Predicting individual blood pressure response to additional antihypertensive treatment: a prototype decision support tool for healthcare professionals in the community

Abstract Nearly one-third of hypertensive adults require >2 antihypertensive (aHTN) medications for blood pressure (BP) control. Clinical guidelines become less prescriptive at this stage, reflecting the high heterogeneity in treatment response at the individual level. Treatment decisions then rely heavily on clinical expertise and understanding patient characteristics. As care shifts to community settings where specialised expertise is limited, a clinical decision support tool to estimate the magnitude of BP lowering, while incorporating patient history, recent BP and medications, could assist in making informed treatment decisions. Purpose Using data from the landmark trials ASCOT and SPRINT, in which treatment was driven by target BP, we estimated the systolic BP (SBP) response associated with changes in intensity across 9 aHTN classes, accounting for current treatment. Methods We developed a linear mixed model using ASCOT to predict future SBP, integrating SBP, aHTN medications, age, sex, ethnicity, BMI, cardiovascular events, and SBP variability from 3 past visits (PVs). Therapeutic intensity was quantified as dose equivalence (prescribed aHTN dose divided by its usual maximal dose) for: diuretics, beta-blockers (BB), ACE inhibitors (ACEi), calcium channel blockers (CCB), angiotensin II receptor blockers (ARB), anti-adrenergics, aldosterone antagonists, potassium (K)-sparing diuretics, and vasodilators. For combination compounds, equivalence was calculated for each component. Doses exceeded maximal were capped at 1.5. Model performance was validated in SPRINT. Results Among 18,318 ASCOT participants (mean age 64y, 23% female), 222,487 observations had complete 3-PV history (median 10 visits per participant), with median time in months to predicted visit: 5.4 for PV1, 9.8 for PV2, and 13.4 for PV3. Table 1 shows, per aHTN class, the number of visits where a participant was 1) on an aHTN at the time of PV1 (‘current medication’); 2) prescribed a change in the aHTN during PV1. Figure 1 shows the predicted effect of aHTN change on SBP, accounting for ongoing treatment and adjusted variables. Statistically significant SBP lowering from -4.5 to -19.5 mmHg, for a medication increase equivalent to the usual maximal dose, was demonstrated for all classes except vasodilators. Aldosterone antagonists produced largest response, then diuretics and CCB; K-sparing diuretics; ACEi and anti-adrenergics; and BB and ARB. The ASCOT model residual SD was 13.35 mmHg, with a comparable prediction error of 12.76 in SPRINT. Conclusions This model may be used as a clinical decision-making support tool to estimate BP response to changes in aHTN dose and combinations. Estimated SBP lowering shown here aligns with earlier studies, except for poor response to vasodilators used as/beyond third-line therapy. This is an observational analysis and may be subject to confounding and selection. Findings require real world validation to confirm clinical utility.