Key result
A joint illness-death regression model for ICD recipients demonstrated adequate predictive performance at 5 years, with an area under the curve of 0.65 for shocks and 0.79 for death.
Why the study?
The risk of death or appropriate therapy varies widely among ICD recipients, prompting the development of a risk prediction tool that jointly considers future probabilities of ICD shock and death.
Can an illness-death regression model accurately predict the joint risk of ICD shock and death in patients receiving ICDs?
Observational (n=803)
Can an illness-death regression model accurately predict the joint risk of ICD shock and death in patients receiving ICDs?
Answers are grounded in this paper and cite their sources.
A joint risk prediction model demonstrates adequate performance for predicting individual risks of ICD shock (AUC 0.65) and death (AUC 0.79) at 5 years, which may aid in individualized counseling prior to device implantation.
BACKGROUND: The risk of death or appropriate therapy varies widely among recipients of implantable cardioverter-defibrillators (ICDs). The goals of this study were to develop a risk prediction tool that jointly considers future outcome probabilities of ICD shock and death. METHODS AND RESULTS: We performed a secondary analysis of patients receiving ICDs as part of the SCD-HeFT trial (Sudden Cardiac Death in Heart Failure Trial). We applied an illness-death regression model to jointly model both ICD shocks and death under the semi-competing risks framework, which predicts for each patient their probability of having received ICD shocks, dying, or both at any given point in time. Among 803 ICD recipients (mean age, 60 years; 23% women) followed for a median of 41.1 months, 430 (53.5%) patients completed the study without dying or receiving an ICD shock, 206 (25.7%) received at least 1 shock but survived, 113 (14.1%) died before experiencing a shock, and 54 (6.7%) received at least 1 shock and subsequently died. Predicted outcome probabilities based on baseline demographic and clinical variables reveal substantial heterogeneity in joint shock and death risks, both between patients at each time point and for each single patient across time. Overall, predictive performance for ICD shock and death individually was adequate, based on area under the curve at 5 years of 0.65 for shocks and of 0.79 for death. CONCLUSIONS: Our analysis of outcomes after ICD implantation provides an alternative predictive model for individual risk of death or ICD shocks. If validated, this may provide a useful tool for individualized counseling regarding likely outcomes after device implantation, while also informing the design of further studies to focus the clinical effectiveness and cost-effectiveness of ICD therapy. CLINICAL TRIAL REGISTRATION: URL: https://www.clinicaltrials.gov. Unique identifier: NCT00000609.
No takes yet. Share an insight, caveat, or question.
Reeder et al. (2019) conducted an observational in Patients receiving implantable cardioverter-defibrillators (n=803). Illness-death regression model (joint shock/death risk prediction) was evaluated on Predictive performance for ICD shock and death individually at 5 years (AUC 0.65 for shocks, 0.79 for death). A joint illness-death regression model for ICD recipients demonstrated adequate predictive performance at 5 years, with an area under the curve of 0.65 for shocks and 0.79 for death.
Synapse has enriched 10 related papers. Consider them for comparative context: