PGx423 Pharmacogenetic evaluation of efficacy of lapatinib in EGF104900, EGF20009, EGF105084 and VEG20007

Trial description: Lapatinib combinations are effective therapy in metastatic breast cancer (MBC) where ErbB2 (HER2) is over-expressed in tumors. As is consistent with other tyrosine kinase inhibitors (TKI), patient response is variable, suggesting that there may be factors that influence sensitivity and resistance to TKIs. Host germline genetic variation has been shown to be associated with response to various TKIs used to treat other cancers. Evidence of association between VEGFA genotypes and median overall survival or progression free survival (PFS) has been shown in patients with breast or ovarian cancer treated with bevacizumab in combination with chemotherapy agents (Schneider et al., 2008). The IL8 -251T>A polymorphism was associated with response rate (RR) in patients with ovarian cancer treated with cyclophosphamide and bevacizumab (Schultheis et al., 2008).Two polymorphisms in VEGFR3 were associated with a shorter progression free survival in renal cell carcinoma (RCC) patients treated with sunitinib (Garcia-Donas, et al., 2011). Polymorphisms in IL-8, HIF1A, NR1I2, and VEGFA showed association with response variables in RCC patients treated with pazopanib (Xu et al., 2011). Therefore, an exploratory pharmacogenetic analysis was undertaken to investigate whether germline genetic variants associate with response to lapatinib in three metastatic breast cancer clinical trials including EGF20009, EGF104900 and EGF105084. Originally, the lapatinib monotherapy arm of VEG20007 was thought to be an appropriate clinical study for inclusion in these analyses. However, due to the nature of the disease definition (skin disease) in this study, a decision was made not to include data for these patients in the analysis. This exploratory pharmacogenetic study was a retrospective, non-interventional analysis to investigate association of genetic polymorphisms with response to lapatinib, either in monotherapy or combination treatment. Genetic data was generated as previously described (Spraggs et al., 2011) and included both candidate gene polymorphisms as well as genome wide association screen (GWAS) data. Markers were assigned to tiers for analysis, where tier I polymorphisms (n=7) were those previously associated with expression, alternative signaling or ADME in TKI treatment response, or breast cancer susceptibility; tier II polymorphisms (n=48) were functional variants in 24 candidate genes; tier III polymorphisms (n=1472) were tagging single nucleotide polymorphisms (SNPs) in 24 candidate genes; and tier IV SNPs were the remaining GWAS SNPs not mapped to candidate genes. All associations were evaluated using an additive test. For any significant marker identified from the additive test, specific contrasts of interest between different genotypes were explored to determine risk genotype(s). Cox proportional hazards model was used to assess the association of genotype with time to event outcome (PFS and OS). For candidate gene data, the Firth method was used to control false positives with small sample size. For GWAS data, genomic control adjustment was used to control false positives. Covariates were evaluated for inclusion in the analysis model. Any covariate significant at p≤0.05 was included in the multivariate Cox model for each endpoint/study. Any covariate that violates the proportional hazard assumption is included as an interaction term. Summary statistics (HR) from EGF104900 monotherapy and trastuzumab combination arms, and EGF105084 monotherapy were meta-analyzed using the inverse variance method which allows for combining effect sizes and weighting by their variance. Pre-specified multiple testing adjustments were applied that accounted for number of SNP tests, but not for number of endpoints and groups examined. Patients from EGF20009 had a different ethnic makeup (6% Caucasian) and HER2 status from EGF104900 and EGF105084 (>80% Caucasian). The data was analyzed; however, no results met the pre-specified significance threshold and are not discussed further in this summary.