Peritoneal Cancer Index Correlates with Radiographic Assessment of Colorectal Carcinomatosis

Abstract

Background:

Peritoneal Cancer Index (PCI), calculated intra-operatively, has previously yielded mixed results when correlated with Computed Tomography (CT). We aimed to quantify variation in this scoring modality comparing radiologists’ and surgeons’ radiologic PCI (rPCI) assessment.

Methods:

The rPCI of 104 patients treated at a single institution with peritoneal carcinomatosis was calculated by an abdominal radiologist and a surgeon. An additional 36-patient cohort was studied to compare pre-operative rPCI versus intra-operative gold standard PCI. Agreement was compared using Kappa statistics.

Results:

Of 104 patients studied, rPCI ranged from 2 to 39 (median 12, IQR 6–23) by radiologist’s analysis and 2 to 37 (median 9, IQR 6–15) by surgeon’s analysis. There was good agreement for PCI cutoffs of 15 (77.48%; Kappa 0.40) or 20 (78.63%; Kappa 0.24). The 36-patient cohort undergoing surgical exploration showed a median rPCI of 4 (IQR 2–5.75); median intra-operative PCI as 11 (IQR 6–12), with significant difference in score by modality (p<0.001, Wilcoxon signed rank test.

Conclusions:

Among rPCI cutoffs of >15 and >20, there was strong concordance between surgeon’s and radiologist’s rPCI, interobserver reliability of rPCI. Moreover, concordance with intra-operative PCI translated to radiographic assessment. The rPCI consistently underestimated intra-operative PCI, suggesting that rPCI may be a useful conservative tool for assessing peritoneal burden. While surgical exploration is needed to “rule in” patients as candidates for CRS, we suggest that rPCI can be used to “rule out” patients as CRS candidates based on institutional PCI cutoffs.

INTRODUCTION:

Peritoneal metastases from colorectal cancer (CRC) pose a significant challenge to patients and a burden to the healthcare system in terms of rising incidence and necessity for complex care at specialized centers.^{1, 2} While there are evolving standards of care of peritoneal carcinomatosis of colorectal origin, current consensus is that select patients may benefit significantly from cytoreductive surgery (CRS) with or without heated intraperitoneal chemotherapy (HIPEC).^3–6 Importantly, burden of disease as measured by peritoneal cancer index (PCI) plays a key role in patient selection and defining those patients who may benefit from CRS+/−HIPEC.^7–9

The concept of patient suitability for CRS/HIPEC in the setting of metastatic colorectal disease has been previously studied.^10–15 A number of PCI score cutoffs have been proposed to define those patients that may benefit from CRS+/−HIPEC.¹⁶ The calculation of this score, however, classically requires a formal laparotomy and exploration.^{7, 8, 17} The use of diagnostic laparoscopy (DL) has reduced the need for a formal exploration and decreased the rate of non-therapeutic laparatomies.^{14, 18–22} While limitations of cross sectional imaging exist for peritoneal disease, there is significant interest in the potential utility of Computed Tomography (CT) scans in defining resectability and even outcomes.^23–27 Furthermore, CT has been proposed as a safe and useful adjunct to DL in assessing the extent of peritoneal disease and candidacy for future CRS.^{28, 29} Despite its known limitations, radiographic PCI (rPCI) is often calculated and discussed by radiologists and surgical oncologists in assessing these patients for candidacy in the clinic and at multi-disciplinary tumor boards.³⁰ The reproducibility of this score across observers and its correlation with operatively calculated PCI has not been consistently defined.²³

As a non-invasive surrogate, CT scans have yielded mixed results. Given the lack of consistent study of the reproducibility and correlation between radiologic PCI (rPCI) and standard PCI scores, we evaluated the utility and reproducibility of rPCI in patients with colorectal carcinomatosis in a cohort of patients from our institution. Moreover, we correlated these scores with PCI calculated intraoperatively to define the clinical value of these in surgical decision making.

METHODS:

Study Design

CT scans of patients with peritoneal metastatic disease from CRC were retrospectively identified from a single institution database at an NCI-designated cancer center. The rPCI of 104 patients was calculated by a fellowship-trained abdominal radiologist. Reproducibility and concordance were determined by review of CT scans by a complex general surgical oncology fellow board-certified in general surgery. An additional 36-patient cohort was analyzed for pre-operative rPCI compared to PCI calculated during surgical assessment and noted on retrospective review of a prospectively maintained database.

Image reviewers were blinded to the clinical details of the patients so as not to bias the scoring. We also employed inter-reviewer blinding so that scores from the radiologist were not known by the scoring surgeon, and vice versa. Following score assignment, reviewers were unblinded to further analyze parameters such as inter-rater agreement and clinical outcomes.

Radiographic PCI Scores

To calculate the radiographic rPCI score, we followed the methodology of the standard intra-operative PCI scoring system.^{7, 8} The abdominal cavity assessment was divided into nine segments with the small bowel being divided into four segments. Reviewing physicians correlated disease-burden with each rPCI region as assigned a score from 0 to 3. CT scans were reviewed in the axial, coronal and sagittal planes and scored immediately at time of image review by region with summation resulting in final PCI score.

Retrospective Review

This study was performed under an Institutional Review Board (IRB)-approved protocol. Patients seen at M.D. Anderson Cancer Center from the years 2015 to 2019 with peritoneal dissemination of their primary cancer were included. Colorectal disease site primary anatomic location was subdivided into left colon, right colon, and rectum. On review, three cases were ultimately appendiceal cancers and one additional case included an appendiceal cancer synchronous with a metastatic rectal cancer. A second synchronous primary case included both cecal and sigmoid lesions.

CT scans of patients with known peritoneal spread from their primary site were retrospectively identified from a database of patients treated at a single NCI-designated comprehensive cancer center. The rPCI score of each patient was determined using CT scans obtained on various scanners from different vendors with slice thickness varying between 2.5–5.0 mm, performed either at our institution or an outside facility. When multiple scans were available for review the most recent CT scan before date of treatment initiation with thin cuts and contrast performed at M.D. Anderson was selected.

Statistical Analysis

JMP software version 15 (Cary, NC) and Stata version 13.1 (College Station, TX) were utilized to analyze the data. Kappa analysis was utilized to compare differences between reader scores for the CT scans by the radiologist versus surgeon. Descriptive statistics were utilized for patient demographics and disease site breakdown. Paired t test Wilcoxon rank was utilized for matched comparison of rPCI and PCI scores across each individual participant. Kaplan-Meier with log-rank and Wilcoxon was utilized for survival analysis.

RESULTS

Demographics

There were 104 patients identified from the prospectively maintained institutional database of patients with peritoneal carcinomatosis from colorectal cancer. This cohort of patients were all unresectable and treated with systemic therapy alone. The median age at peritoneal carcinomatosis diagnosis was 57.8 years with a range of 22.7 to 92.4 years of age. There was a male predominance with 53.8% (n=56) being men and 45% (n=48) being women (p<0.001). Self-reported race included predominantly Caucasian or White at 73.1% (n=76), African-American or Black comprising 14.4% (n=15), Asian at 3.8% (n=4) and other ethnicity at 8.7% (n=9). Self-reported ethnicity breakdown included predominantly Non-Hispanic at 83.7% (n=87), Hispanic or Latino at 12.5% (n=13), and unknown or declined to answer at 3.8% (n=4).

Median survival from the time of their carcinomatosis diagnosis was 16 months with a range of 1 to 116 months. The predominant site of primary cancer diagnosis was left colon comprising 48% (n=50), followed by right colon at 35.6% (n=36), rectum at 12.4% (n=13), primary appendix at 2.8% (n=3) and lastly two synchronous primaries (1.9%). Regarding microsatellite instability (MSI), 87 participants tested as MSI-stable, 6 as MSI-high, and 11 did not have MSI status abstracted within searchable text of the electronic medical record. Patient demographics are summarized in Table 1.

Table 1.

Demographic summary of 104 patients with Peritoneal Carcinomatosis identified through the institution’s prospectively maintained database at MD Anderson Cancer Center.

Category	% (n) unless otherwise specified
Median Age (range)	57.8 years (22.7 – 92.4)
Sex
Male	53.8 (56)
Female	46.2 (48)
Race
Caucasian or White	73.1 (76)
African-American or Black	14.4 (15)
Asian	3.8 (4)
Other	8.7 (9)
Ethnicity
Non-Hispanic	83.7 (87)
Hispanic or Latino	12.5 (13)
Unknown / Declined to Answer	3.8 (4)
Marital Status
Married	73.1 (76)
Single	11.5 (12)
Divorced	7.7 (8)
Widowed/Other	7.7 (8)
Primary Site of Cancer
Appendix	2.9 (3)
Right Colon	34.6 (36)
Left Colon	48.1 (50)
Rectum	12.5 (13)
Two Synchronous Primaries*	1.9 (2)
Median Survival from Unresectable Carcinomatosis Diagnosis	16 months (1 – 116 months)
MSI Status
MSI-Stable	83.7 (87)
MSI-High	5.8 (6)
Unknown / Not Tested	10.5 (11)

^*Synchronous primaries included appendix/rectum & cecum/sigmoid

Radiographic Peritoneal Cancer Index (rPCI) Scoring

Of 104 patients studied, rPCI ranged from 2 to 39 (median 12, IQR 6–23) as analyzed by a radiologist and 2 to 37 (median 9, IQR 6–15) as analyzed by a surgeon. Modest overall agreement of 57.14% (Kappa 0.38) existed between the reviewers (Figure 1, Supplementary Table 1). Radiological review determined 42 (40.4%) had a PCI greater than 15, and 31 (29.8%) had a PCI greater than 20. Surgical review determined 27 patients (25.7%) had a PCI greater than 15, and 16 patients (15.2%) had a PCI greater than 20. There was good agreement for patients with PCI cutoffs of 15 (77.48%; Kappa 0.40) or 20 (78.63%; Kappa 0.24), commonly utilized cutoffs for determining resectability. The areas of highest concordance were the left upper quadrant (68.57%) and the upper jejunum (70.48%). Carcinoembryonic antigen (CEA) levels proximate to the time of PC diagnosis did not correlate with rPCI levels (Supplementary Figure 1).

Figure 1.

Inter-rater agreement between Surgeon’s rPCI score and Radiologist’s rPCI score.

Survival Analysis

Kaplan-Meier survival analysis was performed based on an rPCI cutoff of 15. We utilized a composite assessment whereby if either surgeon-assessor or radiologist-assessor scored the PCI as greater than 15, that score was categorized in a high-risk (PCI >15) cutoff. This group of unresectable metastatic patients had a worse overall survival from PC diagnosis date compared to those with lower PCI’s (Figure 2). This was statistically significant by both log rank (p=0.043) and Wilcoxon (p=0.0034) analysis.

Figure 2.

Survival analysis from date of carcinomatosis diagnosis as stratified rPCI.

rPCI Validation with Surgical Comparators

An additional cohort of 36 patients who went to the operating room for planned cytoreduction was assessed to compare rPCI to the intra-operative gold standard PCI. Among those patients, there was a female predominance of 55.6% (n=20), a Caucasian predominance of 75% (n=27) with Non-Hispanic ethnicity as the majority at 83.3% (n=30) as summarized in Table 2. A majority of patients were being treated for a right colon primary (52.8%), followed by left colon primary (44.4%).

Table 2.

Demographic summary of 36 patients with Peritoneal Carcinomatosis as assessed by intra-operative PCI scoring in addition to radiographic assessment.

Category	% (n) unless otherwise specified
Median Age (range)	55.3 years (23.2 to 78 years)
Sex
Male	44.4 (16)
Female	55.6 (20)
Race
Caucasian or White	75 (27)
African-American or Black	5.6 (2)
Asian	11.1 (4)
Other	8.3 (3)
Ethnicity
Non-Hispanic	83.3 (30)
Hispanic or Latino	13.9 (5)
Unknown / Declined to Answer	2.8 (1)
Marital Status
Married	75 (27)
Single	16.7 (6)
Divorced	2.8 (1)
Widowed/Other	5.5 (2)
Primary Site of Cancer
Right Colon	52.8 (19)
Left Colon	44.4 (16)
Rectum	2.8 (1)
Median PCI score
Intra-operative score	7.5 (2 – 12)
Radiographic score	4 (0 – 9)

This additional set of patients was identified as those who underwent operative cytoreduction with or without HIPEC from the institutional database. Among these participants, 18 (50%) of patients were later diagnosed with peritoneal recurrence following initial cytoreduction. Among patients with extra-peritoneal sites of metastasis, regions included abdominal wall, lung, retroperitoneal nodes, and intra-parenchymal liver metastases. Median rPCI was 4 (IQR 2 – 5.75); median intra-operative PCI was 7.5 (IQR 5.5 – 9) with a significant difference between groups (p<0.001, Wilcoxon signed rank test). Overall, the rPCI underestimated the true intra-operative PCI (Figure 3).

Figure 3.

rPCI (blue) compared to gold standard intra-operative PCI (red) amongst the 36-patient cohort who went to surgery following initial CT-based assessment. Gray lines represent the individual patient’s score. t-Ratio 6.06; p < 0.001; 95% CI 2.13 – 4.27.

DISCUSSION:

This work demonstrates that, among higher-risk stratifications of >15 and >20 PCIs, there was strong concordance between the surgeon’s and radiologist’s rPCI, underscoring the ability of cross-sectional imaging to define patients unlikely to benefit from CRS. Moreover, concordance with intra-operative PCI translated to radiographic assessment was noted as seen in previous work as well.²⁴ The rPCI, however, regularly underestimated intra-operative PCI, although not unanimously, suggesting that rPCI may be a useful conservative tool for assessing peritoneal disease burden. While CT does not “rule in” patients as candidates for CRS, it does suggest that patients can be reliably “ruled out” as CRS candidates based on the rPCI and common surgical cutoffs. This is important in making surgical recommendations, avoiding unnecessary laparoscopy, and minimizing delays in systemic chemotherapy initiation for clearly unresectable patients.³¹

These results suggest the DL may be indicated only for patients with a rPCI of less than 20. This is supported by a previous study which analyzed the combination of DL as an adjunct to CT scan assessment only in patients with higher PCI.²⁶ Previous work has included a retrospective review with two radiologists reading CT scans of a similar patient population to our study and concluded that patients with a rPCI of >10 may benefit the most from DL.²⁶ Although in our study the overall rPCI was in only modest agreement between the radiologic reader and the surgical reader, they were in high agreement for the patients that would or would not be candidates for CRS/HIPEC based on common PCI cutoffs of 15 or 20. Although controversy exists in the ability of CT to reliably define resectability,^32–34 the majority of previous work looked to define ability to resect rather than using CT as a tool to ‘rule out’ the potential for CRS.

Regarding OS as stratified by rPCI score, it is perhaps not surprising that those patients with a higher disease burden had diminished overall survival. However, the group of patients with a rPCI<15 that did not go on to resection may have had either a more aggressive biology, comorbidities precluding resection, or otherwise technically unresectable disease -- all of which my lead to a diminished overall survival. Our ability to quantify the importance of this factor, however, is limited by our number of patients and power to perform a multivariable regression to understand its ability as an independent predictor.

This data has significant implications for patients and their surgeons considering cytoreduction for their colorectal cancer with peritoneal metastases and higher rPCI. While the rPCI may stratify patients with high volume disease as able to forego DL in favor of definitive chemotherapy, those patients with underestimated rPCI or borderline PCI require more nuanced attention. We would recommend liberal use of diagnostic laparoscopy in those patients with low or borderline rPCI, or in those who uncertainty exists. Notably we found here that true PCI was oftentimes significantly higher than rPCI and laparoscopy remains critical for staging potentially patients; allowing for direct visualization of the peritoneal cavity and calculating of a PCI score.

With regard to the choice of imaging modality in measuring peritoneal disease burden, a number of other tools such as MRI and Positron Emission Tomography (PET) have been studied. For ovarian cancer specifically, Positron Emission Tomography (PET) imaging has been proposed over CT scans reflecting the unique biology of these two histologies.³⁵ Magnetic Resonance Imaging (MRI) has similarly been studied and occasionally utilized at select institutions as a reliable technique for peritoneal assessment and surveillance.^{25, 36–42} For the purposes of this study, due to the high predominance of CT scan utilization for assessment and surveillance of these patients with peritoneal disease, we chose to focus on CT scans alone.

This is not to say that other modalities aside from CT scans are not valid modalities for assessment of peritoneal carcinomatosis, MRI is often specifically used for surveillance of patients with PC, and similar pre-operative PCI estimations have been performed utilizing PET MRI, albeit with limitations.⁴³ Our use of CT as our primary imaging modality is due to its ubiquitous and generally-high quality images when patients are scanned within the Cancer Center as part of their initial assessment and can be compared more readily to imaging from external facilities. Importantly within this cohort, although a small minority, some externally-obtained pre-operative CT scans were scored for the rPCI assessment. Finally, it has been anecdotally noted, dependent on the subspecialty, surgeon’s increased comfort with interpreting CT images of the abdomen over the equivalent MRI series.⁴⁴

We have also preferred CT scan due to its spatial resolution as MRI, particularly below a 3 Telsa magnet, and F18-Flourodeoxyglucose PET CT (FDG-PET) have a limited spatial resolution when compared to high quality CT. This has the potential to limit detection of smaller peritoneal implants. In addition, PET is not commonly utilized as some histologic variants of peritoneal malignancy do not show FDG uptake.⁴⁵ Furthermore, sites vulnerable to motion artifact (e.g. respiratory motion adjacent to the diaphragm, bowel motion from peristalsis) lead to lower sensitivity for peritoneal implants at these sites.⁴⁵ Motion artifact impacts PET-CT and MRI more than CT images because PET images are typically obtained over a 4-minute period and MRI requires several seconds of breath hold for each sequence. Conversely, modern CT machines can scan the abdomen and pelvis in 4 seconds. Dromain et al. compared CT with PET/CT vs. surgical exploration and reported higher sensitivity for CT compared to PET-CT (82% vs. 57%). In addition, the sensitivity for implants involving the small bowel was 32% for CT vs. 0% for PET-CT.⁴⁵ However, visualization of FDG avid peritoneal lesions has a high specificity (>92%) for peritoneal carcinomatosis.^46,47–49 MRI has no significant difference in overall diagnostic outcomes compared to CT,^{36, 49} but the accuracy in estimating surgical PCI increases when both CT and MRI are utilized.⁵⁰ Therefore, CT is the preferred modality for diagnosing and quantifying peritoneal carcinomatosis due to its accuracy, availability, low cost, and rapid acquisition.⁴⁹

At the authors’ institution, FDG PET-CT and MRI are typically used as problem solving tools in patients with inconclusive CT findings or patients with increasing tumor markers but unremarkable CT in the case of PET scan. While MRI is most commonly used to stage pelvic peritoneal disease when concern for invasion necessitating multivisceral resection is being considered. Our general recommendation for imaging modality is, regardless of the exact technology, there should be structured parameters across high-quality imaging, reliable expertise radiologist and surgeon interpretation, and consistency over time for individual patients to allow for durable serial comparisons.

Here we note significant agreement between surgeon-read and radiologist-read CT scans, particularly those of higher PCI’s. Importantly, rPCI overall as underscored on CT compared to operatively measured PCI. This implies that if both consulting surgeon and reading radiologist read a CT with a rPCI >20, a patient is unlikely to benefit from DL and instead may proceed with definitive medical management. Ultimately this work will require multi-institutional validation of these findings to confirm the trend noted herein and solidify the proposed patient-care algorithm for forfeiting DL for patients with high score rPCI’s. We suggest consideration of a uniform report within radiologist’s dictated impression to include regional and total PCI scoring based upon the system we previously described. Integration of this synoptic reporting structure would allow for more accurate data reporting moving forward and potentially aid in future multicenter study of this clinical dilemma.

Importantly as well, future study need not be limited to high-volume centers only. We suggest that, as centers managing patients with PC for CRC grow, it would benefit the community of surgeons caring for these patients to encourage more rigorous and ordered interpretation of cross-sectional imaging. With more patient scans, more hospitals and more clinicians (surgeons and radiologists alike) committed to this endeavor, we are more likely to be successful with broad application of the rPCI to benefit patients with more personalized surgical and cancer care. Despite our findings, DL continues to play a significant role in staging patients with peritoneal carcinomatosis both in obtaining pre-operative biopsies and washings, the results of which can be indicative of biology, as well as defining resectability beyond simple PCI cutoffs alone -- such as extensive pelvic involvement, involvement of porta hepatitis or envelopment of other critical structures that may limit a resection to above a CCR0 or 1.

This work should be viewed in light of its limitations as a single institution retrospective study within a tertiary referral center. It addresses common problems discussed at tumor boards across the country and we hope it can aid informing and streamlining care of these cancer patients. Unfortunately, a clinical trial to address these questions is unlikely, although we hope that further prospective study of CT in PC will continue to improve care. Referral bias is noted and significant as patients in the surgical cohort had a significantly lower PCI and no patients were noted to cross PCI thresholds of 15 or 20 on radiographic or surgical calculation. This speaks to our ability to select patients but limits our ability to make a more definitive statement regarding reproducibility of these correlations across the disease spectrum. Unfortunately, however, given the patient selection algorithm utilized by our group, no patients in the cohort had an rPCI score >10 to fully delineate the potential for operative exploration to downstage PCI. Another limitation, in part as a function of this being a single institution retrospective study within a high-volume referral center, is the homogeneity of the population in terms of self-reported race and ethnicity. As we move forward with prospective study of this topic and propose multi-institutional involvement, a critical aspect of this recruitment will be to include greater diversity, racial/ethnic and otherwise, to better reflect the heterogeneity of the PC in CRC patient population that we serve. The biologic differences in racial and ethnic groups affected by CRC need further exploration, as our group has recently highlighted in terms of varying driver gene mutations linked to disparities in OS.⁵¹

We have herein found that, among higher-risk stratifications of a PCI score of >15 or >20, there was strong concordance between the surgeon’s and radiologist’s rPCI, underscoring the precision of CT scan in defining disease burden in colorectal carcinomatosis. Moreover, concordance with intra-operative PCI translated to radiographic assessment. While CT lacked accuracy, the rPCI overall underestimated intra-operative PCI, albeit not unanimously. This suggests that rPCI may be a reproducible and conservative tool for assessing peritoneal disease burden. While CT does not “rule in” patients as candidates for CRS, and our group are strong proponents of diagnostic laparoscopy, it does suggest that patients can be reliably “ruled out” as CRS candidates based on common surgical cutoffs. This data has been practice-changing for early-adopting surgeons at our institution, with rPCI being a tool utilized in the clinic and tumor boards to counsel and direct patients on future diagnostics and treatment options including DL. This is important in making surgical recommendations, avoiding unnecessary laparoscopy, and minimizing delays in systemic chemotherapy initiation for clearly unresectable patients.

Synopsis:

CT scans to calculate radiologic PCI (rPCI) as a proxy for intra-operative PCI assessment is proposed. Surgeon’s scrutiny of imaging has good agreement with the radiologist’s read; rPCI >20 may suggest that surgery may not be beneficial.