In this current study, we investigated the sensitivity of biopsy and/or TURBT for detecting the histological differentiation on RC and assessed the prognostic significance of variant histology on urothelial carcinoma outcomes after RC. Our results indicate that overall sensitivity of biopsy or TURBT to detect variant differentiation on RC is relatively low. Presence of variant differentiation in urothelial carcinoma at cystectomy portends inferior survival.
It is not an uncommon phenomenon that UCB has a great propensity to undergo variant differentiation. In accordance with previous studies [3, 15], the most prevalent mixed differentiation in this current series was squamous, accounting for 31 % of all cases. Squamous differentiation, characterized by the presence of intercellular bridges, keratin pearls or keratinization, has an unfavorable response to radiation and chemotherapy compared with pure UCB . Previous studies have shown that its frequency runs parallel with grade and stage [16, 17]. Glandular differentiation, as the second common variant, is exemplified by the presence of true glandular spaces within the tumor, presented in 28 % of all cases . After squamous and glandular subtypes, there is sarcomatoid differentiation, which is featured by biphasic malignant neoplasms with the evidence of expressing epithelial and mesenchymal markers, accounting for 12 % of all cases. In a sarcomatoid carcinoma, molecular evidence strongly argues for a monoclonal origin of both epithelial and mesenchymal components [18, 19]. Additionally, history of previous radiation or cyclophosphamide can be a valuable clue in ascertaining the diagnosis . Other unusual architectural patterns (small cell, lymphoepithelioma-like, clear cell, microcystic, undifferentiated) of UCB only accounted for a little proportion of the whole, namely 8 %. Those variants, compared with squamous and glandular differentiation, are more complicated for the pathologist to reach a direct diagnosis and can sometimes mimic reactive processes, benign lesions or metastasis of other tumors. For instance, lymphoepithelioma-like carcinoma, for which the sine que non feature is the presence of a pronounced lymphoid infiltrate, may be partially or substantially obscured by a variable desmoplastic response, a brisk inflammatory infiltrate and even a prominent chronic inflammatory cell infiltrate .
Although some studies have contributed to raising the awareness of these entities and improved diagnosis, the diagnosis of UCB variants may still be challenging for (a) the limitation of biopsy samples and transurethral resectates available to the pathologist, (b) the artifact caused by tangential sectioning, cautery and mechanical injury, and (c) even more importantly the difficulty of discriminating over selected mimics [3–5, 12]. In this study, the most sensitive variants detected by biopsy and LTURBT are squamous and glandular subtypes. This finding corroborates a previous study in which Ahmed et al. reported that the sensitivity of TURBT to detect squamous and glandular variant was 53 % and 25 % respectively . Similar to the squamous differentiation, the sensitivity of LTURBT to detect glandular variant is nearly twice as that of biopsy. Compared to other variants, the sensitivity of biopsy and LTURBT to detect squamous and glandular differentiation was relatively higher. For most variants, Table 1 suggests an unfavorable sensitivity. However, considering their nature of rarity, more studies based on multicenter and/or international collaboration with large sample size are warranted to further validate these conclusions.
It is well known that sensitivity lies in correct diagnoses. In order to make a correct diagnosis, several points should be noted. First, recognition of the immunochemical profile is quite important for pathologists to distinguish certain variants from the confounding variables. Mahul et al.  summarized the immunohistochemical markers associated with urothelial differentiation in detail. Second, morphology should be made the best advantage. One case in point is that immunohistochemistry is less significant compared with morphology in discrimination nested variant over florid von Brunn’s nests, as the reliable immunohistochemical cut-point is difficult to be determined . Besides, in some cases, clinicopathological correlation is helpful in excluding an extravesical primary tumor . Recently, Hughes et al.  have reported the effect of Fourier transform infrared (FTIR) microspectroscopy to diagnose some selected variants, which might open a new door for variant diagnosis.
In this study, we found that 49 % patients who underwent TURBT at least once were not diagnosed with variants on RC specimens. That is to say, approximately 50 % of the patients could be in variant-free condition partially due to complete resection(s). Dissimilarly, Ahmed et al. reported that 6 % (9/159) patients demonstrated variant(s) only on precystectomy biopsy or TURBT . The factors that could influence the variant diagnosis could also explain the discrepant results between the two studies. Interestingly, among the 13 patients who underwent TURBT at least twice, 12 of them presented variants detected in some but not in all TURBTs (Table 2). Similarly, Matthew et al. found that some patients had a different or additional mixed histologic type on cystectomy than they did on transurethral resection of bladder tumor . This raises several practical questions encountered in clinical work. First, for those who are diagnosed with different variant differentiation at biopsy, TURBT(s) or RC, what treatment algorithms should be applied in clinical management? Besides, for those who are diagnosed with variant differentiation at biopsy/TURBT and without variant subtype diagnosis on RC, should we adopt the same follow-up strategies used on the UCB patients without a diagnosis of variant differentiation during the disease course? Further investigations are needed to validate the paradigms.
Despite a number of individual studies have evaluated the impact of histological variants on prognosis of patients with UCB, this question remains in debate [14, 22]. We also evaluated the impact of variant differentiation on clinical outcomes. Of the 139 patients with available follow-up information, 95 UCB patients were identified with variant differentiation on RC specimens. Patients with variant differentiation on RC specimen have inferior survival both in univariate analysis (P = 0.005) and multivariate analysis (HR4.48, 95 % CI:1.03–19.53). Of note, our data were limited by an overall shorter median follow-up time of 31 months and small sample size. The statistical power was weakened by very few recurrent and death events. More investigations with a large number of patients, with a longer follow-up time and with a centralized pathologic re-review are warranted to validate these conclusions. Ultimately, our understanding of urinary bladder cancer biology should not be limited to histologic variants. The underlying genetic and molecular drivers of tumor induction, promotion, and progression and as well as makers of chemosensitivity also need to be investigated .
Our study is not devoid of potential limitations that need to be addressed. First and foremost are the limitations inherent to its retrospective nature. Besides, the follow-up time and sample size were the limitations of this study.