Subcategorization of T1 Bladder Cancer on Biopsy and Transurethral Resection Specimens for Predicting Progression Open Access

This retrospective study was approved by our Institutional Review Board and included patients who were diagnosed with urothelial carcinoma with “lamina propria invasion” or “suspicious for lamina propria invasion” on a bladder biopsy or TUR at our institution between 2004 and 2014. Review of our pathology database identified an initial cohort of 140 patients diagnosed with T1 bladder cancer. On further review of patient medical history, the following cases were excluded: previous diagnosis of T1 urothelial carcinoma made on a prior specimen (n = 18); patients lost to clinical follow-up within 6 months of initial diagnosis (n = 18); cases that were restaged as T2 on early re-TUR performed to ensure that muscularis propria was obtained (n = 17); slides no longer available for review (n = 9); and cases revealing ambiguous patterns of infiltration that are best classified as Ta on review (n = 5). This resulted in a final cohort of 73 patients (58% of the cases considered) with confirmed original diagnosis of T1 urothelial carcinomas and adequate clinical follow-up for inclusion in this study. We allowed for the inclusion of patients with prior noninvasive urothelial carcinoma, previous administration of BCG therapy, and focal variant morphology. Of note, re-TUR was not routinely performed in patients where muscularis propria was present in the initial diagnostic specimen. Of the 73 cases included in this study, 15 patients underwent re-TUR and did not upgrade to muscle-invasive disease.

All hematoxylin-eosin slides were re-reviewed by a genitourinary pathologist (A.N.) to confirm the diagnosis and associated histopathologic features, such as tumor grade, focality, presence of CIS, lymphovascular invasion, muscularis propria, and presence of variant morphology. All tumors were classified according to the 2016 World Health Organization grading system and staged according to the 2018 American Joint Committee on Cancer staging scheme.²⁵ 

To determine which method of substaging best predicts tumor recurrence and progression, we analyzed invasive component(s) of each T1 specimen by using several described and novel methods (Table 1) and compared their respective ability to predict patient outcome. Figure 1, A through F, shows examples of histologic application of each measurement approach.

The following methods were applied to each specimen: (1) relationship of carcinoma to MM, defined as the presence of invasive component either “above” or “into/below” MM or VP as its anatomic substitute in cases where definitive MM could not be identified, as previously described (Figure 1, A and B)^11,12 ; (2) estimation of burden of invasive component, defined as “focal” if no more than 2 invasive foci were present, each measuring less than 1 mm (Figure 1, C), otherwise T1 disease was classified as “extensive” (Figure 1, D); (3) number of invasive foci: counting invasive foci present either in different bladder anatomic locations or noncontiguous invasive foci within the same specimen (Figure 1, C and D); (4) micrometer measurement of depth of invasion, defined as a distance from the basement membrane to the deepest point of invasion, as previously described (Figure 1, A)¹³ ; (5) micrometer measurement of largest contiguous focus of invasion in any direction (Figure 1, B, C, and E); and (6) aggregate linear length of invasion, which is a sum of micrometric measurements of greatest dimension of each invasive focus, as previously described (Figure 1, C, D, and F).¹⁵ 

Immunohistochemistry (IHC) was performed on 5-μm formalin-fixed, paraffin-embedded tumor sections with adequate controls, using antibody against GATA3 (PM405AA, clone L50-823, Biocare Medical, Pacheco, California) and CK5/6 (IR780, clone D5/16B4, Agilent/Dako, Santa Clara, California). Staining was done on a Leica Bond-IIITM instrument using the Bond Polymer Refine Detection System (Leica Biosystems DS9800, Buffalo Grove, Illinois). Heat-induced epitope retrieval was done for 20 minutes with ER2 solution (Leica Biosystems AR9640). The experiment was performed at room temperature. Slides were washed 3 times between each step with bond wash buffer or water.

The invasive component of the T1 cancer was scored for GATA3 (nuclear) and CK5/6 (cytoplasmic) expression. Percentage of tumor cells positive for staining was scored as an average of total invasive component. A cutoff percentage of 20%, based on the study by Dadhiana et al,³¹  was used to call a staining positive or negative. Further, cases were classified into 4 categories: luminal (GATA3⁺/CK5/6⁻), basal (GATA3⁻/CK5/6⁺), mixed (double-positive for GATA3⁺/CK5/6⁺), and null (double-negative for GATA3⁻/CK5/6⁻).

Demographic, clinical, and radiologic information of the study population was retrieved from the electronic medical records. This included data on the patients' sex, ethnicity, age, previous or concurrent history of other malignancies, urothelial lesion imaging findings, prior history of Ta/Tis and date of diagnosis, date of the initial T1 diagnosis, lesion topography, date of last follow-up, survival status, and precedent and subsequent BCG administration. Clinical follow-up data obtained also included number of recurrences and time to the disease recurrence and progression. Recurrence and progression were considered as the study primary endpoints. “Recurrence” was defined as the detection of non–muscle-invasive urothelial neoplastic lesion (<pT2 stage) after at least 6 months following complete removal of the primary tumor by TUR. “Progression” was defined as the recurrence of urothelial carcinoma with pT2 or higher stage and/or with distant metastasis. Recurrence-free and progression-free survival were calculated by using the time of initial T1 diagnosis as the start point.

Parametric and nonparametric statistical tests were used to compare the baseline clinical, demographic, and histopathologic features, as well as the results of T1 substaging methods and immunohistochemical findings, among the patients with and without disease recurrence and progression. Shapiro-Wilk test was used to determine the distribution normality of the data. t test and Mann-Whitney test were used in normally and nonnormally distributed continuous variables, respectively. Categorical variables were assessed by using χ² test. The receiver operating characteristic (ROC) analysis was performed to compare the discriminative power of each of the quantitative substaging methods in predicting recurrence or progression, using the area under the curve (AUC). A test was considered to discriminate better than chance if the AUC was above 0.5. The optimal cutoff values for the statistically significant substaging methods were derived from the ROC curves to divide patients into progressors (those whose condition progressed to muscle-invasive disease or metastatic disease) and nonprogressors (those whose condition did not progress and remained at pT1 disease on follow-up). Kaplan-Meier analysis was performed to assess the effect of individual cutoffs on survival. All statistical analyses were performed with STATA software (Stata/SE 13.1., College Station, Texas). A P value less than .05 was regarded as statistically significant.

Seventy-three patients with the confirmed first-time diagnosis of pT1 urothelial carcinoma on bladder biopsies or TUR specimens were included in this study. The cohort included 61 males and 12 females. The mean age at the time of the initial diagnosis was 70 years (range, 39–94 years). The mean follow-up time was 46 months (range, 1–157 months). Twenty-nine patients (40%) had a prior or concurrent medical history of nonurothelial malignancies, while 2 patients (3%) had a prior or concurrent history of noninvasive upper tract urothelial carcinoma. Sixteen patients (22%) had a prior diagnosis of Ta or Tis bladder urothelial carcinoma, with a median duration of 13.5 months (range, 1.0–73.1 months) before the initial diagnosis of T1 disease. In total, 32 patients (44%) had received prior BCG administration exposure, and 1 patient (1.4%) had received prior laser therapy. Subsequent to T1 diagnosis, 47 patients (64%) were treated with BCG and 17 (23%) received chemotherapy. Of the total 73 patients, 3 patients (4%) underwent partial cystectomy/diverticulectomy and 16 (22%) underwent a disease-necessitating cystectomy during the course of disease.

High-grade morphology was present in most cases (65 of 73, 89%). All 73 cases (100%) were papillary urothelial carcinoma by histology, with focal variant morphology noted in 10 cases (14%) (most commonly showing focal glandular, micropapillary, or squamous cell carcinoma differentiation). Out of 73, there were 34 cases (47%) with multifocal disease. Concurrent CIS was identified in 24 cases (33%).

After a mean follow-up of 46 months (1–157 months), 36 patients (49%) had no recurrence, 21 patients (29%) experienced at least 1 recurrence without progression, and 16 patients (22%) had recurrence with disease progression. Of note, 4 patients had a follow-up period of less than 6 months, but we chose to include these patients in our cohort because all 4 had well-sampled muscularis propria in the initial TUR and elected for immediate cystectomy instead of conventional BCG therapy and follow-up. Eleven of 16 patients (69%) with progressive disease died of disease, whereas no patient in the nonprogressive disease group died of urothelial carcinoma (Table 2). On univariate analysis, total cystectomy status was significantly associated with progression (odds ratio, 10.3; 95% CI, 2.8–37.8; P < .001). None of the other clinical, demographic, or histopathologic features (sex, age, prior treatment status, past medical history, tumor grade, variant morphology, or presence of CIS) demonstrated a significant correlation with recurrence or progression status. No correlation with tumor grade could be partly explained by predominance of high-grade tumors in our cohort (65 versus 8), which is in alignment with the recent literature based on 2-tier grading system.

As for the applicability of proposed substaging methods, measurements relying on anatomic landmarks and tissue orientation proved to be less user-friendly. Assessment of MM was feasible in only 49 of the 73 cases (67%). In an additional 7 cases (10%), the invasive front of the tumor was assessed in relation to the VP of the lamina propria as a surrogate of the MM. Therefore, this anatomy-based method was successfully applied to 56 cases (77%). Depth of invasion requires tissue orientation to the basement membrane and was successfully applied to 70 cases (96%). On the contrary, measurements independent of anatomic landmarks and tissue orientation were successfully applied to all cases. The mean number of invasive foci was 2.9 (range, 1–19), with 21 cases showing focal invasion (defined as 2 or fewer foci of invasion, each measuring <1 mm), and 52 cases with extensive invasion (defined as more than focal). The mean tumor depth from basement membrane was 1.2 mm (range, 0.1–5.7 mm). The mean diameter of the largest invasive focus was 2.4 mm (range, 0.2–13.1 mm), and the mean aggregate length of invasion was 8.9 mm (0.25–113.4 mm). None of the substaging methods correlated with disease recurrence. However, univariate analysis revealed that substaging using aggregate linear length of invasion (mean, 20.37 mm versus 4.65 mm in progressors versus nonprogressors; P = .009), number of invasive foci (mean, 5.56 versus 2.54 foci in progressors versus nonprogressors; P = .008), largest invasive focus (mean, 4.35 mm versus 2.26 mm in progressors versus nonprogressors; P = .01), and depth of invasion (mean, 2.08 mm versus 1.20 mm in progressors versus nonprogressors; P = .02) could predict carcinoma progression (Table 1). The 4 quantitative methods retained their significant predictive value on multivariate analysis after adjusting for age, sex, ethnicity, tumor focality, presence of carcinoma in situ, and prior treatment status (Table 1). Notably, we could not adjust for grade as most of our cases were high grade (65 versus 8).

Further, we performed an ROC analysis to calculate the optimal cutoff values for the statistically significant substaging methods. The depth of invasion of 1.4 mm or greater (n = 29/70, 41%) was significantly associated with progression of disease (hazard ratio [HR], 4.2; P = .01). The cutoff values that correlated best with progression for other methods were as follows: 3.6 mm or greater (n = 19/73, 26%) for the largest contiguous focus of invasive carcinoma (HR, 3.1; P = .03) and 8.9 mm or greater (n = 15/73, 21%) for aggregate length of invasion (HR, 4.9; P = .002) (Table 3). Lastly, the presence of 3 or more foci of invasion (n = 28/73, 38%) correctly predicted progression rate in 70% of cases (HR, 4.5; P = .007) (Figure 2, A through D). Individual cutoffs derived from the ROC analysis correlated significantly with the reduced progression-free survival on Kaplan-Meier analysis (Figure 3, A through D).

We were able to perform the immunohistochemical analysis for GATA3 and CK5/6 in 67 patients owing to the loss of invasive component of interest on deeper sections in 6 patients with focal lamina propria disease. The IHC analysis revealed that 48 of the 67 tumors (72%) were luminal [GATA3⁺, CK5/6⁻] and 19 tumors (28%) were nonluminal, including 3 basal [GATA3⁻, CK5/6⁺] and 16 double-positive [GATA3⁺, CK5/6⁺] (Figure 4, A through I). None of the tumors stained double-negative [GATA3⁻, CK5/6⁻]. Therefore, the luminal subgroup was a predominant phenotype in our cohort of NMIBC. Our data showed no significant difference between luminal and nonluminal tumors with regard to sex, ethnicity, grade, and presence of CIS. The χ² analysis did not reveal statistically significant differences in the recurrence rate between luminal and nonluminal tumors. Twenty-six of 48 luminal tumors (54%) and 9 of 19 nonluminal tumors (47%) recurred on follow-up (P = .37). Thirteen of 48 luminal tumors (27%) exhibited progression on follow-up compared to 2 of 19 (11%) nonluminal tumors; however, the difference did not reach statistical significance (P = .14).

Given the unmet need for a user-friendly and a robust method of T1 bladder cancer risk stratification, we sought to compare various methods of substaging T1 disease in bladder biopsies or TUR specimens. Additionally, this study aimed to investigate whether stratifying T1 cancers into luminal or basal phenotypes, using immunohistochemical surrogates (GATA3 and CK5/6), could have prognostic implications. Our results suggest that quantitative substaging methods (aggregate linear length of invasion, number of invasive foci, largest diameter of contiguous invasive focus, and depth of invasion) were superior in predicting cancer progression compared to 2 other methods (above versus into/below MM or VP, focal versus extensive) on univariate and multivariate analysis adjusted for age, sex, ethnicity, tumor focality, presence of carcinoma in situ, immunohistochemical phenotype, and prior treatment status. We could derive the optimal cutoff values best separating the cohort into “progressors” and “nonprogressors,” using ROC analysis for each method: 1.4 mm or greater for depth of invasion, 3.6 mm or greater for the diameter of largest contiguous invasive focus, 8.9 mm or greater for aggregate length of invasion, and presence of 3 or more foci of invasion. Survival analysis revealed that the above cutoffs correlated significantly with reduced progression-free survival. None of the substaging methods were, however, able to predict recurrence.

In agreement with other studies,^13,37–40  we did not find any significant correlation between MM invasion and progression. Similarly, we could not apply this method to all of the cases, even though we used the large vessels in the lamina propria as a substitute in 10% of the cases. MM or VP invasion could be assessed only in 77% of cases in our cohort, consistent with the broad identification range (from 32% to 100%) reported in the literature.^{16,18,37–40} 

The depth of invasion measured by a micrometer, another well-studied histologic method of T1 substaging, while considered to be more objective and reproducible by some,^13,24  has its own challenges. A major criticism against this method is that in poorly oriented TUR specimens, an overlying urothelium is not always present to measure the distance between the basement membrane and the deepest invasive focus. Surprisingly, in our cohort we could successfully apply this approach to 96% of cases and found a higher rate of progression in patients with a depth of invasion of 1.4 mm or greater. Our results validate the findings of the original study by Cheng et al¹³  that found a depth of invasion of 1.5 mm or greater to be a predictor of T1 cancer progression. To date, few other groups have confirmed the prognostic significance of this approach; however, the cutoffs proposed have varied. Brimo et al⁴⁰  reported a depth of invasion of 2 mm or greater to correlate with recurrence, and 3 mm or greater to be predictive of tumor progression.

More recently, newer methods such as the size of the maximum diameter of the invasive focus or the aggregate linear length of invasive carcinoma have become of interest because they do not rely on specimen orientation and therefore can be widely applied to all cases encountered in the routine pathologic practice.^{14–16,38–40}  Not surprisingly, we were able to apply them to 100% of cases in our cohort. A recent study by Leivo et al¹⁶  performed a comprehensive comparison of some of the methods described above in a cohort of 118 patients. They found the aggregate length of all invasive foci to be a superior prognostic variable for progression with an optimal threshold of 2.3 mm. The depth of invasion was the second-best method in their study. We found these T1 quantification methods to be significant as well. However, our optimal cutoff value for the aggregate linear length of invasion was much higher (≥8.9 mm), possibly due to our relatively smaller study cohort.

In our cohort, the maximum diameter of the invasive focus that could predict progression reliably was 3.6 mm (HR, 3.1; P = .03). In our opinion, this method is simpler to apply in clinical practice as it is not influenced by tissue artifacts and poor orientation. However, the threshold derived in this study is much higher than that in previously published studies by van Rhijin et al³⁸  and Patriarca et al,³⁹  who have proposed 0.5-mm and 1-mm cutoffs, respectively, for subclassifying T1 disease. We believe larger studies are required for validating these thresholds.

Another simpler method of substaging invasive tumor burden could be counting of infiltrative foci, with presence of 3 or more invasive foci indicating adverse clinical outcome as noticed in our cohort (HR, 4.5; P = .007). To the best of our knowledge, no previous studies have specifically reported this criterion for substaging T1 disease. Indeed, this cutoff turned out to be second best in predicting progression after aggregate linear length of invasive carcinoma in ROC analysis. We believe this is the best and most user-friendly method for substaging T1 cancers in day-to-day pathology practice.

The second aim of this study was to investigate whether there is any role for molecular subtyping (luminal versus basal) in prognostic stratification of T1 bladder cancers, as suggested by some recent studies.^26,28,33,35  Limited literature available on molecular subtyping of NMIBCs indicates that the molecular subtypes of MIBC and NMIBC have opposite effects on prognosis. For example, studies have shown that MIBC enriched with the basal-type proteins CK5/6 and CD44 carry a significantly worse prognosis than MIBC expressing the luminal-type protein CK20^26,28 ; however, on the contrary, luminal-type NMIBCs enriched with CK20 show more aggressive phenotypes and shorter progression-free survival than basal-type NMIBCs.^33,35  Rebola et al³⁵  in their study that included 37 T1 cancers and used CK20 and CK5/6 as luminal and basal markers, respectively, concluded that molecular subcategorization could assist in prognostic stratification of NMIBCs, with luminal phenotype being an independent predictor of most aggressive behavior and mixed phenotype resulting in less aggressive tumors. Similarly, Breyer et al³³  assessed CK5 and CK20 mRNA expression using RT-qPCR (real-time quantitative polymerase chain reaction) in a cohort of 284 patients with T1 cancer and demonstrated that luminal subtype (combination of high CK20 and low CK5) was associated with worse progression-free and recurrence-free survival. Instead of the CK5 and CK20 combination, we chose to use GATA3 and CK5/6 as our 2 IHC classifiers in light of a recent study by Dadhania et al³¹  that elegantly demonstrated that more than 90% of bladder cancers can be accurately phenotyped as luminal or basal when using GATA3 (luminal) and CK5/6 (basal) as immunohistochemical surrogates. Our results show that luminal subgroup (GATA3⁺/CK5/6⁻) is a predominant phenotype in T1 cancers (48 of 67; 72%) followed by mixed/double-positive (GATA3⁺/CK5/6⁺) (16 of 67; 24%), and basal (GATA3⁻/CK5/6⁺) (3 of 67; 5%). Owing to a limited number of cases with basal phenotype, statistical analysis was done in the context of luminal and nonluminal phenotype. No significant difference was observed between luminal and nonluminal phenotypes in relation to sex, ethnicity, grade, presence of CIS, recurrence, and survival. In alignment with past studies, we observed a positive trend for progression in patients with luminal tumors when compared with patients with nonluminal tumors (27% versus 11%), although the difference did not reach statistical significance (P =.14). Furthermore, CK5/6 expression in tumors did not significantly correlate with recurrence (P = .37) and shorter survival (P = .18).

Our study is one of a few that attempted to compare published and emerging T1 substaging methods in a well-characterized cohort of patients with bladder cancer. Potential limitations of the current study are a relatively small number of cases and the low pathologic progression rates observed (22%) compared to other studies.^34–37  The former is due to the number of strict exclusion criteria we applied during the process of case selection, such as presence of previous history of T1 disease or advanced urothelial carcinoma from another site in the urinary tract, morphologic variants other than urothelial carcinoma, absence of clinical follow-up, and cases without the presence of detrusor muscle in either their initial TUR or re-TUR specimen. The low pathologic progression rates observed in our study may in part be due to a more aggressive institutional approach in treating patients with recurring T1 disease with early cystectomy.

To conclude, substaging should be attempted in routine pathology practice for prognostic stratification of T1 bladder cancers. Although aggregate linear length of invasive carcinoma is statistically the strongest method, it could be cumbersome and time-consuming. Hence, we recommend either measuring the largest contiguous focus of invasive carcinoma (3.6 mm or greater) or quantifying the number of invasive foci (3 or more) in pathology reports, the latter being simplest and most practical. These methods are easy to apply in routine pathology practice and are not affected by tissue orientation/artifacts. The role of molecular subtyping in prognostic stratification of T1 bladder cancers needs to be further tested in larger cohorts.