Table 4. Unadjusted and Adjusted Relative Hazards of Incident Diabetes During Follow-up From 1986 to 2016 Among Lactation Duration Categories.

Multivariate Models	Time-Dependent Lactation Duration Categories, Adjusted Relative Hazard (95% CI)				P Value for Trenda
	None (n = 322)	>0 to 6 mo (n = 418)	>6 to <12 mo (n = 268)	≥12 mo (n = 230)
Model 1 Unadjusted	1 [Reference]	0.60 (0.43-0.83)	0.36 (0.23-0.57)	0.29 (0.17-0.49)	<.001
Model 2 (model 1 plus adjustment for race time-dependent parity and GD status during follow-up)	1 [Reference]	0.74 (0.53-1.04)	0.55 (0.34-0.87)	0.45 (0.26-0.80)	.003
Model 3 (model 2 plus age, baseline covariates [body mass index,b waist circumference, and fasting glucosec], and familiy history of diabetes)	1 [Reference]	0.84 (0.60-1.20)	0.56 (0.35-0.91)	0.50 (0.27-0.90)	.006
Model 4 (model 3 plus time-dependent physical activity scored and Diet Quality scoree during follow-up)	1 [Reference]	0.81 (0.56-1.19)	0.53 (0.31-0.88)	0.53 (0.29-0.98)	.01
Model 5 (model 4 plus time-dependent weight changef during follow-up)	1 [Reference]	0.75 (0.51-1.09)	0.52 (0.31-0.87)	0.53 (0.29-0.98)	.01

Abbreviation: GD, gestational diabetes.

^aNo evidence of effect modification by race, GD status, or parity groups.

^bCalculated as weight in kilograms divided by height in meters squared.

^cFasting serum glucose at baseline (year 0); there was missing data for 5 participants, and 10 participants had fasting glucose from year 7 before diabetes onset (all values at year 7 <100 mg/dL).

^dQuartiles for time-dependent physical activity score for race-specific quartiles were evaluated as the average of scores from years 0, 7, and 20 (19 women were missing follow-up physical activity score).

^eQuartiles for the a priori Diet Quality score were evaluated as the average of scores from years 0, 7, and 20 (10 women were missing average Diet Quality score).

^fTime-dependent weight change equals weight change above the median vs at or below from baseline through the end of follow-up.