ObjectiveTo systematically evaluate the effectiveness of survival prediction models for lung cancer patients and analyze factors influencing model performance. MethodsRelevant literature was retrieved from PubMed, EMbase, China National Knowledge Infrastructure (CNKI), and Wanfang Data up to March 2025 using computerized searches. The risk of bias and applicability assessment tool for prognostic studies was applied to evaluate methodological quality and applicability of included studies. Meta-analysis was conducted using R software (version 4.4.3). ResultsA total of 11 studies involving 23134 patients published between 2017 and 2025 were included. Meta-analysis demonstrated a pooled C-index of 0.72 [95%CI (0.70, 0.74)] for lung cancer survival prediction models. Subgroup analysis revealed that studies with sample size >1,000 cases showed greater stability in model performance with a C-index of 0.72 [95%CI (0.71, 0.72)]; non-small cell lung cancer (NSCLC) models exhibited a C-index of 0.71 [95%CI (0.69, 0.73)] compared to small cell lung cancer (SCLC) models at 0.70 [95%CI (0.64, 0.76)]. Studies with survival rate <50% had a C-index of 0.71 [95%CI (0.69, 0.73)] while those with survival rate ≥50% showed 0.73 [95%CI (0.70, 0.75)], with no statistically significant difference between groups (P=0.2601). Tumor staging, surgical intervention, and chemotherapy were identified as key prognostic predictors. Risk of bias assessment indicated 7 studies with high or unclear risk of bias, 3 with low risk, though overall applicability remained satisfactory (only 1 study had unclear applicability). ConclusionLung cancer survival prediction models demonstrate good prognostic discrimination and calibration capabilities, but significant heterogeneity exists across models. Large sample size (>1000 patients) is crucial for reducing heterogeneity and improving stability. NSCLC models show slightly better predictive performance than SCLC models, with higher discriminative ability observed in populations with higher survival rates. Future research should focus on optimizing model design, minimizing bias risks, and enhancing both predictive accuracy and clinical utility.