Materials and Methods

This is a nested case-control study conducted from 1996 through 2000 within a 17-center retrospective cohort study of pregnant women with at least 1 previous cesarean delivery. To identify factors associated with uterine rupture, all cases (women who attempted TOLAC and experienced uterine rupture) were matched on hospital site with 5 control subjects, chosen by a random number generator, who attempted TOLAC but did not have a uterine rupture. Institutional review board approval was obtained from all study sites. A detailed description of the parent study has been published previously, but a brief description follows.

International Classification of Disease, Ninth Revision codes for “previous cesarean delivery, delivered,” were used to identify subjects at each site and data were extracted from medical charts by trained research nurses using standardized, closed-end data collection forms. Three percent of charts were reextracted for quality control. Data collected included maternal demographics, medical and obstetric history, antepartum course, labor and delivery events, complications, and maternal outcomes. Data for patients selected for the case-control study were reextracted in further detail, including all procedures, medications, and exam details in 15-minute time increments throughout labor. Only women with ≥1 LTCS were included in the parent cohort; patients were excluded if their prior cesarean was not low transverse.

Uterine rupture was explicitly defined a priori as a full-thickness disruption of the uterine wall accompanied by at least one of the following clinical signs: nonreassuring fetal heart rate tracing immediately preceding surgery, hemoperitoneum, or signs of maternal hemorrhage (systolic blood pressure <70 mm Hg, diastolic blood pressure <40 mm Hg, or heart rate >120 beats/min). This definition was used to distinguish a clinically significant uterine rupture from an asymptomatic or incidental finding of uterine scar separation or “uterine window.”

For this analysis, women who attempted TOLAC were identified as having a labor induction by a directly extracted dichotomous variable for “induce.” Subjects were excluded if they had >1 prior LTCS. Cases (uterine rupture) were compared with control subjects (no uterine rupture) with respect to baseline characteristics: χ ² or Fisher exact tests, as appropriate, for dichotomous variables and Student t test or Mann-Whitney U test, as appropriate, for continuous variables. Additionally, a sensitivity analysis of sociodemographics was performed, comparing the controls used for this analysis with the group of patients who did not experience a uterine rupture in the larger cohort to ensure that the controls chosen at random were representative (data available upon request). Because controls for this analysis were representative of the larger cohort, weights for the final covariates were not used.

For the time-to-event analysis, patients were classified as having the event of interest (uterine rupture) or censored (delivered). Imputed values were not used because data were nearly complete; <2% of data points were missing for any given variable. Subjects were grouped according to induction of labor (exposure) or spontaneous onset of labor. We anticipated that the admission exam would typically be a smaller cervical dilation for subjects admitted for induction compared to those admitted in labor, which would introduce left censoring for those presenting in spontaneous labor. Therefore, time zero was defined as the first exam at 4 cm to minimize left censoring. An exam of 4 cm was chosen as a cutoff, not as a surrogate marker for labor, but because the majority of laboring subjects were admitted with an initial exam ≤4 cm. In this study 12 uterine ruptures occurred prior to 4 cm: 7 in the induction group and 5 in the spontaneous onset of labor group. As these uterine ruptures were evenly distributed between the exposed and unexposed groups, we believe that the exclusion of these subjects did not significantly bias our results.

Because some women who present in spontaneous labor eventually require oxytocin augmentation and because oxytocin has been linked in some studies to an increased risk of uterine rupture, a secondary analysis was performed defining labor as induced, augmented, or spontaneous. An additional secondary analysis was performed to examine the effect of cervical dilation (the extent of cervical ripening) at initiation of induction. As Bishop score was not routinely available for all subjects, cervical dilation at the time of starting oxytocin was used as a surrogate marker. Cervical dilation at the time of starting oxytocin was categorized as <2 cm, 2-3.9 cm, 4-5.9 cm, and ≥6 cm.

Kaplan-Meier plots were used to graphically illustrate the risk of uterine rupture over time by whether or not labor was induced. Log rank tests were used to compare the plots. Univariable analyses were used to identify potentially confounding factors in the labor induction-uterine rupture risk relationship. Cox proportional hazard regression was used to model the effect of induction of labor on the risk of uterine rupture; adjustment was made for potentially confounding effects identified in the univariable analysis and those historically proposed, such as prior vaginal delivery, race, and oxytocin dosing. The proportional hazards assumption was tested using cumulative martingale residuals and the Kolmogorov-based supremum test. All statistical analyses were completed with SAS (version 9.2; SAS Institute Inc, Cary, NC) and STATA (version 10 Special Edition; StataCorp, College Station, TX).