In order to identify novel non-invasive biomarkers with high accuracy for the screening of non-small cell lung cancer (NSCLC), we investigated the predictive power of 4 microRNAs (miR-146, miR-204, miR-106a and miR-124) in plasma samples obtained from patients with NSCLC and healthy controls (n=50; training phase) by reverse transcription-quantitative polymerase chain reaction (RT-qPCR). We found that the levels of miR-204 in the patients with NSCLC were significantly dysregulated compared with the healthy controls, and thus this miRNA was selected for further validation. For the validation phase, RT-qPCR was performed on plasma samples from patients with NSCLC and healthy controls (n=176) in order to examine the expression levels of the candidate miRNA, miR-204. The results revealed that the plasma levels of miR-204 were significantly downregulated in the patients with NSCLC compared with the healthy controls (p<0.001). The value of the area under the receiver operating characteristic (ROC) curve obtained for miR-204 was 0.809 (sensitivity, 76%; specificity, 82%), which was higher than the values obtained for carcinoembryonic antigen (CEA) and carbohydrate antigen 19-9 (CA19-9). The expression of miR-204 in plasma significantly correlated with the tumor stage (p=0.009) and distant metastasis (p=0.048). A log-rank test revealed that lower plasma levels of miR-204 were associated with a shorter overall survival and disease-free survival (p=0.006 and 0.0065, respectively). Both univariate and multivariate Cox regression analyses indicated that a lower miR-204 expression level in plasma was a prognostic factor with a relative risk of death of 1.936 and 1.712, respectively. On the whole, our results suggest that the decreased expression of miR-204 in plasma is a promising biomarker for the detection of NSCLC and the prediction of poor survival in patients with the disease.