| アブストラクト | OBJECTIVE: Syncope is one of the most common side effects associated with antihypertensive medication. In patients at increased of syncope, the additional risk of harm from antihypertensive medication may outweigh the potential benefits of treatment in terms of cardiovascular risk reduction. However, it is unclear how to identify patients at most risk of syncope events. This study aimed to develop a clinical prediction model for risk of hospitalisation or death from syncope. DESIGN AND METHOD: This was a cohort study using data from the Clinical Practice Research Datalink (CPRD) in the UK. The electronic health records of patients aged greater than 40, with at least one blood pressure measurement between 130-179 mmHg were included. Outcomes were defined as a syncope event resulting in hospitalisation or death within 10 years of baseline. Predictors of syncope were based on the literature and expert opinion and included patient characteristics, past medical history and prescribed treatment (including antihypertensive prescription). A Fine-Gray model was used to adjust for competing risk of mortality and results are reported as subdistribution hazard ratios (SHR). RESULTS: A total of 1,772,617 patients were eligible for the study, with mean age of 59 and 48% males. Median follow up was 6.2 years with 39898 syncope events (2.3%). The antihypertensive medications most strongly associated with syncope were alpha blockers (SHR: 1.21, 95%CI 1.15 to 1.28) and ACE inhibitors (SHR: 1.19, 95%CI 1.16 to 1.22). Other important predictors included age (Figure 1), male sex, high social deprivation, heavy alcohol consumption, previous syncope, diabetes, dementia, structural cardiac problems, arrhythmias, spinal cord injuries, parkinsonism, cardiopulmonary disease and prescription of antidepressants, antipsychotics and opioids (table 1). CONCLUSIONS: This prediction model identified a number of strong predictors of syncope which are routinely available in an individual's electronic health records. The accuracy of this model will examined in a further approximately 3,000,000 patients from a different electronic health record database. If it is found to perform well, such a model could be used to provide personalised estimates of an individual's risk of harm from antihypertensive treatment, thus facilitating more informed treatment choices. |
|---|
