What happened
The Paris Transplant Group built iBox, a score that predicts whether a transplanted kidney will fail, from data on 4,000 patients at four French centers. The Transplant Therapeutics Consortium, which C-Path formed with two US transplant societies, asked the European Medicines Agency (EMA) to qualify it for drug trials. EMA's medicines committee adopted its opinion on 15 December 2022. It qualified the full and abbreviated scores as a secondary endpoint prognostic for graft loss in trials of new immunosuppressive medicines. In four external validation datasets, every c-statistic was 0.70 or greater. The committee found that the models predict individual graft loss with good performance. It did not qualify iBox as a surrogate primary endpoint, because too few trials and events were available. In 2025, Catalonia's assessment agency AQuAS reviewed iBox for Spain's network of assessment agencies. In the one study it found, iBox predicted seven-year graft survival better than nine transplant physicians did. Its discrimination was 0.789, against 0.638 to 0.786 for the physicians. The group that built iBox ran that study on the records of 400 past patients.
Sources
-
Peer-reviewed study, by the makers.
Divard G, Raynaud M, Tatapudi VS, et al. Comparison of artificial intelligence and human-based prediction and stratification of the risk of long-term kidney allograft failure. Communications Medicine 2022, 2:150. doi:10.1038/s43856-022-00201-9. Written by the Paris Transplant Group, which developed iBox. One author declares shares in Cibiltech.
The iBox system demonstrates the best predictive performance with a discrimination of 0.79 and a median calibration error of 5.79%, while physicians tend to overestimate the risk of graft failure.
-
Regulatory document, by authors with no tie to the system.
European Medicines Agency, Committee for Medicinal Products for Human Use (CHMP). Qualification opinion for the iBox Scoring System as a secondary efficacy endpoint in clinical trials investigating novel immunosuppressive medicines in kidney transplant patients. EMADOC-1700519818-946771, 19 December 2022, adopted by CHMP on 15 December 2022. Section 1, CHMP qualification opinion statement.
CHMP qualifies the iBox Scoring System (Composite Biomarker Panel) as a secondary endpoint prognostic for death-censored allograft loss (allograft failure) in kidney transplant recipients to be used in clinical trials to support the evaluation of novel immunosuppressive therapy applications.
-
Assessment report, by authors with no tie to the system.
Agència de Qualitat i Avaluació Sanitàries de Catalunya (AQuAS), for the Red Española de Agencias de Evaluación de Tecnologías Sanitarias y Prestaciones del SNS. Prieto Durán I, Moltó-Puigmartí C, Mestre Lleixà B, Ibeas J, Estrada Sabadell MD, Vivanco-Hidalgo RM. Algoritmo predictivo para monitorizar y evaluar el impacto de los tratamientos en trasplante de riñón. Ficha de evaluación de tecnologías nuevas y emergentes. Madrid: Ministerio de Sanidad, 2025. DOI 10.62727/DSalut.AQUAS/13303. Puntos clave.
concluye que el algoritmo predice de manera más realista y homogénea el pronóstico de supervivencia del aloinjerto renal, en comparación con las predicciones de los profesionales sanitarios, que son más heterogéneas y menos fiables.
Details
AI system
- Reported configuration
-
Full and abbreviated models
- Workflow
-
IBox combines kidney function (eGFR), proteinuria, donor-specific antibodies, biopsy findings and the time since transplant, and the abbreviated score leaves out the biopsy. For trials, the score is measured one year after transplant to predict death-censored graft loss at five years. In the study that AQuAS reviewed, iBox and nine physicians worked from the same anonymized records of 400 past patients.
- Relation to AI
-
IBox is a statistical prediction model. EMA describes a multivariable Cox proportional hazards model derived from prospectively collected data on 4,000 patients, with variables selected from 31 candidates and 8 retained in the full score. AQuAS calls it a classical statistical model built on Cox regression coefficients. Its marketer calls it AI, and AQuAS notes that no publication makes clear whether it has since gained more advanced AI elements. The comparison with physicians isolates the model's predictions.
Follow-up
- Conditions for replication
-
Trial sponsors use iBox as a secondary endpoint, since EMA did not qualify it as a surrogate primary endpoint. The required measurements are available between 6 and 24 months after transplant. Physicians use the prediction as support alongside their own clinical judgment.
Entry details
- Sector
-
Healthcare (kidney transplantation).
- Purpose
-
The Paris Transplant Group built iBox to predict the risk that a transplanted kidney will fail within three, five or seven years from routine clinical, immunological and biopsy data.
References
- Other identifiers
-
Clinicaltrials_gov: NCT04918199.
28 more questions not yet checked
model, provider, role, configuration, inputs and outputs, autonomy, human oversight, beneficiaries, reach, distribution and access, duration, every measurement question, every money question, ownership and control, dependencies, cross-border connections, geopolitical implications, replication attempts, availability, institutional decision, grounds, action and result, success criterion, every evidence question, observed or intended, alternative assessments, related entries, category mappings.
Cite this record
Peter H. Vartanian. (2026). EMA backs the iBox score to forecast kidney graft loss in trials (AIBD-0927). Artificial Intelligence Benefits Database, edition of 1 November 2026. https://petervartanian.xyz/aibd/records/AIBD-0927.html (archived at https://doi.org/10.7910/DVN/AADSII, version 7.0)