Background: Correlations between overall survival (OS) and objective response rate (ORR) or progression-free survival (PFS) are poor. We aimed to evaluate the impact of crossover and unbalanced subsequent treatments on ORR and PFS as surrogate endpoints for OS in patients with advanced NSCLC receiving second- and further-line therapy. Methods: A systematic literature review of randomized clinical trials of systemic treatment for patients with stage IIIB/IV NSCLC receiving second- and further-line therapy was performed. Weighted (by trial size) linear regression models were fitted with the absolute difference in ORR or median PFS as an independent variable and the absolute difference in median OS as a dependent variable. The analysis was repeated in predefined subsets based on crossover and balance of post-progression therapies. Surrogate threshold effect (STE) was estimated using prediction intervals. Results: 146 trials (43,061 patients) fulfilled the eligibility criteria. In all treatment arms, the mean ORR, median PFS, and median OS were 12.2% (standard deviation [SD] 11.2%), 3.2 months (SD 1.3), and 9.6 months (SD 4.1), respectively. ORR and PFS had very weak (R = 0.181; 95% CI: 0.016-0.337) and weak (R = 0.254; 95% CI: 0.074-0.418) relationships, respectively, with OS. However, within trials that did not allow crossover and reported balanced post-progression treatments, ORR and PFS had moderate (R = 0.528; 95% CI: 0.081-0.798) and strong (R = 0.778; 95% CI: 0.475-0.916) relationships, respectively, with OS. STE estimation indicated that trials that show statistically significant treatment effect size of ≥41.0% ORR or ≥4.15 median PFS months can be expected to show significant OS benefit with sufficient certainty. Conclusions: Surrogacy of ORR and PFS for OS might be better estimated in trials that do not allow crossover and report balanced post-progression treatments. Presented STE calculation can be used to estimate the expected effect on OS when either ORR or PFS are used as primary endpoints.