Lifestyle characteristics of a cohort of 1904 Germans adhering mainly to a vegetarian diet were examined in relation to their mortality after 11 years of follow-up. Poisson regression modelling was performed to consider the simultaneous effects of different variables on mortality from all causes, cancer (ICD 140-208) and cardiovascular diseases (ICD 390-459). Compared to a low level of self-reported physical activity, those with a medium or high level of activity experienced only half the mortality from all causes and from cardiovascular diseases. Physical activity showed no beneficial effect for cancer mortality in this cohort. The body mass index (BMI) was an independent risk factor for mortality among men but essentially unrelated to mortality among women. Those in the middle third of the BMI distribution experienced the lowest mortality. A negative association between BMI and cancer mortality lost statistical significance when the first 5 years of follow-up were deleted, suggesting that a lower BMI was a consequence of prevalent disease. Both the duration of vegetarianism and the vegetarian status (strict versus moderate) showed a moderate effect on all cause and cancer mortality. A longer duration of vegetarianism (> or = 20 years) was associated with a lower risk, pointing to a real protective effect of this lifestyle. A lower risk of death among moderate vegetarians suggests that sound nutritional planning may be more important than absolute avoidance of meat.