Correlates of weight patterns during middle age characterized by functional principal components analysis

UMMS Affiliation

Department of Quantitative Health Sciences

Publication Date


Document Type



Adult; Body Mass Index; Body Weight; Cohort Studies; Female; Humans; Male; Middle Aged; Obesity; Weight Gain; Weight Loss


Bioinformatics | Biostatistics | Epidemiology | Health Services Research


PURPOSE: Describing weight trajectories using functional methods may further our understanding of how weight impacts health. We characterize weight patterns and describe correlates of these patterns.

METHODS: Using a subset of the Framingham Heart Study original cohort limited-access data set (n=1,429), we conducted a functional principal components analysis (PCA) of body mass index from 40 to 55 years of age. Scores from the principal component functions defined weight patterns. Gender-specific logistic regression models provided estimates of association with sociodemographic and lifestyle factors.

RESULTS: Overall weight status, weight changes, and cycling emerged as weight patterns during middle age. Overweight/obesity at 25 years was the most consistent correlate of weight patterns (adjusted odds ratios [AORs] for men and women were 14.2 and 3.7 for overall overweight, 99.5 and 28.3 for overall obese, and 1.4 and 3.9 for cycling). Weight status at 25 years was not associated with weight gain in either gender or weight loss in men; for women the AOR was 2.7 for weight loss.

CONCLUSIONS: Functional PCA described weight patterns during middle age. The strong associations between weight status at 25 years and overall weight status and cycling during middle age underscore the importance of addressing weight earlier in life.

DOI of Published Version



Ann Epidemiol. 2010 Mar;20(3):201-9. Link to article on publisher's site

Journal/Book/Conference Title

Annals of epidemiology

PubMed ID


Related Resources

Link to Article in PubMed