Impact of Social Determinants of Health on Overall Survival in Patients with AML By European Leukemianet Risk Group

Introduction

The European LeukemiaNet 2022 (ELN) criteria are a validated genetic risk classification for patients with newly diagnosed acute myeloid leukemia (AML) treated with intensive induction chemotherapy. While previous studies demonstrated that social determinants of health (SDOH; e.g., education, income) are associated with overall survival (OS) in AML, their impact within ELN risk groups remains unclear. Hence, we evaluated the impact of SDOH on OS by ELN risk group.

Methods

We extracted patient, disease, and treatment characteristics from electronic medical records for all patients with newly diagnosed AML treated with intensive chemotherapy (cytarabine plus daunorubicin or CPX-351) between 2015-2022 at Dana-Farber Cancer Institute. The national Area Deprivation Index (nADI), a place-based, composite measure of socioeconomic disadvantage built from 17 census variables at the census block group level, was used to assess SDOH. nADI is an ordinal measure from 1-100 with 100 indicating the highest level of disadvantage. For United States (US) residents nADI was analyzed as a continuous or categorical variable grouped as binary (low [<=50] vs. high [>50]) or by tertiles (T1 [1-33], T2 [34-66], T3 [67-100]). Analysis was performed with the primary outcome of overall survival (OS) determined by the Kaplan-Meier method and between-group comparisons by the log-rank test; and secondary outcome as allogeneic hematopoietic stem cell transplantation (HCT) rate. Cox regression models were fitted to evaluate SDOH in relation to OS within each ELN risk group. We conducted a sensitivity analysis censoring follow-up at HCT.

Results

Overall, 365 of 434 patients were included (3% were missing ELN data; 13% could not be linked to nADI score) and classified as favorable (n=115, 32%), intermediate (n=84, 23%) and adverse (n=166, 45%) ELN risk. The median age was 60 years (interquartile range [IQR] 50-66) and median nADI was 25 (IQR 14-39), which did not differ significantly by ELN risk group (p=0.11). Most patients (n=327, 90%) were non-Hispanic (NH) White, followed by NH-Black (n=12, 3.3%), Hispanic (n=12, 3.3%), NH-Asian (n=8, 2.2%), and others (n=6, 1.6%). Most patients were in the more advantaged nADI groups when evaluated as binary (low: n=323 [88%]; high n=42 [12%]) or by tertiles (T1: n=235 [64%]; T2: n=115 [32%]; T3: n=15 [4%]).

ELN risk was associated with OS (log-rank p<0.0001), with median OS for favorable: 57 months (95% confidence interval [CI] 49-not reached [NR]); intermediate risk: 28 months (95% CI 21-NR); and adverse risk: 18 months (95% CI 11-26). Among the favorable risk group, OS was similar by binary (high vs. low hazard ratio [HR]: 1.03 [95% CI 0.36-2.92], p=0.96) and tertile-based nADI (HR T2 vs T1 1.40 [95% CI 0.71-2.76], HR T3 vs T1: 1.00 [95% CI 0.13-7.48], p=0.64). Similarly, OS did not differ by nADI among the adverse risk group (binary HR: 0.79 [95% CI 0.40-1.57], p=0.50; tertile HR: T2 vs T1 0.69 [95% CI 0.43-1.11], T3 vs T1: 1.09 [95% CI 0.44-2.98], p=0.27). In both ELN groups, HCT rates did not differ between different nADI categories.

In contrast, among intermediate risk patients, OS was higher for more advantaged groups (low vs. high nADI median OS 33 vs 17 months, p=0.035), with a HR for the high nADI group of 2.23 (95% CI 1.04-4.76; p=0.039). This difference remained significant after adjustment for age, sex, and race (adjusted HR 2.98; 95% CI 1.24, 7.17; p=0.02), and was redemonstrated when nADI was categorized by tertiles (T1: median OS not reached, T2: median OS 24 months, T3 median: OS 7 months, p=0.007). The HCT rates were similar by binary (71% vs. 67%, p=0.74) and tertile-based nADI (T1: 67%; T2: 79%; T3: 50%, p=0.29). When censored at HCT, there was no significant difference in OS by nADI when categorized as binary (p=0.19), or by tertiles (p=0.59).

Conclusion

In a relatively affluent and NH-white cohort with newly diagnosed AML, stratification by nADI was significantly associated with OS for ELN intermediate risk patients but not for those with favorable or adverse risk disease. The prognostic value of nADI for intermediate risk patients was mitigated when censoring at transplant, although realization rates of HCT were similar. Together, this suggests that the powerful effects of favorable and adverse risk disease biology may overcome the influence of SDOH on survival, and that the influence of SDOH may be most impactful for intermediate risk patients in the peri/post-HCT period.