Reassessing voxelotor safety and efficacy: Real-world outcomes from the ASH research collaborative data hub

Abstract Voxelotor binds the α-globin subunit of HbS, stabilizing it in the oxygenated state, which may reduce polymerization. Voxelotor was approved in 2019 as a disease-modifying therapy (DMT) for sickle cell disease (SCD) based on a modest increase in Hb (median 1.1 g/dL) and improved hemolysis markers, but no benefit on clinical outcomes. In September 2024, the manufacturer voluntarily withdrew voxelotor, citing in a press release an “imbalance in vasoocclusive events and fatal outcomes” that no longer justified its continued use. However, no detailed safety or efficacy data were disclosed, and independent real-world analyses are lacking. We evaluated the real-world effectiveness and safety of voxelotor before its withdrawal in the ASH Research Collaborative (RC) Data Hub. The ASH RC Data Hub aggregates electronic health records (EHR) from 14 US SCD center. Patients with PI-attested diagnoses from 2015-2024 were included. DMT use was defined as EHR-reported treatment with hydroxyurea, voxelotor, or L-glutamine for ≥90 days/year. Transfusions were categorized as DMT if ≥6 occurred/ year. Crizanlizumab was excluded due to insufficient data. Clinical and lab data were calculated per patient-year to account for follow up duration variability and were compared pre- and on-treatment. Hb data were included if ≥1 year pre- and on-treatment were available. EHR mortality data were used, though their accuracy needs further validation. For longitudinal analysis, random effects repeated measures models were used, reported as least-square means ±SE. Among 10,412 patients with confirmed SCD, 7,033 (68%) had HbSS, 2,602 (25%) HbSC, and 777 (7%) other genotypes. Among 388 patients (3.7%) with identified voxelotor prescriptions, 97% had HbSS. Compared to other DMTs, patients on voxelotor were older (24±16 vs 20.1±15 years, P<0.001), more often female (63% vs 52%, P<0.001), had lower Hb (7.5±1.3 vs 8.5±1.5 g/dL, P<0.001), higher reticulocyte counts (9.7±5% vs 8.2±4.8%, P<0.001), more VOE/ACS (0.2±0.6 vs 0.1±0.3 /patient-year, P<0.001), more ED/hospitalizations (2.3±9.3 vs1.9±6.2 /patient-year, P=0.002), and higher CKD prevalence (18.2% vs 8.2%, P<0.001). In a longitudinal analysis (721 patient-years pre-treatment vs. 614 on-treatment), Hb increased on voxelotor from 7.36±0.06 to 7.51±0.06 g/dL (mean difference 0.15±0.04, P=0.001), reticulocyte count decreased from 10.2±0.3% to 9.6±0.3% (P=0.01), and Hb increased by ≥1 g/dL in 44 patients (11.3%). No significant changes were observed on voxelotor in VOE/ACS rates (P=0.15), ED /hospitalizations (P=0.46), or hospital length of stay (P=0.09). Eight voxelotor-treated patients (2.1%) died, compared to 125 (3.8%) on other DMTs (P=0.07). In univariate analyses, Hb varied by age group (≥ vs <18 yo, P=0.01), genotype (P=0.004), and CKD status (P=0.014), but not hydroxyurea co-therapy (P=0.1). In multivariate modeling including age, genotype, and CKD, only genotype significantly interacted with Hb change on voxelotor (HbSS +0.2, non-HbSS -0.4, mean difference 0.6±0.27 g/dL, P=0.03). Hb change in non-HbSS patients was not significant (P=0.11). In this real-world cohort, voxelotor use was associated with a very small but statistically significant increase in Hb (+0.15 g/dL) and decrease in reticulocyte count (-0.6%). The Hb increase was higher among patients with HbSS genotype. It remains unclear whether these small changes are clinically meaningful or compensate for the potential reduction in O2delivery from voxelotor's allosteric modulation of Hb. Notably, despite older age and more severe SCD in the voxelotor cohort, rates of VOE, ACS, acute care utilization, or crude mortality were not increased during treatment, but also not reduced. While this cohort was not designed to capture rare events, no clear signal for harm was identified in 614 patient-years of exposure, though benefit for most patients remains limited. Continued analysis of post-withdrawal trends in the Data Hub will reveal any potential changes in outcomes. Differences between these real-world data and clinical trial results likely reflect the controlled rigor of trial protocols versus the variability of broader clinical application across diverse populations and should be viewed as complementary. In conclusion, the ASH RC Data Hub is a valuable resource for post-marketing pharmacovigilance studies. Independent analyses can complement industry data, providing additional value through increased transparency and more comprehensive dataset.