Heterogeneous reno-protective effects of hypoxia-inducible factor prolyl hydroxylase inhibitors versus erythropoietin-stimulating agents in kidney dysfunction: a target trial emulation with causal forest using real-world data

Erythropoietin-stimulating agents (ESAs) have long been the only drug class available for treating anemia in chronic kidney disease (CKD). Recently, hypoxia-inducible factor prolyl hydroxylase inhibitors (HIF-PHIs) have emerged as a new therapeutic option and have shown efficacy even in patients with hypo-responsiveness to ESAs. While both therapies raise hemoglobin levels, it remains unclear whether their reno-protective effects differ across heterogeneous patient subgroups. We aimed to compare the initiation of HIF-PHIs in comparison with ESAs with respect to progression to kidney replacement therapy (KRT) induction in individuals with varying degrees of kidney dysfunction, using real-world data and advanced causal inference methods.

We conducted a target trial emulation using a population-based administrative claims database (DeSC database) including individuals with CKD (estimated glomerular filtration rate, eGFR 15–60 mL/min/1.73 m²) and anemia (hemoglobin < 12 g/dL) who initiated either a HIF-PHI or an ESA between 2021 and 2024. Patients with prior ESA/HIF-PHI use or KRT were excluded. A generalized causal forest (GCF) algorithm estimated heterogeneous treatment effects on KRT initiation within a 2-year follow-up. Baseline covariates included demographics, kidney function, albuminuria (predicted by dipstick proteinuria), blood pressure, body mass index, comorbidities, concomitant drug use, and baseline hemoglobin. Confounding and censoring were adjusted using inverse probability weighting. The average treatment effect (ATE) of HIF-PHIs in comparison with ESAs was estimated as the difference in KRT-free probability, and heterogeneity of treatment effect was examined across subgroups with the GCF. The intention-to-treat analysis was primary, and per-protocol analyses were secondary.

A total of 4,060 patients (HIF-PHI n = 1,393; ESA n = 2,667) were included. The baseline mean age was 84.0 years, mean eGFR 34.4 mL/min/1.73 m², and mean hemoglobin 10.3 g/dL. During follow-up, 210 patients initiated KRT, corresponding to an incidence of 44.7 per 1,000 person-years (95% confidence interval [CI]: 39.1–51.2). The overall estimated ATE indicated a 2-year KRT-free probability difference of 2.7% (95% CI: 0.2–5.0) in favor of HIF-PHIs compared with ESAs. The GCF revealed that patients with lower baseline eGFR, higher predicted albuminuria, higher systolic blood pressure, and higher body mass index derived greater reno-protective benefit in HIF-PHI initiation, whereas baseline hemoglobin and age did not show such heterogeneity. Results from per-protocol analyses were consistent.

In this real-world target trial emulation, initiation of HIF-PHIs for anemia in CKD demonstrated heterogeneous reno-protective patterns across patient subgroups, particularly among those with lower eGFR and higher albuminuria—subgroups that are generally considered more likely to be resistant to ESA treatment. These findings highlight the potential importance of patient stratification when selecting anemia therapies. Limitations included the older study population and the focus on treatment initiation rather than switching between agents. Future prospective studies are warranted to validate these findings and to inform precision-medicine approaches in nephrology.