Introduction

Bladder cancer (BCa) is the second common cancer in urinary and male genital system and the 10th most common cancer in the world [1]. It is estimated that, in 2018, BCa accounts for 81,190 new cases and 17,240 cancer-related deaths in the United States (US) [2]. Furthermore, BCa is more common in men than in women. Especially in men older than 80 years, BCa is the fourth leading cause of cancer-associated mortality in the US [2]. The stage and grade of tumor have largely accepted for directing therapy and predicting outcome of BCa [3, 4]. In addition, histological type is another important factor should be considered to relate the prognosis of BCa [5].

More than 90% of BCa are classified as urothelial carcinoma, which include pure urothelial carcinoma of bladder (PUCB) and urothelial carcinoma with divergent differentiation or histologic variants. Urothelial carcinoma with divergent differentiation is defined by tumors arising within the urothelial tract, in which some percentage of ‘‘usual type’’ urothelial carcinoma is present along with other morphologies [6]. The incidence of histologic variants in studies ranges from 10.4 to 33% [6,7,8,9,10,11,12,13]. Previous studies have reported that urothelial carcinoma with histologic variants was associated with advanced tumor stage, extravesical disease, lymph node invasion, and positive soft tissue surgical margin [8, 11, 14, 15]. The presence of histologic variants was also associated with worse cancer-specific survival (CSS) or overall survival (OS) [9, 15, 16]. However, other studies showed that variant histology did not significantly influence the outcomes of BCa compared to PUCB [8, 10,11,12, 14]. These conflicting results might be explained by different types of histologic variants.

Urothelial carcinoma with squamous differentiation (UCSD) is the most common histologic variant and defined by the presence of intercellular bridges and/or keratinization [6, 17]. Previous studies showed that urothelial carcinoma with other variants except squamous differentiation (SD) was associated with increased disease recurrence and cancer-specific mortality (CSM) compared to PUCB [8, 13, 15] and patients with non-SD variants had shorter OS and poorer CSS compared to those with SD [18]. When the role of SD in BCa was examined, inconsistent results were obtained with some studies demonstrated that patients with UCSD had a significantly higher risk for poorer survival outcome [19,20,21], while other reports found no differences in survival outcomes [8, 22,23,24]. At present, the prognostic and clinical value of SD in BCa have not been fully determined.

In this meta-analysis, we aimed to clarify the clinical characteristics of patients with UCSD and to assess whether the presence of SD in patients with UCSD indicates a worse prognosis when compared to those with PUCB.

Methods

This meta-analysis was conducted in accordance to the Preferred Reporting Items for Systematic Reviews and Meta-analysis (PRISMA) guidelines [25]. Protocol of this study was registered in PROSPERO under number CRD42018116633.

Search strategy

A systematic literature search was performed in electronic databases including PubMed, Embase, Chinese National Knowledge Infrastructure (CNKI) and Wanfang Data until October 2018. The following terms were used for the literature search: “bladder cancer” OR “bladder tumor” OR “bladder carcinoma” OR “urothelial carcinoma” OR “transitional cell carcinoma” AND “squamous differentiation”.

Inclusion and exclusion criteria

Studies included in this meta-analysis met the following criteria: (1) all the patients were diagnosed as urothelial bladder cancer that was pathologically confirmed; (2) studies investigated the clinicopathological features in patients with UCSD compared to PUCB; (3) treatment outcomes were reported in studies, such as rate of recurrence, disease-free survival (DFS), CSS, OS. The exclusion criteria were as follows: (1) case reports, non-published materials, editorials, reviews and animal experiments; (2) patients with upper tract urothelial carcinoma or other cancers; (3) studies investigated BCa with other histologic variants or pure squamous carcinoma. In addition, the most recent report or study with larger population and more complete information was used in this analysis when the same population was published in multiple studies.

Data extraction and study quality assessment

The following data were extracted from eligible studies: first author, year of publication, study country, recruitment period, patient age, gender ratio, pathological tumor stage, pathological tumor grade, treatment strategy, positive surgical margin rate, lymph node involvement rate, lymphovascular invasion (LVI) rate, follow-up time, and related survival outcomes.

For the assessment of study quality, all the included studies were evaluated by the Newcastle–Ottawa Scale (NOS) [26], which includes 8 items for the patients selection, comparability of cohorts and assessment of outcome. The literature search, data extraction and study quality assessment were conducted by two reviewers independently (XL and TD).

Study outcomes and statistics analysis

The primary outcomes in this meta-analysis were recurrence and CSS. The secondary outcomes were DFS and OS. For recurrence, we performed subgroup analysis according to different treatments.

Dichotomous variables were calculated using summarized odds ratios (ORs) with 95% confidence intervals (CIs). Mantel–Haenszel estimates were performed to compare the percentage of female, tumor stage T3/T4, tumor grade 3, carcinoma in situ (CIS), tumor multiplicity, rate of positive surgical margin, positive lymph node and LVI. Available multivariable adjusted hazard ratios (HRs) of UCSD prognostic survival were pooled in eligible studies. The heterogeneity among studies was analyzed by using chi-square test based Q- and I2-statistic [27]. The fixed-effect model was used when heterogeneity was not significant with a P value > 0.10; otherwise, the random-effect was applied. All results in this meta-analysis were considered as significant with a two-sided P value < 0.05. The publication bias among included studies was assessed through the inverted funnel plot visual inspection and Egger’s test [28]. All the analyses were performed by RevMan (version 5.3; Cochrane Collaboration, Oxford, UK) and STATA (version 13.0; StataCorp, College Station, Texas, USA) software.

Results

Characteristics and quality assessments of eligible studies

After the initial search of the databases, 947 studies were identified following the search strategy. Excluding the duplicate reports, reviews, letters, case reports, irrelevant studies and other articles, a total of 19 studies [8, 11, 15, 20,21,22,23,24, 29,30,31,32,33,34,35,36,37,38,39] were included in this analysis (Fig. 1). The basic characteristics of all eligible studies are presented in Table 1. 13,284 patients were enrolled in these studies which were published from 2007 to 2018. All the included studies were retrospective cohort study. Most studies focused on the BCa in tumor stage T1–T4, 3 studies focused on the BCa in tumor stage T1. The main treatment approaches for patients in these studies included transurethral resection of bladder tumor (TURBT) in 3 studies and radical cystectomy (RC) in 16 studies. For the prognostic outcomes, recurrence was reported in 6 studies, DFS, CCS and OS were reported in 3, 4 and 5 studies, respectively. The results of quality assessment of included studies are presented in Supplementary Table 1. One study was in intermediate quality with six stars, the other studies were in relatively high quality with seven or more stars according to NOS assessment.

Fig. 1
figure 1

Procedure of literature search and selection of included studies

Full size image
Table 1 Characteristics of the included studies
Full size table

Clinicopathological characteristics

Results of meta-analysis of clinicopathological characteristics are presented in Table 2. Our results indicated that the percentage of female patients who had UCSD was significantly higher than those with PUCB (random-effect model, OR 1.70, 95% CI 1.36–2.13, P = 0.0004, Supplementary Fig. 1). High grade tumors and pathological stage T3/T4 tumors were more common in UCSD (random-effect model, OR 1.92, 95% CI 1.28–2.86, P = 0.001; random-effect model, OR 2.57, 95% CI 2.07–3.19, P < 0.0001; Supplementary Fig. 2 and Supplementary Fig. 3). Positive surgical margin rate and lymph node involvement rate was higher in UCSD group (random-effect model, OR 1.65, 95% CI 1.10–2.46, P = 0.01; random-effect model, OR 1.29, 95% CI 1.07–1.56, P = 0.009; Supplementary Fig. 4 and Supplementary Fig. 5). The percentage of CIS was higher in PUCB (random-effect model, OR 0.71, 95% CI 0.51–0.99, P = 0.04, Supplementary Fig. 6). However, the percentage of LVI and tumor multiplicity showed no significant difference between the two groups (random-effect model, OR 1.52, 95% CI 0.84–2.72, P = 0.16; random-effect model, OR 1.03, 95% CI 0.66–1.60, P = 0.89; Supplementary Fig. 7 and Supplementary Fig. 8). There were no differences in the number of patients who received neoadjuvant chemotherapy (NAC) or adjuvant chemotherapy (AC) (fixed-effect model, OR 0.77, 95% CI 0.48–1.25, P = 0.29; random-effect model, OR 1.10, 95% CI 0.55–2.20, P = 0.78; Supplementary Fig. 9 and Supplementary Fig. 10).

Table 2 Results of meta-analysis of clinicopathological characteristics in patients with UCSD vs. PUCB
Full size table
Fig. 2
figure 2

Forest plots of the recurrence rate of patients who underwent TURBT in patients with UCSD vs. PUCB. UCSD  urothelial carcinoma with squamous differentiation, PUCB   pure urothelial carcinoma of bladder

Full size image
Fig. 3
figure 3

Forest plots of the recurrence rate of patients who underwent RC in patients with UCSD vs. PUCB. UCSD  urothelial carcinoma with squamous differentiation, PUCB   pure urothelial carcinoma of bladder

Full size image
Fig. 4
figure 4

Forest plots of disease-free survival in patients with UCSD vs. PUCB from multivariable analysis. UCSD  urothelial carcinoma with squamous differentiation, PUCB   pure urothelial carcinoma of bladder

Full size image
Fig. 5
figure 5

Forest plots of cancer-specific survival of patients who underwent RC in patients with UCSD vs. PUCB from univariable analysis. UCSD  urothelial carcinoma with squamous differentiation, PUCB   pure urothelial carcinoma of bladder

Full size image
Fig. 6
figure 6

Forest plots of cancer-specific survival of patients who underwent RC in patients with UCSD vs. PUCB from multivariable analysis. UCSD  urothelial carcinoma with squamous differentiation, PUCB   pure urothelial carcinoma of bladder

Full size image

Recurrence

Disease recurrence was reported in six studies. To explore the recurrence rate after TURBT or RC, subgroup analysis was performed according to different treatments (two studies in TURBT group and four studies in RC group). The result of subgroup analysis showed the recurrence rate was significantly higher in patients who had UCSD than patients with PUCB in either the patients who underwent TURBT (fixed-effect model, OR 1.37, 95% CI 1.08–1.75, P = 0.01; Fig. 2) or RC (random-effect model, OR 1.73, 95% CI 1.02–2.92, P = 0.04; Fig. 3).

Prognosis

Disease-free survival

Multivariable adjusted HR of UCSD in predicting DFS of patients with BCa after RC was available in three included studies. The result showed that UCSD was associated with significantly poorer DFS (fixed-effect model, HR 1.42, 95% CI 1.08–1.86, P = 0.01; Fig. 4)

Cancer-specific survival

There were three studies that performed univariable analyses and multivariable analyses for CSS of patients with BCa after RC, respectively. The results revealed that patients who had UCSD had worse CSS on univariable analysis (random-effect model, HR 1.65, 95% CI 1.01–2.70, P = 0.04; Fig. 5). However, UCSD was not associated with CSS on multivariable analysis (random-effect model, HR 1.70, 95% CI 0.98–2.95, P = 0.06; Fig. 6).

Overall survival

Multivariable analysis for OS of patients with BCa after RC was available in five studies. The result demonstrated no significant difference of OS was observed between patients with UCSD and those with PUCB (random-effect model, HR 1.12, 95% CI 0.84–1.5, P = 0.45; Fig. 7).

Fig. 7
figure 7

Forest plots of overall survival in patients with UCSD vs. PUCB from multivariable analysis. UCSD  urothelial carcinoma with squamous differentiation, PUCB   pure urothelial carcinoma of bladder

Full size image

Publication bias

No publication bias was detected in survival outcome comparisons through inverted funnel plot or Egger’s test.

Discussion

As the second common cancer in urinary and male genital system, the BCa survival rate has not changed significantly for the last 30 years [40]. BCa showed high recurrence rate for tumor stage T1 and poor survival for tumor stage T3/4 [40,41,42]. The prognostic factors for patients with BCa were investigated in many studies including patients’ age, gender, prior recurrence rate, number of tumors, tumor size, tumor stage and grade, and the presence of concomitant CIS [43, 44]. However, the impact of histological variants should not be ignored. To the best of our knowledge, this is the first meta-analysis to investigate the clinicopathological characteristics and prognosis of different outcomes in patients with UCSD. We found that the percentage of female patients who had UCSD was significantly higher than those with PUCB. UCSD was associated with high tumor grade, pathological tumor stage T3/T4, positive surgical margin and lymph node involvement. The percentage of CIS was more common in PUCB. For the survival outcomes, we revealed that patients who had UCSD had a higher recurrence rate and worse DFS prognosis than patients with PUCB after RC, but no significant difference of CSS and OS was found between these two groups on multivariable analyses.

Although the incidence of BCa is higher in men than in women, female patients with BCa are more likely diagnosed with advanced stage tumors and have higher bladder CSM rate [45, 46]. For pathological tumor stage T3–T4 urothelial cancer of bladder, Liberman et al. [47] reported that female gender was independently associated with higher CSM. From the analysis of patients with pT4 BCa who underwent RC, Tilki et al. [48] reported that female gender was independently associated with significantly higher risk of disease recurrence and CSM, and May et al. [49] found that 5-year CSS in female patients was significantly lower than that in male patients. Interestingly the percentage of female patients and pathological tumor stage T3/T4 in our analysis was higher in patients with UCSD than those with PUCB, which could partly explain the difference in BCa recurrence rate and CSS between patients with UCSD and PUCB. We also found the percentage of high tumor grade, positive surgical margin rate and lymph node involvement rate were higher in patients with UCSD, which could contribute to the worse survival compared to patients with PUCB. These findings indicated that the adequate resection range and lymphadenectomy may both be helpful to improve the survival outcomes for patients with UCSD. Furthermore, we found that the percentage of CIS was lower in patients with UCSD than PUCB. Mazzucchelli et al. [50] found that the presence of CIS was associated with a 60% reduction of the risk of death for patients with BCa treated with RC. Moschini et al. [51] reported that the presence of CIS was associated to an increased risk of urothelial recurrence and worse CSM in pT0–pT2 patients. Yafi et al. [52] reported that the presence of concomitant CIS on cystectomy specimens did not independently affect outcomes of patients treated with RC with pelvic lymph node dissection. The results indicated that concomitant CIS maybe a predictive factor for a favorable outcome, but the impact of CIS in UCSD is still unclear.

Recently, a meta-analysis [53] specifically assessed the clinical implications of histological variants in urothelial carcinoma of the bladder. They performed a subgroup analysis according to the histological variants type and found that the pooled HR was 0.97 (95% CI 0.83, 1.11, P = 0.701) for UCSD without any statistical significance. But they did not analyze different endpoints (CSS and OS) separately and put the data of HR from univariable and multivariable analyses together. We conducted a subgroup analysis and found that UCSD was associated with CSS on univariable analysis but was not an independent prognostic factor for CSS and OS on multivariable analysis. Several previous studies also reported that the presence of SD was associated with survival outcomes on univariable analyses [21, 33, 54,55,56,57], but not on multivariable analyses [8, 22, 24, 31, 56, 57]. These results indicated that the poor survival outcomes in patients with UCSD were associated with the unfavorable clinicopathological characteristics of tumor, such as high percentage of pT3/T4 and high tumor grade. Another potential reason for different results obtained from previous studies could be due to the different extent of SD in different patients. Liu et al. [29] found that the lymph node-positive rate was significantly higher in urothelial carcinoma with extensive SD when compared with that seen in PUCB, but not in urothelial carcinoma with focal SD. Mitra et al. [23] assessed the outcome predictors in patients with UCSD and found that wide spread presence of SD elements was associated with poor OS on univariable analysis but not on multivariable analysis. Currently, the clinical effect of different extent of SD is still unknown [17]; further study is required to make it clear.

Honma et al. [58] performed a retrospective study included 145 patients with muscle-invasive BCa treated with RC and revealed that a concomitant squamous cell carcinoma component in the specimen was the only independent predictor of local recurrence after surgery. In our meta-analysis, the recurrence rate was significantly higher in UCSD than PUCB regardless of patients who underwent TURBT or RC. Furthermore, we detected that UCSD was not associated with CSS for patients who underwent RC only. For patients who underwent TURBT, Li et al. [21] found that patients treated with re-TURBT had shorter median CSS than those who underwent RC in patients with UCSD. These results suggested that patients with UCSD treated with RC may have better outcome than patients treated with TURBT, especially for recurrence after TURBT. Our results suggest close follow-up will be necessary for patients with UCSD to detect the recurrence early, therefore, to improve CSS.

Regarding NAC or AC treatment, there were five studies that included patients who received NAC or AC in this meta-analysis, but the effect of chemotherapy for patients with UCSD is undetermined. To assess the impact of NAC and AC in patients with variant histology at RC, two studies from Vetterlein et al. [59] and Berg et al. [60] showed that there was no significant OS benefit from NAC for patients with histological variants except neuroendocrine tumors, and no survival benefit of AC for patients with concomitant variant histology or pure variant histology. Minato et al. [20] reported 9 patients with UCSD treated with NAC followed by RC. None of the patients with UCSD achieved pathologic complete response and only one patient had pathologic downstaging. Mitra et al. [23] also found that NAC and AC administration were not significantly associated with recurrence-free survival or OS. However, Buisan et al. [61] revealed that treatment with NAC was a significant prognostic factor in multivariate analysis for PFS and CSS in patients with squamous-cell-feature muscle-invasive bladder, and found that patients with low neutrophil-to-lymphocyte ratio treated with NAC had significantly better outcomes. Thus, the patients with UCSD may benefit from multimodality therapy combined with surgery; however, further research will be required to determine the group of patients who will have better response to chemotherapy.

Several limitations should be acknowledged in this study. Firstly, there is no standard of the percentage of SD for diagnosis of variant histology, the extent of SD between different studies may be inconsistent, and the samples from TURBT may be not enough to accurately evaluate the presence of SD [62]. The different criteria for diagnosing SD may contribute to potential bias. Secondly, some studies with a small population of patients were included in our analysis, and the population from different countries. The data were not enough to perform the subgroup analysis according to race. Thirdly, the presence of heterogeneity between the enrolled studies still existed in many data analyses, which could influence the results. Furthermore, the publications which were not written in English or Chinese, or whose complete data or full text were not available were not included; this may introduce a potential enrollment bias. Finally, subgroup analyses on different gender, tumor stage or grade were not performed due to the lack of sufficient data available.

Conclusion

Our study demonstrated that the presence of SD in bladder urothelial carcinoma was associated with poor clinicopathological features, high risk of recurrence and worse clinical outcome for disease-free survival. SD should be considered as a marker for prognosis and treatment for patients with BCa. In the future, a well-designed randomized study is warranted to verify our findings and to further explore the role of SD in urinary bladder cancer.