Simplifying Diagnosis of Rejection After Kidney Transplant

An automated system for the diagnosis of rejection after kidney transplant simplifies the assessment of increasingly complex criteria, significantly reducing the common occurrence of misclassification and potentially improving outcomes, according to new research.

“To date, no study in transplantation and other medical fields has developed and validated an automated multimodal disease classification,” report the French authors in their article published online May 4 in Nature Medicine.

“In this study, we built a kidney allograft diagnostic automation system and demonstrated its ability to overall avoid 40% of misdiagnoses of allograft rejection and improve patient risk stratification in adult and pediatric kidney transplant recipients,” they observe.

First author Daniel Yoo, a PhD student at the Paris Institute for Transplantation and Organ Regeneration, France, said in a press statement that the system could be compared to “a ChatGPT specialized for rejections.”

“We have developed an intelligent and very user-friendly system,” he said. “Doctors can obtain a correct diagnosis for their patients with just a few clicks.”

The online, user-friendly system, which has been validated by international transplant societies including the American Society of Transplantation and European Society for Organ Transplantation, has been made available for free to the medical community as an open-access application, which will be updated regularly.

Asked to comment, Andrew Malone, MBBCh, an associate professor of medicine in transplant nephrology at Washington University in St. Louis, Missouri, said the study underscores the complexities and confusion that exist without the automation.

“At a philosophical level, the [take-home message] is that we don’t fully understand what rejection is or how to diagnose/classify it and its various forms,” he told Medscape Medical News.

Further commenting, Roslyn B. Mannon, MD, of the University of Nebraska Medical Center, in Omaha, noted that “a major challenge [in transplant medicine] is borderline rejection. These [cases] may or may not be treated and are associated with long-term graft loss.”

“This [new automated] system actually classified 57% of the borderlines to a higher grade of rejection, and that would more than likely have triggered additional immunosuppressive treatments,” she told Medscape Medical News.

Overall, she noted that “this is a well-done paper, innovative technology, and an opportunity to improve patient care. [The results are] very exciting for the future, and in considering technology as an important piece of personalized care management.”

System Highlights Errors Made by Pathologists

The Banff classification system has been used for decades to diagnose kidney allograft rejection, recently incorporating rapidly advancing technologies and precision medicine processes to improve accuracy.

However, with the advances, the system has become highly complex, using histological, immunological, and other data, leading to substantial observer variability, and ultimately, misclassification, potentially compromising outcomes.

As a result, international transplant societies have underscored an urgent need for a more reliable and reproducible system to help predict rejection.

In response, Yoo and colleagues set out to develop a more simplified system, first conducting a compilation of all published and up-to-date Banff rules, decoding them, and translating the data into a coded, automated algorithm covering all possible diagnostic scenarios, called the Banff Automation System.

To analyze the new system’s ability to reclassify rejection diagnoses, the algorithm was tested in adult and pediatric kidney transplant recipients in three international multicentric cohorts and two large prospective clinical trials.

The trials included 4409 biopsies from 3054 patients followed at 20 transplant referral centers in Europe and North America. Overall, 62.1% of patients were male.

“The system identified rejections not identified by pathologists, which could have benefited from a treatment, and also nonrejection cases, initially identified as rejections, which should have not been treated by physicians,” senior author Alexandre Loupy, MD, PhD, also of the Paris Institute for Transplantation and Organ Regeneration, told Medscape Medical News.

In the adult kidney transplant population, use of the Banff Automation System resulted in reclassification of 83 of 279 (29.8%) cases as antibody-mediated rejection, and 57 of 105 (54.3%) cases as T cell-mediated rejection.

In addition, 237 of 3239 (7.3%) biopsies diagnosed by pathologists as nonrejection were reclassified with the automated system as rejection.

These results were confirmed in the pediatric population, with 8 of 26 (30.8%) cases reclassified as antibody-mediated rejection and 12 of 39 (30.8%) reclassified as T cell-mediated rejection.

Clinical Factors Beyond Scope of Algorithm May Be Considered by Pathologists

To evaluate the clinical outcomes of the reassignments, an analysis of adult and pediatric transplant recipients showed that 144 (6.6%) lost their graft after a median follow-up of 2.7 years post-biopsy.

Importantly, patients diagnosed by a pathologist as nonrejection but reclassified by the Banff Automation System as rejection showed significantly worse graft survival compared with patients without rejection (hazard ratio, 6.4; P < .0001), suggesting that the rejection diagnosis from the automated system was indeed correct.

In addition, those diagnosed by a pathologist as rejection but reclassified by the automated system as nonrejection showed excellent graft survival, similar to that of patients without rejection (P = .754).

While the automated system works to prevent the main causes of pathologist misclassification, clinical factors beyond its scope may nevertheless ultimately override decision-making, Loupy underscored.

“It is important to note that this tool does not consider the patient’s clinical history, which can sometimes cause the pathologist to decide on a rejection diagnosis without meeting all the criteria of the Banff classification,” he explained.

Malone noted: “From a pragmatic point of view, this manuscript tells us that there is still huge variation in the accuracy of…or fidelity to the use of the Banff system by pathologists when making transplant biopsy diagnoses.”

That being said, “our understanding of the pathobiology of rejection, in particular antibody-mediated rejection, continues to evolve. Therefore, our approach to diagnosis will continue to change and hopefully improve. Automated systems such as this are also useful for consistency in defining outcome measures in clinical trials,” Malone concluded.

The authors and Malone have reported no relevant financial relationships. Mannon is a trustee of the Banff Allograft Pathology Foundation.

Nature Med. Published online May 4, 2023. Abstract

For more news, follow Medscape on Facebook, Twitter, Instagram, YouTube, and LinkedIn.

Source: Read Full Article