Whether Cell block (CB) samples are applicable to detect anaplastic lymphoma kinase (ALK), c-ros oncogene 1 (ROS1) and ret proto-oncogene (RET) fusion genes in lung adenocarcinoma is still unknown. In this study, 108 cytological samples that contained lung adenocarcinoma cells were collected, and made into CB. The CB samples all contained at least 30% lung adenocarcinoma cells. In these patients, 48 harbored EGFR mutation. Among the 50 EGFR wild type patients who detected fusion genes, 14 carried EML4-ALK fusion (28%), 2 had TPM3-ROS1 fusion (4%), and 3 harbored KIF5B-RET fusion (6%). No double fusions were found in one sample. Patients with fusion genes were younger than those without fusion genes (p = 0.032), but no significant difference was found in sex and smoking status (p > 0.05). In the thirty-five patients who received first-line chemotherapy, patients with fusion gene positive had disease control rate (DCR) (72.7% VS 50%, p > 0.05) and objective response rate (ORR) (9.1% VS 4.2%, p > 0.05) compared with those having fusion gene negative. The median progression free survival (mPFS) were 4.0 and 2.7 months in patients harbored fusion mutations and wild type, respectively (p > 0.05). We conclude that CB samples could be used to detect ALK, ROS1 and RET fusions in NSCLC. The frequency distribution of three fusion genes is higher in lung adenocarcinoma with wild-type EGFR, compared with unselected NSCLC patient population. Patients with fusion genes positive are younger than those with fusion gene negative, but they had no significantly different PFS in first-line chemotherapy.