Methods
Patient Cohort and Data Collection
A total of 1015 patients were initially enrolled, and 940 patients were involved in the final analysis. Patients enrolled between June 2014 and May 2017 underwent scar defect repair in Shanghai First Maternity & Infant Hospital, and those enrolled between May 2017 and May 2021 underwent repair in Xinhua Hospital (The enrolled subjects were not replicated with the previous articles published by our team). All VR surgeries were performed by a single doctor. Seventy-five patients were lost to follow-up at 3 or 6 months after CSD repair. According to the results of our previous studies, the scar stabilizes 3 months after CSD repair surgery, and there is no change after 6 months16. Therefore, we used MRI measurements of CSD at 6 months postoperative, in combination with a statistical analysis of age, uterine position, number of C-sections, the timing of cesarean section(selective or emergency operation), menstrual cycle, menstrual period and CSD size before and after repair surgery. According to the analysis results, the key factors determining the prognosis of diverticulum repair were identified, and the scoring model was established (eFigure 1 in the Supplement ).
This study was approved by the Ethics Committee of Xinhua Hospital Affiliated with Shanghai Jiao Tong University School (XHEC-H-2018-002) of Medicine and Shanghai First Maternity & Infant Hospital affiliated with Tongji University (KS1512). All patients signed written informed consent to participate in this study.
Surgical Procedures
The surgical techniques are described in detail in our previous publication15, and the most important steps of the procedure are summarized below.
At a distance of 0.5 cm below the site of the reflexed vesica-cervical area, an anterior incision was made from the 3 o’clock to the 9 o’clock position using an electric knife. After entry into the abdominal cavity and complete exposure of the cervix, the location of the uterine defect was determined, and the thickness of the lower uterine segment of the CSD could be gauged through contact with the surgeon’s index finger and sounding. The CSD tissue was resected, the incision tissues were trimmed to healthy myometrium with dissecting scissors, and the incision was closed with a double layer of 1–0 absorbable interrupted sutures. All operations were performed by the same surgeon with extensive vaginal surgical experience.
Statistical Analysis
In the analyses, β coefficients and ORs are presented as estimates of effect, and the reported statistical significance levels were all two-sided, with statistical significance set at 0.05. Statistical analysis was conducted with R software (3.5.1; R Foundation for Statistical Computing) and SAS version 9.4 (SAS Institute Inc., Cary, NC).
A descriptive analysis of patients and a demographic comparison between the suboptimal healing group and the optimal healing group were performed. The distributions of the baseline characteristics of the patients are represented by the mean ± SD or as a number (n) with a percentage (%). Differences between groups were assessed as follows: t-test and ANOVA were used for continuous variables and multiple comparisons, and the χ2 test was used for categorical variables.
Multivariate linear regression was used to assess the association between CSD parameters and optimal healing, CSD parameters and menstrual improvement, and CSD parameters and TRM after adjustment for potential confounding variables.
These patients were randomly divided into two cohorts: a training cohort (634, 70%) and an internal validation cohort (306, 30%). After multivariable logistic regression analysis, clinical candidate predictors were applied to develop an individualized prediction model by using the training cohort. Forward and backward stepwise regression was applied by using the likelihood ratio test with Akaike’s information criterion as the stopping rule. To provide a quantitative tool for prediction, we built nomograms based on multivariate logistic analysis of the training cohort. We selected the point that yielded the highest Youden’s index (i.e., specificity + sensitivity - 1) on the receiver operating characteristic (ROC) curve of the training set as the optimal cutoff value and used it to calculate which score combination for every patient represented the sum of scores for each risk factor.
Calibration curves were plotted to assess the calibration of the nomogram using the Hosmer‒Lemeshow test. To quantify the discrimination performance of the nomogram, Harrell’s C-index was measured. The nomogram was subjected to bootstrapping validation (1,000 bootstrap resamples) to calculate a relatively corrected C-index. We also evaluated the nomogram using the area under the curve (AUC) of the ROC with a 95% confidence interval (95% CI). Internal validation was carried out using data from 306 patients. Decision curve analysis was performed to determine the clinical usefulness of the nomogram by quantifying the net benefits at different threshold probabilities.