Year: 2003
Type: letter
Abstract: To the Editor: I read with interest the report by Kasiske et al. concerning an analysis of new-onset diabetes mellitus, based on data from the United Renal Data System (1Kasiske BL Snyder JJ Gilbertson D Matas AJ Diabetes mellitus after kidney transplantation in the United States.Am J Transplant. 2003; 3: 178-185Abstract Full Text Full Text PDF PubMed Scopus (1071) Google Scholar). The authors' analysis of risk factors for new-onset diabetes and its clinical implications provides valuable evidence on this important issue, and highlights the need for management strategies that minimize risk of diabetes developing after transplantation. Their results concerning the incidence of new-onset diabetes and the increased relative risk of diabetes associated with tacrolimus-based immunosuppression are broadly in line with our own findings (2Woodward RS Schnitzler MA Baty J et al.Incidence and cost of new onset diabetes mellitus among U.S. wait listed and transplanted renal allograft recipients.Am J Transplant. 2003; 3: 590-598Abstract Full Text Full Text PDF PubMed Scopus (336) Google Scholar). However, I believe that we must be cautious when interpreting the results concerning the likelihood of graft failure. For example, in Table 3, ‘Independent effects of post-transplant diabetes on graft failure’, the authors report that use of tacrolimus confers a significant 30% improvement in graft survival. However, by including both tacrolimus and new-onset diabetes in the model presented in Table 3, the analysis revealed the effect of tacrolimus on graft survival excluding the increased risk of graft loss resulting from tacrolimus-associated diabetes. The analysis, as performed, effectively removes from tacrolimus the adverse effect on graft survival it causes through increased risk of new-onset diabetes. This entangles the causal relationship between tacrolimus and new-onset diabetes into the estimated ‘benefit’ of tacrolimus. To expect the estimated 30% survival benefit from tacrolimus we must know with certainty that a patient will not develop new-onset diabetes. We can not know that future, and therefore can not directly interpret the estimated effect for practical use. This interpretation problem exists for any variable presented in Table 3 that is related to the development of new-onset diabetes, an intermediate outcome, as shown in Table 2. A more appropriate use of the results of Table 3 is to assess the impact of new-onset diabetes on graft survival accounting for other factors as a ‘robustness’ check. It must be clearly described that the estimated effects of the factors beyond new-onset diabetes are difficult to interpret, especially those correlated with the development of new-onset diabetes. Unfortunately, this interpretation was not described in the manuscript. To determine the independent effect of a factor such as tacrolimus, one should exclude intermediate outcomes from the model. Registries offer tremendous advantages in sample size and unprecedented views of real-world practice unavailable in strictly controlled clinical trials. However, misinterpretations like those made by Kasiske et al. (1Kasiske BL Snyder JJ Gilbertson D Matas AJ Diabetes mellitus after kidney transplantation in the United States.Am J Transplant. 2003; 3: 178-185Abstract Full Text Full Text PDF PubMed Scopus (1071) Google Scholar) are not uncommon. This, in part, results from the uncontrolled design inherent in a registry, which necessitates multivariate modelling. The causality problems seen here, as a result of the mixing of characteristics and intermediate outcomes as ‘independent’ effects, are just one of many mistakes that are easy to make. I do not wish to undermine the usefulness of the work presented by Kasiske et al. (1Kasiske BL Snyder JJ Gilbertson D Matas AJ Diabetes mellitus after kidney transplantation in the United States.Am J Transplant. 2003; 3: 178-185Abstract Full Text Full Text PDF PubMed Scopus (1071) Google Scholar) and others who have and will continue use registries. However, the transplant community must be prepared to interpret registry analyses as pieces of evidence and not definitive statements, incorporating registry findings into practice with care.
Author: Mark Schnitzler
Institution: Washington University in St. Louis
Language: English
Subfield: Transplantation
Field: Medicine
Domain: Health Sciences
Open Access status: hybrid
Estimated APC paid: $4,300
OpenAlex ID: W2071921294