AUTHOR=Jin Haijiao , Huang Lulu , Ye Jinling , Wang Jinkun , Lin Xinghui , Wu Shaun , Hu Weiguo , Lin Qisheng , Li Xiaoyang TITLE=Enhancing nutritional management in peritoneal dialysis patients through a generative pre-trained transformers-based recipe generation tool: a pilot study JOURNAL=Frontiers in Medicine VOLUME=11 YEAR=2024 URL=https://www.frontiersin.org/journals/medicine/articles/10.3389/fmed.2024.1469227 DOI=10.3389/fmed.2024.1469227 ISSN=2296-858X ABSTRACT=Background

Patients undergoing peritoneal dialysis (PD) often face nutritional deficiencies due to inadequate intake, nutrient loss, insufficient dialysis, and a state of micro-inflammatory. Traditional nutritional management methods have not fully met personalized needs. Therefore, this study aimed to develop and evaluate an application for generating recipes based on Generative Pre-trained Transformers to improve the nutritional status of these patients.

Methods

This self-controlled prospective study included 35 patients undergoing PD from January to February 2024. The study was divided into two phases: the initial phase involved conventional dietary education under PD management, followed by a second phase where a new GPT-based dietary guidance tool was introduced. Patients adhered to the diets recommended by the tool. Nutritional intervention effects were assessed by comparing serum prealbumin, albumin, and phosphate levels before and after the intervention.

Results

After the intervention, the mean prealbumin levels significantly improved from 289.04 ± 74.60 mg/L to 326.72 ± 78.89 mg/L (p = 0.001). Although there was no statistical significance, the serum albumin levels in patients increased from 34.70 ± 5.94 g/L to 35.66 ± 5.14 g/L (p = 0.153). Serum phosphate levels remained stable and within safe limits (p = 0.241).

Conclusion

The AI-based recipe generation application significantly improved serum prealbumin levels in PD patients without causing adverse changes in phosphate levels, confirming its efficacy and safety in nutritional management for these patients. This study highlights the potential and practical value of AI technology in nutritional management for patients with chronic disease, providing important evidence for future clinical applications.