There has been a focus on the role of dietary patterns in obesity aetiology in recent years(Reference Hu1,Reference Ocke2) . The Mediterranean dietary pattern, with its focus on whole fruits, vegetables and olive oil, may be protective against long-term weight gain(Reference Beunza, Toledo and Hu3–Reference Mendez, Popkin and Jakszyn6) and incident obesity(Reference Mendez, Popkin and Jakszyn6). The Dietary Approaches to Stop Hypertension (DASH) dietary pattern, high in Ca and K-rich foods such as fruits and vegetables, has been associated with significantly lower weight gain in adolescents(Reference Berz, Singer and Guo7,Reference Paula Bricarello, Poltronieri and Fernandes8) and adults(Reference Soltani, Shirani and Chitsazi9). The 2015 Dietary Guidelines for Americans Adherence Index (DGAI), which measures adherence to the 12 USDA Food Patterns for chronic disease prevention described in the 2015–2020 Dietary Guidelines for Americans (DGA)(Reference Jessri, Lou and L’Abbe10,11) , has been cross-sectionally associated with a lower likelihood of obesity among Canadian adults(Reference Jessri, Lou and L’Abbe10,Reference Jessri, Wolfinger and Lou12) and in the SU.VI.MAX French cohort(Reference Lassale, Fezeu and Andreeva4).
Many of the studies applying comparisons to multiple dietary patterns using rigorous, evidence-based diet quality indexes are based in the USA and utilise prospective cohort data. There is a lack of research that exists on the role of healthy dietary patterns and the likelihood of being affected by obesity utilising nationally representative health and nutrition survey data, particularly in the Canadian context, and specifically for promising dietary patterns for chronic disease prevention such as DASH and the Mediterranean diet(Reference Jessri, Lou and L’Abbe10,Reference Jessri, Wolfinger and Lou12,Reference Jessri, Ng and L’Abbe13) . This is a crucial knowledge gap, as an estimated one in four Canadian adults are currently living with obesity and an estimated one in thirty are living with multiple chronic diseases(14).
The objectives of this study were to apply three diet quality indexes – the 2015 DGAI(Reference Jessri, Lou and L’Abbe10), the Mediterranean-Style Dietary Pattern Score (MSDPS)(Reference Rumawas, Dwyer and McKeown15) and a DASH index(Reference Matsunaga, Hurwitz and Li16) to two cycles of nationally representative Canadian health and nutrition survey data in order to examine the relationship between adherence to healthy dietary patterns and concurrent likelihood of being affected by obesity.
Studies utilising a nationally representative Canadian population can offer international audiences insight into how various dietary patterns perform in varied settings and offer a distinct population for global researchers to compare study findings across public health benchmarks. Furthermore, this study provides key methodological considerations and insights into the application of diet quality indexes in the context of national health and nutrition survey data, including application of the National Cancer Institute (NCI) method for estimating usual intakes using repeat 24-h dietary recall data and applying validated correction factors to partially account for systemic biases introduced in self-reported BMI values.
Subjects and methods
Study population and data collection
Data for this study were collected under the authority of the Statistics Act of Canada. Analyses were conducted at the Statistics Canada’s Research Data Center in Toronto, Canada. All data were de-identified and accessible through Statistics Canada’s Research Data Centers.
Cross-sectional data were taken from 2004 and 2015 collections of the Canadian Community Health Survey (CCHS)-Nutrition, which provided the most-recent, nationally representative data available for Canadians’ food and beverage intakes(17). The sample frame included Canadians ≥ 0 year for CCHS 2004 and ≥ 1 year for CCHS 2015; members of the Canadian Forces and those living in the territories, on reserves, in prisons or in long-term care facilities were excluded from both cycles. In CCHS 2004, 35 107 respondents were sampled and weighted by Statistics Canada to represent 98 % of the Canadian population with a survey response rate of 76 %; in CCHS 2015, 20 487 respondents were sampled with a response rate of 62 %.
In both cycles, one respondent per household was selected to complete a general health questionnaire; anthropometric measurements were also taken with respondent consent. One 24-h recall was completed with the assistance of a trained Statistics Canada interviewer using a computerised, Canadian modification of the USDA’s Automated Multiple Pass Method(18); recall interviews were conducted from January 2015 to December 2015 during all 7 d of the week. Approximately 35 % of respondents completed a second recall by phone 7–10 d later. The nutrient and energy composition of consumed foods were obtained from the Canadian Nutrient File, Supplement 2001b(19) and 2015(20) for CCHS 2004 and CCHS 2015, respectively.
Additional details on CCHS-Nutrition survey design and the dietary recall component can be accessed elsewhere(17,18) .
Outcome measures and exclusion criteria
The primary outcome of interest was BMI ≥ 30 kg/m2.
CCHS 2004 data were accessed, and all analyses completed in Summer 2017. Exclusion criteria included: respondents < 18 years, pregnant or breast-feeding women, underweight individuals (BMI < 18·5 kg/m2) and respondents with missing values for smoking, physical activity (PA), measured or self-reported height and measured weight. The analytical sample size for CCHS 2004 was 11 748 adults. Analyses of CCHS 2015 commenced in Fall 2017, after release and access of the data in June 2017. The same exclusion criteria were applied for a sample size of 12 110. The aforementioned sample sizes were used for all analyses of the likelihood of being affected by obesity across survey cycles.
For analyses presented in Tables 1–3 only, further exclusions included respondents missing the following socio-demographic variables: highest household education (below secondary education, secondary education, trade school/college or university), immigrant status (yes/no) and marital status (single, married or widowed), for a smaller sample size of 12 049. These results were not compared with CCHS 2004 data.
Table 1 Analysis of socio-demographic and lifestyle characteristics across quintile categories of DASH, DGAI 2015 and MSDPS scores among Canadian adults from CCHS 2015 (n 12 049)*
DASH, Dietary Approaches to Stop Hypertension(Reference Matsunaga, Hurwitz and Li16); DGAI, Dietary Guidelines for Americans Adherence Index(Reference Jessri, Lou and L’Abbe10); MSDPS, Mediterranean-Style Dietary Pattern Score(Reference Rumawas, Dwyer and McKeown15); CCHS, Canadian Community Health Survey. Values are means or percentages ± standard errors (se). Estimates are weighted least squares means or percentages from a regression model adjusted for age and sex with bootstrapping to ensure accurate standard errors when using survey data. P trends were estimated with the use of DASH, DGAI and MSDPS in their continuous form and represent the P-value associated with the linear regression coefficient for continuous variables and the logistic regression coefficient for categorical variables. All P trends were < 0·0001 unless otherwise noted. Analyses were conducted on the first day of 24-dietary recall data only.
* Maximum possible scores for each diet quality index were: 90 out of 90 for DASH, 19 out of 19 for DGAI and 100 out of 100 for MSDPS.
† P trend = 0·0368.
‡ P trend = 0·0137.
§ P trend = 0·0018.
|| Current Canadian physical activity guidelines for adults states reaching a goal of 150 min of moderate/vigorous-intensity physical activity per week.
¶ P trend = 0·3394.
** P trend = 0·8003.
†† P trend = 0·1109.
‡‡ P trend = 0·5025.
Table 2 Mean daily intake of macro- and micronutrients across quintile categories of DASH, DGAI 2015 and MSDPS scores among Canadian adults from CCHS 2015 (n 12 049)
DASH, Dietary Approaches to Stop Hypertension; DGAI, Dietary Guidelines for Americans Adherence Index; MSDPS, Mediterranean-Style Dietary Pattern Score; CCHS, Canadian Community Health Survey; RAE, retinol activity equivalents. Values are means or percentages ± standard errors (se). Estimates are weighted least squares means or percentages from a regression model adjusted for age, sex and energy misreporting status (under-reporters, plausible reporters and over-reporters) with bootstrapping to ensure accurate standard errors when using survey data.
P trends were estimated with the use of DASH, DGAI and MSDPS in their continuous form and represent the P-value associated with the linear regression coefficient.
Analyses were conducted on the first day of 24-dietary recall data only.
All P-trends were < 0·0001 unless otherwise noted (†).
* Naturally occurring folate includes various forms of folate found naturally in food.
‡ Sum of quantities of naturally occurring folate in addition to folic acid without considering their differing bioavailability.
Table 3 Mean daily intake of select food groups and dietary components across quintile categories of DASH, DGAI 2015 and MSDPS scores among Canadian adults from CCHS-Nutrition 2015 (n 12 049)
DASH, Dietary Approaches to Stop Hypertension1; DGAI, Dietary Guidelines for Americans Adherence Index2; MSDPS, Mediterranean-Style Dietary Pattern Score3; CCHS, Canadian Community Health Survey. Values are means or percentages ± standard errors (se). Estimates are weighted least squares means or percentages from a regression model adjusted for age, sex and energy misreporting status (under-reporters, plausible reporters and over-reporters) with bootstrapping to ensure accurate standard errors when using survey data.
P trends were estimated with the use of DASH, DGAI and MSDPS in their continuous form and represent the P-value associated with the linear regression coefficient.
All P trends were < 0·0001 unless otherwise noted (†).
Analyses were conducted on the first day of 24-dietary recall data only.
Physical activity and energy misreporting
In CCHS 2004, derived variables were provided for daily energy expenditure of respondents ≥ 12 years, from which respondents could be categorised into PA levels. In CCHS 2015, this was not provided. To categorise PA levels in 2015, respondents’ average PA per day in minutes was computed and cut-offs were applied to define respondents as either sedentary, low active, active or very active(21). To preserve power and sample size, if individuals could not be categorised due to missing values for PA (n 48), levels were assigned according to Garriguet et al. (i.e. < 14 years was categorised as low active, ≥ 14 years was categorised as sedentary)(Reference Garriguet22).
To account for energy misreporting, estimated energy requirements (EER) for each respondent were calculated using Institute of Medicine equations, which require an individual’s age, sex, PA level, height and weight(21). To determine energy misreporting, respondents’ reported energy intakes from CCHS were compared with their calculated EER(Reference Garriguet22–Reference Jessri, Lou and L’Abbe24). An energy intake < 70 % of the EER indicated under-reporting, and an energy intake > 1·42 of EER indicated over-reporting; those in between were considered plausible reporters(Reference Garriguet22–Reference Jessri, Lou and L’Abbe24). In CCHS 2004 analyses, those whose EER could not be calculated were not included in analyses.
In CCHS 2015 analyses, if individuals self-reported their height and weight but did not consent to their height and weight being measured (n 2965), a correction factor was applied to estimate their BMI from their self-reported data, as determined by Statistics Canada(Reference Shields, Connor Gorber and Janssen25). This was done to preserve power and sample size. The USDA’s categorisation of individuals into energy levels in the 2015–2020 DGA was used to calculate EER for this subset of individuals, as the USDA specifications require only age, sex and PA level(11) (see Appendix 1).
Calculation of diet quality index scores
Three diet quality indexes were chosen for analysis in this study: the MSDPS measured adherence to a Mediterranean-type dietary pattern(Reference Rumawas, Dwyer and McKeown15), a DASH index developed by Matsunaga et al. measured adherence to a DASH-style dietary pattern(Reference Matsunaga, Hurwitz and Li16), and the 2015 DGAI measured adherence to a dietary pattern in-line with recommendations in the 2015–2020 DGA(Reference Jessri, Lou and L’Abbe10). Each index’s scoring criteria can be found in Appendices 2–4.
The Mediterranean-Style Dietary Pattern Score
The MSDPS developed by Rumawas et al. assesses a Mediterranean-type dietary pattern among Western populations(Reference Rumawas, Dwyer and McKeown15). Individuals were scored proportionally across thirteen dietary components commonly found in this dietary pattern; overconsumption received a penalty proportional to the amount consumed over the recommendation. Final scores were totalled out of 100 and weighted by a factor 0–1·0 representing the proportion of energy from Mediterranean diet-type foods. The range of possible scores was 0–100, with higher scores representing better adherence.
The Dietary Guidelines for Americans Adherence Index
The 2015 DGAI measures adherence to the twelve energy-based USDA Food Patterns found in the 2015–2020 DGA(Reference Jessri, Lou and L’Abbe10,11) . The 2015 DGAI is divided into a ‘food intake’ sub-score with eleven components and a ‘healthy choices’ sub-score with eight components. Scoring was assigned proportionally with a penalty for overconsumption. The trans-fat component was omitted in this study due to a lack of trans fat data in CCHS. The range of possible scores was 0–19, with higher scores representing better adherence.
Dietary Approaches to Stop Hypertension index(Reference Matsunaga, Hurwitz and Li16)
The DASH index by Matsunaga et al. assesses adherence to a DASH-style dietary pattern, with scoring criteria modelled after the Healthy Eating Index 2010 to maintain a robust scoring algorithm(Reference Matsunaga, Hurwitz and Li16). Scores were assigned proportionally across nine dietary components using an energy density approach. The range of possible scores was 0–90, with higher total scores denoting better adherence.
Statistical methods
All analyses were performed using SAS (version 9.4; SAS Institute Inc.). Analyses were age- and sex-adjusted and weighted using sample survey weights provided by Statistics Canada to ensure representative estimates. All standard errors were bootstrapped using the balanced repeated replication method with 500 replications to account for the complex sampling design used in CCHS(17).
To assess the association between DASH, DGAI 2015 and MSDPS scores with lifestyle characteristics, least squares means were estimated across quintiles of each diet quality index and various socio-demographic variables, adjusting for age and sex in the analysis. To assess the association between scores and quality of the respective dietary patterns, least squares means were estimated across quintiles and selected macro- and micronutrients (to assess nutrient quality of the dietary patterns) and selected dietary components (to assess overall diet quality), adjusting for age, sex and energy misreporting status. Misreporting status in CCHS has been shown to adjust for implausible recalls and selective misreporting of healthy v. unhealthy foods in examining the association between dietary intakes and the likelihood of obesity and indirectly adjusts for socio-economic characteristics correlated with misreporting status, including education and smoking status(Reference Jessri, Lou and L’Abbe24). Macro- and micronutrient intakes were reported either as a percentage of total energy or per 1000 kcal where appropriate. A P trend < 0·0001 was considered significant.
Pearson correlation coefficients between pairs of indexes were calculated to examine the similarity between indexes. To evaluate concordance or agreement between pairs of indexes, total scores from each index were divided into quintiles, and the proportion of the sample falling into quintile categories for pairs of indexes was examined. These analyses were performed using 1 d of dietary recall from the CCHS 2015 sample only.
For examining the association between dietary patterns and the likelihood of being affected by obesity, both days of dietary recall were used, from both CCHS 2004 and CCHS 2015.
After calculating respondents’ index scores, the NCI method was used to produce total ‘usual’ index scores that were adjusted for age, sex, weekend/weekday and sequence of dietary recall (first/second)(Reference Tooze, Midthune and Dodd26,27) . SAS macros for the NCI method are publicly available(27). To use the NCI-adjusted index scores in logistic regression models with obesity prevalence as the outcome (binary yes/no, where ‘yes’ denotes BMI ≥ 30 kg/m2), scores were first regression calibrated using the SAS macro INDIVINT(27,Reference Kipnis, Midthune and Buckman28) . After input into the regression model, the beta estimate and its standard error were combined with the output from the DISTRIB macro to manually calculate OR and 95 % CI. To avoid misclassification of respondents, OR and 95 % CI were computed at the median of each quintile from the distribution of continuous total scores(Reference Yang, Liu and Kuklina29).
In the logistic regression model with DGAI and obesity, covariates included age, sex (male/female), PA level (sedentary, low-active, moderately active and very active), smoking status (daily smokers, occasional smokers and non-smokers) and sequence of dietary recall analysed. In the regression model with DASH and obesity, alcohol was also adjusted for, as alcohol intake is not included in Matsunaga et al.’s DASH index(Reference Matsunaga, Hurwitz and Li16). Because MSDPS does not account for energy in its scoring criteria, the regression model with MSDPS and obesity was additionally adjusted for energy. To do this, energy intake was also estimated using the NCI method before entering it into the regression model to avoid the erroneous use of usual intake-estimated and single-day food intake variables in the same model. The NCI method univariate SAS macros were used for DGAI and DASH, while the NCI method bivariate SAS macros were used for MSDPS.
Results
In CCHS 2004, estimated mean MSDPS, DASH and DGAI scores were 12·43 ± 0·14 out of 100, 44·18 ± 0·29 out of 90 and 8·82 ± 0·05 out of 19; in CCHS 2015, scores were 13·9 ± 0·13, 44·99 ± 0·25 and 8·99 ± 0·04, respectively. These mean scores translated to a roughly 12–14 % adherence to the recommendations in a Mediterranean style-dietary pattern, 49–50 % adherence to the recommendations in a DASH diet and 46–47 % adherence to the recommendations in the 2015–2020 DGA for both samples of CCHS 2004 and 2015.
Socio-demographic and lifestyle variables in Canadian Community Health Survey-Nutrition 2015
Table 1 reports the socio-demographic and lifestyle characteristics for the sample of Canadian adults from CCHS 2015 (n 12 049) across quintile categories of DASH, DGAI 2015 and MSDPS scores using 1 d of recall. Compared with Q1, those in Q5 (i.e. higher index scores) were more likely to be older, non-smokers, married and with post-secondary education (P trend < 0·0001 for all indexes). Except for MSDPS, those with the best adherence to a healthy dietary pattern were also more likely to be female, immigrants and less likely to be affected by obesity (P trends < 0·0001 for DASH and DGAI 2015). Those who had the best adherence to DASH were also more likely to meet current Canadian PA guidelines (P trend < 0·0001); however, the average percentage of respondents meeting PA guidelines in Q5 of DASH was low at 49·5 ± 2·1 %.
Association of Dietary Approaches to Stop Hypertension, Dietary Guidelines for Americans Adherence Index 2015 and Mediterranean-Style Dietary Pattern Score with diet quality
Table 2 provides mean intake of specific macro- and micronutrients for the sample of Canadian adults from CCHS 2015 (n 12 049) for all indexes and using 1 d of recall. In general, compared with those in Q1, respondents in Q5 of DASH and DGAI 2015 (i.e. higher scores) were more likely to consume beneficial macro- and micronutrients (P trends < 0·0001). Those most adherent to DASH and DGAI 2015 were more likely to consume less fat, cholesterol and Na and were less likely to consume an energy-dense diet (P trend < 0·0001 for all). In general, there were fewer significant trends among macro- and micronutrient intake across quintile categories for MSDPS. Compared with Q1, respondents in Q5 consumed fewer carbohydrates and alcohol and greater intakes of fibre, protein, MUFA and Ca intake (P trend < 0·0001 for all). For all indexes, those most adherent to the respective dietary pattern were less likely to consume energy from added sugars (P trend < 0·0001 for all).
Focusing on the consumption of food groups (Table 3), in general, compared with Q1, respondents in Q5 for all three indexes were more likely to consume a variety of whole fruits and vegetables, more of their total grain intake as whole grains, more low-fat dairy and meat products, and overall had a greater total vegetable and plant protein intake (P trend < 0·0001 for all). There was no significant trend across quintile categories for red meat or dairy intake in the DGAI 2015, or for red meat, Na or sweets intake for the MSDPS (P trend > 0·0001 for all).
Performance of Dietary Approaches to Stop Hypertension, Dietary Guidelines for Americans Adherence Index 2015 and Mediterranean-Style Dietary Pattern Score
Appendices 5–8 show the correlations between the three indexes as well as concordance plots. Moderate correlations were observed between MSDPS and DASH, and MSDPS and DGAI (r = 0·44 for both), with the highest correlation between DASH and DGAI (r = 0·69). All correlations were significant (P < 0·0001).
Association between Dietary Approaches to Stop Hypertension, Dietary Guidelines for Americans Adherence Index 2015 and Mediterranean-Style Dietary Pattern Score with likelihood of being affected by obesity
Figures 1–3 illustrate the OR and 95 % CI for the likelihood of being affected by obesity at various percentiles of continuous index scores. Low scores for all indexes were associated with a greater likelihood of being affected by obesity in CCHS 2004 (dotted line; OR 10th v. 90th percentile and CI for DASH: 2·23 (95 % CI 1·50, 3·32), DGAI: 3·01 (95 % CI 1·98, 4·57) and MSDPS: 2·02 (95 % CI 1·14, 3·58)). Similar results were observed in CCHS 2015; however, results for MSDPS were not significant (solid line; OR 10th v. 90th percentile for MSDPS: 1·30 (95 % CI 0·82, 2·06)).
Fig. 1 OR and 95 % confidence intervals for the association between the Dietary Approaches to Stop Hypertension (DASH) scores and likelihood of obesity (BMI ≥ 30 kg/m2) for the CCHS 2004 sample (n 11 748) and CCHS 2015 sample (n 12 110) of Canadian adults. CCHS 2004: dashed line CCHS 2015: solid line. The logistic regression model was adjusted for age, sex (male/female), physical activity level (sedentary, low-active, moderately active and very active), smoking status (daily smokers, occasional smokers and non-smokers), sequence of dietary recall analysed and alcohol intake; DASH scores were entered as continuous and the 90th percentile was used as reference. Analyses were conducted on both days of 24-dietary recall data
Fig. 2 OR and 95 % confidence intervals for the association between Dietary Guidelines for Americans Adherence Index (DGAI) scores and likelihood of obesity (BMI ≥ 30 kg/m2) among a Canadian Community Health Survey (CCHS) 2004 sample (n 11 748) and CCHS 2015 sample (n 12 110) of Canadian adults. CCHS 2004: dashed line CCHS 2015: solid line. The logistic regression model was adjusted for age, sex (male/female), physical activity level (sedentary, low-active, moderately active and very active), smoking status (daily smokers, occasional smokers and non-smokers) and sequence of dietary recall analysed; DGAI scores were entered as continuous and the 90th percentile was used as reference. Analyses were conducted on both days of 24-dietary recall data
Fig. 3 OR and 95 % confidence intervals for the association between Mediterranean-Style Dietary Pattern Scores (MSDPS) and likelihood of obesity (BMI ≥ 30 kg/m2) among a Canadian Community Health Survey (CCHS) 2004 sample (n 11 748) and CCHS 2015 sample (n 12 110) of Canadian adults. CCHS 2004: dashed line CCHS 2015: solid line. The logistic regression model was adjusted for age, sex (male/female), (under-reporters, plausible reporters and over-reporters), physical activity level (sedentary, low-active, moderately active and very active), smoking status (daily smokers, occasional smokers and non-smokers), sequence of dietary recall analysed and energy intake; MSDPS scores were entered as continuous and the 90th percentile was used as reference. Analyses were conducted on both days of 24-dietary recall data
Discussion
This is the first study to utilise two cycles of nationally representative Canadian health and nutrition survey data to assess the association between dietary patterns for chronic disease prevention and likelihood of being affected by obesity. Our results suggested that both a DASH-like diet and a diet in-line with the 2015–2020 DGA were associated with better diet quality and lower likelihood of obesity among Canadian adults, with an over 2·4-times greater likelihood of being affected by obesity when comparing those with weak adherence to those with best adherence to these dietary patterns. In contrast, our results indicated that a Mediterranean-type dietary pattern was not associated with obesity among the Canadian adult population.
Findings for obesity examined with DASH and DGAI 2015 have been echoed in the scientific literature(Reference Lassale, Fezeu and Andreeva4,Reference Berz, Singer and Guo7–Reference Jessri, Lou and L’Abbe10) ; however, results on the association between Mediterranean-type diets and obesity are mixed(Reference Beunza, Toledo and Hu3–Reference Mendez, Popkin and Jakszyn6,Reference Woo, Cheung and Ho30) . While some studies have reported reduced risks for those who are overweight or obese in those who follow a Mediterranean diet(Reference Beunza, Toledo and Hu3–Reference Mendez, Popkin and Jakszyn6), other studies have reported insignificant or null associations in European and East-Asian cohorts(Reference Lassale, Fezeu and Andreeva4,Reference Woo, Cheung and Ho30) . When looking at studies which specifically use the MSDPS to define a Mediterranean-type diet, a positive inverse association was reported between higher MSDPS scores and waist circumference in a study by the developers of the MSDPS(Reference Rumawas, Dwyer and McKeown15); however, no significant association was found between body weight or risk of obesity when it was applied to the SU.VI.MAX cohort(Reference Lassale, Fezeu and Andreeva4). Insignificant findings for MSDPS and the likelihood of being affected by obesity in this study were not completely unexpected and suggest poor discriminatory ability of MSDPS to distinguish between those with ‘healthier’ and ‘poorer’ diets. Furthermore, the mean total MSDPS score for both samples in this study ranged from 12 to 14 out of 100, suggesting extremely low adherence to this dietary pattern in the Canadian population. This may point to the impracticality of studying associations between Mediterranean-type diets and health outcomes using CCHS data because very few Canadians follow such a pattern. Dietary patterns for chronic disease prevention and management designed and tested in Canadian populations include the Portfolio diet(Reference Chiavaroli, Nishi and Khan31,Reference Jenkins, Josse and Wong32) (plant-based and focusing on cholesterol-lowering foods for cardiometabolic risk reduction) and the Prairie diet(Reference Asaad, Soria-Contreras and Bell33) (diet pattern for diabetes management, designed to be in-line with Diabetes Canada’s clinical guidelines). Both the Portfolio and Prairie diets emphasise the use of healthy oils over animal fat; the Prairie diet also emphasises the consumption of dairy products. To study a Mediterranean-type diet in Canada, similar considerations for Canadian intakes (e.g. availability and preference for rapeseed oil instead of olive oil, and higher overall dairy intake) would have to be taken into consideration.
The methodology and reporting for this study were influenced by the Dietary Patterns Methods Project(Reference Liese, Krebs-Smith and Subar34). As was the case in Dietary Patterns Methods Project findings, results from our study suggested that there is no one approach to healthful eating for weight management. In our study, close adherence to all three dietary patterns was associated with a lower likelihood of obesity in both 2004 and 2015 cycles of CCHS data. Among the dietary patterns with the strongest association with obesity in both cycles (i.e. DASH and DGAI), those who ranked highly for adherence to DASH also ranked highly for adherence to DGAI (and vice versa), suggesting that a combination of beneficial dietary components, macronutrients and micronutrients may exist within different dietary patterns, while conferring possible benefits in weight management.
There are several strengths to this study. The use of two cycles of comprehensive, nationally representative data on Canadians’ intakes improved the generalisability and robustness of findings. Care was taken during methodological research design to reduce the effects of bias on results: the NCI method was used to estimate index scores in the context of ‘usual’ dietary intake, self-measured BMI was adjusted for using a validated correction factor to approximate measured BMI for those with missing measured height and weight data and all indexes were checked within the scientific literature for their validity, reliability and overall quality for application in this study. Despite these strengths, data from CCHS are cross-sectional and do not imply causal relationships. Residual confounding and reverse causation were also a possibility; however, the use of two cycles of CCHS data lowered the chances of this occurring. Additionally, BMI may not be the most precise marker of excess adiposity, especially for older adults who may lose height with age. However, as there are no additional markers of adiposity available in CCHS, our use of measured height and weight to examine obesity is justified.
While we were able to estimate index scores with the NCI method to capture some of the complexities due to long-term dietary intake, some within-person bias could not be accounted for in this study. This includes variation in individuals’ day-to-day intake and systematic choices in individuals’ preferences for combining foods (e.g. steak and potatoes). Additionally, the 24-h dietary recalls provided in CCHS 2015 were conducted by trained Statistics Canada interviewers and reviewed by dietitians; this may have introduced error into the collection of respondents’ dietary data.
In conclusion, the findings from this study of Canadian adults using comprehensive, nationally representative data suggested that, among Canadian adults, a DASH-like diet and a diet in-line with the 2015–2020 DGA may be associated with a lower likelihood of also being affected by obesity across two distinct time-points of data. These results aid in the understanding of diet and obesity in the Canadian context and highlight that multiple, distinct dietary patterns can be associated with weight management. Future prospective research is required to confirm these study findings.
Acknowledgements
Acknowledgements: The authors give thanks to the team at the Toronto Statistics Canada Research Data Centre and Dr. Wendy Lou at the Dalla Lana School of Public Health, Biostatistics Division, for their support and expertise; and Didier Garriguet at Statistics Canada and Dr. Kevin Dodd and the team at the National Cancer Institute for their advice. Authorship: A.P.N., M.J. and M.R.L. conceptualised and designed the study; A.P.N. and M.J. developed the coding for the analyses at the Research Data Center of Statistics Canada; A.P.N. conducted all statistical analyses at the Research Data Center of Statistics Canada; M.J. and M.R.L. reviewed the analyses and data tables; M.J. and M.R.L. provided scientific input into the statistical analyses and manuscript; A.P.N. drafted and revised the manuscript; all authors approved the final version to be published. Ethics of human subject participation: This study was a secondary analyses of CCHS-Nutrition – 2015 data. All information was de-identified and vetted by a trained member of Statistics Canada prior to release from the Toronto Research Data Center.
Financial support:
This research was supported by funds to the Canadian Research Data Centre Network (CRDCN) from the Social Sciences and Humanities Research Council (SSHRC), the Canadian Institute for Health Research (CIHR), the Canadian Foundation for Innovation (CFI) and Statistics Canada. MRL was supported by a research grant from the Burroughs Wellcome Fund Innovation in Regulatory Science Award. MJ was supported by the Canadian Institutes of Health Research (CIHR) Banting Fellowship, a CIHR Fellowship and Banting Discovery Award. APN was supported by the Canadian Institutes of Health Research (CIHR) Master’s Fellowship.
Conflict of interest:
There are no conflicts of interest.
Supplementary material
For supplementary material accompanying this paper visit https://doi.org/10.1017/S1368980023000903
Open access
