INTRODUCTION
Liver transplantation (LT) is the optimal lifesaving operation for patients with end-stage liver failure[1]. It was achieved for the first time by Starzl in March 1963, and the exponential rate of medical discoveries have aided and facilitated this procedure, but barriers remain in exploring its full potential[2]. These include organ demand and allocation, donor-recipient matching, post-transplant complications prediction, personalization of immunosuppressive regimens and other issues deriving from the complexity and multidisciplinary nature of LT[3]. In recent years, technology has started to permeate the medical field, providing solutions to overcome such barriers but also facilitating research and education[4]. Innovations such as artificial intelligence (AI) driven learning platforms, electronic health records, medical literature summarization tools, virtual clinical skill simulators and remote healthcare applications like telepathology might prove to be of valuable assistance in such fields.
Large language models (LLM) are intricate AI systems, that can understand, generate and interact with human language[5]. Using a structure called transformers, they may process and learn from large amounts of data[6]. ChatGPT is a natural language generative AI, created through sophisticated tuning and modeling of LLM[7]. Since its initial release in November of 2022, ChatGPT has demonstrated significant user growth with 100 million users during the first 64 days, positioning it as one of the most rapidly growing applications in history[8]. This tool created by OpenAI as well as other LLMs have attracted significant interest from the medical community. Their applications range from support in diagnostic decision-making to assistance in various aspects of LT, including preoperative planning, intraoperative guidance and education of the surgeon[9].
ChatGPT has the potential to facilitate multiple tasks for transplant clinicians, researchers, educators, learners and patients, from enhancing organ allocation to the generation of personalized immunosuppressive medications. This mini review focuses on current and potential ChatGPT applications in LT medicine, with the goal of exploring its potential and defining its limitations to ultimately enhance the transplant process. We performed a search of the English language literature in PubMed, Scopus and ScienceDirect for ChatGPT applications in LT published between 2021 and 2025. The applications included areas of clinical practice, education and research.
CHATGPT APPLICATIONS IN LIVER TRANSPLANTATION
Since its invention in November 2022, it has rapidly gained popularity, with a rising role in the medical field, including solid-organ transplantation. Its applications span across various areas including decision-making support, data analysis and patient education. Figure 1 provides an overview of the current uses of ChatGPT in the field of LT.
Figure 1 provides an overview of the current uses of ChatGPT in the field of liver transplantation.
FAQS: Frequently asked questions. This Figure has been designed using resources from Flaticon.com.
Clinician-facing applications of ChatGPT
One well documented application of LLMs including ChatGPT is the usage of these platforms to facilitate patient record analysis and documentation[10]. Administrative tasks and documentation continue to consume a significant percentage of clinicians’ time, resulting in added work overload, burnout and reduced direct patient contact time[11]. It is reported that tasks outside of directed clinical time accumulate up to 39% of doctors time, with increased times in complicated cases such as LT[12]. LLMs have successfully tackled administrative tasks, particularly, Garcia et al[13] used two ChatGPT models for automated patient communication. Specifically, GPT-3.5 Turbo was used for the categorization of emails and GPT-4 for draft reply generation[13]. Overall, this intervention resulted in significant reduction of work exhaustion for the professionals in the field of hepatology and its utilization was welcomed in clinical settings[13].
Furthermore, the practicality in usage of ChatGPT has led to its emerging role as a clinical aid for healthcare professionals. In the context of LT, Christou et al[14] investigated the role of two ChatGPT models in providing reliable clinical recommendations. In their observational study they found that overall ChatGPT correctly answered 50.3%, with the latest more advanced version of ChatGPT-4 demonstrating higher performance (70.7%)[14]. Regarding the nature of the scenario, ChatGPT demonstrated higher agreement in questions regarding general information, with lower performance on questions regarding treatment and prognosis[14]. Malik et al[15] also assessed the applicability of ChatGPT-4 model as a clinical aid in the field of acute liver failure. In this study, ChatGPT exhibited great performance in terms of proficiency, clarity and relevance in its answers on clinical scenarios questions[15].
Patient-focused applications of ChatGPT
The successful outcome of LT is directly related to patient education, from acknowledging lifestyle changes to understanding warning signs[16]. To ensure this, guidelines from organizations such as the American Medical Association and the National Institutes of Health recommend that patient education materials be written at or below a 6th-grade reading level to maximize accessibility[17,18]. Optimal education depends on both patient variables, such as educational level, and provider variables, such as time spent on providing patient education, availability and empathy. Nowadays, there is great availability of health-related information online and LLMs have the potential to provide scientifically based and accurate information in a patient friendly language. Patients express various questions through the process of LT, starting from diagnosis of liver failure to years after their LT. The potential of ChatGPT in answering patient questions in the field of hepatology and LT has started to be explored, with a total of six published studies[19-25]. Yeo et al[19] were the first to study the potential role of ChatGPT in answering patient questions regarding Cirrhosis and hepatocellular carcinoma, two of the most common diagnoses leading to LT. The study revealed that ChatGPT could offer general information about those diseases. However, it struggled with pinpointing exact cut-off values or integrating guideline recommendations into its responses[19]. In addition, ChatGPT was also able to provide emotional support to patients and their family members[20]. In the same year, Endo et al[20] also assessed the quality of frequently asked questions by patients regarding LT. Experts in the field evaluated over 75% of the provided answers as “very good” or “excellent”[20]. Li et al[21] evaluated the usage of ChatGPT answering patients’ questions regarding cirrhosis and their study showed great performance regarding accuracy. Pradhan et al[22] used LLMs, including ChatGPT, to create patient information materials and compare them with human derived ones. In this study, ChatGPT outperformed the other 3 chatbots regarding readability of the provided answers based on objective score scales[22]. Daza et al[23] also compared ChatGPT performance to other LLMs and found that ChatGPT model achieved great results ranking 2nd in overall performance and 1st in quality of answers regarding clinical decision questions about four autoimmune liver diseases. Pugliese et al[24] selected 15 common questions about non-alcoholic fatty liver disease and posed them to ChatGPT. Their findings show great accuracy (mean score, 4.84 on a 6-point scale), comprehensiveness (mean score, 2.08 on a 3-point scale) and understandability (mean score, 2.87 on a 3-point scale)[24]. While the majority of studies evaluates ChatGPT performance in English Language, Alqahtani et al[25] evaluated the accuracy of ChatGPT in the creation of Arabic responses to patients with metabolic dysfunction-associated steatotic liver disease. Lastly, while the majority of the studies presented good results regarding the application of ChatGPT on the creation of patient educational materials, Cao et al[26] reported that ChatGPT provided inaccurate information about liver cancer surveillance and diagnosis.
Research-focused applications of ChatGPT
ChatGPT has numerous applications in medical research, from summarizing texts to providing research ideas. Particularly, Akabane et al[27] studied the potential of ChatGPT as a research tool in the field of LT. In this study, ChatGPT successfully generated a series of clinical questions related to LT, which after expert evaluation were identified as novel ideas, highlighting the potential role of ChatGPT in early stages of research[27]. ChatGPT’s ability to gather data from ongoing studies, synthesize information and even identify research gaps that may not yet be apparent to human investigators, ultimately supports a smoother and more consistent research workflow[27].
Education-facing applications of ChatGPT
LT is an operation requiring an efficient multidisciplinary approach. Doctors of medical and surgical specialties, nurses, occupational therapists, psychologists and other healthcare professionals are required for the long-term success of LT[28]. LLMs and particularly ChatGPT have been used for the creation of patient-education materials in many studies[19-26]. In a similar way, Christou et al[14] used two ChatGPT models as a clinical aid but also as a tool for the simulation of clinical scenarios. In this way, ChatGPT can provide knowledge but also test healthcare professionals with the method of simulation, providing sufficient feedback regarding their level of knowledge. Moreover, Malik et al[15] also used ChatGPT to create and answer clinical scenario simulations regarding acute liver failure disease. It is important to highlight that currently there is some controversy regarding the performance of ChatGPT in answering medically related questions. Particularly, when Knoedler et al[29] analyzed the performance of ChatGPT in answering USMLE Step 1 questions, they found overall accuracy rate of 55.8%, while Yaneva et al[30] in a similar study found that ChatGPT achieved scores greater than 60%, an indicator of meeting the passing standard. In regards to the field of LT, Suchman et al[31] evaluated ChatGPT’s performance on Multiple-Choice American College of Gastroenterology Self-Assessment Test. In this study, both ChatGPT models, ChatGPT-3.5 and ChatGPT-4, that were tested failed, reaching scores below the cut-off point of 70%[31].
FUTURE APPLICATIONS
Applications of LLMs have the potential to be applied in the creation of discharge summaries for transplant patients[32]. Particularly, ChatGPT successfully processed data deriving from text, summarized pertinent issues and clearly stated the follow-up plans in a patient-friendly language and format[33]. While ChatGPT is an LLM generating text answers, its latest versions can also analyze various forms of data information. Based on this function, ChatGPT-4 has been used successfully to create structured medical notes from audio recordings of doctor-patient encounters[34]. Furthermore, AI applications have already emerged in the field of digital pathology in LT, with studies reporting great sensitivity and specificity in diagnosis of liver pathologies, such as cirrhosis, in liver grafts from biopsies and phone images[9,35]. In the same context, LLM models with their ability for image analysis can offer real-time support to transplant pathologists[36]. Moreover, AI applications are dominating in the field of radiology, with existing applications regarding liver imaging already in use[37].
Another area where ChatGPT can be applied is in assisting surgeons during the LT operation. ChatGPT and other LLM have the potential to generate warnings and suggestions, as well as serve as a patient monitoring tool. Furthermore, implementation of a “talk to me” functionality can be beneficial particularly for surgeons-in-training enabling them to ask questions and receive information from the dataset that the system will be trained on[38].
In terms of patient education ChatGPT applications are not limited to question answers or creation of patient information leaflet but it can also generate material based on clinic letters provided by doctors into language patients understand, without loss of clinical information, increasing satisfaction and understanding[39]. This integration has increased potential in the field of LT, where good patient understanding is crucial for the long-term success of the transplantation. Moreover, it can aid LT recipients and their caregivers regarding safe-living and long-term care for their transplant[40].
Another important aspect is the potential use of ChatGPT in the field of research supporting non-native English speakers with academic writing. These individuals often face challenges in organizing logical flow and crafting key sections such as the introduction and discussion. Given their strong capabilities in summarizing and synthesizing text and data, LLMs like ChatGPT may serve as valuable tools in the writing process, ultimately promoting a more efficient and consistent global research output in the field of LT.
LIMITATIONS OF CHATGPT APPLICATION IN LT
Despite the strong potential ChatGPT shows in the field of LT, there are certain limitations and obstacles that restrict its broad application. Limitations and considerations include ethical and legislation issues as well as data accuracy liability[41,42].
Ethics and privacy concerns
Application of ChatGPT in the field of healthcare and particularly in decision making on complicated cases such as LT patients raises critical ethical and privacy considerations. Data privacy is one of the main issues arising from the application of ChatGPT and patient data entry. To ensure compliance with local regulation such as HIPAA in the United States or General Data Protection Regulation in Europe, healthcare professionals should refrain from sharing personal data when using ChatGPT, as well as to include end-to-end encryption, strict access controls and updated consent protocols[43]. There are also ethical issues related to the applications of LLMs, such as ChatGPT, in research, with some of the considerations including bias, spread of inaccurate information and plagiarism[44]. Specifically, the phenomenon of “artificial hallucination”, where the model generates inaccurate or non-existent references, has arisen, creating important considerations regarding the reliability of uncontrolled ChatGPT applications in research[45]. Additionally, legislation should be adapted to the use of ChatGPT by healthcare professionals. Specifically, clear guidelines should be established to outline their boundaries of application to govern their use[45].
Data accuracy
Data accuracy and liability remains one of the main limitations of ChatGPT in the field of healthcare. As previously mentioned, despite ChatGPT exhibiting great accuracy, this never reaches 100%, with lower performance particularly in difficult or complex questions[14,15,19-26]. Inaccurate data and limited training datasets may pose a potential risk to the LT process and the overall health of patients. Clinical decision-making in this field requires a high degree of precision to ensure that therapeutic strategies lead to the best possible patient outcomes. In this context, even subtle misinformation or misclassification can compromise critical aspects of the transplantation process. For instance, preoperative errors in the evaluation of a patient might lead to the incorrect clinical prioritization of patients and to a false candidate selection. Moreover, data used in the training sets of ChatGPT remain limited, with potential limitations in generalization of data interpretation, particularly regarding ethnical minority patients[41,42].
It is evident from current research that even though ChatGPT constitutes a valuable supportive tool for healthcare professionals, it cannot replace or outperform medical professionals and researchers. Likewise, although ChatGPT may offer a degree of emotional support or patient guidance, it would never be as efficient as the empathy, the ethical reasoning and the complex human understanding that are intrinsic to the patient–doctor relationship.
For the issues mentioned above to be addressed, healthcare providers and policymakers should aim to establish a controlled framework for the safe use of ChatGPT. Regular privacy impact assessments, data collection consent applications, strict assessment of ChatGPT accuracy, legislation related to AI usage in clinical settings and research are only few of the steps needed to be taken to design a dedicated environment for the responsible use of ChatGPT.
CONCLUSION
This is the first review, to the best of our knowledge, presenting and summarizing all the existing literature in the field of ChatGPT application in LT. Applications of LLMs in LT have the potential to revolutionize not only clinical practice but also research and education. ChatGPT applications are beneficial both for the clinicians, by reducing the workload by automating administrative tasks, but also for the patients, providing them accurate and understandable education material. Our review of the existing literature revealed a noticeable lack of comprehensive data on ChatGPT’s performance in addressing challenges that arise during the intraoperative and postoperative phases of LT. We also detected an absence of standardized evaluation frameworks and limited research addressing the potential risk of generating responses that lack patient comprehensibility. We strongly urge the research community to prioritize the development of standardized evaluation models and real-world clinical assessment. Future investigation should also focus on how equal access, patient-focused outcomes and the ethical integration of AI in critical decisions during the LT process can be ensured. Lastly, multidisciplinary collaboration is essential to overcome obstacles such as legislation limitation and lack of readability.
Despite the current obstacles such as ethical issues, data accuracy and legislation, healthcare professionals should embrace the applications of this powerful tool and use its technology to its full potential. With the recent progress of AI, LT can now be reinvented, facilitated and overall revolutionized.
Provenance and peer review: Invited article; Externally peer reviewed.
Peer-review model: Single blind
Specialty type: Transplantation
Country of origin: Greece
Peer-review report’s classification
Scientific Quality: Grade C, Grade D
Novelty: Grade B, Grade D
Creativity or Innovation: Grade B, Grade C
Scientific Significance: Grade C, Grade C
P-Reviewer: Correa TL, MD, Academic Fellow, Senior Postdoctoral Fellow, Senior Research Fellow, United States; Zhang N, MD, PhD, Postdoctoral Fellow, China S-Editor: Liu JH L-Editor: A P-Editor: Wang CH
