GWAS-guided pharmacogenomic risk scores predict premedication-related pegaspargase hypersensitivity and toxicities in pediatric and adolescent young adults with acute lymphoblastic leukemia or lymphoma

Abstract Background: The addition of asparaginase to chemotherapy regimens has played a key role in significantly improving outcomes in pediatric and adolescent/young adult (AYA) patients with acute lymphoblastic leukemia (ALL), with >90% overall survival (OS). The use of pegaspargase (PEG), the PEGylated form of asparaginase commonly used in such regimens, is mainly limited by hypersensitivity reactions and associated toxicities. This frequently leads to switching to an alternate form of asparaginase isolated from Erwinia chrysanthemi (ERW). While response to ERW use after hypersensitivity is equivalent to PEG, limitations such as higher cost, greater dose requirement due to pharmacokinetic differences, pain when administered intramuscularly, and higher nausea make this substitution less convenient. A prospective multi-institutional study initiated by Rady Children's Health/Children's Hospital of Orange County (CHOC) evaluated the efficacy of pre-medicating pediatric and AYA patients with B or T-ALL or lymphoblastic lymphoma (Lly) with histamine-1 (H1) and histamine-2 (H2) blockers before PEG administration in preventing hypersensitivity reactions. Here, we describe a genome-wide association study (GWAS) leading to the development of pharmacogenomic risk scores for PEG-related hypersensitivity and Grade ≥ 3 toxicities in this cohort. Patients and Methods: GWAS was performed using genomic and clinical data from a cohort of 87 pediatric or AYA patients with ALL/Lly. Median age was 8.3 years, 65.5% were male, and 55.2% identified as Hispanic. Overall, 18 (20.7%) patients had a grade >2 PEG hypersensitivity reaction and 44 (50.6%) had an incidence of CTCAE, V5 grade ≥3 toxicities (hyperbilirubinemia, hyper/hypoglycemia, hypertriglyceridemia, pancreatitis, or thrombosis) within ≤30 days of PEG infusion. Infinium GSA-24 v3.0 was used to genotype 650k SNPs and post standard QC, 233,798 SNPs were tested for association with PEG hypersensitivity and toxicities by logistic regression analysis using PLINK. All SNPs significantly associated with each endpoint were used in multi-SNP predictor models testing all possible 2-SNP combinations across additive, dominant or recessive modes of inheritance. 1000 permutations/model were performed to determine likelihood of obtaining them falsely, and models with low Bayesian Information Criterion (BIC; <90) and a 1000 permutation p<0.05 were selected as significantly associated with the outcome. Genotype scores for unique SNPs within top models were assigned based on modes of inheritance and direction of their association (negative for favorable and positive for detrimental) and composite pharmacogenomic risk scores were calculated as the sum of genotype scores. Results: In our GWAS, 77 SNPs associated with the incidence of grade >2 PEG hypersensitivity (p < 0.002) and 81 SNPs associated with grade ≥3 toxicities (p < 0.001). Eight SNPs from 5 2-SNP combinations associated with hypersensitivity and were used to develop the composite PEG Hypersensitivity Risk Score. Similarly, 11 unique SNPs from six 2-SNP combinations for toxicity were used to develop a composite PEG Toxicity Risk Score. Both scores were stratified into low and high score groups based on the median. The distribution of these groups did not differ by sex, ethnicity (Hispanic or non-Hispanic), or NCI risk groups (standard or high). However, all patients with a hypersensitivity reaction were within high Hypersensitivity Risk Score group, while significantly higher incidences of grade >3 toxicities were observed in those within high Toxicity Risk Score group compared to those in low toxicity risk score group (toxicities incidence: 85% v 14% in high score or low score, respectively; p = <0.0001). Conclusion: Our results describe the GWAS-guided development of polygenic risk scores predicting PEG hypersensitivity and toxicity incidences. This approach eliminates certain biases associated with a candidate selection approach to build pharmacogenomic risk scores. Validation of these scores in a similar cohort as well as functional validations will be performed in the near future. In conclusion, we provide promising pharmacogenomic SNP models which could allow clinically relevant pre-medication or asparaginase format choice to minimize hypersensitivity and toxicities in pediatric or AYA patients with ALL/Lly.Funding for this study was provided by CHOC PSF Tithe Grant, CHOC Hyundai Cancer Institute Research Grant, and Servier Pharmaceuticals.