Diagnosis of “cribriform” prostatic adenocarcinoma: an interobserver reproducibility study among urologic pathologists with recommendations

Introduction

The term “cribriform” is derived from the Latin word cribrum (i.e. sieve). It describes glands composed of sheets of tumor cells that form cohesive rounded or irregularly shaped trabeculae with perforations or multiple “punched out” lumina [1,2]. There is growing evidence that cribriform pattern 4 (CrP4) prostate adenocarcinoma (PCa) in both biopsies and radical prostatectomies (RP) is associated with adverse clinical outcomes, including worse biochemical recurrence-free, metastasis-free, and cancer-specific survival than those without [3,4]. Specifically, CrP4 carries a much higher risk of disease progression compared to the other Gleason grade 4 patterns [5-7]. Among men with grade group (GG) 2 at biopsy, some studies have demonstrated a higher risk of failure in the presence of cribriform morphology, while patients without cribriform architecture carried the same risk as GG 1 cancer, implying that CrP4 diagnosis might affect clinical decision-making [4,8-10]. Whereas the value of cribriform architecture has mostly been studied for Gleason score (GS) 7 patients, some studies have demonstrated its independent prognostic value in men with GS 8 and GS 9-10 PCa [6,11,12]. However, the majority of studies addressing the significance of cribriform carcinoma have not distinguished between intraductal carcinoma of the prostate (IDC-P) and CrP4 to determine outcomes, and have referred to these lesions collectively as “cribriform growth” [2-4,10]. Cribriform PCa also has unique adverse molecular features with significantly more frequent PTEN and p27 loss at RNA and protein levels, defects in DNA repair genes, and higher frequency of mutational events similar to metastatic prostate cancer as opposed to non-cribriform Gleason pattern 4 PCa [3,13-17]. Due to multiple lines of evidence supporting the association with adverse outcomes, both the Genitourinary Pathology Society (GUPS) and the International Society of Urologic Pathology (ISUP) recommended reporting of CrP4 in needle biopsy and RP specimens [1,18].

Although the cribriform pattern has the best interobserver reproducibility among genitourinary pathologists (ranging from 54%-79%) compared with other Gleason 4 patterns, there is still a significant variability and ambiguity in diagnosing this pattern [19]. Specifically, differentiation of CrP4 from complex fused glands, complex glands with papillary proliferation, glands with partial or roman bridging, complex “glomeruloid-like” growth pattern, or PCa exhibiting complex cribriform-like morphology but with intra- or extraglandular mucin or involving a nerve creates significant challenges in classification, diagnostic reproducibility, and reporting of CrP4. In this study, we focused specifically on the classification and practice-related issues that may impact the diagnostic reproducibility of CrP4. Our objectives were to assess the diagnostic reproducibility amongst urological pathologists with expertise in prostate cancer using a wide range of potential CrP4 lesions, identify morphological features that are associated with consensus for and against CrP4, and provide recommendations for diagnostic criteria for CrP4.

Materials and methods

Results

Discussion

A growing body of evidence suggests that cribriform morphology may potentially affect clinical decision-making in PCa management. van Leenders et al. have recently proposed significant modification to the Gleason system by incorporating cribriform morphology to improve its prognostic utility [22]. In the current study, nearly 85% of participants indicated that they routinely report the presence or absence of CrP4 in prostate biopsies and RP specimens, indicating an acceptance of CrP4 as an adverse pathological feature that should be reported. Using any morphological marker for important clinical decision-making requires that its diagnosis should be reproducible among pathologists. The overall interobserver agreement for a diagnosis of CrP4 by a group of urological pathologists was only fair (k=0.40), which is lower than the reproducibility (53-79%) reported by earlier studies [19], due to the inclusion in this study of more difficult cases. Nevertheless, it suggests that there is significant interpretational disagreement in diagnosing cribriform PCa. To address these issues, we correlated various morphologic features of CrP4 with consensus diagnosis, attempting to extract the key morphological features that urologic pathologists rely on to diagnose or rule out CrP4.

Cribriform cancer pattern is defined as sheets of tumor cells that form cohesively rounded or irregularly shaped trabeculae with perforations or multiple “punched out” lumina. Cribriform glands may be spherical or oblong and may sometimes have irregular borders [1]. However, this definition does not provide clear guidance regarding how to differentiate CrP4 from complex fused glands, complex glands with papillary proliferation or partial bridging, complex glomeruloid-like growth pattern, or PCa exhibiting complex cribriform-like morphology but with intra- or extraglandular mucin. In order to provide guidance regarding these issues, there have been recent attempts to improve the histological definition of the cribriform gland, specifically pertaining to CrP4. Van Leenders et al. proposed a definition of cribriform architecture as an epithelial sheet in which the majority of tumor cells do not contact the surrounding stroma, with a gland-like space surrounding less than half of the sheet circumference, and with regular intercellular lumens clearly visible on hematoxylin and eosin (H&E)-stained sections [2]. Recently, van der Kwast et al. proposed a consensus definition of CrP4 as “a confluent sheet of contiguous malignant epithelial cells with multiple glandular lumina that are easily visible at low power. There should be no intervening stroma or mucin separating individual or fused glandular structures” [23]. To derive this consensus definition, the Delphi method was applied among the 12 panelists on a set of 32 images during 2 initial rounds of the study. Using the same set of 32 images, in a subsequent study they analyzed morphological features to identify those that define the essence of the CrP4 [24]. They found the presence of intervening stroma, mucin, predominant papillary pattern, or an irregular outer boundary detracted participants from calling CrP4 while all consensus cases had ≥9 lumens.

Compared with the two aforementioned publications, this study has several important differences and strengths. First, more (27) pathologists participated in this study, therefore, this study represents a broader opinion of expert urologic pathologists’ diagnostic criteria of cribriform lesions. Second, more images (60) with a much broader morphological spectrum were used to represent a wide range of lesions that may be considered CrP4, from unequivocal CrP4 to those with morphology overlapping between CrP4 and non-cribriform pattern 4, pattern 3, or 5. These lesions included variably sized PCa glands exhibiting “glomeruloid-like” morphology, those with intraluminal cellular proliferation with a variable number of attachments to the peripheral layer of the gland, prominent mucinous features (specifically mucinous fibroplasia, mucin extravasation, or extracellular mucin pool), cribriform-like morphology with a central nerve, poorly formed “rosette-like” formations without clear lumina, papillae with fibrovascular cores, and complex fused glands (Figures 1, 2, 3 and 4). Third, more morphological features were investigated to identify that were associated with consensus for and against CrP4 diagnosis, including those that were not addressed by the ISUP study, such as glandular bridging (transluminal vs. partial), number of contacts of intraluminal cells with gland periphery, whether peripheral luminal space occupies more or less than 50% of glandular circumference and whether a nerve was present within the cribriform lesions. In addition to following two morphological features: a lack of contact of the intraglandular cells with stroma and lack of intraglandular mucin, we found that transluminal bridging and a clear luminal space along the periphery of the gland accounting for <50% of the glandular circumference were two additional important criteria for CrP4 diagnosis (Figures 1 and 2). Specifically, all cases that reached consensus for CrP4 demonstrated transluminal bridging, a lack of contact of the intraglandular cells with stroma and a clear luminal space along the periphery of the gland accounting for <50% of the glandular circumference. These features are not new diagnostic criteria for CrP4; rather they identify a group of cribriform lesions that have achieved consensus among GU pathologists. Importantly, we also identified diagnostic features that were associated with the consensus against CrP4: partial bridging, a majority of intraglandular cells in contact with the stroma, mucinous fibroplasia, only one attachment of tumor cells to gland border for “glomeruloid-like” proliferation, and a clear luminal space along the periphery of gland accounting for >50% of the glandular circumference (Figure 3). Two cases showing partial bridging were diagnosed as “against CrP4”, and all 3 cases exhibiting complex mucinous fibroplasia also reached a consensus against CrP4 (Figure 3). Therefore, our study not only attempts to address issues pertaining to the definition but also for interpretation aspects of CrP4. We show that the diagnosis of CrP4 is not only based on the presence of the classic definitional features of the CrP4 diagnosis but importantly depends on ruling out features against the CrP4 diagnosis. One may still classify a lesion as CrP4 if all five morphological features associated with CrP4 consensus are not present as long as it lacks features against CrP4 consensus diagnosis.

In some studies, CrP4 has been divided into small and/or large CrP4, based on its size in relation to adjacent benign glands, the number of lumina, and the greatest cross-sectional diameter [6,21,25]. Specifically, a cribriform gland twice the size of the adjacent benign glands and exhibiting >12 lumina has been proposed as large CrP4 [6,21]. Hollemans et al. used a criterion of twice the size of the diameter of adjacent benign glands and demonstrated that, in a multivariable analysis, only large CrP4 was an independent predictive factor for biochemical recurrence-free survival in GG 2 patients [21]. However, most studies, when assessing for outcomes associations, have either not differentiated CrP4 based on size, or have utilized varying definitions for the size cut-off. In this study, we showed that large CrP4 had a better agreement (k=0.49) compared with small CrP4 (k=0.40). However, we found that the size of CrP4, when compared with the adjacent benign glands, or defined based on the number of lumina, was not associated with a consensus diagnosis for CrP4 or non-CrP4. This is further evident from the variable responses on this issue as only 26% of participants indicated that they routinely distinguish between small vs. large CrP4. For these reasons, in our opinion, routine reporting of CrP4 as two separate categories (large or small) requires additional studies demonstrating its value of such division.

We have also demonstrated that for certain PCa morphologic patterns, there remains considerable diagnostic variability among experts. Examples in this category included variably sized “glomeruloid-like” PCa glands with more than one point of attachment to gland periphery, complex cribriform-like proliferation around nerve and PCa with extravasated mucin imparting complex patterns (Figure 4). The glomeruloid pattern has been defined as dilated cancer glands of variable size, with a cellular mass protruding into the lumen yet not attaching to the opposite side of the gland wall (one attachment to gland periphery), superficially resembling a glomerulus [20]. Notably, because of its resemblance to the cribriform pattern, particularly when the gland is large, some have suggested that a glomeruloid pattern may represent an early stage of cribriform cancer [26]. However, subsequent studies have suggested that a glomeruloid pattern, regardless of size, has outcome associations intermediate between pattern 3 and CrP4 [4,5,7]. For this “glomeruloid-like” category, two examples with a single attachment of tumor cells to the gland periphery were uniformly associated with consensus for a glomeruloid pattern 4 but for 9 lesions (two classified as small gland and 7 as large glands) showing multiple contacts to the gland border had no consensus for CrP4 vs. non-cribriform pattern 4 diagnosis (Figure 4A). In a significant number of these lesions (89%), a clear luminal space along the periphery of the gland that occupied >50% of the glandular circumference was present, a feature that in this study was correlated “against CrP4” consensus diagnosis (Figure 4A). Another significant area of diagnostic variability was the presence of a nerve in the center of the PCa glands showing complex architecture that resembles CrP4 (Figure 4B). Of 5 such examples, one case reached a consensus for Gleason pattern 3, while the remaining 4 cases had no consensus. Interestingly, two participants indicated that they do not grade components of PCa in regions of perineural invasion. Traditionally, some experts have cautioned against over-grading when the tumor is present around the nerve and recommend a conservative approach, particularly in needle biopsy. One case had all and remaining 3 had variable of five diagnostic features of CrP4. Importantly, all lacked features that were associated against CrP4 diagnosis, arguing that CrP4 can be encountered around nerve. Finally, all 6 cases of PCa exhibiting mucin extravasation had no consensus classification (Figure 4C). It is well known that PCas with extracellular mucin, extravasated mucin, and/or fibroplasia may represent patterns 3 or 4, dependent upon the complexity of the glandular structures, and the overall grade of these lesions should be based on the underlying architecture [20]. Interestingly, the one included example of PCa with cribriform architecture embedded in extracellular mucin pools did reach consensus for CrP4 (Figure 2A). This finding is similar to the architectural pattern described by McKenney et al., which they termed “mucinous adenocarcinoma with marked epithelial complexity, either solid or complex confluent cribriform” [7]. In their study, this specific mucin pattern was associated with aggressive outcomes, compared to other mucin patterns which tend to have better outcomes [7,20].

The current study has a few limitations. First, the study was conducted using prospectively collected cases and was enriched for difficult-to-classify cases, which may not represent the routine morphologic spectrum of CrP4. Likewise, the survey did not address the issue of extent of CrP4, which may potentially influence outcomes. Despite these limitations, this study showed that certain morphologic features are quite reproducible for the diagnosis of CrP4 and may be used to establish consensus diagnostic criteria for CrP4. The recommendations from this study are provided in Table 3. We also found that there is considerable overlap between the spectrum of CrP4 within certain morphologic patterns with the potential for over-grading or diagnosing CrP4. Finally, further validation of our proposed diagnostic criteria for CrP4 with cases that are annotated with clinical outcome data would be desirable.

In summary, we have demonstrated only fair diagnostic agreement amongst urologic pathologists for CrP4 diagnosis, suggesting a need for standardization of diagnostic criteria. We identified the following morphologic features associated with consensus CrP4 diagnosis: “a dense sheet of tumor cells forming multiple lumens with transluminal bridging, imparting a “sieve-like” architecture, in which a majority of intraglandular cells are not in direct contact with stroma or mucin, and a clear luminal space along the periphery of gland accounts for <50% of the glandular circumference”. We hope that these suggestions will represent a step toward standardization of prostate cancer grading based on reproducible morphologic features.

Disclosure of conflict of interest

None.