Pharmacogenomics of serious drug induced toxicity in pediatric oncology

Abstract

Survival of children with cancer has been increasing over the last decades for all malignancies due to improvements in treatment protocols and supportive care. However, this increasing survival has come at the cost of severe lifelong adverse effects of treatment for many survivors. Identifying patients at highest risk for these adverse effects might offer ways to prevent or reduce associated morbidity and mortality. Many factors, including genetics, influence the effects of drugs. Pharmacogenomics aims to identify these genetic variants. Several studies have previously investigated the role of genetic variants in the development of adverse drug reactions (ADRs) for a variety of anticancer drugs, including two important ADRs in pediatric oncology: cisplatin-induced ototoxicity (CIO) and anthracycline-induced cardiotoxicity (ACT), which carry high morbidity and for anthracyclines also mortality. Many of these studies, however, had methodological shortcomings. The studies in this thesis were aimed to overcome these issues and to discover and replicate genetic variants associated with CIO and ACT to provide a basis towards developing pharmacogenetic tests to predict and prevent these severe ADRs. First, it was established that principal component analysis can be used in pharmacogenomic studies to detect and correct for population stratification, a potential confounder in such studies. Next, 2977 single nucleotide polymorphisms (SNPs) in 220 genes involved in drug transport, metabolism and toxicity using a custom-made SNP genotyping panel were examined in a discovery cohort of 54 children treated with cisplatin. Two SNPs in TPMT and COMT were identified to be significantly associated with CIO, with replication in another cohort of 112 children. In a follow-up study of 155 children, the variant in TPMT was further replicated, while an effect in the same direction was seen for COMT, though this was not statistically significant. In addition, a variant in ABCC3 was replicated that was marginally associated in the first study. Combining these variants with clinical risk factors in a single model predicted CIO better than clinical factors alone. Similarly, using the same SNP genotyping panel in a cohort of children (n=156) treated with anthracyclines, one variant in SLC28A3 was found to be highly associated with ACT, with replication in a second cohort (n=188), and further replication in a smaller cohort (n=96). In a follow-up study in an extended cohort of 218 patients this variant was further validated as well as another variant in UGT1A6 and suggestive evidence was found for variants in 3 other genes. Combining all variants with clinical risk factors into a single risk prediction model predicted ACT better than clinical factors alone. This model was further extended with two novel variants in SLC22A17 and SLC22A7 that were identified in a subsequent study. In summary, multiple novel genetic risk factors for CIO and ACT were identified and replicated. Combining these genetic risk factors with clinical risk factors into a single model improved prediction of patients at risk for these toxicities. When further replicated and validated, this could allow for better risk stratification in individual patients to provide the best cancer treatment while minimizing toxicity.