Conflicting associations between dietary patterns and changes of anthropometric traits across subgroups of middle-aged women and men

Abstract

Background: Individuals respond differently to dietary intake leading to different associations between diet and traits. Most studies have investigated large cohorts without subgrouping them. Objective: The purpose was to identify non-uniform associations between diets and anthropometric traits that appeared to be in conflict with one another across subgroups. Design: We used a cohort comprising 43,790 women and men, the Danish Diet, Cancer and Health study, which includes a baseline examination at age 50e64 years and a follow-up about 5 years later. The baseline examination involved anthropometrics, body fat percentage, a food frequency questionnaire and information on lifestyle. From the questionnaire data we computed association rules between the intake of food groups and changes in waist circumference and body weight. Using association rule mining on subgroups and gender-specific cohorts, we identified non-uniform associations. The two gender-specific cohorts were stratified into subgroups using a non-linear, self-organizing map based method. Results: We found 22 and 7 cases of conflicting rules in 8 participant subgroups for different anthropometric traits in women and men, respectively. For example, in a subgroup of women moderate waist loss was associated with a dietary pattern characterized by low intake in both cabbages and wine, in conflict with the association trends of both dietary factors in the female cohort. The finding of more conflicting rules in women suggests that inter-individual differences in response to dietary intake are stronger in women than in men. Conclusions: This combined stratification and association discovery approach revealed epidemiological relationships between dietary factors and changes in anthropometric traits in subgroups that take food group interactions into account. Conflicting rules adds an additional layer of complexity that should be integrated into the study of these relationships, for example in relation to genotypes.

Dietary Intake of Adult Residents in Luxembourg Taking Part in Two Cross-Sectional Studies—ORISCAV-LUX (2007–2008) and ORISCAV-LUX 2 (2016–2017)

“…However, what remains to be more fully explored is why dietary patterns have shifted toward rather less healthy attributes in westernized countries, if not globally. In general, it is believed that factors such as the globalization of the economy and food production, widespread advertising of fast-food companies, and a lack of physical activity play important roles [66][67][68][69]. In addition, with globalization, staple foods have been shifting from local to industrial products, which entail, to a large extent, low-cost and highly processed foods and, consequently, result in a deterioration in healthy dietary patterns [70].…”

Linking Bisphenol S to Adverse Outcome Pathways Using a Combined Text Mining and Systems Biology Approach

“…Another type of data-driven approach is the use of TM and frequent item-set mining (FIM) to extract fragmented information in texts (e.g., abstracts or the full text of publications) (Jensen et al. 2012) or large datasets (clinical data, biobank, electronic patient records) (Jiang et al. 2019) and establish relevant associations.…”

Semen quality and waiting time to pregnancy explored using association mining

“…In this study, participants were not clustered using Self‐Organizing Maps (SOMs) due to the small study population. Instead, we used an updated version of Compass (2.0) 9 to pre‐process and perform association mining on the classical semen parameters, and female partner age (years) as independent variables to explain the dependant variable, TTP. Compass and the software used for association mining statistical analyses are implemented in the Python programming language (version 2.7).…”

“…The specifications of the Compass method suggest that variables with more than 90% of missing values should be removed, as well as dichotomous variables with more than 90% of the values being 0 (or 90% being 1) 8,9 . In addition, variables with a Pearson correlation r > .9 should be removed.…”

“…Association mining has previously been used to detect relations between variables in different kinds of datasets and research areas, such as market basket analysis, web usage mining, or bioinformatics 5‐7 . For example, Compass is a method for clinical and biobank data mining that employs Self‐Organizing Maps and Association Mining 8,9 . Here, we used the modified Apriori Association Mining algorithm to uncover statistically significant combinations of semen parameters from fertile men that influence the waiting times for couples to achieve pregnancy (time to pregnancy, TTP).…”