Modeling prepregnancy body mass index as a continuous exposure for obstetric and neonatal outcomes: a population-based analysis of 3.4 million births.

BACKGROUND

Maternal body mass index (BMI) is associated with adverse pregnancy outcomes but is commonly categorized, imposing thresholds on a continuous exposure. Within-category risk variation across outcomes is uncertain.

OBJECTIVE

To characterize continuous associations between prepregnancy BMI and obstetric and neonatal outcomes, estimate absolute risks, and compare spline with categorical models.

STUDY

DESIGN

We analyzed 3,428,969 singleton live births at 20 to 42 weeks in the 2024 U.S. Natality Public Use File. BMI was modeled with restricted cubic splines using knots at 18.5, 25, 30, 35, and 40 kg/m² and reference BMI 22. Models adjusted for demographic, obstetric, and clinical covariates. Relative estimates were translated into incidence-calibrated absolute risks. Spline and 6-category models were trained in 80% and evaluated in 20% using c-statistics and Brier scores.

RESULTS

Associations were nonlinear and outcome specific. At BMI 40 vs 22, adjusted odds ratios were 2.36 for cesarean delivery, 3.86 for gestational diabetes, 3.37 for gestational hypertension, and 1.22 for preterm birth. Absolute risks increased from 24.8% to 43.7%, 4.6% to 15.8%, 6.2% to 18.2%, and 8.1% to 9.7%, respectively. Within class I obesity, risk increased by 5.0 percentage points for cesarean delivery, 3.6 for gestational diabetes, and 3.3 for gestational hypertension, compared with approximately 0.1 point across the BMI 30 boundary. Held-out c-statistic changes were 0.000 to 0.004 and Brier-score changes 0.0000 to -0.0002.

CONCLUSION

Continuous BMI modeling better describes outcome-specific gradients and within-category variation but minimally improves prediction metrics. Category boundaries should not be interpreted as biological transitions; application requires external validation.