Real-World Generalizability of Clinical Trial Cytomolecular Features By Race and Ethnicity in Pediatric Acute Myeloid Leukemia (AML)

Background:
Black and Hispanic children with acute myeloid leukemia (AML) have consistently demonstrated worse overall survival (OS) compared to non-Hispanic White children. We have previously demonstrated that these children are also less likely to enroll in clinical trials, and that children receiving treatment off-trial experience lower OS. While race/ethnicity are social constructs, some have debated that differential biology may drive some of the survival disparities due to race/ethnicity functioning as a poor proxy for genetic ancestry. There is clinical trial data demonstrating differential racial and ethnic distribution of cytomolecular features which play a critical role in pediatric AML prognostication and risk-stratified treatment determination (Conneely, 2021). However, the generalizability to the real-world setting has not been assessed. We evaluated the distribution of cytomolecular features by race/ethnicity and the contribution of these features to OS disparities among trial-enrolled and non-enrolled patients in a large nationally representative real-world pediatric AML cohort.

Methods:
This interim analysis of a retrospective cohort included all pediatric (age <19 years) patients treated for AML from January 2011 to April 2023 at 10 institutions in the United States. Data were abstracted from the electronic medical record by trained study personnel, including individual cytomolecular reports (FISH, karyotype, targeted sequencing for CEBPA and NPM1). Patients were categorized into 4 cytomolecular risk profiles: 1) only favorable markers; 2) only unfavorable markers; 3) both favorable and unfavorable markers; 4) neutral. Four favorable and 19 unfavorable markers were defined according to the Children’s Oncology Group’s most recent classification in AAML1831. Kaplan-Meier survival curves were generated for OS and event-free survival (EFS) by cytomolecular risk profiles. Adjusted Cox models compared the hazard of death and relapse/refractory disease by cytomolecular markers. The frequency of favorable and unfavorable markers was compared by trial enrollment status and race/ethnicity.

Results:
A total of 654 pediatric patients with AML (54% non-White and non-Hispanic, 39% trial-enrolled) were included in the analytic cohort. Hispanic and non-Hispanic Black (p<0.01) patients were significantly less likely to be enrolled on a clinical trial. Regarding cytomolecular features overall, 27% had only favorable markers, 5% had both favorable and unfavorable markers, 42% had neutral markers, and 26% had only unfavorable markers. Compared to patients with only favorable markers (3-year OS 88.31%; 95% CI: 81.93%-92.54%), patients with both favorable and unfavorable (hazards ratio [HR] = 2.44, p=0.04), neutral (HR=3.64, p<0.001), and only unfavorable (HR=5.47, p<0.001) markers all had increased hazards of death. The hazards for relapse and/or refractory disease were similarly increased (HR=1.92, p=0.03; HR=2.38, p<0.001; HR=3.04, p<0.001, respectively). A test for trend indicated a significant association between cytomolecular risk profile and OS (p<0.001) and EFS (p<0.001). There was no significant difference in cytomolecular risk profiles or individual markers by trial enrollment status. Despite observed trial enrollment disparities, there was also no statistically significant difference in cytomolecular features by race/ethnicity. There was a trend towards Hispanic and non-Hispanic Black patients being more likely to have only favorable cytomolecular compared to non-Hispanic White patients (30.6% and 33.0% vs. 24.5%, p=0.10).

Discussion:
OS and EFS differed by cytomolecular risk profiles. Black and Hispanic patients were less likely to enroll on clinical trials as previously observed. However, distributions of cytomolecular features were comparable for patients not enrolled on a clinical trial and those enrolled, as well as by race/ethnicity. Thus, differential disease biology does not seem to be the major driver of previously observed survival differences by trial enrollment or race/ethnicity. The alternative mechanism is that social drivers of health (i.e. structural racism) independently impact intermediates such as minimal residual disease (MRD) through differential health care access, delays to presentation, and timely treatment. This will be a critical focus of our future work.