Original quantitative research – Gender differences in the longitudinal association between multilevel latent classes of chronic disease risk behaviours and body mass index in adolescents
This article has been peer reviewed.
Correspondence: Nour Hammami , McGill University, Institute for Health and Social Policy, 1130 Pine Avenue West, Montréal, QC H3A 1A3; Tel: 514-398-6764; Email: email@example.com
Introduction: Few studies have assessed the relationship between chronic disease risk behaviours and body mass index (BMI) in a longitudinal, sex/gender-specific context. This study used gender-specific analyses to assess the extent to which chronic disease risk behaviour latent classes are associated with BMI and weight status at follow-up.
Methods: Longitudinal data from 4510 students in Grades 9 to 12, tracked from 2013–2015, who participated in the COMPASS study were used to assess gender differences in the lagged association between previously determined latent classes (of physical activity and substance use) with BMI using multilevel mixed-effects models. Our multilevel regression models assessed the association between two latent classes, active experimenters and inactive non-using youth, with BMI when stratified by gender.
Results: Male inactive non-substance-using youth were associated with a 0.29 higher continuous BMI (95% CI: 0.057, 0.53) and odds of overweight/obesity increased by 72% (OR = 1.72, 95% CI: 1.2, 2.4) for binary BMI at follow-up relative to active youth who experiment with substance use. No significant associations were detected in females.
Conclusion: Over time, physical activity has a protective role on BMI in male youth. Both substance use and physical inactivity should be addressed in obesity prevention efforts. Gender stratification in analyses is also important since females and males have different contributing factors to increases in BMI.
Keywords: chronic disease risk behaviours, substance use, physical activity, sex, gender, BMI, obesity, overweight, adiposity
- Approximately 20% of youth who were overweight/obese were in the normal weight body mass index (BMI) category 1 year later.
- Male inactive non-substance users had increases in BMI and increases in the odds of becoming overweight/obese relative to their active substance-using peers. No associations were identified in females.
- Physical activity seems to play a role in males in prospectively maintaining BMI despite their engagement with substance use.
Overweight and obesity are escalating in youth: 27% of Canadian children were classified as overweight or obese in 2013.Footnote 1 Childhood obesity tracks into adulthood, increasing the risk of adult chronic diseases.Footnote 2Footnote 3 That body mass index (BMI) decreases in youth over time (in males more than in females), but not in adults, highlights the importance of prevention efforts directed at this age group.Footnote 1Footnote 4Footnote 5
Despite public health prevention and intervention efforts, BMI is increasing in certain populations, including in Canadian youth.Footnote 5Footnote 6 Overweight/obesity have many determining factors, of which chronic disease risk behaviours (CDRB; e.g. physical inactivity, binge drinking) are major contributors.Footnote 7Footnote 8 Laxer et al.Footnote 6 used longitudinal data (2012–2014) from the Cohort Study on Obesity, Marijuana Use, Physical Activity, Alcohol Use, Smoking and Sedentary Behaviour (COMPASS) to assess the effect of 15 CDRB (including physical inactivity, dietary choices and sedentary and substance use behaviours) at baseline (via latent classes) on BMI at concurrent years, while controlling for gender. The authors found that BMI increased by an average of 0.61 units per year; however, the researchers were unable to identify a specific latent class that had higher risk of increasing BMI.
Devis-Devis et al.Footnote 4 and Jackson & CunninghamFootnote 9 did not find an association between CDRB latent classes and obesity, and recommended further investigation.Footnote 4 A limitation of these two studies is that they only included physical activity, sedentary behaviours and diet,Footnote 4Footnote 10 despite evidence that substance users (i.e. smokers and marijuana users) are most likely to have overweight/obesity relative to their non-substance-using counterparts.Footnote 11Footnote 12Footnote 13
Another notable shortcoming of previous research has been the lack of assessment of the role of sex/gender in the association between CDRB and obesity.Footnote 6Footnote 9Footnote 14Footnote 15 For instance, Laxer et al. reported that males were associated with a higher increase in BMI, but did not conduct sex/gender-specific modelling.Footnote 6 Published studies have demonstrated sex/gender differences in BMIFootnote 16Footnote 17 and in CDRB engagement—including physical inactivity and substance useFootnote 17Footnote 18, indicating that sex/gender-specific models are warranted. Research has also found that physical inactivity and substance use play roles in youth overweight and obesity,Footnote 14Footnote 15Footnote 17Footnote 19Footnote 20 showing the importance of incorporating substance use, as well as sex/gender-stratified analyses, into research of obesity in youth.
Given the notable gaps in obesity literature, Hammami et al.Footnote 17 identified latent classes of CDRB (physical inactivity, binge drinking, marijuana use and tobacco smoking) in 2013–2015 and regressed BMI onto these classes (in repeated cross-sectional analyses) in youth in Ontario, Canada, participating in COMPASS. The authors found that latent classes with inactivity and substance use in females were associated with higher odds of overweight/obesity relative to active and non-substance-using females; in contrast, activity and experimenting with substance use in males were associated with higher odds of overweight/obesity relative to inactive non-substance-using males.Footnote 17
However, whether individuals in these latent classes had higher BMIs, for both sexes/genders, relative to their counterparts at follow-up remains unknown. This information is crucial for obesity and substance use prevention programs because CDRB are modifiable, and addressing them while youth are at school can help mitigate the impact of factors associated with high/increasing BMI. As such, the aim of this study was to investigate the prospective association of CDRB latent classes, namely physical inactivity, binge drinking, marijuana use and tobacco smoking, with BMI, while accounting for gender, among participants in COMPASS in Ontario, Canada.
COMPASS is a large longitudinal study (2012–2021) collecting behaviour and outcome data from secondary school students in Canada. Further information on COMPASS (http://www.compass.uwaterloo.ca) is available elsewhere.Footnote 21
This study used three waves of COMPASS data from Ontario, Canada. Wave 1 was collected in the 2013/14 school year, Wave 2 in the 2014/15 school year and Wave 3 in the 2015/16 school year. Consistent with earlier research,Footnote 6Footnote 17 we chose to focus our attention on Ontario data as these constitute 92% of the observations for these waves of COMPASS data.
A total of 41 734 youth in Grades 9, 10, 11 or 12 responded to the student questionnaire in Wave 1, 39 013 responded in Wave 2 and 37 106 responded in Wave 3. Most of the students who did not respond (20.9%, 21.6% and 20.1% in Waves 1, 2 and 3, respectively) were absent from school the day the questionnaire was administered. Students were recruited from schools that permit active-information, passive-consent protocols (n = 79, 78 and 72 in Waves 1, 2 and 3, respectively). In addition to the approval of the schools and school boards, the University of Waterloo Office of Research Ethics approved all procedures. Passive consent was obtained from participants.
Schools that participated in at least two of the three waves (n = 70 in each wave) were included in this study. Youth who responded to the student questionnaire more than once (n = 6594) were included in the study.
Body mass index (dependent variable)
We calculated BMI from the self-reported height and weight measures. We used the World Health Organization sex-specific BMI-for-age cut-off values corresponding to the age of our sample.Footnote 22 The measures used to determine BMI in COMPASS participants have previously been validated (intraclass correlation coefficient [ICC] = 0.84).Footnote 23
BMI calculated at each time-point ranged from 10.0 to 49.9 kg/m2, which suggests the presence of outliers. On removal of outliers at the 1% and 99% ends of the range, BMI was 15.5 to 35.9. BMI was used as a continuous and as a binary outcome (weight status) for comparative purposes. We used weight status based on BMI cut-offs (overweight/obese versus normal weight) because youth classified as overweight or obese have similar risks of future chronic diseasesFootnote 24; in addition, doing so is consistent with previously published studies.Footnote 15Footnote 17Footnote 25
Chronic disease risk behaviours (independent variables)
The CDRB measures and multilevel latent class analysis procedure are briefly described below. (For more information, see Hammami et al.Footnote 17)
Substance use behaviours
To identify current cigarette smokers, the questionnaire included questions asking (1) if respondents had ever smoked 100 or more whole cigarettes in their life; and (2) on how many days respondents had smoked one or more cigarettes in the past 30 days. Students who answered “yes” to the first question, and reported any smoking in the previous 30 days were identified as current smokers.Footnote 17Footnote 27
To identify binge-drinking behaviour, respondents were asked how often they had had five or more drinks of alcohol on one occasion during the past 12 months. Current binge drinkers were identified as those who had had five or more drinks at least once in the last month.Footnote 17Footnote 28Footnote 29
Respondents were asked how often they used marijuana or cannabis during the past 12 months. They were classified as current marijuana users if they had used marijuana in the last month.Footnote 17Footnote 28Footnote 29
Chronic disease risk behaviour (CDRB) latent classes
We previously conducted multilevel latent class analysis using gender-specific models for Waves 1–3 of the data in this study to independently assess the consistency of the CDRB profiles identified over time.Footnote 17 The findings suggested either two latent classes, active experimenters and inactive non-users, or three latent classes, active experimenters, inactive non-users and inactive substance users.
To ensure that the classes studied over time (in terms of their association with BMI) were comparable, we assumed that the classes were fixed in number and type.Footnote 30 Since all waves in our study had at least two latent classes, we performed our longitudinal analyses with a parsimonious model of two student latent classes; in addition, this parsimony makes for easier interpretation and communication of findings.
Other variables: ethnicity and gender
We identified ethnicity based on responses to the question “How would you describe yourself?” in the student questionnaire. Options for answers were “White,” “Black,” “Asian,” “Aboriginal (First Nations, Métis, Inuit),” “Latin American/Hispanic” or “Other.” We grouped all the non-White ethnicities as they constituted only about 25% of the sample. In these analyses ethnicity is only used as a control variable.
We identified gender based on the answer to the question “Are you female or male?” with the options “female” and “male” as answers.
For variable descriptive statistics, we calculated gender-specific frequencies and percentages for categorical variables of interest and reported means and standard deviations for continuous variables. All analyses in this study used statistical package SAS version 9.4 (SAS Institute Inc., Cary, NC, USA) with a threshold of significance p < .05.
Bivariate exploratory analysis was conducted using McNemar test to assess the degree to which youth changed weight status categories and latent classes (i.e. transitions) across consecutive waves.
We used gender-stratified mixed-effects regression models to assess the longitudinal association between (lagged) CDRB latent classes and BMI (at follow-up). These models considered the outcome BMI in two ways: as continuous BMI and as binary BMI (overweight/obese versus normal weight). All mixed models adjusted for the following predictors while accounting for the hierarchical structure of the data: BMI (in the previous wave), ethnicity (at baseline), grade (at current wave) and year.
All mixed-effects regression models restricted analysis to monotone patterns of missingness in the outcome variable via maximum likelihood, based on the assumption that the data are “missing at random.”Footnote 31 Altogether 4510 youth participated in at least two student questionnaires across the three waves (with a monotone pattern of missing BMI) and were included in the analyses.
For each of the three Waves, females accounted for 51.1% of secondary school students responding (Table 1). The mean BMI at Wave 1 was 21.0 in females and 21.6 in males (classified as normal weight). Estimates were similar at Waves 2 (females, 21.6; males, 22.4) and 3 (females, 21.9; males, 22.9).
Table 1. Summary statistics for longitudinal outcomes and covariates, by gender, Grade 9–12 secondary school students, COMPASS (Ontario, Canada)
|Demographics||Study participants, % (n)Footnote a of Table 1|
|Wave 1||Wave 2||Wave 3|
|Gender distribution||51.1 (2307)||48.9 (2203)||51.1 (2307)||48.9 (2203)||51.1 (2307)||48.9 (2203)|
|White||51.5 (1826)||48.5 (1721)||–||–||–||–|
|Non-White||50.2 (474)||49.8 (470)||–||–||–||–|
|9||50.7 (1551)||49.3 (1505)||–||–||–||–|
|10||53.0 (724)||47.0 (643)||50.7 (1541)||49.3 (1500)||–||–|
|11||36.8 (32)||63.2 (55)||53.1 (728)||46.9 (643)||50.7 (1532)||49.3 (1488)|
|12||–||–||37.4 (31)||62.6 (52)||52.3 (747)||47.7 (682)|
|Weight status (binary BMI)Footnote b of Table 1|
|Normal||55.2 (1896)||44.8 (1541)||55.0 (1782)||45.0 (1457)||55.2 (1642)||44.8 (1334)|
|Overweight/obese||38.3 (411)||61.7 (662)||39.9 (413)||61.1 (648)||38.5 (362)||61.5 (578)|
|Continuous BMI (SD)||21.0 (3.1)||21.6 (3.4)||21.6 (3.2)||22.4 (3.5)||21.9 (3.2)||22.9 (3.5)|
|Active experimenters||59.0 (128)||41.0 (89)||49.9 (253)||50.1 (254)||46.0 (354)||54.0 (415)|
|Inactive non-users||50.8 (2179)||49.2 (2114)||51.3 (2054)||48.7 (1949)||52.2 (1953)||47.8 (1788)|
More than half of the female youth reported a normal BMI (55.2%, 55.0% and 55.2% for Waves 1, 2 and 3, respectively). This was higher than the proportion of males reporting a normal BMI (44.8%, 45.0% and 44.8%, respectively). In tandem, males reported higher overweight/obesity rates (61.7%, 61.1% and 61.5%, respectively) than their female counterparts (38.3%, 39.9% and 38.5%, respectively).
Weight status transitions
Most youth who were classified as having a normal weight remained in this category at follow-up (Wave 1 to 2: females = 92.8%, males = 89.5%; Wave 2 to 3: females = 95.0%, males = 90.3%) (Table 2). Youth with overweight/obesity also tended to remain in the same category across consecutive waves (Wave 1 to 2: females = 75.1%, males = 78.4%; Wave 2 to 3: females = 78.4%, males = 77.7%).
Table 2. Transitions in weight status across consecutives waves (Waves 1 to 2 and 2 to 3), by gender, Grade 9–12 secondary school students, COMPASS (Ontario, Canada)
|Categories||Binary BMI statusFootnote a of Table 2, % (n)||McNemar chi-square|
|Normal||92.8 (1688)||7.2 (130)||5.78Footnote * of Table 2|
|Overweight/Obese||24.9 (94)||75.1 (283)|
|Normal||89.5 (1321)||10.5 (155)||1.24|
|Overweight/Obese||21.6 (136)||78.4 (493)|
|Normal||95.0 (1565)||5.0 (82)||0.16|
|Overweight/Obese||21.6 (77)||78.4 (280)|
|Normal||90.3 (1205)||9.7 (129)||0|
|Overweight/Obese||22.3 (129)||77.7 (449)|
A lower proportion of females than of males reported transitioning from a normal weight to overweight/obesity at follow-up (Wave 1 to 2: females = 7.2%, males = 10.5%; Wave 2 to 3: females = 5.0%, males = 9.7%). A greater proportion of female youth transitioned from overweight/obesity to normal weight at follow-up (Wave 1 to 2: females = 24.9%, males = 21.6%; Wave 2 to 3: females = 21.6%, males = 22.3%). These transitions were found to be significant only in females for Wave 1 to 2.
Latent class transitions
Most active experimenters were the same at follow-up (Waves 1 to 2: females = 71.9%, males = 71.9%; Waves 2 to 3: females = 70.7%, males = 70.9%) (Table 3). Similarly, inactive non-users remained largely non-using at follow-up (Waves 1 to 2: females = 92.6%, males = 91.0%; Waves 2 to 3: females = 91.5%, males = 87.9%).
Table 3. Transitions in latent classes across consecutive waves (Waves 1 to 2 and 2 to 3), by gender, Grade 9–12 secondary school students, COMPASS (Ontario, Canada)
|Categories||Chronic disease risk behaviour latent classes, % (n)||McNemar chi-square|
|Active experimenters||Inactive non-users|
|Active experimenters||71.9 (92)||28.1 (36)||79.3Footnote *** of Table 3|
|Inactive non-users||7.4 (161)||92.6 (2018)|
|Active experimenters||71.9 (64)||28.1 (25)||126.6Footnote *** of Table 3|
|Inactive non-users||9.0 (190)||91.0 (1924)|
|Active experimenters||70.7 (179)||29.3 (74)||41.0Footnote *** of Table 3|
|Inactive non-users||8.5 (175)||91.5 (1879)|
|Active experimenters||70.9 (180)||29.1 (74)||83.9Footnote *** of Table 3|
|Inactive non-users||12.1 (235)||87.9 (1714)|
Transition from active experimenting to inactive non-user status occurred at higher rates than the reverse, that is, from inactive non-using to active experimenting status (Wave 1 to 2: females = 28.1%, males = 28.1%; Waves 2 to 3: females = 29.3%, males = 29.1%). A significant McNemar chi-square test statistic (p < .0001) suggests that there are statistically significant transitions in youth’s CDRB latent classes across a 1-year period.
Longitudinal regression analyses
Male inactive non-users in the previous wave were associated with an average increase of 0.29 in continuous BMI at follow-up relative to their active experimenter counterparts (95% confidence interval [CI]: 0.057, 0.53) (Table 4). When weight status was used as an outcome, inactive non-using males were associated with 72% higher odds of overweight/obesity relative to their active experimenter counterparts (OR = 1.72, 95% CI: 1.2, 2.4). No significant associations were identified in females.
Table 4. Adjusted estimates from mixed-effects modelsFootnote a of Table 4 that regressed BMI onto lagged latent classes, by gender, Grade 9–12 secondary school students, COMPASS (Ontario, Canada)
|Latent class in previous wave||Regression coefficients (95% CI)Footnote b of Table 4||OR (95% CI)Footnote c of Table 4|
|Model||Model 1||Model 2|
|Active experimenter (Ref.)||–||–|
|Inactive non-user||-0.0087 (-0.20, 0.19)||0.85 (0.55, 1.32)|
|Model||Model 3||Model 4|
|Active experimenter (Ref.)||–||–|
|Inactive non-user||0.29Footnote * of Table 4 (0.057, 0.53)||1.72Footnote ** of Table 4 (1.2, 2.4)|
Building on earlier research,Footnote 17 we conducted a longitudinal analysis assessing for gender differences in the association of CDRB with BMI at follow-up. Our assessment shows that, at follow-up, BMI was higher by 0.29 in inactive non-using males than among active experimenters, with no such significant association in females. In addition, inactive non-using males were associated with 72% higher odds of overweight/obesity relative to their more active counterparts who experiment with substances.
Our findings emphasize the importance of stratified analyses that assess the association between CDRB and longitudinal BMI because the results of cross-sectional analyses in our earlier researchFootnote 17 are not likely to be consistent with findings from longitudinal analyses.
Substance use is associated with higher prevalence and incidence of obesity.Footnote 12Footnote 14Footnote 19Footnote 32Our findings can be partially explained by the difference in physical activity across the two classes. Physical activity was found to be protective against obesity in male but not female youth in the United States.Footnote 33
Physical activity is not the sole behaviour that contributes to differences in overweight/obesity. Research shows that unhealthy CDRB collectively contribute to higher BMI in youth. Low physical activity does not occur in isolation; it is usually associated with poor dietary intake and sedentary behaviour, and how research is conducted should reflect that.Footnote 34Footnote 35Footnote 36Footnote 37
The problem behaviour theory (1977) explains that youth who engage in one problem behaviour are at a higher risk of other problem behaviours due to the shared meanings and the social influences surrounding the behaviours.Footnote 38 Research indicates that behaviour change (positive or negative) is more effective when the behaviours are addressed simultaneously rather than each in isolation.Footnote 39 Furthermore, peer effects are reportedly associated with differences in diet, exercise and BMI.Footnote 37Footnote 40 Recent findings from Europe also suggest that active experimenters likely have experimenter friends who tend to be physically active in their free time. It is reported that athletes’ perceived social norms and increased exposure to alcohol through alcohol advertising during sporting events play a prominent role in their higher drinking habits relative to their peers.Footnote 41Footnote 42 Similarly, inactive non-using youth likely have non-using friends with sedentary pastimes (e.g. TV viewing, video games).Footnote 37
Consistent with earlier research,Footnote 5Footnote 6 our analyses indicate that there are annual increases in BMI across genders but that the predictors of the annual increase differ. Physical activity and substance use are not likely predictors of increasing BMI in females. Studies show that adult women have healthier dietary patterns than adult men.Footnote 43 Nevertheless, a study of adults in Scotland, England and Northern Ireland reported that variance in BMI was explained by physical activity (by 10.3%) and dieting behaviours (by 10.3%), while healthy eating explained only 1.6% of the variance.Footnote 43
We found that about one-quarter of youth transition to a “healthier” weight: 28.1% of youth with overweight/obesity in Wave 1 transitioned to a normal weight in Wave 2 and 27.3% of youth with overweight/obesity in Wave 2 transitioned to a normal weight in Wave 3. Similar findings were reported from Spain, where 26% of youth classified as obese transitioned to overweight status.Footnote 4 Over one-quarter of youth transitioning to a lower BMI category indicates the need for future research into attitudes, behaviours and peer and school effects in these youth. This will provide valuable lessons as to how youth successfully achieve a healthier weight.
BMI also tends to increase in adults, and there are fewer reported decreases than in youth. Over a span of 18 years, BMI in adults in the USA increased by 13% (equivalent to 3.1), with only 1.9% of women and 0.5% of men having a 1 unit decrease in BMI.Footnote 5 Our and others’ findingsFootnote 4 that a proportion of youth with overweight/obesity tend to transition to lower BMI categories emphasizes the importance of healthy weight loss and maintenance during adolescence.
From the point of obesity and chronic disease prevention, school-based interventions are warranted as they also associated with decreases in substance use.Footnote 44Footnote 45 Participatory approach programs, which are gaining popularity because of their success, encourage youth to participate and have succeeded at retaining studentsFootnote 46 as well as leading to decreases in BMI.Footnote 47 A meta-analysis shows that problem-solving training and techniques from cognitive behavioural therapy were beneficial, as were programs based on a social influences approach that teaches refusal skills.Footnote 48
Tailored prevention and intervention programs (such as gender-specific programs) are reportedly more effective than those intended for the general population of youth.Footnote 49 Our findings are important because they show that youth have gender-specific longitudinal predictors of BMI, warranting targeted gender-specific prevention and intervention efforts. We recommend gearing school-based interventions for inactivity and substance use that promote healthy food intake specifically towards male students and unhealthy dieting behaviours specifically towards female students. The scientific literature supports that addressing more than one health behaviour simultaneously is associated with more desired intervention-outcomes especially when a recommendation for increasing physical activity or limiting screen time predominates.Footnote 50Footnote 51
Strengths and limitations
Our study contributes to the discussion that CDRB are associated with youth health differently over time and between the genders. Our study adopted a novel approach by taking into consideration the dependence of students in schools in both the gender-specific multilevel latent class analysis and the gender-specific multilevel longitudinal regression analyses. Our findings suggest that increases in BMI at follow-up were significantly associated with the latent class, inactive non-using youth in males; there were no such associations in females. Similar studies that only adjusted for gender might have not found any association because they did not stratify by gender.Footnote 6 In addition, we accounted for monotone type missingness by using maximum likelihood models, based on the assumption that data are missing at random. These models are preferable over those that are based on complete case analysis because the latter assume the outcome is missing entirely at random.Footnote 31
In terms of limitations, the student questionnaire is entirely self-reported and therefore subject to social desirability bias. However, previous analyses have shown that there are no significant differences in the prevalence of BMI in COMPASS self-reports versus those measured by a trained professional across a national sample of youth in Canada.Footnote 17 COMPASS participants’ self-reported BMI have a high validity compared with measurements made by a trained professional (ICC = 0.84).Footnote 23
In addition, the use of passive consent likely mitigates social desirability bias. Active consent procedures are discouraged when measuring substance use to avoid limiting the participation of substance users who are most likely to benefit from these programs.Footnote 52
Only two latent classes were used in the analysis: active experimenters and inactive non-using youth. This does not mean that other youth were left out of the analyses; rather, it indicates that some youth may have had a better fit with another latent class, for example, inactive substance user.Footnote 17 To conduct our longitudinal analysis, the classes had to be fixed in this manner across the 3 years so that comparisons could be made over time.
Lastly, COMPASS is not generalizable to youth across Canada since it uses purposeful sampling. However, the prevalence of substance use and of BMI was comparable to those found in a nationally representative sample.Footnote 53
Although previous cross-sectional analyses show that youth in latent classes with substance use are associated with higher BMI and higher odds of overweight/obesity,Footnote 17 our longitudinal findings indicate that inactive male youth who do not use substances are at 72% higher odds of overweight/obesity than their active peers who experiment with substance use. This indicates that physical activity plays a longitudinal role in male youth BMI. No longitudinal predictors of increase in BMI were identified in female youth when considering latent classes of physical activity and current substance use.
The authors thank Trevor Bain, Kate Battista and Chad Bredin for their assistance with accessing the data and SAS software.
The COMPASS study has been supported by a bridge grant from the Canadian Institutes of Health Research (CIHR) Institute of Nutrition, Metabolism and Diabetes (INMD) through the “Obesity – Interventions to Prevent or Treat” priority funding awards (OOP-110788; awarded to STL), an operating grant from the CIHR Institute of Population and Public Health (IPPH) (MOP-114875; awarded to STL), a CIHR project grant (PJT-148562; awarded to STL), a CIHR bridge grant (PJT-149092; awarded to KP/STL), a CIHR project grant (PJT-159693; awarded to KP) and by a research funding arrangement with Health Canada (#1617-HQ-000012; contract awarded to STL).
Conflicts of interest
The authors have no conflicts to report.
Authors’ contributions and statement
NH conceived this study and its design, conducted the analysis and drafted the manuscript as part of her PhD dissertation at the University of Waterloo.
AC supervised NH in the drafting of the manuscript, co-designed the analyses, interpreted the results and edited the manuscript’s content.
STL conceived the COMPASS study and wrote the funding proposal, developed its tools and is leading its implementation and coordination.
AC, STL and PB provided ideas and thoughts for discussion.
All authors supported NH in the development of the study design, revision of earlier manuscript drafts and approved the final manuscript.
The content and views expressed in this article are those of the authors and do not necessarily reflect those of the Government of Canada.
Report a problem or mistake on this page
- Date modified: