Description

\# Study Summary

\## Background

Kidney transplantation is currently the gold standard treatment for end-stage renal disease, with over 170,000 transplants performed each year worldwide. Despite major advances in short-term survival, long-term graft survival remains limited: approximately 40% of patients return to dialysis within 10 to 15 years following transplantation.

In recent years, artificial intelligence (AI) has emerged as a promising tool for predicting graft outcomes and supporting clinical decision-making. The iBox system, developed by the Paris Institute for Transplantation and Organ Regeneration, calculates a personalized probability of graft loss up to ten years after evaluation of the kidney transplant recipient.

This algorithm has been validated internationally across multiple cohorts and clinical trials, outperforms clinicians in predicting risk, and was qualified by the European Medicines Agency as an endpoint for clinical trials in 2022.

However, while the clinical value of the iBox and similar predictive tools is well documented, patient perceptions remain largely unexplored. Understanding how kidney transplant recipients perceive the prediction of their individual risk of graft loss (in terms of usefulness, acceptability, fears, or impact on their daily life), as well as their broader views on the role of artificial intelligence in medicine - including their hopes, expectations, and concerns about its deployment - is essential to ensure an ethical, transparent, and truly patient-centered implementation.

\## Study period

January 2026: Distribution of the questionnaire to the collaborating transplant centers and patient associations. February 2026: Data collection and management. April 2026: Analysis of results. June 2026: Presentation and discussion of results with the various participating centers/associations. September 2026: Publication.

\## Sample size

The questionnaire will be distributed to a sample of between 10,000 and 20,000 kidney transplant recipients. Based on response rates observed in comparable studies using patient questionnaires, an estimated response rate of 10% to 15% is expected, which would yield a sufficient volume of responses to ensure statistical robustness and diversity of represented profiles.

\## Study type

Cross-sectional, international, and multicenter survey, conducted via an anonymized online questionnaire. The study adopts a mixed-methods approach, combining qualitative and quantitative analysis of responses. No medical or biological data will be used, nor will data collected during clinical care be reused.

• Questionnaire development:\*\ The questionnaire was developed based on a review of the existing literature on patient perceptions of artificial intelligence and predictive medicine. An initial version of the questionnaire was developed and tested with four French patients. After adjustments, a bilingual translation was carried out by French-speaking and English-speaking collaborators. The final questionnaire was then shared with American patients for linguistic and cultural validation (\\questionnaire presented in the appendix of this document\\).
• Questionnaire distribution:\\ Distribution will be carried out through partner patient associations and collaborating transplant centers in France and the United States. The questionnaire will be distributed via a secure and anonymized REDCap platform, hosted on institutional servers compliant with data protection standards (GDPR). Participants will receive a unique link allowing them to access the online questionnaire directly. Participants will be able to respond at their own pace and discontinue their participation at any time without justification or consequence.
• Data analysis:\\* The collected responses will be analyzed using a quantitative approach. Qualitative analyses will rely on thematic analysis assisted by large language models (LLMs). These models will automatically extract recurring themes, associated emotions, and nuances of perception in the free-text responses, using a standardized methodology developed by the investigators. This type of LLM-based thematic analysis was already successfully conducted in the investigators' group. To ensure reliability and scientific rigor, a systematic manual verification will be performed on a representative sub-sample of responses. This independent review will compare human and automated coding and refine the thematic categories.
  • Data collected
  • Age
  • Sex
  • Year of last transplant
  • Country of last transplant
  • Transplant follow-up center
  • Transplant rank
  • Current occupation or last occupation held
  • Perceptions of the role of AI in medicine
  • Perceptions regarding a graft loss risk prediction system
  • Data flow

Data will be collected via the REDCap platform (HDS-certified hosting, GDPR-compliant). The questionnaire is anonymous: no nominative or identifiable medical data will be collected. Each participant will receive a unique link generating an untraceable alphanumeric identifier. Data will be transferred from REDCap and stored in encrypted form on institutional servers. Access will be strictly limited to the project investigators (Marc Raynaud, Alexandre Loupy) and the data manager (Thibaut Thalamas) via secure authentication. Data analyses will be performed on an internal environment using R. Fully anonymized datasets may be archived for secondary research or scientific replicability purposes.

\## Participant information

Participants will be contacted by email via their follow-up center, using contact databases already existing at these centers. The invitation message will briefly present the study and its sponsor, the Paris Institute for Transplantation and Organ Regeneration, as well as the responsible researchers. It will also contain an information notice outlining the study objectives, confidentiality procedures, and a direct link to the online questionnaire hosted on REDCap. Participation will be entirely voluntary, with no impact on medical follow-up or the relationship with the transplant team.