Understanding Masked Hypertension and Its Hidden Dangers

Masked hypertension (MHT), a condition where blood pressure (BP) readings appear normal during clinical visits—typically below 140/90 mmHg—but elevate to hypertensive levels (24-hour ambulatory BP monitoring, or ABPM, ≥130/80 mmHg) outside the office, poses a stealthy threat to cardiovascular health. This discrepancy arises due to factors like 'white coat' anxiety inversely affecting out-of-clinic measurements or true underlying elevations undetected in brief office checks. MHT affects roughly 10-18% of normotensive individuals globally, but in young South African adults, studies indicate prevalence up to 18%, heightening risks of heart disease, stroke, and organ damage comparable to or exceeding sustained hypertension.

In resource-constrained settings like South Africa, where traditional diagnostics rely on cumbersome 24-hour ABPM devices—costly and uncomfortable, limiting widespread use—MHT often evades detection until advanced complications emerge. Untreated, it contributes to South Africa's alarming hypertension burden, one of the world's highest, with over 8 million adults affected and low control rates exacerbating non-communicable disease epidemics.

The Hypertension Crisis in South Africa

South Africa grapples with a hypertension epidemic, with prevalence rates soaring to 78% in some adult cohorts, far outpacing global averages. Among young adults aged 20-30, even normotensives harbor subclinical risks, influenced by urbanization, dietary shifts, stress, and genetic factors prevalent in black South African populations. The World Health Organization notes sub-Saharan Africa's rapid rise, projecting 125 million cases by 2025, straining public health systems amid competing priorities like HIV.

Masked forms amplify this, as office BP misses ambulatory spikes linked to sympathetic overactivity and inflammation. Local data from cohorts like African-PREDICT reveal early vascular remodeling and left ventricular hypertrophy in these 'healthy' youths, underscoring the need for accessible screening. Government initiatives target awareness, but diagnostic gaps persist, particularly in rural areas lacking ABPM infrastructure.

For academics and researchers eyeing impact, South Africa's cardiovascular research landscape offers fertile ground. Institutions like North-West University lead prospective studies, fostering collaborations that yield actionable insights. Explore research jobs or South African university opportunities to contribute.

The African-PREDICT Study: Foundation of Breakthrough Research

Launched by North-West University's Hypertension in Africa Research Team (HART) and the Medical Research Council (MRC) Research Unit for Hypertension and Cardiovascular Disease in Potchefstroom, the African-PREDICT study (NCT03292094) prospectively tracks early cardiovascular pathophysiology in 1,202 black and white South African adults aged 20-30, initially normotensive. Participants underwent comprehensive phenotyping: office and 24-hour ABPM, biomarkers (lipids, cytokines, hormones like DHEA-S), body composition, physical activity via accelerometers, and cardiac ultrasounds assessing left ventricular mass.

This bi-racial cohort captures socioeconomic and ethnic diversity, enabling nuanced risk profiling. Ethical standards were rigorous, with North-West University Health Research Ethics Committee approval (NWU-00001-12-A1). A 10-year follow-up promises longitudinal insights, positioning NWU as a hub for predictive cardiology. Such datasets empower global AI innovations, highlighting South African higher education's pivotal role.

Birth of the AI Diagnostic Tool: A Transatlantic Collaboration

Published November 17, 2025, in Frontiers in Physiology (doi: 10.3389/fphys.2025.1684693), the study "Machine learning model for detecting masked hypertension in young adults" unites U.S. and South African scholars: Brendyn Miller (Wake Forest University), Samuel J. Coeyman and William J. Richardson (University of Arkansas), alongside NWU's Annemarie Wentzel and Carina M.C. Mels. Leveraging African-PREDICT's rich dataset, they engineered ML models to predict MHT from a single clinic visit, bypassing ABPM barriers.

This partnership exemplifies higher education's cross-border synergy, with SA providing gold-standard data and U.S. expertise in computational modeling. Lead author Richardson emphasizes pattern recognition across 335 preprocessed features, mirroring clinical ratios but scaling to hundreds of biomarkers. For aspiring researchers, such collaborations open doors—check postdoc positions in biomedical engineering or epidemiology.

Read the full paper here.

Step-by-Step: How the Machine Learning Model Operates

The methodology unfolds methodically: Raw African-PREDICT data (526 features) underwent preprocessing—removing incomplete cases (>10% missing), engineering categories (e.g., overweight BMI ≥25 kg/m²), imputing via Multiple Imputation by Chained Equations (MICE), and scaling with MinMaxScaler. Five feature selection techniques pruned to optimal sets: Recursive Feature Elimination-SVM (RFE-SVM), BorutaSHAP, LASSO (Least Absolute Shrinkage and Selection Operator), literature-based manual, or none.

• Five classifiers tested: Logistic Regression (LR), Random Forest (RF), Extreme Gradient Boosting (XGBoost/XGB), Artificial Neural Network (ANN), Stacking (STK).
• Hyperparameters optimized via Bayesian search with 5-fold cross-validation on 1,042 participants (160 MHT cases).
• SHAP (SHapley Additive exPlanations) and Partial Dependence Plots (PDPs) ensured interpretability.

LASSO + XGB emerged superior, selecting 21 features for robust predictions. This transparent pipeline suits clinical integration, demystifying 'black box' AI.

Results Spotlight: 83% Accuracy and Key Biomarkers

The LASSO-XGB model dazzled with 82.8% accuracy, 0.855 ROC AUC, 0.649 sensitivity, 0.866 specificity, and F1-score 0.571 on unseen test data—outpacing simple systolic BP >120 mmHg cutoffs (74.2% accuracy, 0.719 AUC). Odds ratio of 11.96 underscores predictive power, correctly reclassifying 19% false positives.

Top quartet: systolic BP (highest SHAP), body weight, left ventricular (LV) mass at systole (echo-derived), dehydroepiandrosterone sulfate (DHEA-S, adrenal stress hormone). PDPs reveal non-linear risks: higher values correlate with MHT probability, tying to sympathetic activation and early remodeling.

Transformative Implications for South African Healthcare

This tool democratizes MHT detection, integrable into electronic health records for automated flagging during routine visits. In South Africa, where ABPM access lags, it slashes costs and discomfort, enabling targeted ABPM only for high-risk cases. Early intervention—lifestyle mods, meds—could avert progression, vital amid rising youth CVD.

Stakeholders praise: NWU's HART pioneers data sovereignty, while UArk's modeling accelerates translation. Challenges include external validation and generalizability beyond young Africans, but low false positives minimize overtreatment risks. For clinicians, it frees time for patient rapport, as Richardson notes.

Learn more via UArk's coverage here. NWU details here.

Higher Education's Vanguard Role in AI Health Innovations

North-West University exemplifies SA academia's thrust in precision medicine, with HART's longitudinal cohorts fueling AI advancements. Faculty like Wentzel and Mels mentor postgrads in phenotyping, while global ties amplify impact. UArk's chemical engineering lens innovates biomarkers, training interdisciplinary talent.

• Benefits: Accelerates PhD theses, grants (NRF, MRC).
• Risks: Data privacy, AI bias mitigation.
• Comparisons: Echoes UAE's AI-health pushes, but SA emphasizes equity.

Aspiring lecturers? View lecturer jobs or professor jobs in health sciences.

Photo by Margaret Jaszowska on Unsplash

Future Horizons: Validation, Expansion, and Careers

Prospects gleam: 10-year African-PREDICT follow-up refines models; external trials across ages/ethnicities beckon. Longitudinal causality probes DHEA-S's stress link. Scalable to wearables, it heralds proactive care.

For careers, SA unis seek AI-savvy researchers. Internal links: Craft your CV; browse Rate My Professor. Final CTA: Discover higher ed jobs, university jobs, professor ratings, career advice.