Peritoneal regression grading score (PRGS) in peritoneal metastasis: how many biopsies should be examined?

Introduction

Peritoneal metastasis (PM) is common in gastrointestinal and gynecological cancer patients, with an estimated incidence of one million cases/year worldwide [1]. Upon diagnosis, current therapy guidelines recommend palliative systemic chemotherapy or the best supportive care [2]. Due to the limited soft-tissue contrast and the mobility of the small bowel, CT depiction of small peritoneal tumors remains unsatisfactory [3, 4]. Against this framework, surgical oncologists increasingly use laparoscopy to assess the extent of tumor burden at initial diagnosis and to obtain tissue biopsies for therapy response assessment during treatment [5].

The peritoneal regression grading score (PRGS) was proposed in 2016 for reporting the regressive changes under palliative chemotherapy in PM of gastrointestinal and gynecological origin [6]. The PRGS considers two main tumor regression criteria: partial or complete disappearance of malignant cells and fibrous or fibro-inflammatory tissue replacement. It also includes acellular mucin pools and infarct-like necrosis (ILN) as additional features of response [6]. PRGS was shown to be reproducible for assessing the response to intraperitoneal chemotherapy in PM [6]. The best reproducibility was achieved by the mean, not the maximal PRGS.

In the meantime, multiple clinical trials and cohort studies used the PRGS as a surrogate outcome criterion in intraperitoneal chemotherapy. These trials included ovarian cancer [7], gastric cancer [8], colorectal cancer [9], hepatobiliary-pancreatic cancer, and various histologies [10–14]. Combining the highest PRGS with peritoneal cytology (so-called CPI) was independently associated with worse overall and progression-free survival [15]. This finding follows the initial proposal of the PRGS authors, who recommended sampling peritoneal fluid for cytological analysis in the presence of tumor scarring or the absence of macroscopic peritoneal lesions [6].

However, different peritoneal biopsies in other locations might deliver different PRGS scores due to tumor clonality, cell cycle phases, or various expositions to chemotherapy. Moreover, due to intraabdominal adhesions, it is not always possible to take biopsies in every four abdominal quadrants, and the minimal number of biopsies needed to determine PRGS remains unknown.

We aimed to examine the minimum number of biopsies required to obtain reliable information on therapy response using the PRGS score and provide a corresponding, evidence-based recommendation.

Materials and methods

Study flow

The study flow is displayed in Figure 1. To quantify the loss of information caused by missing biopsies, it was first necessary to decide on a reference value (gold standard). In the absence of an external comparator, we chose the gold standard to be the mean PRGS in all patients, where results from four biopsies from all abdominal quadrants were available. The loss of information was calculated by excluding one, two, and three biopsies. All combinations were tested, and the mean was calculated. In parallel, the mean difference in PRGS between consecutive chemotherapy cycles was documented. When the loss of information outweighed 95% CI of the absolute difference in PRGS observed between repeated biopsies, the diagnosis of regression vs. progression was considered false (Figure 2).

Figure 1:

Study flow. We determined the mean PRGS from all four peritoneal biopsies, and the loss of information (LOI) associated with one and two biopsies missing (left part). In parallel, the mean difference in PRGS between consecutive chemotherapy cycles was documented (right part). When the loss of information outweighed the 95% CI of the observed difference, the diagnosis of regression vs. progression was considered false. Δ PRGS = difference in mean PRGS between consecutive therapy cycles.

Figure 2:

Histological tumor response and informational loss induced by missing biopsies. The tumor response is measured as the absolute difference of the mean PRGS between two therapy cycles. Progression of disease (PD) is defined by an increase in the mean PRGS (red area). Tumor response (green area) is defined by an unchanged PRGS (stable disease, SD), a decrease in the mean PRGS (partial response, PR), or the absence of vital tumor cells in all biopsies (PRGS 1, complete response, CR). The pink curve shows the diagnostic uncertainty induced by one missing biopsies (= only three biopsies available), the blue curve by two missing biopsies (= only two biopsies available). Diagnosis of tumor progression is wrong when the diagnostic uncertainty is superior to the difference in the mean PRGS.

Results

A cohort of 89 patients with PM of various origins treated with PIPAC was available for analysis (Table 2). Altogether, 641 biopsies from 210 laparoscopies were available for histological examination. Fifty-four patients had at least two laparoscopies (= PIPAC cycles) and were eligible for tumor response assessment.

Discussion

The lack of a radiological gold standard highlights the challenges for assessing tumor response in PM patients. A histological score, the PRGS, was developed to address this challenge. PRGS is a simple four-tied system [6], is reproducible [17], and has been used in independent cohorts on an international level [7]. PRGS is tailored to peritoneal metastasis, taking into account their frequent mucinous nature [6]. In analogy to the revised RECIST guideline [18], histological PM progression was defined as an increase in mean PRGS between therapy cycles and response to therapy as a decrease or stable mean PRGS [13].

This study aimed to examine the minimum number of peritoneal biopsies required to obtain a reliable diagnosis of tumor progression with PRGS and provide a corresponding, evidence-based recommendation. For this purpose, we determined the information loss per missing biopsy and compared this information loss with the difference in PRGS observed between therapy cycles (histological tumor response). We calculated that a minimum of three peritoneal biopsies is required to reach at least 80% accuracy for determining tumor progression using PRGS. Against this framework, it is essential to note that accuracy is not recommended as a quality metric for diagnostic tests. The cut-off of 80% used for clinical significance is arbitrary. Specificity and sensitivity metrics show that missing biopsies often result in false-positive results and, thus, false diagnosis of non-response. This knowledge is essential for proper clinical decision-making, and pathologists should be cautious about diagnosing tumor progression with less than four peritoneal biopsies.

We decided to use the mean PRGS obtained from all biopsies and not the maximal PRGS from a single biopsy to prevent loss of information. Dropping three out of four biopsies (=75% of the information available) might cause an unacceptable risk of bias. Our preference is in line with the prior art in regression grading, recommending considering all sections of the tumor bed rather than the worst section (highest grade) [19].

Prior work has highlighted that spatial and temporal tumor heterogeneity increases during disease progression, probably because of clonal selection [20]. Tumor heterogeneity is a significant reason for chemoresistance [21], which in turn worsens patients’ prognosis. Therefore, since PM is a manifestation of advanced disease, consideration of tumor heterogeneity appears essential to evaluating tumor response to chemotherapy and prognostic assessment. Against this framework, there is an obvious need for a uniform sampling method for determining tumor regression grading [19]. In particular, a minimal number of biopsies needs to be determined that consider tumor heterogeneity between sampling locations. To our knowledge, this is the first study showing that the histological analysis of a single or even two tumor biopsies does not provide accurate information on the response of PM to chemotherapy.

The current lack of evidence on the number of biopsies needed to classify histological tumor regression is in sharp contrast with the binding recommendations of the UICC on the minimal number of lymph nodes needed to determine the pN category in tumor staging. When a minimal number of lymph nodes are not available for microscopic analysis and determination of vital tumor cells, then the pN category cannot be safely determined, and the case might be classified pNx. In analogy, based on our data, we recommend ranking PRGS as undetermined (“PRGSx”) in any case where a minimum of three peritoneal biopsies is not available.

In contrast to other studies [22], no external comparator such as, for example, a surgical specimen was available for this study. RECIST criteria were not considered as a reliable external comparator either, since they are, to our knowledge, never been validated for assessing the response of peritoneal metastasis to chemotherapy. As an alternative, we decided to use an internal comparator as the gold standard and calculated for this purpose the mean PRGS in patients with histological information from all abdominal quadrants. Then, it was easy to calculate the loss of information per missing biopsy and correlate this value with the difference of PRGS observed between therapy cycles.

In summary, this study indicates that the use of PRGS may address the need for reliable, objective, reproducible tumor response assessment in patients with PM. PRGS adds valuable information to the RECIST 1.1 criteria, which can barely reflect intraperitoneal tumor heterogeneity. Notably, this is the first study to our knowledge to investigate the minimal number of biopsies required to assess histological tumor response under palliative chemotherapy. Our results provide compelling evidence for intraperitoneal tumor heterogeneity and various degrees of tumor response depending on tumor location. Thus, PRGS appears effective for assessing the response of PM to chemotherapy.

Future research should follow three avenues. First, prospective cohort studies are needed to validate PRGS as a predictive or prognostic parameter in PM of various origins. Second, the potential role of additional staining in PRGS1 (no vital tumor cells) cases should be investigated. Last but not least, molecular parameters might be integrated into the PRGS, paving the way for individualized therapy of patients with PM and improving the prognosis of this challenging patient population.