CK20 and p53 Immunohistochemical Staining Patterns in Urinary Bladder Specimens With Equivocal Atypia: Correlation With Outcomes Open Access

This study was reviewed and approved by the Henry Ford Health System Institutional Review Board. The surgical pathology database was searched for bladder specimens with equivocal urothelial atypia from 2008 to 2015. We retrieved and analyzed 69 specimens from 65 patients from our surgical pathology files, including biopsies and transurethral resections of bladder tumor. CK20 and p53 immunohistochemistry was performed to analyze staining pattern and intensity at the time that the initial diagnosis was rendered. Upon histologic and clinicopathologic review, cases were included in the study if the immunohistochemical staining was used to evaluate atypia in the setting of a flat urothelial lesion. Using 4-μm–thick tissue sections, immunohistochemistry was performed using a Dako Link Autostainer automated system (Dako Corp, Carpinteria, California). Immunostains were performed for anti-CK20 monoclonal mouse antibody (1:50; Dako), and anti-p53 monoclonal mouse antibody (1:400; Dako). Hematoxylin was used as a counterstain.

Staining pattern for CK20 was defined as: (1) negative—negative staining, or labeling of only the superficial “umbrella” cells; (2) partial or equivocal—patchy or partial-thickness labeling; and (3) positive—strong and uniform labeling of the area of concern (full thickness, or otherwise, dependent on the distribution of the abnormal cells). Staining pattern for p53 was defined as (1) negative—no staining or predominantly basally located positivity; (2) partial or equivocal—patchy labeling, with some cells above the basal layer showing strong positivity; and (3) positive—strong and diffuse labeling of the area of concern (full thickness or otherwise, dependent on the distribution of the abnormal cells). Combined staining interpretation of the overall case was classified as CIS pattern (both stains positive), discordant (1 of the 2 stains yielding a positive pattern but not the other), indeterminate (1 or both yielding partial or equivocal labeling but not meeting the criteria for positive), or benign (both stains yielding a negative pattern). Prior and subsequent bladder specimen diagnoses were also reviewed, and follow-up was assessed using the electronic medical record. Statistical analysis with t test and Fisher exact test was performed and a P value of <.05 was considered statistically significant.

The pathologic characteristics of the overall study population are shown in Table 1. There were a total of 69 specimens from 65 patients with equivocal atypia that were retrieved and analyzed for histomorphology and for CK20 and p53 immunohistochemical staining pattern (Figure).

Twenty specimens (29%) had positive staining, 11 (16%) had patchy or equivocal staining, and 38 (55%) had negative staining or labeling of the “umbrella” cells only.

Sixteen specimens (23%) had positive staining, 41 (60%) had patchy or equivocal staining, and 12 (17%) had negative staining or labeling restricted only basally located cells.

There were 9 specimens (13%) that had a CIS staining pattern (both stains with positive pattern), 18 (26%) were discordant (1 stain yielding a positive pattern but not the other), 31 (45%) were indeterminate (partial or equivocal positivity but no diagnostic positive pattern), and 11 (16%) showed a benign staining pattern. Of the discordant category, 13 specimens were positive for CK20 but not p53, whereas 5 were positive for p53 but not CK20. Most of the patients (n = 47; 72%) had a concurrent or known history of bladder cancer, from which 27 had recurrence, with an average interval of 37 months (ranging from 2 to 216 months). A smaller fraction of patients (n = 22; 34%) had no prior history of bladder cancer. Of these, only 1 patient with CK20-positive/p53-equivocal staining eventually developed diagnostic urothelial carcinoma (Table 2). This particular patient had a history of prostate adenocarcinoma, for which he underwent radiation therapy, complicated with radiation cystitis. Seventeen years later, spanning an interval with multiple urine cytology examinations, a bladder biopsy included in this study contained urothelial atypia with positive CK20 and equivocal p53. Two years after this biopsy, the patient developed a noninvasive, high-grade, papillary urothelial carcinoma. Within the subset of patients with no prior history of bladder cancer, only 1 patient had a CIS staining pattern. This patient had a past medical history of radical prostatectomy followed by radiation therapy for a Gleason score of 5 + 4 = 9 prostate adenocarcinoma diagnosed 6 years before his urothelial biopsy diagnosis. The biopsy showed focal increased urothelial thickness and few enlarged atypical cells. Both CK20 and p53 stains were positive, and the case was interpreted as focal urothelial dysplasia and radiation-induced changes. Subsequent follow-up urine cytology specimens showed atypical urothelial cells of unknown significance with no otherwise diagnostic bladder cancer in 6 years of follow-up.

Twenty-seven cases had recurrent urothelial lesions associated with a biopsy with urothelial atypia. There were 18 cases (67%) that manifested as noninvasive cancer, of which 8 were pTis and 10 were pTa; and 6 cases (22%) progressed to invasive cancer, of which 2 were pT1, 3 were pT2, and 1 was pT3. There were 3 cases (11%) that were demonstrated as recurrent by positive urine cytology with no available tissue diagnosis.

There was a statistically significant association between history of bladder cancer and subsequent development of cancer, regardless of staining pattern (P < .001), and no association was found between history of cancer and immunohistochemical staining pattern (P = .17).

Of the discordant cases with CK20 positivity, 7 of 13 (54%) had recurrence of bladder cancer. Of the discordant cases with p53 positivity, 2 of 5 (40%) had recurrence of bladder cancer. There was no statistically significant difference (P = .60) between positivity for p53 versus CK20 with recurrence in the discordant category.

The risk of progression from urothelial CIS to invasive carcinoma is relatively well established.^2,4,5  Urothelial dysplasia, although less clearly understood, is also thought to have a significant risk for the development of urothelial CIS and subsequent invasive carcinoma. Thus, close follow-up in these patients should be undertaken.^7,15  On the other hand, patients with reactive atypia or urothelial atypia of unknown significance are largely not associated with adverse clinical outcomes.^7,8  Thus, the most critical distinction in flat urothelial lesions of the bladder is between reactive and neoplastic atypia. Because the management and prognosis of both ends of the spectrum are different, accurate histologic characterization of flat urothelial lesions is of great importance. In our cohort, we studied the use of CK20 and p53 in bladder specimens with borderline or suspicious features for CIS and correlated them with prior or subsequent cancer diagnoses.

The staining pattern of benign urothelium with CK20 is usually limited to the superficial or “umbrella” cells in a variable distribution or is entirely negative. Similarly, staining with p53 is usually negative or weak, and is usually limited to the basal cell layers in benign or reactive epithelium.^10,11,13  In urothelial CIS, the staining pattern with CK20 is aberrantly positive through all cell layers (full thickness). Conversely, with p53 there is strong, intense, and diffuse positivity in the nuclei of the atypical cells.^11,13,16 

Staining patterns among different studies in different flat urothelial lesions have shown varied results. In benign and reactive atypia, a CK20-negative or benign pattern has ranged from 100% to 92%.^12,13,16  Positive CK20 has been reported in a few examples of benign reactive urothelium,¹⁶  and this was anecdotally noted by participants in the 2013 International Society of Urologic Pathology Consensus Conference on immunohistochemistry in urologic organs.¹⁴  In one study, Jung et al¹⁶  found 6 benign or reactive specimens with at least 1 positive marker (CK20 or p53). Of these, 3 had positivity only for p53, 2 had positivity only for CK20, and 1 had positivity for both CK20 and p53. In contrast to the current study, however, these authors selected specimens with previously established diagnoses of CIS or benign or reactive urothelium. Only 1 study to our knowledge has reported on staining pattern in cases with urothelial atypia of unknown significance, showing positive CK20 in 30% of cases, and p53 positivity in 73% of cases.¹¹ 

This variation between studies could result from the different thresholds used or the scoring system applied for categorization of staining pattern. Most studies dichotomized results into positive or negative, occasionally noting cases with difficult interpretations. In practice, however, it is challenging to strictly categorize specimens as either positive or negative, and immunohistochemistry in general is notable for a wide array of staining patterns that cannot be consistently categorized as positive or negative, including differences in cellular distribution, intensity, and other parameters.

Other additional markers have been less frequently evaluated for the same purpose, including Ki-67, CD44, and Her2/Neu. The use of Ki-67 as a third positive marker for urothelial CIS was proposed by Mallofré et al.¹¹  Ki-67 staining in benign urothelium was negative or weakly positive in less than 10% of basal cells, and in urothelial CIS it reacted in 94% of cases, showing more than 50% of positive cells in 38% of cases.¹¹  Staining for CD44 in combination with CK20 and p53 has also been studied, showing basal cell labeling in benign urothelium, increased reactivity in reactive urothelium, and no labeling of atypical cells of urothelial CIS.¹²  Similar findings for CD44 staining pattern have been described in the setting of radiation-induced atypia.¹⁷  Her2/Neu is another marker that has been suggested for use in combination with CK20 and p53. Normal or reactive urothelium has been found to have a negative or weak Her2/Neu expression only in the superficial umbrella cells in 93% of cases. In contrast, urothelial CIS showed moderate to strong full-thickness membranous expression in 63% of cases, although the remaining urothelial CIS cases had staining patterns similar to those of normal or reactive urothelium.¹⁶ 

In our cohort, we included a patchy or equivocal staining pattern category for cases in which clear-cut benign or CIS immunohistochemical staining pattern could not be appreciated. We also correlated with results in previous or subsequent specimens. In general, most patients (n = 47; 72%) had concurrent or known history of bladder cancer, of whom 27 had recurrence, with an average interval of 37 months (ranging from 2 to 216 months). A subset of patients (n = 22; 34%) had no prior history of bladder cancer. Of these, only 1 patient with CK20-positive staining and p53-equivocal staining later developed diagnostic carcinoma (Table 2). Within the subset of patients with no prior history of bladder cancer, only 1 patient had a CIS staining pattern. This patient had a history of radical prostatectomy followed by radiation therapy. Despite positivity for CK20 and p53, subsequent follow-up urine cytology specimens revealed only atypical urothelial cells of unknown significance and no otherwise diagnostic cancer in 6 years of follow-up.

Only 20 of our cases (29%) showed both markers to be concordant (both stains showing either urothelial CIS or a benign, reactive pattern), a result that is thus far underemphasized in the existing literature. A total of 18 cases (26%) were categorized as discordant, in which only 1 of the 2 markers yielded the previously described urothelial CIS staining pattern (either CK20 or p53), with the other marker being either equivocal or negative. Of this subgroup, 13 were positive for CK20 but not p53, whereas 5 were positive for p53 but not CK20. Of the patients who had a history of bladder cancer and a discordant staining pattern, 8 of 9 (89%) had recurrence of cancer. On the other hand, of the patients with discordant pattern and no history of bladder cancer, only 1 of 9 (11%) developed subsequent bladder cancer. A total of 31 specimens (45%) showed an indeterminate staining pattern, of which 21 patients had a history of bladder cancer and 7 subsequently had recurrence, whereas none of the patients with indeterminate staining and lack of cancer history later had diagnostic cancer. A total of 11 specimens (16%) showed a benign staining pattern, of which 9 were from patients with a history of bladder cancer. Of these, 7 subsequently had recurrence, whereas none of the patients without a history of cancer had later development of diagnostic cancer.

Atypia of unknown significance is a diagnostic term used when distinction between benign or reactive urothelium and urothelial dysplasia is difficult. Histologically, the degree of pleomorphism or hyperchromasia is regarded as being out of proportion with the amount of inflammation when compared with reactive atypia, making it difficult to rule out dysplasia or CIS with certainty. This was previously considered as a diagnostic category, but there is currently scant evidence showing a negative outcome, and therefore its use is currently discouraged as much as possible.^6–8,18,19  Nonetheless, in some cases a borderline diagnosis may be inevitable.

Our high rate of discordant and indeterminate cases reflects our patient cohort, in which only specimens with equivocal atypia were selected for analysis. This is partly a strength of the study, because in diagnostic practice immunohistochemistry would rarely be needed when morphology is already diagnostic of benign epithelium or CIS. Overall, however, our findings support the 2014 International Society of Urologic Pathology consensus statement that histologic features remain the gold standard for diagnosis of urothelial CIS.¹⁴  Ultimately, to provide the best management for patients with bladder neoplasia, close cooperation and collaboration among urologists, oncologists, radiologists, and pathologists is necessary.²⁰ 

Potential limitations of our study include that urothelial atypia may have intraobserver and interobserver variability. Therefore, there may be cases that we included as borderline that may have been considered straightforward by another observer, and conversely there may be cases that we considered straightforward, and did not include, yet others may have considered them to be borderline. An inherent problem of bladder biopsy sampling in general is that the biopsy sample may not always be representative of the disease in the urinary tract overall, and therefore multiple modalities are typically used in conjunction, including urine cytology, as well as other possible adjunctive assays. Although correlation of subsequent cancer diagnoses is a strength of this study, the numbers of patients in each staining pattern subgroup are smaller, and thus it was not possible to draw definitive conclusions in each pattern subgroup. Another area of potential future study is that bladder cancer is now increasingly recognized to have several molecular phenotypes^21,22  and likely varying immunohistochemical phenotypes.^23–25  As such, it is possible that some bladder cancer phenotypes by their nature have positive staining for CK20 but not p53, demonstrate the opposite, or are negative for both of these markers, reflecting different molecular pathways and cell phenotypes. With growing understanding, this may reshape the expected staining patterns of bladder cancer by dividing it into multiple molecular subgroups.

CK20 and p53 are among the most commonly used immunohistochemical markers for the discrimination of urothelial CIS from benign or reactive urothelium. However, most previous studies on the use of these antibodies were conducted on specimens with established histologic diagnoses. In our cohort of patients with borderline urothelial atypia, only a minority demonstrated a straightforward CK20-positive/p53-positive (CIS) pattern or negative (benign) pattern. Most demonstrated either a single positive marker or an indeterminate result. Regardless of staining pattern, very few patients without a prior diagnosis of bladder cancer progressed to diagnostic cancer (1 of 22), suggesting that staining results should be interpreted with caution in the setting of de novo atypia. Unclassifiable urothelial atypia in a patient with prior history of urothelial cancer is more worrisome than in the absence of such history, because the rate of recurrence is very high in this subset.

The authors would like to thank Daniel S. Schultz, MD, for his critical review of the manuscript and assistance with statistical analysis.