Development of machine learning prediction models to explore nutrients predictive of cardiovascular disease using Canadian linked population-based data.

Machine learning may improve use of observational data to understand the nutritional epidemiology of cardiovascular disease (CVD) through better modelling of non-linearity, non-additivity, and dietary complexity. Our objective was to develop machine learning prediction models for exploring how nutrients are related to CVD risk and to evaluate their predictive performance. We established a population-based cohort from the Canadian Community Health Survey and measured CVD incidence and mortality from 2004 to 2018 using administrative databases of national hospital discharges and deaths. Predictors included 61 nutrition variables and fourteen socioeconomic, demographic, psychological, and behavioural variables. Conditional inference forest models were interpreted and evaluated by permutation feature importance, accumulated local effects, and predictive discrimination and calibration. A total of 12 130 individuals were included in the study. Use of supplements, caffeine, and alcohol were the most important nutrition variables for prediction of CVD. Supplement use was associated with decreased risk, caffeine was associated with increasing risk, and alcohol had a u-shaped association with risk. The model had an out-of-sample c-statistic of 0.821 (95% confidence interval = 0.801-0.842). Exploratory findings included both known and novel associations and predictive performance was competitive, suggesting that further application of machine learning to nutritional epidemiology may help elucidate risks and improve predictive models. Novelty: Machine learning prediction models were developed for CVD using dietary data. Models were interpreted with interpretable machine learning techniques, revealing diverse associations between diet and CVD. Models achieved comparable or superior predictive performance to existing CVD risk prediction models.

Validation of an automated self-administered 24-hour dietary recall web application against urinary recovery biomarkers in a sample of French-speaking adults of the province of Québec, Canada.

The objective of this study was to validate an automated self-administered 24-hour dietary recall web application (R24W) against recovery biomarkers for sodium, potassium and protein intakes and to identify individual characteristics associated with misreporting in a sample of 61 men and 69 women aged 20-65 years from Québec City, Canada. Each participant completed 3 dietary recalls using the R24W, provided two 24-hour urinary samples and completed questionnaires to document psychosocial factors. Mean reported intakes were 2.2%, 2.1% and 5.0% lower than the urinary reference values, respectively, for sodium, potassium and proteins (significant difference for proteins only (p = 0.04)). Deattenuated correlations between the self-reported intake and biomarkers were significant for sodium (r = 0.48), potassium (r = 0.56) and proteins (r = 0.68). Cross-classification showed that 39.7% (sodium), 42.9% (potassium) and 42.1% (proteins) of participants were ranked into the same quartile with both methods and only 4.8% (sodium), 3.2% (potassium) and 0.8% (proteins) were ranked in opposite quartiles. Lower body esteem related to appearance was associated with sodium underreporting in women (r = 0.33, p = 0.006). No other individual factor was found to be associated with misreporting. These results suggest that the R24W has a good validity for the assessment of sodium, potassium and protein intakes in a sample of French-speaking adults. Novelty: The validity of an automated self-administered 24-hour dietary recall web application named the R24W was tested using urinary biomarkers. According to 7 criteria, the R24W was found to have a good validity to assess self-reported intakes of sodium, potassium and proteins.

Protein consumption in Canadian habitual diets: usual intake, inadequacy, and the contribution of animal- and plant-based foods to nutrient intakes.

The 2019 Canada's Food Guide (CFG) emphasizes consumption of plant protein with implications for protein adequacy and nutrient intakes, yet a baseline with which to compare future dietary trends that may result from its adoption is not available. The objectives were to assess usual protein intake, inadequacy, and the contribution of animal- and plant-based foods to intake of protein, nutrients, and energy in Canada. Twenty-four-hour dietary recalls from the 2015 Canadian Community Health Survey - Nutrition were used to assess dietary intake among adults (n = 13 616). The National Cancer Institute method was used to estimate usual protein intake and inadequacy. Population ratios were used to determine the contribution of animal- and plant-based foods to intake of protein, nutrients, and energy. Usual protein intake averaged 79.47 ± 0.70 g/d; inadequacy was highest for females ≥71 y (9.76 ± 2.04%). Top protein contributors were red and processed meat (21.6 ± 0.55%), poultry and eggs (20.1 ± 0.81%), cereals, grains, and breads (19.5 ± 0.31%), and dairy (16.7 ± 0.38%). Dairy contributed most to calcium (53.4 ± 0.61%), vitamin D (38.7 ± 1.01%), but also saturated fat (40.6 ± 0.69%), whereas cereals, grains, and breads contributed most to iron (46.5 ± 0.57%) and vegetables and fruit to potassium (32 ± 0.45%). Given that animal sources contributed overwhelmingly to protein intake in 2015, dietary shifts towards plant protein needed to meet the 2019 CFG recommendations may pose a challenge, particularly for populations most at risk of inadequacy. Novelty: Older adults and females are most at risk of not meeting protein recommendations. Animal sources contribute two-thirds of the protein consumed by Canadian adults.