Prediction model for cardiac allograft vasculopathy: Comparison of three multivariable methods

BackgroundCardiac allograft vasculopathy (CAV) remains an important cause of graft failure after heart transplantation (HT). Although many risk factors for CAV have been identified, there are no clinical prediction models that enable clinicians to determine each recipient's risk of CAV. MethodsWe studied a cohort of 14328 heart transplant recipients whose data were reported to the International Society for Heart and Lung Transplantation Registry between 2000 and 2010. The cohort was divided into training (75%) and test (25%) sets. Multivariable modeling was performed in the test set using variables available at the time of heart transplant using three methods: (i) stepwise Cox proportional hazard, (ii)regularized Cox proportional hazard, and (iii) Bayesian network. ResultsCardiac allograft vasculopathy developed in 4259 recipients (29.7%) at a median time of 3.0years after HT. The regularized Cox proportional hazard model yielded the optimal performance and was also the most parsimonious. We deployed this model as an Internet-based risk calculator application. ConclusionsWe have developed a clinical prediction model for assessing a recipient's risk of CAV using variables available at the time of HT. Application of this model may allow clinicians to determine which recipients will benefit from interventions to reduce the risk of development and progression of CAV.