CHAPTER 1 8 INTRODUCTION Clinical practice guidelines are developed for the general heart failure population and their evidence generally depend on randomized controlled trials. However, clinical trials are often criticized for poor generalizability, and this is among the cited reasons for underuse of effective treatments. For instance, lack of trial data in older patients explains in part the under-prescribing of warfarin in patients aged 75 years and above; the group who has the highest prevalence for non-rheumatic atrial fibrillation and also at highest risk without treatment.1–3 Historically, younger, white men were considered the normative population in heart failure trials whereas women or older persons were expected to have similar responses.4–6 However, recent posthoc analyses of cardiac resynchronization therapy (CRT) trials revealed that females were more likely to respond to CRTs and at shorter QRS complex duration than men7,8 whereas observational studies demonstrate that women benefited from lower doses than guideline recommended target doses of beta-blockers and RAASinhibitors,9 suggesting that this one-size-fits-all approach is flawed. Extrapolation of clinical trial results to its population of interest is often broadly described as external validity. It reflects a complex assessment of patient selection, trial setting, differences between trial and routine practice, outcome measurement and statistical methods.1,10 In this context, terms including generalizability, applicability, transferability, transportability and extrapolation have been used with overlapping meanings.1,10,11 The present thesis focuses on generalizability, defined as making inference from a study sample back to the target population, e.g. the domain (inclusive of the study population).11 While it is neither realistic nor reasonable to expect generalizability to every patient and setting, it can be assessed and described to allow clinicians, regulators and policy makers decide to whom trial results are applicable.1 Moreover, standardizing assessments of generalizability in current trials will pave the way for improving future trials. One way to formally ensure generalizability is by replicating the study in its new target population as effectiveness or pragmatic trials but this is impractical and very expensive to implement widely.10,12 In the absence of guidelines on trial generalizability assessment, another approach would be to conduct secondary data analyses in heart failure trial datasets, trial registers and observational cohort data.1,12 These methods can be classified into those involving emulation of
RkJQdWJsaXNoZXIy MjY0ODMw