Introduction: Confirmatory mediastinoscopy after negative endosonography findings is advised by the guidelines on patients with resectable NSCLC and suspected intrathoracic nodes on fludeoxyglucose F 18 positron emission tomography-computed tomography. Its role however is under debate owing to its limited nodal metastasis detection rate, morbidity, associated treatment delay, and unknown impact on survival. Methods: Systematic review and meta-analysis of studies on invasive mediastinal staging in patients with (suspected) NSCLC. The Medline, Embase, and Cochrane databases were searched until September 19, 2018, without year or language restrictions. The Quality Assessment Tool for Diagnostic Accuracy Studies, version 2, was used to evaluate the risk of bias and applicability of the included studies. Rates of unforeseen N2 disease were assessed for endobronchial ultrasound and/or endoscopic ultrasound staging strategies with or without confirmatory mediastinoscopy. Additionally, the complication rates of cervical video mediastinoscopy for mediastinal staging of NSCLC were investigated. Results: A total of 5073 articles were found, of which 42 studies or subgroups (covering a total of 3248 patients undergoing the surgical reference standard of treatment) were considered in the analysis. Random effects metaanalysis of endosonography with or without confirmatory mediastinoscopy showed rates of unforeseen N2 disease of 9.6% (95% confidence interval [CI]: 7.8%-11.7%, I 2 ¼ 30%) versus 9.9% (95% CI: 6.3%-15.2%, I 2 ¼ 73%), respectively. Random effects meta-analysis of mediastinoscopy (eight studies [1245 patients in total]) showed a complication rate of 6.0% (95% CI: 4.8%-7.5%), with laryngeal recurrent nerve palsy accounting for 2.8% (95% CI: 2.0%-4.0%). Conclusion: The rate of unforeseen N2 disease after negative endosonography findings was similar in patients undergoing immediate lung tumor resection to those undergoing confirmatory mediastinoscopy first, at the cost of 6.0% rate of complications by mediastinoscopy.