 35 31 32 35 0.10£ 3 4 1 3 0 0 30/A 1.01 0.83 1.41* 1.19 1.48* 1.55* 1.2126 N Thompson 47 50 50 48 32 33 47 01T£ 5 7 5 3 1 1 50/A 1.01 0.87 1.08 1.26 1.45* 1.52* 1.0529 Lillooet 12 13 12 5 12 12 11 1T.TW£ 3 3 2 0 3 3 20/A 1.15 0.94 1.77* 3.60* 1.32 1.54 2.01*30 S Cariboo 33 36 35 34 35 36 33 U]YW£ 5 5 3 3 2 2 50/A 0.97 0.80 1.39 1.29 1.38* 1.48* 1.1731 Merritt 34 36 7 35 34 35 6 1YOW£ 4 6 0 4 1 1 00/A 1.16 0.99 2.42* 1.19 1.39* 1.61* 349*33 Chilliwack 39 42 41 40 41 30 41 1t4V£ 3 3 3 3 2 0 30/A 0.88 0.97 0.82 0.91 1.38* 1.44* 0.7936 Surrey 113 121 121 119 124 132 104 O5£ 3 3 3 4 3 2 10/A 1.15 1.08 1.01 0.87 1.09 1.17* 1.23*38 Richmond 38 264 258 246 260 278 269 1Y5W£ 0 2 3 3 3 3 30/A 1.39* 1.17* 1.04 0.98 1.00 1.03 0.9539 Vancouver 202 211 297 283 299 320 310 03T£ 0 0 4 4 4 4 50/A 1.17* 1.20* 0.99 0.97 0.97 0.96 0.9240 New Westminster 1 26 120 26 114 121 129 125 WT5t 0 2 0 3 3 3 30/A 1.48* 1.17* 1.53* 0.82 0.99 1.03 1.0141 Burnaby 1 85 289 281 268 284 304 296 WTW£ 1 2 2 4 3 3 40/A 1.26* 1.17* 1.11* 1.00 1.06 1.02 0.9242 Maple Ridge E 71 76 20 20 77 82 77 1t5W£ 3 3 0 0 3 3 30/A 1.11 1.05 1.52* 1.65* 1.10 1.15 1.0443 Coquitlam k 55 38 119 114 121 129 125 1T3W£ 1 0 3 3 3 3 30/A 1.33* 1.41* 1.00 0.82 0.99 1.03 1.0144 N Vancouver k 233 245 260 247 262 281 273 1J5£ 1 1 3 3 3 3 40/A 1.12* 1.12* 1.01 0.98 1.01 0.95 0.8947 Powell River k 37 40 30 38 30 31 38 ]T]3W£ 3 3 1 3 1 1 30/A 1.14 1.01 1.40* 1.20 1.41* 1.45* 1.2571 Courtenay s 37 40 30 38 22 22 38 1YU£ 4 3 1 3 0 0 30/A 1.03 1.18 1.40* 1.24 1.55* 1.56* 1.1572 Campbell River k 37 40 14 30 31 32 38 ]T01£ 5 4 0 1 1 1 40/A 1.02 1.19 1.89* 1.41* 1.45* 1.52* 1.1984 Van IsI West k 30 32 14 31 32 33 14 1]JTt 4 3 1 3 2 2 10/A 0.82 1.18 1.77* 1.36 1.46* 1.48* 1.72*144Table 10.16: Male General Yearly Results Site 16 Lungs: non-small cells 1983—89School District 1983 1984 1985 1986 1987 1988 198924 Kamloops F 31 33 30 31 27 27 30 0.10£ 3 4 1 3 0 0 30/A 0.97 0.83 1.43* 1.18 1.50* 1.53* 1.2226 N Thompson r 41 44 43 42 28 29 41 1323£ 5 7 5 5 1 1 50/A 1.00 0.86 1.12 1.14 1.48* 1.50* 1.0631 Merritt 30 6 30 30 30 30 6 0]3W£ 5 0 1 4 1 1 00/A 1.19 2.37* 1.43* 1.24 1.43* 1.50* 3.66*33 Chilliwack 1 35 37 36 35 36 26 36 W5T£ 3 3 3 3 2 0 30/A 0.88 1.03 0.83 0.90 1.40* 1.58* 0.9036 Surrey 98 107 104 103 109 114 90 1YIW£ 3 3 3 4 3 2 10/A 1.11 1.04 1.05 0.92 1.13 1.18* 1.26*39 Vancouver 176 186 254 245 261 275 267 13W£ 0 0 4 4 4 4 50/A 1.15* 1.19* 0.99 0.97 0.97 0.96 0.9440 New Westminster 1 98 105 23 99 106 111 26 1T]T£ 2 3 0 3 3 3 00/A 1.19* 1.04 1.67* 0.82 1.02 1.04 1.54*41 Burnaby k 240 254 241 233 248 262 255 IYUW£ 2 2 2 4 3 3 40/A 1.16* 1.17* 1.13* 1.02 1.07 1.04 0.9443 Coquitlam ic 98 51 50 99 106 111 108 13TW£ 2 1 1 3 3 3 30/A 1.19* 1.33* 1.35* 0.82 1.02 1.04 1.0247 Powell River k 33 35 34 33 35 27 33 1J]JT£ 3 3 2 2 2 1 30/A 1.23 1.05 1.49* 1.41* 1.43* 1.55* 1.1471 Gourtenay 33 35 34 19 35 19 33 T]£ 4 3 2 0 2 0 40/A 1.13 1.19 1.49* 1.69* 1.43* 1.69* 1.1472 (Jarnpbell River k 33 35 12 27 27 28 33 ]T0r£ 5 4 0 1 1 1 40/A 1.11 1.20 2.09* 1.65* 1.47* 1.49* 1.0684 Van Isi West k 26 28 13 27 28 28 27 13]3£ 4 3 1 2 2 2 40/A 0.86 1.22 1.96* 1.61* 1.49* 1.46* 1.06145CD © CD CD CD c:1C C C CD C CC, CCD p I. CD CD CDCD p I. CD cc.) CD CD I H CiD CD (ID 0 cc ccI,I-—CC0CCDCD0.I—CCcCCD±000CD0 C0-C0 0 E222??222->-----..I—i-03***C.._—.)—I—i-***C)C)C).I—I—’0)*****C.I—I—’:i—i—1)-a’CO*****C..—I***:t..)-i01CL\)I- ***0_IsC.Q1 ***0CCCCCCee-eeeCIsCCCC 0 ci C 0‘•l 0--IL1cL:i.0I—*I—II—)-.,I——)—0CIC)I—CCCIIsOC—C 80CCo822??2222.-.--.---..i—I—kC)1CDI—I—000)010I—I—,_Iè)!cCII—II:.•*****010II—I—‘I-.-z0)****C)CIs:IsI.DIsCCI—’Is)Is0O*****CIIsIs..C)i******CCCCCCCCèCCCCCI00 0)C CIC C. CI- CC CI CCTable 1O.20b: Male General Yearly Results Site 20 Prostate 1983—89School District 1983 1984 1985 1986 1987 1988 1989 j28 Quesnel 34 36 41 44 49 27 13 OUW£ 4 2 3 5 2 1 00/A 1.23 1.99* 1.19 1.18 1.33* 1.42* 1.84*29 Lillooet 12 12 9 9 15 15 15 W5T£ 3 3 1 1 3 3 30/A 1.11 1.13 2.06* 1.98* 0.83 1.18 1.2030 S Cariboo .—— 33 34 7 9 46 47£ 5 5 0 1 5 3 30/A 1.30 1.21 2.92* 1.98* 0.79 1.18 1.2031 Merritt . 33 35 37 39 47 48 47 13]J5£ 5 4 1 1 5 3 40/A 1.05 1.20 1.41* 1.50* 0.94 1.21 1.1534 Abbotsford 34 71 41 86 99 53 53 1YW£ 0 3 0 3 3 0 00/A 1.41* 1.07 1.34* 1.05 0.95 1.28* 1.48*44 N Vancouver 272 259 45 48 354 61 434£ 3 1 0 0 1 0 30/A 1.04 1.12* 1.40* 1.47* 1.13* 1.52* 0.9045 W Vancouver k 60 281 72 36 93 101 434 TYUW£ 1 2 1 0 1 1 30/A 1.26* 1.13* 1.35* 1.58* 1.24* 1.38* 0.9047 Powell River k 37 31 14 46 53 55 42 131T£ 4 1 0 3 4 3 10/A 0.80 1.52* 1.78* 0.92 0.72 1.11 1.54*48 Howe Sound k 63 67 76 40 97 105 107 0]3r£ 2 3 2 1 3 2 30/A 1.25* 1.17 1.34* 1.45* 1.23 1.35* 1.1049 Central Coast 9 9 10 10 4 11 11 lflT£ 4 4 4 4 0 2 40/A 1.17 1.12 1.68 1.22 3.72* 1.86* 1.6255 Burns Lake k 13 13 14 15 17 6 16 13]JW£ 3 3 3 3 2 0 30/A 1.12 1.31 1.08 1.37 1.69* 347* 1.1056 Nechako k 26 25 8 31 32 33 9 13]JW£ 3 1 0 3 1 1 00/A 1.15 2.35* 2.18* 1.10 1.85* 1.45* 2.16*57 Prince George k 30 21 35 37 27 28 32 wor£ 4 0 4 4 0 0 10/A 1.21 2.61* 1.18 1.06 2.01* 1.43* 1.55*59 Peace River S k 32 33 12 39 44 45 43 TJ3T£ 4 2 0 4 2 3 30/A 1.13 2.00* 1.74* 1.09 1.63* 1.29 1.3560 Peace River N k 32 33 19 39 44 45 43 ]TOT£ 5 2 1 5 2 4 40/A 1.11 2.00* 1.66* 1.07 1.63* 1.27 1.3375 Mission k 59 16 53 75 86 91 67 1TlT£ 3 0 1 3 3 3 10/A 1.11 1.79* 1.36* 1.05 1.00 1.05 1.33*77 Summerland k 85 12 100 41 15 133 53 ]3U0£ 4 0 6 1 0 3 10/A 1.17 2.04* 0.89 1.56* 1.64* 1.18 1.28*1470 CC C CD CD C 0 CD0 0 Ti’ C’ 0 0 C00CD I.CD CD CD I00—CCOCC..C<c9)-—CCD00:z:.‘22?2222->----)-)——.-pICD.20 0CDCD-CCCz-Q1C’0CT!10ISO****IS:Its’I-0OI—CIsLCD1***4••1CCCOCO—OI._0CiCT!0ISO.C”CT!***.I‘-0COCC’CiC’1,_.,_,_I—I;—.07!07!—CD0ISOCDCO.0CT!COCO.‘-0I00***0CCCCCeeeCC0—CPI‘—‘I..CT!4,_,..•IO,QCC”C.I—.CT!CT!CI—’I.‘-‘C’C,-**C—IS.j**Cc I—IS0 CD I. CD CD CD0—-‘c0CIsI--CC‘‘0,S0q:CDI’:‘0C•1CD0‘-—1CDC.222222-:.--.--.-I—’I—’I—’I.-’.COiLcoiCC?C,00C”C”*0ISPCj,00CO0CD.•CD00ccCD***I—’.I—,_CD0—40?b.4C0c;—‘cg-COCOC”IOISZCO00OC’boCDIàoàoC’DC’DC’1)-CD-COCOIS.)CD.0CC00CISO-0—I.—IS—i—-‘-C0CCO-L—COCY!CDCO•__CCCCCeeeeC,4IsIsCD CD CD CD CD Ti’ C” CD I.C CD C CD 0 CD(1. 2’ 0 0 CO C) I Cc ccCC0(1CCCCL:•Co2’CD—>---..ào•0**C)‘-.II—0i—,.I—)àoC)1:.0.I;1:‘•--C)’**0I—....C,’,..I—’0CCCeC$CD CD CD CD CDCD p - CD CD CD CD-C)C.CCJCCc0C?Cr00—CC)•cCDCDa..-—CD2’--CD222222------.C,0C)’*00I—-‘—-II—- i—CC?1C—c>—00Z-00C0000****—-‘:‘CD0L0è-100I—r- CCCCe•e-I-C;CD I. CD CD CD CD CD CO CO C/) CD 0) z 0 0 oq CO 0 00 ccC))I*C I C00C)CI—IC0Ca0)0Ls e IH•pI__CD0C,’00b..)-00cI’C0C)’C)II—C)is:C*)ISI--*I—.IC)Càoi0C C CCD p CD CD CD CDI CDC CC) C C’CD CD CD CDCD CD CD CDCCCI-I-CJCII-CG7-o,0,.S0CD-,0jp,2222222>----->--—I—..b0CDO°àoQ01C)C**:I•CCO**.—.I—.‘IC7CC1C)1C)’CC)’CC I—CCTOO’oC3C,CC?CDC)cI—’ j’ào°0)Q..C?C)1C.CI—I—,pC3C—CD—CD.C)zC-1-C3**0)_.‘C3b)CèC)’0CCDC)C)1C)1ccCccCCCCCCCcIISCC)CCCjC.c ,,—0_C 0••‘-.I_jr)CD .2??--C)): C)C)i—-‘•ISC—**CCI—)—‘i_cC)-4cDISICLICDI-vCCCCL’C)CI-ICD CD C) CD CD CD Co 0 Co Cl) CD C) C ICD CD C) CD CD CD Co Co Cl) CD 00 00 CC0-—CCCJCIsC0C.C£, DCo- Co -.0(D Co +, )-2?2?22222-->----—I-‘QOfCCCCC7CC)1IsZIào-CCCCC1.ICC’,0C)1boC—1-C)CC.I-i—I—Iez.‘-C)C.T’*C.I.-’_zCCCIsCD.IsCCCC**I—i—’.:‘I—_C)—bocC-CCCCC*****....I—I-cDCC*****CCCCCC—eC—C.’C.C.’C.’C.’-DI-CCCC4C(CO..Ct’C.) 0C.)C0‘-222222‘-‘-‘-:.-Is:?.I—’Ii—’‘- C a.C’i_.•.,—I—’I—’Ca—‘CC.’)C..)CCI—..i.—’I—’_I—’_-CCI—’‘I—’I—’I-C?cjC..I—’I-’I-’00,’IsZ?I_.-IC.3Is)IsCI-..I-CCCCCI-CC****—I-’I-’I—’I-’Is.)C)-C)CCC****CCI-I•-I-—‘I—CCcIs:?CIsIs.)C1s.’CC***•CCCCCe.’eeC.’IsCI—Table 10.38: Male General Yearly Results Site 38 All Cancers Except Lung 1983—89School District 1983 1984 1985 1986 1987 1988 1989 j14 S Okanagan F 130 135 51 145 156 163 164 1J]Yi.T£ 2 3 0 2 2 2 3OR 1.18* 1.11 1.31* 1.16* 1.22* 1.19* 1.0615 Penticton 80 148 155 159 98 103 182 1EUT£ 0 3 4 2 0 0 4OR 1.29* 1.11 0.88 1.16* 1.39* 1.28* 0.9616 Keremeos r— 130 135 62 145 156 163 164 irutr£ 2 3 1 2 2 2 30/A 1.18* 1.11 1.25* 1.16* 1.22* 1.19* 1.0628 Quesnel 150 161 167 170 174 175 170 1UJ5£ 5 2 2 6 2 3 3OR 0.97 1.17* 1.14* 0.94 1.18* 1.11 1.0838 Richmond 1023 1060 1094 1103 1193 1259 296 11OT£ 2 2 2 3 3 3 10/A 1.06* 1.09* 1.10* 1.02 1.00 1.02 1.12*39 Vancouver 816 836 857 857 930 1446 1464£ 0 0 0 0 0 3 3OR 1.08* 1.16* 1.13* 1.06* 1.07* 1.03 0.9841 Burnaby 329 1157 1190 1196 1299 1375 1395 1T]W£ 1 2 2 2 3 3 30/A 1.14* 1.09* 1.09* 1.07* 1.02 0.99 0.9744 N Vancouver 148 976 1004 167 1093 189 1282 0]JW£ 0 1 1 0 1 0 30/A 1.16* 1.15* 1.13* 1.25* 1.08* 1.15* 0.9845 W Vancouver 1027 1061 1093 263 1195 1264 1282 13]3T£ 2 2 2 1 2 3 30/A 1.07* 1.14* 1.11* 1.20* 1.07* 1.03 0.9846 Sunshine Coast Ic 244 259 272 280 53 54 53 WTh£ 3 3 3 3 0 0 00/A 0.94 0.98 0.96 0.97 1.30* 1.43* 1.41*56 Nechako Ic 110 108 111 122 115 126 123 WIT£ 4 1 1 3 1 3 30/A 0.95 1.25* 1.32* 0.93 1.24* 1.15 1.1157 Prince George Ic 127 91 94 142 97 147 143 13]£ 4 0 0 4 0 3 30/A 0.96 1.37* 1.32* 0.93 1.28* 1.15 1.1159 Peace River S Ic 138 146 151 153 156 158 154 1JTT£ 4 2 2 4 2 4 30/A 0.92 1.20* 1.17* 0.93 1.15* 1.04 1.0660 Peace River N Ic 138 146 151 153 156 158 154 1J]T£ 5 2 2 5 2 5 40/A 0.92 1.20* 1.17* 0.92 1.15* 1.02 1.0677 Summerland Ic 288 303 318 327 126 133 365 1YIW£ 4 4 6 4 1 1 40/A 0.98 0.98 0.88 1.02 1.34* 1.25* 0.96152Table 10.46: Male General Yearly Results Site 46 All Cancers 1983—89School District 1983 1984 1985 1986 1987 1988 198914 S Okanagan k 158 166 61 175 187 196 196 1J.1J9£ 3 3 0 2 2 3 30/A 1.05 1.07 1.31* 1.16’ 1.15* 1.11 1.0115 Penticton Ic 97 182 188 192 116 122 218 13]YT£ 0 3 4 2 0 0 40/A 1.19* 1.07 0.87 1.15* 1.26* 1.17* 0.9516 Keremeos Ic 158 166 171 175 187 196 196 1T1T£ 3 3 3 2 2 3 30/A 1.05 1.07 0.97 1.16* 1.15* 1.11 1.0128 Quesnel Ic 180 194 201 201 206 208 199 )]V£ 6 2 2 5 2 3 30/A 0.95 1.16* 1.23* 0.99 1.20* 1.10 1.1031 Merritt Ic 153 23 157 168 175 167 23 0]JW£ 3 0 1 3 3 1 00/A 1.08 1.69* 1.15* 1.09 1.06 1.18* 1.59*38 Richmond Ic 1254 1304 1332 1328 1432 222 351 ]30T£ 2 2 2 3 3 0 10/A 1.08* 1.11* 1.09* 1.02 1.00 1.12* 1.12*39 Vancouver Ic 1000 1028 1042 1033 1117 1742 1751 O]3T£ 0 0 0 0 0 3 40/A 1.09* 1.16* 1.13* 1.06* 1.08* 1.02 0.9840 New Westminster Ic 110 113 595 113 645 682 685 •1JIT£ 0 0 3 0 3 3 30/A 1.29* 1.20* 0.94 1.30* 0.97 0.98 1.0041 Burnaby Ic 305 1424 1450 1442 1561 1656 1668 1YOT£ 0 2 2 2 3 3 30/A 1.12* 1.10* 1.10* 1.06* 1.02 1.00 0.9642 Maple Ridge Ic 339 362 89 218 408 428 425 1J57£ 3 3 0 1 3 3 30/A 0.93 1.02 1.24* 1.16* 1.04 1.01 1.0343 Coquitlam Ic 547 272 595 597 645 682 685 1t5£ 2 1 3 3 3 3 30/A 1.10* 1.14* 0.94 0.95 0.97 0.98 1.0044 N Vancouver Ic 1161 1201 1222 200 1312 1523 1534 OZT£ 1 1 1 0 1 3 30/A 1.09* 1.15* 1.11* 1.23* 1.08* 1.02 0.9645 W Vancouver Ic 1260 1306 1332 316 1436 1523 1534 1TO]E£ 2 2 2 1 2 3 30/A 1.07* 1.14* 1.09* 1.15* 1.07* 1.02 0.9646 Sunshine Coast Ic 300 318 331 338 62 64 63 WT7£ 3 3 3 3 0 0 00/A 0.96 0.95 0.97 0.98 1.25* 1.35* 1.30*56 Nechako Ic 131 129 133 144 136 150 144 1YT£ 4 1 1 3 1 3 30/A 0.92 1.25* 1.38* 0.97 1.27* 1.16 1.1057 Prince George Ic 153 109 112 168 114 174 167 13]3£ 4 0 0 4 0 2 20/A 0.93 1.36* 1.41* 0.97 1.31* 1.14* 1.14*59 Peace River S Ic 165 176 181 180 185 188 180 13]JW£ 4 2 2 4 2 3 30/A 0.88 1.20* 1.20* 0.95 1.14* 1.03 1.0560 Peace River N Ic 165 176 181 180 185 188 180 1T11T£ 5 2 2 5 2 4 40/A 0.89 1.20* 1.20* 0.95 1.14* 1.02 1.0568 Nanaimo Ic 264 280 292 298 315 325 183 1T5W£ 4 4 4 4 4 2 00/A 1.03 0.89 0.94 0.95 0.95 1.11* 1.26*153C Cl 0Cd 0 0 Cl) C) C’ 0cJqCD I. CD CD CD0 0 0 0 C) I C’ 0CCCCJC0-C-C)c%-->-I—ID-:C 0cxC)1CipI-.00cC’000**.I—)chC’0-4C)Ct’.,I- C’cCI- C’ 0900*-:-—iC’C’ 0pC)C’tZI—’C’I—. *C’CCCI—CCD CD CD CD CD 0 Cl) Cl)CD CI)C S 00 cxIs:’_*ICD0 C IC0CCII-C’OC)D< ,—o<‘-I——8C)C00.C)D -‘b?----I—II—’::.zc.n3C)t00I—’C1I—’Il-I00I—’.•-‘iI—I‘•C)CC)C1-4t:’C—._I—I-zcCi.)•I-4-I—•I—’I.-’-****0!•‘:‘——‘I—ICCCCeeeCI--,CD S CD CD CD CD 0 Cl) Cl) CD CA 0 0 0 CC 00 00 CCcCCI0CC)00 C)->-I—’Is:’Ii—I.,.-I5:’àoIs:’C IC?’..ICCCCO0CD I I. S CD CD CDCD S CD CD CD IboC)C C? CI- C’ 0 C’ C’I— •C’ 000 C,:’_*C?*0IS:’I0)•‘C’ 0-40C C 0I- C’ 0 C’ C)C (C0 0 0C,’C”C0IsII-C—CCCIC,.—DL•I,,00090(•CflCj00+—00,2?2222-----03IcD!bo’è,’LC*..CD03I-I—I.cào0-.1.CD-00*00CDCD,—I0**CI—-000CDCCT!Is00CDIs.0CT!0**CI—IC00I—’03C,IsZIs-Is!D0‘D0CD0CCCCee0000CD CD CD CD I.I-OCI I- 0!0 ,+— aq cq 0 22 -- 0! 0 0303 I CD0 0! .C,’ C, 03 0cj -0 T!CC àcC)bC CD- CD— 03C 0 Is’ —0 0303 C C CCD I C CD CD CDCD I’ CD CD CDC-0c : rCI3c ‘-I0-- II_I03C30bCOCT’C00ISZ—15309 I-CCCD CD CD CD.,I-••0!CT!I-0!—C,03 *CCT!i_ 0! 0!C CI C’ 0 C,CD I. I. CD CD CDCD I CD CD CDCD I. CD CD CD I0—C C—ce,)c00—CDr,———C•C02 ----DC000C)ZC)I—I—’C)‘--‘..c.rc.C)’cooccC)C)’00I0—cj,coooo0)CDC)I- 0)tC)) C)CCCDC)’tsCsZC)zC.C—C)I—C.)—1)CD.‘0.——.I.CC)’-C)0)00CCCCCCCC)C)C)C)i-CC 0CDD C -•..CDEa.,-— +C ÷ CD2222--).--‘I—I—00C)3I—’C)I—I—’•ISO00I—.:••__0C)30C CI—:I1ISCOC)100C)0)C..I—.•‘CO****...I—Cc CI***—CII—I—C)’ISZ00CDCDCDCCCCeee—C0C)01—CCISI—C0-CC<cc 0C0CoC0J)CD+—2222222-)--.-:--I—IiI.,co0)0)CDIS) 0I-00**IS:)IS)C)’ISC)IS) C1C)C)) 01CC))C)i—‘,:)iS:)?C)) C)IS) CIC)C)’010—C)C)*CD—**:I—’i—i—i—‘i—’0)00)-C)-èYIISIS)00CDCDC))—CDC))cIs:)Is)Is)Is)Is)IS)Is)CI—’CCIC))—i—0*****II—1)—ICb--——I—C—0):0C)’I-1,I—iI0)0C)’-CDCDC))C))C)’•C)CCCCCCCèèèèC)CCI.-CCCD I. CD CD CDCD CD CD CD0——CCI—CC—CO tLCC‘op,,oC0OCDCD—;. CDo,-— ,-22222222-)--:-—---I—I—I•bo-C.’CC***:.-IsZjàoàooc,Co***I—D’0’CoC.3C—*::..I—iI—.-.4I—._,I—****I—’-‘I—’I—)—:•-‘-.cj—4-4Co*CcCcCCee-ecs-IsCC0c’ccjC-CO..‘CDCD•1Q’Q’—CDCDCDCD1 CDCDCD0CO.o_rj2222>-—.-I—=i.-ICo-.C—cx’C.’Cc‘i—i—boC’1.àoQC*‘I—IsCoC—4C.C10.4Ccg--1*CI— ..C)’ 0‘X’Cc.C.”—Cc-4C.)’CIs.:cIsè,4--C.3—CCC)jCcCIs0-I—Ix0-4o.00.‘•‘0i—I‘ICoC—iC0C’I—’CCCCCeee‘Is0V IC’ 0 000I C’ 0.I*CCD I. I. CD CD CD CD ‘I0CI—:‘•0--C o----I—!“—‘I--C***I—C’C***C,-’*C.,C.3LZ-0-****C-‘:-Ii—I-o’boboCào0&,C’..I-.’CDCCCCCCeCCCCCD 00 CD CD CD CD0--—CD-CCCC.CCCClCo0,,J0—,;C-+c.,-—0. .+0 E?--.-----I—I—I3CDC?CCDCC,.‘—:IàoCpD’CDC7CDCc—c--C-40I—’:‘:-ICC.I—I-.’I—’C,30‘—-)—I0101C:-i—I—pIi-4àoCC’C,z:‘CD*CCCCCCeeeDeC0CC’CC’CD CD CD CD CD0 0 0 CCD—C’*C C CI. C,’CD I. 9, CD CD CD 9, CD CD CD CD CD C 9, cC ccCD 9, CD CD CD 9,CD I. 9, CD CD CD I-CCC?CI.-1.-C)—CCC.CC,•‘:pci0CDCDCDZCD-OZ.—CDCDCD p9,ccaqCDCDQ,.2222222-------II-I-I.I,’,o1.—CO0C?CT****C?C?3-C?.1D.1cv00-00-C?I—IsZI1.aoC-CCI—:bCC00CàoC7’C?’C70000CC00c:CC?C?CL,CCC?CZCaol9àobo00C?ccccccCC.1.-CC?CCCCC?C?C?C?CjC-CDCD9’Or9’2 ---I—I**C?I—ICC—1**I--.4:1.-bo’4boC?’-4I.IS.sCC?2I——1=I‘00C.I—_C?**00CCCCC11.CC?00C?CJp::9,CC - CD---— L1LjCC?C?C?—I0oIs.C—‘I—I..1C?C*I1.C—C?*—IC?’C?CCCt..CC*-I—I00C..li.C?CC0I—IC CCC?’CC?0 C e C1 ©- C ‘.- ch2%a2a2---I—.‘)-00C.•: :Icz**C.—-—-a*C3*DcrCCCC0Cci 0 0 ICD tz CD CD CD CD CD ccl C cc 00 00 ccCC0Cc,<D.<Q’c’ C, 0+CCC,+222a22a->.--,--I—I—II-I Cc-.4I-CIi—’I.-’•:.4—1CC1CCcCclLCCC.•CCC0Cc*****Cc—CCcI—’Ii—.•IDCC)Cc00Cl-4-D-I—CCC.I—C’--CCCCCC-.èc<ISCCCI—CcjoCC,ci0oCCCC**CCCCi_**‘I—I00CCOCc IccCC I00LcIC——iCàoC-0CD I. S CD CD CD e-clCD I. •ISZcc S CD CD CD CD clCD . 00 S CD CD CD CD cclI—ICCC I-—0-.4C’ Cc* CC_*Cc I C’ 0 C’ C)I— -.4C I.. CCOC ISI C’ 0 C)I’CD I. C)-’CD CD CDCD I3CD CD CD—13C?C?ClCI—0CE c .,0Eo0C 0 .2222222-------I—’..l.-:1C?•C?’13C.I—i:-I0--4C.*13I—C?1l—I—I—iI—-***C?l—3•-IbC.Z<0I—*I—’lI—’:-I—••I-CC)-.*—l,3:l—C)•C)è.z()C‘j**:-:i—C)113àOC***C?’CCCCCC.‘ets0I-CCC-1313C.CCICC)13C.OC:eqD0<—‘EN—‘‘0.( l022222--.-->-I—’•C?C?C?ZIDc?IIII-I—jc.CJI-*****13C?C?413LjC?LCc00CC?C?C.C.OC.30C)1CCCC‘—C‘C?C:-••I-II—130C.3CC.CC?13-1S1cCCCC13C13cCCCI-C.0C.CJCCtD:CoCD0‘O:;CDoCDCD C. 00÷÷ CD—.C-----t,..,.II—00Iec1I—II—’-**.-I•-*CI—..I**—ào°°00 I—Ioccc.ccC-1CtSc‘.01..ct Ct,Table 11.38: Female General Yearly Results Site 38 All Cancers Except Lung 1983—89School District 1983 1984 1985 1986 1987 1988 198913 Kettle Valley 9 9 65 63 67 75 76 038£ 0 0 3 3 3 3 30/A 2.73* 2.43* 0.85 0.86 1.03 0.97 1.0632 Hope Ic 104 109 119 115 121 135 136 T57£ 2 2 4 4 4 4 40/A 1.21* 1.21* 0.85 0.97 1.07 1.04 1.0135 Langley Ic 92 427 111 477 442 130 586 O]3T£ 0 3 0 3 1 0 30/A 1.28* 1.04 1.27* 1.03 1.08* 1.16* 1.0736 Surrey Ic 432 460 517 507 435 606 612 ]YTiT£ 2 3 3 3 1 2 20/A 1.10* 1.02 1.00 0.98 1.13* 1.08* 1.07*37 Delta Ic 321 103 116 277 386 431 429 1TOT£ 3 0 0 1 3 3 20/A 1.02 1.21* 1.18* 1.17* 1.08 1.04 1.09*38 Richmond Ic 1043 1087 282 175 1185 1317 1303 ]T57£ 3 3 1 0 3 3 30/A 1.06 1.02 1.13* 1.16* 1.01 1.03 1.0239 Vancouver Ic 974 1264 925 1323 1364 1510 1489 ]flT£ 1 3 0 3 3 2 30/A 1.06* 1.05 1.10* 1.04 1.04 1.05* 1.0140 New Westminster Ic 346 491 541 525 109 435 599 ]TTW£ 1 3 3 3 0 1 30/A 1.15* 1.00 0.99 0.97 1.32* 1.11* 1.0341 Burnaby Ic 260 1199 1298 1248 1286 329 1404 lflV£ 0 3 2 3 3 0 30/A 1.17* 1.00 1.08* 1.01 1.01 1.11* 1.0044 N Vancouver Ic 164 171 1097 1153 1190 1318 1301 ]J]JT£ 0 0 1 3 2 3 30/A 1.15* 1.26* 1.09* 1.04 1.05* 1.05 1.0145 W Vancouver Ic 253 107 1198 1153 118 1318 1301 1TOT£ 1 0 2 3 0 3 30/A 1.11* 1.20* 1.09* 1.04 1.18* 1.05 1.0176 Agassiz-Harrison Ic 94 109 119 115 121 135 136 T15W£ 1 2 4 3 3 3 30/A 1.23* 1.21* 0.87 0.95 1.06 1.06 1.00 =Table 11.39: Female General Yearly Results Site 39 Endometrium 1983—89School District 1983 1984 1985 1986 1987 1988 198938 Richmond Ic 69 15 90 84 25 100 86 1JO£ 3 0 2 3 1 3 30/A 1.04 1.75* 1.22* 1.07 1.67* 1.06 0.9945 W Vancouver Ic 10 81 91 12 12 101 87 1Y0T£ 0 3 2 0 0 3 40/A 2.25* 1.03 1.22* 2.53* 2.24* 1.10 1.0248 Howe Sound Ic 11 24 27 13 13 30 26 1YTW£ 1 4 4 1 1 4 40/A 1.99* 1.06 1.23 2.35* 1.96* 1.11 0.8957 Prince George Ic 7 12 9 12 9 14 13 IflIT£ 0 13 0 15 0 14 50/A 2.15* 0.77 1.97* 0.54 2.22* 0.83 1.1164 Gulf Islands Ic 14 17 5 11 19 22 20 IJTY£ 4 3 0 1 3 3 30/A 0.79 1.06 3•37* 2.13* 1.34 1.08 1.21 =163Table 11.46: Female General Yearly Results Site 46 All Cancers 1983—89School District 1983 1984 1985 1986 1987 1988 1989 j3 Kimberley k 68 71 77 76 25 27 84 TEW£ 3 3 3 3 0 0 30/A 1.11 0.94 0.87 0.90 1.48* 1.46* 0.8013 Kettle Valley 9 9 71 70 75 84 85 O5£ 0 0 3 3 3 3 30/A 2.89* 2.15* 0.85 0.88 0.99 0.97 1.0232 Hope k 115 17 130 128 135 150 152 1J37£ 2 0 4 4 4 4 40/A 1.19* 1.60* 0.88 1.01 1.04 1.04 1.0235 Langley k 101 477 120 530 565 638 662 OTW£ 0 3 0 3 2 3 30/A 1.25* 1.03 1.29* 1.02 1.08* 1.05 1.0736 Surrey k 478 514 565 564 484 672 551 •O]W£ 2 3 3 3 1 2 10/A 1.10* 1.01 1.02 0.98 1.13* 1.08* 1.08*37 Delta k 355 114 307 306 428 476 476 02T1£ 3 0 1 1 3 3 30/A 1.03 1.18* 1.11* 1.14* 1.07 1.05 1.0738 Richmond k 1154 1214 307 193 1319 1463 1464 1tW£ 3 3 1 0 3 3 30/A 1.06 1.01 1.11* 1.15* 1.01 1.04 1.0139 Vancouver ic 1078 1413 1014 1474 1520 1679 1663 ltW£ 1 3 0 3 3 2 30/A 1.06* 1.04 1.09* 1.04 1.04 1.04* 1.0040 New Westminster k 383 548 592 583 121 483 667 0.18£ 1 3 3 3 0 1 30/A 1.16* 1.00 1.01 0.98 1.33* 1.12* 1.0341 Burnaby k 287 1340 1422 1392 1435 365 1569 0.20£ 0 3 2 3 3 0 30/A 1.17* 1.00 1.08* 1.02 1.02 1.12* 0.9944 N Vancouver ic 1078 190 1201 1285 1326 1466 1453 0.04£ 1 0 1 3 2 3 30/A 1.06* 1.24* 1.08* 1.04 1.05* 1.04 1.0045 W Vancouver ic 280 297 1312 1285 328 1466 1453 ]3]3T£ 1 1 2 3 1 3 30/A 1.11* 1.21* 1.08* 1.04 1.12* 1.04 1.0076 Agassiz-Harrison Ic 104 121 130 128 135 160 152 1337£ 1 2 3 3 3 3 30/A 1.20* 1.20* 0.88 0.99 1.04 1.06 1.02164Table 12a: Male Focused Yearly Results by Site 1983—89Site No. Mill Location 1983 1984 1985 1986 1987 1988 19892 Castlegar 1 3 8 7 7 7 3 3 0.12£ 1 12 12 11 11 1 1Q/\ 449* 0.65 0.72 0.72 0.98 3•73* 4.31*Port Mellon 1E 41 43 39 39 41 49 4 flUW£ 3 3 3 4 3 5 10/A 1.78* 1.85* 1.48* 1.19 1.36* 1.04 3.11*Prince Itupert U 4 3 6 6 7 6£ 11 2 1 11 10 14 100/A 0.70 3.15* 4.41* 0.71 0.82 0.59 0.994 Prince Rupert 1E 6 6 6 4 3 6 6 1J]JP£ 9 11 7 2 1 10 100/A 1.29 0.76 1.63 359* 4.38* 0.85 1.20Port Alice 6 5 6 5 6 U -rr:T3-£ 6 4 5 3 3 6 40/A 1.11 1.38 1.16 2.61* 2.64* 1.14 1.22Ocean Falls 5 5 5 5 5 3 1J7j£ 6 11 8 4 4 2 110/A 1.55 1.08 1.45 2.22 2.37 5.89* 0.826 Nanaimo 18 22 23 15 24 15 15 13]JW£ 8 6 7 1 6 1 10/A 0.82 1.01 0.94 1.66* 1.12 1.97* 1.75*Gold River 10 12 13 13 13 7 13 UTW£ 6 6 6 6 7 2 30/A 1.10 0.96 0.99 1.25 0.78 2.94* 1.87*Campbell River 13 15 16 lii 16 6 13 WTT£ 6 6 6 5 7 1 20/A 1.13 1.09 0.96 1.32 0.99 2.74* 1.78*7 Port Mellon 1 17 15 16 20 18 20 20 OW£ 4 3 3 5 3 3 30/A 1.43 1.65* 2.10* 1.30 1.72* 1.51* 2.14*10 Powell River 16 16 16 12 12 15 15 13]3W£ 4 5 4 2 2 4 40/A 1.17 1.09 1.62 1.92* 1.95* 1.26 1.59Gold River 1E 13 13 14 13 13 13 12 U1T£ 7 6 5 3 3 6 60/A 0.97 1.10 1.28 2.24* 1.83* 1.19 1.06Campbell River 16 16 16 13 13 15 15 UUT£ 6 6 5 2 2 5 60/A 1.01 1.13 1.27 2.28* 1.87* 1.25 1.4211 Crofton 1 3 3 2 3 3 2 2 IJTJW£ 10 9 10 5 5 3 150/A 0.74 0.96 0.86 5.08* 3.84* 6.14* 1.15Port Alberni 2 3 2 3 3 2 2 UU£ 2 7 15 13 9 4 160/A 6.49* 1.88 0.95 2.40 2.61 6.11 1.78?anaimo 1 3 4 2 2 3 2 2 WUW£ 9 9 9 1 5 1 120/A 0.96 1.02 1.13 6.69* 4.10* 9.51* 1.90Powell River ic 3 3 2 2 2 2 2 0.05t 4 7 13 7 7 7 140/A 5.25* 1.75 1.06 2.09 2.10 4.53 1.9512 Gold River k 10 10 10 11 10 11 9 0.10£ 8 6 4 6 3 4 70/A 0.95 1.03 1.70 1.18 1.88* 1.81* 0.92Campbell River ic 12 12 12 13 12 10 11 0.31£ 8 6 6 6 4 2 80/A 0.94 1.08 1.16 1.32 1.74* 1.86* 0.97165Table 12b: MaJe Focused Yearly Results by Site 1983—89Site No. Mill Location 1983 1984 1985 1986 1987 1988 198913 Campbell River k 8 8 9 9 9 4 7 0.24£ 8 7 7 8 6 1 20/A 0.96 1.00 0.94 0.98 1.38 2.97* 2.49*14 Prince George k ii 13 12 8 9 13 9 O]iT£ 15 5 5 1 1 5 10/A 0.68 1.29 1.26 2.80* 2.23* 1.21 1.96*Kitimat k 7 7 7 6 7 3 7 UT3£ 9 9 9 6 3 1 80/A 0.79 1.34 0.82 1.80 2.22* 5.04* 1.34Kamloops k 16 18 17 17 12 12 20 fl3W£ 6 7 6 7 1 1 60/A 1.10 0.87 0.97 0.83 2.00* 1.74* 1.22Quesnel k 13 15 14 13 15 15 15 •U]W£ 12 9 6 4 3 9 30/A 0.65 1.12 1.44 1.65 2.07* 1.35 1.71*MacKenzie k 11 12 12 8 9 13 9 U3£ 17 4 9 1 1 6 10/A 0.73 1.72 1.15 2.80* 2.23* 1.21 1.96*15 Powell River A 37 40 30 38 30 31 38 ]5W£ 4 4 2 4 2 2 40/A 1.14 1.01 1.40* 1.20 1.41* 1.45* 1.25Kamloops k 48 52 34 50 31 32 49 UU5£ 6 6 2 6 1 1 50/A 0.99 0.89 1.41* 1.20 1.48* 1.55* 1.18Gold River k 30 32 14 31 32 33 14 UUW£ 6 5 2 4 3 3 20/A 0.80 1.09 1.77* 1.30 1.46* 1.48* 1.72*Campbell River k 37 40 14 30 31 32 38 OUT£ 6 5 1 2 2 2 50/A 1.02 1.19 1.89* 1.41* 1.45* 1.52* 1.1916 Powell River k 33 35 34 33 35 27 33 13]W£ 4 4 3 3 3 2 40/A 1.23 1.05 1.49* 1.41* 1.43* 1.55* 1.14Kamloops k 43 46 30 43 27 27 42 WDW£ 6 6 2 6 1 1 50/A 0.97 0.88 1.43* 1.15 1.50* 1.53* 1.19Gold River k 26 28 13 27 28 28 27 WDr£ 6 5 2 3 3 3 50/A 0.84 1.11 1.96* 1.61* 1.49* 1.46* 1.04Campbell River k 33 35 12 27 27 28 33 WDr£ 6 5 1 2 2 2 50/A 1.11 1.20 2.09* 1.65* 1.47* 1.49* 1.0617 Port Mellon k 20 22 24 22 22 26 32 TJTJW£ 4 4 3 3 3 3 30/A 1.44 1.32 1.70* 1.56* 2.47* 2.24* 2.16*166Table 12c: Male Focused Yearly Results by Site 1983—89Site No. Mill Location 1983 1984 1985 1986 1987 1988 1989 j520 Prince George k 30 21 35 37 27 28 32 1Y]3T£ 5 1 5 5 1 1 20/A 1.21 2.61* 1.18 1.06 2.01* 1.43* 1.55*Port Mellon A 252 260 284 49 58 21 398 WU2£ 4 3 4 2 2 1 50/A 1.02 1.12* 1.04 1.48* 1.32* 2.01* 0.93Powell River 37 31 14 46 53 55 42 W1T£ 5 2 1 4 5 4 20/A 0.80 1.52* 1.78* 0.92 0.72 1.11 1.54*Squamish 63 67 76 40 97 105 107 UOT£ 3 4 3 2 4 3 40/A 1.25* 1.17 1.34* 1.45* 1.23 1.35* 1.10Kaniloops 48 30 37 35 69 72 69 1JUW£ 3 1 2 1 6 4 30 A 1.34* 1.44* 1.41* 1.45* 0.77 1.20 1.26*Quesnel 34 36 41 44 49 27 13 UUW£ 5 3 4 6 3 2 10/A 1.23 1.99* 1.19 1.18 1.33* 1.42* 1.84*MacKenzie 29 21 34 35 27 28 32 UUW£ 4 1 3 4 1 1 20/A 1.12 2.61* 1.41* 1.13 2.01* 1.43* 1.55*21 Powell River 6 3 5 4 5 6 7 13]3W£ 5 1 7 2 12 9 50/A 1.85 6.98* 1.05 3.23* 0.93 1.22 1.74Port Alice 4 4 4 4 4 3 5 lflW£ 8 6 4 4 15 1 40/A 1.56 2.29 1.97 2.37 0.65 9.46* 1.79Campbell River 6 6 5 5 5 6 7 U£ 6 4 8 3 13 11 50/A 1.78 2.81* 1.03 2.54* 0.89 1.25 1.9322 Port Alice ••• 7 7 6 7 6 7 3 JJW£ 6 3 3 3 4 5 10/A 1.02 2.60* 2.39* 3.04* 1.03 1.10 5.96*Gold River k 11 6 5 6 5 11 6 UUW£ 7 2 2 2 2 7 20/A 1.02 333* 2.54* 335* 2.59* 1.09 3.03*Campbell River 14 5 12 5 5 13 s irur£ 6 1 5 1 1 6 10/A 0.99 3.58* 1.64 3.05* 2.72* 1.13 3.22*23 Kitimat 5 6 5 5 5 6 3 iY£ 9 12 9 9 9 3 10/A 0.96 0.90 1.26 1.35 1.16 2.80* 7.16*Ocean Falls 1 4 5 4 4 4 3 3 -tJ:U7-1 12 11 9 11 5 2 20/A 0.65 0.86 1.15 1.15 1.72 5.19* 534*26 Port Alberni 12 12 13 6 13 15 14 1TTT1 6 4 4 1 3 7 50/A 1.12 1.25 1.14 2.88* 1.80* 0.90 1.02Powell River 4 11 11 12 9 13 12 1TIW1 1 3 5 5 2 7 60/A 3.13* 2.08* 1.07 1.20 2.38* 0.78 1.1027 Prince George 5 4 4 5 5 5 5 13]371 21 1 2 19 10 21 60/A 0.85 3.48* 347* 0.72 1.62 0.56 1.99MacKenzie —k— 5 4 4 5 5 5 5 13]Jr1 26 1 2 21 13 26 62L.. 0.67 3.48* 347* 0.69 1.41 0.54 2.19167Table 12d: Male Focused Yearly Results by Site 1983—89Site No. Mill Location 1983 1984 1985 1986 1987 1988 1989 E28 Nanaimo k 13 15 12 13 14 13 9 1J1J4£ 5 5 8 5 3 5 10/A 1.83 1.21 0.81 1.45 1.67* 1.45 2.08*30 Skookumchuck k 5 5 5 5 5 4 5 1OW£ 8 9 9 5 13 2 30/A 1.42 1.10 1.05 1.83 0.70 3•93* 3.29*32 Port Mellon Ic 79 77 91 89 92 90 105 1YOY£ 5 3 3 4 3 3 40/A 0.96 1.22* 1.26* 1.14 1.22* 1.21* 1.11Powell River k 6 12 17 17 17 16 18 WW£ 1 2 6 6 6 7 60/A 2.76* 2.08* 0.87 0.88 0.81 0.67 0.92Quesnel k iti 15 18 19 11 5 20 1J])T£ 9 7 3 9 2 1 90/A 1.04 1.29 1.94* 1.15 1.96* 2.69* 1.1236 Nanaimo Ic 41 28 50 49 51 27 30 1J]5W£ 7 1 7 5 5 1 10/A 0.81 1.60* 0.89 1.10 1.10 1.47* 1.78*Gold River 21 24 25 25 26 11 25 U7£ 6 6 5 5 6 2 30/A 0.96 0.91 1.10 1.08 0.99 1.99* 1.55*Uampbell River 1 27 30 31 31 32 lii 24 UW£ 6 6 5 5 6 1 20/A 0.95 0.89 1.07 1.05 0.99 1.92* 1.47*38 Prince George k 127 91 94 142 97 147 143 1TO£ 5 1 1 5 1 4 40/A 0.96 1.37* 1.32* 0.93 1.28* 1.15 1.11Port Mellon 931 95 985 990 53 54 53 13UW£ 4 3 3 3 1 1 10/A 1.07 1.14* 1.10* 1.06* 1.30* 1.43* 1.41*Quesnel k 150 161 167 170 174 175 170 U]J5£ 6 3 3 7 3 4 40/A 0.97 1.17* 1.14* 0.94 1.18* 1.11 1.08MacKenzie 1 125 91 94 138 97 144 140 1JT£ 5 1 1 5 1 4 40/A 0.94 1.37* 1.32* 0.89 1.28* 1.02 1.0646 Prince George 153 109 112 168 114 174 167 IJOr£ 5 1 1 5 1 3 30/A 0.93 1.36* 1.41* 0.97 1.31* 1.14* 1.14*Port Mellon k 1142 1180 1201 1194 62 64 63 UUW£ 3 3 3 3 1 1 10/A 1.06* 1.14* 1.10* 1.05* 1.25* 1.35* 1.30*Nanaimo F 264 280 292 298 315 325 183 U5W£ 5 5 5 5 5 3 10/A 1.03 0.89 0.94 0.95 0.95 1.11* 1.26*Quesnel Ic isO 194 201 201 206 208 199 00T£ 7 3 3 6 3 4 40/A 0.95 1.16* 1.23* 0.99 1.20* 1.10 1.10MacKenzie k 149 109 112 163 114 170 164 TT£ 5 1 1 4 1 4 40/A 0.90 1.36* 1.41* 0.92 1.31* 1.03 1.05168Table 13a: Female Focused Yearly Results Site 1983—89Site No. Mill Location 1983 1984 1985 1986 1987 1988 19892 Port Mellon 5 29 30 23 28 21 30 OU5£ 2 4 4 5 3 6 60/A 2.60* 1.24 1.28 1.31 1.61* 0.97 0.904 Crofton k 7 8 9 8 5 9 9£ 8 8 8 3 1 9 7OR 1.02 1.48 0.83 2.18* 3.26* 0.46 1.175 Port Mellon 7 8 123 122 113 123 129 TU£ 1 1 5 3 5 6 5OR 2.54* 2.40* 0.93 1.18* 1.00 0.96 0.977 Port Mellon k 10 12 11 13 12 14 16 1J]3T£ 12 5 5 6 7 3 5OR 0.97 1.48 1.52 1.31 1.17 1.77* 1.418 Castlegar 6 4 6 5 6 7 5 TO£ 12 2 11 3 6 11 2OR 0.83 3.12* 1.05 2.63* 1.89 0.67 333*9 Kamloops 3 3 2 2 3 2 3 1J]J£ 33 20 1 3 20 1 7OR 0.61 0.88 943* 8.59* 0.95 6.97* 1.3410 Nanaimo 9 6 8 10 7 7 11 13]3W£ 7 1 5 5 1 1 7OR 1.33 2.72* 1.61 1.57 2.98* 3.24* 0.95Powell River 6 3 5 7 7 7 7 UUW1 3 1 2 6 4 7 4OR 2.91* 5.31* 3.13* 1.29 1.74 1.61 1.41Gold River k 3 5 5 6 6 6 6 UU11 2 5 4 7 3 8 7OR 4.51* 2.92 2.75* 1.08 2.85* 1.75 1.29Campbell River k 3 6 6 7 6 7 7 UUW1 1 3 3 7 2 8 60/A 4.68* 3.13* 2.75* 1.29 2.89* 1.61 1.3313 Powell River k 5 6 6 6 6 7 3 1YOT1 4 4 9 5 2 7 10/A 2.12 1.80 0.69 2.12 2.93* 1.31 4.17*Gold River k 4 5 5 5 6 6 6 0UF1 5 4 8 6 3 6 50/A 2.20 2.59* 0.80 2.06 2.73* 1.54 1.80Campbell River k 5 6 6 6 6 7 7 UU51 5 6 12 5 2 8 60/A 2.21 1.45 0.66 2.15 2.77* 1.30 1.6014 Prince George k 7 6 6 7 5 8 9 1JDT1 7 5 6 6 1 7 70/A 1.40 1.64 2.10 1.75 2.91* 1.22 1.14Port Mellon k 46 4 3 47 45 54 59 1ETT1 6 1 1 6 4 5 50/A 0.99 3.96* 3.72* 0.90 1.17 1.05 1.00MacKenzie k 6 6 5 7 5 8 9 WtJT1 5 6 6 6 1 8 90/A 1.72 1.79 1.66 1.72 2.91* 1.24 0.97169Table 13b: Female Focused Yearly Results Site 1983—89Site No. Mill Location 1983 1984 1985 1986 1987 1988 198915 Crofton 1E 24 28 25 29 31 15 16 0.46£ 5 6 5 5 5 1 10/A 0.89 0.86 1.01 1.12 1.02 1.64* 1.63*Port Mellon 1 117 8 8 133 138 149 153 075£ 4 1 1 5 4 5 50/A 1.04 2.34* 2.63* 1.00 1.07 0.99 0.93Fort Alberni 22 25 23 18 10 10 31 UU7£ 4 4 4 2 1 1 40/A 1.34 1.15 0.93 1.65* 2.08* 2.35* 1.00Powell River 18 8 19 21 22 24 9 WDT£ 3 1 5 4 3 5 10/A 1.56* 2.70* 0.93 1.33 1.55* 1.08 2.20*Gold River 15 17 16 17 18 19 21 WUW£ 3 3 6 5 3 6 50/A 1.82* 1.85* 1.04 1.19 1.84* 1.34 1.28Uampbell River 15 17 19 21 18 24 25 WTJT£ 2 2 6 5 2 5 50/A 1.74* 1.88* 0.93 1.27 1.86* 1.31 1.1716 Crofton 21 24 22 25 26 13 14£ 5 6 5 5 6 1 10/A 0.95 0.82 0.91 1.08 0.91 1.72* 1.74*Port Mellon 1E 102 7 7 115 115 127 129 •OTW£ 4 1 1 4 4 5 50/A 1.06 2.52* 3.24* 1.03 1.12 1.03 0.97Port Alberni 19 21 19 23 8 9 27 U3£ 4 4 4 4 1 1 40/A 1.38 1.10 0.96 1.32 2.31* 2.13* 0.87Gold River 13 15 13 15 15 17 18 00W£ 3 3 6 5 3 6 50/A 1.86* 1.76* 1.06 1.16 1.62* 1.30 1.17Campbell River 13 14 16 19 15 21 22 U5£ 2 2 6 6 2 5 60/A 1.90* 1.79* 0.96 1.26 1.65* 1.25 1.0918 Port Mellon 8 36 10 4 48 58 51 1JZ£ 2 6 2 1 4 5 60/A 2.43* 1.06 2.30* 375* 1.18 0.96 0.87Squamish 7 14 9 8 18 22 19 WOV£ 2 5 2 2 3 4 50/A 2.33* 0.89 2.52* 2.19* 1.56* 1.40 0.9125 MacKenzie 5 5 3 4 4 4 5 OW£ 22 28 2 3 26 23 280/A 1.05 0.62 3•99* 3.08* 0.82 0.99 0.7027 MacKenzie k 3 4 4 3 3 4 4 IflY£ 1 3 17 18 24 17 170/A 4.48* 3.82* 0.97 0.83 0.81 0.79 0.9928 Port Alice 4 5 4 4 2 4 5 iTrt 6 6 10 8 1 3 80/A 1.12 1.27 0.66 0.67 7.21* 373* 0.7729 Nanaimo k 4 4 5 4 4 5 5 0.18£ 1 1 9 9 9 8 80/A 3.98* 3.66* 1.34 0.99 0.70 1.46 1.07170Table 13c: Female Focused Yearly Results Site 1983—89Site No. Mill Location 1983 1984 1985 1986 1987 1988 1989 j531 Nanaimo k 16 14 15 11 19 17 19 03£ 6 8 3 1 7 7 70/A 1.14 0.75 1.70* 1.96* 0.98 1.04 0.9635 Port Mellon ic 46 31 43 39 35 43 44 0:0W£ 4 3 4 4 4 4 30/A 1.23 1.39* 1.26 1.16 1.29 1.12 1.73*Port Alberni Ic ii 8 10 9 6 7 10£ 14 9 4 4 2 2 140/A 0.48 0.79 1.21 1.09 2.50* 2.77* 0.6136 Kitimat Ic 10 11 8 11 11 11 11 1TOW£ 9 9 2 3 8 9 30/A 0.95 0.87 2.15* 1.98* 1.05 0.88 1.80*Port Mellon k 9 10 188 10 174 181 183 lflT£ 1 1 5 1 5 5 50/A 2.13* 2.17* 0.96 1.96* 0.94 0.95 0.9839 Prince George Ic 7 12 9 12 9 14 13 13111£ 1 14 1 16 1 15 60/A 2.15* 0.77 1.97* 0.54 2.22* 0.83 1.11Port Mellon k 12 72 82 15 15 90 78 0:0W£ 2 4 4 2 2 4 50/A 1.81* 1.07 1.21 2.44* 2.18* 1.11 0.97Squamish Ic 11 24 27 13 13 30 26 11T3£ 2 5 5 2 2 5 50/A 1.99* 1.06 1.23 2.35* 1.96* 1.11 0.89MacKenzie Ic 7 12 9 12 9 14 13 11IT£ 1 10 1 18 1 13 80/A 2.15* 0.84 1.97* 0.52 2.22* 0.73 1.0146 Port Mellon ic 1045 155 48 1138 1176 1301 1291 5£ 4 2 1 4 4 4 40/A 1.06 1.15* 1.34* 1.04 1.05 1.02 1.00Skookumchuck k 68 71 77 76 25 27 84 0:5T5 5 5 5 1 1 50/A 1.09 0.91 0.87 0.90 1.48* 1.46* 0.81171
- Library Home /
- Search Collections /
- Open Collections /
- Browse Collections /
- UBC Theses and Dissertations /
- Cancer cluster detection in British Columbia school...
Open Collections
UBC Theses and Dissertations
Featured Collection
UBC Theses and Dissertations
Cancer cluster detection in British Columbia school districts, 1983-1989 Rosychuk, Rhonda Jean 1994
pdf
Page Metadata
Item Metadata
Title | Cancer cluster detection in British Columbia school districts, 1983-1989 |
Creator |
Rosychuk, Rhonda Jean |
Date Issued | 1994 |
Description | A disease cluster is an aggregation of occurrences of a disease. The observa tion of a perceived excess number of similar illnesses is termed disease clustering. Statistical tests for disease clustering investigate if the observed pattern of cases in at least one geographical area could possibly have happened by chance alone. This pattern may be spatial, temporal, or both. Investigating possible cancer clusters in British Columbia for the period of 1983—1989 inclusive is the objective of this thesis. Whether or not cancer clustering appears near pulp and paper mills within the province is of specific interest. The geographical units upon which our investigation will be based are the B.C. school districts. The variation in size and population demographics among districts requires a cluster detection method which is considerate of the underlying population distribution within the study region. School district population size diversity requires a modification to the Besag and Newell (1991) method. This modification is implemented with B.C. school district data and several possible clusters are detected for various types of cancer. |
Extent | 4874060 bytes |
Genre |
Thesis/Dissertation |
Type |
Text |
FileFormat | application/pdf |
Language | eng |
Date Available | 2009-02-24 |
Provider | Vancouver : University of British Columbia Library |
Rights | For non-commercial purposes only, such as research, private study and education. Additional conditions apply, see Terms of Use https://open.library.ubc.ca/terms_of_use. |
DOI | 10.14288/1.0087354 |
URI | http://hdl.handle.net/2429/5008 |
Degree |
Master of Science - MSc |
Program |
Statistics |
Affiliation |
Science, Faculty of Statistics, Department of |
Degree Grantor | University of British Columbia |
GraduationDate | 1994-05 |
Campus |
UBCV |
Scholarly Level | Graduate |
AggregatedSourceRepository | DSpace |
Download
- Media
- 831-ubc_1994-0166.pdf [ 4.65MB ]
- Metadata
- JSON: 831-1.0087354.json
- JSON-LD: 831-1.0087354-ld.json
- RDF/XML (Pretty): 831-1.0087354-rdf.xml
- RDF/JSON: 831-1.0087354-rdf.json
- Turtle: 831-1.0087354-turtle.txt
- N-Triples: 831-1.0087354-rdf-ntriples.txt
- Original Record: 831-1.0087354-source.json
- Full Text
- 831-1.0087354-fulltext.txt
- Citation
- 831-1.0087354.ris
Full Text
Cite
Citation Scheme:
Usage Statistics
Share
Embed
Customize your widget with the following options, then copy and paste the code below into the HTML
of your page to embed this item in your website.
<div id="ubcOpenCollectionsWidgetDisplay">
<script id="ubcOpenCollectionsWidget"
src="{[{embed.src}]}"
data-item="{[{embed.item}]}"
data-collection="{[{embed.collection}]}"
data-metadata="{[{embed.showMetadata}]}"
data-width="{[{embed.width}]}"
async >
</script>
</div>
Our image viewer uses the IIIF 2.0 standard.
To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.831.1-0087354/manifest