Ovarian endometrioma recurrence after laparoscopic surgery: First assessment of ACSAP predicting score

Endometriosis is one of the most common benign gynaecological diseases and is characterised by ectopic endometrial-like tissue outside the uterus.[ 1 ] Ovarian endometrioma (OMA) is the most common form of endometriosis, affecting 2%–10% of women of childbearing age and nearly 50% of women with infertility.[ 2 3 ] Moreover, laparoscopic ovarian cystectomy is a treatment option for women with OMA who have intolerable symptoms or other medical conditions.[ 4 ] However, the relatively high rate of post-operative recurrence has posed significant challenges to clinicians.[ 5 ] The recurrence rate remains high after reoperation for OMA recurrence,[ 6 ] highlighting the urgent need for more effective comprehensive treatments. OMA recurrence may require repeated surgeries and lead to premature ovarian failure, which causes great physical and mental burden to patients.[ 7 ] However, the recurrence risk differs according to patient characteristics, and one-size-fits-all treatment strategies may not be appropriate. Several studies have investigated the risk factors for recurrence in patients with OMA.[ 8 9 10 11 12 ] A recent meta-analysis demonstrated that risk factors such as age at surgery, cancer antigen 125 (CA125) level, cyst size, dysmenorrhea, history of endometriosis-related surgery and revised American Society for Reproductive Medicine (rASRM) score significantly influenced the risk of OMA recurrence.[ 13 ] However, to the best of our knowledge, no predictive model has been established using these factors to stratify the individual risk of OMA recurrence. Therefore, in the present study, we aimed to develop and validate a predictive score for estimating the risk of OMA recurrence after laparoscopic surgery.

This study included two cohorts of patients with OMAs. Training and testing sets comprising 431 and 185 consecutive women who underwent laparoscopic surgery for OMA were obtained from our institution between January 2015 and September 2017 and October 2017 and October 2018, respectively. The inclusion criteria were: age > 18 years, histologically confirmed OMA, no residual lesions confirmed by the first ultrasonography within 6 months after surgery and a minimum follow-up period of 60 months. Patients were excluded from the study if they underwent concurrent resection of other organs (e.g. hysterectomy) ( n = 16), had revised (rASRM) classification I or II ( n = 34), had a concurrent borderline or malignant ovarian tumour ( n = 2) or had incomplete clinical data ( n = 10). Clinical data were collected retrospectively from the medical records of each patient. This study was approved by the local ethics committee. Written informed consent was obtained from all the patients. This study followed the Transparent Reporting of a multivariate prediction model for Individual Prognoses Or Diagnosis (TRIPOD) reporting guidelines. In our institution, all patients were followed up postoperatively by routine gynaecological examinations and ultrasonography at 6–12-month intervals. Additional follow-up was performed if the patient developed symptoms. The follow-up period was completed in October 2023. The primary outcome of this study was recurrence of OMA, which was defined according to the following ultrasonography findings: a round cystic mass ≥20 mm with thick walls, irregular margins, homogeneous low-echogenic fluid content, scattered internal echoes or negative papillary proliferation. Multiple cysts with the sum of cyst sizes measuring over 20 mm were also considered as a recurrence.[ 14 ] The duration of treatment was calculated using the interval between medication use and initial recurrence. Post-operative medications included gonadotropin-releasing hormone agonists, oral contraceptive pills, the levonorgestrel-intrauterine system or a combination of these. Patients who received medical treatment for more than 15 months were considered to have good compliance.[ 14 ] Post-operative pregnancy was defined as the visualisation of a gestational sac with demonstration of embryonic cardiac activity after surgery. Candidate variables (age at surgery, pre-operative CA125, cyst size, dysmenorrhea, history of OMA surgery and rASRM stage) were selected from the risk factor analysis performed in a meta-analysis by Jiang et al .[ 13 ] We used the least absolute shrinkage and selection operator (LASSO) Cox regression model to select the most useful factors for predicting the scores in the training set.[ 15 ] Ten cross-validations were used to determine the optimal combination of factors and corresponding coefficients through 1-standard error criteria. Based on the LASSO coefficients, we established a simplified version of the scoring system (the ACSAP score) that could be implemented in real-life practice while maintaining predictive performance and the ability to calculate absolute risk.[ 16 ] The predictive performance of the ACSAP score was assessed based on discrimination, calibration and clinical usefulness. The time-dependent receiver operating characteristic curve and area under the curve (AUC) were used to assess the discriminatory power. The calibration of the score was evaluated using calibration plots and the Hosmer–Lemeshow test. Decision curve analysis was used to evaluate the clinical usefulness of these scores. Data were summarised as mean (standard deviation) or frequency (percentage) and compared using the Chi-square test or t -test. Univariate and multivariate Cox regression analyses were used to identify independent predictors of recurrence. Variables with a value of P < 0.05 in the univariate analysis were included in the subsequent multivariate analysis. The optimal cut-off values for the ACSAP score for recurrence were determined using the X-tile programme (3.6.1 software 20, http://medicine.yale.edu/lab/rimm/research/software.aspx ). The relationship between the ACSAP score and risk of recurrence was evaluated using restricted cubic splines. Statistical significance was defined as a two-tailed P < 0.05. All statistical analyses were performed using the SPSS (version 22.0; IBM Corporation, Armonk, NY, USA) and R software version 4.1.3 (R Foundation for Statistical Computing, Vienna, Austria).

The baseline characteristics of the training and testing sets are presented in Table 1 . The median (interquartile range) follow-up durations were 84 (78–93) months and 63 (54–70) months for the training and testing sets, respectively. Overall, 101 of the 431 patients (23.4%) in the training set and 28 of the 185 patients (15.1%) in the testing set experienced OMA recurrence. Clinical characteristics of patients in the training and testing sets Data are expressed as mean (SD) or n (%). BMI: Body mass index, SD: Standard deviation, OMA: Ovarian endometrioma, rASRM: Revised American Society for Reproductive Medicine, AMH: Anti-Müllerian hormone Supplementary Table 1 shows the distribution of the clinical characteristics between the recurrence and non-recurrence groups in the training set. Patients who experienced recurrence were more likely to be younger ( P < 0.001), have larger cysts ( P < 0.001), have undergone previous surgery for OMA ( P = 0.003) and have rASRM stage IV disease ( P < 0.001). In addition, patients who experienced post-operative pregnancy were less likely to experience recurrence ( P = 0.018). Clinical characteristics of patients according to ovarian endometrioma recurrence in the training set Data are expressed as mean (SD) or n (%). BMI: Body mass index, SD: Standard deviation, OMA: Ovarian endometrioma, rASRM: Revised American Society for Reproductive Medicine, AMH: Anti-Müllerian hormone Amongst the 431 patients in the training set, age, cyst size, history of OMA, rASRM stage and post-operative pregnancy were significant predictors of recurrence (all P < 0.05). In the multivariate analysis, younger age ( P = 0.015), larger tumour size ( P < 0.001), previous surgery for OMA ( P = 0.001) and stage IV disease ( P = 0.005) were independently associated with a higher risk of recurrence, whereas post-operative pregnancy was an independent protective factor for recurrence [ P = 0.044; Table 2 ]. Univariate and multivariate analyses for ovarian endometrioma recurrence in the training set HR: Hazard ratio, BMI: Body mass index, CI: Confidence interval, OMA: Ovarian endometrioma, rASRM: Revised American Society for Reproductive Medicine, AMH: Anti-Müllerian hormone After the LASSO regression, five factors were retained in the prediction score [ Supplementary Figure 1 ]. A simplified scoring system was established based on LASSO coefficients [ Table 3 ]. The relationship between the ACSAP score and OMA recurrence is shown in Supplementary Figure 2 . Higher scores were linearly associated with a higher risk of recurrence in the training and testing sets (both P for non-linear >0.05). Points assigned for the ACSAP predicting score OMA: Ovarian endometrioma, rASRM: Revised American Society for Reproductive Medicine, ACSAP: Age, Cyst size, previous Surgery for OMA, rASRM stage, and postoperative Pregnancy Using X-tile plots, we classified patients into three risk groups with scores of 6 and 12 as the cut-off values [ Supplementary Figure 3 ]. In the training set, the 5-year cumulative recurrence rates were 5.3%, 20.2% and 48.0% for the low-, intermediate- and high-risk groups, respectively [log-rank P < 0.001; Figure 1a ]. In the testing set, the 5-year cumulative recurrence rates were 0%, 16.5% and 36.5% for the low-, intermediate- and high-risk groups, respectively (log-rank P < 0.001; Figure 1b ]. Cumulative probability of ovarian endometrioma recurrence in the low-, intermediate- and high-risk groups in the training (a) and testing sets (b). We calculated the P values using the log-rank test In the training set, the AUC of the ACSAP scores at 3 and 5 years were 0.741 and 0.727, for recurrence of OMA [ Figure 2a ]. In the testing set, the AUC values of the scores were 0.765 and 0.795 for the 3- and 5-year recurrence rates, respectively [ Figure 2b ]. Calibration plots showed that the score performed well compared with the performance of an ideal model in both the training and testing sets [all Hosmer–Lemeshow P > 0.05, Supplementary Figure 4 ]. The promising clinical utility of the score is depicted in Supplementary Figure 5 using decision curve analysis. The ACSAP predicting score measured by time-dependent receiver operator characteristic curves in the training (a) and testing sets (b). We used the area under the curve values at 3 and 5 years to assess predictive accuracy. AUC: Area under the curve In univariate and multivariate analyses, post-operative medical treatment was not significantly associated with OMA recurrence [ P > 0.05, Table 2 ]. When stratified by the ACSAP score, the patients in the intermediate-risk [15.0% vs. 24.6%, log-rank P = 0.011, Figure 3b ] and high-risk groups [32.1% vs. 80.0%, log-rank P < 0.001, Figure 3c ] experienced a significant reduction in 5-year cumulative recurrence with medical treatment for at least 15 months, whereas those in the low-risk group did not benefit from post-operative treatment [log-rank P = 0.872, Figure 3a ]. Cumulative probability of ovarian endometrioma recurrence in the low (a), intermediate (b) and high-risk groups (c) stratified by the duration of post-operative medical treatment in the entire cohort. We calculated the P values using the log-rank test

In this study, we present the first attempt to develop a scoring system to predict the probability of OMA recurrence after laparoscopic excision. The score incorporates five clinical characteristics and successfully stratifies patients based on their individual recurrence risk. Validation of the score in the testing set confirmed its performance and robustness. In a meta-analysis involving 2331 patients with OMA, eight variables were identified as significant risk factors for endometrioma recurrence.[ 13 ] Using LASSO regression, we selected five variables and calculated the ACSAP scores. All parameters included in the score were independently associated with OMA recurrence. Of these, a younger age at surgery was closely related to an increased risk of recurrence.[ 10 17 18 ] In a retrospective cohort study by Seo et al. , the cumulative recurrence rate at 40–45 years (10.2%) was significantly lower than that at 20–29 years (43.3%) and 30–39 years (22.5%).[ 19 ] There are two possible explanations for this: one is that oestrogen levels decline with increasing age[ 10 ] and the other is that the biological features of OMA in younger women seem more aggressive.[ 8 18 ] Cyst size and disease stage were significant predictors of OMA recurrence. Large cysts and extensive adhesions may increase the difficulty of complete OMA excision, thereby increasing the possibility of recurrence.[ 20 21 ] Our results also showed that patients with a surgical history of endometrioma were more likely to experience recurrence. A possible explanation is that these patients may have more aggressive disease.[ 8 ] In addition, post-operative pregnancy had a protective effect against OMA recurrence, which is consistent with previous studies.[ 14 17 22 23 ] This finding could be explained by pregnancy-associated anovulation and amenorrhea, which promote the necrosis of endometriotic lesions. Our study did not support the prognostic value of pre-operative symptoms or CA125 levels. Tobiume et al . reported a significant association between pre-operative dysmenorrhea and disease recurrence.[ 20 ] However, these symptoms were subjectively assessed, and the evaluation criteria were inconsistent in this retrospective study, which limited the reliability of their results. For pre-operative CA125 levels, the timing of blood sampling may have influenced the results because of its dynamic changes throughout the menstrual cycle.[ 13 ] However, further studies are needed to confirm this association. The association between post-operative medical treatment and the risk of OMA recurrence is still under debate. In a recent meta-analysis of 2137 patients, Zakhari et al . demonstrated a significantly decreased risk of OMA recurrence in patients who received post-operative medical treatment.[ 24 ] However, other studies have not supported the prognostic value of post-operative treatment.[ 13 17 22 ] In a large study involving 756 patients, Choi et al . showed that an adequate duration of medical treatment resulted in lower recurrence after adjusting for disease characteristics.[ 14 ] Our study used 15 months as a cut-off value for treatment compliance[ 14 ] and found that good compliance significantly reduced cumulative recurrence in the intermediate- and high-risk groups. Based on these results, we recommend long-term medical treatment for patients with higher ACSAP scores who do not desire pregnancy after surgery. This study had some limitations. First, this was a retrospective study, and selection bias could not be prevented. Moreover, we could not analyse pain recurrence or quality of life as secondary outcomes to strengthen our findings. Second, this study analysed only the duration of post-operative treatment. The type of treatment may also have influenced the results; however, this information could not be collected because of the retrospective nature of the study. Third, pre-operative medication for OMA was not routinely administered in our institution and was not included in the analysis. The addition of this parameter to the ACSAP score requires further investigation. In conclusion, the ACSAP score is the first validated scoring system for the risk stratification of OMA recurrence. This objective, easy-to-use clinical tool provides significant information on patient prognosis and should be applied to guide individual management strategies for patients with OMA after laparoscopic surgery. External validation is required to confirm the robustness of this score. The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request. This study was conducted in accordance with the Declaration of Helsinki (revised in 2013) and was approved by the institutional review board of the Xingtai Third Hospital. Written informed consent was obtained from all patients for being included in the study. Niu C, Bai X and Gui X conceived of the study and designed the study; Liang Y and Zhang L helped collect data; Niu C analysed the data; Niu C wrote the manuscript; Bai X and Gui X helped revise the manuscript critically for important intellectual content. All authors read and approved the final manuscript. There are no conflicts of interest.

Feature selection using the least absolute shrinkage and selection operator (LASSO) Cox regression model. (a) Ten-fold cross-validation for tuning parameter selection in the LASSO model. Solid vertical lines represent binomial deviance ± standard error. The vertical lines are drawn at the optimal values by minimum criteria and 1 - S.E. criteria. (b) LASSO coefficient profiles of the 7 features. A coefficient profile plot was produced against the log (λ) sequence. A vertical line was drawn at the value selected using ten-fold cross-validation, where optimal λ resulted in 5 non-zero coefficients Univariate Cox analysis of ovarian endometrioma recurrence with restricted cubic splines in the training (a) and testing sets (b) X-tile plots Plots depict the calibration of the ACSAP predicting score in terms of agreement between predicted and observed outcomes in the training (a) and testing sets (b). Model performance is shown by the plot, relative to the 45° line, which represents perfect prediction Decision curve analysis of the ACSAP predicting score in the training (a) and testing sets (b)