Tingting Hu

Primary diabetes care and diabetic retinopathy (DR) screening persist as major public health challenges due to a shortage of trained primary care physicians (PCPs), particularly in low-resource settings. Here, to bridge the gaps, we developed an integrated image-language system (DeepDR-LLM), combining a large language model (LLM module) and image-based deep learning (DeepDR-Transformer), to provide individualized diabetes management recommendations to PCPs. In a retrospective evaluation, the LLM module demonstrated comparable performance to PCPs and endocrinology residents when tested in English and outperformed PCPs and had comparable performance to endocrinology residents in Chinese. For identifying referable DR, the average PCP's accuracy was 81.0% unassisted and 92.3% assisted by DeepDR-Transformer. Furthermore, we performed a single-center real-world prospective study, deploying DeepDR-LLM. We compared diabetes management adherence of patients under the unassisted PCP arm (n = 397) with those under the PCP+DeepDR-LLM arm (n = 372). Patients with newly diagnosed diabetes in the PCP+DeepDR-LLM arm showed better self-management behaviors throughout follow-up (P < 0.05). For patients with referral DR, those in the PCP+DeepDR-LLM arm were more likely to adhere to DR referrals (P < 0.01). Additionally, DeepDR-LLM deployment improved the quality and empathy level of management recommendations. Given its multifaceted performance, DeepDR-LLM holds promise as a digital solution for enhancing primary diabetes care and DR screening.

BACKGROUND: Low muscle mass likely results in reduced capacity for glucose disposal, leading to a significant but under-appreciated contribution to increasing the risk of diabetes. But few prospective studies have investigated the association between the loss of muscle mass and the occurrence of diabetes. We aimed to investigate whether short-term changes in muscle mass affect the incidence of diabetes in a Chinese population. METHODS: This study included 1275 individuals without evident diabetes at baseline. In the baseline and re-examination, individuals completed the risk factors survey and underwent body composition measurement. Muscle mass index was defined as the percentage skeletal muscle mass, which was measured by an automatic bioelectric analyzer. RESULTS: After a median follow-up of 2.1 years, 142 individuals developed diabetes (11.1%). There was an inverse association between basal skeletal muscle mass index and the risk of diabetes in participants with impaired glucose regulation but not in those with normal glucose tolerance. Multivariate-adjusted hazard ratios for the risk of developing diabetes were 0.85 (95% CI: 0.74-0.98) and 1.15 (95% CI: 0.98-1.34), respectively. Furthermore, Cox regression analysis revealed that a two-year change in skeletal muscle mass was also inversely associated with the incidence of diabetes in both participants with normal glucose tolerance and with impaired glucose regulation (HR: 0.76, 95% CI: 0.65-0.89; HR: 0.81, 95% CI: 0.71-0.91). CONCLUSIONS: These findings emphasized the importance of early detection and control of muscle mass loss for the prevention of diabetes.

Heterogeneous nuclear ribonucleoprotein A1 (HNRNPA1) is involved in lipid and glucose metabolism via mRNA processing. However, whether and how HNRNPA1 alters adipocyte function in obesity remain obscure. Here, we found that the obese state downregulated HNRNPA1 expression in white adipose tissue (WAT). The depletion of adipocyte HNRNPA1 promoted markedly increased macrophage infiltration and expression of proinflammatory and fibrosis genes in WAT of obese mice, eventually leading to exacerbated insulin sensitivity, glucose tolerance, and hepatic steatosis. Mechanistically, HNRNPA1 interacted with Ccl2 and regulated its mRNA stability. Intraperitoneal injection of CCL2-CCR2 signaling antagonist improved adipose tissue inflammation and systemic glucose homeostasis. Furthermore, HNRNPA1 expression in human WAT was negatively correlated with BMI, fat percentage, and subcutaneous fat area. Among individuals with 1-year metabolic surgery follow-up, HNRNPA1 expression was positively related to percentage of total weight loss. These findings identify adipocyte HNRNPA1 as a link between adipose tissue inflammation and systemic metabolic homeostasis, which might be a promising therapeutic target for obesity-related disorders.

Diabetes poses a considerable global health challenge, with varying levels of diabetes knowledge among healthcare professionals, highlighting the importance of diabetes training. Large Language Models (LLMs) provide new insights into diabetes training, but their performance in diabetes-related queries remains uncertain, especially outside the English language like Chinese. We first evaluated the performance of ten LLMs: ChatGPT-3.5, ChatGPT-4.0, Google Bard, LlaMA-7B, LlaMA2-7B, Baidu ERNIE Bot, Ali Tongyi Qianwen, MedGPT, HuatuoGPT, and Chinese LlaMA2-7B on diabetes-related queries, based on the Chinese National Certificate Examination for Primary Diabetes Care in China (NCE-CPDC) and the English Specialty Certificate Examination in Endocrinology and Diabetes of Membership of the Royal College of Physicians of the United Kingdom. Second, we assessed the training of primary care physicians (PCPs) without and with the assistance of ChatGPT-4.0 in the NCE-CPDC examination to ascertain the reliability of LLMs as medical assistants. We found that ChatGPT-4.0 outperformed other LLMs in the English examination, achieving a passing accuracy of 62.50%, which was significantly higher than that of Google Bard, LlaMA-7B, and LlaMA2-7B. For the NCE-CPFC examination, ChatGPT-4.0, Ali Tongyi Qianwen, Baidu ERNIE Bot, Google Bard, MedGPT, and ChatGPT-3.5 successfully passed, whereas LlaMA2-7B, HuatuoGPT, Chinese LLaMA2-7B, and LlaMA-7B failed. ChatGPT-4.0 (84.82%) surpassed all PCPs and assisted most PCPs in the NCE-CPDC examination (improving by 1 %-6.13%). In summary, LLMs demonstrated outstanding competence for diabetes-related questions in both the Chinese and English language, and hold great potential to assist future diabetes training for physicians globally.