Partial Least Squares Regression Discriminated High and Low Total Weight Loss Groups in Overweight and Obese Adults Consuming an Energy Restricted Diet.

Abstract

Multivariate analysis can reduce data to lower dimensional graphics that display group clusters not observed with other techniques. We used PCA and PLS to interpret data from a controlled feeding trial with an energy restricted diet in 71 overweight and obese adults. Clinical, endocrine, and inflammatory markers, body composition, and energy intake and energy expenditure measurements were included in the analyses. Analyses were conducted in R (ver 2.11.1). PCA found two clusters of participants by sex when post‐intervention data were plotted on the first two components. Subsequent PLS analysis of the sex adjusted data discriminated participants grouped by high and low weight loss along Latent Variable 1 (LV1) and LV2. Loadings related to low weight loss included total fat mass, blood lipids and sedentary activity. Loadings related to higher weight loss included physical activity. PLS projections indicated that subjects who had greater weight loss engaged in more physical activity. Since subjects maintained normal activities, our results suggest innate person‐to‐person differences in sedentary behavior and physical activity have a large impact on weight loss even with a highly‐controlled weight loss paradigm.Funding: National Dairy Council; USDA, ARS, WHNRC; Dairy Council of California; CTSC, University of California (1M01RR19975–01), National Center for Medical Research (UL1 RR024146)