Pharmacogenomics-guided therapy aims to personalize antihypertensive prescribing to improve efficacy and medication safety. Real-world evidence from South Asian populations, and specifically from Pakistan, is limited despite a high burden of cardiovascular disease in the region. The study analyzed routine electronic health record (EHR) data from Khyber Pakhtunkhwa to characterize genotype–drug response relationships in an underrepresented clinical population.
The primary objective was to evaluate associations between selected pharmacogenomics variants and antihypertensive therapy response using EHR data from patients treated at Kuwait Teaching Hospital (KTH), Peshawar. A secondary objective was to assess the incremental predictive value and practical limitations of genotype-based stratification in routine clinical care.
This retrospective cohort included 1,247 hypertensive adults treated at KTH between January 2023 and August 2025. Eligible patients initiated antihypertensive monotherapy, had at least 12 months of EHR follow-up, and had available genotyping results to reduce misclassification of treatment response.
Five genetic variants were genotyped using TaqMan assays: CYP3A5 rs776746, ADD1 rs4961, NEDD4L rs4149601, ADRB1 rs1801253, and ACE rs4340. Primary outcomes were change in systolic blood pressure (SBP) and achievement of target blood pressure (< 140/90 mmHg).
Statistical analyses included ANOVA for drug-class and genotype comparisons, multivariable logistic regression to identify independent predictors of target BP achievement, and receiver operating characteristic (ROC) analysis to assess model discrimination. Multiple testing for variant–drug class associations used a Bonferroni-corrected significance threshold (α = 0.002). Internal validation of model discrimination used bootstrap resampling with 1,000 iterations.
Cohort characteristics: mean age 63.1 ± 11.5 years; 661 (53.0%) female. Baseline BP averaged 157.1 ± 18.8 / 91.7 ± 11.1 mmHg.
Drug-class comparisons showed that calcium channel blockers (CCBs) produced a greater mean SBP reduction than other antihypertensive classes (22.32 ± 8.78 vs. 19.68 ± 8.67 mmHg; t = 4.43, p = 1.0 × 10-5). The overall drug-class ANOVA was significant (F = 5.27, p = 3.3 × 10-4) and survived Bonferroni correction.
Variant-specific findings identified two Bonferroni-significant genotype–drug associations. Patients with **CYP3A5 rs776746 3/3 genotype experienced greater SBP reduction with CCB therapy (23.96 ± 8.79 vs. 19.78 ± 8.17 mmHg; p = 9.7 × 10-5). Patients with ADD1 rs4961 GG genotype showed larger SBP reduction when treated with diuretics (24.02 ± 8.84 vs. 17.66 ± 8.69 mmHg; p = 1.7 × 10-6).
On direct recalculation from reported genotype counts, four of five variants deviated from Hardy–Weinberg equilibrium: CYP3A5 (p = 0.0058), NEDD4L (p = 0.0146), ADRB1 (p < 0.0001), and ACE (p = 0.0379). Only ADD1 (p = 0.554) was in equilibrium. The authors highlight that such disequilibrium may reflect population stratification, genotyping issues, or other sample characteristics that require caution in interpretation.
In adjusted logistic regression, independent predictors of achieving target BP included older age (adjusted odds ratio [aOR] = 1.019, 95% CI 1.007–1.031), lower baseline SBP (aOR = 0.915 per mmHg, 95% CI 0.906–0.925), and absence of diabetes (diabetes aOR = 0.670, 95% CI 0.499–0.901). CCB therapy (aOR = 1.362, 95% CI 0.975–1.902) and favorable CYP3A5/ADD1 genotypes (aOR ≈ 1.14 each) showed positive trends but were not statistically significant after adjustment.
The predictive model achieved an area under the ROC curve (AUC) of 0.847. After bootstrap-derived optimism correction, the AUC was 0.841 (95% CI 0.823–0.866), indicating good internal discrimination.
EHR data from Khyber Pakhtunkhwa demonstrate significant associations between genetic variants and antihypertensive response for CCB and diuretic therapy in this underrepresented South Asian population. The findings suggest potential value in exploring pharmacogenomics-informed prescribing to improve treatment response and medication safety in similar settings.
However, the authors explicitly caution that the observed associations are hypothesis-generating. The extensive deviations from Hardy–Weinberg equilibrium for multiple variants, possible population stratification, genotyping considerations, the observational design, and the absence of external validation impede immediate translation into routine clinical implementation.
Key limitations reported by the authors include the retrospective observational design, reliance on a single-centre EHR dataset, and lack of external validation. The Hardy–Weinberg disequilibrium observed for four variants raises concerns that must be addressed before clinical application; these may reflect population structure, sampling bias, or genotyping artifacts. The authors recommend that findings be replicated in independent cohorts, that population genetic structure be rigorously assessed, and that prospective or interventional studies be conducted to define the clinical utility of pharmacogenomics-guided antihypertensive prescribing in this and other underrepresented populations.
Ethics and conflicts: The Institutional Ethics Review Board of Kuwait Teaching Hospital approved the study (KTH-ERB-2023-045) with waiver of individual consent for retrospective analysis. The authors declared no competing interests.