ObjectivesBevacizumab and erlotinib inhibit different tumour growth pathways, and both exhibit beneficial effects in the treatment of non-small-cell lung cancer (NSCLC). However, the efficacy of bevacizumab in combination with erlotinib remains controversial. Therefore, we conducted a meta-analysis to compare combination treatment with bevacizumab and erlotinib to bevacizumab or erlotinib monotherapy in the treatment of NSCLC.MethodsRandomised controlled trials (RCTs) published in PubMed, Web of Science and EMBASE were systematically reviewed. The main outcome measures included overall survival (OS), progression-free survival (PFS), overall response rate (ORR) and adverse events. Results were expressed as HRs or risk ratios (RRs) with 95% CIs.Results5 RCTs involving a total of 1736 patients were included in this meta-analysis. The combination of bevacizumab and erlotinib significantly improved PFS (HR=0.63, 95% CI 0.53 to 0.75; p=0.000) and the ORR (RR=1.91, 95% CI 1.19 to 3.06; p=0.007) in the second-line treatment of NSCLC compared with bevacizumab or erlotinib alone. However, no significant difference in OS was observed between the combination and monotherapy groups (HR=0.96, 95% CI 0.83 to 1.11; p=0.573). A subgroup analysis has shown that the greatest PFS benefit was associated with an age of <65 years(HR=0.74, 95% CI 0.57 to 0.96; p=0.026), Asian/Pacific Islander ethnicity (HR=0.23, 95% CI 0.10 to 0.54; p=0.001), Eastern Cooperative Oncology Group performance status (ECOG PS) 1 (HR=0.82, 95% CI 0.68 to 0.98; p=0.033), stage IIIB or IV disease (HR=0.68, 95% CI 0.57 to 0.82; p=0.000) and no history of smoking (HR=0.48, 95% CI 0.32 to 0.71; p=0.000). The incidence of grade 3/4 adverse events such as rash and diarrhoea was higher in the combination group than in the monotherapy group.ConclusionsThe addition of bevacizumab to erlotinib can significantly improve PFS and the ORR in the second-line treatment of NSCLC with an acceptable and manageable risk of rash and diarrhoea. Further well-conducted, large-scale trials are needed to validate these findings.