Food substitution models for nutritional epidemiology

Abstract

The advantage of using specified substitution analysis in nutritional epidemiology has been clearly demonstrated in studies of macronutrient intake and disease risk. However, the method has not been widely applied in studies of food intake. The aim of this article is to describe and compare the interpretation and application of different food substitution models in epidemiologic studies on diet and disease development. Both theoretically and in the context of a specific example, we discuss methodologic issues to be considered, including modeling of food substitutions using diet at a single time point or at multiple time points (focusing on dietary changes), choice of substitution unit, adjustment for total energy intake, and adjustment for confounding. We argue that specified food substitution analyses can be used to identify optimal food composition of the diet and that these analyses are thus highly relevant to inform public health policy decision makers.

Intake of dairy products and associations with major atherosclerotic cardiovascular diseases: a systematic review and meta-analysis of cohort studies

“…Thus, summarizing findings on substitutions between dairy products were possible from only two studies. The lack of specifying substitutions in most previous cohort studies has implications for the interpretation of the results of our systematic review and meta-analysis as the effect of a specific dairy product on CVD depends on the replaced foods 62 .…”

The association of minerals intake in three meals with cancer and all-cause mortality: the U.S. National Health and Nutrition Examination Survey, 2003–2014

“…Food substitutions model is used to study the relationship between food or nutrient substitution and related diseases or health outcomes, and to provide dietary recommendations for disease prevention and treatment [ 36 ]. At present, many studies have adopted substitution analysis to replace one food or nutrient with another food or nutrient under the condition of equal energy or equal intake, and then observe the changes of epidemiological indicators [ 37 , 38 ].…”

Machine Learning Prediction of Food Processing

“…Overall, individuals with a higher food processing score exhibit higher blood pressure, trunk fat and subscapular skinfold, measures of obesity, and blood insulin and triglyceride levels; and lower "good" HDL cholesterol. Further novel findings indicate a higher prevalence of type 2 diabetes (C-peptide), inflammation (C-Reactive Protein), vitamin deficiency (homocysteine, methylmalonic acid), and inflammatory arthritis [40][41][42][43][44][45]. We also find an inverse association between iF P ro and telomere length, which can be affected by diet through inflammation and oxidation [46], suggesting a higher biological age for individuals with higher reliance on more ultra-processed foods (SI Section 4).…”