Risk prediction in patients with heart failure with preserved ejection fraction: the LIFE-Preserved model.

Figures

• 

  Figure 1. C-statistics for heart failure hospitalization or cardiovascular death in external validation data. Forest plot displaying the discrimination in all external validation data is based on Harrell’s C-statistic for heart failure hospitalization or cardiovascular death adjusted for non-cardiovascular death. Discrimination was assessed at the 75th percentile of follow-up duration (2-year risks for EMPEROR-Preserved and the NHSE SDE, 4-year risks for TOPCAT-Americas, and 8-year risks for HF-Particles) and at 10-year risks for the Department of Veterans Affairs. Pooled C-statistics were calculated using inverse-variance weighting. Cohort sizes here may be lower than the total cohort population listed in Table 3 as risks can only be estimated up to age 90 (e.g. 89-year-olds can have 1-year risk but no risks for 2-year or higher and are consequently excluded at the horizon for validation). CI, confidence interval

• 

  Figure 2. Calibration for heart failure hospitalization or cardiovascular death in external validation data. Calibration of the LIFE-Preserved model in external validation. Risks were recalibrated in the NHSE SDE and EMPEROR-Preserved using expected–observed ratios specified in Supplementary data online, Table S11. Smoothed calibration plots of predicted risks vs observed risks and a histogram of the predicted risks is shown. Cohort sizes here may be lower than the total cohort population listed in Table 2 as risks can only be estimated up to age 90 (e.g. 89-year-olds can have 1-year risk but no risks for 2-year or higher and are consequently excluded at the horizon for validation). HF, heart failure; CV, cardiovascular

• 

  Figure 3. Clinical example of estimation of individualized treatment benefit. Clinical example of estimation of individual short-term and lifetime risk and individualized treatment benefit after starting a sodium–glucose cotransporter 2 inhibitor for two different heart failure with preserved ejection fraction patients. Event-free survival in the curves is defined as survival without any heart failure hospitalization or cardiovascular death. Treatment benefits were estimated by combining the individual absolute risks from the LIFE-Preserved model with the published hazard ratios from trials or meta-analyses. In this example, the published hazard ratio of .81 was used as treatment effect of starting a sodium–glucose cotransporter 2 inhibitor.41 Treatment benefits do not reflect observed treatment use in the cohorts and should be interpreted as projections of potential benefit. The individualized treatment benefit can be shown as an absolute risk reduction (bar chart, bottom row) or as gain in heart failure hospitalization-free life-years (darker area in survival plot, top row). HFpEF, heart failure with preserved ejection fraction; NYHA, New York Heart Association; BMI, body mass index;…