Log in to MyACC
Prognostic Models for Mortality and Morbidity in HFpEF
Quick Takes
- The PREDICT-HFpEF risk prediction model, developed and validated in this study, uses readily assessed clinical variables.
- For patients with HFpEF, this risk model accurately predicts clinical outcomes at 1 and 2 years (HF hospitalizations and CV death, CV death, all-cause death).
- The PREDICT-HFpEF risk model performs similarly to biomarker-driven models and better than the MAGGIC risk score.
Study Questions:
Can clinical outcomes be accurately predicted for patients with heart failure and preserved ejection fraction (HFpEF) using routinely collected variables?
Methods:
Data from the DELIVER trial (dapagliflozin use in HFpEF) from January 2023 to June 2023 were used to derive the prediction models. The outcomes studied were a composite of HF hospitalization (HFH) or cardiovascular (CV) death, CV death, and all-cause death. Multivariable analysis was used to select the predictor variables. The models were then validated using data from the PARAGON-HF (sacubitril/valsartan use in HFpEF) and I-PRESERVE (irbesartan use in HFpEF) trials. Comparisons between the new prediction model (PREDICT-HFpEF) and existing models like the Meta-Analysis Global Group in Chronic (MAGGIC) risk score and the biomarker-driven EMPEROR-Preserved risk score were made.
Results:
The PREDICT-HFpEF prediction models were created from a cohort of 6,263 patients in the DELIVER trial, and subsequently validated in 4,796 patients in the PARAGON-HF trial and 4,128 patients in the I-PRESERVE trial.
The model for the composite of HFH and CV death included 11 common variables, including HF duration, history of HFH, history of diabetes, history of chronic obstructive pulmonary disease, history of transient ischemic attack (TIA)/stroke, geographic region, heart rate, N-terminal pro–B-type natriuretic peptide (NT-proBNP) level, creatinine level, and use of a sodium-glucose cotransporter 2 inhibitor (SGLT2i). The model demonstrated good discrimination at 1 year (C-statistic, 0.73; 95% confidence interval [CI], 0.71-0.75) and 2 years (C-statistic, 0.71; 95% CI, 0.70-0.73) and validated in the PARAGON-HF (C-statistic at 1 year, 0.71; 95% CI, 0.69-0.74; C statistic at 2 years, 0.68; 95% CI, 0.66-0.70) and I-PRESERVE (C-statistic at 1 year, 0.75; 95% CI, 0.73-0.78; C-statistic at 2 years, 0.73; 95% CI, 0.71-0.75) trial cohorts. The PREDICT-HFpEF model performed similarly to the biomarker-driven EMPEROR-Preserved model and better than the MAGGIC risk score.
The models created for CV death and all-cause death prediction similarly performed well. Predictor variables differed between the three models, but all used HFH in the past 6 months, history of TIA/stroke, NT-proBNP level, and creatinine level.
Conclusions:
The PREDICT-HFpEF prediction model, developed and validated in this study, uses readily assessed clinical variables and accurately predicts clinical outcomes at 1 and 2 years in patients with HFpEF.
Perspective:
Accurate risk predictions for patients with HFpEF are difficult given the heterogenous nature of the condition but remain important for clinical practice. While risk models should not be the sole factor in deciding individual-level risk and treatment choices, they can be very powerful tools. Being able to use objective data to predict risk may help to reduce inherent biases and misconceptions about a patient’s clinical trajectory and can help to enhance shared decision making with patients, providing additional key information for patients to understand their condition and make future decisions. The PREDICT-HFpEF model developed and validated in this study provides an important tool to be used in the care of patients with HFpEF and seems to perform as well or better than some of the existing risk models available. Future studies in risk prediction are likely on the horizon; this current work marks an important advancement in the field.
Clinical Topics: Heart Failure and Cardiomyopathies
Keywords: Heart Failure, Preserved Ejection Fraction, Risk Assessment
< Back to Listings
