Medication use and cardiovascular outcomes in patients with suspected angina due to coronary artery disease: a data-driven analysis of the SCOT-HEART trial linked to national prescribing data

Background

Coronary artery disease is a leading cause of death and disability. The Scottish COmputed Tomography of the HEART (SCOT-HEART) trial is a multicentre randomised controlled trial which assessed whether the use of coronary computed tomography angiography (CCTA) could improve diagnosis and management in patients with chest pain suspected to be due to coronary artery disease. SCOT-HEART randomised 4,146 patients with suspected angina to standard care or standard care plus coronary CT angiography (CCTA). Over 10 years, CCTA-guided management reduced coronary artery disease death or non-fatal myocardial infarction (MI) (hazard ratio 0.79, 95% CI 0.63 to 0.99) and was associated with more frequent preventive therapy prescribing.^{1, 2} However, only aspirin and statin use has been examined, in a limited analysis at yearly timepoints. The wider use of other cardiovascular medications (antihypertensives, beta-blockers, anticoagulants, anti-anginals), non-cardiovascular medications, adherence over time, and their relationship to outcomes and to health inequalities remain unknown.

This PhD will use information from the Scottish national Prescribing Information System (PIS), which captures every community-dispensed prescription in Scotland. The PIS records every community-dispensed prescription nationally and has been linked to the SCOT-HEART cohort, enabling longitudinal reconstruction of each patient's medication history. The student will build a reproducible medication feature-extraction pipeline (drug classes, repeats, gaps, persistence and adherence metrics such as proportion of days covered). The student will characterise 10-year medication trajectories across trial arms and disease severity, quantify how medication use and adherence relate to CCTA outcome benefit, and examine socioeconomic and sociodemographic variation in prescribing and adherence to address health inequality.

The project sits at the interface of cardiovascular medicine, health data science and pharmacoepidemiology, providing training in large-scale linked-data analysis, machine learning, and causal/survival methods.

Aims

1. Develop a reproducible, data-driven medication feature-extraction module from PIS linked to SCOT-HEART. This will classify drugs and derive repeat, gap, persistence and adherence metrics (e.g. proportion of days covered).
2. Characterise medication-use trajectories over the 10 years of the trial across randomisation arms and by baseline risk, pre-randomisation treatment and CT-defined disease severity.
3. Determine the association between medication use/adherence and the primary outcome (CHD death or non-fatal MI), and assess the extent to which medication use mediates the benefit of CCTA-guided management.
4. Examine socioeconomic and sociodemographic variation in prescribing and adherence to quantify and address health inequalities in preventive cardiovascular care.

Training Outcomes

Health data science: analysis of large-scale linked trial and routine prescribing data, data management and governance, data engineering and reproducible analysis pipelines (R/Python, version control).

Quantitative and machine-learning methods: feature engineering for medication adherence, trajectory/clustering analysis, survival analysis and causal mediation.

Clinical and translational understanding: cardiovascular disease and prevention, pharmacoepidemiology, and health-inequalities research.

Transferable skills: scientific writing and publication, presentation at national/international meetings, research ethics and information governance, and an internship placement.

References

1. Williams et al. Coronary CT angiography-guided management of patients with stable chest pain: 10-year outcomes from SCOT-HEART. Lancet 2025:405(10475);329-337. 
2. SCOT-HEART Investigators. Coronary CT angiography and 5-year risk of myocardial infarction. N Engl J Med 2018;379:924–33.