Korean J Transplant.  2022 Nov;36(Supple 1):S197. 10.4285/ATW2022.F-3320.

The future of tacrolimus dosing: harnessing the potential of CURATE.AI for tacrolimus dose optimization: retrospective data analysis

• Gan T
• Tan SB
• Kumar KS
• Truong A
• Tan L
• Blasiak A
• Aw M
• Ho D
• Mali V

• KMID: 2535713
• DOI: http://doi.org/10.4285/ATW2022.F-3320

Abstract

Background
Living donor liver transplantation (LDLT) has become a gold standard treatment in pediatric end-stage liver disease. Tacrolimus forms the cornerstone of immunosuppression after pediatric LDLT. Standard-of-care for tacrolimus dose ti-tration is conventionally based on physician-guided drug dosing. This, however, leads to frequent deviations from target trough levels due to inter- and intra-patient variability, particularly during the critical early postoperative phase. Tacrolimus has a narrow therapeutic index and under or overexposure leads to clinically significant adverse effects. We explored the applicability of CURATE.AI, a small data, clinically validated artificial intelligence-derived platform, for guiding tacrolimus dosing towards achieving desired therapeutic levels.
Methods
This is a retrospective study of 16 pediatric LDLT recipients (13 males; median age, 2 years) at the National University Hospital Singapore from 2011–2018. Each patients' clinical data including tacrolimus dose and corresponding tacrolimus trough was used to generate a personalized CURATE.AI response profile that identifies and recommends an optimal dose to achieve the target treatment outcomes. CURATE.AI is both disease mechanism-independent and indication-agnostic and has dynamic ability to evolve with time. CURATE.AI's predictive performance was then evaluated with metrics that assessed both technical performance and clinical relevance.
Results
CURATE.AI-guided dosing fared better than standard-of-care physician-guided dosing in terms of percentage days within clinically acceptable tacrolimus levels of 6.5–12 ng/mL (54.55% vs. 49.08%). With CURATE.AI-guided dosing, patients could potentially achieve therapeutic range earlier (Fig. 1A) and better maintain therapeutic range with dynamic dose adjustments (Fig. 1B).
Conclusions
CURATE.AI was able to enhance the accuracy of tacrolimus dosing compared to unaided physician-guided deci-sions. Prospective studies may reveal its full potential as a clinical decision support system to balance tacrolimus dose optimization with drug-related toxicities.