ESC Congress 2026 Abstract · Oct 26, 2026

Artificial intelligence-assisted personalised dose titration for hypertension: safety and feasibility results of the CURATE.AI pilot randomised controlled trial

By Wang L, Tadeo X, Yeung W et al.

Background: Hypertension is the leading modifiable cardiovascular risk factor for global mortality and disability1. Despite effective pharmacological therapies, fewer than 1 in 4 patients achieve blood pressure (BP) targets2. This discrepancy may be partly attributed to the limitations of standardised but non-personalised dosing strategies3. Digital health and artificial intelligence (AI) platforms such as CURATE.AI offer an opportunity to move beyond limitations of population-based, "one-size-fits-all" dosing by using the individual’s unique dose-response profile (N-of-1) to dynamically optimise BP control4. Purpose: To evaluate the safety and clinical feasibility of integrating CURATE.AI, an individualised, AI-assisted clinical decision support system, for antihypertensive dose titration in routine outpatient care. Methods: In this open-label, randomised trial, 45 participants with primary hypertension were randomised (1:1:1) into control (standard care; CT), home BP monitoring (HBPM), or HBPM with AI-assisted titration (CURATE.AI) arm. CURATE.AI generated individualised quadratic dose response profiles based on dose-BP pairs, guiding physician dose recommendations. Primary endpoints assessed logistical and scientific feasibility with prespecified “traffic light” performance targets5. Exploratory secondary endpoints included mean change in office BP and time to target BP. Results: 42 participants were included in the final analysis (CT n=12, HBPM n=15, CURATE.AI n=15). No adverse events or serious adverse events occurred. Clinical feasibility was high, with strong adherence to home BP monitoring (82%; green) and drug dosing (99%; green), and physician acceptability of CURATE.AI recommendations (64%; yellow). However, rapid achievement of BP targets in all 3 randomised groups meant that CURATE.AI successfully generated dose-response profiles for only 33% of participants (yellow), increasing to 73% (green) when including participants already within the target BP range. Although cumulative CURATE.AI guided doses differed substantially (≥10%) from standard care in 100% of participants (green), we observed no significant between group differences in 1-month office BP (SBP, ANCOVA: F(2,38)=1.84, p=0.173; DBP, F(2,38)=2.73, p=0.078). Time to target BP between HBPM and CURATE.AI did not differ significantly (SBP, log-rank p=0.889; HR=0.92, 95% CI 0.26–3.24; DBP, log-rank p=0.181; HR=0.47, 95% CI 0.14–1.55). Time in therapeutic range for home BP also did not differ significantly. Conclusions: This proof-of-concept trial demonstrates that AI-assisted titration of antihypertensive medication is both safe and feasible for collaborative clinical use. Although BP outcomes were not superior to standard care, the study was not powered for efficacy. The pilot findings highlight areas for refinement of the study design and support the feasibility of conducting a larger randomised controlled trial to evaluate efficacy and cardiovascular outcomes.Curate.AI guided titrationoutcomes

doi.org/10.1093/eurheartj/ehag785.3053Open accessInterventions