The Annals of Thoracic Surgery
Volume 111, Issue 6, June 2021, Pages 2033-2040
Laura Seese MD, MS, Victor O. Morell MD, Melita Viegas MD, Mary Keebler MD, Gavin Hickey MD, Yisi Wang MPH, Arman Kilic MD
The United Network for Organ Sharing registry identified patients with ACHD (≥18 years of age) who underwent OHT between 1987 and 2018. The primary outcome was 1-year mortality. Associated covariates (univariate P < .2) were entered into a multivariable logistic regression model. Variable inclusion in the model was assessed by improvement in the McFadden pseudo-R2, likelihood ratio test, and C-index. A risk score was created using the absolute magnitude of the odds ratios from the derivation cohort, and its ability to predict 1-year mortality was tested in the validation cohort.
A total of 1388 recipients were randomly divided into derivation (66.7%, n = 950) and validation (33.3%, n = 438) cohorts. A 13-point risk score incorporating 4 pretransplant variables (age, dialysis dependence, serum bilirubin level, and mechanical ventilation) was created. The predicted 1-year mortality ranged from 14.6% (0 points) to 49.9% (13 points) (P < .001). In weighted regression analysis, there was a strong correlation between predicted 1-year mortality and observed 1-year mortality in the validation cohort (r = 0.85, P < .001). Logistic regression also demonstrated a significant association (odds ratio, 1.18; 95% confidence interval, 1.1-1.3; P = .004). The Brier score of the composite score in the validation cohort was 0.14. Kaplan-Meier analysis demonstrated that risk scores of 4 points or higher portended worse survival at 1-year posttransplant (P < .001).
This 13-point risk score for ACHD is predictive of mortality within 1 year after OHT and has potential utilization in improving recipient selection for OHT in adult patients with CHD.
Seese and coworkers’ 13-point risk score for predicting 1-year mortality after orthotopic heart transplantation (OHT) has potential utility in improving recipient selection for
OHT in adult patients with congenital heart defects (CHD). Current outcome trends in treating children born with CHD suggest the adult CHD (ACHD) population needing OHT will grow substantially in the coming decades. Despite the effectiveness of OHT, its availability is limited making optimal selection a crucial issue. A risk score that can predict the likelihood of survival beyond one year is thus helpful and likely to become essential in the next few decades. But can a risk score really predict the likelihood of mortality within a year of OHT, or indeed determine which ACHD potential recipient will benefit the most from a donor heart? For such risk scores to be clinically beneficial, they must be able to identify which ACHD potential recipients are at increased risk of death within a year of OHT, be able to assign the appropriate weight to the individual risk factors of mortality, and then help stratify that risk to determine who will benefit the most (in this case survive the longest) from an available donor heart. In the current study, the predicted rate of 1-year mortality ranged from 14.5% (0 points) to 49.9% (13 points) [P < .0001] suggesting a fairly good performance of the risk score. Among the key determinants of performance of a risk estimation system are the discriminatory power and calibration. The discriminatory power of the scoring system is the ability of the system to distinguish those who will develop the end point (mortality within a year after OHT in this case) from those who will not. On this point the current system has a span of 35.4% (range of mortality between 14.5% and 49.9%). Then comes calibration potential – the measure of how closely predicted outcomes agree with actual outcomes. In this case, the study reports that the weighted regression analysis showed a strong correlation between predicted 1-year mortality and observed 1-year mortality in the validation cohort (r = 0.85, P < .001). We must bear in mind however that the method of validation used in this study was internal – based on the same dataset from which the derivation and validation cohorts were drawn. Such internal validation has important limitations – the baseline survival of the whole population is included in the derivation, and identical risk factor definitions are used in the validation and derivation datasets. Validation using an externally derived dataset would be more secure. Furthermore, the authors have also pointed out limitations in the dataset that restrict prediction of mortality with regards to specific CHD diagnoses and prior operations, as well as the effect of temporal changes in OHT allocation policy, data recording, or center volume. These limitations imply that the model must be used with caution. That notwithstanding, the study is an important contribution to our ability to predict ACHD potential recipients who are likely to be the beneficiaries of the so called ‘survival paradox’ – high early mortality risk after OHT in ACHD but superior long-term survival in those who survive the early hazard phase.