Risk Factors for Anastomotic Leakage in Advanced Ovarian Cancer Surgery: A Large Single-Center Experience

The enrolled population included all patients with histological diagnosis of epithelial ovarian, fallopian, or peritoneal cancer (FIGO stage IIIB–IVB) who underwent rectosigmoid resection and anastomosis with curative intent. All patients received preoperative mechanical bowel preparation.

Patients with end-colostomy or end-ileostomy were excluded from the study.

Several system scores, clinical and surgical variables, helpful in predicting operative and postoperative risk, were used to classify patients’ risk.

Patients with American Society of Anesthesiologists (ASA) score > 2, Eastern Cooperative Oncology Group-Performance Status (ECOG-PS) ≥ 2, and Age-Adjusted Charlson Comorbidity Index (ACCI) score > 2 were considered at high risk of postoperative complications.

Preoperative albumin level below 30 mg/dL and preoperative hemoglobin values below 10.0 g/dL were indicative respectively of severely poor nutritional status and moderate to severe anemia.

Nutritional status was also assessed by body mass index (BMI) classification (divided into the following categories: underweight patients with BMI < 18 kg/m², normal weight–overweight with BMI 18–30 kg/m², and obese patients with BMI ≥ 30 kg/m²).

CA-125 value and Predictive Index Value (PIV) at initial diagnosis were considered as potential indicators of tumor burden. To identify the disease with the greatest tumor burden, the cutoffs were set to CA-125 ≥ 1000 U/mL and PIV ≥ 6.

A cutoff of 60 years was used, based on the median age of the study population. Suspicion of AL, suggested by clinical signs such as abdominal pain or distension, leukocytosis, fever, presence of gas, pus, or feces in the drains, the abdominal incision, or vagina was confirmed by computed tomography (CT) scan with rectal contrast enema and/or immediate relaparotomy.

AL was defined as communication between the intra- and extraluminal compartments due to a defect in the integrity of the intestinal wall originating from the staple line of the neorectal reservoir between the colon and rectum. Pelvic abscess adjacent to the anastomosis were also categorized as anastomotic leakage even if no communication could be demonstrated with the colonic lumen at the anastomosis.²¹

The Common Terminology Criteria for Adverse Events v3.0 (CTCAE) was used to classify intraoperative complications (CTCAE 0–1 versus ≥ 2). The Extended Clavien–Dindo classification of surgical complications was used for the grading definition of early complications.²²

OS was calculated from date of primary diagnosis to date of death or last follow-up.

All colorectal resections and anastomoses were performed by a dedicated team of general surgeon who routinely collaborate with our department. Regarding the resection technique, when possible, we tended to perform low ligation of the lower mesenteric artery, thus preserving the left colic artery. In some cases, selective ligation of the sigmoid arteries was performed, preserving the superior rectal artery to improve the vascularization of the rectal stump. In case of doubt regarding intestinal vascularity, we used a coarse evaluation of the pulsatile flow and/or a marginal artery section to the proximal colon.

The mobilization of the colon for the execution of tension-free anastomosis involved the following steps: The first surgical maneuver was mobilization of the splenic flexure through the development of the avascular plane between the Gerota’s fascia and the left Toldt’s fascia. Second, the mobilization continued by dissecting the pancreatic–colic ligament. The ligation and section of the mesenteric vessels (first sacrificing the inferior mesenteric vein and/or the sigmoid vessels up to the ligation and sectioning of the inferior mesenteric artery) were maneuvers considered when the tension at the level of the anastomosis was still considered excessive by the surgeon. Colorectal anastomosis was performed using a circular stapling device for either end-to-end or side-to-end anastomosis. An air leakage test was performed immediately after completion of anastomosis. In case of intraoperative leakage, additional overstitches were placed.

In rare cases where the tissues were damaged or excessive leakage was found at the intraoperative air leakage test, the anastomosis was repackaged.

The level of the anastomosis was considered medium–low if the distance from the anal verge was less than 10 cm as measured by the rigid probe. The cutoff was set at 10 cm as this was considered and assumed to be the average length of the “extraperitoneal” rectum (medium–low rectum).

Diverting ostomy was performed based on the clinical status of the patient before surgery (compromised clinical condition, judged mainly by ASA and EOCG scores), the surgical procedures performed (multiple bowel resection), specific characteristics of the anastomotic complex (such as the level of anastomosis, the quality of the tissues, and their vascularization), and in general the surgical complexity (also judged by operative time and need for intraoperative transfusions).

Considering that the leakage rate for OC ranges between 2.7% and 16.9%,^7,10‐19 a sample size of N = 515 patients was calculated to be appropriate to detect a 16.9% AL (conservative approach), with α = 0.05 and a margin of error of 3.23%.

Descriptive statistics were used to describe the surgical procedures, patients, and pathological characteristics.

Absolute frequency and percentage were adopted for qualitative variables, or median and interquartile range (IQR) for continuous variables. The normality of data was verified via the Kolmogorov–Smirnov test. Groups were compared using the Mann–Whitney U test for continuous variables and the Pearson χ² test or Fisher exact test for categorical variables.

Binomial logistic regression was performed to ascertain the effects of independent variables, judged as possible risk factors by clinicians, on the occurrence of AL and on the risk of diverting ostomy.

Survival analysis was performed using the Kaplan–Meier method and the Cox regression models.

Statistical analyses were performed using SPSS version 24.0 (IBM, Armonk, New York, NY) software for Windows.

Five-hundred fifteen (515) patients [372 (72.2%) PDS and 143 (27.8%) IDS] were identified in the study period.

Clinical features of the study population are presented in Table 1.

The study population was almost equally divided between women under the age of 60 years and (53.8%) and older women (46.2%). Of the study population, 12.6% was obese (BMI ≥ 30 kg/m²) while 4.9% was underweight (BMI < 18 kg/m²). Forty-six patients (10.8%) had preoperative serum albumin values below 30 mg/dl, while 52 (10.1%) had preoperative hemoglobin values below 10.0 g/dl. Almost all patients had an American Society of Anesthesiologists (ASA) score of 2 or less (95.7%), and an Eastern Cooperative Oncology Group-Performance Status (ECOG-PS) equal to or less than 1 (96.1%).

Surgical characteristics of the studied population are presented in Table 2.

Optimal debulking was achieved in 95.9% of the enlisted population (CGR, 79.6%; RT (residual tumor) ≤ 1, 16.3%) while RT > 1 cm was left in the remaining population (4.1%). The Surgical Complexity Score (SCS) was generally elevated, with 69.9% of patients belonging to group 3 (high complexity score group).²³ Regarding the specific characteristics of the rectosigmoid resection, high resection was usually performed (75.1%), with left colic artery sparing in 75.1% of cases. Based on patients’ clinical characteristics, surgical complexity, and the surgeon’s decision, protective ostomy (ileostomy or colostomy) was performed in 230 patients (44.7% of cases). Twenty-one patients (4.1%) had intraoperative complications of grade 2–4. Estimated blood loss (EBL) above 500 ml was recorded in 61.4% of cases, while intraoperative transfusions were required in 24.9% of patients.

Postoperative features are described in Table 3.

The median hospital stay was 8 days, with 100 patients (19.4%) suffering severe postoperative complications (grade III–V). Fourteen patients (2.7%) died within 90 days from surgery.

Median time to chemotherapy was 39 days, but 17 patients (3.3%) could not start or resume chemotherapy due to postoperative complications (deterioration of physical condition or death).

Of the 230 patients receiving protective ostomy, 153 (66.5%) underwent ostomy reversal, with a median time of 7 months. The remaining 77 had no reversal due to disease recurrence or refusal to undergo further surgery.

The registered AL rate was 2.9% (15 out of 515 patients) (Table 3).

Diagnosis of AL was generally made on postoperative day 5 (interquartile range (IQR): 4–10.5 days).

Seven of the 15 patients with AL (46.7%) had diverting ostomy during first surgery; 3 could be treated conservatively with drainage plus broad-spectrum antibiotic therapy, while the rest underwent reoperation with resection of the anastomotic complex and colostomy according to Hartmann procedures. All eight patients with AL and without diverting ostomy required surgical reintervention (six were treated with the Hartmann procedure, and two received ileostomy) (Supplementary Table S1). The following variables showed a statistical significant association with AL on univariate analysis (Table 4): age at surgery ≥ 60 years (odds ratio (OR): 3.307, 95% confidence interval (CI): 1.039–10.527, p = 0.043), body mass index (BMI) < 18 kg/m² (OR: 16.461, 95% CI: 5.191–52.196, p < 0.001), preoperative albumin value < 30 mg/dL (OR: 5.671, 95% CI: 1.772–18.143, p = 0.003), pelvic lymph node resection (OR: 4.269, 95% CI: 1.297–14.051, p = 0.017), section at the origin of the inferior mesenteric artery (IMA) (OR: 9.002, 95% CI: 2.814–28.801, p < 0.001), hypogastric vessels section (OR: 11.758, 95% CI: 3.354–41.220, p < 0.001), distance of the anastomosis from the anal verge < 10 cm (OR: 21.761, 95% CI: 4.840–97.838, p < 0.001), intraoperative transfusions (OR: 3.619, 95% CI: 1.286–10.187, p = 0.015), and postoperative anemia (OR: 5.132, 95% CI: 1.431–18.412, p = 0.012).

Among the variables included in multivariate logistic regression analysis, only the following were identified as independent predictors of AL: BMI < 18 kg/m² (OR: 19.621, 95% CI: 3.394–113.447, p = 0.001), preoperative albumin value < 30 mg/dL (OR: 20.639, 95% CI: 2.345–181.669, p = 0.006), section at the origin of the IMA (OR: 10.732, 95% CI: 1.862–61.846, p = 0.008), and distance of the anastomosis from the anal verge < 10 cm (OR: 14.673, 95% CI: 2.340–92.006, p = 0.004).

Supplementary Table S2 reports several pre- and intraoperative risk factors for diverting ostomy. Among all the analyzed variables, only the following showed increased risk on multivariate analysis: more than one bowel resection (OR: 5.412, 95% CI: 3.097–9.456, p < 0.001), medium–low anastomosis at less than 10 cm from the anal verge (OR: 2.414, 95% CI: 1.446–4.030, p = 0.001), operative time longer than 300 min (OR: 1.680, 95% CI: 1.023–2.758, p = 0.040), and need for intraoperative transfusions (OR: 1.688, , 95% CI: 1.031–2.763, p = 0.037).

Table 5 compares risk factors for diverting ostomy and anastomotic leak.

Regarding the early postoperative course (Table 6), a significant difference among patients with and without AL was noted in length of hospital stay (8 versus 23.5 days, p = 0.001) and time to chemotherapy (38 versus 49.5 days, p = 0.013). Three patients with AL (20%) died from multiorgan failure (MOF) (Table 6) within 90 days versus 2.2% in women without AL (2.2% versus 20%, p = 0.001).

Supplementary Table S3 reports ostomy-related complications. We detected an overall rate of ostomy-related complications of 33.9% (78 out of 230 patients), of which only 16 were grade ≥ 3 (7%).

The most frequently reported complication was dehydration (23.9%), with most patients requiring i.v. fluid therapy (22.2%), and four patients (1.7%) underwent early ostomy reversal.

Forty-five patients (19.6%) were admitted to the emergency department due to ostomy-related complications, but as manty as 87 patients (37.8%) reported difficulties in ostomy management and impaired quality of life.

AL was confirmed as a rare postoperative complication in OC surgery, with only 15 reported cases in the entire population of 515 patients (2.9%).

Of the ALs, 80% were classified as “severe” since they required reintervention, and three patients died during post-operative course, leading to a significant increase in early postoperative mortality rate. Generally, patients with AL had statistically longer hospital stay and prolonged time to start of chemotherapy.

Concerning variables that could increase the AL risk, the state of cachexia (demonstrated by low BMI and low preoperative albumin value) and specific characteristics of the colorectal anastomosis (ligation and section at the origin of the IMA and “mid–low level” anastomosis) resulted as relevant risk factors.

Several studies have shown how the consequences of AL can have a profound and negative impact on patients’ postoperative course, resulting in prolonged hospital stay, reduced probability of starting chemotherapy, and increased time to start of chemotherapy.^7,10‐15

Furthermore, some authors have recognized it as a significant negative prognostic factor in terms of 90-day mortality and OS, although these latter data did not reach statistical significance in our series.^15,24

However, the identification of the population at greatest risk for AL, and proven strategies to prevent it, are still lacking today. Patients with OC represent a particular population since they usually present with critical clinical conditions and malnourishment, due to abundant ascitic fluid and widespread carcinomatosis.^25‐27 Moreover, given the predominantly peritoneal spread of OC, resections are usually higher than those performed for rectal cancer, and as suggested by Richardson et al.,¹⁷ this may explain the slightly lower rate of AL in OC patients.

When looking specifically at the OC literature, the studies appear inhomogeneous in assessing AL risk factors.^7,12,15 This could be due both to its rarity and to the heterogeneity of the surgical technique used to perform the intestinal resection and anastomosis.^10‐17,24

According to literature data, we confirmed the low preoperative albumin value (≤ 30 mg/dL)^{16,17,20,28‐30} and, in addition, identified low BMI (< 18 kg/m²) as independent preoperative risk factors for AL.

Among intraoperative variables, in agreement with several other studies, we confirmed medium–low level of anastomosis as a relevant risk factor of AL.^16,17,20,28 The hypotheses proposed to explain this association are various, such as the major surgical difficulty when performing an anastomosis in the narrower and deeper portion of the pelvis, with greater tissue trauma, increased tension, and lower blood flow.³¹ In this context, section of the IMA at its origin was another important negative predictor identified, suggesting that a reduced blood supply could hinder correct healing of the anastomosis. This hypothesis was also suggested by Son et al.,³² who demonstrated reduced incidence of AL in patients with preservation of superior rectal artery.

As anticipated, conflicting data are reported on the efficacy of ostomy as a strategy to prevent AL,^33‐44 although a broad consensus exists on its role in preventing the catastrophic consequences of AL, with a reduction in morbidity and mortality.^{34,36‐38,40,45‐48}

In our series, diverting ostomy did not prove to be a protective factor for AL, but it allowed conservative treatment in three patients who received it, which means a number needed to treat of 91 (91 ostomy to conservatively treat 1 AL). Another key to understanding these results could be that we correctly identify only 46.7% (7/15) of high-risk AL patients, in which we performed a protective ileostomy. Nevertheless, 57.1% (4/7) of these patients required a Hartmann procedure because of complete or nearly complete detachment of the anastomosis. These findings may indicate that there is a very small group of patients with an extremely increased risk of AL, where probably the best choice should be to perform a Hartmann procedure directly.

The greatest bias in this study certainly lies in its retrospective nature, and we are aware that the limited number of AL events makes the risk factor analysis less reliable. Furthermore, this prevented an evaluation of interesting surgical maneuvers for the reduction of AL, such as techniques for mobilization of the descending colon to obtain an anastomosis as vascularized and tension free as possible. However, this is currently the largest monocentric study to specifically report the rate of AL and its main risk factors for patients with advanced OC undergoing PDS or IDS. Moreover, all the resections and anastomoses were performed by a dedicated team of general surgeons who routinely collaborate with gynecologic oncologists, and all the procedures were performed using the same surgical technique, further reducing the inherent limitations of a retrospective surgical study.

The findings of our study totally support the hypothesis that cancer cachexia, malnutrition, and chronic inflammatory activity (documented by low serum albumin value)^10,49 are the most predictive preoperative factors of surgical morbidity, and strongly influence the postoperative course of OC patients,^50,51 also increasing the incidence of AL. Therefore, increasing attention must be paid to preoperative optimization of the patient by developing programs that include physical exercise, and nutritional counseling with possible protein integration.^51,53 Recently, in fact, a protocol was adopted in our department for preoperative optimization of the patient, providing both the association between mechanical bowel preparation and antibiotics, and careful nutritional assessment of the patient, protocols that were not active in the years considered in this study. Regarding the intraoperative details, data on the best colorectal resection and anastomosis in patients with advanced OC are lacking and the exact amount of blood flow needed to heal the intestinal wall remains to be determined.³¹ For this, it would be advisable to accurately tailor surgery according to tumor dissemination, to ensure maximal vascular supply to the anastomosis through preservation of the left colonic artery or, when possible, the upper rectal artery.^31,54,55 For this purpose, intraoperative evaluation of anastomosis perfusion using indocyanine green could be an interesting topic for future studies.

Another currently open question is the usefulness of ostomy for AL prevention. We are aware of the high rate of ostomies performed, which we could explain with the high rate of optimal cytoreduction achieved (95.9% of cases, with 79.6% of CGR) and the high surgical complexity of the operations performed. In fact, in these cases, the surgeon, in an attempt to minimize the incidence and severity of postoperative complications, could be reassured by the diverting ostomy, although our results show that this did not prevent the onset of AL and allowed conservative treatment in a relatively low number of cases (NNT 91). In support of this, with the exception of the distance of the anastomosis from the anal verge, risk factors for ostomy and those for AL differed for the most part (Table 5). This means that the decision-making process that leads the surgeon to perform a protective ostomy is quite heterogeneous and based on personal experience and surgeon’s feelings, rather than actual risk factors for AL. However, considering the nonnegligible complication rate directly related to the ostomy itself, the nonreversal rate of 33.5%, and its negative psychological consequences,^56,57 further studies are needed to better define the population that could really benefit from it and to make intraoperative decision-making as objective as possible