Author + information
- John E. Madias, MD∗ ()
- ↵∗Division of Cardiology, Elmhurst Hospital Center, 79-01 Broadway, Elmhurst, New York 11373
I enjoyed studying the contribution by Ahmad et al. (1), and the accompanying editorial by Francis et al. (2), published in the October 28, 2014 issue of the Journal, about the application of cluster analysis to the data from 1,619 participants of HF-ACTION (Heart Failure: A Controlled Trial Investigating Outcomes of Exercise Training) study. The investigators reported that cluster analysis provided an advantage over traditional phenotyping, based on the subjective symptomatic assessment of the patients via the New York Heart Association functional classification I to IV, and A to D stages, and imaging-based left ventricular ejection fraction (LVEF) in predicting outcomes (all-cause and cardiovascular, mortality, and hospitalization risks) and response to surrogate parameters (change in peak oxygen consumption and in standard 6-min walk test). Clinicians following longitudinally large numbers of heart failure (HF) patients with a reduced LVEF in cardiology clinics have long been accustomed to the incongruity between the functional classification/staging (New York Heart Association class II to IV) and LVEF (≤35%) and outcomes of their patients, stemming from our current coarse phenotyping of a highly heterogeneous disease as HF and the impact of comorbidities.
Ahmad et al. (1) arbitrarily employed 45 pre-specified clinical variables and identified 4 phenotypic clusters, with intracluster similarities and intercluster differences, in which they showed diverse mortality and hospitalization rates. It is of interest that in the exhaustive list of variables used (1), a measure of the patients’ overall compliance with their management in general, and with drug taking in particular, is missing (issues of frequent concern in cardiology clinics), for which the investigators are not responsible. Ahmed et al. (1) and Francis et al. (2) cited the reasons why a number of trials (refs. 7, 8, and 35 in Ahmed et al. ) and patients with HF with a reduced LVEF (refs. 5 to 8 in Francis et al. ) “have failed to meet their endpoints” (2), and Ahmed et al. (1) stated that “we have seen such little progress in developing new treatments for this disorder.”
Although we need to adopt the philosophy of enhanced and refined phenotyping in designing future HF clinical trials (2), what should have precedence now are the following objectives: 1) a “culture” change in practice and research, conducive to distancing ourselves from the attachment and the “false security” provided by the New York Heart Association functional classification I to IV, A to D staging and LVEF, when phenotyping our HF patients, and adopting more liberal classification schemes; and 2) application in research of cluster analysis to existing data from previously published randomized HF clinical trials, and data from electronic medical records, modeled after the present study (1), in an attempt to come up with winning sets of a few clusters that will outperform our current HF classification systems. These 2 objectives will pave the way to a future when design of “rational” clinical trials (2) will be feasible and cost-effective.
- 2015 American College of Cardiology Foundation
- Ahmad T.,
- Pencina M.,
- Schulte P.J.,
- et al.
- Francis G.S.,
- Cogswell R.,
- Thenappan T.