Balancing the Tightrope of Cardiac Allograft Rejection

> “The only relevant test of the validity of a hypothesis is comparison of prediction with experience.”> > Milton Friedman Let's face it. We love clinical prediction rules that forecast outcomes in cardiovascular medicine. Our clinical armamentarium is suffused with a variety of scores that seek to predict unique outcomes such as the Framingham risk score for future ischemic events,1 the Congestive Heart Failure, Hypertension, Age ≥75, Diabetes Mellitus, and Prior Stroke or Transient Ischemic Attack (CHADS2) score to ascertain the risk of neurological events in atrial fibrillation,2 or clinical decision rules to exclude acute coronary syndromes in patients presenting to the emergency department.3 Generally, for a prediction rule to be successfully inserted into our clinical lexicon, it should be highly relevant, easy to use, and exercise sufficient discrimination between extremes to allow for deployment of decisional strategies. This frequently requires an artificial reductionism of the clinical outcome of interest into yes or no categorical variables. Often, such an oversimplification of an otherwise complex outcome variable can render it to be less informative.Article see p 3013In this issue of Circulation , Kilic and colleagues4 sought to develop a clinical prediction rule for the risk stratification of propensity for cardiac allograft rejection. Using a relatively large and contemporary cohort of adult heart transplant recipients (1998–2008) drawn from the United Network for Organ Sharing registry, these investigators meticulously correlated >450 clinical variables per patient with the dichotomous primary outcome of cardiac transplant rejection treated at the clinician's discretion by 1 year after transplant. After a laborious statistical process, they found 4 variables worthy of entry into a scoring system (range 0–13), including age at transplant, recipient sex, recipient race, and the number of donor-recipient HLA mismatches. This scoring system, which assigned numeric values to variables based on the magnitude …