Pharmacogenetic variants in the molecular characterization initiative: a report from the children's oncology group.

Germline variants in 13 pharmacogenes define diplotypes, built from star alleles, that map onto actionable metabolizer phenotypes requiring treatment modification; HLA alleles were inferred separately.

The authors analysed germline exome sequencing data from 3,177 patients enrolled in the Children's Oncology Group Molecular Characterization Initiative, 69% of whom had a central nervous system tumour. Pharmacogenomic diplotypes for 13 genes deemed clinically actionable under CPIC guidelines were extracted with PharmCAT, HLA alleles were inferred with the T1K method, and genetic ancestry was estimated with iAdmix using 1000 Genomes as reference. The cohort was ancestrally diverse: 51% European, 21% Admixed American and 9% African. Ninety-three percent of patients (n = 2,940) carried at least one actionable pharmacogenomic phenotype requiring modification of one or more medications, with no significant difference across ancestral populations (p = 0.07). Observed allele frequencies closely matched those of gnomAD and All of Us (R2 > 0.99).

Synthesis written by Geno'X. For the full original abstract, please refer to the source publication.

The 93% figure mainly shows that almost every patient carries an actionable phenotype once 13 pharmacogenes are interrogated: it does not say how many children were actually prescribed the relevant drug, nor how many prescriptions would have changed. The close match with gnomAD and All of Us confirms there is nothing distinctive about this population and instead strengthens the logistical point — the germline exome is already sequenced, so the issue is pharmacogenomic annotation and its return to the prescriber, not an extra test. What remains to be shown, through follow-up of prescriptions and toxicities, is that returning these results actually changes decisions and adverse events.

Analysis by Dr Thibaut Benquey