3.42 General discussion 201 representative populations to trials based on HF epidemiology and novel methods for recruitment of underrepresented patient groups.86,87 A widely accepted solution to generalizable clinical evidence is pragmatic trials but this option can be unattractive when returns on investment are uncertain, particularly to industry sponsors.55 Perhaps more stands to be gained by embedding pragmatic elements earlier in phase III explanatory HF trials.55,88 Multi-national HF registries represent promising platforms for cost-efficient and more inclusive patient identification and screening for double-blind explanatory trials. A pioneering example is the DAPA-MI trial for myocardial infarction, the first indication-seeking registry-based RCT which enrolled patients from cardiovascular disease registries in Sweden and the UK. On a similar note, adaptive trial designs have been proposed in a guidance for industry by the FDA in situations where a drug is expected to have larger effects in a targeted subpopulation, whether by demography or pathophysiology.89 Rather than an all-or-none rule, a trial may enroll populations with and without the characteristic of interest up till an interim analysis period. Then, a decision can be made based on pre-specified terms whether to continue with the overall study population or restrict to the targeted group.56 An advantage of such adaptive enrichment designs is that data on the intervention will be available for the non-targeted or complementary subpopulation.89 Lastly, generalizability metrics provide a quantifiable means to benchmark representativeness of trial samples against the intended target population as well as infer expected treatment effects at a population level. Numerous methods have been proposed for calculating eligibility based on eligibility criteria, assessing overlap between study samples and target population with regard to demographic and prognostic characteristics and statistical extrapolation of effects from narrow study samples to broader populations by applying weights derived from propensity scores to the RCT sample to mimic the target population and estimate population average treatment effects.90–96 Pre-requisites for the abovementioned methods include access to individual-level data to both RCTs and the target population with comparable measure and sufficient overlap of covariates.28 It is exciting and challenging times for therapeutic development in heart failure now that generalizability and representation of trial populations is brought into focus. The path forward requires multi-faceted and -stakeholder strategies, both working in tandem.
RkJQdWJsaXNoZXIy MjY0ODMw