Genetically Defined Racial Differences Underlie Risk of Relapse in Childhood Acute Lymphoblastic Leukemia

The effect of race or ethnicity on cure and survival outcomes among children with acute lymphoblastic leukemia (ALL) is controversial. For those studies that have demonstrated variability in survival associated with race or ethnicity, the cause of these differences remains unclear. We therefore used genome-wide germline single nucleotide polymorphism (SNP) genotypes to quantitatively estimate racial ancestral composition in children with ALL and examined associations between ancestries and the probability of ALL relapse. To infer population genomic structures, we applied principal components analysis (PCA) to genotypes at 219,955 germline SNPs in 893 individuals: 683 patients with ALL (450 from St. Jude Children’s Research Hospital [St. Jude] Total XIIIB and Total XV studies and 233 from Children’s Oncology Group [COG] P9906 protocol), and 210 HapMap samples (60 CEU, 60 YRI, 90 CHB/JPT, serving as references for white, black, and Asian races, respectively). The top ranked principal component (PC1) separated self-reported black patients (n=92) and the YRI HapMap group from all other racial/ethnic groups; PC2 separated self-reported Asian patients (n=19) and the CHB/JPT HapMap samples from non-Asian populations. PC3, on the other hand, primarily captured genetic variation characteristic of American Indian ancestry (assessed using publicly available data from American Indians, n=105). Interestingly, Hispanic patients with ALL (n=75) showed a continuous cline between the American Indians and whites, displaying a gradient of these two ancestries among Hispanics. The relationships between ancestries (PC1, PC2, and PC3) and outcome were assessed in the St. Jude cohort first and validated in the COG cohort, using Fine and Gray’s regression test and after stratification for risk-adapted treatment. Of the top three PCs, only PC3 exhibited a significant association with cumulative incidence of relapse in St. Jude (P=0.038), with higher PC3 values linked to higher risk for relapse. This relationship between PC3 and relapse was validated in the COG cohort (P=0.003), which included more Hispanic patients than the St. Jude cohort. When the St. Jude and COG cohorts were combined, only the American Indian/Hispanic-informative PC3 (not PC1 or PC2) was related to relapse (P=4.5×10⁻⁴). Further, PC3 remained significant even after accounting for self-reported race/ethnicity (P=0.044), or when the analysis was restricted to self-reported white patients (P=0.006). The proportions of patients with high risk clinical or biological features (i.e. high leukocyte count, unfavorable age, unfavorable genetic subtypes, and minimal residual disease at the end of induction therapy) did not differ between the high and low PC3 groups (PC3 > or < 0.005). Therefore, the higher relapse rates in patients with higher proportions of American Indian ancestry are not derived from overrepresentation of unfavorable prognostic features in this group. In a multivariate analysis, PC3 remained significantly associated with the risk of relapse (P=0.041) after adjusting for known risk factors, indicating a possible independent prognostic value of PC3. In conclusion, germline genetic variation that is related to American Indian ancestry is associated with increased risk of leukemia relapse, providing evidence for a genetic basis for racial/ethnic differences in cancer treatment outcome.