In the USA and Canada, dietary supplements encompass vitamins and minerals, as well as non-vitamin non-mineral (NVNM) products including herbal preparations, probiotics, amino acids, fatty acids and miscellaneous combinations of these and other ingredients(1, 2).
It has been estimated that approximately half of the adult population in the USA are users of dietary supplements(Reference Radimer, Bindewald, Hughes, Ervin, Swanson and Picciano3). Although no comparable, large population-based studies have reported on supplement intake in Canada, existing prevalence estimates are broadly similar to those observed in the USA(Reference Troppmann, Johns and Gray-Donald4).
Despite the apparently widespread consumption of dietary supplements in North America, the long-term health effects in the general population are far from clear(Reference Neuhouser5). For many NVNM supplements in common use, the evidence for health benefits or harms is inconclusive or lacking(Reference Ernst6, Reference Tesch7). Similarly, for vitamin and/or mineral supplements, the conflicting and insufficient evidence for health benefits of supplementation in the general population(8–Reference Huang, Caballero and Chang11) has led several recent reviews to conclude that public health policies should continue to recommend consumption of a variety of foods rather than supplements(Reference Bender12–14).
Failure to account fully for the effects of potential confounders when attempting to determine health effects of supplementation may partially explain why hypotheses generated in observational studies and tested in randomized controlled trials have been inconsistent(Reference Reedy, Haines and Campbell15–Reference Lawlor, Davey Smith, Bruckdorfer, Kundu and Ebrahim17). For example, relative to non-users, supplement users tend to report more physical activity(Reference Radimer, Bindewald, Hughes, Ervin, Swanson and Picciano3, Reference Fennell18–Reference Ishihara, Sobue, Yamamoto, Sasaki and Tsugane26), greater intakes of fruits and/or vegetables(Reference Lyle, Mares-Perlman, Klein, Klein and Greger19, Reference Foote, Murphy, Wilkens, Hankin, Henderson and Kolonel20, Reference Patterson, Neuhouser, White, Hunt and Kristal24, Reference Frank, Bendich and Denniston27) and lower intake of dietary fat(Reference Frank, Bendich and Denniston27). Higher income and education have also been associated positively with supplement use(Reference Radimer, Bindewald, Hughes, Ervin, Swanson and Picciano3, Reference Fennell18–Reference Foote, Murphy, Wilkens, Hankin, Henderson and Kolonel20, Reference Marques-Vidal, Arveiler and Evans22, Reference Patterson, Neuhouser, White, Hunt and Kristal24, Reference Shikany, Patterson, Agurs-Collins and Anderson25, Reference Block, Cox, Madans, Schreiber, Licitra and Melia28).
However it has been observed recently that the dichotomous approach, used by the majority of studies that have examined the characteristics of supplement users v. non-users, is likely to mask additional health-related and sociodemographic differences that may exist among users of different types of dietary supplements(Reference Reedy, Haines and Campbell15, Reference Hoggatt, Bernstein and Reynolds29). This observation has prompted some authors to suggest that there is a clear need for more research to investigate similarities and differences among sub-categories of dietary supplement users(Reference Reedy, Haines and Campbell15, Reference Fennell18).
Thus, the aims of the present study were to estimate the prevalence of dietary supplement use, at baseline, in adults taking part in a longitudinal cancer cohort study in Alberta, Canada (The Tomorrow Project(Reference Bryant, Robson, Ullman, Friedenreich and Dawe30)) and to explore the extent to which differences in self-reported sociodemographic, dietary and lifestyle characteristics exist between users of different types of dietary supplements and non-users of supplements.
Subjects and methods
Subject recruitment
Full details describing subject recruitment and enrolment to The Tomorrow Project are described elsewhere(Reference Bryant, Robson, Ullman, Friedenreich and Dawe30). In brief, participants were recruited to the longitudinal cohort study using a two-stage random sampling design. The first stage identified households using a telephone random digit dial (RDD) method, while the second stage identified an eligible individual from within each household.
Eligibility criteria were: age between 35 and 69 years; no personal history of cancer, other than non-melanoma skin cancer; planning to stay in Alberta for one year; and able to complete written questionnaires in English. In households with more than one eligible participant, the person with the most recent birthday was selected(Reference Lavrakas31). In the first of five recruitment waves, we piloted an approach of recruiting two eligible adults from each household, but this approach was not continued in the latter waves.
The study was described to potentially eligible individuals as a long-term project that would help researchers learn about the causes of cancer. Eligible adults who expressed interest in taking part received a consent form and a Health and Lifestyle Questionnaire (HLQ) by mail. Questionnaires designed to ascertain past year diet and physical activity were sent to participants who returned a completed HLQ and consent form. All questionnaires were self-administered. Data analysed in the present study were obtained between February 2001 and November 2004.
Dietary assessment
Dietary habits were assessed using the US National Cancer Institute’s Diet History Questionnaire (DHQ), adapted for use in Canada(Reference Csizmadi, Kahle, Ullman, Dawe, Zimmerman, Friedenreich, Bryant and Subar32, 33).
Dietary supplements
Forty-four different supplements, divided into three separate categories, were queried on the DHQ: multivitamins (with or without minerals); sixteen specific nutritional supplements not consumed as part of a multivitamin/mineral; and twenty-seven herbal/other supplements. Use of multivitamins (with or without minerals) was queried in a single question, and ‘One-a-day-, or Centrum-type multivitamins (as pills, liquids, or packets)’ were given as examples. For the purposes of the present analysis, we defined specific nutritional supplements as those products that contained at least one component with established nutritional value, but that were not consumed as part of a multivitamin/mineral. The term ‘herbal/other’ was used to describe supplements of predominantly herbal/botanical origin and those for which no firm nutritional value has yet been established in healthy adults. Products included in the specific nutritional and herbal/other groups of dietary supplements queried on the DHQ are presented in Table 1.
Table 1 Classification of dietary supplements queried on the Diet History Questionnaire(Reference Csizmadi, Kahle, Ullman, Dawe, Zimmerman, Friedenreich, Bryant and Subar32, 33) completed by participants in The Tomorrow Project
Regular supplement users were defined as participants who reported consumption of at least one type of dietary supplement at least once weekly in the year prior to DHQ completion. For the purposes of the present analysis, the remaining participants, who used supplements less than once weekly or not at all, were combined into one group and referred to as non-users. Following dichotomization, regular users were assigned to one of seven discrete categories: (i) multivitamins (with or without minerals) only (M); (ii) specific nutritional supplements only (N); (iii) herbal/other supplements only (H); (iv) multivitamins and specific nutritional supplements (MN); (v) multivitamins and herbal/other supplements (MH); (vi) specific nutritional and herbal/other supplements (NH); and (vii) multivitamins, specific nutritional and herbal/other supplements (MNH). These categories represented all possible combinations of supplement type queried on the DHQ.
Food and nutrient intakes
Intakes of foods and beverages reported by each person using the DHQ were analysed using the Diet*Calc software version 1·4·2 (National Cancer Institute, Bethesda, MD, USA) to provide information on mean daily nutrient intakes and servings of selected food groups.
Lifestyle and sociodemographic variables
A validated Past Year Total Physical Activity Questionnaire (PYTPAQ)(Reference Friedenreich, Courneya, Neilson, Matthews, Willis, Irwin, Troiano and Ballard-Barbash34) was used to estimate the total number of hours per week spent by each participant performing occupational, transportation, household and recreational activities at vigorous intensity (>6 MET(Reference Ainsworth, Haskell, Leon, Jacobs, Montoye, Sallis and Paffenbarger35, Reference Ainsworth, Haskell and Whitt36), where MET=metabolic energy equivalent task). Smoking status, height, body weight, sex, age, marital status, highest level of educational attainment and total household income before tax were assessed using the self-administered HLQ.
Statistical analyses
In order to account for the complex sampling procedure used in The Tomorrow Project, sample weights were calculated in five steps: adjustments for (i) the probability of household selection, (ii) telephone under-coverage, (iii) household non-response, (iv) person non-response and (v) post-stratification by health region of residence at time of recruitment, sex and age, using population estimates for Alberta(37). The approach of calculating sample weights was based on the weighting strategy used in the Canadian Community Health Survey(38). As the first step in the analysis, Rao–Scott design-adjusted χ 2 tests(Reference Rao and Scott39–Reference Rao and Scott41) were used to explore associations between regular supplement use and sociodemographic, dietary and lifestyle characteristics that had been selected a priori as being consistent with general health guidelines. Multinomial logistic regression models(Reference Hosmer and Lemeshow42) were then used to investigate differences and similarities in these characteristics between non-users and the seven discrete categories of supplement use.
All analyses were performed with weighted data, using the PROC SURVEYFREQ and PROC SURVEYLOGISTIC procedures available in the Statistical Analysis System (SAS) statistical software package version 9·1·3 (SAS Institute Inc., Cary, NC, USA). These procedures used the Taylor expansion method to estimate standard errors of estimates(Reference Binder43, Reference Woodruff44). All tests for significance of estimated logit coefficients were two-sided and performed with Wald tests at a 5 % significance level. Owing to the complex sampling design, the use of likelihood ratio tests was not appropriate, as these tests assume independence between observations(Reference Roberts, Rao and Kumar45, 46).
Ethical approval
Ethical approval for The Tomorrow Project was obtained from the Research Ethics Committees of the Alberta Cancer Board and the University of Calgary, Alberta, Canada.
Results
Of the 29270 individuals recruited on the basis of the eligibility criteria outlined in the RDD protocol, 15046 (51·4 %) enrolled in the cohort by returning a consent form and HLQ. Assuming that the ratio of eligible to ineligible in those for whom a screening interview could not be completed was the same as that in the successfully screened group, it is estimated that the enrolled sample represented about 31 % of all potential participants. DHQ and PYTPAQ were not returned by 1859 (12·4 %) participants. Using the raw data, no significant differences in education or BMI were found between those who returned DHQ and PYTPAQ and those who did not. However, women, people aged 55–69 years and people living with a partner were over-represented in the group that returned all three baseline questionnaires relative to the group that returned the HLQ only.
Following exclusion of those who did not return DHQ and PYTPAQ, we also excluded the remaining subjects who were recruited as ‘second in household’ in the first recruitment wave (n 344), people outside the 35–69 year age range at the time of completing the HLQ (n 19), pregnant women (n 24), transgender subjects (n 2), people with BMI < 18·5 kg/m2 (n 85) and those with incomplete data (n 36). In addition we undertook a data linkage with the Alberta Cancer Registry, and subsequently excluded those participants who were flagged as having had a previous diagnosis of cancer before enrolment into the study (n 171). We also considered excluding participants who reported implausible energy intakes (n 504), as defined by Hung et al.(Reference Hung, Joshipura, Jiang, Hu, Hunter, Smith-Warner, Colditz, Rosner, Spiegelman and Willett47), but as this exclusion had no significant impact on our results (data not shown) these records were retained for analysis, giving a final sample of 12506 respondents.
Just over two-thirds of participants reported using at least one dietary supplement at least once weekly during the 12 months prior to DHQ completion (Table 2). Within the study sample use of multiple supplements was common, with just over a quarter of users reporting that they had consumed five or more different supplements at least once weekly (data not shown). Among women who were regular supplement users, the most commonly reported specific nutritional and herbal/other supplements were calcium, reported by 60·3 %, and glucosamine, reported by 20·3 % of users. In contrast, vitamin C (37·4 %) and garlic (18·9 %) were the specific nutritional and herbal/other supplements reported most commonly by men.
Table 2 Categories of dietary supplement use reported by participants in The Tomorrow ProjectFootnote †
† Excludes ‘second in household’ recruits and those who were <35 years or >69 years, pregnant, transgender, underweight, had had prior cancer or had incomplete data.
‡Weights calculated in five steps: adjustments for the probability of household selection, telephone under-coverage, household non-response, person non-response and post-stratification adjustment by health region of residence at time of recruitment, sex and age.
§ Regular users of dietary supplements were defined as those participants who reported consumption of at least one supplement at least once weekly in the year prior to questionnaire completion. All other participants were defined as non-users.
Table 3 presents weighted estimates of percentages with 95 % confidence intervals for the sociodemographic, dietary and lifestyle characteristics reported by all participants included in the current analysis, and also presents the percentages of people in each category who were regular supplement users. Rao–Scott χ 2 tests (data not shown) revealed significant associations between dietary supplement use and most of the characteristics presented in the table, with the exception of educational attainment and participation in vigorous physical activity.
Table 3 Sociodemographic, dietary and lifestyle characteristics reported by adults participating in The Tomorrow Project† and percentage who are regular supplement users
†Excludes those who did not return all questionnaires, ‘second in household’ recruits, and those who were <35 years or >69 years, pregnant, transgender, underweight, had had prior cancer or had incomplete data.
‡Weights calculated in five steps: adjustments for the probability of household selection, telephone under-coverage, household non-response, person non-response and post-stratification adjustment by health region of residence at time of recruitment, sex and age.
§Total time (hours per week) performing occupational, transportation, household and recreational activities at intensities greater than 6 MET (MET=metabolic energy equivalent task).
To investigate differences and similarities in dietary and lifestyle characteristics reported by participants in each of the seven discrete categories of regular supplement use, relative to those in the non-use group, we estimated a multinomial logistic model. Preliminary explorations of our data demonstrated that income was significantly positively associated with education and, as we had fewer missing values for the education variable, we chose to use the latter as our indicator of socio-economic status. Thus the final models controlled for sex, age, marital status and education. As our preliminary analyses indicated significant interaction effects for supplement use between age and sex, and between marital status and sex (data not shown), the final models presented in Table 4 were estimated separately for men and women. The odds ratios presented in Table 4 can be interpreted as follows: men aged 55–69 years, relative to men aged 35–44 years, were three times more likely to be in the MNH group than in the non-users group. Similarly, men who were married/living with someone, relative to men who were not cohabiting, were 0·4 (i.e. 1·0 minus 0·6) times less likely to be in the MNH group than in the non-user group.
Table 4 Sociodemographic, dietary and lifestyle characteristics reported by adults in each of seven discrete categories of dietary supplement use compared with those reported by supplement non-usersFootnote †, Footnote ‡
M, multivitamins only; N, specific nutritional supplements only; H, herbal/other supplements only; MN, multivitamins and specific nutritional supplements; MH, multivitamins and herbal/other supplements; NH, specific nutritional and herbal/other supplements; MNH, multivitamins, specific nutritional and herbal/other supplements.
* P < 0·05, **P < 0·01 and ***P < 0·001, assessed using Wald tests. In the first column, asterisks beside the description of the characteristic indicate the range for the P value resulting from Wald tests for the null hypothesis that there is no relationship between the characteristic and the outcome (i.e. supplement use). Asterisks that follow odds ratios in the body of the table represent the range for the P values resulting from the Wald tests for the null hypothesis that a particular estimated logit coefficient is zero.
† Analyses were based on weighted data from baseline questionnaires completed by participants in The Tomorrow Project, a longitudinal cohort for the study of cancer aetiology.
‡ Comparisons were undertaken by estimating multinomial logistic regression models. Full models were estimated separately for men and women, using non-users as the comparison group. All variables presented in the table were included in the final model estimated for each sex. The first row within each category of the sociodemographic, dietary and lifestyle characteristics is the reference category.
§ The odds ratios can be interpreted as follows: men aged 55–69 years, relative to men aged 35–44 years, were three times more likely to be in the MNH group than in the non-user group. Similarly, men who were married/living with someone, relative to men who were not cohabiting, were 0·4 (i.e. 1·0 minus 0·6) times less likely to be in the MNH group than in the non-user group.
Men and women in the N, MN, NH and MNH groups were more likely than non-users to be in the older age group (Table 4). However, men and women who used multivitamins only (M) and multivitamins with herbal/other supplements (MH), and women who used herbal/other supplements only (H), were not significantly different from non-users with respect to age. Furthermore, the majority of dietary and lifestyle characteristics reported by men and women who used herbal/other supplements only (H) were not significantly different from those of non-users. Similarly, women who reported use of multivitamins only (M) did not differ significantly from non-users with respect to most sociodemographic characteristics or dietary and lifestyle behaviours.
Men and women in the MN, NH and MNH groups were more likely than non-users to report consumption of at least five servings of fruit and vegetables per day. In addition, women in the MNH and MN groups were less likely than non-users to be smokers, and more likely to report at least one daily serving of whole-grain foods, less than 10 % energy from saturated fat and more participation in vigorous physical activity. A broadly similar pattern was observed for men in the MNH group. With respect to marital status, men and women in the MNH group were less likely than non-users to be married or living with someone. In addition, men in the MN and NH groups were less likely than non-users to report cohabitation.
Discussion
Approximately 70 % of participants in The Tomorrow Project were classified as regular users of dietary supplements. This estimate is somewhat higher than the 45–50 % reported in two smaller Canadian surveys(Reference Troppmann, Johns and Gray-Donald4, 48) and the two most recent large US population-based studies(Reference Radimer, Bindewald, Hughes, Ervin, Swanson and Picciano3, Reference Fennell18). One possible reason for the apparent disparity in prevalence estimates could be the fact that we assessed supplement use in adults aged 35–69 years, whereas other North American surveys have included people from the age of 18 years(Reference Radimer, Bindewald, Hughes, Ervin, Swanson and Picciano3, Reference Troppmann, Johns and Gray-Donald4, Reference Fennell18, 48). As supplement use tends to increase with increasing age(Reference Radimer, Bindewald, Hughes, Ervin, Swanson and Picciano3, Reference Fennell18, Reference Block, Cox, Madans, Schreiber, Licitra and Melia28, Reference Satia-Abouta, Kristal, Patterson, Littman, Stratton and White49, Reference Yu, Kogan and Huang50), this could account, at least in part, for the observation that our estimate of regular use was higher than has been reported by others.
However, a more likely explanation is the fact that our subjects were participants in a long-term cohort being established for the study of cancer aetiology. Compared with the two most recent US population-based surveys, which reported response rates of 72·1 % (National Health Interview Survey; NHIS(Reference Fennell18)) and 82 % (National Health and Nutrition Examination Survey(Reference Radimer, Bindewald, Hughes, Ervin, Swanson and Picciano3)), our estimated response rate was relatively low (31 %). Furthermore, it is likely that our participants may be more health-conscious than the general population, suggesting that our estimate of the prevalence of supplement use is likely to be somewhat biased.
The lower response rate obtained in The Tomorrow Project is not particularly surprising, given that the participants were invited to enrol in a longitudinal cohort study, rather than a one-off cross-sectional survey. Previous comparisons of the sample with the Albertan population have demonstrated no significant differences with respect to marital status and income. In contrast, however, our participants reported greater educational attainment, but they also reported higher BMI, suggesting that there is likely to be more than a simple ‘healthy user’ bias operating within our sample(Reference Bryant, Robson, Ullman, Friedenreich and Dawe30). Furthermore, we attempted to account for selection bias by calculating sample weights and by using statistical approaches appropriate for surveys with complex sampling procedures.
Although our prevalence of use estimates may be somewhat biased, we did observe sociodemographic, dietary and lifestyle differences among discrete groups of supplement users in our study population. These results are difficult to compare directly with previous work because the methods used to group subjects on the basis of use of different types of supplements have varied widely between studies(Reference Troppmann, Johns and Gray-Donald4, Reference Reedy, Haines and Campbell15, Reference Fennell18, Reference Messerer, Johansson and Wolk23, Reference Hoggatt, Bernstein and Reynolds29). The lack of consistency in the published literature means that there is no strong theoretical framework to guide the methods that should be used in this type of work. Therefore, the approach used in the current study was driven by our need to try to understand the extent to which users of different types of supplements may differ from non-users and from each other. Thus the decision to create seven discrete groups was a pragmatic one; it represented all possible combinations of the groups of supplements and did not rely on the creation of arbitrary groupings.
However, despite the difficulties encountered in making direct comparisons between studies, our finding that men and women who used multiple types of supplements tended to report characteristics that are more consistent with established health guidelines supports the results of previous studies(Reference Reedy, Haines and Campbell15, Reference Hoggatt, Bernstein and Reynolds29). Similarly, our finding that women who used multivitamins only tended to report characteristics that were not significantly different from those reported by non-users is a theme that has emerged in the few previous studies that have investigated this issue(Reference Troppmann, Johns and Gray-Donald4, Reference Reedy, Haines and Campbell15, Reference Block, Cox, Madans, Schreiber, Licitra and Melia28, Reference Hoggatt, Bernstein and Reynolds29). It is possible that the lack of statistically significant differences observed in the current study could be a function of the relatively small size of the multivitamin only group. However, previous studies that have reported similar findings have speculated that users in this category may be taking a multivitamin/mineral as ‘nutrient insurance’, without considering the need to participate in other healthy lifestyle behaviours(Reference Reedy, Haines and Campbell15, Reference Foote, Murphy, Wilkens, Hankin, Henderson and Kolonel20). None the less, this hypothesis is unlikely to explain our observation that users of herbal/other supplements only reported characteristics that were little different from those reported by non-users. Clearly, this is an area that requires further research.
One other intriguing finding was the association between marital status and supplement use. Users of all types of dietary supplements (MNH) were less likely than supplement non-users to be married or living with someone. Of the few other studies that have examined associations between marital status and supplement use, results have been conflicting. For example, the Multiethnic Cohort Study reported that marital status was weakly and inconsistently associated with supplement use. Specifically, those participants who were divorced, separated or widowed were slightly more likely to report supplement use than people who were married. However, this relationship was apparent only in some ethnic groups(Reference Foote, Murphy, Wilkens, Hankin, Henderson and Kolonel20). Another study of US female physicians reported that women who had never married were least likely to use supplements, whereas widowed women were most likely to be supplement users(Reference Frank, Bendich and Denniston27). Analysis of NHIS 2000 data reported that unmarried people were more likely to use herbal supplements than married people, but no such associations were observed for vitamins and minerals(Reference Fennell18). Others have reported no association between marital status and supplement use(Reference Marques-Vidal, Arveiler and Evans22, Reference Wallström, Elmståhl, Hanson, Östergren, Johansson, Janzon and Larsson51).
In the absence of data concerning motivations for use of dietary supplements in the present study, the reasons for the apparent differences in health-related and sociodemographic characteristics reported by discrete groups of supplement users can only be speculated upon. It has been suggested that people use dietary supplements for health maintenance, to reduce risk of developing chronic diseases or to treat existing conditions(Reference Neuhouser5, Reference Satia-Abouta, Kristal, Patterson, Littman, Stratton and White49, Reference Conner, Kirk, Cade and Barrett52). Although participants in The Tomorrow Project were asked about previous diagnoses of chronic conditions, we could not ascertain whether those who indicated that they had ever received such a diagnosis still had that condition. Therefore, while we observed that those who took all types of dietary supplements were more likely than non-users to report behaviours associated with healthier lifestyles, we cannot determine the extent to which this phenomenon was associated with health maintenance, disease prevention or treatment of pre-existing disease. This is an area that needs further examination, and it is anticipated that future follow-up surveys undertaken with The Tomorrow Project cohort will attempt to explore motivations for supplement use, in addition to capturing more in-depth data concerning types, doses, frequency and duration of use of different types of supplements.
The present study does have several other limitations that should be borne in mind when attempting to compare its findings with previous and future studies. First, dietary habits and supplement use were assessed using the National Cancer Institute’s DHQ that had been modified to account for Canadian food fortification practices(Reference Csizmadi, Kahle, Ullman, Dawe, Zimmerman, Friedenreich, Bryant and Subar32). The supplement section was not modified, owing to a paucity of up-to-date, national population-based data concerning the types of supplements used by Canadian adults at that time. Thus we cannot be certain that the DHQ list adequately reflects the types of supplements likely to have been consumed in our population. However, the supplements reported most commonly in the study were broadly in line with those described in the Food Habits of Canadians Study(Reference Troppmann, Johns and Gray-Donald4) and the British Columbia Nutrition Survey(48), both of which used open-ended questions to assess types of supplements consumed by their participants.
Despite these limitations, the results of the present study support previous observations that the relationships between use of different types of dietary supplements and sociodemographic, dietary and lifestyle characteristics are complex and cannot be accounted for by simply dichotomizing subjects as users or non-users of dietary supplements(Reference Reedy, Haines and Campbell15, Reference Hoggatt, Bernstein and Reynolds29). If this issue is to be investigated further, there is a clear need for the implementation of standardized methods designed specifically to assess supplement use, in order to better describe and assess the prevalence, types, doses, frequencies and motivations for use of different types of supplements. Standardized methods of assessing these aspects of supplement use, as well as categorizing study participants, will help researchers to compare results more easily between studies, thereby moving this area of research forward. In the long run, such information will also be useful when trying to disentangle the effects of use of different types of supplements, as well as dietary and lifestyle habits, on chronic disease risk.
Acknowledgements
Funding: The Tomorrow Project receives support from the Alberta Cancer Board and the Alberta Cancer Foundation.
Contributions: P.J.R. conceptualized the study and P.J.R. and G.L.S. developed the analysis plan. G.L.S. undertook the analysis and P.J.R. and G.L.S. interpreted the data. H.E.B. was responsible for obtaining funding for The Tomorrow Project, as well as overseeing the choice of data acquisition methods. R.U. oversaw subject enrolment and data acquisition and was responsible for the day-to-day operations of the cohort study. H.E.B. contributed substantial intellectual content and critically reviewed the manuscript.
Conflicts of interest: None of the authors has any conflict of interest to declare.
Thanks: The authors would like to thank the participants of The Tomorrow Project for their participation in the study. In addition, we acknowledge the contributions made by study staff in data entry and processing. Finally, we wish to thank Drs Elizabeth McGregor, Christine Friedenreich, Ilona Csizmadi and Karen Kopciuk of the Alberta Cancer Board for their helpful comments on the manuscript. The latter four people have given permission to be acknowledged.
