All analyses were planned a priori when this study was designed. Cox regression models were used to estimate hazard ratios (HRs) with 95% confidence intervals (CIs) of incident hospital-diagnosed hypertension and CVD according to weight change from prepregnancy to 18 months postpartum. We saw similar associations for ischemic heart disease and stroke, and they were therefore treated as a composite endpoint (CVD) to increase power. Women were considered at risk from the date of second postpartum interview (approximately 18 months postpartum) until the time of diagnosis, emigration, death (n = 336 in the hypertension analyses, n = 348 in the CVD analyses), or end of follow-up (September 10, 2018), whichever came first. To investigate how GWG may modify the association between overall weight change and risk of hypertension and CVD, we estimated how permutations of GWG (below, within, and above) and overall weight change (<−1, ±1, >1 BMI units) in 9 categories were associated with both outcomes. Women who gained weight according to the GWG recommendation and had no overall weight change serve as reference. We further examined how the 5 different postpartum weight change patterns were associated with both outcomes. Women who had returned to their prepregnancy BMI at 18 months postpartum served as reference. Finally, we examined whether the associations were modified by prepregnancy BMI (<25 and ≥25 kg/m2).

The assumption of proportional hazards was examined graphically by log-minus-log plots for the main exposure, and no violation was observed. All analyses were adjusted for a priori selected covariates. We adjusted for BMI, parity, and alcohol intake before the index pregnancy, maternal age at conception, socio-occupational status, dietary intake, leisure-time exercise, diabetes, preeclampsia, preterm birth, smoking status during index pregnancy, and total duration of breastfeeding. To be able to evaluate the baseline risk, we estimated adjusted incidence rates using Poisson regression models for a reference woman (characteristics presented in Table 1).

DNBC, Danish National Birth Cohort; GWG, gestational weight gain; IOM, Institute of Medicine.

In a sensitivity analysis, we examined risk of incident self-reported hypertension, and findings were similar to those presented (S2 Table). Also, we did a sensitivity analysis adjusted for births (yes/no) during follow-up, and findings were similar to those presented. We further examined our main exposure continuously by restricted cubic splines with 4 knots (fifth, 35th, 65th, and 95th percentiles) and a reference value set to 0 in weight change [27]. The splines supported the findings from the categorical analyses and are presented in S1 Fig.

To address the problem of missing data in covariates, we used multiple imputation [28]. Variables with complete data (prepregnancy weight, height, age at conception, gestational age, and weight 18 months postpartum) were included in the imputation step as explanatory variables in addition to the variables included for imputation. Furthermore, the outcome variable for hypertension, ischemic heart disease, and stroke were included together with the Nelson–Aalen estimator, an approximation of the cumulative baseline hazard, as suggested by others [29]. For women still breastfeeding at the time of the interview, total breastfeeding duration was imputed using interval imputation with a lower limit set to the time of the interview and a universal upper limit set to 3 years. A total of 50 copies of the dataset were generated by chained equations. The imputation and subsequent analyses were conducted using standard mi procedures available in STATA/SE 15 (StataCorp, College Station, Texas, US). We also carried out complete case analyses and observed results of same direction and approximate magnitude as those presented (S3S5 Tables).

Finally, death may be a potential competing risk in the present study. Therefore, as suggested in the peer review, we did a sensitivity competing risk analysis of our main exposure using the Fine–Gray approach [30] with death as a competing risk (S6 Table). Results were similar to those observed for the complete case analysis using the Cox regression model.

Note: The content above has been extracted from a research article, so it may not display correctly.

Please log in to submit your questions online.
Your question will be posted on the Bio-101 website. We will send your questions to the authors of this protocol and Bio-protocol community members who are experienced with this method. you will be informed using the email address associated with your Bio-protocol account.

We use cookies on this site to enhance your user experience. By using our website, you are agreeing to allow the storage of cookies on your computer.