Zand, Jaleh

Background/Objective: Hypertension is the most prevalent chronic disease in primary care and a leading cause of cardiovascular morbidity and mortality. Despite existing guidelines, therapeutic inertia and suboptimal control persist. Large language models (LLMs) offer a potential valuable addition to augment clinical decision-making, yet their reliability for guideline-driven tasks remains unverified. This study evaluated the accuracy and safety of hypertension management recommendations generated by three LLMs compared to expert responses. Methods: Fifty-one clinical vignettes representing 17 core hypertension management concepts were constructed by hypertension experts. Each case was submitted to three LLMs (GPT-4, Gemini, MedLM) and a hypertension expert also wrote the “gold standard” answers. Three blinded expert reviewers rated each response on a 4-point accuracy scale, a binary safety (safe/unsafe) scale, and attempted to identify the source (LLM vs. expert) providing the response. Ratings were analyzed using mean scores, percentages of accurate and safe responses, and inter-rater agreement. Results: GPT-4 had the highest accuracy (83%) and safety (86%) scores among LLMs but remained inferior to expert responses (92% accuracy, 93% safety). Gemini and MedLM performed significantly worse (accuracy: 64% and 35%; safety: 73% and 39%, respectively). GPT-4 generated the most guideline-concordant responses (46%) among the three LLMs (Gemini 35%, MedLM 14%), but remains lower than experts’ responses (68%). Evaluators misidentified LLM responses as expert-written in 10 to 25% of cases, particularly with GPT-4. Inter-rater reliability for accuracy ratings was highest for expert-generated responses (ICC 0.81), with progressively lower agreement for GPT-4 (0.76), Gemini (0.70), and MedLM (0.68). A similar pattern was observed for safety and source discrimination ratings. The agreement was strongest for safety assessments and weakest for source discrimination. Conclusion: Among three tested LLMs, GPT-4 demonstrated closer agreement to expert decisions thereby showing greater potential for supporting hypertension management. However, current LLMs’ versions frequently produce inaccurate or unsafe recommendations and remain inferior to expert judgment. Human-in-the-loop supervision remains essential when deploying LLMs for clinical decision-making.

BACKGROUND: Hypertension, the leading cause of cardiovascular mortality, remains suboptimally controlled. Large language models (LLMs) could improve hypertension control by augmenting clinical decision-making, but their reliability for guideline-driven tasks is unverified. This study evaluated the accuracy and safety of hypertension management recommendations generated by 3 LLMs. METHODS: Fifty-one vignettes were constructed and submitted to the LLMs (GPT-4, Gemini, Medical Large Language Model [by Google; MedLM]) and a hypertension expert to generate the responses. Three blinded reviewers rated each response on a 4-point accuracy scale, a binary safety (safe/unsafe) scale, and attempted to identify the source (LLM versus expert) providing the response. RESULTS: GPT-4 had the highest accuracy (83%) and safety (86%) scores among LLMs but remained inferior to expert responses (92% accuracy, 93% safety). Gemini and MedLM performed significantly worse (accuracy: 64% and 35%; safety: 73% and 39%, respectively). GPT-4 generated the most guideline-concordant responses (46%) among the 3 LLMs (Gemini 35%, MedLM 14%) but was lower than expert responses (68%). Interrater reliability for accuracy ratings was higher for LLM-generated responses (GPT-4 [intraclass correlation coefficient, 0.30], Gemini [intraclass correlation coefficient, 0.61], and MedLM [intraclass correlation coefficient, 0.58]), with lower agreement for expert responses (intraclass correlation coefficient, 0.23). A similar pattern was observed for safety and source discrimination ratings. The agreement was strongest for safety assessments and weakest for source discrimination. CONCLUSIONS: Among the 3 tested LLMs, GPT-4 demonstrated closer agreement to expert decisions, thereby showing greater potential for supporting hypertension management. Despite their potential, current LLM versions are inferior to expert recommendations. Human-in-the-loop supervision remains essential when deploying LLMs for clinical decision-making.

ABSTRACT Background Membranous nephropathy (MN) associated with phospholipase A2 receptor (PLA2R) antibodies is the most common cause of nephrotic syndrome in non-diabetic adult patients. This study investigated the relationship between clinical phenotypes and renal outcomes in this population, emphasizing the potential for phenotype-based treatment stratification. Methods We conducted a retrospective, multi-center cohort study of PLA2R-positive MN. Unsupervised cluster analysis grouped patients based on clinicopathological characteristics. Primary outcomes included complete or partial remission within 2 years of biopsy and end-stage kidney disease (ESKD) or death during follow-up. Results Among 178 patients, three distinct clusters emerged (n = 89, 70, and 19). Within 2 years of biopsy, 102 patients (57%) achieved complete or partial remission. Cluster 1, characterized by the mildest disease markers, including the lowest body mass index, serum anti-PLA2R titer, serum creatinine, proteinuria, and triglycerides, had the highest remission rate of 72%, compared with 50% in cluster 2 and 54% in cluster 3. Cluster 2 had significantly lower remission compared to cluster 1 (HR 0.64, 95% CI 0.42–0.96, P = .03); results were similar in cluster 3, albeit not statistically significant (HR 0.50, 95% CI 0.23–1.09, P = .08). Elevated anti-PLA2R levels (>100 U/ml) and proteinuria (>8 g/24-hour) predicted reduced remission (HR 0.56, 95% CI 0.36–0.88, P = .01; HR 0.43, 95% CI 0.26–0.73, P = .002, respectively), while high high-density lipoprotein levels were protective (HR 1.01, 95% CI 1.00–1.02, P = .048). Overall, 21 patients (12%) developed ESKD or died. ESKD-free survival differed significantly across clusters (log-rank test P = .04). Conclusions Unsupervised clustering identified distinct clinical phenotypes in PLA2R-positive MN, each associated with different renal prognoses. Phenotype-based risk stratification could enhance treatment precision, improve patient outcomes, and potentially reduce treatment-related adverse effects.

AbstractBackgroundAccurately estimating glomerular filtration rate (GFR) is crucial for dosing medications in hospitalized patients. Due to limitations of serum creatinine for GFR estimation, serum cystatin C (CysC) has been explored as an alternative functional kidney biomarker. This study assessed discordance between eGFRCr and eGFRCysC in a large sample of hospitalized patients and examined the frequency of renally eliminated medications affected by eGFR discordance.MethodsThis multisite historical study included adults hospitalized between 2011 and 2023 with CysC and serum creatinine reported within 24 h of each other. The first concurrent biomarker pair for each patient was analyzed. eGFR discordance and use of renally eliminated medications were described.Results17,718 hospitalized patients with concurrent creatinine and CysC assessments were included. The median eGFRCr was 65 mL/min, and the eGFRCysC was 46 mL/min. The median absolute difference of eGFRCr−eGFRCysC was 15 mL/min, and 7972 patients (45%) had a > 30% absolute difference. There was a significantly greater percentage of patients with an eGFR <30 mL/min based on eGFRCysC (26%) compared to eGFRCr (15%) (p < 0.001). Patients were prescribed an average of 20 medications in the 24 h surrounding the concurrent biomarker assessment. Renally eliminated medications accounted for 39% ± 13% of medication orders, and 80% of patients with eGFR discordance were prescribed five or more renally eliminated medications.ConclusionSubstantial eGFR discordance between eGFRCysC and eGFRCr was observed in hospitalized patients, which directly affects the dosing of renally eliminated medications. Further research is needed to optimize the pharmacotherapy of renally eliminated medications with discordant GFR assessments to improve medication safety and effectiveness.