Interobserver, intraobserver, and interlaboratory variability in reporting pT4a colon cancer

A considerable group of colon cancer patients presents with T4 tumors. The T4 category represents the most advanced category with respect to local invasion [1] and is related to a high risk of developing peritoneal metastases [2, 3]. Pathological (p)T4 includes two main entities of locally advanced growth, categorized as pT4a (peritoneal penetration) and pT4b (adjacent organ/structure invasion) according to the TNM 8th ed. [4]. Intensified treatment strategies for patients with pT4 colon cancer are currently under investigation, including adjuvant hyperthermic intraperitoneal chemotherapy (HIPEC) and second-look laparoscopy, aiming at prevention and early detection of peritoneal metastases [5‐7]. Furthermore, current clinical guidelines recommend adjuvant chemotherapy in stage II colon cancer in the presence of pT4 [8]. Based on a recent pooled analysis from six clinical trials, pT4 is now used to inform the duration of adjuvant chemotherapy in stage III colon cancer [9]. Thus, the pT4 diagnosis is becoming an increasingly important parameter for patient management.

Parameters used for clinical decision-making should be reliable and reproducible. Pathologists have been aware of the problems with regard to the pT4 category [10‐12], with peritoneal penetration (pT4a) being a less straightforward diagnostic feature than often assumed. The TNM definition implies that full penetration of tumor cells through the peritoneum is required for pT4a. However, some national guidelines on histopathology reporting of colorectal cancer also allow tumors that are close to or at the peritoneal surface to be regarded as pT4a [13, 14]. This practice originates from studies showing that tumors that are close to or at the peritoneal surface, especially in combination with certain reactive changes, also carry negative prognostic impact and risk of peritoneal metastases similar to tumors showing full peritoneal penetration [11, 15]. From these data, it has been concluded that certain features could be allowed to represent pT4a in order to prevent underdiagnosis of pT4a [4, 10, 11, 15, 16]. Another difficulty with regard to the pT4a category is that microscopic detection of pT4 is dependent on macroscopic specimen evaluation and meticulous sampling of resection specimens [10, 12]. With the increasing clinical importance of the pT4a category, it should be clear how pathologists currently diagnose and define pT4a. Studies exploring current pathological practice and variability in diagnosing pT4a colon cancer are lacking.

The aim of this study was to evaluate inter- and intraobserver variability in diagnosing pT4 among pathologists on preselected slides, focusing on the distinction between deeply invasive pT3 and pT4a. Furthermore, interlaboratory variability and the average number of blocks taken from pT3 tumors compared with those from pT4a tumors for microscopic analysis in case of histologically verified metachronous peritoneal metastases were determined.

In this study, we explored the current pathological practice of diagnosing pT4a colon cancer. The main findings are a moderate interobserver agreement in distinguishing pT4a from deeply invasive pT3 and a substantial intraobserver agreement. Most disagreements were found in cases with tumor cells at a distance of 0–25 μm to the surface, but even preselected slides with tumor cells on the peritoneal surface resulted in disagreement. Also, after adjustment for case mix, the proportion of pT4a colon cancers differed significantly between the median laboratory and eight other laboratories (24%). Furthermore, pT3 tumors from patients who subsequently developed peritoneal metastases were diagnosed using significantly lower number of tissue blocks than for the diagnosis of pT4a tumors from patients who developed peritoneal metastases. Our findings highlight inconsistencies in diagnosing pT4a in colon cancers. Because the pT4 category increasingly bears clinical and therapeutic consequences, there is an urgent need for a better definition of the pT4a category, which can be applied by pathologists in a reliable and reproducible fashion.

Studies evaluating the practice of diagnosing pT4a colon cancer at a pathologist and/or laboratory level are scarce. Littleford et al. [23] determined interobserver variability of the four categories of Shepherd’s LPI classification, using a single-center cohort of 138 cases of pT3/pT4 cases. Kappa values ranged between 0.45 and 0.64 among four pathologists assessing the separate LPI categories, which is similar to the present study [23]. Agreement increased when only LPI1–2 had to be discriminated from LPI3–4, with Kappa values ranging from 0.74 to 0.89.

Detailed macroscopic examination and subsequent extensive sampling of suspected areas have been described to likely improve the accuracy of pT4a assessment [12, 17], although studies on the subject are still limited. Our study is the first to show that pT4a is likely missed in some pT3 cases due to inadequate sampling. In a recent guideline from the Royal College of Pathologists, UK, it is recommended that a minimum of 5 blocks should be taken from the primary tumor for an accurate assessment of various pathology parameters [14]. Data from the present study appear to support this, with an average of approximately 5 tissue blocks being taken from primary pT4a tumors that developed peritoneal metastases, as opposed an approximate average of 4 tissue blocks from pT3 tumors that developed peritoneal metastases. Currently, the Dutch colorectal cancer guideline does not contain any recommendation on the minimum number of blocks from the primary tumor for accurate assessment of pT status. In the current study, 4 tumor blocks or less, i.e., less than the minimum of 5 blocks according to the UK guideline, were submitted in 61% of analyzed cases (183/299). Also, the low frequency of pT4a in some of the laboratories (Fig. 6) likely represents underdetection. An optimal pT4a detection rate for stage II and III tumors is not known but a threshold of 20% for pT4a+b in all stages combined has been recommended in the UK [4]. The current data raise the question if a minimum number of blocks from the primary tumor for pT assessment need to be put forward in the Dutch guidelines and implemented as an audit and quality criterion, similar to the minimum number of examined lymph nodes.

In this study, we observed different approaches among pathologists to diagnose pT3 or pT4a. Various criteria were applied, some of which being highly subjective and often inconsistently used. Most discrepancies among pathologists can be traced back to two main schools of thoughts, one of them being more strict and requiring cells to be growing through the peritoneum and be present on the peritoneal surface (Fig. 1, LPI4), while the other is less prescriptive with pT4a already being considered when tumor cells are close to or at the peritoneal surface (LPI2/3), especially in the presence of particular reactive changes.

Attempts have been made to further define the pT4a category in the literature. The main UICC and AJCC TNM definition of pT4a, i.e., tumor that perforates visceral peritoneum, has hardly changed during the last 20–30 years except that in the AJCC 7th and 8th edition, the word perforates was exchanged with the word penetrates and invades, respectively. In some national guidelines, further details on the definition of pT4a have been put forward. In both the US and UK guidelines, this is based on the LPI categorization by Shepherd et al. [15, 17]. In the College of American Pathologists guideline from 2017, the pT4a definition includes tumor present at the serosal surface and free tumor cells on the serosal surface with underlying erosion/ulceration of mesothelial lining, mesothelial hyperplasia and/or inflammatory reaction (roughly equates to LPI3 and 4) [13]. In the current UK guideline from 2018, pT4a is defined as tumor cells visible either on the peritoneal surface, free in the peritoneal cavity or separated from the peritoneal surface by inflammatory cells only (also roughly LPI3 and 4) [14]. In the Netherlands, the national guideline on colorectal cancer does not give any details on the definition of pT4a beyond the TNM literature (www.​oncoline.​nl). The LPI classification might be suboptimal because the distinction between LPI2 and LPI3 may be unclear in many cases where tumor cells approach the peritoneal surface. Also, a gradient in prognostic impact within the LPI3 category has been described [24]. Recently, some authors have proposed that colon cancers 1 mm or less from the serosal surface should be regarded as pT4a when additionally displaying certain features (serosal fibroinflammatory reaction; peritumoral abscesses that communicate with the serosa; serosal hemorrhage; and serosal fibrin) [11]. Other authors have proposed that invasion beyond the peritoneal elastic lamina should be regarded as pT4a [25]. Various proposed additional criteria for diagnosing pT4a may, however, contradict one of the general rules of the TNM system that states that if there is doubt concerning the correct T, N or M category to which a particular case should be allotted, then the lower (i.e., less advanced) category should be chosen [4]. Variation in guidelines and literature suggestions regarding the pT4a definition has likely contributed to the confusion in what represents pT4a [26].

Defining pT4 has been mainly based on survival, and seldom on the risk of metachronous peritoneal metastases. The upcoming treatment strategies for peritoneal metastases justify consideration of the risk of peritoneal metastases when defining pT4. Shepherd [15] reported that peritoneal recurrences all occurred in the LPI3 and 4 group except for one case in the LPI2 group (1%). In a recent series [27] of 159 patients, the 5-year peritoneal recurrence rate was 33% for true peritoneal penetration (LPI4), as opposed to 21% (p = 0.057) for peritoneal reaction with tumor less than 1 mm from the peritoneum (LPI2–3). Using peritoneal scrape cytology [11], tumor cells were found in 46% and 55% of the deeply invasive pT3 and pT4a tumors (vs. 19% in all pT3), translating into peritoneal recurrence rates of 11% and 18%, respectively. These data show that the presence of tumor cells on the peritoneal surface carries a higher risk of peritoneal metastases than when tumor cells are close to or at the surface but without full penetration. Although the risk of local, peritoneal, or systemic recurrence is also increased in these deeply invasive pT3 cases, it remains unclear whether that justifies including them into the pT4a category.

There are some limitations related to the present study. Regarding the interobserver variability analysis, the selected samples might not have been a realistic representation of daily clinical practice, also since deeper levels and analyzing/adding more tumor sections was not possible in the present research setting, potentially leading to an underestimation of interobserver agreement. We rather choose this design in order to identify pitfalls and points of attention. In addition, virtual slide analysis requires training and may be less efficient on this kind of material. In the interlaboratory analysis, the use of the median laboratory as a reference is convenient for describing the level of variation that exists between laboratories [20, 21]. This study does not attempt to provide information on which frequency of pT4a would be optimal. Finally, the sampling analysis might be confounded by for example fewer blocks being taken from areas grossly suspicious of T4a and by variability in microscopic assessment. Despite these shortcomings, we were able to demonstrate a difference in the number of tissue block between pT3 and pT4a cases, which may even strengthen our finding. It should also be mentioned that the frequency of peritoneal metastases is most likely underreported in this series, as in daily clinical practice, not all peritoneal metastases are histologically confirmed.

We conclude that the current pathology practice leaves room for subjectivity and variable interpretation when distinguishing pT3 from pT4a colon cancer. Also, the current literature on the topic is limited and does not offer enough data on how pT3 and pT4a should be distinguished. Considering the potential therapeutic and prognostic implications, the reproducibility of pT4a diagnosis should be improved, both with regard to sampling and microscopic assessment. Especially, the gray area of peritoneal involvement should be clarified with explicit criteria to distinguish pT4a from pT3. To achieve this, future research should aim at assessing the histopathology of pT3-pT4a within clinical trials with detailed follow-up regarding peritoneal recurrences.