Impact of Surgeons’ Experience on Oncologic Outcomes in Women Undergoing Laparoscopic Radical Hysterectomy for Cervical Cancer: A Comparison of the First 50 and last 50 Cases

Objectives: To compare the oncological outcomes of the rst 50 laparoscopic radical hysterectomy (LRH) surgeries with the last 50 LRH, performed by high volume surgeons, for cervical cancer patients. Design: A nationwide multicentre retrospective cohort study Setting: Clinical diagnosis and treatment of cervical cancer patients in mainland China (Four C) database. Population: women with early cervical cancer undergone LRH. Methods: We retrospectively analyzed the oncological outcomes of 1004 cervical cancer patients who underwent LRH performed by 19 surgeons. They were divided into two groups according to the sequence of operations, the rst 50 and the last 50 patients with LRH. Kaplan-Meier survival analysis and log-rank test, Cox proportional risk regression model and propensity score matching were used. Main Outcome Measures: 5-year overall survival (OS) and disease-free survival (DFS) rates. Results: There were no signicant differences in the 5-year OS and DFS between rst 50 patients with LRH group (n=413) and last 50 patients with LRH group (n=591) (OS: p=0.388; DFS: p=0.226). The last 50 cases of LRH was not an independent risk factor for OS and DFS in early cervical cancer patients (p=0.830, p=0.300). After propensity score matching, similar outcomes were observed (n=364:364 (cid:0) OS:P = 0.764; DFS:P = 0.705). Conclusions: The oncological outcomes of the rst 50 LRH surgeries were similar to those of the last 50 surgeries in patients with early-stage cervical cancer. Increase in the surgeons’ experience did not improve signicantly with oncological outcomes of patients with early stage cervical cancer after LRH.


Introduction
Radical hysterectomy in combination with pelvic lymphadenectomy is the standard treatment for nonfertility sparing stage IA1 cervical cancer with positive lymphovascular space invasion (LVSI) to stage IIA2 cervical cancer. NCCN guideline 1 used to indicate that laparotomy or minimally invasive surgery (MIS) are acceptable for patients with early stage cervical cancer. Due to these suggestions, MIS was widely used. However, in November 2018, Ramirez et al. 2 published a multicentre, phase III randomized controlled trial in the New England Journal of Medicine, which reported that laparoscopic radical hysterectomy(LRH) was associated with worse OS and DFS rates after 4.5 years in patients with early stage cervical cancer than was abdominal radical hysterectomy (ARH). In the same period, retrospective cohort data from Melamed et al. 3 were also published, and LRH was reported to result in a shorter OS than ARH. These two studies overturned the traditional understanding of laparoscopic surgery in gynecologic oncology. Cervical Cancer Guidelines (Version 3.2019) released by NCCN Clinical Practice Guidelines in Oncology in 2019 4 no longer recommends laparoscopic surgery as the preferred surgical approach for the treatment of early stage cervical cancer .
Although these two studies suggested that the long-term prognosis was poor for patients that underwent LRH with early stage cervical cancer, the study did not provide reasons for the poor e cacy of laparoscopy. Among many possible causes, surgical experience was often thought to in uence oncological outcomes. Therefore, we suggest a hypothesis: was the surgeons' experience one of the reasons for the poor outcomes of LRH surgery? However, the previous studies on the learning curve of LRH for early stage cervical cancer focused on operative techniques such as operation time, bleeding volume, incidence of intraoperative and postoperative complications and perioperative outcomes, rarely focused on long-term oncological outcomes and ignored the in uence of surgeons' experience on the long-term oncological outcomes of patients with early stage cervical cancer. In addition, in most of the studies, single-center, single-surgeon and small samples were included.
To test this hypothesis, we explored whether the experience of surgeons affects the e cacy of LRH based on the clinical diagnosis and treatment of cervical cancer in China database. This study is a multicenter retrospective study. We compare the oncological outcomes of the rst 50 LRH surgeries and the last 50 LRH surgeries of surgeons for patients with early stage cervical cancer.

Data Source
The data used in this study is from the Clinical Diagnosis and Treatment for Cervical Cancer in China All follow-up procedures were carried out by trained gynecological oncology staff at each centre to keep the patients' personal data con dential and to provide disease management guidance at the same time.
The follow-up information was gathered through the return visit system or telephone follow-up, including survival status, time of death, recurrence time, recurrence site, treatment after recurrence. The oncological outcomes were estimated according to the recorded information, and the last day of the return visit or telephone follow-up was de ned as the last follow-up. We follow-up all patients, but the follow-up rate of oncological outcomes was 72.7% in this database. Detailed data collection requirements and database establishment processes were described in our previously published study. 5,6 Laparoscopic cervical cancer surgery started to be conducted in all participating institutions after 2004,so the rst LRH cases performed by surgeons after this year were included in this database. In this study, more than 600 surgeons were enrolled. In the database, 19 surgeons who completed more than 100 cases of LRH were enrolled. The operation cases were sorted according to the operation date. The rst 50 patients with LRH were considered a group and the last 50 patients with LRH were considered another group. The patients were grouped according to the date of the operation, regardless of whether they met the inclusion or exclusion criteria.
In this database, the FIGO stage was recorded and corrected by tumour size according to the FIGO 2009 staging system. Tumour size was evaluated using magnetic resonance imaging, computed tomography, ultrasound, physical examination or postoperative pathological records.

Inclusion and exclusion criteria
The patients' inclusion criteria were as follows: age 18 years or older; cervical cancer diagnosis of FIGO stage IA1 with positive LVSI to stage IIA2; squamous cell carcinoma, adenocarcinoma, or adenosquamous cell carcinoma; a history of laparoscopic surgery; type QM-B or QM-C 7 hysterectomy with pelvic lymphadenectomy with or without paraaortic lymphadenectomy; and a postoperative survival outcome. Surgeons who performed 100 or more LRH surgeries were also included, and we excluded patients if they had neoadjuvant chemotherapy or radiotherapy before surgery. Patients with pregnancy or other types of malignant tumours were excluded, too.
In subgroup analyses, the standard postoperative adjuvant therapy was de ned as cervical cancer patients with one or more high-risk factors (lymph node metastasis, parametrial involvement, and surgical margin invasion) or patients with two or more intermediate-risk factors (deep cervical stromal invasion, tumor size > 4 cm, and LVSI were received radiotherapy or chemo-radiotherapy. 8

Outcome measures
The overall 5-year survival rate (OS) and disease-free survival rate (DFS) were calculated. OS was de ned as the time from diagnosis to death for any reason. DFS was de ned as the time from diagnosis to death or recurrence for any reason. Data regarding patients with no evidence of recurrence or death were censored at the date of last follow-up.

Statistical methods
Metrological data were expressed as the mean (+ standard deviation) and categorized data by percentage (%). The Kaplan-Meier curve was used to describe the changes in survival outcomes, and the log-rank test was used to compare the differences in survival curves among groups. A Cox proportional risk regression model was used to adjust for case mix, determine independent risk factors, and calculate the relevant risk and 95% con dence interval. The statistical software used was SPSS 23.0 (SPSS Inc., Chicago, IL, USA). P < 0.050 was considered statistically signi cant. In the multivariate model, we also included the following known variables that may affect prognosis to reduce confounding effects: FIGO stage, age, histology, tumour size, depth of the stromal invasion, LVSI, lymph node metastasis, surgical margin invasion, parametrial tumour invasion, and postoperative adjuvant treatment.
To control confounding factors, each patient in the rst 50 patients with LRH group was matched to one patient in last 50 patients with LRH group using the known risk factors for recurrence or death using propensity score matching (PSM).

Results
A total of 46 313 patients with cervical cancer were included from the large database on the clinical diagnosis and treatment of cervical cancer in China. 7670 patents received LRH, and there were 19 surgeons operating more than 100 LRH. The mean LRH of 19 surgeons is 201.2 ± 85.2 (range 102 to 375). According to the inclusion and exclusion criteria, 1004 patients with early stage cervical cancer were nally included. The data ltering process is shown in Fig. 1.
Median follow-up of censored cases was 35.3 months (interquartile range, 26.0-51). Clinico-pathologic characteristics of two groups before and post matching are shown in Table 1. The rst 50 patients with LRH group was younger than another group. In both groups, patients with IB1 stage disease accounted for most(66.1% vs. 65.0%). The rst 50 patients with LRH group were more likely to have lower stage disease, whereas the last 50 patients with LRH group were more likely to have higher stage disease. Tumours in the rst 50 patients with LRH group were less likely to have lymphovascular space invasion and deep stromal invasion (all, P < 0.05).  Fig. 2). Cox multivariate analysis (see in Table 3) showed that surgical experience was not an independent risk factor for poor OS in patients with early stage cervical cancer (p = 0.830) or an independent risk factor for poor DFS (p = 0.300).
Multicollinearity test and cox proportional hazard regression models were used for analysis.
Proportional hazard assumption was tested and showed no interaction with time.
Bold indicates signi cant P-value.
After propensity score matching, a total of 728 patients (364 in rst 50 patients with LRH group and 364 in last 50 patients with LRH group) met the criteria for inclusion. The baseline data of patients after matching were consistent (Table 1). OS was 93.8% in group rst 50 patients with LRH group and 96.9% in group last 50 patients with LRH group (see in Fig. 2). There was no signi cant difference in OS between the two groups (p = 0.764). Cox multivariate analysis (see in  Multicollinearity test and cox proportional hazard regression models were used for analysis.
Proportional hazard assumption was tested and showed no interaction with time.
Bold indicates signi cant P-value.
The same results were obtained from the subgroup analysis of standard postoperative adjuvant therapy (Fig. 3, Table 5-6). The baselines of FIGO staging, histological type, tumour size, LVSI, cervical stromal invation depth and surgical margin were not consistent between the two groups ( Table 2). The OS was 95.6% in rst 50 patients with LRH group (n = 187) and 95.8% in last 50 patients with LRH group (n = 305) before standard treatment (see in Fig. 3). There was no signi cant difference in OS between the two groups (p = 0.189). Cox multivariate analysis (see in Table 5) showed that LRH surgical experience was not associated with OS in patients with early stage cervical cancer (p = 0.189). DFS was 90.2% in rst 50 patients with LRH group and 92.2% in last 50 patients with LRH group (see in Fig. 3). There was no signi cant difference in DFS between the two groups (p = 0.280). Cox multivariate analysis showed that LRH operation experience was not related to DFS in patients with early stage cervical cancer (p = 0.516). The similar result were obtained after matching (see in Table 2 and Table 6).  Multicollinearity test and cox proportional hazard regression models were used for analysis.
Proportional hazard assumption was tested and showed no interaction with time.
Bold indicates signi cant P-value. Multicollinearity test and cox proportional hazard regression models were used for analysis.
Proportional hazard assumption was tested and showed no interaction with time.
Bold indicates signi cant P-value.

Discussion
Our study found that long-term oncological outcomes in patients with early stage cervical cancer did not improve signi cantly with increased surgical experience for high LRH volume surgeons.
The inclusion and exclusion criteria are relatively loose in the general analysis, and the results could re ect the in uence of surgeons' LRH experience on survival outcomes of early stage cervical cancer patients. In matched cohort studies, the patients' baseline data were more uniform than those in the RWD, and mixed bias could be effectively controlled. However, the external validity of the results was poor, which meant that the ability to generalize the results to other groups of patients with cervical cancer was limited. 21,22 In this study, general analysis and matched cohort studies were consistent and mutually validated, which improved the reliability of the results. 23 This further con rmed that the surgical experience of LRH surgeons was not related to the oncological outcomes of the patients.
At present, there are no original articles to study the reasons for the worse laparoscopic oncologic outcome of women with cervical cancer from the perspective of the experience of surgeons, but after the publication of LACC study, some articles 9-12 discussed whether the experience of surgeons could be one of the reasons for the worse laparoscopic oncologic outcome of women with cervical cancer. However, previous studies analyzed fewer institutions, surgeons and fewer patients than this study. For example, Chong et al. 13 compared the earliest 50 patients with LRH of a surgeon in a single center with the next 50 patients with LRH in patients with locally advanced cervical cancer. With an increase in surgical experience, the operation time, hospitalization days, time to recover normal residual urine and blood transfusion decreased, the number of lymph nodes obtained increased, and the intraoperative and postoperative complications decreased signi cantly. For each group, the ve-year OS rates were 96% and 92%, and the ve-year DFS rates were 92% and 90%. The differences were not statistically signi cant, and the study did not provide a reason for the lack of signi cant differences in long-term oncological outcomes. However, our nding differ from another study of surgeon experience on prostate cancer.
Vickers et al. 14 conducted a retrospective cohort study of 4702 patients with prostate cancer treated by laparoscopy. A total of 29 surgeons from 7 hospitals participated in the study. It was found that the 5- year recurrence rate was 16.3%, 11.0% and 7.1% (p = 0.038) for patients treated by surgeons with 10, 250 and 750 laparoscopic surgeries, respectively, and that the increase in surgical experience was associated with the oncological outcomes of patients with prostate cancer.
Matsuo et al. 15 conducted a nationwide multicenter retrospective study on the RH surgical volume of early stage cervical cancer. A total of 5964 patients with stage IB1-IIB cervical cancer were enrolled from 2004 to 2008. It was found that RH surgical volume may be a prognostic factor for early stage cervical cancer. High surgical volume hospitals may be associated with a reduced risk for local recurrence and an improved survival rate. In this study, there were fewer than 32 RH operations in low volume hospitals, 32 to 104 RH operations in medium volume hospitals and more than 105 RH operations in high volume hospitals over a ve-year period. However, the number of surgeries in hospitals could not fully equal to the experience of the surgeons. High-volume hospitals also have doctors with less experience in surgery, and low-volume hospitals also have surgeons with a high surgical volume. The lack of information on surgical approaches in this study means that researchers cannot adjust for the confounding factors of surgical approaches to oncological outcomes. Recent studies such as the study by Ramirez et al. 2 suggest that LRH is associated with poor oncological outcomes in patients with early stage cervical cancer. This study addresses the lack of research on surgeons' experience and surgical approaches highlighted in Matsuo et al.. 15 Our study found that the oncological outcomes of patients with early stage cervical cancer was not affected by the LRH experience of surgeons. According to literatures on the learning curve of laparoscopic surgery for cervical cancer, surgeons need 20-50 surgeries to master laparoscopic surgery. 13 . Therefore, the standards adopted in this study are not completely consistent with Sedlis standard.
The similar survival outcome were observed in both general and subgroup analysis, before or after matching. It might be explained by the following reason. Availability of data and material The datasets used and/or analysed during the current study are available from the corresponding author upon reasonable request.

Competing interests
The authors declare that they have no competing interests to disclose.

Funding
The National Science and Technology Support Program of China (2014BAI05B03) The Natural Science Foundation of Guangdong Province(2015A030311024)