Objective:

We evaluated the influence of coffee consumption on diabetes incidence among the Hawaii component of the Multiethnic Cohort (MEC).

Design:

Prospective cohort.

Setting:

‍Population-based sample residing in Hawaii.

Subjects:

After exclusions, 75 140 men and women of Caucasian, Japanese American and Native Hawaiian ancestry aged 45–75 years were part of the current analysis. All participants provided information on diet and lifestyle through an FFQ. After 14 years of follow-up 8582 incident diabetes cases were identified using self-reports, medication questionnaires and health plan linkages. Hazard ratios (HR) and 95% confidence intervals were calculated using Cox regression while adjusting for known covariates.‍

Results:

The risk for diabetes associated with total coffee consumption differed by sex (P interaction,<0.0001). Women consuming >3 cups of any type of coffee daily had a significantly lower risk (HR = 0.66; 95% CI 0.58, 0.77; Ptrend, <0.0001) than those reporting ,1 cup/d, whereas the relationship in men was borderline (HR = 0.89; 95% CI 0.80, 0.99; Ptrend = 0.09). The same difference by sex was seen for regular coffee consumption, with HR of 0.65 (95% CI 0.54, 0.78; Ptrend,<0.0001) and 0.86 (95% CI 0.75, 0.98; Ptrend = 0.09) in men and women, respectively. No significant association with diabetes was apparent for decaffeinated coffee in women (HR = 0.85; 95% CI 0.72, 1.01; Ptrend = 0.73) or men (HR = 1.07; 95% CI 0.93, 1.23; Ptrend = 0.71). Despite small differences by ethnicity, the interaction terms between coffee intake and ethnicity were not significant.

Conclusions:

In this multiethnic population, regular, but not decaffeinated, coffee intake was much more protective against diabetes in women of all ethnic groups than in men.