Dynamic biomarker trajectories predict outcomes in AML-associated hemophagocytic lymphohistiocytosis

Abstract Introduction: Hemophagocytic lymphohistiocytosis (HLH) in adults with acute myeloid leukemia (AML) represents a devastating hyperinflammatory syndrome with median survival often limited to months (Löfstedt A et al.). Early diagnosis is challenging due to overlapping clinical features with leukemia and limitations of current diagnostic criteria in cancer patients. Moreover, standard diagnostic frameworks (HLH-2004 criteria and HScore) were not specifically validated for adults with hematologic malignancies. The aim of this study was to investigate temporal biomarker changes around HLH diagnosis in patients (pts) with AML so as to identify early predictive signatures for timely intervention and improve outcomes. Methods: We retrospectively analyzed 91 consecutive pts with a diagnosis of leukemia AML and HLH treated between 2010–2024 at MD Anderson Cancer Center utilizing the MD Anderson Context Engine. The cloud-based Syntropy Foundry Platform enabled secure integration of clinical, translational, and research data under institutional review board approval, with informed consent waived. Demographic data, leukemia genetics, therapy details, and results from over 50 laboratory tests were extracted. Laboratory values were standardized, with extreme values captured within 14 days prior to HLH diagnosis. Statistical analysis included longitudinal median trajectories (±15 days from diagnosis), Kaplan-Meier and Cox proportional-hazards modeling for survival predictors, and dichotomization of continuous laboratory variables at their medians for analyses. Data visualization was performed using R ComplexHeatmap and ggplot2, with statistical significance set at p<0.05. Results: Median age was 51 years with 57% male predominance. Based on ELN 2017 criteria, 71% of pts had adverse-risk leukemia; HLH was diagnosed concurrently in 13%. Molecular analysis revealed somatic mutations in TP53 (32%), DNMT3A (27%), TET2 (25%), and RUNX1 (18%). Median overall survival was 1.9 months (95% CI 1.4–2.6). Bone marrow (BM) at baseline showed a median blast of 22% (IQR 5-70). BM around 30 days available in 54 pts with complete response (CR) in 23%, persistent or residual disease in 29%, and residual, hypocellular, or necrotic marrow 7.7%. Pts who achieved CR in BM had markedly improved OS compared to those who did not (p < 0.0001). Multivariate analysis identified four independent predictors of poor overall survival (all p < 0.01): ferritin ≥34,004 ng mL⁻¹, LDH ≥1,339 U L⁻¹, prolonged prothrombin time ≥20 seconds, and hypoalbuminemia <2.9 g dL⁻¹. Longitudinal biomarker analysis revealed distinctive patterns preceding clinical diagnosis, with median ferritin and LDH levels beginning to rise approximately 7 days before HLH diagnosis and crossing high-risk thresholds by day -2. This trend was paralleled by elevations in liver enzymes (AST and ALT) showing similar trajectories. While Optimized Hematologic Inflammatory Index (OHI) calculation using ferritin and soluble interleukin 2 receptor (sIL2r) was attempted, limited availability of sIL2r testing precluded meaningful trend analysis. Leukemia-directed chemotherapy was administered to 75% of the patients. Treatment patterns showed that non-survivors received more intensive chemotherapy, etoposide, and corticosteroids in the pre- and early post-HLH period, while anakinra administration was associated with trends in improved survival. Conclusions: In our cohort of pts with concurrent high-risk leukemia and HLH, median survival was just 1.9 months. Ferritin, LDH, prothrombin time, and albumin emerged as independent biomarkers predicting poor survival. Longitudinal biomarker analysis identified a critical 5-day pre-diagnostic window where ferritin and LDH show accelerated elevation before reaching high risk levels. Furthermore, the molecular landscape, dominated by mutations in TP53, DNMT3A, TET2, and RUNX1, commonly associated with clonal hematopoiesis of indeterminate potential (CHIP), suggests a possible mechanistic link between pre-leukemic clonal evolution and HLH susceptibility. Monitoring biomarker trajectories, potentially triggering diagnostic workup or preemptive intervention based on the rate of rise or crossing moderate thresholds may offer a window of opportunity to improve survival by intervening before irreversible organ damage occurs. Updated analysis will be presented at the upcoming ASH meeting in Orlando.