Heterogeneous Effects of Labor Analgesia on Labor Duration and Perineal Trauma: A Causal Forest Analysis of a Real-World Obstetric Cohort
Abstract Background. Whether epidural analgesia prolongs labor and increases operative delivery remains debated, with randomized trials and real-world studies reaching conflicting conclusions. Both literatures estimate average effects over heterogeneous populations. We asked instead whose labor is prolonged by intrapartum interventions, and whether those women bear the downstream risks of perineal trauma and intrapartum cesarean. Methods. We analyzed a single-center prospective delivery registry of 2,111 singleton parturients who attempted vaginal birth (June 2025 onward). The primary exposure was maintained epidural analgesia versus Lamaze breathing-based analgesia (active comparator); epidural versus no analgesia served as sensitivity contrast. A secondary exposure was intrapartum uterotonic augmentation (oxytocin or artificial rupture of membranes) versus none. Adjusted average effects were estimated by cross-fitted double machine learning (continuous outcomes) and doubly robust estimation (binary outcomes). Heterogeneous effects (CATEs) on total labor duration were estimated with an honest causal forest using DML residualization, and validated by quartile calibration, calibration regression, and best linear projection; variable importance stability was assessed across five full-pipeline re-runs. Secondary obstetric outcomes were analyzed within CATE quartiles. Results. Epidural analgesia was associated with 125.0 minutes longer labor on average (SE 11.3, P < 0.001), but quartile-calibrated effects ranged from 58.2 (Q1) to 273.3 minutes (Q4; Q4–Q1 difference 215.1, P < 0.001). Maternal age, gestational age at delivery, and gestational age at prenatal registration were the dominant effect modifiers, ranking in the top three in all 15 stability re-runs. Women in the lowest CATE quartile had lower adjusted episiotomy rates than Lamaze controls (− 14.2 percentage points [pp]), whereas women in the highest quartile had higher rates (+ 13.1 pp; Q4–Q1 difference 27.3 pp, P = 0.010) and a trend toward more intrapartum cesarean (+ 7.0 pp, P = 0.101). An independent analysis of uterotonic augmentation reproduced the heterogeneity pattern (Q1 7.5 vs Q4 109.4 minutes, P < 0.001). Conclusions. The labor-prolonging association of intrapartum interventions is strongly heterogeneous and predictable from routine prenatal characteristics. The magnitude of prolongation itself stratifies downstream perineal-trauma risk. Reframing intrapartum intervention research around heterogeneity rather than averages may inform individualized labor management; prospective validation is required.
Authors
- Qiulian Xu (ORCID: https://orcid.org/0000-0002-0918-3566)
- Jiaping Lu
- Heng Xu
Publication Details
- Journal
- Research Square
- Published
- 2026-11-17
- DOI
- https://doi.org/10.21203/rs.3.rs-11320016/v1
- Primary Topic
- Anesthesia and Pain Management
- Type
- preprint