Fig 1.
Schematic illustration of the variable-ranking process for the derivation cohort.
Panel (A). Step I: training of the survival model using fivefold cross-validation and a committee machine with 10 members. Step II: variable-ranking using the trained survival model from step I. Each variable is omitted from the model and the decrease in performance is recorded. The variable resulting in the least reduction in performance is removed. Steps I–II are repeated until only one variable is left. A ranking list is constructed using the elimination order. Panel (B). Performance as a function of number of variables included. The C-index is plotted against the number of input variables, where order is in terms of decreasing importance. Variables with a high index number are least important. Panel (C). The relative importance of 20 of the 43 top-ranked variables. The box plot for each variable is created from the series of C-index changes that was the result of removal of the variable during the ranking procedure.
Table 1.
Baseline characteristics of recipients in the derivation and validation cohorts.
Table 2.
Baseline characteristics of donors in the derivation and validation cohorts.
Table 3.
Time-dependent hazard ratios for the 32 recipient risk variables included in the IHTSA model.
Table 4.
Time-dependent hazard ratios for the 11 donor risk variables included in the IHTSA model.
Fig 2.
Cumulative mortality for the internal validation cohort (IVC).
The black solid line shows the observed cumulative mortality and dotted lines show the 95% confidence interval (estimated with Kaplan-Meier failure function) in the IVC. The red solid line shows the predicted survival for transplanted patients in the IVC.
Table 5.
Comparison of C-index between different risk models used to predict overall survival.
Table 6.
Comparison of AUROCs between different risk models used to predict one- year mortality.
Fig 3.
Influence of age and duration of ischemia on survival.
Panel (A) illustrates, in matched cohorts, the influence of donor age and recipient age on survival when the duration of ischemia is fixed to 2 hours, and panels B, C, and D illustrate the influence on survival at 3, 4, and 5 hours of ischemia, respectively. The colored solid lines show the predicted survival in years (dark blue: worst; dark red: best).
Fig 4.
Influence of weight and gender match.
Panel (A) illustrates, in matched cohorts, the influence of donor weight and recipient weight on survival for male-male donor-recipient pairs, and panels B, C, and D illustrate the influence on survival for female-male, female-female, and male-female donor-recipient pairs, respectively. The colored solid lines show the predicted survival in years (dark blue: worst; dark red: best).
Fig 5.
Relative importance for the IHTSA variables.
The numbers within the nodes and leaves represent the predicted median survival together with standard deviation (SD) in years (y), and the number of patients the results is based on (n). The decision variable and the criterion are shown on the dotted lines between the nodes. The most important variable is the top node. Light blue coloring of the nodes shows an internal node and red represents a leaf node. Dark blue, dark red, yellow, and green coloring represent the four leaf nodes for which the observed cumulative mortality is plotted against time in Fig. 6. The tree has been pruned to a cost-complexity parameter of 0.0035. Cong, congenital; ICM, ischemic cardiomyopathy; NICM, non-ischemic cardiomyopathy; GF, graft failure; Valv, valvular; Oth, other.
Fig 6.
Validation of the IHTSA variable decision tree.
This graph illustrates the difference in observed cumulative mortality for patients belonging to leaf nodes 16, 23, 24, and 127 in the decision tree. The line color corresponds to the leaf node color in Fig. 5. The solid lines show the observed cumulative mortality for transplanted patients in the derivation cohort and dashed lines show the observed cumulative mortality for transplanted patients in the internal validation cohort (estimated with Kaplan-Meier failure function).
Fig 7.
The International Heart Transplantation Survival Algorithm (IHTSA) as a web application.
The IHTSA has been implemented as an interactive program that estimates median, 1-, 5-, and 10-year survival, and the benefit of adding (or removing) properties from an individual recipient and the potential donor (http://www.ihtsa.med.lu.se). ECMO, extracorporeal membrane oxygenation; HLA, human leukocyte antigen; PRA, panel reactive antibody; PVR, pulmonary vascular resistance; SPP, systolic pulmonary pressure. †Drug treated systemic hypertension. ‡Infection requiring intra venous antibiotic therapy within two weeks prior to transplant. *Previous transplant——previous kidney, liver, pancreas, pancreas islet cells, heart, lung, intestine and/or bone marrow transplant.
Fig 8.
Predicted cumulative mortality for different organ allocation models.
In panel (A) the graph shows the number of transplanted patients for The IHTSA model (red lines), Clinical model (blue lines) and Control (green lines) influenced by the size of the waiting list. The internal validation cohort (IVC) is presented with solid lines, temporal validation cohort (TVC) dashed lines and external validation cohort NTTD dotted lines. In panel (B) the graph shows the difference in predicted cumulative mortality as a function of time since heart transplantation for a waiting list including 50 patients in the IVC (N = 8,569). The solid black lines present the observed mortality and the dotted lines the predicted mortality for all patients. Panel (C) shows the difference in predicted cumulative mortality for IHTSA model influenced by the number of patients on the waiting list (NW). In panel (D) the results from the sub analysis including patients from the IVC, who were not treated in the intensive care unit and were not on life support prior to transplantation, are presented (N = 4,868).