UBC Faculty Research and Publications

Dietary patterns in Canadian men and women ages 25 and older: relationship to demographics, body mass… Langsetmo, Lisa; Poliquin, Suzette; Hanley, David A; Prior, Jerilynn C; Barr, Susan; Anastassiades, Tassos; Towheed, Tanveer; Goltzman, David; Kreiger, Nancy Jan 28, 2010

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

Item Metadata

Download

Media
52383-12891_2009_Article_766.pdf [ 449.23kB ]
Metadata
JSON: 52383-1.0223762.json
JSON-LD: 52383-1.0223762-ld.json
RDF/XML (Pretty): 52383-1.0223762-rdf.xml
RDF/JSON: 52383-1.0223762-rdf.json
Turtle: 52383-1.0223762-turtle.txt
N-Triples: 52383-1.0223762-rdf-ntriples.txt
Original Record: 52383-1.0223762-source.json
Full Text
52383-1.0223762-fulltext.txt
Citation
52383-1.0223762.ris

Full Text

RESEARCH ARTICLE Open AccessDietary patterns in Canadian men and womenages 25 and older: relationship to demographics,body mass index, and bone mineral densityLisa Langsetmo1, Suzette Poliquin1, David A Hanley2, Jerilynn C Prior3, Susan Barr4, Tassos Anastassiades5,Tanveer Towheed5, David Goltzman6, Nancy Kreiger7,8*, the CaMos Research GroupAbstractBackground: Previous research has shown that underlying dietary patterns are related to the risk of many differentadverse health outcomes, but the relationship of these underlying patterns to skeletal fragility is not wellunderstood. The objective of the study was to determine whether dietary patterns in men (ages 25-49, 50+) andwomen (pre-menopause, post-menopause) are related to femoral neck bone mineral density (BMD) independentlyof other lifestyle variables, and whether this relationship is mediated by body mass index.Methods: We performed an analysis of 1928 men and 4611 women participants in the Canadian MulticentreOsteoporosis Study, a randomly selected population-based longitudinal cohort. We determined dietary patternsbased on the self-administered food frequency questionnaires in year 2 of the study (1997-99). Our primaryoutcome was BMD as measured by dual x-ray absorptiometry in year 5 of the study (2000-02).Results: We identified two underlying dietary patterns using factor analysis and then derived factor scores. The firstfactor (nutrient dense) was most strongly associated with intake of fruits, vegetables, and whole grains. The secondfactor (energy dense) was most strongly associated with intake of soft drinks, potato chips and French fries, certainmeats (hamburger, hot dog, lunch meat, bacon, and sausage), and certain desserts (doughnuts, chocolate, icecream). The energy dense factor was associated with higher body mass index independent of other demographicand lifestyle factors, and body mass index was a strong independent predictor of BMD. Surprisingly, we did notfind a similar positive association between diet and BMD. In fact, when adjusted for body mass index, eachstandard deviation increase in the energy dense score was associated with a BMD decrease of 0.009 (95% CI: 0.002,0.016) g/cm2 for men 50+ years old and 0.004 (95% CI: 0.000, 0.008) g/cm2 for postmenopausal women. Incontrast, for men 25-49 years old, each standard deviation increase in the nutrient dense score, adjusted for bodymass index, was associated with a BMD increase of 0.012 (95% CI: 0.002, 0.022) g/cm2.Conclusions: In summary, we found no consistent relationship between diet and BMD despite finding apositive association between a diet high in energy dense foods and higher body mass index and a strongcorrelation between body mass index and BMD. Our data suggest that some factor related to the energy densedietary pattern may partially offset the advantages of higher body mass index with regard to bone health.* Correspondence: Nancy.Kreiger@cancercare.on.ca7Dalla Lana School of Public Health, University of Toronto, 155 College St,Toronto, ON, CanadaLangsetmo et al. BMC Musculoskeletal Disorders 2010, 11:20http://www.biomedcentral.com/1471-2474/11/20© 2010 Langsetmo et al; licensee BioMed Central Ltd. This is an Open Access article distributed under the terms of the CreativeCommons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, andreproduction in any medium, provided the original work is properly cited.BackgroundThe traditional approach to assessing the potential influ-ence of diet is to determine the relationship of a parti-cular nutrient to a given outcome after controlling forother nutrients. Another approach that gives comple-mentary information is to identify underlying dietarypatterns and determine the relationship of a particularpattern to a given outcome[1]. There is growing evi-dence that this second approach yields some strong andconsistent predictors of multiple health outcomes [2-7].Furthermore controlled trials have shown that it is pos-sible to modify underlying dietary patterns[8,9].Fractures related to osteoporosis contribute toincreased mortality[10], lower quality of life [11], as wellas substantial direct and indirect costs [12]. Secularchanges in nutrition and lifestyle may contribute to anincreased fracture burden if they are associated withbone fragility. Bone mineral density (BMD) is a strongpredictor of fracture[13]. Certain nutrients, most notablycalcium and vitamin D, are related to BMD[14]. Theassociation between dietary patterns and BMD is lessclear, although several studies have noted the potentialbenefit of fruits and vegetables [15-18]. A further com-plication is that while some dietary patterns mightdirectly affect BMD, other effects may be dependent onintermediary changes in fat and/or lean mass.The objective of the present analysis was to determinewhether dietary patterns in men (ages 25-49, 50+) andwomen (pre-menopause, post-menopause) are related toBMD independently of other lifestyle variables and howthe association is mediated by body mass index.MethodsSubjectsWe included participants in an on-going cohort study,the Canadian Multicentre Osteoporosis Study (CaMos)who completed the food frequency questionnaire with10 or fewer missing responses in the food and drinksection. A total of 9423 participants were enrolled.Among the 7315 participants who returned the food fre-quency questionnaire, 6539 had sufficiently completedata from the food frequency questionnaire to beincluded in this analysis.The methodological details of the study have beendescribed elsewhere [19]. Briefly, eligible participants wereat least 25 years old at the beginning of the study, livedwithin a 50-kilometre radius of one of nine Canadian cities(St John’s, Halifax, Quebec City, Toronto, Hamilton, King-ston, Saskatoon, Calgary, and Vancouver) and were able toconverse in English, French, or Chinese (Toronto or Van-couver). Households were randomly selected from a list ofresidential phone numbers and participants were ran-domly selected from eligible household members usingstandard protocol. Of those selected, 42% agreed to parti-cipate in the study resulting in a baseline cohort of 9423participants. Ethics approval was granted through McGillUniversity and the appropriate research ethics boards foreach participating centre. Signed informed consent wasobtained from all study participants.Data collectionAll participants were given a standardized interviewer-administered questionnaire (CaMos questionnaire©1995) at baseline. The questionnaire covered demo-graphics, health, nutrition, lifestyle, as well as a medicalhistory that included both a detailed history of fractureand major risk factors for fracture. The questionnairealso included the Medical Outcomes Study 36-itemShort Form (SF-36) with summary score for physicaland mental components[20]. Medication and supple-ment use was assessed by a complete inventory of pre-scriptions and bottles brought to the interview. Baselineclinical assessment included height, weight, and BMD.Height was measured without shoes, using a height rodmounted on beam balance scale, a wall-mounted stadi-ometer, or ruler on the wall. Weight was measured inlight clothing using a beam balance or electronic scale.A food frequency questionnaire was mailed to all parti-cipants in the second year of the study (1997-99). Fol-low-up visits were scheduled in the third year (1998-2000) for those between 40 and 60 years old and in thefifth year (2000-02) for all participants. These visits alsoincluded an interviewer-administered questionnaire,height and weight measurements, and BMD testing.Year 5 body mass index (kg/m2) and BMD were used asthe main outcomes, as the Year 5 time point was thefirst clinical measurement after the food frequency ques-tionnaire with no age exclusion criterion.Nutrition questionnaireThe self-administered nutrition questionnaire used inthis study was based on the food frequency question-naires developed and tested by Block (short form) andWillett [21,22], slightly modified for Canadian foods[23].We note that energy intake estimates derived from theshort Block questionnaire had a high correlation (r =0.9) with energy intake estimates from the full Blockquestionnaire[21]. A total of 106 questions assessed useof nutritional supplements (n = 11), beverages (n = 18),foods (n = 51), condiments/fat (n = 7), summary items(n = 5), and change in diet over 20 years (n = 15) withfood and beverage items selected from the Block FFQ.Each food and beverage item had a specified usual por-tion size as done in the Willett FFQ. Response optionswere one of nine mutually exclusive ordinal frequencycategories ranging from never/less than once a monthLangsetmo et al. BMC Musculoskeletal Disorders 2010, 11:20http://www.biomedcentral.com/1471-2474/11/20Page 2 of 11to 6 or more times per day. The FFQ has not been vali-dated in an independent study.Bone mineral densityBone mineral density was measured at the lumbar spine(L1-L4), femoral neck, trochanter, Ward’s triangle, andtotal hip. Seven centres used Hologic densitometers andtwo centres used Lunar densitometers. All Lunar mea-surements were converted to equivalent Hologic valuesusing standard reference formulas; this study requiredthe formulas for femoral neck[24]. We did not use thelumber spine data in this analysis as degenerativechanges and vertebral deformity both affect measure-ment at this site. Femoral neck was chosen as the refer-ence hip site as it is used more frequently in theliterature. All BMD values were adjusted to anthropo-morphic phantoms that were scanned in each centre inthe year of initial and all follow-up examinations.Statistical methodsThe nutritional questionnaire was first screened formissing responses and those questionnaires with morethan 10 missing responses in the food and beverage sec-tion were excluded. Missing responses for individualfood items in the remaining questionnaires werereplaced with the median response for the study sample[25]. Two variables were created to assess impact ofimputing missing responses in further analyses: totalnumber of missing responses and total differencebetween imputation using median intake and imputationusing lowest intake. Total energy intake was based onfrequency and specified portion size from the question-naire together with caloric information from the Cana-dian Nutrient File[26].After assessing the distributions and correlationsbetween 69 food and beverage items, the responses weregrouped into 34 categories (see Table 1 for list of sum-mary categories). The total consumption for each cate-gory was determined by summing the monthlyfrequency of each item as measured by servings permonth. The resulting variables were log transformed toadjust for skewness, and further rescaled to have meanof zero and standard deviation of 1. Factor analysis wasperformed using the 34 variables derived in the abovefashion. Factor analysis is used to assess underlying pat-terns of variation and the derived factor score is a mea-sure of common variation. Factor loading scoresindicate the strength and direction of the association.Thus a high factor score indicates greater consumptionof those foods with high factor loadings relative to withthose with lower factor loadings.Models were run with different numbers of factors,and were assessed by an eigenvalue criterion where fac-tors with eigenvalue less than 1 were dropped fromconsideration. Factors were rotated using varimax rota-tion to achieve uncorrelated factors with betterinterpretability.A two-factor model resulted in factors that were bothstatistically important (eigenvalue > 1) and clinicallyrelevant. Moreover, nearly identical factor loadings andfactor scores were obtained when factor analysis withtwo factors was done separately for a randomly chosensub-sample, for men, for women, and for those with/without imputed responses. The above stability of factorloadings across different subgroups was not present inmodels with 3 or more factors. The factor scores result-ing from the analysis on the whole sample with two fac-tors were used in all subsequent analyses. This choiceenables between group comparisons.It was hypothesized that behaviours would not be lim-ited to a single domain and that other lifestyle variableswould be associated with eating habits. Multiple linearregression was used to assess the relationship betweenfactor scores and baseline variables, with factor scoresbeing the dependent variable. The baseline variablesconsidered were age, education (< 12 years schooling,high school diploma, post-secondary education), smok-ing (non-smoker, former smoker, current smoker), alco-hol (non-drinker, moderate intake(< 1 drink/day inwomen, < 2 drink/day in men), high intake (1+ drink/day in women, 2+ drink/day in men))), activity (kcalorieper week spent on moderate activity, vigorous work, orstrenuous sports calculated from weekly inventory ofactivities in these three categories), sedentary time (timeper day spent sitting or sleeping), daily milk consump-tion, daily use of supplements (vitamin D, calcium).It was hypothesized that there would be a relationshipbetween diet and BMD and that the relationship of dietto BMD would be partially mediated by body massindex. We assessed this hypothesis using a series ofregression models testing both direct and indirect asso-ciations. The first set of models used body mass indexas the dependent variable; the second set of modelsused femoral neck BMD as the dependent variable,adjusting for height but not for body mass index; andthe final set of models used femoral neck BMD as thedependent variable, adjusting for both height and bodymass index. We also looked at the direct associationbetween body mass index and BMD using univariateregression. All regression analyses were a priori strati-fied by sex, and by age category in men and menopausalstatus in women based on known differences in bothdiet and bone mineral metabolism. All models includeda priori specified potential confounders including allbaseline variables listed above, together with study cen-tre, medication (antiresorptives, corticosteroids),oophorectomy, and recent menopause (last 5 years); asrelevant. No adjustments were made for multipleLangsetmo et al. BMC Musculoskeletal Disorders 2010, 11:20http://www.biomedcentral.com/1471-2474/11/20Page 3 of 11comparisons. In addition to regression diagnostics, weused robust regression to determine whether regressionwas sensitive to extreme values and found only minordifferences not impacting overall interpretation.ResultsThe baseline characteristics of the 1928 men and 4611women in the study sample are shown in Table 2. Ouranalyses excluded 956 men and 1928 women becausethe Year 2 food frequency questionnaire was missing orincomplete. Men who were excluded were on average3.1 (95% CI: 1.9, 4.3) years older, had lower SF-36 physi-cal health and mental health scores (mean difference 3.1(95% CI: 2.3, 3.9) and 1.2 (95% CI: 0.6, 1.9) respectively),but had similar body mass index and total hip BMDcompared with those in the study. Women who wereexcluded were on average 6.2 (95% CI: 5.6, 6.9) yearsolder, had lower SF-36 physical health and mentalhealth scores (mean difference 3.9 (95% CI: 3.4, 3.6) and0.9 (95% CI: 0.4, 1.4) respectively), had 0.032 (95% CI:Table 1 Food Categories and Factor Loadings Based on CaMos Year 2 Food Frequency QuestionnaireCategory Items Factor LoadingsNutrient dense Energy denseCoffeeTeaWater Bottled water, tap water 0.29Juice Fresh fruit juice, frozen concentrated fruit juice 0.28Low fat milk Skim and 1% milk 0.22High fat milk 2% and whole milk 0.21Beer 0.26Alcohol Wine, liquorSoft drinks Soft drinks, powdered drink mix 0.42Fruit 1 Apples, oranges, bananas 0.48Fruit 2 Cantaloupe, other fruit 0.54Tomatoes Tomatoes, tomato juice 0.49Green vegetables Broccoli, spinach, other leafy greens 0.61Yellow vegetables Cabbage, cauliflower, sweet potato, squash, brussel sprouts 0.56Other vegetables Carrots, corn, peas, green beans, soup, other vegetables 0.59Whole grains Dark whole grain bread, bran/granola, shredded wheat 0.46White bread White bread/rolls -0.24 0.37Cereal Cold cereal, cooked cereal 0.25Rice 0.26Pasta Macaroni, spaghetti, noodles 0.22 0.26Potatoes 0.25 0.24Meat 1 Beef, pork, lamb, poultry (main dish or mixed) 0.28 0.28Meat 2 Hamburger 0.50Meat 3 Hotdog, lunch meat, smoked meat 0.55Meat 4 Bacon, sausage 0.56Fish Fish (fresh, frozen, smoked, dried) 0.40Eggs 0.32Cheese 0.25 0.30Nuts Nuts, peanut butter 0.21 0.20Legumes Beans, lentils, tofu, soybeans 0.37Sweets 1 Cake, pie, cookies 0.28Sweets 2 Ice cream, chocolate, doughnuts 0.47Added fats Margarine, butter, mayonnaise 0.35High fat potatoes Fries, Potato chips 0.58Percentage of overall variance explained by factor 9.3% 7.8%Factor loadings with absolute value less than 0.2 indicated by blanks and those above 0.4 indicated in bold to visually emphasize strength of association. Sign offactor score indicates direction of association.Langsetmo et al. BMC Musculoskeletal Disorders 2010, 11:20http://www.biomedcentral.com/1471-2474/11/20Page 4 of 110.024, 0.039) g/cm2 lower femoral neck BMD, but hadsimilar body mass index compared with those in thestudy.The two factors and the corresponding factor loadingsare shown in Table 1. The first factor (nutrient dense)was most strongly associated with intake of fruits, vege-tables, and whole grains. The second factor (energydense) was most strongly associated with intake of softdrinks, potato chips and French fries, certain meats(hamburger, hot dog, lunch meat, smoked meat, bacon,and sausage), and certain desserts (doughnuts, chocolate,ice cream). The items listed above all had positive load-ing with the dietary patterns, indicating that above aver-age intake contributed to a positive score on theassociated factor and below average intake contributedto a negative score. The distributions of both factorscores were approximately normal with mean = 0 andstandard deviation = 0.9.Both factor scores were related to sex, but in oppositedirections. The mean score for the nutrient dense factorwas 0.43 (95% CI: 0.37-0.48) standard deviations higherin women than men. The mean score for the energydense factor was 0.52 (95% CI: 0.46-0.57) standarddeviations lower in women than men. The results ofmultivariate models predicting nutrient dense andenergy dense factor scores in men and women by demo-graphic and lifestyle variables are shown in Table 3. Agewas a strong linear predictor, with younger participantshaving on average lower nutrient dense scores andhigher energy dense scores than older participantsamong both men and women.Higher educational attainment, vitamin D supplementuse, and non-smoking were independently associatedwith higher nutrient dense scores and lower energydense scores. Higher alcohol intake and greater physicalactivity were both independently associated with highernutrient dense scores and higher energy dense scores.Finally, higher milk consumption and greater use of cal-cium supplements was associated with a higher nutrientdense score and longer sedentary time was associatedwith a higher energy dense score.We assessed the relationship of total energy intake tothe factor scores. Both factor scores were positively cor-related with energy intake (data not shown). Since thefactor scores were derived using log-transformed vari-ables these relationships were not linear. The Pearsoncorrelation of the log-transformed energy intake and thefactor scores was r = 0.54 for the nutrient dense factorscore and r = 0.49 for the energy dense factor score.Furthermore, among those with a given energy intakethere was an inverse relationship between the nutrientdense factor score and the energy dense factor score. Inview of these strong correlations, we performed a sec-ondary analysis including the difference between thetwo factor scores and the log transformed energy intakeas adjustment factor. The difference was calculated asthe energy dense factor score minus the nutrient densefactor score and can be interpreted as the direct com-parison between a diet high in energy dense foods ver-sus a diet high in nutrient dense foods.The estimated parameters for factor scores, differencebetween scores, and energy intake as predictors of bodymass index in each of the four groups (men 25-49 years,men 50+ years, women pre-menopause, women post-menopause) are shown in Figure 1. Higher nutrientdense factor scores were associated with a similar orlower body mass index. Higher energy dense factorscores were associated with a higher body mass index.The difference between scores (energy dense score-nutrient dense score) was positively associated withbody mass index, but energy intake was not associatedwith body mass index. There was no evidence ofbetween group heterogeneity of the regression coeffi-cients as assessed by analysis of variance (p-valuesbetween 0.24 to 0.75). The above associations wereadjusted for age, height, study centre, education, smok-ing, alcohol, activity, sedentary time, milk consumption,supplements (vitamin D and calcium), antiresorptiveTable 2 Baseline Characteristics of the Study SampleMen(N = 1928)Women(N = 4611)Mean SD Mean SDAge (years) 58.8 13.5 61.2 12.2Height (cm) 174.1 7.1 160.1 6.4Weight (kg) 82.2 13.7 69.1 13.8Body mass index (kg/m2) 27.1 4.0 26.9 5.1Femoral neck BMD (g/cm2) 0.81 0.13 0.72 0.12SF-36 physicala 49.9 8.6 47.6 10.0SF-36 mentala 54.5 7.7 53.3 8.8Sedentary timeb (hours/day) 14.6 3.1 13.9 3.0Activityc(1000 kcal/week) 5.8 5.1 4.4 3.2N % N %Current smoker 300 15.6 630 13.7Current alcohol use 1479 76.7 2754 59.7Antiresorptives 5 0.3 1317 28.6Corticosteroids 178 9.2 623 13.5Vitamin D from supplements 338 17.5 1551 33.6Calcium from supplements 659 34.2 2574 55.8Caucasian (white) 1822 94.5 4438 96.2Chinese 54 2.8 80 1.7North American Indian 24 1.2 56 1.2South Asian 15 0.8 18 0.4a Scores range from 0 to 100 with 0 indicating worst possible health and 100indicating best possible healthb Total time spent sitting or sleepingc Moderate, strenuous or vigorous activityLangsetmo et al. BMC Musculoskeletal Disorders 2010, 11:20http://www.biomedcentral.com/1471-2474/11/20Page 5 of 11therapy, corticosteroids, oophorectomy, and recentmenopause (final menstrual flow within the last 5 years).The estimated parameters for factor scores, differencebetween scores, and energy intake as predictors of BMDadjusted for the same confounders noted above areshown in Figure 2 (no adjustment for BMI) and Figure 3(with adjustment for BMI). The associations between thefactor scores and BMD without adjusting for body massindex were not consistent across subgroups and did notfollow the previously noted relationships between dietaryfactors and body mass index. There was some unex-plained between group heterogeneity of the regressioncoefficients (p-values between 0.03 and 0.46). There wasa positive association between energy intake and femoralneck BMD among young men, but very weak and notstatistically significant associations in all other subgroups.There was a strong positive correlation between bodymass index and femoral neck BMD in all subgroups,both in univariate and multivariate analyses. As a resultof the serial correlations, the association between thenutrient dense factor score and BMD was more positiveand the association between the energy dense factorscore and BMD was more negative in each subgroup inthe analysis adjusted for body mass index comparedwith the unadjusted analysis. The between group hetero-geneity in the body mass index adjusted analysis wasslightly less than that of the unadjusted analysis (p-values between 0.06 and 0.31).DiscussionWe identified two dietary patterns (nutrient dense andenergy dense) in Canadian men and women analogousTable 3 Regression Coefficients for Baseline Variables as Predictors of Dietary Factor Scores in Men and Women fromMultivariate ModelIndependent variables (Baseline) Outcome variables (Year 2)Nutrient densescore aEnergy densescoreaMen Women Men WomenAge (10 years) 0.200.16, 0.230.150.13, 0.17-0.15-0.19, -0.12-0.13-0.15, -0.11Education b < 12 years -0.24-0.36, -0.13-0.20-0.27, -0.14-0.01-0.13, 0.120.05-0.01, 0.11Post-secondary 0.300.19, 0.400.260.19, 0.33-0.13-0.24, -0.02-0.09-0.15, -0.02Smoking c Current -0.27-0.40, -0.13-0.36-0.44, -0.280.390.24, 0.530.250.17, 0.33Former -0.02-0.12, 0.080.02-0.04, 0.080.130.02, 0.230.00-0.06, 0.06Alcohol d Moderate 0.260.15, 0.360.210.15, 0.260.330.21, 0.440.120.06, 0.18High 0.14-0.05, 0.320.140.02, 0.260.500.30, 0.690.270.15, 0.39Activity e(2500 kcal/week)0.050.03, 0.070.040.02, 0.060.030.01, 0.050.040.02, 0.06Sedentary time f(hours/day)0.00-0.01, 0.02-0.01-0.02, 0.000.020.00, 0.030.020.01, 0.03Milk consumption(250 mL/day)0.030.00, 0.060.070.05, 0.090.030.00, 0.060.00-0.02, 0.03Calcium from supplements (500 mg/day) 0.06-0.02, 0.140.030.00, 0.070.02-0.06, 0.11-0.03-0.06, 0.00Vitamin D from supplements (200 IU/day) 0.080.03, 0.130.030.00, 0.05-0.04-0.10, 0.01-0.04-0.06, -0.01a Outcome variables are unitless but standardized to have mean = 0 and SD = 1. Each individual’s diet was characterized by both a nutrient dense score and anenergy dense score. A positive regression coefficient for the nutrient dense score indicates category or increase in the independent variable is associated with agreater consumption of fruits, vegetables and whole grains relative to other foods. A positive regression coefficient for the energy dense scores indicatescategory or increase in the independent variable is associated with a greater consumption of chips/fries, processed meat, soft drinks, and certain desserts relativeto other foods.Reference or comparison category:b Reference category = high school diploma.c Reference category = never-smokerd Reference category = non-drinker,e Moderate, strenuous or vigorous activityf Total time spent sitting or sleepingConfidence intervals not crossing 0 prior to rounding are indicated in bold type (equivalent to p < 0.05)Langsetmo et al. BMC Musculoskeletal Disorders 2010, 11:20http://www.biomedcentral.com/1471-2474/11/20Page 6 of 11Figure 1 Regression coefficients for dietary patterns and energy intake as predictors of body mass index. The parameter estimates arefor each 1 SD increase of the nutrient dense factor score, the energy dense factor score, the difference between energy dense and nutrientdense factor score, and the log-tranformed energy intake (1 SD is roughly 36% change in energy intake). P-values for null hypothesis (from topto bottom) Younger Men: 0.080, 0.001, 0.001, 0.884; Older Men: 0.294, < 0.001, 0.002, 0.961; Premenopausal Women: 0.077, 0.126, 0.019, 0.683;Postmenopausal Women: 0.842, < 0.001, < 0.001, 0.096. Analyses were run for the two factor scores and for the difference between factor scoresand energy intake separately due to multicollinearity between intake and factor scores. All models are adjusted for age, height, center,education, smoking, alcohol consumption, activity, sedentary time, milk consumption, supplements (vitamin D, calcium); and antiresorptives,corticosteroids, recent (< 5 years) menopause, oophorectomy, as relevant. A high nutrient dense score indicates a greater consumption of fruits,vegetables and whole grains relative to other foods, a high energy dense scores indicates a greater consumption of chips/fries, processed meat,soft drinks, and certain desserts relative to other foods. A high difference indicates more energy dense food relative to nutrient dense foods.Figure 2 Regression coefficients for dietary patterns and energy intake as predictors of femoral neck BMD without adjustment forbody mass index. The parameter estimates are for each 1 SD increase of the nutrient dense factor score, the energy dense factor score, thedifference between energy dense and nutrient dense factor score, and the log-tranformed energy intake (1 SD is roughly 36% change in energyintake). P-values for null hypothesis (from top to bottom) Younger Men: 0.057, 0.120, 0.770, 0.008; Older Men: 0.381, 0.284. 0.202, 0.357;Premenopausal Women: 0.907, 0.232, 0.449, 0.874; Postmenopausal Women: 0.607. 0.451, 0.905, 0.324.Langsetmo et al. BMC Musculoskeletal Disorders 2010, 11:20http://www.biomedcentral.com/1471-2474/11/20Page 7 of 11to dietary patterns ("Prudent” and “Western”) noted inother studies [2-5]. Surprisingly, we did not find anyconsistent relationship between diet and BMD withoutadjustment for body mass index, and among postmeno-pausal women, this association if present was very small.Genetics and early environment play a strong role in thedevelopment of peak bone mass and genetics may alsoimpinge on the rate of bone loss[27]. Later environmen-tal determinants may have an effect on the rate of boneloss, but these effects may be small in comparison rela-tive to other determinants of BMD. The early determi-nation of overall bone mass may explain the overall veryweak associations between diet and BMD despite thenoted relationship between diet and body mass index.We posit there may be an association between theincreased consumption of nutrient dense foods relativeto energy dense foods and BMD after further adjust-ment for body mass index. After adjustment for bodymass index, higher intake of nutrient dense food wasassociated with a higher BMD among men ages 25-49.Weaker but still positive relationships were foundamong older men and women, but none of these resultswere statistically significant. A higher energy dense fac-tor score, adjusted for body mass index, was associatedwith lower BMD among men ages 50 and over, andpost-menopausal women. Albeit non-significant, thereverse association was found among younger men andwomen in our study, but it is not clear whether this wasa reflection of overall uncertainty or true heterogeneity.Viewing these results together suggests a comparativeadvantage of nutrient dense foods over energy densefoods, except among premenopausal women. In contrastto the null results for pre-menopausal women, Okubo etal., using a factor analysis approach similar to the analy-sis we have used in this paper, found that a dietary pat-tern including fish, fruits, and vegetables and low inmeat and processed meat was associated with higherBMD in pre-menopausal Japanese farm women [28].Components of the nutrient dense dietary pattern, nota-bly fruits and/or vegetables, have also been shown to beassociated with BMD [15-18].The consistency of dietary patterns between our studyand other studies enable better between-study compari-sons, and suggests that our results may be applicable todifferent populations. We found that the factor scoreswere related to several demographic factors, includingsex, age, and education. The lower nutrient dense scoresand higher energy dense scores among younger agegroups is alarming from the view of population health,given the relationships found between “Prudent” and“Western” dietary patterns and adverse health outcomesincluding coronary heart disease and stroke [2-5].Comparison of our results with studies based onassessment of nutrients is more difficult. The energydense dietary pattern included items high in carbohy-drates, high in fat, and high in both. It has beenFigure 3 Regression coefficients for dietary patterns and energy intake as predictors of femoral neck BMD with adjustment for bodymass index. The parameter estimates are for each 1 SD increase of the nutrient dense factor score, the energy dense factor score, thedifference between energy dense and nutrient dense factor score, and the log-tranformed energy intake (1 SD is roughly 36% change in energyintake). P-values for null hypothesis (from top to bottom) Younger Men: 0.028, 0.552, 0.258, 0.030; Older Men: 0.118, 0.007, 0.005, 0.399;Premenopausal Women: 0.300, 0.374, 0.904, 0.425; Postmenopausal Women: 0.353, 0.032, 0.028, 0.110.Langsetmo et al. BMC Musculoskeletal Disorders 2010, 11:20http://www.biomedcentral.com/1471-2474/11/20Page 8 of 11previously noted that dietary patterns high in sweets areassociated with lower BMD[18], however the impact ofdietary fat is unclear. Thus, one study demonstrated anegative association between intake of unsaturated fatand BMD, and a positive association between intake ofsaturated fat and BMD[29]. In contrast, another studyfound a negative association between intake of saturatedfat and BMD[30]. The observed associations maydepend on complex interactions between fat and othernutrients and an analysis of these interactions woulddepend both on identification of the nutrients involvedand proper model specification.We confirmed the hypothesis that diet was a predic-tor of body mass index. Notably, we found that ahigher intake of energy dense foods relative to nutrientdense foods, as seen directly by use of the differencescore, was associated with increased body mass indexacross all subgroups. In contrast, there was no associa-tion between overall energy intake and body massindex. Energy intake is related to energy density offoods consumed, and increasing the energy density ofa diet has been shown to increase total energy intake[31-33]. Some of the foods associated with the nutrientdense factor have low energy density, notably fruitsand vegetables[6,8]. Therefore, the observed associationbetween dietary patterns and body mass index may becausally linked by increased energy intake without aconcomitant increase in energy expenditure. Assess-ment of this causal path is problematic in observa-tional studies since it is difficult to measure energybalance with sufficient accuracy[34]. Metabolic effi-ciency is usually unknown and typical measures ofintake and activity are susceptible to both bias anderror. The most probable source of bias in this casewas the use of a standard portion size for all fooditems. Underestimation of portion size would lead togreater underestimation in the calculation of totalenergy intake for those who consumed more energydense foods. This mechanism could introduce suffi-cient bias to mask any association between energyintake and body mass index.We also confirmed the hypothesis that body massindex was strongly associated with BMD. This is par-tially attributable to the fact that those with larger bonesize have both greater body mass index and higherBMD. Body mass index is also associated with both leanand fat body mass, which are also predictors of bonemineral content[35].The strengths of our study include the fact that wewere able to assess lifestyle and demographics, includingdietary patterns, together with measured assessment ofbody mass index and BMD values in a large randomlyselected population, including both men and womenover a wide age range. This allowed us to study therelationship between diet and BMD after adjusting formany of the variables that are related to dietary patterns.Our use of dietary patterns takes into account interac-tions between nutrient and foods not possible using thesingle nutrient approach.This study has limitations. The factor analysis used isexploratory in nature and involves decisions that aresubjective. Not all members of the cohort completedthe food frequency questionnaire nor had BMDassessed at year 5. Those with a poor diet might havehad a missing FFQ or might have died before secondBMD measurement, with bias most likely toward thenull. The limited scope and specified portion size ofthe FFQ may yield biased estimates of absolute energyintake. Some ethnic groups included in the study (e.g.Chinese) may have had dietary habits not adequatelycaptured by the food frequency questionnaire. Further-more, under-representation of ethnic minorities in thestudy may limit generalizability. Finally, we cannot ruleout the possibility of residual confounding since diet-ary patterns may be related to other unmeasuredhealth behaviours.ConclusionsIn summary, we found no consistent relationshipbetween diet and BMD despite finding a positive asso-ciation between a diet high in energy dense foods andlow in nutrient dense foods and higher body massindex. Because body mass index is strongly associatedwith BMD it was expected that similar associationswould be true for diet and BMD. There may be associa-tions between dietary patterns and BMD after adjustingfor body mass index, which partially offset the expectedpositive effect of body mass index on BMD.AcknowledgementsWe thank all those participants in CaMos whose careful responses andattendance made this analysis possible. The Canadian MulticentreOsteoporosis Study was funded by the Canadian Institutes of HealthResearch (CIHR), Merck Frosst Canada Ltd., Eli Lilly Canada Inc., NovartisPharmaceuticals Inc., The Alliance for Better Bone Health: sanofi-aventis andProcter & Gamble Pharmaceuticals Canada Inc., The Dairy Farmers of Canada,and The Arthritis Society. These funding sources had no role in theconception of this analysis, statistical methods, or interpretation of the data.CaMos Research GroupDavid Goltzman (co-principal investigator, McGill University), Nancy Kreiger(co-principal investigator, Toronto), Alan Tenenhouse (principal investigatoremeritus, Toronto).CaMos Coordinating Centre, McGill University, Montreal, Quebec: SuzettePoliquin (national coordinator), Suzanne Godmaire (research assistant),Claudie Berger (study statistician), Lawrence Joseph (consultant statistician).Memorial University, St. John’s Newfoundland: Carol Joyce (director),Christopher Kovacs (co-director), Emma Sheppard (coordinator).Dalhousie University, Halifax, Nova Scotia: Susan Kirkland, Stephanie Kaiser(co-directors), Barbara Stanfield (coordinator).Laval University, Quebec City, Quebec: Jacques P. Brown (director), LouisBessette (co-director), Marc Gendreau (coordinator).Queen’s University, Kingston, Ontario: Tassos Anastassiades (director), TanveerTowheed (co-director), Barbara Matthews (coordinator).Langsetmo et al. BMC Musculoskeletal Disorders 2010, 11:20http://www.biomedcentral.com/1471-2474/11/20Page 9 of 11University of Toronto, Toronto, Ontario: Bob Josse (director), Sophie Jamal(co-director), Tim Murray (past director), Barbara Gardner-Bray (coordinator)McMaster University, Hamilton, Ontario: Jonathan D. Adachi (director),Alexandra Papaioannou (co-director), Laura Pickard (coordinator).University of Saskatchewan, Saskatoon, Saskatchewan: Wojciech P. Olszynski(director), K. Shawn Davison(co-director), Jola Thingvold (coordinator).University of Calgary, Calgary, Alberta: David A. Hanley (director), Jane Allan(coordinator).University British Columbia, Vancouver, British Columbia: Jerilynn C. Prior(director), Yvette Vigna (coordinator); Brian C. Lentle (radiologist).Author details1CaMos National Coordinating Centre, McGill University, 687 Pine Ave W,Montreal, QC, H3A 1A1, Canada. 2Department of Medicine, University ofCalgary, 3330 Hospital Dr NW, Calgary, AB, T2N 4N1, Canada. 3Department ofMedicine, University of British Columbia, 2775 Laurel St, Vancouver, BC, V5Z1M9, Canada. 4Department of Human Nutrition, University of BritishColumbia, 2205 East Mall, Vancouver, BC, V5Z 1M9, Canada. 5Department ofMedicine, Queen’s University, Etherington Hall, Kingston, ON, K7L 3N6,Canada. 6Department of Medicine, McGill University, 687 Pine Ave W,Montreal, QC, Canada. 7Dalla Lana School of Public Health, University ofToronto, 155 College St, Toronto, ON, Canada. 8Cancer Care Ontario, 620University Avenue, Toronto, ON M5G 2L7, Canada.Authors’ contributionsLL, SP, DAH, JCP, TA, TT, and NK worked on the study design. LL, SP, NKperformed the data analysis. LL, SP, JCP, SB, DG, NK worked on theinterpretation of results. All authors were involved in drafting and revisingthe manuscript and have read and approved the final manuscript. Othermembers of the CaMos Research Group were involved in the initial studydesign, recruitment of participants, data collection, quality control, review ofprojects, retention of the cohort, and other projects.Competing interestsDavid Hanley, MD; advisory board, honoraria, grants: Abbott Laboratories,Amgen, Eli Lilly, Merck, Novartis, Proctor & Gamble, sanofi-aventis, Servier,Wyeth-Ayerst, Nycomed, Paladin. Susan Barr, PhD; consulting: InternationalDairy Foods Association Tassos Anastassiades, MD; honoraria: Merck, Proctor& Gamble, Schering Plough, Servier Tanveer Towheed, MD; honoraria, grants:Abbott Laboratories, Bristol-Myers Squib, Novartis, sanofi-aventis DavidGoltzman, MD; consulting: Eli Lilly, Novartis, Merck-Frosst, Proctor & Gamble,sanofi-aventis, Servier Lisa Langsetmo, Suzette Poliquin, Jerilynn Prior andNancy Kreiger have no competing interests.Received: 9 October 2009Accepted: 28 January 2010 Published: 28 January 2010References1. Jacobs DR Jr, Tapsell LC: Food, not nutrients, is the fundamental unit innutrition. Nutr Rev 2007, 65:439-450.2. Fung TT, Willett WC, Stampfer MJ, Manson JE, Hu FB: Dietary patterns andthe risk of coronary heart disease in women. Arch Intern Med 2001,161:1857-1862.3. Fung TT, Stampfer MJ, Manson JE, Rexrode KM, Willett WC, Hu FB:Prospective study of major dietary patterns and stroke risk in women.Stroke 2004, 35:2014-2019.4. Hu FB, Rimm EB, Stampfer MJ, Ascherio A, Spiegelman D, Willett WC:Prospective study of major dietary patterns and risk of coronary heartdisease in men. Am J Clin Nutr 2000, 72:912-921.5. Lutsey PL, Steffen LM, Stevens J: Dietary intake and the development ofthe metabolic syndrome: the Atherosclerosis Risk in Communities study.Circulation 2008, 117:754-761.6. Newby PK, Muller D, Hallfrisch J, Andres R, Tucker KL: Food patternsmeasured by factor analysis and anthropometric changes in adults. Am JClin Nutr 2004, 80:504-513.7. Handa K, Kreiger N: Diet patterns and the risk of renal cell carcinoma.Public Health Nutr 2002, 5:757-767.8. de Oliveira MC, Sichieri R, Venturim MR: A low-energy-dense diet addingfruit reduces weight and energy intake in women. Appetite 2008,51:291-295.9. Ello-Martin JA, Roe LS, Ledikwe JH, Beach AM, Rolls BJ: Dietary energydensity in the treatment of obesity: a year-long trial comparing 2weight-loss diets. Am J Clin Nutr 2007, 85:1465-1477.10. Katelaris AG, Cumming RG: Health status before and mortality after hipfracture. Am J Public Health 1996, 86:557-560.11. Adachi JD, Ioannidis G, Berger C, Joseph L, Papaioannou A, Pickard L, et al:The influence of osteoporotic fractures on health-related quality of lifein community-dwelling men and women across Canada. Osteoporos Int2001, 12:903-908.12. Burge R, Dawson-Hughes B, Solomon DH, Wong JB, King A, Tosteson A:Incidence and economic burden of osteoporosis-related fractures in theUnited States, 2005-2025. J Bone Miner Res 2007, 22:465-475.13. Johnell O, Kanis JA, Oden A, Johansson H, De Laet C, Delmas P, et al:Predictive value of BMD for hip and other fractures. J Bone Miner Res2005, 20:1185-1194.14. Tang BM, Eslick GD, Nowson C, Smith C, Bensoussan A: Use of calcium orcalcium in combination with vitamin D supplementation to preventfractures and bone loss in people aged 50 years and older: a meta-analysis. Lancet 2007, 370:657-666.15. Macdonald HM, New SA, Golden MH, Campbell MK, Reid DM: Nutritionalassociations with bone loss during the menopausal transition: evidenceof a beneficial effect of calcium, alcohol, and fruit and vegetablenutrients and of a detrimental effect of fatty acids. Am J Clin Nutr 2004,79:155-165.16. Prynne CJ, Mishra GD, O’Connell MA, Muniz G, Laskey MA, Yan L, et al: Fruitand vegetable intakes and bone mineral status: a cross sectional studyin 5 age and sex cohorts. Am J Clin Nutr 2006, 83:1420-1428.17. Zalloua PA, Hsu YH, Terwedow H, Zang T, Wu D, Tang G, et al: Impact ofseafood and fruit consumption on bone mineral density. Maturitas 2007,56:1-11.18. Tucker KL, Chen H, Hannan MT, Cupples LA, Wilson PW, Felson D, et al:Bone mineral density and dietary patterns in older adults: theFramingham Osteoporosis Study. Am J Clin Nutr 2002, 76:245-252.19. Kreiger N, Tenenhouse A, Joseph L, MacKenzie T, Poliquin S, Brown JP, et al:Research notes: the Canadian Multicentre Osteoporosis Study (CaMos) -background, rationale, methods. Can J Aging 1999, 18:376-387.20. Ware JE Jr, Kosinski M, Bayliss MS, McHorney CA, Rogers WH, Raczek A:Comparison of methods for the scoring and statistical analysis of SF-36health profile and summary measures: summary of results from theMedical Outcomes Study. Med Care 1995, 33:AS264-AS279.21. Block G, Hartman AM, Naughton D: A reduced dietary questionnaire:development and validation. Epidemiology 1990, 1:58-64.22. Willett WC, Sampson L, Stampfer MJ, Rosner B, Bain C, Witschi J, et al:Reproducibility and validity of a semiquantitative food frequencyquestionnaire. Am J Epidemiol 1985, 122:51-65.23. Villeneuve PJ, Johnson KC, Kreiger N, Mao Y: Risk factors for prostatecancer: results from the Canadian National Enhanced CancerSurveillance System. The Canadian Cancer Registries EpidemiologyResearch Group. Cancer Causes Control 1999, 10:355-367.24. Genant HK: Universal standardization for dual X-ray absorptiometry:patient and phantom cross-calibration results. J Bone Miner Res 1995,10:997-998.25. Parr CL, Hjartaker A, Scheel I, Lund E, Laake P, Veierod MB: Comparingmethods for handling missing values in food-frequency questionnairesand proposing k nearest neighbours imputation: effects on dietaryintake in the Norwegian Women and Cancer study (NOWAC). PublicHealth Nutr 2008, 11:361-370.26. Health Canada: Canadian Nutrient File.http://205.193.93.51/cnfonline/, 4-18-2008. 7-18-2008. Ref Type: Electronic Citation.27. Brown LB, Streeten EA, Shapiro JR, McBride D, Shuldiner AR, Peyser PA,et al: Genetic and environmental influences on bone mineral density inpre- and post-menopausal women. Osteoporos Int 2005, 16:1849-1856.28. Okubo H, Sasaki S, Horiguchi H, Oguma E, Miyamoto K, Hosoi Y, et al:Dietary patterns associated with bone mineral density in premenopausalJapanese farmwomen. Am J Clin Nutr 2006, 83:1185-1192.29. Macdonald HM, New SA, Golden MH, Campbell MK, Reid DM: Nutritionalassociations with bone loss during the menopausal transition: evidenceof a beneficial effect of calcium, alcohol, and fruit and vegetablenutrients and of a detrimental effect of fatty acids. Am J Clin Nutr 2004,79:155-165.Langsetmo et al. BMC Musculoskeletal Disorders 2010, 11:20http://www.biomedcentral.com/1471-2474/11/20Page 10 of 1130. Corwin RL, Hartman TJ, Maczuga SA, Graubard BI: Dietary saturated fatintake is inversely associated with bone density in humans: analysis ofNHANES III. J Nutr 2006, 136:159-165.31. Rolls BJ, Bell EA, Castellanos VH, Chow M, Pelkman CL, Thorwart ML: Energydensity but not fat content of foods affected energy intake in lean andobese women. Am J Clin Nutr 1999, 69:863-871.32. Cuco G, Arija V, Marti-Henneberg C, Fernandez-Ballart J: Food andnutritional profile of high energy density consumers in an adultMediterranean population. Eur J Clin Nutr 2001, 55:192-199.33. de Castro JM: Dietary energy density is associated with increased intakein free-living humans. J Nutr 2004, 134:335-341.34. Willett W, Stampfer MJ: Total energy intake: implications forepidemiologic analyses. Am J Epidemiol 1986, 124:17-27.35. Khosla S, Atkinson EJ, Riggs BL, Melton LJ III: Relationship between bodycomposition and bone mass in women. J Bone Miner Res 1996,11:857-863.Pre-publication historyThe pre-publication history for this paper can be accessed here:http://www.biomedcentral.com/1471-2474/11/20/prepubdoi:10.1186/1471-2474-11-20Cite this article as: Langsetmo et al.: Dietary patterns in Canadian menand women ages 25 and older: relationship to demographics, bodymass index, and bone mineral density. BMC Musculoskeletal Disorders2010 11:20.Submit your next manuscript to BioMed Centraland take full advantage of: • Convenient online submission• Thorough peer review• No space constraints or color figure charges• Immediate publication on acceptance• Inclusion in PubMed, CAS, Scopus and Google Scholar• Research which is freely available for redistributionSubmit your manuscript at www.biomedcentral.com/submitLangsetmo et al. BMC Musculoskeletal Disorders 2010, 11:20http://www.biomedcentral.com/1471-2474/11/20Page 11 of 11

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

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>
                        
                    
IIIF logo 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.52383.1-0223762/manifest

Comment

Related Items