UC Davis

We conducted latent class analyses to identify women with homogeneous combinations of lifestyle and behavioral variables and tested whether latent classes were prospectively associated with diabetes incidence for women with or without baseline obesity. A total of 64,710 postmenopausal women aged 50-79 years without prevalent diabetes at baseline (years 1993-1998) were followed until 2018 with a mean follow-up of 14.6 years (sd = 6.4). Lifestyle variables included smoking, diet quality, physical activity, and sleep quality. Psychosocial variables included social support, depression, and optimism. Multivariable Cox proportional hazards regression models tested associations between latent classes and diabetes incidence controlling for age, race/ethnicity, and education. During follow-up, 8076 (12.4%) women developed diabetes. For women without baseline obesity, five latent classes were identified. Compared with a lower risk referent, diabetes incidence was higher in classes characterized by high probability of multiple lifestyle and psychosocial risks (HR = 1.45; 95% CI 1.28, 1.64), poor diet and exercise (HR = 1.23; 95% CI 1.13, 1.33), and psychosocial risks alone (HR = 1.20; 95% CI 1.12, 1.29). For women with baseline obesity, four latent classes were identified. Compared with a lower risk referent, diabetes incidence was higher for women with obesity in classes characterized by high probability of multiple lifestyle and psychosocial risks (HR = 1.48; 95% CI 1.32, 1.66), poor diet and exercise (HR = 1.32; 95% CI 1.19, 1.47), and intermediate probabilities of multiple risks (HR = 1.17; 95% CI 1.05, 1.30). Diabetes prevention efforts that focus on diet and exercise may benefit from attention to how lifestyle behaviors interact with psychosocial variables to increase diabetes risks, and conversely, how psychological or social resources may be leveraged with lifestyle changes to reduce the risk for women with and without obesity.