In the second half of 2018, our group innovatively introduced the TLAP technique into ileostomy reversal. Since then, TLAP has been performed in >10 procedures/year by the same surgery team. From October 2018 to October 2021, a total of 65 consecutive patients were retrospectively enrolled. All patients with a history of laparoscopic colorectal cancer surgery underwent TLAP at the National Cancer Center/National Clinical Research Center for Cancer/Cancer Hospital, Chinese Academy of Medical Sciences and Peking Union Medical College.

In this study, any patients suited to undergo classic open reversal were regarded as potential candidates for TLAP. Eligible patients were those ≥18 years of age who received TLAP ≥3 months after former colorectal surgery or 8 weeks after postoperative chemotherapy/radiotherapy. Moreover, study participants also underwent both a colonoscope examination and enhanced computed tomography imaging of the thoracic, abdominal, and pelvic cavities to guarantee acceptable anastomotic stoma healing and exclude tumor recurrence or metastasis. Patients who underwent TLAP combined with additional procedures, such as additional intestine resection, anastomotic reconstruction, or parastomal hernia/abdominal wall repair, and those with other surgical contradictions for the traditional open ileostomy reversal were excluded from subsequent analysis. This study was in accordance with the Declaration of Helsinki and the informed contents were signed before the TLAP surgery. This research was also approved by the Ethical Committee of the Cancer Hospital (Institute), Chinese Academy of Medical Sciences, Beijing, People's Republic of China.

The surgical team included a single experienced surgeon and 2 constant assistants throughout the study period. The participating senior surgeon has trained as an oncology surgeon for 15 years with extensive laparoscopic colorectal surgery experience (>100 procedures/year since 2015). Besides, TLAP was necessarily aided by a first assistant surgeon and a laparoscope holder. Both assistants were surgery residents and had completed 3 years of standardized training of residency after 2017. The primary duties of the first assistant surgeon included retraction and suction when necessary. All team members understood the details of TLAP technique, supported each other with effective methods, and would infuse their experience into their future performance.

After general anesthesia, patients were placed in a supine lithotomy position and the previous stoma was closed in a one-layer continuous Lembert pattern wherein the needle exited the tissue within 1 mm of the stoma edge and engaged the submucosa with each bite. Here, a 4-port technique was employed for trocar placement (Figure 1). First, a 10-mm trocar was inserted upon the umbilicus as an observation port. Then, a 12-mm supraumbilical port was placed at the left anterior axillary line as a principal operating port. Next, a 5-mm operating port located at the left lower-quadrant McBurney's point was used for auxiliary operating. Another 5-mm port for assistant was located in the right anterior axillary line 10 cm superior to the stoma.

After the establishment of pneumoperitoneum, the lysis of adhesions around the stoma and dissection of the mesenteries were performed using an ultrasonic scalpel. A 60-mm endoscopic linear stapler (Johnson ECR60B) was subsequently used to transect the proximal and distal ileum for digestive tract reconstruction. First, a pair of 1-cm incisions located at the anti-mesenteric side of the proximal and distal intestines were made, respectively, and a side-to-side anastomosis was created with a 60-mm endoscopic linear cutter stapler (Johnson ECR60B). Then, the common opening of both intestines was closed by another linear cutter stapler and the mesenteric defect was closed by absorbable sutures routinely. Following examination of the anastomotic blood supply, the stoma remnant was removed and the incision was sutured conventionally.

In this study, operation time (OT) was regarded as a reflection of surgical competency. To explore the association between surgeon experience and OT, CUSUM and moving average analyses were completed. CUSUM analysis is an analytic technique employed in surgical procedures for the quantitative estimation and visualization of the learning curve (14). Briefly, the CUSUM is the total accumulated value of differences in OT between each data point and the mean OT of all data points. In the CUSUM analysis, all 65 cases were ordered chronologically from the earliest to the latest date of TLAP. For the first patient, the CUSUM_OT was the difference between the OT for the first patient and the mean OT for all cases. Similarly, the CUSUM_OT for the second patient was the difference between the second OT and the mean OT of all cases plus the CUSUM_OT for the first patient (15). This recursive process continued until the 65th patients was treated, and the results of all CUSUM_OT analyses were plotted graphically thereafter, revealing the trend of deviation from the mean OT. Of note, the inflection points indicated at each set of ≥3 consecutive negative values were used to divide patients into separate phases. A linear regression model was then fitted to match the CUSUM curve. In addition, we also used a moving average of 5 to eliminate individual variations and highlight the long-term trends of OT (16). Specifically, the moving average of the i cases was the mean value from the i cases to the i + 4 cases (17).

To analyze the learning curve from multiple dimensions, we designated each case as a success or failure. Conventionally, surgical failure was defined as conversion to open surgery. However, since there was no instance of conversion to open surgery, surgical failure was defined as any intraoperative or postoperative complications according to a previous report (18). Similar to the OT analysis, the CUSUM of complications was displayed graphically and showed the cumulative total of a mixture of increments with each surgery failure and decrements with each surgery success (19). Univariate and multivariate logistic analyses were then developed based on baseline variables (gender, age, body mass index et al.) to evaluate the potential confounders on surgical failure, respectively.

Furthermore, RA-CUSUM analysis was applied to depict the success or failure of the TLAP technique. First, baseline variables with P < 0.20 in the univariate association were considered for inclusion, and the predicted probabilities of each case were calculated according to the regression coefficients of the variables in the final multivariate regression model (20). Then, for each failure case, the RA-CUSUM value was incremented by (1-predicted probability of failure). In contrast, for each success, the value was decreased by the predicted probability of failure (21). Patients were again grouped into distinct phases according to the inflection points.

The demographic and baseline variables included gender, age, body mass index (BMI), American Society of Anesthesiologists (ASA) score, duration after previous laparoscopic colorectal cancer surgery and comorbidities. Perioperative results included operation time, estimated blood loss, length of incision, time to ground activities and flatus passage, postoperative hospitalization, and perioperative complications. Estimated blood loss was the sum of the blood in the suction canister (the total volume after subtracting the amount of irrigation fluid) and the segment of increased weight of swabs during operation phase (1 ml of blood is about weighs 1 g) according to previous randomized controlled trails (22, 23). The time to ground activities and flatus was reported by patient. Postoperative hospitalization was defined as the number of nights from TALP to discharge. Perioperative complications were calculated within 30 days of surgery.

The SPSS version 26.0 software program (SPSS Inc., Chicago, IL, USA) and Microsoft Office Excel were used for statistical analysis and data visualization, respectively. For quantitative variables with normal distribution as determined by the Shapiro-Wilk test, data are presented as mean ± standard derivation (SD) values and compared by One-way analysis of variance followed by Bonferroni's test. In contrast, data with skewed distribution are presented using median and interquartile range (IQR) values and compared by the Kruskal-Wallis test. For categorical variables, data are presented using numbers and percentages, and the chi-squared test or Fisher's exact test was applied to reveal group discrepancy. Polynomial regression models were selected according to a best-fitted model. A P value < 0.05 was considered to indicate a significant difference in all tests.

From 2018 to 2021, a total of 65 consecutive patients who underwent TLAP were enrolled in this study. The overall perioperative data are presented in Table 1. There were 43 male and 22 female patients treated with this innovative technique with a median age of 63 years. The mean BMI of the TLAP patients was 23.46 kg/m². Most patients (87.69%) were classified as ASA class I or II cases, and the previous laparoscopic colorectal cancer surgery occurred a median of 9 months ago. Among comorbidities, hypertension was most common (21.54%), followed by diabetes mellitus, affecting 13.85% of enrolled patients. Other comorbidities included hyperthyroidism, coronary disease, and renal insufficiency.

Intraoperative and postoperative data are also presented in Table 1. We found that the median operation time was 94 min, which was adopted as a crucial indicator for subsequent learning curve analyses. The estimated blood loss ranged from 10 ml–100 ml, with a median of 30 ml. The median incision length was 6 cm. Of note, in this study, the first ground activities (median = 1), first flatus passage (median = 2), and number of postoperative hospitalization days (median = 4) were used as reflections of postoperative recovery. Any complication during or after surgery was also estimated to assess the safety of the TLAP technique. In our series, a total of 7 patients suffered from intraoperative/postoperative complications (trocar site bleeding, n = 1; pyrexia, n = 4; incisional infection, n = 2).

The raw operation time was plotted according to chronological case order and exhibited a tendency of steady reduction with the best-fitted logarithmic model [y = −21.44 ln(x) + 165.13, R²= 0.7425, P < 0.001], indicating a complex non-liner relationship between the OT and surgeon experience (Figure 2). CUSUM analysis was subsequently applied, and the mean operation time (96 min) was used as a critical reference. As shown in Figure 3A, the CUSUM of OT was best modeled as a third-order polynomial (y = 0.005x³–0.907x² + 34.673x + 99.112, R² = 0.9604, P < 0.001), which showed a gradual upward slope until the 24th case, followed by small fluctuations between the 25th and 39th cases and a subsequent steep downward trend after the 39th case. Similarly, we determined that the OT decreased significantly after the 20th case and reached a steady state after the 36th case after fitting a logarithmic model of y = −18.56 ln(x) + 153.43 (R² = 0.8806, P < 0.001) in a moving average curve (Figure 3B).

Based on the learning curve of the CUSUM of OT, we were able to separate the learning curve into the following 3 phases: phase I (an initial phase, including cases 1–24), phase II (a transition phase, including cases 25–39), and phase III (the proficient phase, including cases 40–65). Best-fitted lines for each phase were also acquired (Figure 4). The positive slope in phase I indicated a longer OT during the initial learning phase (R² = 0.8026, P < 0.001). However, a flat slope in phase II (R² = 0.3246, P = 0.027) revealed an increased degree of surgery competency in the transition phase. More importantly, the negative slope seen in phase III (R² = 0.9879, P < 0.001) confirmed the proficiency of the TLAP technique.

The interphase comparisons of patient characteristics are presented in Table 2. With regard to demographics, no statistical difference was found in gender (P = 0.742), age (P = 0.863), BMI (P = 0.067), ASA (P = 0.891), duration of ileostomy (P = 0.239), postoperative adjuvant therapy history (P = 0.535), or comorbidities (P = 0.187) among the initial, transition, and proficiency phases. Most notably, our results revealed that the OT between each phase was significantly different (108.5 min vs. 92 min vs. 80 min, P < 0.001). Phase I had the longest OT; meanwhile, a significant difference in OT was also revealed between phase II and III (P = 0.016). We additionally observed a significant downtrend in the number of postoperative hospitalization days (4 vs. 5 vs. 3 days, P < 0.001). In contrast, there was no significant difference in estimated blood loss (P = 0.988), surgical incision length (P = 0.798), time of first ground activities (P = 0.143), or time of first flatus passage (P = 0.663). Rates of intraoperative and postoperative complications between the 3 phases were not significantly different either (3 vs. 2 vs. 2, P = 0.778).

To analyze the relationship between surgery experience and surgery success, CUSUM analysis was also performed. The CUSUM result based on intraoperative/postoperative complications showed a small fluctuation without a significant change between the zero line until approximately the 29th case, followed by an upward slope until the 32nd case and a subsequent downward slope thereafter (Figure 5A). To adjust for potential confounding effects of baseline covariables, univariable and multivariable logistic regression analyses were conducted (Table 3). The univariable analyses indicated that the gender (OR = 0.038, CI: 0.068–1.667; P = 0.183) and BMI (OR = 1.567, CI: 1.070–2.296; P = 0.021) were associated with a potential increased risk of surgery failure, with which age (P = 0.230), ASA score (P = 0.999), duration of ileostomy (P = 0.891), postoperative adjuvant therapy history (P = 0.722), and comorbidities (P = 0.540) were not significantly correlated.

In the multivariable analysis, BMI was the only factor independently associated with perioperative complications (OR = 1.538, CI: 1.044–2.265; P = 0.029). A further RA-CUSUM analysis was conducted based on the predicted odds ratio (Figure 5B); similar to the results of the CUSUM analysis, a small fluctuation was observed until the 22nd case and a downward tendency occurred after the 32nd case, suggesting an acceptable range with regard to perioperative complications during the learning period.