Assessing Treatment Effect in Pancreatic Cancer

A few histopathologic schemes for grading response to therapy in pancreatic cancer have been proposed by Ishikawa et al,34 Evans et al,24 Pendurthi et al,35 and White et al11 (Table 1).

In 1989, Ishikawa et al34 assessed radiation treatment effect based on radiologic imaging and histologic findings. Histologically, they examined the amount of severely degenerative cancer cells (SDCC) in a series of 18 patients (12 pancreatic, 4 biliary, and 2 ampullary cancers). By their criteria, SDCC had either (1) absent, pyknotic, or irregular-shaped nuclei, or (2) acidophilic or vacuolated cytoplasm. Treatment effect was separated into 3 categories: (1) one-third or less of the cancer cell population represented SDCCs, (2) between one-third and two-thirds of cancer cell population represented SDCCs, and (3) greater than two-thirds of the cancer cell population represented SDCCs. These histologic categories correlated with posttherapy radiologic imaging. Additionally, the authors described 3 histologic patterns: type A, predominantly fibrosis with a few SDCCs, but nonaffected cancer cells undetected; type B, a fibrous connective tissuelike capsule (1–3 mm thick), with predominantly SDCCs at the periphery of the lesion and nonaffected cancer cells in the center; and type C, SDCCs randomly distributed within nonaffected cancer cells. The difference between these histologic patterns correlated with the percentage of SDCCs. Interestingly, in a clinical trial reporting response to neoadjuvant therapy, nearly 77% of the cases met the type A histologic pattern.36 

The main outcomes of the Ishikawa et al34 study were based on operative parameters (a low anastomotic leak rate, no operative mortality, all patients underwent surgery, and less morbidity than a control population of patients not given radiation therapy) and showed that the preoperative radiation was well tolerated. The limited survival data in the initial study reported only one death from cancer 12 months after operation (follow-up period was between 2 and 39 months). In the clinical trial that used the Ishikawa et al34 grading scheme, 13 of 26 cases (50%) demonstrated greater than 80% SDCCs (defined as a major response according to the clinical trial), but the assessment of histologic response was not associated with overall or disease-free survival.25,36 A local relapse was reported in one patient from the clinical trial who had negative resection margins after neoadjuvant therapy, and that patient lacked evidence of a histologic response.25 

The Evans et al24 grading scheme for determining treatment effect is the most widely used. This scheme is based on grading response to therapy in other organ systems.37 This grading scheme consists of a 4-tiered system that assesses the percentage of viable cells remaining in the lesion, with the lowest grade (grade I) representing little or no response (<10% to no tumor cells destroyed) and the highest grade (grade IV) representing the greatest response (no viable tumor cells or acellular pools of mucin); grades II and III represent a spectrum of tumor cell destruction from 10% to more than 90% tumor cells destroyed (see Table 1). The authors felt that necrosis, especially coagulative necrosis, could not be interpreted as evidence of treatment effect because it is commonly observed in untreated tumors.

In practice, although the detection of residual tumor cells is typically straightforward, determining the viability of these cells based on histomorphologic features can be quite challenging, if not impossible. Ishikawa et al34 based their morphologic assessment of viability on the description in the 2nd edition of Robbins and Cotran Pathologic Basis of Disease, which is similar to the description in the 7th edition: Necrotic (irreversibly injured) cells show increased eosinophilia (acidophilic cytoplasm) and 1 of 3 patterns of nuclear changes—pyknosis, karyorrhexis, or karyolysis.38 According to the grading scheme proposed by Evans et al,24 ,viable pancreatic cancer cells can show cytoplasmic or nuclear swelling, multiple nuclei, and cytoplasmic vacuolization, whereas nonviable cancer cells have bizarre, hyperchromatic, pyknotic nuclei usually with swollen, vacuolated, or deeply eosinophilic cytoplasm and exhibit karyorrhexis. Because neither illustrations nor photomicrographs of tumor cell destruction have been published, aside from one photomicrograph of SDCCs in the Ishikawa paper,34 defining cells as viable or nonviable is purely subjective and potentially prone to significant interoberserver and intraobserver variability. This is problematic when attempting to use a grading scheme that includes viability. For example, if a few tumor cells remain, the distinction between an Evans et al24 grade III and grade IV response will depend on the viability of the cells. If the cells are deemed nonviable, it is a grade IV response, but if they are deemed viable, then the response is grade III.

Examples of potential therapy effect in pancreatic ductal adenocarcinoma are shown in Figure 1, A through E. It is rather easy to determine that these cells are significantly injured (SDCC), but are these cells viable or nonviable? Similarly, is the cytopathic effect because of the prior chemoradiation therapy, which can cause damage by both direct cytotoxic effect and treatment-induced ischemia, or is the cytopathic effect related to typical tumor outgrowth of its vascular supply, resulting in ischemic injury? Examples of nontreated pancreatic cancer are shown in Figure 2 (A through C) to emphasize the challenge in distinguishing therapy-induced injury from untreated tumor cells that may or may not be affected by inherent tumor-related ischemic injury.

Although these limitations exist, the Evans et al24 grading scheme has been the most widely used system to assess therapy in various clinical studies and clinical trials. Most patients evaluated by the Evans et al24 grading scheme in published reports have fallen into the grade II or grade III category19,24,28,30,39,40; only a handful of patients in the studies reviewed were reported to have a complete (grade IV) response to preoperative chemoradiation therapy.28,30 One study that correlated histologic response to outcome showed that patients whose tumors demonstrated minimal pathologic effect from neoadjuvant therapy (score IIa) had more than twice the risk of death as did patients with a partial or complete pathologic response (score IIb, III, or IV) (hazard ratio, 2.74; 95% confidence interval, 1.27–5.89; P  =  .01).30 Although treatment effect was found to be associated with overall survival, the authors could not be confident that treatment effect can predict survival because the number of patients in this study was too small.30 In another study, the histologic assessment of treatment effect did not correlate with survival duration by multivariate analysis.41 

Another scheme for grading response to chemoradiation has been proposed by Pendurthi et al.35 They presented their grading scheme at a meeting of the American Gastroenterological Association in 1996, and this scheme has been used by some authors to assess therapy effect.27,42 To the best of our knowledge, this grading scheme has not been published. Per the meeting abstract, this scheme assessed 3 histologic features: (1) the ratio of tumor cells to fibrosis, (2) the percentage of either liquefactive or coagulative necrosis in the tumor, and (3) the presence of treatment effect.35 Characteristics cited as indicative of treatment effect included (1) nuclear gigantism, (2) hypereosinophilic cytoplasm, (3) vacuolated cytoplasm, (4) nuclear dropout and/or degeneration, and (5) cell dropout. In the study analyzing histopathologic outcome after preoperative therapy, multiple sections representing all areas of the tumor were examined, and the fibrosis level was reported as the mean value for the tumor.27 In that study, patients with a significant response to therapy, defined as a ratio of fibrosis to tumor greater than or equal to 80%, had fewer positive margins and fewer positive lymph nodes, but on multivariate analysis, the percentage of fibrosis alone did not affect survival. Another study using this grading scheme found that, in patients who lived longer than 3 years, 3 out of 4 (75%) showed at least 75% tumor replacement.42 

White et al11 created a grading scheme that combined the Pendurthi et al35 and Evans et al24 grading schemes. They modified the fibrosis assessment by categorizing it into 3 grades: mild, moderate, and extensive. Additionally, they examined the presence of necrosis within the tumors. The definition provided by the authors was “lysed cells with necrotic debris.” This assessment is in contrast to the Evans et al24 scheme, which specifically excluded necrosis. Necrosis was assessed as extensive, moderate, focal, or absent, and the presence of necrosis was found to be a worse prognostic factor. For the final assessment, the authors estimated the extent of the tumor mass occupied by residual, viable tumor cells. A large tumor load (>90% viable, which is equivalent to <10% destruction or Evans et al24 grade I) and the presence of tumor necrosis were both worse prognostic factors. Fibrosis was not associated with survival by either univariate or multivariate analysis.

Aside from the inherent variability among observers, the application of these schemes in clinical studies has been somewhat inconsistent compared with the original articles. For instance, some studies adjust the level of fibrosis that is considered significant, whereas other studies adjust the cutoff for SDCCs or interpret necrosis as evidence of treatment effect. In fact, there is limited data to suggest that any of these grading schemes correlate with patient outcomes. Essentially, the grading schemes are reduced to measuring one of the 3 variables: viable tumor cell mass, fibrosis, and/or necrosis.

Tumor cell mass is evaluated in the Ishikawa et al34 scheme by assessing the percentage of cells with SDCC changes. This histologic finding correlated with shrinkage in tumor size by radiologic imaging but does not correlate with outcomes. In assessing tumor cell mass, Evans et al24 referenced Shimosato and Oboshi,37 a 1971 article describing the histologic effects of radiotherapy and chemotherapy on carcinomas from the head and neck. It is unclear whether the assessment of treatment effect for head and neck cancers (usually squamous cell lesions) applies to pancreatic cancer (usually glandular lesions). Evans et al24(p1336) proposed quantifying the treatment effect by determining cell viability. However, no depiction or photographs of viable and nonviable cells was provided. Viable cells are described as those cells “whose morphologic features were fairly well-preserved, including those with swelling of the nucleus or cytoplasm, multiple nuclei, or cytoplasmic vacuolation.” Changes considered irreversible included “cells with bizarre, hyperchromatic, or pyknotic nuclei, usually associated with markedly swollen, vacuolated, or deeply eosinophilic cytoplasm.” Karyorrhexis was considered a sign of cell death, but whether this occurs in untreated cancers was not discussed. In the study by White et al,11 a large, residual tumor load (corresponding to Evans et al24 grade I) was an independent, negative prognostic factor, as were tumor necrosis, lymph nodes with cancer cells, and poor tumor differentiation.11 To our knowledge, no study has examined these histologic variables of viability, tumor load, and necrosis to assess whether they are inherent to the neoplastic process or to the neoadjuvant therapy.

Pendurthi et al35 attempted to quantify the therapy effect by the degree of fibrosis, presumably because fibrosis may replace the treated tumor. However, to date, no guidelines for distinguishing tumor-generating, desmoplastic fibrosis from therapy-induced fibrosis have been proposed. Additionally, in pancreatic cancer, the background pancreatic parenchyma frequently shows fibrosis from the obstructive effect of the lesion. In fact, a particularly interesting observation was made by Ishikawa et al34: The most significant factor that predicted the radioresistance, especially at the periphery of the carcinoma, was the presence of coexisting chronic pancreatitis. They went on to recommend that preoperative radiation be abandoned in patients who show evidence of pancreatolithiasis (a marker of chronic pancreatitis) on radiographic studies. On the other hand, some pancreatic cancers also elicit no desmoplastic response. These differences in the presentation of pancreatic cancer likely represent underlying differences in biology that are as yet not understood. From a clinical perspective, pancreatic cancer is often initially diagnosed based on fine-needle aspiration, precluding evaluation of the architecture in which the pancreatic cancer is situated (whether the neoplastic cells are eliciting a fibrotic response or not). Therefore, determining the response to therapy in patients with pancreatic cancer is extremely challenging.

Another controversial area in assessing the therapy effect is the significance of necrosis. Evans et al24(p1336) disregarded “necrosis, especially coagulation necrosis” as a marker of treatment effect, given its presence within untreated cases, whereas White et al11 felt that necrosis was evidence of a treatment effect, and they found its presence correlated with a worsened median survival.11 Until the tumor microenvironment can be visualized preoperatively, many of these questions will go unanswered.

A confusing issue that further makes the application of these grading schemes challenging is the inconsistency in the literature regarding the assignment of grade to treated cancers throughout the gastrointestinal tract. Several studies have assessed neoadjuvant therapy response in esophageal and rectal cancer (Table 2). Some of these schemes, such as the one proposed by Dworak et al,43 for treated rectal cancer, assign the lowest grade to tumors that show no response to treatment and the highest grade to tumors that show a complete histopathologic response. However, several other popular grading schemes use a reverse classification, with tumors that show a complete histopathologic response being assigned the lowest grade and tumors that show no response being assigned the highest grade.44–48 The presence of these reversed grading schemes has generated confusion.49 As noted in Table 2, both 5-tier and 3-tier grading schemes have been proposed. A scoring system that uses 3 tiers apparently correlates with clinicoradiographic response, predicts prognosis, and is more easily implemented and more reproducible.45,47,48 

The College of American Pathologists recommends the use of a 4-tier system for assessing tumor response to previous chemotherapy and/or radiation therapy. This grading scheme is based on one by Ryan et al47 that examined treated rectal cancers; the scheme of Ryan and colleagues was itself based on a tumor-regression grading system proposed by Mandard et al46 for treated esophageal cancer. Interestingly, the term viable was introduced and not defined by Ryan et al47; the study by Mandard and colleagues46 did not comment on the viability of the tumor cells. For presumed ease and consistency, the College of American Pathologists recommended the same grading system for all pancreatic and gastrointestinal cancers for which neoadjuvant therapy may apply and assigned the best (most complete) response to the lowest grade (grade 0).

As more patients with pancreatic cancer receive preoperative chemotherapy and/or radiation therapy, more surgical resections of treated cancers will be performed. Because the College of American Pathologists recommends that tumor response to previous chemotherapy and/or radiation therapy be reported, surgical pathologists need to determine which grading system they will use in their clinical practice. In addition, until specific markers of treatment effect are developed, surgical pathologists will have to assess therapy effect on routine hematoxylin-eosin stained sections.

At our institution, we have chosen to report treatment effect according to a “simplified,” 3-tiered system that does not require an assessment of viability. This grading scheme, which has not been validated, is a modification of the scheme recommended by the College of American Pathologists with some integration of the Ishikawa et al34 and the well-established Evans et al24 grading schemes (Table 3). If extensive residual tumor is present, there was a poor response to therapy; this is relatively straightforward. If there is evidence of treatment effect, such as residual cancer outgrown by fibrosis, we divide the response into 2 groups: (1) no significant residual tumor (marked response), and (2) residual cancer present (minimal to moderate response). Examples of these 3 different responses are shown in Figure 3 (A through F). For a marked response, rare, single cancer cells, or small groups of cancer cells/glands with marked cytopathic effect, may be present within a fibrotic stroma. This category acknowledges the observation that cases of treated pancreatic cancer, even cases with the most dramatic fibrogenic response, typically have at least rare residual tumor cells present; only rarely, do pancreatic cancers demonstrate a complete absence of tumor cells following neoadjuvant therapy.28,30,36 In the rare instances when absolutely no residual tumor cells are present in the entirely sampled lesion, a note can be added to the report that indicates evidence of a complete response.

In our classification scheme, we define a minimal to moderate response as small groups of cells/glands without evidence of cytopathic effect, cells/glands outside of the main fibrotic mass, or more than 5% of the main fibrotic mass with cancer/glands with or without cytopathic effect. Necrosis is reported separately (Figure 4, A and B). Although a grading system such as this one avoids assessing viability on routine histologic sections, it cannot differentiate therapy effect from baseline tumor ischemia. Although still subjective, this grading scheme is based on a simple, 3-tiered system that should reduce intraobserver and interobserver variability, but this has not yet been studied. In addition, the use of a descriptive classification system should help avoid confusion in assessing the meaning of a numeric grade.49 

Before histologic assessment of treatment effect can be determined, a careful gross examination must be performed with a thorough sampling of the involved pancreas. Sampling of the lesion should be performed in such a manner that adequate comment can be made on the surgical margins of resection in relation to the lesion and one block of tissue per centimeter of lesion should be sampled (to representatively sample the lesion). We typically submit the entire tumor or fibrotic area if it is 3 cm or smaller. We have noted that tumor within the duodenal wall often appears unaffected, so sampling of the duodenal wall, if present, is also performed. Additionally, if no tumor is identified in the initial sections, additional sections are obtained, sometimes requiring the entire pancreas to be submitted to rule out residual neoplastic cells. Because many cases often show dense fibrosis away from the main lesion, care should be taken to accurately provide a size for the lesion and to correlate that with the histologic findings. However, the inherent discrepancies between imaging techniques and gross examination is well recognized.50–52 As recommended by others, we suggest the entire peripancreatic fat be examined for grossly visible lymph nodes and the remainder of the tissue be submitted for microscopic lymph node examination.53 Although a precise guideline for the number of lymph nodes to be sampled has not been established, examination of 15 lymph nodes appears to be optimal for accurate staging of node-negative pancreatic cancer after pancreaticoduodenectomy.54 

By reviewing the pathologic grading schemes that have been published in the literature and providing illustrations and guidelines for grading response to preoperative therapy in pancreatic cancer, we hope we have provided useful information that surgical pathologists can incorporate into their daily practices. Accurate reporting of treatment effect is essential, not only to the patients whose specimens are being assessed but also to help determine which clinical and/or pathologic variables may affect the adequacy of response so that optimal therapies can be developed in the future.