Background

Bladder cancer (BCa) is the most common malignancy of the urinary tract and occurs with a relatively high incidence in developing countries [1], with annual mortality rates ranging from approximately 1–5 deaths per 100,000 men and 0.5–1.5 deaths per 100,000 women [2]. Radical cystectomy (RC) with bilateral pelvic lymph node dissection is the gold standard for patients with localized muscle-invasive tumors. Despite a better understanding of BCa biology and the use of adjuvant therapies, BCa continues to have high mortality rates, and the oncological outcomes following RC have not changed in the last 30 years [3].

BCa prognoses vary widely. Many factors have been investigated as potential predictors of clinical outcome in BCa. Positive soft tissue surgical margins (STSM) [4], lymphovascular invasion (LVI) [5], lymph node metastasis (LNM) [6], concomitant carcinoma in situ (CIS) [7], and failure to receive adjuvant chemotherapy (ACT) [8] have been reported to be associated with poor prognoses for BCa after RC. Although these predictive variables have contributed to estimating the BCa recurrence risk and survival outcomes, additional variables that can integrate with well-established prognostic models and provide accurate risk grading for BCa patients after RC are critical.

A major problem for urologists is identifying prognostic factors that can predict cancer progression. The ability to determine cancer-specific survival (CSS) and provide integrated patient survivorship and better estimates of survival probability at each follow-up may lead to more informative prognostic information in patient monitoring [9].Therefore, we aimed to provide a comprehensive systematic review and meta-analysis of previous studies to investigate the prognostic roles of pathological status and clinical variables for CSS in patients following RC. We identified ten common clinicopathological characteristics that should be systematically assessed to guide postoperative decision-making after RC.

Methods

Search strategy

In line with the guidelines of Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) [10], the electronic database of PubMed, EMBASE and Web of Science were searched for studies published prior to November 2018. The following search term combinations were used: ‘urinary bladder neoplasms’, ‘bladder and neoplasms’, ‘radical cystectomy’, ‘cancer-specific survival’, ‘clinical’, and ‘pathological’. The publication language was restricted to English. In addition, the reference lists of the identified studies were also searched manually.

Inclusion and exclusion criteria

The inclusion criteria were as follows: (1) all patients with BCa were pathologically confirmed; (2) the study included prognostic factors for CSS following radical cystectomy; (3) treatment was limited to RC in all studies; and (4) the authors provided the hazard ratios (HRs) and 95% confidence intervals(CIs). The exclusion criteria were: (1) duplicates; (2) lack of sufficient data (HRs and CIs) for further analysis; and (3) case reports, reviews, letters, author replies, expert opinions or meeting abstracts. If the data overlapped across several different articles, only the most recent and informative article was selected.

Data extraction and qualitative assessment

Two authors extracted the information from the selected studies. Any disagreement between the reviewers was resolved by discussion with a third author. The following information were collected from eligible studies: first author’s name, publication date, country, recruitment period, follow-up time, sample size, patient’s age, pathological stage, tumor grade, histopathological subtype in transitional cell cancer (TCC) and the HR and 95% CIs for CSS.

We evaluated the study quality using the 9-star Newcastle-Ottawa Scale (NOS) [11]. Scores of 7–9 indicated a high-quality study, and scores < 7 indicated a low-quality. The cohort study quality was assessed as follows: object selection, inter-group comparability, and outcome measurement. Dichotomous variables were presented as HRs with 95% CIs. If the data results were calculated by multivariate and univariate analysis simultaneously, the multivariate analyses were used.

Statistical analysis

All calculations were performed using STATA 12.0 software (Stata Corp LP, College Station, TX, USA). Heterogeneity was estimated using the Higgins I-squared statistic test, and Pheterogeneity ≤ 0.1 or I2 > 50%. indicated heterogeneity among studies. When significant heterogeneity was observed among the studies, a random-effect (RE) model was used; otherwise, we adopted a fixed-effect (FE) model. To explore the source of heterogeneity, subgroup analysis was performed for CSS. Sensitivity analysis was conducted by excluding single studies one by one to examine the stability and reliability of the pooled results. A funnel plot and Egger’s test were used to statistically evaluate the publication bias between studies. Two-tailed P < 0.05 was considered statistically significant.

Results

Literature search

From the search criteria, 887 articles were identified from the databases and the manual search. Of these articles, 664 studies were excluded based on their titles and/or abstracts, resulting in 223 studies for further analysis. The full texts were then screened, and 190 papers were excluded because of insufficient survival information or duplicated cohorts. Finally, 33 studies [3, 5, 6, 8, 12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40] containing 19,702 patients (range 51–2,944) were included as per the eligibility criteria. Figure 1 presents a flowchart of the study selection process.

Fig. 1
figure 1

Flowchart of the literature search used in this meta-analysis

Full size image

Characteristics of eligible studies

Tables 1 and 2 summarize the main characteristics and clinicopathological outcomes of the 33 included studies. All studies were performed retrospectively, and all were published between 2007 and 2018. Of the included studies, 11 were conducted in Asia, 8 in Europe, 7 in North America, 4 at international multicenters, 3 in Turkey and 1 in Australia. Histopathological examinations were performed on resected tumor specimens. All studies used CSS as a common endpoint to evaluate the prognostic value of the clinicopathological indicators of survival. The quality scores of the studies ranged from 7 to 9.Therefore, all included studies were of high quality (studies with a score ≥ 7; Additional file 2: Table S1).

Table 1 Main characteristics of the studies included in the meta-analysis
Full size table
Table 2 Tumor characteristics of all studies included in the meta-analysis
Full size table

Meta-analysis

Our meta-analysis demonstrated that advanced age (old vs. young: pooled HR = 1.01; 95% CI: 1.00–1.01; P < 0.001; I2 = 68.2%, Pheterogeneity < 0.001; Fig. 2A), higher tumor grade (3 vs. 1/2: pooled HR = 1.29; 95% CI: 1.15–1.45; Pheterogeneity < 0.001; I2 = 76.9%, Pheterogeneity < 0.001; Fig. 2B), higher pathological stage (3/ 4 vs. 1/ 2: pooled HR = 1.60; 95% CI: 1.37–1.86; P < 0.001; I2 = 92.2%, Pheterogeneity < 0.001; Fig. 2C), LNM (positive vs. negative: pooled HR = 1.51; 95% CI: 1.37–1.67; Pheterogeneity < 0.001; I2 = 95%, P < 0.001; Fig. 2D), LVI (positive vs. negative: pooled HR = 1.36; 95% CI: 1.28–1.45; P < 0.001; I2 = 68.4%, Pheterogeneity < 0.001; Fig. 2E), and STSM (positive vs. negative: pooled HR = 1.42; 95% CI: 1.30–1.56; P < 0.001; I2 = 71.7%, Pheterogeneity < 0.001; Fig. 2F) in BCa were associated with poor CSS. However, no significant correlations were observed regarding gender (male vs. female: pooled HR = 0.98; 95% CI: 0.96–1.01; P = 0.278; I2 = 34.9%, Pheterogeneity = 0.036; Fig. 3A), CIS (positive vs. negative: pooled HR = 0.98; 95% CI: 0.88–1.10; P = 0.753; I2 = 78%, Pheterogeneity < 0.001; Fig. 3B), histology (TCC vs variant: pooled HR = 0.90; 95% CI: 0.79–1.02; P = 0.089; I2 = 71.6%, Pheterogeneity = 0.003; Fig. 3C) or ACT (yes vs. no: pooled HR = 1.16; 95% CI: 1.00–1.34; P = 0.054; I2 = 93.8%, Pheterogeneity < 0.001; Fig. 3D).

Fig. 2
figure 2

Meta-analysis of studies that examined the association between: (2A) advanced age, (2B) higher tumor grade, (2C) higher pathological stage, (2D) LNM, (2E) LVI, (2F) STSM and CSS following radical cystectomy (RC)

Full size image
Fig. 3
figure 3

Meta-analysis of studies that examined the association between: (3A) gender, (3B) CIS, (3C) histology, (3D) ACT and CSS following radical cystectomy (RC)

Full size image

To explore the source of heterogeneity for advanced age, tumor grade, pathological stage, LNM, LVI, STSM, CIS and ACT, their significance levels were further evaluated via subgroup analysis based on geographical region (Asia vs. non-Asia), year of publication (≥2015 vs. < 2015), number of patients (≥500 vs. < 500) and median follow-up (≥36 months vs. < 36 months). Because few studies were included in the histology group, no subgroup analysis was conducted for histology. Table 3 presents the subgroup analysis results for CSS. Notably, we observed a significant decline in heterogeneity for CSS in some categories, such as in articles published before 2015, studies with sample sizes of < 500 cases and median follow-ups of < 36 months. The subgroup analysis results were consistent with the primary findings.

Table 3 Summary and subgroup results of the association between common clinicopathological characteristics and BCa
Full size table

Sensitivity analysis

The pooled HR for CSS for advanced age ranged from 1.01 (95% CI:1.00–1.01) to 1.01 (95% CI:1.00–1.01), for gender ranged from 0.98 (95% CI: 0.94–1.02) to 0.99 (95% CI: 0.99–1.04), for tumor grade ranged from 1.25 (95% CI: 1.11–1.41) to 1.34 (95% CI: 1.16–1.54), for pathological stage ranged from 1.53 (95% CI: 1.31–1.79) to 1.68 (95% CI: 1.45–1.95), for LNM ranged from 1.49 (95% CI: 1.35–1.64) to 1.52 (95% CI: 1.37–1.68), for LVI ranged from 1.34 (95% CI: 1.26–1.42) to 1.38 (95% CI: 1.30–1.47), for STSM ranged from 1.34 (95% CI: 1.26–1.43) to 1.44 (95% CI: 1.29–1.61), for CIS ranged from 0.95 (95% CI: 0.86–1.05) to 1.01 (95% CI: 0.89–1.14), for histology ranged from 0.86 (95% CI: 0.76–0.97) to 0.94 (95% CI: 0.82–1.07), and for ACT ranged from 1.12 (95% CI: 0.97–1.29) to 1.19 (95% CI: 1.02–1.38) (Additional file 1: Figure S1).These results indicated that our findings were reliable and robust.

Publication bias

Figure 4 shows the funnel plots for publication bias. Egger’s test demonstrated that no publication bias existed regarding advanced age (p Egger = 0.427, Fig. 4A), gender (p Egger = 0.487, Fig. 4B), CIS (p Egger = 0.172, Fig. 4C), LVI (p Egger = 0.797, Fig. 4D), pathological stage (p Egger = 0.330, Fig. 4E), STSM (p Egger = 0.134, Fig. 4F), histology (p Egger = 0.648, Fig. 4G) and ATC (p Egger = 0.266, Fig. 4H). However, publication biases were found for tumor grade (p Egger = 0.023, Fig. 4I) and LNM (p Egger< 0.001, Fig. 4J), suggesting that publication bias may have played a potential role in tumor grade and LNM.

Fig. 4
figure 4

Funnel plots for the publication bias test. Each point represents a separate study for the indicated association. The vertical line represents the mean effects size: (4A) advanced age; (4B) gender; (4C) CIS; (4D) LVI; (4E) pathological stage; (4F) STSM; (4G) histology; (4H) ATC; (4I) tumor grade and (4 J) LNM

Full size image

Discussion

Despite modern advancements in surgical techniques, the oncological outcomes of BCa remains poor. The 5-yr overall survival rates were only 60% according to a multicenter database [41]. Determining the probability of CSS after RC is difficult because it can vary according to the different clinical features and various tumor characteristics. The traditional clinicopathological features, such as sex [34], pathological tumor stage or grade [25] and LNM [6], have been identified as important parameters with prognostic predictive value and contribute to postoperative clinical decision making based on some nomograms.

Currently, the TNM staging system, which is based on pathological tumor stage and grade, tumor histological subtype, and lymph node status [42] is the most commonly used preoperative model for predicting CSS in BCa patients. Another predictive model is the European Organisation for the Research and Treatment of Cancer (EORTC) risk stratification scheme [43], which uses grade (World Health Organization [WHO] 1973), stage, CIS, multiplicity, size and previous recurrence rate to determine the risk of CSS after RC. Although these two traditional prognostic models have been externally validated, significant variations were founded in some studies. Variations in tumor outcomes may have been related to the heterogeneity of BCa biology and different clinicopathological features in patients with BCa.

Tumor markers that can accurately predict the oncological outcomes in BCa patients when applied with other pathological parameters are essential for clinical decision making. Some published studies on molecular biomarkers, such as luminal and basal subtypes [44], the gene alterations nuclear matrix protein number 22 [45], and the bladder tumor antigen (BTA) stat test [46], have been adopted in recent years to improve diagnosing and managing patients receiving RC. However, none of these biomarkers have been shown to be sufficiently sensitive or specific in predicting survival outcomes. Therefore, in this study, we exploited more validated prognostic factors, including clinical variables (age, gender), pathological information (tumor stage and grade, LNM and STSM, LVI, CIS, and histology), and whether adjuvant therapy (ACT) was received for predicting CSS in BCa patients.

This is the first study to systematically assess the association between ten clinicopathological features and CSS of BCa in a single study. To improve the statistical power and provide more credible results, 33 cohort studies with a large combined sample size of 19,702 BCa patients who underwent RC were pooled in our study. Strictly adhering to the inclusion and exclusion criteria, we extracted the raw data from the relevant studies. The results revealed that advanced age, higher tumor grade, LNM, LVI, and positive STSM significantly predicted the CSS of BCa patients (all P ≤ 0.05). Hence, these clinicopathological findings were independent risk factors in this meta-analysis. Besieds, all the results were reliable and robust via the subgroup and sensitivity analyses.

Interestingly, our results indicated that gender, CIS, histology and ACT may not be associated with CSS. Studies on gender, histology and CIS as prognostic factors for BCa patients have stimulated considerable interest, but the results remain controversial and ambiguous for managing BCa. Some investigators reported that gender and CIS had independent prognostic significance [14, 34, 47], while others considered that gender and CIS may not be significant factors in determining terminal prognosis compared with other widely used prognostic indicators [18, 48, 49]. Additionally, administering ACT after RC in patients with high-risk BCa remains a challenge for clinical urologists. Despite numerous studies being published, no level 1 evidence has demonstrated that ATC confers a significant survival benefit to BCa patients after RC [50]. In the present study, rigorous data analysis indicated that these three factors may not affect the CSS prognosis of patients with BCa.

Although this was a comprehensive meta-analysis, the present study had several limitations. First, most included studies were retrospective cohort studies, and data extracted from those studies may have led to inherent bias. Thus, a prospective multicenter trial providing more definite answers is needed. Second, substantial heterogeneity was observed in some studies. Although we found no possible source of heterogeneity after several subgroup analyses, the conclusions drawn from this meta-analysis should be approached with caution. However, the pooled results in most of the subgroup analyses were consistent with the overall findings. Third, the studies retrieved for our analysis were limited to those published in English, which may result in a language bias. Studies with negative results are not often published in English-language journals [51]; thus, our research may contain some publication bias.

Conclusions

In summary, the data from this meta-analysis indicate that BCa patients with advanced age, higher tumor grade, LNM, LVI, and positive STSM are likely to have poorer CSS, suggesting that these parameters may be independent indicators of BCa in patients following RC. In contrast with what is seen clinical practice, gender, CIS, histology and postoperative ACT were not predictors of CSS in patients with BCa. We identified significant patient-specific (age) and tumor-specific (higher tumor grade, LNM, LVI, and positive STSM) predictors of CSS to propose a risk-based strategy for choosing surveillance and postoperative treatment options. Despite our rigorous systematic approach, further large, prospective studies are needed to confirm our findings considering the inherent limitations of the included studies.