P.R.O.P.S. — A novel Pre-Operative Predictive Score for unresectability in patients with colorectal peritoneal metastases being considered for cytoreductive surgery (CRS) and hyperthermic intraperitoneal chemotherapy (HIPEC)

Cytoreductive surgery (CRS) and hyperthermic intraperitoneal chemotherapy (HIPEC) have been showed to increase survival in patients with colorectal cancer peritoneal metastases (pCRC) [1, 2]. A proportion of CRS and HIPEC cases are abandoned intra-operatively due to extensive disease (i.e. high peritoneal carcinomatosis index (PCI) score) resulting in unnecessary laparotomy [3‐5]. A way to circumvent this is to identify the unresectable cases pre-operatively. Pre-operative radiological investigations alone fail to accurately predict PCI score accurately [6]. One can also extrapolate the factors founds in prognostic scores such as Peritoneal Surface Disease Severity Score (PSDSS), Verwaal Prognostic Score (PS) and Colorectal Peritoneal Metastases Prognostic Surgical Score (COMPASS) for pre-operative selection outlined in Table 1 [7‐9]. However, these prognostic scoring systems were established from studies that only included patients with pCRC that underwent successful CRS and HIPEC (i.e. completeness of cytoreduction score of 0 or 1), and were designed to select patients for complete resection or improved survival. In addition, both scores include intra-operative factors in their scoring model. Therefore, these scores are not primed to predict for unresectability pre-operatively. We have published an earlier study on preoperative factors associated with unresectability included all tumour types [10]. For this study, we only selected patients with pCRC. The aim was to develop a Pre-Operative Predictive Score (PROPS) for unresectability in patients with pCRC, and compare it with the PSDSS and PS, to determine which score best predicts for unresectability.

This is a follow-up to a study that was conducted at the National Cancer Centre Singapore from April 2004 to May 2014 and was approved by the local Centralised Institutional Review Board. Data was retrospectively collected from a prospective CRS and HIPEC database of patients. Only patients with pCRC were included.

The patients included in this study had documented colorectal cancer with peritoneal metastases either on radiological imaging or during previous surgery. All cases were presented in a multidisciplinary meeting where surgical, medical and radiation oncologists, radiologist and pathologists were present, and the decision to proceed with CRS and HIPEC was determined after a consensus was reached. Clinical factors taken into consideration to formulate a decision include patients’ presentation, grade (e.g. presence of signet ring, mucinous or poorly differentiated cells) and stage of tumour, disease free-interval (DFI), response to previous therapies (e.g. chemotherapy or surgery) and radiological images. Patients recommended to undergo CRS and HIPEC were all without distant metastases as determined on imaging. In addition, they were of Eastern Cooperative Oncology Group (ECOG) status 0 or 1 [11].

Patients were grouped into two groups: the unresectable group was defined by inoperability determined intra-operatively, which resulted in the abandonment of planned CRS and HIPEC; the resected group was defined by completion of CRS and HIPEC regardless of the completeness of cytoreduction (CC) score, although it was noted that all patients who underwent CRS and HIPEC achieved either CC-0 or CC-1 [12]. The two groups were compared to identify pre-operative factors that will be useful to detect potentially unresectable patients which can help guide pre-operative decision-making and counselling and improve patient selection to decrease the chance of unnecessary exploration.

The clinical pre-operative factors that were analysed included the patients’ presentation, previous response to chemotherapy and/or surgical intervention as well as blood and radiological investigations. Tumour response to chemotherapy was evaluated with the Response Evaluation Criteria in Solid Tumours (RECIST) criteria [13]. PCI score was calculated intra-operatively during laparotomy. In addition, supplementary data was collected from clinical notes, electronic medical records and surgical records to complete both the PSDSS and PS for the same group of patients.

Overall, the rate of unresectable cases was 13% (7/56). The demographics between the unresectable and successful groups were comparable except for PCI score and histology (Table 2). All unresectable cases were due to high PCI score with mean score of 24 (SD = 2.6). Of note, the unresectable cases had more tumours with signet ring cell and mucinous histology. Otherwise, the majority of the successful group achieved adequate cytoreduction (93% of CC-0 and CC-1).

Univariate analysis of the discovery set identified ten pre-operative factors significant for unresectability (Table 3). With regards to clinical presentation, patients in the unresectable group were more likely to complain of bloatedness (75% vs. 15%, p = 0.03), and were found to have palpable abdominal masses (25% vs. 0%, p = 0.04) on physical examination. In terms of disease factors, there were a greater proportion of high-grade tumours (50% vs. 4%, p = 0.01) in the unresectable group. For patients who had received treatment prior to the consideration of CRS and HIPEC, more patients from the unresectable group underwent multiple lines of chemotherapy (50% vs. 29%, p = 0.04), displayed progression of disease during chemotherapy (33% vs. 0.0%, p = 0.00) and/or had suboptimal initial resection (25% vs. 7%, p = 0.04). For pre-operative investigations, elevated tumour markers (100% vs. 50%, p = 0.03), as well as CT scan findings of ascites (100% vs. 9%, p = 0.00), omental thickening (100% vs. 4%, p = 0.00) and/or small bowel disease (25% vs. 8%, p = 0.01) were also more common in the unresectable group.

In addition to the above factors, those factors with p < 0.1 were chosen for multivariate analysis. All factors besides progression of disease during chemotherapy were still found to be significant. The remaining nine variables were categorised into three groups to generate PROPS (Table 4). Using the beta-coefficient value, individual scores were assigned to each variable in every group: (i) poor tumour biology (1 point each): suboptimal resection (BC 1.0, OR 0.17, 95% CI 0.05–0.63, p = 0.01), underwent multiple lines of chemotherapy (BC 1.0, OR 0.18, 95% 0.03–0.94, p = 0.07), and high-grade tumour (BC 1.0, OR 0.16, 95% CI 0.04–0.68, p = 0.03); (ii) heavy tumour burden (2 points each): sensation of bloatedness (BC 2.5, OR 0.16, 95% CI 0.04–0.58, p = 0.01), palpable abdominal mass (BC 2.5, OR 0.01, 95% CI 0.00–0.24, p = < 0.01) and computed tomography findings of ascites (BC 2.5, OR 0.02, 95% CI 0.00–0.33, p = < 0.01), small bowel disease (BC 2.5, OR 0.01, 95% CI 0.00–0.17, p = < 0.01) and omental thickening (BC 2.5, OR 0.14, 95% CI 0.02–0.78, p = 0.05); and (iii) active tumour proliferation (2 points): elevated tumour markers (BC 2.5, OR 0.10, 95% CI 0.01–0.84, p = 0.03).

Using the validation set for unresectability prediction (Table 5 and Fig. 1), PROPS achieved 86% accuracy with 100% sensitivity and 68% specificity at a cut-off of 3 (YI 0.68). PSDSS achieved 85% accuracy with 100% sensitivity and 63% specificity at a cut-off of 10 (YI 0.63). PS achieved 73% accuracy with 100% sensitivity and 68% specificity at a cut off of 3 (YI 0.68). And lastly, COMPASS achieved 61% accuracy with 27% sensitivity and 100% specificity at a cut off of 90 (YI 0.27). Of note, at a cut-off of 6, PROPS was able determine unresectability to near absolute certainty (specificity 95%).

Current pre-operative selection tools such as PSDSS, PS and COMPESS are useful in predicting survival outcomes, but may not be prime in identifying patients with unresectable disease as these studies excluded unresectable cases in their inclusion criteria. Moreover, these studies included intra-operative factors in their model (e.g. PCI in PSDSS and COMPASS and extent of carcinomatosis in PS) that preclude applicability in the pre-operative setting. The one other score available to predict unresectability in pCRC is the mCOREP [14]. We have considered incorporating mCOREP score into the study. However, as this is a retrospective study, we have a significant amount of missing data, in particular the CA125, which rendered the calculation of mCOREP score to be incomplete. The reason why CA125 was not part of our routine pre-operative was because it is specific for ovarian cancer instead of colorectal cancer. In a recent study conducted by Enbald et al., it was found that the mCOREP score (p = 0.9) and PSDSS (p = 0.09) are not predictive of opened and closed laparotomy. Instead, COMPASS score was found to have better predictive value for opened and closed laparotomy as compared to mCOREP score (p < 0.001) [14]. Additionally, Demey et al. have externally validated COMPASS to be superior over PSDSS in its prognostic ability. However, in our paper, we found COMPASS to perform the poorest in predicting unresectability compared to PROPS, PSDSS and PS. This may be attributed to the difference in our patient demographics as the patients in Demey et al. paper were considerably older (57 years old vs. 49 years old) and had lower PCI score (18 vs. 24), which result in discrepancy in the aggregation of COMPASS by approximately 25 points.

Our article demonstrated specific factors for the pre-operative identification of opened and closed cases that we condensed into a novel model—PROPS. At a cut-off of 3, we found that PROPS was best able to detect unresectability (specificity 68%) with the lowest rate of false positive (i.e. complete CRS/HIPEC, sensitivity 100%) compared to PSDSS and PS. At a cut-off of > 6, PROPS can distinguish unresectable cases close to absolute certainty (specificity 95%). The nine factors in PROPS were categorised into three main groups: (i) poor tumour biology, (ii) heavy tumour burden and (iii) active tumour proliferation.

The PROPS scoring system is a novel pre-operative scoring system that relies solely on pre-operative factors, and is as effective in predicting unresectability compared to the PSDSS, PS and COMPASS.

Moving forward, it will be important to perform external validation for PROPS, and to consider if PROPS can be applied to other tumour types besides colorectal to decrease the incidence of unresectability in planned CRS and HIPEC.