4) Logistic regression analysis is appropriate, as BMI was entered into the equation as a categorical variable, in three categories, where the rate in the two extreme ones is compared to the middle one. 5) Our questionnaire referred to all types of habitual activity, including work time as well as at leisure time. The rate of persons involved in any type of sport was minimal. "High" physical activity was defined as activity score in the highest tertile of the study group. The score consisted of the weekly number of hours spent in moderate activity (heavy activity was not reported by any of the 502 individuals constituting the subsample interviewed for dietary intake and physical activity) + half of the hours spent in light activity defined according to the Committee on Dietary Allowances . Using this crude measure, and accounting for age, BMI and total caloric intake, the risk for Type 2 diabetes for "highly" active individuals of both sexes was 0.67 with 90% confidence limits 0.4%0.96 (p = 0.04) . This was validated by significantly lower insulin response in the "highly" active individuals in all sex and BMI categories (unpublished data).