Study selection
The flowchart for studies selection process was presented in Figure 1. Initially, 683 articles were found after searching in the three databases. Two hundred and four (204) duplicated articles were excluded. For the rest 479 articles, after titles and abstracts review, 406 were further excluded (reasons for exclusion are: not bladder cancer, editorial review, did not analyze CRP, reviews or meta-analysis, published in other languages rather than English, animal studies, conference abstracts, not prognostic studies). Then 73 articles were reviewed in full text to carefully assess the eligibility, 60 of them were further excluded (reasons for exclusion are: other types of urological cancer patients mingled with bladder cancer patients, did not provide measurements of CRP, did not investigate the outcomes of study interest, case-control or cross-sectional design studies, did not use Cox proportional hazards model). In the end, 13 studies were included into our meta-analysis [12–15, 18–26].
Study characteristics
The characteristics of included studies were summarized in Table 1. A total of 3072 patients were analyzed in these 13 studies. The sample size ranged from 34 to 1043. The majority of the patients were men (77%). Nine studies were conducted in Asian countries (5 in Japan, 2 in China, 2 in Korea), and 4 were from European countries (2 in Germany, 1 in Belgium, 1 in the United Kingdom). The age of the patients ranged from 65 to 72 years old, with the medians of follow-up from 7 to 63 months.
CRP with bladder cancer survival
A total of 8 included studies investigated OS, high heterogeneity was observed among them ( =97%, < 0.001), therefore a random-effect model was fitted. The pooled results indicated that, an elevated CRP was in general associated with poor OS of bladder cancer patients (HR=2.24, 95% CI: 1.16-4.34, p=0.017). We also combined studies which reported prognostic significance of CRP in CSS and DFS of bladder cancer by using fixed-effect model, we found that the elevated CRP was also significantly associated with poor CSS (HR=1.53 95% CI: 1.36-1.72, p<0.001) and DFS (HR=2.07, 95% CI: 1.24-3.35, p=0.005) of bladder cancer (Figure 2).
Considering that different cut-offs of CRP may influence the combined results, we performed a meta-analysis by only including studies which adopted the cut-off of 0.5mg/dl or close. Altogether 6 and 3 studies were combined for CSS and OS, and the results of random-effect models revealed that, an elevated CRP level was only prominently associated with CSS (HR=1.53, 95%CI: 1.26-1.86, p<0.001) (Figure 3).
Subgroup analysis
In order to test for robustness of the combined results, we performed a series of subgroup analysis based on different characteristics of included studies for CSS, the most popular outcome of interest among included studies. Different cut-offs for CRP (0.5mg/dl or close VS. 1 mg/dl, Figure 4A), different types of bladder cancer (Unspecified VS. MIBC, Figure 4B), different sample sizes of studies (dichotomized by the mean, <236 VS. ≥236, Figure 4C), different origins of studies (Europe VS. Asia, Figure 4D) were used sequentially to perform subgroup analyses. Results were generally robust with regard to different dichotomization characteristics.
Publication Bias
Begg’s and Eggs tests were used to evaluate publication bias. Funnel plots for CSS (Figure 5A), OS (Figure 5B) and DFS (Figure 5C) suggested insignificant publication bias in general.