Researchers at The Johns Hopkins Medicine say they have developed a new electronic medical records-based tool that should help doctors predict which patients are most at risk of losing a transplanted kidney graft.
The study, which was federally funded by the National Institutes of Health and published in the Oct. 6 edition of the Clinical Journal of the American Society of Nephrology, introduces a dynamic risk prediction model that uses routine lab results — specifically, changes in kidney function over time — to predict whether a transplanted kidney (graft) will fail within three years after surgery.
With kidney disease on the rise with an estimated 15% of adults in the United States having chronic kidney disease (CKD), the need to reduce the chances of patients progressing to end-stage kidney disease (ESKD) has become increasingly important. Kidney transplantation has been seen as the ideal treatment for ESKD as it offers longer survival and better quality of life compared with dialysis treatments.
Although a typical successful kidney transplant should last about 10 years, a fourth of the grafts can fail in the first five years after transplantation. This further illustrates the importance of identifying kidney transplant recipients at risk for graft deterioration to better optimize the outcomes in the long-term for patients who have undergone kidney transplants.
Researchers believe that screening kidney transplant recipients at high risk for allograft failure could enable counseling and potential therapeutic options to prevent progression. Researchers hope that identifying allografts at risk of failing would potentially allow for timely interventions, such as more frequent follow-up with closer monitoring for allograft injury and its causes, modified immunosuppression, and counseling patients about the need for another transplant or the emotional burden of reaching ESKD again.
Conversely, patients with low risk for graft failure can be discharged from the care of their transplant nephrologist and transferred back to their primary nephrologist — frequently closer to home — for continued care and open capacity at the transplant center for other patients in need of transplantation care.
After a kidney transplant, doctors closely monitor how well the organ is working, often using a lab measure called the estimated glomerular filtration rate (eGFR). The new model that researchers used in the study aimed to continuously update a patient's risk of graft failure every time a new eGFR result is measured, allowing for real-time, personalized risk assessment.
To develop the new tool, the researchers analyzed data recorded for 1,114 deceased donor kidney transplant recipients from three registries — the OPTN registry, the Johns Hopkins EMR cohort and the Columbia EMR cohort that combined consisted of roughly 80,000 deceased-donor kidney transplant recipients. Specifically, they looked at repeated eGFR results — a blood test used routinely to determine how well a kidney is working to filter out toxins.
"We developed the prediction model in the Deceased Donor Study, an NIH-funded observational research study," says Heather Thiessen Philbrook, M.Math, assistant director of the Kidney Precision Medicine Center of Excellence at Johns Hopkins Medicine and the study's primary author. The study is a rich data source on deceased-donor kidney transplant recipients with a median of 12 follow-up eGFR measurements within the first three years post-transplant. We validated the model across the broad cohort captured in the U.S. transplant registry and within two real-world data sets leveraging data available in electronic medical record systems
Two stages of modeling were used during the study. One model was the linear mixed-effects model which estimated each patient's eGFR trajectory over time to see how it matched up with graft failure. Graft failure was defined as a return to dialysis or a second transplant within three years of the first procedure. The second was a logistic model predicting graft failure that used the recipient's eGFR trajectories estimated from the first stage linear mixed model.
For all of the validation cohorts used in this study, adult recipients of primary deceased donor kidneys with at least one post-transplant serum creatinine measurement — a measure of how well the kidneys are doing their job of filtering waste from the blood — were included.
Overall, the results of the study showed that the two-stage approach allowed for an efficient estimation of individual eGFR trends and flexible modeling of the association between eGFR and graft failure. Three months after transplant, the model achieved a predictive accuracy of 0.70 and 30 months after transplant, the model achieved a predictive accuracy of 0.90, meaning it was able to distinguish between high and low risk patients.
"The results from this study can be readily implemented at transplant centers to streamline care of the recipients of kidney transplant by providing updated risk predictions as new data became available," says Chirag Parikh, M.D., Ph.D., director of the Division of Nephrology, director of the Kidney Precision Medicine Center of Excellence at Johns Hopkins Medicine and the study's senior author. "There could also be future downstream models developed to predict other infections and immunologic complications."
With the promising results of the study, researchers plan to test the tool in everyday clinical settings and explore additional health data, such as clinical events and other laboratory measurements, to optimize model performance in predicting early graft failure.
Funding for this study was supported by the National Institute of Diabetes and Digestive and Kidney Diseases and the Edward S. Kraus Scholar Award.
No authors had any conflict of interests.
