Comment on Rocha et al, page 4752
In this issue of Blood, Rocha and colleagues provide evidence for the independent prognostic significance of pharmacogenomic parameters on clinical outcome in childhood acute lymphoblastic leukemia.
Current treatment for pediatric acute lymphoblastic leukemia (ALL) results in cure in approximately 80% of patients, but some children still suffer from treatment failure or treatment-related mortality. Apart from leukemia cell drug resistance,1 it is increasingly recognized that treatment may fail due to suboptimal dosing related to inherited variations in drug-metabolizing genes, addressed in the field of pharmacogenomics. This field may concern enzymes that result in either activation or detoxification of drugs. The available data sometimes provide conflicting results regarding the influence of pharmacogenetics on outcome, a problem that may be related to the retrospective nature of most studies.2
In this issue of Blood, Rocha and colleagues describe a prospective study in which 16 genetic polymorphisms were screened in children treated according to a single protocol, although patients were stratified in high-risk and low-risk arms. It appeared that the glutathione S-transferase (GST) GSTM1 non-null genotype was significantly associated with an increased relapse rate in high-risk patients and was further increased by the thymidylate synthetase (TYMS) 3/3 genotype. Both genotypes did indeed result in higher expression levels of the enzymes for which they encode. In multivariate analysis these 2 genotypes showed independent prognostic significance—together with the presence of MLL-AF4 and minimal residual disease (MRD) levels. No genotype was identified associated with relapse in the low-risk group. The authors also addressed pharmacogenetic risk factors for central nervous system (CNS) relapse, and showed that 2 polymorphisms in the vitamin D receptor were associated with CNS relapse in high-risk patients. This relationship is probably indirect, as the vitamin D receptor is linked to the expression of CYP3A4 and possibly also p-glycoprotein. In low-risk patients the TYMS 3/3 genotype predisposed for CNS relapse.
This study clearly demonstrates the proof of principle that suboptimal dosing based on genetic variation in the host plays a role in treatment outcome in pediatric ALL. However, a number of unresolved issues remain. First, researchers must determine whether the influence of genotypes on treatment outcome is protocol specific, as suggested by the lack of prognostic significance of GST genotypes in another large prospective study.3 No relevant pharmacogenomic risk factors were identified in low-risk patients, raising the question of whether pharmacogenomics is more relevant in the context of intensive chemotherapy or whether the relevant genes are in fact unknown,1 which may also explain why many of the polymorphisms did not correlate with outcome. For some of the genotypes, the absence of correlation is rather surprising; the thiopurine methyltransferase genotype, which was recently shown to be related to MRD levels in another prospective study, is one such example.4 Furthermore, no relationships were found with treatment-related mortality.
Despite these limitations, the data do provide an opportunity to further improve outcome in pediatric ALL. From a practical point of view this may be easier to achieve for the TYMS 3/3 genotype than for non-null GSTM1 patients, as the latter genotype is involved in detoxification of many different anti-leukemic agents. The St. Jude's group has already shown that individualized dosing of methotrexate improves outcome in pediatric ALL.5 ▪
