22049 Background: Somatic EGFR mutations have been evaluated in numerous studies as potential predictive factors of response to tyrosine kinase inhibitors, such as gefitinib and erlotinib, for patients with non-small cell lung cancer (NSCLC). Methods: We used information from a comprehensive free-access online analytical database of somatic EGFR mutations in NSCLC (www.egfr-mutations.org) to identify studies reporting on the association between mutations and response to EGFR tyrosine kinase inhibitors. Studies were eligible if they reported on the treatment of unselected patients with advanced NSCLC. Between-study heterogeneity was assessed with the I2 statistic; small study effects were assessed using Begg's and Egger's tests. In the absence of large between-study inconsistency (I2 >50%), RRs were pooled using the Mantel-Haenszel method. Results: Out of a total 202 studies reporting on somatic EGFR mutations in NSCLC patients, 30 studies (1,437 patients) were considered eligible for this analysis; 28 (1,365 patients) assessed the association of mutational status with response to gefitinib and 2 (72 patients) with response to erlotinib. Overall, 482 somatic mutations were detected (33.5%). There was no evidence of large between-study inconsistency (I2 = 30.5%), but there was strong evidence of differential small study effects (p- values for Begg's and Egger's tests <0.01). The presence of any EGFR somatic mutation was highly correlated with response, RR= 6.4 (5.17–7.91). Patient ethnicity did not alter this association: non-Asians, RR= 7.92 (5.31–11.82); Asians, RR = 6.00 (4.68–7.70); interaction p-value = 0.25. Mutational status was a predictor of response both to gefitinib, RR = 6.38 (5.14–7.91) and erlotinib, RR= 7.29 (2.12–25.09); interaction p-value = 0.84. Conclusions: Our analysis provides large-scale empirical evidence that EGFR mutations are predictive of response to gefitinib and erlotinib for patients with advanced NSCLC. The effect of the mutations appears to be independent of ethnicity. A meta-analysis of patient-level data would allow more thorough evaluation of the interaction of clinico-pathological characteristics, mutational status and treatment efficacy. No significant financial relationships to disclose.
Read full abstract