Baseline characteristics
A total of 1118 patients underwent elective PME or TME for primary rectal cancer between February 1991 and December 2020. A total of 942 of these patients were enrolled in this study. Seventy-nine patients were excluded due to metastatic cancer, 36 patients were excluded due to discontinuity resection, 10 patients were excluded due to R1 resection, and 51 patients were excluded due to death in the first 30 days after the operation (Fig. 1).
The mean follow-up time was similar between the leak group and the noleak group (69.9±47.9 months vs. 72.3±48.6 months, p=0.642).
AL occurred in 141 patients (15.0%). Most of these patients underwent TME (n=118, 83.7%) with stoma formation (n=99, 70.2%). According to the ISREC classification [12], 23 patients had grade A AL (16.3%), 48 patients had grade B AL (34.0%), and 70 patients had grade C AL (49.6%). Two patients died due to AL in the first 30 postoperative days. The AL grades were A and C. Both patients initially underwent TME with protective ileostomy.
Patient baseline characteristics and pathological outcome parameters are summarized in Table 1. The leak group had significantly more obese (BMI ≥ 30 kg/m2, n=32 (22.7%) vs. n=97 (12.1%)) and underweight patients (BMI < 18 kg/m2, n=4 (2.8%) vs. n=16 (2.0%), p=0.003) than the noleak group. Sutured coloanal anastomoses (n=16 (11.3%) vs. n=31 (3.9%), p=0.001) were significantly more common in the leak group than in the noleak group.
Logistic regression identified a higher BMI as a significant predictor of AL according to univariate (p=0.005), multivariate (p=0.007), and stepwise analyses (p=0.005). The lack of neoadjuvant chemotherapy was a predictor of AL only in the stepwise analysis (p=0.031) (Table 2).
According to univariate (p=0.004), multivariate (p=0.006), and stepwise (p=0.005) analyses, the cumulative link model ordinal regression also revealed that BMI was a significant predictor of AL severity according to the ISREC classification (Table 3).
Cancer-specific survival (CSS)
Univariate and multivariate analyses (n=942)
Univariate analysis revealed similar 5-year CSS rates in the leak group (90.7% (95% CI: 84.9-96.9%)) and the noleak group (92.6% (95% CI: 90.5-94.8%)) (HR for leakage = 1.34 (95% CI: 0.72-2.50), p=0.371). The severity of AL according to the ISREC classification did not significantly affect the 5-year CSS rate (p=0.642) for patients with Grade A AL (94.7% [95% CI: 85.2-100.0%]), Grade B AL (84.4% [95% CI: 73.5-97.0%]), or Grade C AL (94.2% [95% CI: 86.3-100.0%]).
When the patients were divided into a group with nonoperative treated leak (ISREC grade A and B; 87.9% (95% CI: 79.8-96.9)) and another group (noleak and ISREC grade C; 92.7% (95% CI: 90.7-94.8)), no significant difference in 5-year CSS rates was found (HR = 1.55 (95% CI: 0.74-3.23), p=0.272).
The univariate and unadjusted survival curves for CSS are shown in Fig. 2.
According to Cox regression analyses, neither AL (multivariate p=0.548; stepwise p=n/a) nor its subclasses according to the ISREC classification (multivariate p=0.854; stepwise p=n/a) had a significant negative effect on CSS.
Cox regression analyses indicated that older age, lower BMI, ASA classification III/IV, higher UICC stage, year of operation (2011–2020), and lower tumour height had a significant negative impact on CSS (Table 4).
Propensity score analyses (n=936)
The risk factors were evenly distributed after propensity score matching in the two groups. A well-matched sample of both groups (leak, n=141; noleak, n=795; not matched and excluded n=6 (all from the noleak group)) was generated by propensity score matching (leak 0.151±0.054 vs. noleak 0.151±0.053, p=0.958). There was no significant difference in the 5-year CSS between the leak group (87.9% [95% CI: 80.8–95.7]) and the noleak group (92.2% [95% CI: 90.0–94.6]) after propensity score matching (HR = 1.14 [95% CI: 0.56–2.34], p=0.161) (Fig. 3).
Secondary outcomes
Univariate and multivariate analyses (n=942)
AL had a significant negative influence on 5-year OS (leak 71.4% (95% CI: 63.6-80.1) vs. noleak 78.5% (95% CI: 75.4-81.8); HR = 1.53 (95% CI: 1.11-2.10), p=0.013) and DFS (leak 63.2% (95% CI: 55.0-72.5) vs. noleak 72.8% (95% CI: 69.5-76.3); HR = 1.53 (95% CI: 1.13-2.06), p=0.008) but not on RFR (leak 91.9% (95% CI: 86.9-97.2) vs. noleak 94.3% (95% CI: 92.4-96.1); HR = 1.54 (95% CI: 0.74-3.19), p=0.266). This was confirmed by Cox regression analyses of OS (multivariate: p=0.014, stepwise: p=0.015), DFS (multivariate: p=0.012, stepwise: p=0.025), and RFR (multivariate: p=0.292, stepwise: p=n/a).
Analysis of the severity of AL according to the ISREC classification revealed a significant influence on 5-year OS (p=0.017) and DFS (p=0.014) but not on RFR (p=0.656). Multivariate and stepwise analyses revealed a significant effect on DFS (p=0.039 and p=0.044), but no significant effect on OS (p=0.075 and p=0.082) and RFR (p=0.655 and p=n/a).
The analysis of the division of the patients into a group with nonoperative treated leak (ISREC grade A and B) and another group (noleak and ISREC grade C) revealed no significant differences in 5-year OS (p=0.468), DFS (p=0.495), or RFR (p=0.216).
Cox regression analyses indicated that older age, lower BMI, ASA classification III/IV, higher UICC stage, year of operation (2011–2020), and lower tumour height had significant negative impacts on OS, DFS, and/or RFR (Table 5, Table 6, Table 7).
Propensity score analyses (n=936)
After propensity score matching and analysis of survival as described above, AL had a significant negative influence on OS (HR = 1.67 (95% CI: 1.17-2.39), p=0.005), DFS (HR = 1.57 (95% CI: 1.12-2.20), p=0.011), and RFR (HR = 2.03 (95% CI: 0.99-4.17), p=0.047) (Fig. 4).