The efficacy of large language models (LLMs) in domain-specific medicine, particularly for managing complex diseases such as osteoarthritis (OA), remains largely unexplored. This study focused on evaluating and enhancing the clinical capabilities of LLMs in specific domains, using osteoarthritis (OA) management as a case study. A domain specific benchmark framework was developed, which evaluate LLMs across a spectrum from domain-specific knowledge to clinical applications in real-world clinical scenarios. DocOA, a specialized LLM tailored for OA management that integrates retrieval-augmented generation (RAG) and instruction prompts, was developed. The study compared the performance of GPT-3.5, GPT-4, and a specialized assistant, DocOA, using objective and human evaluations. Results showed that general LLMs like GPT-3.5 and GPT-4 were less effective in the specialized domain of OA management, particularly in providing personalized treatment recommendations. However, DocOA showed significant improvements. This study introduces a novel benchmark framework which assesses the domain-specific abilities of LLMs in multiple aspects, highlights the limitations of generalized LLMs in clinical contexts, and demonstrates the potential of tailored approaches for developing domain-specific medical LLMs.

This retrospective study aimed to compare the effectiveness between concurrent meniscoplasty and conservative management on the asymptomatic side in children with unilaterally symptomatic bilateral discoid lateral meniscus operated on the symptomatic side in a tertiary level of clinical care. Patients aged under 14 with unilaterally symptomatic bilateral discoid lateral meniscus underwent arthroscopic meniscoplasty were eligible. Patients were either underwent meniscoplasty on the symptomatic side, with the asymptomatic side conservatively treated (group 1) or underwent meniscoplasty on both sides concurrently (group 2). Functional outcomes were evaluated using the Lysholm score and Ikeuchi score. Relative cost collected through the hospital data was applied by the Kruskal-Wallis test. The occurrence of symptoms was applied by the Kaplan–Meier model. A total of 50 patients (39 female and 11 male) were eligible. For the average Lysholm scores, groups 1 and 2 of the previously asymptomatic side were 90.86 ± 8.25 and 92.62 ± 8.68. For the symptomatic side, the Lysholm scores were 91.38 ± 8.90 and 95.71 ± 7.45. For the average cost of treatment, a significant difference was discovered between group 1 and 2 (P < .0001). The Kaplan–Meier survival analysis of the occurrence of symptoms showed no statistical difference between the 2 groups (P = .162), and the terminal survival rates in the 2 groups were 86.2% and 81.0%, respectively. Conservative treatment reached the same clinical outcomes as the concurrent meniscectomy, with a potential of longer mean survival time and lower demand of cost in treatment.

Background: In recent studies, there have been no reports on the diagnostic value of musculoskeletal ultrasound in pediatric patients with a discoid lateral meniscus (DLM). This study aims to evaluate the diagnostic value of musculoskeletal ultrasound in pediatric DLM patients and to investigate potential correlations between ultrasound indices and subgroup classifications on the basis of ultrasound findings. Methods: This study included patients aged 18 years or younger from 1 September 2021, to 1 January 2023, who were diagnosed with lateral meniscus (LM) or DLM injuries through arthroscopic surgery. Ultrasound and MRI assessments were performed before surgery. Radiological measurements were conducted by one evaluator to determine the diagnostic value of the two methods. The McNemar–Bowker test was used to compare the sensitivity and specificity between the two methods. Spearman’s correlation was used for correlation analysis of different injury types. Cluster analysis was used to explore the basis for classifying the DLM. Results: This study included 72 patients (37 females and 35 males). The average age was 11.18 ± 4.25 years, and the BMI was 18.50 ± 3.46 kg/m ² . The results revealed the high diagnostic efficacy of ultrasound and MRI. There was no significant difference in sensitivity or specificity in diagnosing DLM via ultrasound compared with MRI ( P >0.05). Subgroup analysis did not reveal a significant difference between complete/incomplete DLM type or child/adolescent age ( P >0.05). Correlation analysis revealed that DLM patients were more likely to suffer from vertical tears (r = 0.34, P = 0.015) and locking symptoms (r = 0.36, P = 0.032). In the 30-degree flexion assessment, body thickness was useful for detecting vertical (r = 0.26, P = 0.049) and bucket tears (r = 0.3, P = 0.026). Cluster analysis of all DLM patients revealed two subtypes: DLM type I, with more anterior horn tears, and Type II DLM, with more popliteal hiatus tears ( P = 0.020). Conclusion: MUS has proven to be a valuable diagnostic tool with comparable diagnostic efficiency to that of MRI in pediatric DLM. Ultrasound indices correlated with different injury types and related symptoms. Cluster analysis revealed that pediatric DLMs can potentially be classified into two types. The two types of injuries differ in terms of injury area.