The Learning Curve in Vaginal Pelvic Reconstruction Surgery for Severe Pelvic Organ Prolapse: Analysis of the Cumulative Summation Test (CUSUM)

Women who underwent vaginal pelvic reconstructive surgery with or without mesh consecutively between 2004 and 2018 were retrospectively analyzed to determine the learning curve in vaginal pelvic reconstructive surgery. With cumulative summation (CUSUM) analysis of surgical failure and operation time, we assessed the learning curve of vaginal pelvic reconstructive surgery, including sacrospinous ligament xation, anterior colporrhaphy, posterior colporrhaphy, and optional vaginal hysterectomy with or without mesh placement. Two hundred and sixty-four women with stage III or IV pelvic organ prolapse underwent vaginal pelvic reconstructive surgery by surgeon A or B. The median follow-up time of 44 months ranged from 24 to 120 months. Surgical prociency was achieved in 32-54 vaginal pelvic reconstructive surgery procedures without mesh and 37-61 procedures in the same surgery with mesh. The surgical success rates for surgeons A and B were 82.2% and 94.1%, with median follow-up times of 60 and 33 months, respectively. The learning phase of vaginal pelvic reconstructive surgery in advanced pelvic organ prolapse in this institutional cohort required 54 and 61 procedures, respectively. A higher number of procedures were required for the learning curve of vaginal pelvic reconstructive surgery with mesh. Having crossed the boundary of prociency, the surgical success rate and operation time were improved. reconstructive surgery without mesh (SSLF, anterior and posterior colporrhaphy, optional hysterectomy) was the operation method of surgeon A. Vaginal reconstructive surgery with mesh was the operation method of surgeon B. Surgical failure was frequently encountered in the early cases of the series in both cohorts. In case of a failure, the graph falls by 0.855. In the chance of success, the graph rises by 0.145. The unacceptable recurrence threshold, the H1 line, is presented as a horizontal line at 2.709. The surgical prociency was stabilized after 54 cases in surgeon A's cohort. A persistent rate of success above the breakthrough H1 line indicates that prociency was achieved. The overall success rates for surgeons A and B were 82.2% and 94.1%, respectively, with median follow-up times of 60 and 33 months.


Introduction
Pelvic organ prolapse (POP) is a complex entity that comprises the anterior, posterior, and apical compartments, each requiring a separate stage of surgical treatment. Pelvic reconstructive surgery is divided into the repair of the different compartments, with or without mesh. Studies addressing the outcomes and complications of pelvic oor reconstructive surgery usually state that the surgical procedure was "performed by an experienced surgeon," yet little data are available to clarify how many surgeries are needed for surgical skill to be considered sophisticated enough to merit this designation. The American Urogynecology Society advocates that surgeons performing intricate pelvic oor reconstructive surgery should have adequate experience and training to manage the inherent complex anatomy and complications 1 .
The cumulative summation (CUSUM) test was originally designed for industrial quality control and later adopted to monitor the learning curve in medicine 2,3 .
The CUSUM test sequentially analyzes changes in the process under scrutiny, determining after each procedure whether the process is "in control" (performing at an acceptable level) or "out of control" (performing at an unacceptable level). The CUSUM curve can readily show the trends and outcomes of consecutive events in the gure. When applied to the concept of a learning curve in surgery, the CUSUM curve can be used to determine when a pro cient surgeon has crossed the ordinary level and entered into a more stable one 4 . Currently, a wide variety of procedures and operations in gynecology, from embryo transfer to robotic sacrocolpopexy, use CUSUM to analyze the pro ciency of doctors 5,6 .
Pelvic reconstructive surgeries can be divided into vaginal and abdominal approaches. Two studies calculated the learning curve of laparoscopic and robotic pelvic reconstructive procedures 6,7 . Very little data regarding the vagnal approach are available. The literature contains more studies of vaginal pelvic reconstructive surgeries, but those focusing on surgical pro ciency of the vaginal approach are lacking 8, 9 . In addition, pro ciency becomes more critical with the application of transvaginal mesh. The use of mesh reinforces pelvic structures and native tissue repairs but also raises issues of mesh-related complications 10,11 . One study analyzed 10,000 transvaginal mesh surgeries and concluded that the surgeon's experience and technique are critical to enhance the success rate and reduce complications 12 .
The objective of this study was to examine the learning curve of vaginal pelvic reconstructive surgery, including sacrospinous ligament xation and concomitant anterior and posterior colporrhaphy with or without mesh. The study could help build evidence-based training programs and predict obstacles they may encounter.

Results
Demographic data A total of 264 patients who underwent pelvic reconstruction surgery were analyzed. Among them, 162 patients received pelvic reconstruction without mesh by surgeon A, and 102 patients received pelvic reconstruction with mesh by surgeon B (Table 1). Surgeon A's patients were 65.6 ± 10.5 (mean ± SD) younger than surgeon B's patients (68.4 ± 11.2). Because of the surgeon's preference, patients who received pelvic reconstruction without mesh were signi cantly more likely to have concomitant and previous hysterectomy than those who received surgery with mesh (p< 0.05). Other parameters, including BMI, parity, menopausal status, hypertension, diabetes mellitus, and POP-Q stage, did not differ signi cantly.   Figure 3A) and that of surgeon B was case No. 37 ( Figure 3B). Corresponding to the CUSUM analysis of the surgical success rate, pelvic reconstruction with mesh requires a greater number of operations to achieve a stable surgical condition. There are more turning points in the graph of surgeon B ( Figure 3B) compared with surgeon A. These turning points re ect minor modi cations made during the pelvic reconstructive surgery, when it took more time to tailor and adjust the mesh position. surgeon B were 6% and 27%, respectively. A total of 20.5% of the complications in surgeon B's group were postoperative higher residual urine, de ned as more than 150 ml after voiding three times. These complications resolved spontaneously during hospitalization. The postoperative days of hospitalization are shown in Figure 4. The mean postoperative days of surgeon A's group was 3.28±0.57 days, and that of surgeon B's group was 4.48±2.62 days.

Characteristics of the surgeon A and B cohorts per 25-surgery tier
Combining the results of the CUSUM analysis of surgical success and operation time, surgical pro ciency can be achieved in 32-54 cases in pelvic reconstructive surgery without mesh. Pelvic reconstructive surgery with mesh was steady after 37-61 patients. Surgical success is the expectation of both the surgeon and patient. Pelvic reconstructive surgery tends to fail over time 18,19 . Our study analyzed three parameters: surgical recurrence, operation time, and hospital stay. We found that hospital stay was not a suitable parameter because the medical and insurance system highly in uences its length. Patients with high postoperative residual urine may be discharged with a Foley catheter or hospitalized with medication and observation. Additionally, the unit of day for hospital stay was relatively too large considering that our average postoperative stay was 3.28 days for surgeon A's group and 4.48 days for surgeon B's group. Surgical recurrence is an essential indicator for surgical success in large studies 10,19 . The operation time is a sensitive indicator for the evolution of surgical pro ciency [20][21][22] . Interestingly, operation time is an early indicator, while surgical recurrence is a later indicator for the learning curve. Surgical pro ciency was achieved in 32 patients by operation time and 54 by surgical success in pelvic reconstructive surgery without mesh. Pelvic reconstructive surgery with mesh was steady after 37 by operation time and 61 by surgical success.
The strength of our study is that we retrospectively analyzed two similar large cohorts of POP. To our knowledge, we are the rst study using CUSUM to evaluate surgical pro ciency in vaginal pelvic reconstructive surgeries. In addition, we considered two parameters, surgical recurrence and operation time, to approach the speci c procedures necessary for maturation.
Not taking into account the common parameters of surgical complications is a limitation of our study. There are two reasons we were unable to factor it into our analysis. First, some complications, such as mesh erosions, may need a more extended follow-up period to be revealed. An adequate period of follow-up for long-term or rare complications is hard to de ne. Second, we gathered many more in-hospital than out-of-hospital complications. The medical records might have had missing values in the retrospective cohort. Logically, the CUSUM learning curve analysis of in-hospital complication rates would nd fewer procedures required than analysis of surgical recurrence found and more procedures than operation time did. The other limitation is that hysterectomy volumes fell in Taiwan during our follow-up period 23 . This caused a signi cant difference in the rates of previous and concurrent hysterectomy in surgeon A's cohort. Because surgeon A's cohort usually included an optional vaginal hysterectomy, this result may have negatively impacted the CUSUM analysis of operation time.
To our knowledge, we are the rst study using CUSUM to analyze vaginal pelvic reconstructive surgery with or without mesh, comprising SSLF and anterior and posterior colporrhaphy. Two approaches of pelvic reconstructive surgeries, abdominal and vaginal, are mainly utilized currently. Most studies have focused on laparoscopic or robotic sacrocolpopexy and pectopexy, showing that pro ciency can be attained in approximately 78 cases 6,7 . Nevertheless, few studies have calculated the learning curve of the vaginal route. De Tayrac analyzed the learning curve of bilateral anterior sacrospinous ligament suspension associated with anterior mesh repair but was based on the number of adverse events, which may not be as sensitive as CUSUM 20 . In our series, we observed that surgical pro ciency was achieved after performing fewer vaginal procedures with or without mesh than were needed for sacrocolpopexy in other studies.
Vaginal pelvic reconstructive surgery for multiple-compartment pelvic organ prolapse is complicated. The CUSUM test was applied to nd the learning curve of surgical performance when surgical outcome, operation time, and hospitalized days reached a steady state where cumulative success rates consistently remained above the acceptable boundary line of the CUSUM analysis. Combining the results of CUSUM analysis of surgical recurrence and operation time, surgical pro ciency can be achieved in 32-54 cases in reconstructive surgery (SSLF, anterior and posterior colporrhaphy) without mesh, and reconstructive surgery with mesh is steady after 37-61 patients. Our data showed that cumulative sum control chart analysis can assist in the training program of urogynecologic fellows. The trainees are able to visualize their performance as they progress toward surgical pro ciency.

Methods
Our study retrospectively analyzed two surgeons (Surgeons A and B) who performed multicompartment repairs of advanced pelvic organ prolapse. The data were collected at a single center longitudinally between 2004 and 2018. Local institutional review board approval was obtained (IRB/REC 202107084RIN). Written informed patient consent was obtained. All methods in this study were performed in accordance with the relevant guidelines and regulations. Patient medical records were retrieved, including patient age, body mass index (BMI), past medical illness, surgical history, surgical method, operation time, hospital stay, complications, and outcomes. The postoperative pain scores were recorded during the period of admission, and voiding function was evaluated. Patients in stable recovery were discharged to an outpatient department for follow-up. The minimal follow-up period was 24 months. Patients were all pelvic organ prolapse quanti cation (POP-Q) stage III or IV. Those who planned to undergo pelvic reconstructive surgeries were divided into two groups based on the performing surgeon (A or B). Recurrent surgeries were not included in our study.
The primary outcome measure was surgical pro ciency, which was based on anatomic success. Anatomic failure was de ned as the objective recurrence of POP-Q stage II or above during follow-up. Surgical pro ciency was de ned as the point at which the CUSUM score rises above the acceptable boundary line H1 of the CUSUM analysis and remains there. Secondary outcomes were stabilization of operation time and the period of postoperative hospitalization.
The CUSUM results for anatomic success were recorded on a graph in which the x-axis represents the number of procedures, and the y-axis represents the "cumulative sum" of successes (s) and failures (1-s). With each anatomic success, the graph rises by "s"; with each anatomic failure, the graph falls by "1-s" (Supplement Table 1). When the proportion of anatomic successes to failures is su ciently high, the CUSUM score rises above the boundary line, H1. The boundary line is set according to the acceptable failure rate, P0, and the unacceptable failure rate, P1. In our study, P0 and P1 were selected to be 10% and 20% according to the previous literature 13,24,25 . When there was no recurrence, the CUSUM graph increased by s = = 0.145, where P = and Q = . When a recurrence arose, the graph fell by 1-s = 0.855. Type 1 error ( ) and type 2 error ( ) represent the probability of falsely de ning the surgeon's performance as "acceptable" or "unacceptable", respectively. Type 1 and 2 errors of 10% were considered acceptable in this study. Pro ciency was obtained when the graph crossed above H1 and remained there. It is assumed that the surgeon's performance matured with a false positive rate of .
CUSUM was also applied to nd the operation time learning curve. CUSUM analysis was used to measure the deviation between the raw data of each individual case and the mean value of the cohort, tracking the accumulation of each deviation in a sequential manner. Thus, CUSUM was de ned as , where is the operation time in each case and is the mean operation time of the cohort. By this method, the CUSUM curve portrays trends in data that are not discernable with other approaches 26 .
Surgeon A performed multicompartment repairs with natural tissue repair (NTR) by unilateral sacrospinous ligament xation (SSLF) and concomitant vaginal anterior and posterior colporrhaphy. Surgeon A did not use mesh in the surgeries. Vaginal hysterectomy is an optional surgery but is usually performed. We preferred to perform SSLF with hysterectomy to have a better surgical eld but preserved the uterus with mesh to avoid complications. SSLF surgery was performed with a Veronikis ligature carrier and Miya hook as previously described in detail [27][28][29] .

Declarations Acknowledgment
We thank the staff of Biotechnology R&D Center, National Taiwan University Hospital, Hsin-Chu Branch, for their assistance in statistical analysis.

Authorship contributions
Chin-Jui Wu designed the concept, analyzed the data, and wrote the manuscript.
Kuan-Ju Huang acquired data and joined the discussion.
Wen-Chun Chang shared her experience and supplied data.
Ying-Xuan Li helped to collect data.
Lin-Hung Wei interpreted the data and discussed and approved the nal draft.
Bor-Ching Sheu designed the concept, discussed the analysis, and served as the corresponding author.

Disclosure of Interests
The authors declare no con ict of interest.

Details of Ethics Approval
The local institutional review board (202107084RIND) was approved by the Research Ethics Committee O ce of National Taiwan University Hospital (https://www.ntuh.gov.tw/RECO/Index.action).

Funding
There was no funding.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.