Miao, Jing

Abstract Objectives ChatGPT (OpenAI, San Francisco, CA) has shown impressive results across various medical examinations, but its performance in kidney pathology is not yet established. This study evaluated proficiencies of GPT-4 Vision (GPT-4V), an updated version of the platform with the ability to analyze images, on kidney pathology questions and compared its responses with those of nephrology trainees. Methods Thirty-nine questions (19 text-based questions and 20 with various kidney biopsy images) designed specifically for the training of nephrology fellows were employed. Results GPT-4V displayed comparable accuracy rates in the first and second runs (67% and 72%, respectively, P = .50). The aggregated accuracy, however—particularly, the consistent accuracy—of GPT-4V was lower than that of trainees (72% and 67% vs 79%). Both GPT-4V and trainees displayed comparable accuracy in responding to image-based and text-only questions (55% vs 79% and 81% vs 78%, P = .11 and P = .67, respectively). The consistent accuracy in image-based, directly asked questions for GPT-4V was 29%, much lower than its 88% consistency on text-only, directly asked questions (P = .02). In contrast, trainees maintained similar accuracy in directly asked image-based and text-based questions (80% vs 77%, P = .65). Although the aggregated accuracy for correctly interpreting images was 69%, the consistent accuracy across both runs was only 39%. The accuracy of GPT-4V in answering questions with correct image interpretation was significantly higher than for questions with incorrect image interpretation (100% vs 0% and 100% vs 33% for the first and second runs, P = .001 and P = .02, respectively). Conclusions The performance of GPT-4V in handling kidney pathology questions, especially those including images, is limited. There is a notable need for enhancement in GPT-4V proficiency in interpreting images.

The integration of large language models (LLMs) into healthcare, particularly in nephrology, represents a significant advancement in applying advanced technology to patient care, medical research, and education. These advanced models have progressed from simple text processors to tools capable of deep language understanding, offering innovative ways to handle health-related data, thus improving medical practice efficiency and effectiveness. A significant challenge in medical applications of LLMs is their imperfect accuracy and/or tendency to produce hallucinations—outputs that are factually incorrect or irrelevant. This issue is particularly critical in healthcare, where precision is essential, as inaccuracies can undermine the reliability of these models in crucial decision-making processes. To overcome these challenges, various strategies have been developed. One such strategy is prompt engineering, like the chain-of-thought approach, which directs LLMs towards more accurate responses by breaking down the problem into intermediate steps or reasoning sequences. Another one is the retrieval-augmented generation (RAG) strategy, which helps address hallucinations by integrating external data, enhancing output accuracy and relevance. Hence, RAG is favored for tasks requiring up-to-date, comprehensive information, such as in clinical decision making or educational applications. In this article, we showcase the creation of a specialized ChatGPT model integrated with a RAG system, tailored to align with the KDIGO 2023 guidelines for chronic kidney disease. This example demonstrates its potential in providing specialized, accurate medical advice, marking a step towards more reliable and efficient nephrology practices.

The accurate interpretation of CRRT machine alarms is crucial in the intensive care setting. ChatGPT, with its advanced natural language processing capabilities, has emerged as a tool that is evolving and advancing in its ability to assist with healthcare information. This study is designed to evaluate the accuracy of the ChatGPT-3.5 and ChatGPT-4 models in addressing queries related to CRRT alarm troubleshooting. This study consisted of two rounds of ChatGPT-3.5 and ChatGPT-4 responses to address 50 CRRT machine alarm questions that were carefully selected by two nephrologists in intensive care. Accuracy was determined by comparing the model responses to predetermined answer keys provided by critical care nephrologists, and consistency was determined by comparing outcomes across the two rounds. The accuracy rate of ChatGPT-3.5 was 86% and 84%, while the accuracy rate of ChatGPT-4 was 90% and 94% in the first and second rounds, respectively. The agreement between the first and second rounds of ChatGPT-3.5 was 84% with a Kappa statistic of 0.78, while the agreement of ChatGPT-4 was 92% with a Kappa statistic of 0.88. Although ChatGPT-4 tended to provide more accurate and consistent responses than ChatGPT-3.5, there was no statistically significant difference between the accuracy and agreement rate between ChatGPT-3.5 and -4. ChatGPT-4 had higher accuracy and consistency but did not achieve statistical significance. While these findings are encouraging, there is still potential for further development to achieve even greater reliability. This advancement is essential for ensuring the highest-quality patient care and safety standards in managing CRRT machine-related issues.

ABSTRACT Background Evidence supporting glucagon-like peptide-1 receptor agonists (GLP-1RAs) in kidney transplant recipients (KTRs) remains scarce. This systematic review and meta-analysis aims to evaluate the safety and efficacy of GLP-1RAs in this population. Methods A comprehensive literature search was conducted in the MEDLINE, Embase and Cochrane databases from inception through May 2023. Clinical trials and observational studies that reported on the safety or efficacy outcomes of GLP-1RAs in adult KTRs were included. Kidney graft function, glycaemic and metabolic parameters, weight, cardiovascular outcomes and adverse events were evaluated. Outcome measures used for analysis included pooled odds ratios (ORs) with 95% confidence intervals (CIs) for dichotomous outcomes and standardized mean difference (SMD) or mean difference (MD) with 95% CI for continuous outcomes. The protocol was registered in the International Prospective Register of Systematic Reviews (CRD 42023426190). Results Nine cohort studies with a total of 338 KTRs were included. The median follow-up was 12 months (interquartile range 6–23). While treatment with GLP-1RAs did not yield a significant change in estimated glomerular filtration rate [SMD −0.07 ml/min/1.73 m2 (95% CI −0.64–0.50)] or creatinine [SMD −0.08 mg/dl (95% CI −0.44–0.28)], they were associated with a significant decrease in urine protein:creatinine ratio [SMD −0.47 (95% CI −0.77 to −0.18)] and haemoglobin A1c levels [MD −0.85% (95% CI −1.41 to −0.28)]. Total daily insulin dose, weight and body mass index also decreased significantly. Tacrolimus levels remained stable [MD −0.43 ng/ml (95% CI −0.99 to 0.13)]. Side effects were primarily nausea and vomiting (17.6%), diarrhoea (7.6%) and injection site pain (5.4%). Conclusions GLP-1RAs are effective in reducing proteinuria, improving glycaemic control and supporting weight loss in KTRs, without altering tacrolimus levels. Gastrointestinal symptoms are the main side effects.

Accurate information regarding oxalate levels in foods is essential for managing patients with hyperoxaluria, oxalate nephropathy, or those susceptible to calcium oxalate stones. This study aimed to assess the reliability of chatbots in categorizing foods based on their oxalate content. We assessed the accuracy of ChatGPT-3.5, ChatGPT-4, Bard AI, and Bing Chat to classify dietary oxalate content per serving into low (<5 mg), moderate (5–8 mg), and high (>8 mg) oxalate content categories. A total of 539 food items were processed through each chatbot. The accuracy was compared between chatbots and stratified by dietary oxalate content categories. Bard AI had the highest accuracy of 84%, followed by Bing (60%), GPT-4 (52%), and GPT-3.5 (49%) (p < 0.001). There was a significant pairwise difference between chatbots, except between GPT-4 and GPT-3.5 (p = 0.30). The accuracy of all the chatbots decreased with a higher degree of dietary oxalate content categories but Bard remained having the highest accuracy, regardless of dietary oxalate content categories. There was considerable variation in the accuracy of AI chatbots for classifying dietary oxalate content. Bard AI consistently showed the highest accuracy, followed by Bing Chat, GPT-4, and GPT-3.5. These results underline the potential of AI in dietary management for at-risk patient groups and the need for enhancements in chatbot algorithms for clinical accuracy.

Chain-of-thought prompting enhances the abilities of large language models (LLMs) significantly. It not only makes these models more specific and context-aware but also impacts the wider field of artificial intelligence (AI). This approach broadens the usability of AI, increases its efficiency, and aligns it more closely with human thinking and decision-making processes. As we improve this method, it is set to become a key element in the future of AI, adding more purpose, precision, and ethical consideration to these technologies. In medicine, the chain-of-thought prompting is especially beneficial. Its capacity to handle complex information, its logical and sequential reasoning, and its suitability for ethically and context-sensitive situations make it an invaluable tool for healthcare professionals. Its role in enhancing medical care and research is expected to grow as we further develop and use this technique. Chain-of-thought prompting bridges the gap between AI’s traditionally obscure decision-making process and the clear, accountable standards required in healthcare. It does this by emulating a reasoning style familiar to medical professionals, fitting well into their existing practices and ethical codes. While solving AI transparency is a complex challenge, the chain-of-thought approach is a significant step toward making AI more comprehensible and trustworthy in medicine. This review focuses on understanding the workings of LLMs, particularly how chain-of-thought prompting can be adapted for nephrology’s unique requirements. It also aims to thoroughly examine the ethical aspects, clarity, and future possibilities, offering an in-depth view of the exciting convergence of these areas.

Background and Objectives: Limited evidence exists regarding the safety and efficacy of glucagon-like peptide-1 receptor agonists (GLP-1RAs) in type 2 diabetes mellitus (T2DM) patients with advanced chronic kidney disease (CKD) or end-stage kidney disease (ESKD). Thus, we conducted a systematic review and meta-analysis to assess the safety and efficacy of GLP-1RAs in T2DM patients with advanced CKD and ESKD. Materials and Methods: We performed a systematic literature search in MEDLINE, EMBASE, and Cochrane database until 25 October 2023. Included were clinical trials and cohort studies reporting outcomes of GLP-1RAs in adult patients with T2DM and advanced CKD. Outcome measures encompassed mortality, cardiovascular parameters, blood glucose, and weight. Safety was assessed for adverse events. The differences in effects were expressed as odds ratios with 95% confidence intervals (CIs) for dichotomous outcomes and the weighted mean difference or standardized mean difference (SMD) with 95% confidence intervals for continuous outcomes. The Risk of Bias In Non-randomized Studies—of Interventions (ROBIN-I) tool was used in cohort and non-randomized controlled studies, and the Cochrane Risk of Bias (RoB 2) tool was used in randomized controlled trials (RCTs). The review protocol was registered in the International Prospective Register of Systematic Reviews (CRD 42023398452) and received no external funding. Results: Eight studies (five trials and three cohort studies) consisting of 27,639 patients were included in this meta-analysis. No difference was observed in one-year mortality. However, GLP-1RAs significantly reduced cardiothoracic ratio (SMD of −1.2%; 95% CI −2.0, −0.4) and pro-BNP (SMD −335.9 pmol/L; 95% CI −438.9, −232.8). There was no significant decrease in systolic blood pressure. Moreover, GLP-1RAs significantly reduced mean blood glucose (SMD −1.1 mg/dL; 95% CI −1.8, −0.3) and increased weight loss (SMD −2.2 kg; 95% CI −2.9, −1.5). In terms of safety, GLP-1RAs were associated with a 3.8- and 35.7-time higher risk of nausea and vomiting, respectively, but were not significantly associated with a higher risk of hypoglycemia. Conclusions: Despite the limited number of studies in each analysis, our study provides evidence supporting the safety and efficacy of GLP-1RAs among T2DM patients with advanced CKD and ESKD. While gastrointestinal side effects may occur, GLP-1RAs demonstrate significant improvements in blood glucose control, weight reduction, and potential benefit in cardiovascular outcomes.

The emergence of artificial intelligence (AI) has greatly propelled progress across various sectors including the field of nephrology academia. However, this advancement has also given rise to ethical challenges, notably in scholarly writing. AI’s capacity to automate labor-intensive tasks like literature reviews and data analysis has created opportunities for unethical practices, with scholars incorporating AI-generated text into their manuscripts, potentially undermining academic integrity. This situation gives rise to a range of ethical dilemmas that not only question the authenticity of contemporary academic endeavors but also challenge the credibility of the peer-review process and the integrity of editorial oversight. Instances of this misconduct are highlighted, spanning from lesser-known journals to reputable ones, and even infiltrating graduate theses and grant applications. This subtle AI intrusion hints at a systemic vulnerability within the academic publishing domain, exacerbated by the publish-or-perish mentality. The solutions aimed at mitigating the unethical employment of AI in academia include the adoption of sophisticated AI-driven plagiarism detection systems, a robust augmentation of the peer-review process with an “AI scrutiny” phase, comprehensive training for academics on ethical AI usage, and the promotion of a culture of transparency that acknowledges AI’s role in research. This review underscores the pressing need for collaborative efforts among academic nephrology institutions to foster an environment of ethical AI application, thus preserving the esteemed academic integrity in the face of rapid technological advancements. It also makes a plea for rigorous research to assess the extent of AI’s involvement in the academic literature, evaluate the effectiveness of AI-enhanced plagiarism detection tools, and understand the long-term consequences of AI utilization on academic integrity. An example framework has been proposed to outline a comprehensive approach to integrating AI into Nephrology academic writing and peer review. Using proactive initiatives and rigorous evaluations, a harmonious environment that harnesses AI’s capabilities while upholding stringent academic standards can be envisioned.