Combination of multiple mRNA markers has been largely investigated for detection of circulating cancer cells. However, current PCR-based methods are relatively expensive and time consuming. The aim of this study was to develop a membrane array-based multimarker assay for detection of circulating cancer cells in nonsmall cell lung cancer (NSCLC) patients. At first, we selected 22 candidate genes by means of suppression subtractive hybridization and Northern blot analysis. The diagnostic value of each candidate gene was then preliminarily evaluated in 50 pairs of blood samples by membrane array method. Accordingly, 17 genes with area under the ROC curve (AUC) > or = 0.8 were selected as target genes to reconstruct the diagnostic membrane array, which was then used to test peripheral blood samples from 100 NSCLC patients and 147 control subjects. ROC curve analysis demonstrated that the optimal threshold number of overexpressed markers on membrane array for discrimination between NSCLC patients and control subjects was 12. As a result, the diagnostic membrane array could detect circulating cancer cells in 90 (90%) of 100 NSCLC patients and in 14 (9.5%) of 147 control subjects (including 6 of 100 normal persons, 3 of 20 breast cancer patients, 3 of 15 colorectal cancer patients and 2 of 12 gastric cancer patients). Moreover, the detection rate was significantly correlated with NSCLC patients' metastatic status and overall stage (p = 0.028 and 0.014, respectively). These results suggested that our blood-based membrane array assay for molecular detection of circulating lung cancer cells has great potential for clinical applications.