Physicians don’t trust predictive tools for heart failure
Several models have been developed to predict mortality among heart failure patients, but clinicians remain reluctant to use them in everyday practice. In fact, fewer than 1 percent of patients received a prognostic estimate from their physicians in a European registry analysis published in JACC: Heart Failure.
“These scores are not routinely calculated in clinical practice, primarily because of their poor reliability at the individual patient level and also because treatments that specifically fit different levels of risk have not been established,” wrote corresponding author Aldo P. Maggioni, MD, with ANMCO Research Center in Florence, Italy, and colleagues.
The study included 6,161 chronic heart failure patients with the necessary clinical information to calculate each of four risk scores: CHARM (Candesartan in Heart Failure-Assessment of Reduction in Mortality), GISSI-HF (Gruppo Italiano per lo Studio della Streptochinasi nell'Infarto Miocardico-Heart Failure), MAGGIC (Meta-analysis Global Group in Chronic Heart Failure) and SHFM (Seattle Heart Failure Model).
At one year of follow-up, 91.8 percent of the patients were alive.
Maggioni and coauthors noted all the risk scores other than SHFM overestimated mortality. However, SHFM demonstrated the worst overall accuracy with an area under the curve (AUC) of 0.714, while MAGGIC had the best accuracy of 0.743.
“We showed a significant difference in the accuracy of four different risk scores predicting all-cause mortality, with the MAGGIC risk score outperforming others, particularly the more popular SHFM,” the researchers noted. “Nonetheless, calibration at different levels of risk was still imperfect for most scores, with a general trend toward overestimation of risk, which could in part justify why less than 1 percent of patients received a prognostic estimate from their enrolling physician.”
Maggioni et al. said the models’ overestimation of mortality may be explained by improvements in heart failure management, including the increased use of renin-angiotensin system blockers and beta-blockers from when these tools were developed. Developing new, more precise risk prediction models based on current practice is “urgently needed,” according to the authors.
Importantly, the study population was relatively young (age 64.9 on average) and about two-thirds male. Analyzing risk scores’ utility in an older population could be important given the “more challenging issues in prognostication and in the choice of the most appropriate therapy,” the authors wrote.
“Critical medical decisions (in cardiology as in other disciplines of medicines) are based on life expectancy, and prognostication remains essential to developing appropriate treatment plans and to relaying truthful information to the patient and his/her family members,” Maggioni and colleagues wrote. “Thus, further research is needed in search of increased accuracy and precision at the individual, more than at the population level.”
The authors of an accompanying editorial agreed, saying accurate risk assessment allows for more informed decision-making on heart transplantation or ventricular assist device implantation and how consistently and intensely a patient should be monitored.
“Clearly, we need new models in contemporary cohorts and these will probably be different for heart failure with reduced compared with preserved ejection fraction,” wrote Joanne Simpson, MBChB, and John J.V. McMurray, MD, both with the British Heart Foundation Cardiovascular Research Centre at the University of Glasgow.
B-type natriuretic peptides and troponin are now commonly measured in clinical practice and could provide incremental predictive value, they noted. Also, serial measurements of these and other variables would be expected to provide additional accuracy over single measurements, particularly if electronic health records allow for these factors to be easily tracked over time.
“New mathematical approaches may also allow us to build better models and deal with the Achilles’ heel of missing data,” Simpson and McMurray wrote. “What we would really like are not just models that predict death from any cause, but rather the two major modes of death—sudden death and death from progressive pump failure separately. Then we could really start to think about targeting specific therapies.”
Daniel joined TriMed’s Chicago editorial team in 2017 as a Cardiovascular Business writer. He previously worked as a writer for daily newspapers in North Dakota and Indiana.