Summary Background MicroRNAs (miRNAs) can be used as prognostic biomarkers in many types of cancer. We aimed to identify miRNAs that were prognostic in patients with nasopharyngeal carcinoma. Methods ...We retrospectively analysed miRNA expression profiles in 312 paraffin-embedded specimens of nasopharyngeal carcinoma from Sun Yat-sen University Cancer Center (Guangzhou, China) and 18 specimens of non-cancer nasopharyngitis. Using an 873 probe microarray, we assessed associations between miRNA signatures and clinical outcome in a randomly selected 156 samples (training set) and validated findings in the remaining 156 samples (internal validation set). We confirmed the miRNAs signature using quantitative RT-PCR analysis in 156 samples from a second randomisation of the 312 samples, and validated the miRNA signature in 153 samples from the West China Hospital of Sichuan University in Chengdu, China (independent set). We used the Kaplan-Meier method and log-rank tests to estimate correlations of the miRNA signature with disease-free survival (DFS), distant metastasis-free survival (DMFS), and overall survival. Findings 41 miRNAs were differentially expressed between nasopharyngeal carcinoma and non-cancer nasopharyngitis tissues. A signature of five miRNAs, each significantly associated with DFS, was identified in the training set. We calculated a risk score from the signature and classified patients as high risk or low risk. Compared with patients with low-risk scores, patients with high risk scores in the training set had shorter DFS (hazard ratio HR 2·73, 95% CI 1·46–5·11; p=0·0019), DMFS (3·48, 1·57–7·75; p=0·0020), and overall survival (2·48, 1·24–4·96; p=0·010). We noted equivalent findings in the internal validation set for DFS (2·47, 1·32–4·61; p=0·0052), DMFS (2·28, 1·09–4·80; p=0·030), and overall survival (2·87, 1·38–5·96; p=0·0051) and in the independent set for DFS (3·16, 1·65–6·04; p=0·0011), DMFS (2·39, 1·05–5·42; p=0·037), and overall survival (3·07, 1·34–7·01; p=0·0082). The five-miRNA signature was an independent prognostic factor. A combination of this signature and TNM stage had better prognostic value than did TNM stage alone in the training set (area under receiver operating characteristics 0·68 95% CI 0·60–0·76 vs 0·60 0·52–0·67; p=0·013), the internal validation set (0·70 0·61–0·78 vs 0·61 0·54–0·68; p=0·012), and the independent set (0·70 0·62–0·78 vs 0·63 0·56–0·69; p=0·032). Interpretation Identification of patients with the five-miRNA signature might add prognostic value to the TNM staging system and inform treatment decisions for patients at high risk of progression. Funding Science Foundation of Chinese Ministry of Health, National Natural Science Foundation of China, Pearl River Scholar Funded Scheme, Guangdong Key Scientific and Technological Innovation Program, Guangdong Natural Science Foundation, Fundamental Research Funds for the Central Universities.
