A systematic literature search through PubMed, Embase, Web of Science, the Cochrane Library and CBM databases was conducted to identify the relevant clinical studies up to November, 2014. The search terms were as follows: (CRP OR C-reactive protein) AND (ureteral neoplasms OR ureter cancer OR ureter carcinoma OR renal pelvis cancer OR renal pelvic carcinoma OR upper tract urothelial carcinoma OR UTUC). The reference lists of the retrieved articles were searched for any additional relevant studies. The initial selection was performed based on the title and abstract by two independent investigators (Y. Luo and D.L. She). Thereafter, the full text was reviewed according to the eligibility criteria. To be eligible for this analysis, studies were required to evaluate the association between preoperative CRP and UTUC prognosis, including recurrence-free survival (RFS), cancer-specific survival (CSS) or overall survival (OS). Studies with duplicated or overlapping patient data and studies without sufficient available data were excluded.

The quality of the studies was independently assessed by the Newcastle-Ottawa scale (NOS) (8). Data were extracted by two independent investigators (Y. Luo and D.L. She) and cross-checked. Disagreements were resolved through discussion. The extracted data included the name of the first author, year of publication, source of patients, cut-off value of CRP, prognostic outcomes delineated, sample size, geographical region and follow-up time. The data were extracted from the original articles. When required data were missing, the original research data were requested directly from the corresponding authors via e-mail. When the original data were obtained, the cut-off value was set at 0.5 mg/dl; alternatively, the provided survival or mortality curves were used to calculate the hazard ratio (HR) and 95% confidence interval (CI), as described by Tierney et al (9).

Review Manager 5.3 software (The Cochrane Collaboration, Copenhagen, Denmark) was used for the analysis. As the number of included studies was <10, quantitative publication bias detection with Egger's regression intercept test or meta-regression analysis was not performed (10–12). HR was the time-to-event effect estimate for survival analysis and the generic inverse variance method was used. First, Cochran's Q test and Higgins I² statistic were used to estimate heterogeneity (11). when P≥0.1 and I²≤50%, there was no significant heterogeneity and the fixed-effects model was used. In any other case, the random-effects model was used to calculate pooled HR. Univariate and multivariate analyses were performed and two-tailed P-values of <0.05 were considered to indicate statistically significant differences. All the results are presented in the forest plots.

A total of 85 citations were selected by the initial search strategy. After reading the title and abstract, 71 articles were excluded for unrelated information. The remaining 14 articles were read and relative information was extracted. A further 3 studies were excluded due to duplicated references or overlapping patients and 4 studies due to insufficient survival data. Finally, 7 retrospective cohort studies including a total of 1,919 patients were entered in the analysis (13–19). The flow chart of the study selection process is presented in Fig. 1 and the basic characteristics of the included studies are summarized in Table I. The NOS scores were used to identify included studies of high methodological quality. A total of 5 articles reported RFS, 6 reported CSS and only 2 reported OS.

The meta-analysis results are presented in Figs. 2–4. As shown in Fig. 2, high CRP level was a significant risk factor for UTUC recurrence in the univariate as well as the multivariate analysis. The comprehensive HR for RFS was 2.90 (95% CI: 1.87–4.51, P<0.00001) in the univariate and 1.57 (95% CI: 1.26–1.97, P<0.0001) in the multivariate analysis. As shown in Fig. 3, high CRP level was a predictive risk factor for CSS of UTUC in the univariate as well as the multivariate analysis, with pooled HRs of 2.78 (95% CI: 1.75–4.43, P<0.0001) and 1.64 (95% CI: 1.32–2.03, P<0.00001), respectively. Patients with a high preoperative CRP level exhibited poorer CSS compared with patients with a low CRP level. Only 2 studies reported OS (Fig. 4); the pooled HRs of univariate and multivariate analysis were not significant [1.24 (95% CI: 0.72–2.15, P=0.43) and 1.24 (95% CI: 0.88–1.75, P=0.22), respectively].

Of note, 2 of the included studies utilized inconsistent cut-off values for CRP level, which caused significant heterogeneity as verified by a sensitivity analysis (data not shown). However, the statistical significance remained the same even following sensitivity analysis. A more significant effect may have been observed if the cut-off value in these two studies was set at 0.5 mg/dl (compared with 0.13 or 0.9 mg/dl; Table I).