Ann Thorac Surg 1998;65:1578-1579
© 1998 The Society of Thoracic Surgeons
Original articles: cardiovascular
Invited commentary
Jeffrey D. Hosenpud, MDa
a Registry Director, ISHLT/UNOS Joint Thoracic Registry, Division of Cardiovascular Medicine, Department of Medicine, Medical College of Wisconsin, 8700 W Wisconsin Ave, Milwaukee, WI 53226, USA
 |
Invited commentary
|
|---|
"Lies, damned lies and statistics"
—Benjamin Disraeli, British Prime Minister 1874–1880
McCarthy and colleagues present the Cleveland Clinic experience in cardiac transplantation and subject their series of patients to univariate and multivariate analysis looking for preoperative risk factors that predict outcome. In their series of 405 patients, they demonstrate excellent 1-year survival and fail to confirm several preoperative risk factors that have been demonstrated to be important in larger series and registry data. They suggest that one should rely on a given center’s experience when making decisions regarding donor and recipient acceptance criteria rather than using registry or multicenter data, which can be difficult to interpret. This obviously has important implications, not the least of which are issues surrounding informed consent. For example, should a patient destined to undergo transplantation with a 55-year-old donor be told that his or her transplant risk is higher based on the national data, or not informed of increased risk given a single center’s not being able to demonstrate a difference in survival outcome using older donors?
Before addressing that question, it is important to briefly review one of the sources of the data being compared, the Joint International Society for Heart and Transplantation/United Network for Organ Sharing Thoracic Registry, for which United States reporting is mandatory. The multivariate analysis published in the 1997 Registry report [1] contained verified data on a total of 13,695 heart transplantations in the United States, two thirds of which were performed from 1990 forward, the same time frame that was used in McCarthy and colleagues [1] more recent data. The analysis takes into account all of the major known risk factors for outcome including but not limited to year of transplantation, ventricular assist device use, retransplantation, center volume, and ischemic time. In contrast to other potentially soft end points used in various studies (rejection incidence, steroid use, coronary disease), in most analyses reported by the Registry, the end points used are not subject to interpretation (death). In summary, these data represent large numbers of patients representing the complete United States experience, the current transplant era, verified data, and hard end points.
The question then remains, is a given center "good enough" in its approach to transplantation to overcome risk factors documented in larger series? Or alternatively, are the numbers of patients who receive transplants in each center, even the largest centers, insufficient to detect differences in outcome when, in fact, they are present? It is important to emphasize that the inability to detect statistical differences is not the same as the lack of difference (ß statistical error). Taking nothing away from the superb transplant track record of the Cleveland Clinic, I believe the latter explanation to be the answer to the question. To illustrate, let us focus on one piece of data presented from the Cleveland Clinic series, donor age (see their Table 5). There is a progressive decrease in 1-year survival from almost 93% to 88% to 86% to 82% as the age of the donor increases through 49 years. The only aberrant group are the donors aged greater than 50 years (the smallest group, containing only 25 patients). If one looks at 5-year survival, the patients receiving hearts from donors less than 20 years of age enjoy an 86% survival rate, the remainder of patients receiving hearts less than 50 years old are in the 70% range, whereas those receiving hearts from donors greater than 50 years old have a survival rate of 60%. The Registry shows that increasing donor age is an independent, highly statistically significant (p < 0.0001) linear risk factor. This relationship between donor age and outcome is certainly suggested by the Cleveland Clinic data. The problem is the inability to detect this relationship because of small numbers.
McCarthy and colleagues have clearly demonstrated that a high-quality transplant program can "push the envelope" and have reasonable success in doing so. Moreover, there are substantial problems with multicenter registries, epidemiologic studies involving retrospectively captured data, and even multicenter prospective studies in which there are soft end points, patient and reporting biases, and multiple interrelated risk factors. Nonetheless, for certain diseases and procedures, heart transplantation being one of these, registry and multicenter data represent the only mechanism for obtaining substantial numbers of patients to generate meaningful statistical analyses.
|
References
|
|---|
- Hosenpud J.D., Bennett L.E., Keck B.M., Fiol B., Novick R.J. The Registry of the International Society for Heart and Lung Transplantation: fourteenth official report—1997. J Heart Lung Transplant 1997;16:691-712.[Medline]
Top
Invited commentary
References