Representativeness of participants included in heart failure trials: a systematic review

Talk Code: 
Pieter MacKeith
Gollop ND, Fleetcroft R, Ford J, Perera K, Shafi AA, Sorefan J, Thurlow C, Wakelin R, Cowie MR, Steel N.
Author institutions: 
University of East Anglia, Norwich Medical School, Norfolk and Norwich University Hospital, JS Medical Practice London, North Central and East London Deanery, National Heart and Lung Institute Imperial College London.


Heart failure is often managed in the community by primary care doctors. Guidelines for the management of heart failure are based on several randomised controlled trials (RCTs). The problem is that most of these RCTs may have been undertaken on secondary care populations with more advanced disease and that these patient do not represent the primary care population with heart failure. Moreover, previous research found that Beta blockers and ACE inhibitors carry significant morbidity and account for hospitalisation because of adverse drug reactions. Our aim was to compare the participants in randomised control trials (RCTs) with a primary care population.


A systematic review was undertaken using Cochrane methodology to identify RCTs of heart failure drugs. Electronic databases were searched from inception to July 2015. Studies with a follow up of less than six weeks, single dose and studies not judged to be generalisable to a primary care population were excluded. A UK-based sample of people aged ≥45 were randomly selected from computerised general practice registers representative of a wide geographical spread of populations and used as a ‘reference population’ to which we could compare the included studies. NYHA class, baseline cardiovascular risk factors, baseline cardiovascular comorbidities and use of heart failure drugs were compared between the reference study and each extracted study. Each patient-specific variable was compared to the reference study in terms of prevalence or frequency of use. If the extracted study population had a <10% deviation from the reference study, it was termed as a ‘close match’, 11-20% deviation was a ‘fair match’, and >20% a ‘poor match’.


3975 abstracts were screened for inclusion. After screening, 30 RCTs were included representing 43,454 patients with heart failure. No study was a close match to all four NYHA classes when compared to the reference population. Furthermore no study was a close match to NYHA class I which comprises 47% of the reference population with heart failure. Six studies (36%) were a close match for NYHA class II, five studies (7%) for NYHA class III and 18 studies (10%) for NYHA class IV.


We found that heart failure studies are not typically representative of the primary care population of patients in England, and that patients with more severe heart failure are over-represented with milder heart failure in the community under-represented. Guidelines based on these trials may overestimate the benefits of heart failure drugs for typical primary care patients. Guideline committees such as NICE should present the evidence for effectiveness in different target populations separately when making recommendations. Meta-analyses using individual patient data from the studies identified in this research would provide more evidence for the effectiveness of treatment in mild heart failure.

Submitted by: 
Pieter MacKeith