The relationships of rate of weight gain and weight fluctuation to inc
idence of non-insulin-dependent diabetes mellitus (NIDDM) were examine
d in Pima Indians. The 1,458 subjects were participants in a prospecti
ve study with examinations approximately every 2 years, Rate of weight
gain was defined as the slope of the regression line of weight with t
ime for two or more consecutive examinations greater than or equal to
2 years apart and weight fluctuation as the root-mean-square departure
from this line for four examinations, Among men, incidence of NIDDM w
as strongly and significantly related to rate of weight gain (e.g., ag
e-adjusted incidence = 56.7/1,000 person-years in those with weight ga
in greater than or equal to 3 kg/year and 16.9/1,000 person-years for
those losing weight [P-trend < 0.01]). In women, weight gain was signi
ficantly related to diabetes incidence only in those who were not init
ially overweight (body mass index < 27.3 kg/m(2)). In contrast to the
relationship with weight gain, weight fluctuation was not associated w
ith incidence of diabetes in either sex. These findings suggest that w
eight control in overweight individuals may be a more effective strate
gy for prevention of NIDDM in men than in women, whereas prevention of
obesity may prevent diabetes in both sexes. Concern about a diabetoge
nic effect of weight fluctuation should not deter weight control effor
ts.