Peritoneal cancer index predicts severe complications after ovarian cancer surgery

Introduction: prediction and importance of severe postoperative complications after ovarian cancer surgery is a strong issue in patient selection and evaluation. Pre-and early peroperative predictors of severe 30-days postoperative complications (Clavien-Dindo class (cid:1) 3) after surgery for primary ovarian cancer are not fully established, neither their impact on patients ’ survival. Materials and methods: A prospective observational study included 256 patients with primary ovarian cancer FIGO stages IIB-IV, operated during 2009 e 2018 in a primary or interval debulking surgery setting. Patient variables were analysed in relation to severe postoperative complications (Clavien-Dindo class (cid:1) 3) and overall survival. Results: High-grade postoperative complications occurred in 24.2% patients. Class 3a complications were observed in 12.5% cases. High-grade complications class (cid:1) 3 were observed in 31.6% after primary debulking surgery compared to 12.2% after interval debulking surgery (p ¼ 0.0004). Peritoneal cancer index (cid:1) 21 and preoperative albumin concentration (cid:3) 33 g/L were independent predictors of high-grade complications. Peritoneal cancer index correlated with the surgical complexity score and completeness of cytoreduction. Increased peritoneal cancer index was a negative predictor of overall survival, but high-grade complications did not in ﬂ uence survival negatively. Conclusions: Peritoneal cancer index (cid:1) 21 was an independent predictor of high-grade complications after ovarian cancer surgery. Increased peritoneal cancer index also impacted overall survival negatively, but high-grade complications did not in ﬂ uence overall survival.


Introduction
Complete cytoreduction is the aim of ovarian cancer surgery, and an important predictor of survival [1,2].Primary debulking surgery (PDS) is the standard of care for operable patients with resectable International Federation of Gynecology and Obstetrics (FIGO) stage IIB-IV disease, followed by platinum-based chemotherapy [3,4].Although two randomized clinical trials suggested that interval debulking surgery (IDS) following neoadjuvant chemotherapy was not inferior to PDS [5,6], the rate of complete cytoreduction in those studies was low [2].IDS is usually used when preoperative investigations indicate poor metabolic condition, low performance status, comorbidities, high age, or expected high surgical complexity score (SCS), altogether comprising a high risk of residual disease and postoperative complications [7e9].
The peritoneal cancer index (PCI) is a measure of the extent of peritoneal carcinosis [10,11].PCI is widely used during the evaluation of the extent of colorectal carcinosis and prediction of surgical complexity [12].Elevated PCI has been associated with higher FIGO stage, prolonged duration of surgery, an increased number of visceral resections, incomplete cytoreduction and postoperative complications [13,14].PCI scoring is, however, to a lesser extent used by ovarian cancer surgeons and it is not fully known which baseline variables are the most important predictors for high-grade postoperative complications.
We therefore aimed to define baseline predictors of high-grade 30-days complications after PDS and IDS for primary ovarian cancer and the subsequent influence on survival.

Study design
This is an observational cohort study including patients undergoing ovarian cancer surgery during 2009e2018 at the Uppsala University hospital, a gynaecologic cancer surgery center for a population of more than 2,100,000 inhabitants.Baseline patient characteristics, surgical outcome variables and follow-up data were consecutively documented.The study was approved by the Swedish Ethical Review Board Dnr.2018/071 and 2019-05513Gk.

Patients and clinical characteristics
Inclusion criteria were primary ovarian (ICD C56.9), fallopian tube (ICD C57.0) or primary peritoneal cancer (ICD C48.2) and operable FIGO stages IIBeIV, based on histopathology and computed tomography as judged by a multidisciplinary conference.Patients with age above 75 years, major ascites with repeated draining or pleural effusion, low albumin concentration, radiologic signs of intestinal carcinosis, unresectable parenchymal metastasis or uncompensated comorbidities were preferably treated with IDS.None of the patients were treated with HIPEC or other intraperitoneal chemotherapies.The PCI was recorded after midline laparotomy at index surgery in the PDS and IDS groups in 13 abdominal regions with tumor lesions categorized as small (1) (<0.5 cm), moderate (2) (>0.5 cm -<5 cm) or large (3) (>5-cm), and the maximum PCI score was 39 [10,11].Completeness of cytoreduction (CC) 0 corresponded to the absence of residual disease, a CC-1 score indicated residual tumor nodules < 2.5 mm, CC-2 corresponded to residual tumor nodules between 2.5 mm and 2.5 cm and CC-3 indicated tumor nodules > 2.5 cm or a confluence of unresected tumor [10,11].The surgical complexity score (SCS) is a sum of the assigned complexity of each surgical procedure where the scores range from 1 to 3, where low SCS corresponds to 3, intermediate to 4 to 7 and high to !8 [9].We did not include inoperable, or openclose surgery, cases into this study.All patients underwent postoperative adjuvant treatment with Carboplatin and Paclitaxel and were followed for five years.Patients who were alive at the last follow-up in September 2020 were censored.The primary outcome was high-grade postoperative complications (Clavien-Dindo class !3) within 30 days of surgery, Clavien-Dindo class !3: 3a e intervention under local anesthesia, 3b e intervention under general anesthesia, 4a e single organ dysfunction, 4b e multi-organ dysfunction and 5 e fatal outcome [15].Secondary outcome was overall survival defined as time from date of diagnosis to death.Patient characteristics are presented in Table 1.

Statistical methods
Receiver-operator curve (ROC) analysis in relation to the binary outcomes for Clavien-Dindo class !3 or <3 was the basis for dichotomization of continuous variables.If no significant area under the curve was identified, we used the 75th percentile.Bowel resections were categorized into colectomy (2), large bowel resection (1) or no resection (reference) and type of anastomosis was categorized as rectosigmoid end-to-end anastomosis, colo-colic anastomosis, ileocolic, small bowel anastomosis (1) or no resection (reference).Only bowel resection variable was included into multivariate logistic regression since it involves anastomosis, end stoma or protective ileostomy.Completeness of cytoreduction was categorized to CC-0 or CC ! 1. Summary statistics comprised range, mean and 95% confidence interval (CI) of the mean.Chi-square test and correlation analyses were used to assess relationships between the variables.Logistic regression modelling was performed for continuous and categorized variables using enter and stepwise methods to identify predictors of high-grade postoperative complications.Here we presented results of the univariate and adjusted multivariate logistic stepwise regression modelling based on analysis of categorized variables.To identify potential confounding, we adjusted multivariate models for all analysed variables, even if not significant in the univariate analysis.Strong correlation between variables (R > 0.6) lead to exclusion of the least clinically important or previously established variable from the multivariate logistic regression modelling.Associations were summarized by calculating the odds ratio (OR) and corresponding 95% CI and pvalue.Prediction model validation comprised Akaike information criterion and Nagelkerke's R2.The Nagelkerke "pseudo" R2 ranges from 0 to 1, with 1 indicating a perfect model fit.Missing data were not included in the analyses.Survival analysis was performed using Cox regression modelling and Kaplan-Meier analysis with the logrank test.No patient was lost to follow-up.The analyses were carried out on the Statistica 13.3 (Tibco Software Inc., USA) and MedCalc 19.1 (MedCalc Software BVBA, Ostend, Belgium).

Results
In total, 403 patients with ovarian cancer underwent surgery during 2009e2018 (Fig. 1).Of those, 37 patients had FIGO stage I-IIA and 74 patients underwent relapse surgery, leaving 292 patients fulfilling the inclusion criteria for the study.Of those, 31 were inoperable, 5 cases were finally diagnosed as a non-gynecological cancer, leaving 256 women for final inclusion.These 256 patients were studied here and comprised 158 patients who underwent PDS and 98 patients, subjected to IDS (Table 1
After large bowel resection (162 patients), high-grade complications were observed in 50/162 (30.9%).Specific intestinal high-  In the multivariate analysis, OR for albumin concentration 33 g/L was 4.35 (95% CI, 1.73e10.94)and for PCI!21 5.78 (95% CI, 2.00e16.73,Table 3 A) (adjusted for age, BMI, platelet count, FIGO stage, PDS or IDS and completeness of cytoreduction).The results did not change after additional adjustment for bowel resection and blood loss.There was a strong correlation between age and ACCI (R ¼ 0.727) and between duration of surgery and SCS (R ¼ 0.665) and PCI (R ¼ 0.747), therefore ACCI and duration of surgery were not included in the multivariate analyses.Another strong correlation between PCI and SCS (R ¼ 0.721, Fig. 2 A) and a moderate correlation between the potential confounder CC and PCI (R ¼ 0.395) was observed (Fig. 2 B).Large bowel resection and colectomy were not associated with high-grade postoperative complications in the multivariate analysis.

High-grade complications predictors and overall survival
High-grade complications did not influence overall survival negatively (Table 3B).After adjustment for age, BMI, platelet count, FIGO stage, PDS or IDS and completeness of cytoreduction, hazard ratio (HR) for increasing age was 1.02 (95% CI, 1.00 to 1.05) and HR for PCI was 1.05 (95% CI, 1.02 to 1.08, Table 3 B).High age and PCI thus predicted a short survival in multivariate analysis.Median overall survival of the patients who underwent complete cytoreduction was 65 months compared to 40 months after incomplete cytoreduction (p ¼ 0.001, Fig. 3 A).Median survival after PDS was longer (59 months, p ¼ 0.037) compared to IDS (51 months, Fig. 3 B).After complete cytoreduction median overall survival after PDS was 81 months while after complete IDS it was 53 months (p ¼ 0.011, Fig. 3 C).A tendency to higher rate of complete cytoreduction was noted after IDS (81/98e82.6%)compared to PDS (72.8% -115/158) (p ¼ 0.055).

Discussion
Our findings suggest that PCI !21 and albumin concentration 33 g/L define patients with increased risk for high-grade postoperative complications, half of which comprise Clavien-Dindo class 3b and higher.Correlation between PCI and completeness of cytoreduction and between PCI and SCS suggests that PCI is the most powerful predictor of surgical outcome.We see an advantage in using PCI as it is estimated straight after the midline laparotomy and it can therefore be used for clinical decision making.According to the previous findings for ovarian cancer, PCI >20 was associated with a high rate of postoperative complications [14].Sensitivity of PCI assessment upon laparotomy is around 70% and is superior to laparoscopic and computer tomography-based evaluation of the carcinomatosis [14].Imaging techniques are rapidly evolving suggesting that the methods for preoperative PCI assessment should be further explored.According to a systematic review, laparoscopy is one of the widely used methods for carcinosis evaluation in ovarian cancer [16].Laparoscopic evaluation during the workup of ovarian cancer patients was also found to be useful to predict major complications through the combination of tumor load pattern and increased ascites production [17].However, a risk for disease   upstaging as well as a documented risk of the abdominal wall metastasis and necessity for abdominal wall resection are arguments against laparoscopic evaluation of PCI [18,19].We previously found that PCI evaluation is useful for assessment of operability and we identified a cut-off of !24 in relation with the risk for incomplete cytoreduction [13].Additionally, pre-operative cancer antigen-125 > 600 U/mL, PCI >20 and intra-operative mapping of ovarian cancer score >6 are reported to be predictors of complete tumour resection, while the combination of all these three values predicted the incomplete resection of disease in up to 90% of patients [20].Considering the above-mentioned findings, it appears important and informative to perform PCI evaluation at the initial stage of surgery.No difference in the PCI was observed when comparing PDS and IDS, which we attributed to our principles of patient selection for PDS for operable patients with resectable disease and the IDS strategy for those with ascites, pleural effusion, impaired nutritional status with low albumin values and signs of extensive intestinal carcinosis.Other previously identified predictors of postoperative complications are high age, BMI above 40, increased ACCI and ASA score, high levels of C-reactive protein, interleukin-6, increased platelet count [21,22] and low albumin concentration [9,23e26].
According to results reported from other high volume centers, high-grade complications after extensive upper abdominal procedures were observed in 19% of cases [27], after incomplete PDS in 23% and after complete PDS in 45% cases [23].A metanalysis, based on 18,579 patients, identified high-grade postoperative complications in 24% patients [28].A comparison of equally long time periods in the present cohort showed a decrease in the rate of highgrade complications and increase in the completeness of cytoreduction (data not shown).
Pleural effusion was a common complication occurring in 21% cases and was always associated with full thickness diaphragmatic resection [27].Stage IVB was diagnosed intraoperatively after pleural cavity exploration in 20/82 (24%) in line with the recommendation to explore pleural cavity in bulky disease IIIC [29,30].We practice intraoperative placement of active thoracic drain after extensive manipulation on the diaphragmatic muscle to prevent postoperative pleural effusion.In a meta-analysis, the rate of 30days postoperative mortality was 4.6% and was related to the increasing patient age, tumor burden and FIGO stage IV and high SCS [28].
Rectosigmoid resection and colectomy are feasible procedures providing good pelvic disease control and improved survival, despite higher rate of complications [31,32].Previous studies suggested that rectosigmoid anastomosis is associated with a higher risk of complications, but not directly with intestinal leaks [33,34].We also observed an increased risk of high-grade postoperative complications of various type after large bowel resection, colectomy and large bowel anastomosis, occurrence of protective ileostomy in 33% after PDS and 18% after IDS.However, bowel related complications were observed in 5/162 (3.1%) patients compared to 30% of various high-grade complications after large bowel resection with and without anastomosis.
Chance for longer overall survival after complete cytoreduction is in several studies thought to outweigh the risk of high-grade complications without significant influence on the quality of life [9,35e38], while other authors advocate IDS when high SCS is expected [8,14,27].Complete cytoreduction increased the probability for improved overall survival [1,2], even at FIGO stage IV [39,40].Situation is however different for the patients surgically treated for colorectal carcinosis, disease with peritoneal spread and limited chemosensitivity.A decrease in overall survival by 9 months was observed in patients who experienced high-grade complications [41].Similarly, a 11.7% five-year overall survival was observed in the group who underwent high-grade complications compared to 58.8% for those who had a complication-free recovery [42].Tumour biology and disease progression pattern of these cancers are rather different which might explain the differences in survival.
The major strengths of our study are consecutive inclusion of patients and multidisciplinary based decision for treatment performed by the same team of surgeons and oncologists.This study would have benefited from a larger sample size, however while a multicenter analysis would allow for a higher number of patients, it would be biased by the variability in treatment approaches and furthermore, the PCI is not a commonly recorded variable.Our identified cut-off for the PCI in relation to high-grade postoperative complications is supported by previous findings [14].A possible selection bias is present in the decision for PDS or IDS for patients with more aggressive ovarian cancer biology and with more severe comorbidities, which makes it difficult to interpret the survival advantage after PDS compared to IDS.An ongoing phase III randomized clinical trial TRUST will hopefully define criteria for more objective patient selection for IDS and PDS [3,4].Recent advances in characterization of tumor biology provide additional selection criteria for PDS or IDS [43].However, a clear predictor of improved overall survival is complete cytoreduction and our observation is in line with previous studies [1,2].Further analysis of tumor biology and their impact on postoperative outcomes in primary ovarian cancer is justified.
In accordance with our findings and based on previous reports, we suggest that ovarian cancer surgeons should evaluate PCI upon middle line laparotomy since it is a tool to predict surgical complications and predict expected surgical outcome.Despite occurrence of the high-rate complications with the higher rate after PDS (32%, including 10% of reoperations), compared to IDS (12%, contain 4% reoperations), high grade complications do not impact long term outcome and should not be a reason to prefer IDS.The worse survival after IDS could be explained by the initially more aggressive tumor biology with massive ascites and pleural effusion, development of chemoresistance during neoadjuvant treatment, larger part of the stage IV disease.Our findings support an individualized selection for PDS or IDS, weighing in the patient condition, disease spread pattern and risk for incomplete cytoreduction against the risks for complications.

Conclusions
1. High PCI with a cut-off !21 and reduced albumin concentration 33 g/L were independent predictors of high-grade postoperative complications.2. High-grade complications were more frequent after PDS compared to IDS.An increasing age and high PCI were independent predictors of impaired overall survival.3. High-grade postoperative complications did not impact survival negatively.4. PCI correlated with the postoperatively evaluated completeness of cytoreduction and SCS.

Funding
This study was supported by the Swedish Cancer Society, project number 170206.

Fig. 1 .
Fig. 1.Flowchart of the patient recruitment into the study.

Fig. 2 .
Fig. 2. Correlation between a) PCI and completeness of cytoreduction and b) PCI and SCS.

Fig. 3 .
Fig. 3. Overall survival of analysed patients: a) after complete compared to incomplete cytoreduction, b) after all PDS compared to all IDS, c) after complete cytoreduction, PDS compared to IDS.

Table 1
).Of 256 patients, 135 were alive at the last follow-up in September 2020 with a mean follow-up of 46 months (range 21e123 months).Patients characteristics, clinical parameters, tumor and surgery-related variables in 256 women with ovarian cancer in relation to management with PDS or IDS.
(continued on next page)

Table 3a
Univariate and multivariate modelling to predict high-grade postoperative complications.

Table 3b
Univariate and multivariate Cox regression analyses to predict overall survival.