Large Language Models (LLMs) show promise in healthcare tasks but face challenges in complex medical scenarios. We developed a Multi-Agent Conversation (MAC) framework for disease diagnosis, inspired by clinical Multi-Disciplinary Team discussions. Using 302 rare disease cases, we evaluated GPT-3.5, GPT-4, and MAC on medical knowledge and clinical reasoning. MAC outperformed single models in both primary and follow-up consultations, achieving higher accuracy in diagnoses and suggested tests. Optimal performance was achieved with four doctor agents and a supervisor agent, using GPT-4 as the base model. MAC demonstrated high consistency across repeated runs. Further comparative analysis showed MAC also outperformed other methods including Chain of Thoughts (CoT), Self-Refine, and Self-Consistency with higher performance and more output tokens. This framework significantly enhanced LLMs’ diagnostic capabilities, effectively bridging theoretical knowledge and practical clinical application. Our findings highlight the potential of multi-agent LLMs in healthcare and suggest further research into their clinical implementation.

Background: Musculoskeletal tissue degeneration impairs the life quality and motor function of many people, especially seniors and athletes. Tendinopathy is one of the most common diseases associated with musculoskeletal tissue degeneration, representing a major global healthcare burden that affects both athletes and the general population, with the clinical presentation of long-term recurring chronic pain and decreased tolerance to activity. The cellular and molecular mechanisms at the basis of the disease process remain elusive. Here, we use a single-cell and spatial RNA sequencing approach to provide a further understanding of cellular heterogeneity and molecular mechanisms underlying tendinopathy progression. Results: To explore the changes in tendon homeostasis during the tendinopathy process, we built a cell atlas of healthy and diseased human tendons using single-cell RNA sequencing of approximately 35,000 cells and explored the variations of cell subtypes’ spatial distributions using spatial RNA sequencing. We identified and localized different tenocyte subpopulations in normal and lesioned tendons, found different differentiation trajectories of tendon stem/progenitor cells in normal/diseased tendons, and revealed the spatial location relationship between stromal cells and diseased tenocytes. We deciphered the progression of tendinopathy at a single-cell level, which is characterized by inflammatory infiltration, followed by chondrogenesis and finally endochondral ossification. We found diseased tissue-specific endothelial cell subsets and macrophages as potential therapeutic targets. Conclusions: This cell atlas provides the molecular foundation for investigating how tendon cell identities, biochemical functions, and interactions contributed to the tendinopathy process. The discoveries revealed the pathogenesis of tendinopathy at single-cell and spatial levels, which is characterized by inflammatory infiltration, followed by chondrogenesis, and finally endochondral ossification. Our results provide new insights into the control of tendinopathy and potential clues to developing novel diagnostic and therapeutic strategies.

Background: Acute gouty arthritis is inflammatory joint arthritis. Gouty arthritis (GA) involves multiple pathological processes. Deposition of joints by monosodium urate (MSU) crystals has been shown to play a critical role in the injury process. Due to the different effects of MSU stimulation on the joints, the exact changes in the synovial fluid are unknown. We want to explore the changes in proteins and metabolites in the joints of gouty arthritis. Regulating various functional substances in the joint can reduce inflammation and pain symptoms. Methods: 10 patients with gouty knee arthritis and 10 normal controls were selected from clinical, surgical cases. The biological function of the metabolome was assessed by co-expression network analysis. A molecular network based on metabolomic and proteomic data was constructed to study critical molecules. The fundamental molecular changes in the relevant pathways were then verified by western blot. Results: Proteomic analysis showed that the expressions of proteases Cathepsin B, Cathepsin D, Cathepsin G, and Cathepsin S in synovial fluid patients with gouty arthritis were significantly increased. Enrichment analysis showed a positive correlation between lysosomal and clinical inflammatory cell shape changes. Untargeted metabolomic analysis revealed that lipids and lipoids accumulate, inhibit autophagic flux, and modulate inflammation and immunity in gouty arthritis patients. It was determined that the accumulation of lipid substances such as phospholipase A2 led to the imbalanced state of the autophagy-lysosome complex, and the differentially expressed metabolites of Stearoylcarnitine, Tetradecanoylcarnitine, Palmitoylcarnitine were identified (|log2 fold change|> 1.5, adjusted P value < 0.05 and variable importance in prediction (VIP) > 1.5). The autophagy-lysosomal pathway was found to be associated with gouty knee arthritis. Essential molecular alterations of multi-omics networks in gouty knee arthritis patients compared with normal controls involve acute inflammatory response, exosomes, immune responses, lysosomes, linoleic acid metabolism, and synthesis. Conclusions: Comprehensive analysis of proteomic and untargeted metabolomics revealed protein and characteristic metabolite alterations in gouty arthritis, it mainly involves lipids and lipid like molecules, phospholipase A2 and autophagic lysosomes. This study describes the pathological characteristics, pathways, potential predictors and treatment goals of gouty knee arthritis.

The combination of lateral extra-articular tenodesis (LET) with primary single-bundle anterior cruciate ligament (ACL) reconstruction (ACLR) remains controversial. To determine whether the combination of LET with single-bundle ACLR provides greater control of anterolateral rotatory instability and improved clinical outcomes compared with ACLR alone. Systematic review; Level of evidence, 2. PubMed, Embase, and the Cochrane Central Register of Controlled Trials databases were searched between inception and July 1, 2020. Level 1 or 2 randomized controlled trials that compared isolated single-bundle ACLR with combined LET with ACLR were included. Data were meta-analyzed for the primary outcome measure of knee stability and the secondary outcome measures of patient-reported outcome scores, return to sports, and graft failure. Dichotomous variables were presented as relative risks (RRs), and continuous variables were presented as mean differences (MDs) and standardized MDs (SMDs). A total of 6 studies involving 1010 patients were included. Pooled data showed that the ACLR+LET group had a lower incidence of the pivot shift (RR, 0.56 [95% CI, 0.45 to 0.69]; P < .00001), a higher postoperative activity level (MD, 0.47 [95% CI, 0.15 to 0.78]; P = .004), and a lower risk of graft failure (RR, 0.35 [95% CI, 0.21 to 0.59]; P < .00001) than did the ACLR group. However, there were no statistically significant differences in primary outcomes including positive Lachman test findings (RR, 0.76 [95% CI, 0.48 to 1.21]; P = .26) or side-to-side differences (SMD, –0.43 [95% CI, –0.95 to 0.09]; P = .11) or in secondary outcomes including International Knee Documentation Committee scores (SMD, 0.25 [95% CI, –0.06 to 0.56]; P = .11) or Lysholm scores (SMD, 0.28 [95% CI, –0.06 to 0.62]; P = .11). Although the overall rate of return to sports was not significantly different between the groups (RR, 0.97 [95% CI, 0.90 to 1.03]; P = .33), the activity level was higher in the ACLR+LET group. The addition of LET to primary single-bundle ACLR produced greater knee stability, a higher activity level, and a lower incidence of graft failure than did ACLR alone. There may be a role for adding LET to ACLR for the treatment of ACL injuries.

Purpose: FAI (femoroacetabular impingement syndrome) is a common cause of hip pain, resulting in a decreased life quality. This study aims to compare the postoperative clinical outcome between arthroscopic surgery (AT) and conservative treatment (CT). Method: The six studies were selected from PubMed, Embase and OVID database. The data were extracted and analyzed by RevMan5.3. Mean differences and 95% confidence intervals were calculated. RevMan5.3 was used to assess the risk of bias. Result: Six observational studies were assessed. The methodological quality of the trials indicated five of six studies had a low risk of bias and one article had a high risk of bias. The differences were statistically significant between AT and CT for HOS (follow-up for 6 months), iHOT-33 (follow-up for 6 months) improvement, iHOT-33 (follow-up for 12 months) improvement, iHOT-33 (follow-up for 12 months), EQ-5D-5L index score (follow-up for 12 months) and AT showed higher benefits than CT. Meanwhile no statistically significant were found in iHOT-33 (follow-up for 6 months), EQ-5D-5L index score (follow-up for 6 months), EQ5D-VAS (follow-up for 6 months) and EQ5D-VAS (follow-up for 12 months). Conclusion: AT and CT both can have clinical effects when facing FAI. In our meta-analysis, hip arthroscopy is statistically superior to conservative treatment in both long-term and short-term effects.