OBJECTIVE: Owing to inconsistency between reports, a meta-analysis was designed to appraise the clinical implications of long non-coding RNAs (lncRNAs) in urine and blood for the diagnosis of bladder cancer. METHODS: Studies that met the criteria were acquired by bibliographic retrieval through PubMed, Embase, and the Cochrane Library. The pooled diagnostic performance was evaluated by calculating the area under the summary receiver operator characteristic (SROC) curve. The potential sources of heterogeneity were approached through meta-regression and subgroup analyses. All statistical analyses and plots were performed by RevMan 5.3, Meta-DiSc 1.4, and STATA 12.0. RESULTS: A total of 43 studies from 15 articles consisting of 3370 bladder cancer patients and 3212 controls were incorporated in our meta-analysis. lncRNAs in urine and blood performed relatively well in diagnosing bladder cancer, with a pooled sensitivity of 0.78, a specificity of 0.79, and an area under the SROC curve (AUC) of 0.86. H19 displayed the best diagnostic accuracy with a pooled AUC of 0.90, followed by UCA1 and MALAT1. The heterogeneity among studies was partly conducted by sample size, lncRNA existence form (cell-free or intracellular lncRNA), lncRNA origin (exosome- or non-exosome-based lncRNA), lncRNA profiling (single- or multiple-lncRNA), specimen types, and ethnicity. CONCLUSIONS: lncRNAs in urine and blood may serve as noninvasive diagnostic biomarkers with great promise for bladder cancer, while their clinical values need to be examined through further synthetic forward-looking studies.