Objective To compare lymph node sampling (LN-S) and lobe-specific lymph node dissection (L-SLD) in the clinical efficacy and safety for early-stage non-small cell lung cancer (NSCLC). Methods PubMed, Medline, EMbase, Web of Science and The Cochrane Library databases were searched up to March 2017 for English language studies. We collected randomized controlled trials (RCTs) and cohort studies (CS) which used the systematic mediastinal lymph node dissection (SMLD) and LN-S or L-SLD for the treatment of NSCLC. Direct meta-analysis was performed using RevMan 5.3 software and indirect meta-analysis with ITC software after two researchers screened the literature, extracted the data and evaluated the risk of bias independently. Results A total of 18 articles were included (4 RCTs and 14 CS, and 10 714 patients). Meta-analysis results showed that in the CS, compared with the the SMLD group, overall survival increased in the L-SLD group (HR=0.99, 95%CI 0.78 to 1.25, P=0.92), and overall survival decreased in the LN-S group with significant difference in CS (HR=1.43, 95%CI 1.17 to 1.75, P=0.000 4), but was not statistically significant in RCT (P=0.35). In terms of disease-free survival, there was no significant difference between the SMLD group and the LN-S group (HR=1.25, 95%CI 0.90, 1.62, P=0.10) as well as the L-SLD group (HR=1.15, 95%CI 0.92 to 1.43, P=0.23) in the CS. There was no significant difference in the local recurrence rate or distant metastasis rate between the non-systematic lymph node dissection (NSMLD) and SMLD in CS and RCTs (CS: P=0.43, P=0.39; RCT: P=0.43, P=0.10). There was no significant difference in the postoperative complications between NSMLD and SMLD in the CS (OR=0.79, 95%CI0.58 to 1.09, P=0.15) and RCTs (OR=0.36, 95%CI 0.09 to 1.45, P=0.15). Indirect meta-analysis showed that risk of death decreased by 31% and risk of recurrence by 35% in the L-SLD group compared with the LN-S group (HR=0.69, 95% CI 0.51 to 0.95, P=0.46; HR=0.65, 95% CI 0.65 to 1.30, P=0.72), but the difference was not statistically significant. Conclusion For early-stage NSCLC, L-SLD is not statistically different from SMLD in terms of survival; however, the overall survival of LN-S is lower than that of systematic lymphadenectomy. Indirect meta-analysis shows that L-SLD reduces the risk of death and recurrence risk compared with LN-S. There is no evidence to support both direct comparison of the prognosis of LN-S and L-SLD, therefore further prospective studies are still needed to verify.