Background MicroRNA-221 (miR-221) has been shown to play an important role in cancer prognosis. In order to evaluate the predictive value of miR-221, we compiled the evidence from 20 eligible studies to perform a meta-analysis. Design All of relevant studies were identified by searching PubMed, Embase, and Web of Science, and were assessed by further quality evaluation. Pooled hazard ratios (HRs) with 95% confidence intervals (CIs) of total and stratified analyses, for overall survival (OS) and recurrence-free survival (RFS), were calculated to investigate the association between high miR-221 expression and cancer prognosis. Results We found that high miR-221 expression can predict a poor OS in malignant tumors (pooled HR = 1.55, P = 0.017) but has no significant association with RFS (pooled HR = 1.02, P = 0.942). Further in stratified analyses, high miR-221 expression was significantly associated with a poor OS in Asians (pooled HR = 2.04, P = 0.010) or serum/ plasma subgroup (pooled HR = 2.28, P<0.001), and even showed significantly poor OS (pooled HR = 1.80, P<0.001) and RFS (pooled HR = 2.43, P = 0.010) in hepatocellular carcinoma (HCC) subgroup, but was correlated to a favorable RFS in prostate cancer subgroup (pooled HR = 0.51, P = 0.004). Conclusions Our findings demonstrate that miR-221 is more suitable to predict cancer prognosis in Asians, and it is a promising prognostic biomarker for HCC. The detection of miR-221 in serum or plasma samples may make it become an effective method for monitoring patients' prognosis and assessing therapeutic efficacy in the future.