Preventing respiratory infection-related cardiovascular disease events in primary care

Background: Cardiovascular disease events (CVD events, comprising coronary and cerebrovascular events) are major causes of morbidity and mortality. CVD can be prevented by medications that target the underlying pathological processes of thrombosis and atherosclerosis. When a patient is diagnosed with a respiratory infection their risk of CVD events is about four times higher than their background risk for the following four weeks. This infection-related CVD event risk is well characterised by epidemiological research, but clinical practice guidelines for primary care do not address it. Prior to this thesis there were no tools for predicting an individual’s risk of an infection-related CVD event and, apart from vaccines, no established interventions for this scenario. Overall aim: To investigate strategies for preventing infection-related CVD events in primary care. Approach: 1. Developing statistical models to identify patients with respiratory infections who are at risk of CVD events 2. Validating the prediction models, using them to derive a clinical risk prediction score 3. Estimating the effect of aspirin on infection-related cardiovascular events 4. Estimating the effect of statin use on infection-related cardiovascular events Methods: Four epidemiological studies using large cohorts from coded UK primary care records held by the Clinical Practice Research Datalink (CPRD). These data were linked to datasets of NHS hospital and Office of National Statistics (ONS) mortality and relative deprivation datasets. The first two studies used prediction modelling methods, and the next two used propensity modelling methods with logistic regression to estimate causal effects. Results: I developed two statistical models and derived a clinical prediction points-based tool, the DASHI score. DASHI comprises five clinical variables: Diabetes, Age, Smoking status, Heart failure and Infection diagnosis. External validation showed DASHI can predict risk of infection-related CVD events with good calibration and discrimination (both C statistic and observed to expected ratios were 0.85 with IQR 0.85-0.85). This performance was very similar to the regression models. Aspirin and statins were estimated to increase infection-related CVD events; Relative Risk 2.52 (95% CI 2.26 to 2.81) for aspirin and 3.17 (95% CI 2.41 to 4.08) for statins. Aspirin increased bleeding with a relative risk of 1.31 (95% CI 1.06 to 1.16). Conclusion: The DASHI score can predict risk of primary infection-related CVD events. The absolute risks are low for most people due to the short prediction period. It is unlikely that aspirin and statins increase CVD events given what we know about their effects in other settings. It is more likely the results are inaccurate because of confounding or coding problems in the datasets. In particular, prescriptions are recorded immediately in the clinical record, but there are delays before CVD events enter the datasets. This timing difference may have led to biases exacerbated by the short follow-up period. A definitive answer is likely to require a different approach, and may require different datasets, or a randomised controlled clinical trial.