Introduction: An abundance of evidence has shown that leisure-time physical activity (PA) is associated with reduced risk of diabetes. Adults, however, engage in various forms of PA, with potentially differential health effects. We tested if specific modes of PA are differentially associated with diabetes prevalence in a population-based adult sample.

Methods: Among 16903 racially/ethnically diverse adults ≥20 years old participating in NHANES 2011-16, we fit separate log binomial models to test associations between mode of PA and diabetes, and whether associations differed by race/ethnicity. Participants completed none, some (1-149 min/week) or met national PA guidelines (150+ min/week) for leisure time, work, transportation and combined modes (’All’). Diabetes was based on laboratory-confirmed HbA1C ≥ 6.5% or ever being told by a health professional respondent had diabetes.

Results: For each mode of PA and All modes combined, respondents completing 150+ minutes of PA were 33% to 39% less likely to have diabetes than respondents who completed no PA, even after controlling for covariates. There was a general gradient of decreasing prevalence with increasing PA. Effects varied by race/ethnicity for all PA modes except work-based PA (P<.01).

Conclusions: Meeting PA guidelines via any mode appears to confer protection from diabetes, but there are differential effects by mode and race/ethnicity.


S. Echeverria: None. A.A. Divney: None. E.K. Tsui: None. C. Mirzayi: None.

Readers may use this article as long as the work is properly cited, the use is educational and not for profit, and the work is not altered. More information is available at