Lung cancer is an aggressive disease with high mortality rate worldwide. Surgical intervention is the most effective treatment; therefore, prompt diagnosis is fundamental to cure this neoplasm effectively. Low-dose computed tomography (LDCT)-based screening programs were proven a valid approach to further increase patients survival rate, but false positives may occur and controversies still exist in the management of subjects with indeterminate or premalignant nodules. Furthermore, some patients may experience disease recurrence and conventional prognostic factors are not able to pinpoint patients at high risk accurately. In this context, the non-invasive measurement of circulating biomarkers represents a desirable opportunity. Here, we report the results of our pilot study on circulating tumor cell (CTC) detection in operable non-small cell cancer (NSCLC) patients. Blood samples, collected from healthy volunteers (N=10), nodule-negative high-risk individuals (N=7) at first LDCT, and NSCLC patients (N=74) before surgery, were enriched for CTC using a size-based approach (Isolation by Size of Epithelial/Tumor cells, ISET®). Cytological samples obtained on porous membranes were stained with May-Grünwald Giemsa (4 membrane spots, corresponding to 4 mL of blood) and analyzed by a referral cytopathologist, without knowledge of the disease status and outcome. Cells with features of malignancy were identified according to the classical morphological criteria used for cytological samples. Spike-in experiments with commercially available lung cancer cell lines helped to optimize the workflow and to verify that the formation of artifacts during the filtration procedure did not occur. Overall the CTC detection rate in patients was 60% and did not show significant differences between patients with early (I-II, N=52) and advanced stage (III-IV, N=22). Cells with features of malignancy were not detected in lung cancer-free donors. In addition to the classical CTC presenting as physically isolated events (single CTC, sCTC) within the size-based enriched fraction of cells, we identified a subpopulation of clusters of CTC and leukocytes (hetCLU), mainly monocytes (55.3%), neutrophil granulocytes (10.5%) and lymphocytes (10.5%), detected with an overall frequency of 31%, without substantial differences between the two cohorts. The status of both CTC subsets (overall median (IQR) number of CTC/4 mL in CTC+ve cases: 2 (1-3)) did not correlate with the patients’ clinico-pathological features. Interestingly, the prevalence of sCTC and the presence of hetCLU predicted the risk of disease recurrence (median (IQR) follow-up time: 28 (11.8-33.5) months) within the cohort of early stage (HR 95%CI: 5.15 (1.10-24.33), p-value=0.0009, for cases with ≥2 sCTC, and 3.99 (0.47-33.57), p-value=0.0216, for cases with ≥3 sCTC) or advanced stage (HR 95%CI: 3.44 (0.76-15.50), p-value=0.0129) tumors, respectively, while neither the grading nor the lymph-node status were able to predict the risk of recurrence. The CTC detection rate obtained in operable NSCLC patients by the ISET® technique was encouraging if considering the small amount of blood analyzed compared to other works, thus laying the foundation for further trials in the diagnostic setting. Interestingly, the prevalence of two distinct subpopulations of CTC informed postoperative prognosis. Based on our observations, the biological and clinical significance of both sCTC and hetCLU in non-metastatic NSCLC would deserve further investigation.