European Journal of Obstetrics & Gynecology and Reproductive Biology 188 (2015) 88–94
Contents lists available at ScienceDirect
European Journal of Obstetrics & Gynecology and Reproductive Biology journal homepage: www.elsevier.com/locate/ejogrb
Predicting the chance of vaginal delivery after one cesarean section: validation and elaboration of a published prediction model Marie C. Fagerberg a,b,*, Karel Marsˇa´l b, Karin Ka¨lle´n c a
Department of Obstetrics and Gynecology, Ystad Hospital, Ystad, Sweden Department of Obstetrics and Gynecology, Clinical Sciences Lund, Lund University, Lund, Sweden c Centre of Reproduction Epidemiology, Clinical Sciences Lund, Lund University, Lund, Sweden b
A R T I C L E I N F O
A B S T R A C T
Article history: Received 12 October 2014 Received in revised form 25 January 2015 Accepted 19 February 2015
Objective: We aimed to validate a widely used US prediction model for vaginal birth after cesarean (Grobman et al. [8]) and modify it to suit Swedish conditions. Study design: Women having experienced one cesarean section and at least one subsequent delivery (n = 49,472) in the Swedish Medical Birth Registry 1992–2011 were randomly divided into two data sets. In the development data set, variables associated with successful trial of labor were identified using multiple logistic regression. The predictive ability of the estimates previously published by Grobman et al., and of our modified and new estimates, respectively, was then evaluated using the validation data set. The accuracy of the models for prediction of vaginal birth after cesarean was measured by area under the receiver operating characteristics curve. Results: For maternal age, body mass index, prior vaginal delivery, and prior labor arrest, the odds ratio estimates for vaginal birth after cesarean were similar to those previously published. The prediction accuracy increased when information on indication for the previous cesarean section was added (from area under the receiver operating characteristics curve = 0.69–0.71), and increased further when maternal height and delivery unit cesarean section rates were included (area under the receiver operating characteristics curve = 0.74). The correlation between the individual predicted vaginal birth after cesarean probability and the observed trial of labor success rate was high in all the respective predicted probability decentiles. Conclusion: Customization of prediction models for vaginal birth after cesarean is of considerable value. Choosing relevant indicators for a Swedish setting made it possible to achieve excellent prediction accuracy for success in trial of labor after cesarean. During the delicate process of counseling about preferred delivery mode after one cesarean section, considering the results of our study may facilitate the choice between a trial of labor or an elective repeat cesarean section. ß 2015 Elsevier Ireland Ltd. All rights reserved.
Keywords: Elective repeat cesarean Prediction model Vaginal birth after cesarean
Introduction The increasing rate of women having undergone cesarean section (CS) stresses the necessity of well founded counseling on subsequent pregnancy delivery mode – trial of labor (TOL) or elective repeat cesarean section (ERCS). To minimize negative outcomes for mothers and babies, the main issue would be to avoid planned vaginal deliveries ending in emergency CS [1–6]. Thus, a custom-built model for prediction of vaginal birth after cesarean (VBAC) would prove useful in pregnancies following one CS.
* Corresponding author at: Department of Obstetrics and Gynecology, Clinical Sciences Lund, Lund University, Lund, Sweden. Tel.: +46 701 70 87 51. E-mail address: [email protected] (M.C. Fagerberg). http://dx.doi.org/10.1016/j.ejogrb.2015.02.031 0301-2115/ß 2015 Elsevier Ireland Ltd. All rights reserved.
Several models aiming to predict VBAC have been published [7–17]. Some models have been externally validated or supplemented with new variables in subsequent publications [2,18–21]. To our knowledge, the most utilized model originates from the Grobman et al.’s study [8], considering only factors available at the first prenatal visit. Later, it was supplemented with conditions appearing in late pregnancy or at admission for delivery [20]. To customize a VBAC prediction model for Swedish conditions, we aimed at a model comprising data present well before labor and delivery. In Swedish midwifery-based obstetric care, otherwise healthy pregnant women with one previous CS would see an obstetrician only once, usually in the beginning of the third trimester. In the USA, a similar pregnant woman would visit her physician about ten times [22]. Referring to our previous studies showing considerable influence of the first CS indication on the
M.C. Fagerberg et al. / European Journal of Obstetrics & Gynecology and Reproductive Biology 188 (2015) 88–94
89
outcomes of the subsequent delivery [3,23,24], we aimed to improve the VBAC prediction ability by adding detailed information about the prior CS indication to the model. Methods We identified women in the Swedish Medical Birth Registry (MBR). The women would have given birth at least twice 1992– 2011, including one CS and at least one subsequent delivery. The last delivery was considered as the index delivery, and only the last infant of each woman was counted. Thus, each woman was included once. The MBR contains information on practically all deliveries in Sweden [25]. Standardized medical records are used at all antenatal clinics, delivery units, and pediatric examinations of newborns. The record forms are sent to the National Board of Health and entered into the MBR. The international classification of diseases (ICD)-9 was used for 1992–1996 and ICD-10 was used from 1997 onwards. Almost all pregnant women in Sweden utilize standardized antenatal care, offered free of charge. At the first prenatal visit, a midwife interviews each woman. Smoking habits, height, and weight are recorded. An ultrasound examination is offered before 20 weeks of pregnancy. The present study variables and inclusion criteria were selected in order to make a validation of the prediction model by Grobman et al. [8] feasible and to optimize it for Swedish conditions. Exclusion criteria for the index delivery were the same as in the Grobman study [8]: antenatal death, multiple gestation, preterm birth, non-cephalic presentation, ERCS, or no indication reported for the previous CS. A planned/elective cesarean delivery was defined as a CS being performed before the start of uterine contractions. The following variables were considered conceivable VBAC predictors (all valid for the last/index pregnancy): any previous vaginal delivery (yes/no), previous VBAC (yes/no), maternal age (continuous, linear variable), body mass index (BMI) (continuous, linear variable), smoking (ordinal variable, 1 = no, 2 =
Contents lists available at ScienceDirect
European Journal of Obstetrics & Gynecology and Reproductive Biology journal homepage: www.elsevier.com/locate/ejogrb
Predicting the chance of vaginal delivery after one cesarean section: validation and elaboration of a published prediction model Marie C. Fagerberg a,b,*, Karel Marsˇa´l b, Karin Ka¨lle´n c a
Department of Obstetrics and Gynecology, Ystad Hospital, Ystad, Sweden Department of Obstetrics and Gynecology, Clinical Sciences Lund, Lund University, Lund, Sweden c Centre of Reproduction Epidemiology, Clinical Sciences Lund, Lund University, Lund, Sweden b
A R T I C L E I N F O
A B S T R A C T
Article history: Received 12 October 2014 Received in revised form 25 January 2015 Accepted 19 February 2015
Objective: We aimed to validate a widely used US prediction model for vaginal birth after cesarean (Grobman et al. [8]) and modify it to suit Swedish conditions. Study design: Women having experienced one cesarean section and at least one subsequent delivery (n = 49,472) in the Swedish Medical Birth Registry 1992–2011 were randomly divided into two data sets. In the development data set, variables associated with successful trial of labor were identified using multiple logistic regression. The predictive ability of the estimates previously published by Grobman et al., and of our modified and new estimates, respectively, was then evaluated using the validation data set. The accuracy of the models for prediction of vaginal birth after cesarean was measured by area under the receiver operating characteristics curve. Results: For maternal age, body mass index, prior vaginal delivery, and prior labor arrest, the odds ratio estimates for vaginal birth after cesarean were similar to those previously published. The prediction accuracy increased when information on indication for the previous cesarean section was added (from area under the receiver operating characteristics curve = 0.69–0.71), and increased further when maternal height and delivery unit cesarean section rates were included (area under the receiver operating characteristics curve = 0.74). The correlation between the individual predicted vaginal birth after cesarean probability and the observed trial of labor success rate was high in all the respective predicted probability decentiles. Conclusion: Customization of prediction models for vaginal birth after cesarean is of considerable value. Choosing relevant indicators for a Swedish setting made it possible to achieve excellent prediction accuracy for success in trial of labor after cesarean. During the delicate process of counseling about preferred delivery mode after one cesarean section, considering the results of our study may facilitate the choice between a trial of labor or an elective repeat cesarean section. ß 2015 Elsevier Ireland Ltd. All rights reserved.
Keywords: Elective repeat cesarean Prediction model Vaginal birth after cesarean
Introduction The increasing rate of women having undergone cesarean section (CS) stresses the necessity of well founded counseling on subsequent pregnancy delivery mode – trial of labor (TOL) or elective repeat cesarean section (ERCS). To minimize negative outcomes for mothers and babies, the main issue would be to avoid planned vaginal deliveries ending in emergency CS [1–6]. Thus, a custom-built model for prediction of vaginal birth after cesarean (VBAC) would prove useful in pregnancies following one CS.
* Corresponding author at: Department of Obstetrics and Gynecology, Clinical Sciences Lund, Lund University, Lund, Sweden. Tel.: +46 701 70 87 51. E-mail address: [email protected] (M.C. Fagerberg). http://dx.doi.org/10.1016/j.ejogrb.2015.02.031 0301-2115/ß 2015 Elsevier Ireland Ltd. All rights reserved.
Several models aiming to predict VBAC have been published [7–17]. Some models have been externally validated or supplemented with new variables in subsequent publications [2,18–21]. To our knowledge, the most utilized model originates from the Grobman et al.’s study [8], considering only factors available at the first prenatal visit. Later, it was supplemented with conditions appearing in late pregnancy or at admission for delivery [20]. To customize a VBAC prediction model for Swedish conditions, we aimed at a model comprising data present well before labor and delivery. In Swedish midwifery-based obstetric care, otherwise healthy pregnant women with one previous CS would see an obstetrician only once, usually in the beginning of the third trimester. In the USA, a similar pregnant woman would visit her physician about ten times [22]. Referring to our previous studies showing considerable influence of the first CS indication on the
M.C. Fagerberg et al. / European Journal of Obstetrics & Gynecology and Reproductive Biology 188 (2015) 88–94
89
outcomes of the subsequent delivery [3,23,24], we aimed to improve the VBAC prediction ability by adding detailed information about the prior CS indication to the model. Methods We identified women in the Swedish Medical Birth Registry (MBR). The women would have given birth at least twice 1992– 2011, including one CS and at least one subsequent delivery. The last delivery was considered as the index delivery, and only the last infant of each woman was counted. Thus, each woman was included once. The MBR contains information on practically all deliveries in Sweden [25]. Standardized medical records are used at all antenatal clinics, delivery units, and pediatric examinations of newborns. The record forms are sent to the National Board of Health and entered into the MBR. The international classification of diseases (ICD)-9 was used for 1992–1996 and ICD-10 was used from 1997 onwards. Almost all pregnant women in Sweden utilize standardized antenatal care, offered free of charge. At the first prenatal visit, a midwife interviews each woman. Smoking habits, height, and weight are recorded. An ultrasound examination is offered before 20 weeks of pregnancy. The present study variables and inclusion criteria were selected in order to make a validation of the prediction model by Grobman et al. [8] feasible and to optimize it for Swedish conditions. Exclusion criteria for the index delivery were the same as in the Grobman study [8]: antenatal death, multiple gestation, preterm birth, non-cephalic presentation, ERCS, or no indication reported for the previous CS. A planned/elective cesarean delivery was defined as a CS being performed before the start of uterine contractions. The following variables were considered conceivable VBAC predictors (all valid for the last/index pregnancy): any previous vaginal delivery (yes/no), previous VBAC (yes/no), maternal age (continuous, linear variable), body mass index (BMI) (continuous, linear variable), smoking (ordinal variable, 1 = no, 2 =
