The 425 women (ethnicity: white 67.2%, Asian 14.6%, other 18.2%) recruited at 1-year postpartum had mean age 35.5 ± 4.3 years, pre-pregnancy BMI 25.1 ± 5.0 kg/m2, with median 12.1 months since delivery (interquartile range 11.7–12.9) and median 10 months breastfeeding (interquartile range 5–12). 50.4% were primiparous and 63.7% had a family history of diabetes. Compared to those who did not complete the visits at 3- and 5-years postpartum, the 330 women who completed these visits were slightly older (mean age 35.7 ± 4.3 vs. 34.8 ± 4.3 years, p = 0.047), with no significant differences in ethnicity, family history of diabetes, parity, pre-pregnancy BMI, duration of breastfeeding, or time since delivery (data not shown).

The 330 study participants were stratified into the following three groups based on their weight change between pre-pregnancy and 5-years postpartum: (i) women with weight gain < 0% (i.e. weight loss; n = 100), (ii) those with weight gain 0–6% (n = 110), and (iii) women with weight gain ≥ 6% (n = 120). As shown in Table 1, there were no significant differences between the groups in age, ethnicity, family history of diabetes, parity, and prevalence of GDM in the recent pregnancy. Of note, pre-pregnancy BMI was higher in the weight loss group (mean 26.9 ± 5.6 kg/m2) than in the weight gain 0–6% (mean 24.2 ± 4.1 kg/m2) and weight gain ≥ 6% group (mean 24.7 ± 4.3 kg/m2) (overall p < 0.0001), and this comparative difference persisted at 3-months postpartum (p = 0.04).

At 1-year postpartum (Table 1), BMI continued to differ between the groups (overall p = 0.005), but now it was highest in both the weight loss and weight gain ≥ 6% groups (both mean BMI 26.3 kg/m2). Waist circumference was highest in the weight loss group (p = 0.01), with no differences between the groups in duration of breastfeeding and physical activity in the preceding year. Of note, there were no significant differences in cardiovascular risk factors at 1-year postpartum, including blood pressure, lipid profile, fasting glucose, 2-hour glucose, HOMA-IR, Matsuda index, CRP and adiponectin.

By 3-years postpartum (Table 2), the weight gain ≥ 6% group had the highest BMI (mean 26.9 kg/m2) (overall p = 0.005), coupled with the largest waist circumference (mean 89.8 cm) (overall p = 0.06). Again, there were no differences in physical activity in the preceding year. However, the weight gain ≥ 6% group now had lower HDL cholesterol (p = 0.03), the highest triglycerides (p = 0.03), highest fasting glucose (p = 0.03), poorest insulin sensitivity (Matsuda index: p = 0.004; HOMA-IR: p = 0.02), and lowest adiponectin (p = 0.04). These emergent differences in cardiovascular risk factors were further amplified at 5-years postpartum, at which time both BMI (mean 27.7 kg/m2) and waist circumference (mean 91.7 cm) were highest in the weight gain ≥ 6% group (overall p < 0.0001 and p = 0.0008, respectively). Indeed, compared to their peers, the women comprising this group now exhibited the highest total cholesterol (p = 0.04), apoB (p = 0.03), triglycerides (p = 0.003), fasting glucose (p = 0.03), 2-hour glucose on the OGTT (p = 0.04), and HOMA-IR (p = 0.0002), coupled with the lowest Matsuda index (p = 0.0004), HDL cholesterol (p = 0.009) and adiponectin (p = 0.01).

We next compared mean adjusted levels of cardiovascular risk factors in the three groups at each of 1-year, 3-years, and 5-years postpartum, respectively, after adjustment for age, ethnicity, family history of diabetes, parity, pre-pregnancy BMI, duration of breastfeeding and time since delivery (Table 3). These analyses revealed that, after covariate adjustment, there was stepwise worsening (from the weight loss group to weight gain 0–6% to weight gain ≥ 6%) of the following CV risk factors at 5-years: triglycerides (p = 0.001), HDL (p = 0.02), LDL (p = 0.01), apoB (p = 0.003), Matsuda index (p < 0.0001), HOMA-IR (p < 0.0001), fasting glucose (p = 0.07), and C-reactive protein (p = 0.01). Moreover, none of these differences were present at 1-year postpartum. Instead, these significant differences in mean adjusted cardiovascular risk factors progressively emerged over time at 3- and 5-years (Table 3). These findings were unchanged on sensitivity analyses adjusting for the area-under-the-glucose-curve on the OGTT in pregnancy (data not shown). Linear mixed models also revealed that significant differences in the averages of cardiovascular risk factors across the weight retention groups depended on the time since delivery, as follows: triglycerides (interaction effect between weight group and time, p = 0.04), HDL (p = 0.01), apoB (p = 0.002), Matsuda index (p < 0.0001) and HOMA-IR (p = 0.0003). On sensitivity analyses in which weight retention group was replaced with repeated measure BMI, the change in BMI showed significant effect on the average rate of change in triglycerides (interaction effect between BMI and time, p = 0.04) and HDL (p = 0.01).

Having identified that greater postpartum weight retention tracks with the early evolution of an adverse cardiovascular risk profile over the first 5-years after pregnancy, we next considered its implications for dysglycemia (i.e. pre-diabetes/diabetes). At 5-years postpartum, 77 women had dysglycemia, of which the vast majority was pre-diabetes (n = 65). On logistic regression analyses adjusted for age, ethnicity, family history of diabetes, and pre-pregnancy BMI, postpartum weight gain ≥ 6% emerged as an independent predictor of pre-diabetes/diabetes at 5-years (adjusted OR = 3.40, 95%CI: 1.63–7.09) (Fig. 1A). These findings were unchanged with further adjustment for duration of breastfeeding and average total physical activity over the preceding 5-years (adjusted OR = 2.96, 95%CI: 1.41–6.23) (Fig. 1B). Similarly, the findings were unchanged with further adjustment for parity (data not shown).