The use of cluster analysis to derive dietary patterns: methodological considerations, reproducibility, validity and the effect of energy mis-reporting

Abstract

Over the last three decades, dietary pattern analysis has come to the forefront of nutritional epidemiology, where the combined effects of total diet on health can be examined. Two analytical approaches are commonly used: a priori and a posteriori. Cluster analysis is a commonly used a posteriori approach, where dietary patterns are derived based on differences in mean dietary intake separating individuals into mutually exclusive, non-overlapping groups. This review examines the literature on dietary patterns derived by cluster analysis in adult population groups, focusing, in particular, on methodological considerations, reproducibility, validity and the effect of energy mis-reporting. There is a wealth of research suggesting that the human diet can be described in terms of a limited number of eating patterns in healthy population groups using cluster analysis, where studies have accounted for differences in sex, age, socio-economic status, geographical area and weight status. Furthermore, patterns have been used to explore relationships with health and chronic diseases and more recently with nutritional biomarkers, suggesting that these patterns are biologically meaningful. Overall, it is apparent that consistent trends emerge when using cluster analysis to derive dietary patterns; however, future studies should focus on the inconsistencies in methodology and the effect of energy mis-reporting.

Evaluation of methodologies for assessing the overall diet: dietary quality scores and dietary pattern analysis

“…The described approaches included hypothesis-driven scores of overall dietary quality, data-driven approaches such as principal component, exploratory factor and cluster analysis, and hybrid approaches such as reduced rank regression, partial leastsquares regression and decision tree analysis. Several reviews have been published that present comprehensive overviews of existing dietary quality scores, empirically derived dietary patterns, and their relationships with demographic characteristics, risk factors, biomarkers, health and disease (1)(2)(3)6,7,(10)(11)(12)20,21,23,24,26,27,38,58,63) . The present paper did not attempt to update these reviews, but focused particularly on methodological aspects.…”

“…Cluster analysis: K-means or Ward's method. The number of clusters to retain or report usually based on interpretability (after trying several options), in combination with the cluster variance ratio or scree plot, and cluster sample size (38) . Labelling of the dietary pattern Can be quantitatively (based on the highest factor loadings; or nutrient composition), or qualitatively (specific combinations of foods and nutrient composition).…”

Study of association between beverage consumption pattern and lipid profile in shift workers

“…Cluster analysis allows understanding of the variations in dietary patterns in the population, compared with more traditional methods of nutrition assessment such as mean nutrient or food group intake (16). Analysis based on dietary intake is more directly related to dietary recommendations, and makes it more feasible for further action to improve health outcomes (17).…”

What do we know about the nutritional status of the very old? Insights from three cohorts of advanced age from the UK and New Zealand

“…Instead of concentrating on one single nutrient or food group when testing diet-disease hypothesis (a single nutrient/food approach), recent developments in nutritional epidemiology advocate the importance of whole diet, DP and consequent interaction of foods and nutrients (a whole diet/DP approach) on health outcomes in general and older adults population (42)(43)(44)(45) . Healthy DP defined either by a greater adherence to predefined dietary scores (e.g.…”