Highlights
Researchers in the Netherlands developed and temporally validated two prediction models for obstetric anal sphincter injury (OASI) in nulliparous women: one for spontaneous vaginal delivery (SVD) and one for operative vaginal delivery (OVD).
Both models showed moderate discrimination but excellent calibration, meaning they were better at estimating absolute risk than at sharply separating who will and will not sustain injury.
Mediolateral episiotomy, estimated fetal birth weight, prolonged second stage of labor, occipitoposterior fetal position, epidural analgesia, Asian ethnicity, and gestational age were central predictors in both models; maternal age, induction of labor, and fetal sex were additionally retained in the SVD model.
The models may help target counseling and prevention strategies, but their modest discriminatory performance means they should support—not replace—clinical judgment.
Background
Obstetric anal sphincter injury is a serious complication of vaginal birth involving trauma to the anal sphincter complex. It is clinically important because it can lead to anal incontinence, chronic pain, dyspareunia, psychological distress, and reduced quality of life. The burden is especially relevant in nulliparous women, who face the highest baseline risk for severe perineal trauma during first vaginal delivery.
Prevention remains challenging. Known risk factors include operative vaginal delivery, fetal macrosomia, occiput posterior presentation, prolonged second stage of labor, and certain maternal characteristics, but the predictive value of these factors in routine practice has been limited. Many prior models lacked external validation, were derived from small or single-center samples, or had limited clinical transportability. A robust, population-based model could help identify women at particularly high risk and guide decisions around intrapartum management, episiotomy technique, and counseling.
Study Design
This was a nationwide, population-based cohort study using the Netherlands Perinatal Registry. The investigators included nulliparous women who delivered a singleton live-born infant in cephalic presentation at term between 2016 and 2020. Separate predictive models were developed for women undergoing spontaneous vaginal delivery and operative vaginal delivery.
Candidate predictors were selected a priori from the literature and clinical expertise, then extracted from registry data. The final models were built using logistic regression with backward selection guided by the Akaike Information Criterion (AIC). Performance was assessed with discrimination, reported as the area under the receiver operating characteristic curve (AUC), and calibration, reported with the Brier score. The models underwent internal and temporal external validation, which strengthens confidence that their estimates were not limited to a single time period.
Key Findings
Among 171,046 women with spontaneous vaginal delivery, the OASI rate was 4.1%. After model selection, ten predictors remained: mediolateral episiotomy (MLE), expected fetal birth weight, duration of the second stage, occipitoposterior presentation, induction of labor, epidural analgesia, Asian ethnicity, maternal age, gestational age, and fetal sex.
The final SVD model had moderate discrimination, with an AUC of 0.67 (95% CI 0.67–0.68). Its calibration was excellent, with a Brier score of 0.039. In practical terms, the model’s predicted risks aligned closely with observed event rates, but its ability to distinguish high-risk from low-risk patients was only fair. This is a common trade-off in obstetric prediction modeling, where the outcome is influenced by many overlapping clinical and procedural factors.
Among 37,547 women with operative vaginal delivery, the average OASI rate was 3.5%. Seven predictors were retained: MLE, expected fetal birth weight, duration of the second stage, occipitoposterior fetal presentation, epidural analgesia, Asian ethnicity, and gestational age. The OVD model also showed moderate discrimination, with an AUC of 0.68 (95% CI 0.67–0.70), and excellent calibration, with a Brier score of 0.032.
Several predictors are clinically intuitive. Larger estimated fetal birth weight and prolonged second stage are consistent with mechanical strain on the pelvic floor and perineum. Occipitoposterior presentation can increase the complexity of delivery and is associated with more difficult descent and rotation. Epidural analgesia may reflect longer labors or altered second-stage dynamics rather than a direct causal role. Asian ethnicity has been associated in some datasets with increased OASI risk, although this likely reflects a mixture of anatomic, obstetric, and care-related factors and should be interpreted cautiously to avoid overattribution to biological difference alone.
The inclusion of mediolateral episiotomy is particularly important. In many settings, appropriately angled MLE is considered protective against OASI during operative vaginal birth and selected high-risk deliveries. Its presence in the model suggests that episiotomy is not merely a procedural detail, but a strong modifier of injury risk that should be incorporated into risk estimation. However, because episiotomy may be preferentially used in more complex deliveries, the direction and magnitude of association in observational data should be interpreted in context.
The modest AUCs indicate that the models are not sufficiently precise to serve as standalone decision tools for all women in labor. Nonetheless, the excellent calibration is clinically meaningful. A well-calibrated model can help estimate individualized absolute risk, which is often more useful for counseling than ranking patients alone. For example, a woman with several risk factors may have a risk estimate that is materially higher than the average baseline risk, even if the model cannot perfectly classify every eventual case of OASI.
One practical strength of this work is that the investigators provide accessible nomograms, making the models more usable at the bedside or in electronic decision support. In high-volume maternity care, even a moderately accurate, calibrated model can help focus attention on women most likely to benefit from preventive measures or from enhanced intrapartum vigilance.
Expert Commentary
This study has several notable strengths. It is large, population-based, and based on a national registry, which improves generalizability within the Dutch maternity system. The inclusion of temporal validation is also important, as models often perform well in derivation samples but degrade over time. By separating SVD and OVD, the investigators acknowledged that the mechanism of OASI differs between these delivery modes and that a single combined model may obscure clinically relevant differences.
At the same time, the limitations are substantial enough that implementation should be cautious. First, the AUC values around 0.67–0.68 indicate only moderate discrimination. Second, registry-based studies depend on the accuracy and completeness of coded data; if predictors such as estimated fetal weight, second-stage duration, or episiotomy type are measured inconsistently, model performance may be affected. Third, the model does not establish causality. Predictors like epidural analgesia and induction of labor may serve as markers of labor complexity rather than direct causes of OASI.
There is also the question of clinical actionability. A prediction model has the most value when it changes care in a way that improves outcomes. For OASI, possible responses include targeted counseling, optimization of delivery technique, correct use of mediolateral episiotomy when indicated, careful perineal support, and heightened readiness for difficult operative delivery. However, whether using the model actually reduces injury rates remains unknown.
Another important issue is external validity beyond the Netherlands. Obstetric practice patterns, episiotomy technique, thresholds for operative vaginal delivery, and population composition vary across countries and hospitals. Before adoption elsewhere, the model should be independently validated in other health systems and ideally tested in prospective implementation studies.
Finally, risk prediction should complement—not replace—shared decision-making. Women deserve clear counseling about the potential benefits and harms of interventions such as operative vaginal birth and episiotomy. A validated model may improve that conversation by quantifying risk, but it cannot capture every relevant clinical nuance, including fetal station, operator experience, pelvic anatomy, and real-time labor progress.
Conclusion
van Bavel and colleagues have developed and temporally validated two population-based prediction models for OASI in nulliparous women, tailored to spontaneous and operative vaginal delivery. The models are well calibrated and incorporate clinically meaningful predictors, but their moderate discrimination limits standalone use. Their main value is likely as decision-support tools to refine counseling and target preventive strategies, especially around delivery technique and episiotomy use.
The next step is not simply wider deployment, but prospective testing: does using these models change obstetric behavior, improve prevention of OASI, and reduce long-term pelvic floor morbidity? Until then, the study offers an important foundation for individualized intrapartum risk assessment.
Funding and ClinicalTrials.gov
The abstract does not report funding details or a ClinicalTrials.gov registration number. As this was a registry-based observational cohort study, trial registration may not apply in the same way as for interventional studies.
References
1. van Bavel J, Ravelli ACJ, Abu-Hanna A, Roovers JWR, Mol BW, de Leeuw JW. Prediction of Obstetric Anal Sphincter Injury in Nulliparous Women: Model Development and Temporal Validation. BJOG. 2026; PMID: 41987414.
2. RCOG Green-top Guideline No. 29. The Management of Third- and Fourth-Degree Perineal Tears. Royal College of Obstetricians and Gynaecologists.
3. Sultan AH, Thakar R, Fenner DE. Perineal and anal sphincter trauma during childbirth and subsequent anal incontinence. Clin Colon Rectal Surg.
4. ACOG Practice Bulletin on Prevention and Management of Obstetric Lacerations at Vaginal Delivery. American College of Obstetricians and Gynecologists.
AI Image Prompt
Professional medical illustration of a maternity labor and delivery setting with a clinician reviewing a digital risk prediction nomogram for obstetric anal sphincter injury, subtle overlay of maternal pelvic anatomy and fetal delivery icons, clean modern hospital environment, informative editorial style, realistic lighting, high-resolution, scientific and clinical tone.
