Background Pulmonary metastasectomy is widely accepted for different malignant diseases. The role of mediastinal lymph node (LN) dissection in these procedures is discussed controversially. We evaluated our results of LN removal at the time of pulmonary metastasectomy with respect to the frequency of unexpected LN disease. Methods This was a retrospective analysis of 313 resections performed in 209 patients. Operations were performed in curative intention. Patients with known thoracic LN involvement and those without lymphadenectomy (n = 43) were excluded. Patients were analyzed according the type of LN dissection. Subgroups of different primary cancers were evaluated separately. Results Sublobar resections were performed in 256 procedures with lymphadenectomy, and 14 patients underwent lobectomy. Patients underwent radical lymphadenectomy (n = 158) or LN sampling (n = 112). The overall incidence of unexpected tumor in LN was 17% (radical lymphadenectomy, 15.8%; sampling, 18.8%). Unexpected LN involvement was found in 17 patients (35.5%) with breast cancer, in 120 (9.2%) with colorectal cancer, and in 53 (20.8%) with renal cell carcinoma. The 5-year survival was 30.2% if LN were tumor negative and 25% if positive ( p  = 0.19). LN sampling vs radical removal had no significant effect on 5-year survival (23.6% vs 30.9%; p  = 0.29). Conclusions Dissection of mediastinal LN in resection of lung metastases will reveal unexpected LN involvement in a relevant proportion of patients, in particular in breast and renal cancer. Routine LN dissection appears necessary and may become important for further therapeutic decisions. On the basis of our data, LN sampling seems to be sufficient.