Linked Comment: Ultrasound Obstet Gynecol 2013:41:672–678 In this issue of the Journal, Naji et al. suggest a prognostic model for the likelihood of successful VBAC, constructed using a single-center sample of 131 women with VBAC planned at ≥ 38 gestational weeks. The outcome measure for this model is successful vaginal delivery and the predictors comprised maternal age, prior successful VBAC, residual myometrial thickness (RMT) in the second trimester and change in RMT from the first to the second trimester (ΔRMT). The authors should be congratulated for having conducted a methodologically rigorous study using appropriate statistical analyses. However, the rationale for such a model and its construction is not obvious and several issues should be borne in mind when considering the take-home message(s) of this work. First, do we wish to predict the failure to deliver vaginally and the resulting morbidity of an emergency Cesarean section or the risk of acute uterine rupture and its undisputed morbidity and mortality for both the mother and newborn? The latter is quite rare and poorly reported in most studies, including the present one, in which uterine rupture was defined as a full-thickness separation of the uterine wall and overlying serosa observed at Cesarean section, regardless of the causal relationship of this finding with maternal and fetal distress. Naji et al. chose to predict VBAC, thereby questionably classifying emergency Cesarean section during labor as an adverse event, regardless of the cause. However, they cleverly investigated the impact of a physician or patient's level of tolerance to the risk of emergency Cesarean on the model-based decision using decision curves, suggesting overall superiority of a model-based decision with mode of delivery as the outcome of interest. Second, in routine obstetrics, a constant goal is to refine the appraisal of future risks. A model therefore aims to predict a complication that may occur, and the choice of covariates to predict this outcome is of crucial importance. In their article, Naji et al. suggest a model based upon first- and second-trimester measurements of RMT together with two simple demographic covariates as predictors for a successful VBAC. In their population, this model was strikingly discriminative, with an area under the model-associated receiver–operating characteristics curve of 0.96, indicating that one should be able to discriminate with very high confidence (i.e. with a very low false-positive rate) those patients who will eventually require a Cesarean section for any reason during labor. This result should be interpreted with caution, as it suggests that this set of covariates strongly determines the occurrence of any peripartum complication, such as arrested cervical dilatation and fetal distress. Moreover, in the final multivariable model, prior VBAC was found to be statistically non-significant, emphasizing the potential role of RMT as the sole determinant of successful VBAC. Although the quality of the scar tissue may be related to arrested dilatation and fetal distress, it cannot be expected that ultrasound scar characteristics in the first and second trimesters explain all possible complications that can occur during labor in VBAC attempts. Third, the interpretation and generalization of results does not depend solely on the choice of outcome, covariates and study population; it also depends on variations in local practice. For example, in the study of Naji et al., 17/121 (14%) women attempting VBAC were induced using prostaglandins, with a visible effect on the outcome, yielding 9/17 subsequent emergency Cesarean sections. Such practice is not universal and would be contraindicated in some centers. Conversely, as expected, successful VBAC was associated with spontaneous onset of labor. Similarly, breech presentation and low-lying placenta may be considered contraindications for VBAC, if not for vaginal birth altogether. Nonetheless, breech and placenta previa accounted for 10.7% of the Cesarean sections performed during labor in the present study. More subtle parameters, such as the opinion of the obstetrician regarding VBAC and elective repeat Cesarean section, are difficult to measure, but are likely to affect the results. These reasons make validation in external populations mandatory and might explain failure of the validation process if this were to occur. This work shows the potential impact on the outcome of VBAC of measurement of RMT by transvaginal ultrasound in the first and second trimesters. This finding adds to our understanding of the mechanisms that underlie obstetric complications in VBAC and, in the future, should help in refining the path of care of women with prior Cesarean section. However, as the authors themselves acknowledge, it would be inappropriate to implement this model in routine practice without further validation. Not only should the model itself be validated but also its impact on public health outcomes in large populations should be investigated in an evidence-based process. J. Stirnemann Obstetrics and Maternal Fetal Medicine, Hôpital Necker–Enfants Malades, AP-HP, Paris, France; Applied Mathematics, MAP5, UMR CNRS 8145 Université Paris Descartes, Paris, France (e-mail: j.stirnemann@gmail.com)
Read full abstract