Schwartz, Gary L.

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.

Background: Hypertension (HTN) affects almost half of the adults in the US, and most do not have their blood pressure (BP) under control. Minorities and individuals living in rural areas experience health disparities. The goal of this study was to assess the rate of BP control among adults in a large health system that included minorities and adults living in rural areas. Methods: Clinical cohort study of adults with HTN who live in Southeast Minnesota and had at least one outpatient clinic encounter within the Mayo Clinic Health System in the last 12 months (5/2021-5/2022). Data was extracted through the HTN registry of the electronic health record and included age, race, ethnicity, last office BP and any active anti-hypertensives drug use. Results: The overall rate of HTN control among 127,644 individuals with HTN, defined as an office BP < 130/80 mmHg, was 25.7%. The rates of BP control between HTN subjects living in rural (64527, 67.3%) versus urban areas (41285, 32.3 %), were significantly lower at 24.9% vs. 27% respectively, p< 0.001. A total of 8,722 (6.8%) were non-white, including 2640 Black, 1919 Asian, and 2867 Hispanics A total of 5,734 (45%) had a SBP ≥ 140 or DBP ≥ 90 mmHg and 7178 (12.5%) did not have any anti-hypertensives in their medication list. The percentage of White and Asian persons who were not on anti-hypertensives was 12.4% and 12.5%, respectively, while a larger percentage of Black (13.8%) and Hispanic (14%) persons were not (p<0.001). The corresponding percentage was 16.7% among individuals who did not disclose their race (p<0.001). Conclusion: Among hypertensive adults, BP was not at goal in most patients based on their most recent office BP measurement with a higher rate among those living in rural areas. About 12.5% were not on anti-hypertensives with a disproportionate larger percentage of Black and Hispanic compared to White and Asian persons.

The combination of endocrine therapy and cyclin-dependent kinase (CDK) 4/6 inhibitors is currently one of the preferred first-line regimens for patients with hormone receptor (HR)-positive, human epidermal growth factor receptor 2 (HER2)-negative advanced or metastatic breast cancer. CDK 4/6 inhibitors can cause serum creatinine (SCr) elevation and corresponding estimated glomerular filtration rate (eGFR) reduction without affecting kidney function by inhibiting tubular secretory transporters in the kidneys—a phenomenon called pseudo-acute kidney injury (pseudo-AKI). We report a case of a 69-year old woman found to have elevated SCr while on ribociclib for treatment of metastatic breast cancer. Further evaluation with cystatin C-based eGFR and measured GFR by urinary iothalamate clearance revealed values close to her baseline Cr-based eGFR prior to initiation of treatment with ribociclib. She was therefore diagnosed with ribociclib-induced pseudo-AKI. Treatment with ribociclib was continued with steady favorable response. This case highlights that, in addition to true AKI, ribociclib can cause pseudo-AKI. Alternative methods of kidney function assessment may be necessary in patients who develop elevated SCr while on this medication to help discriminate between artifactual and true kidney function impairment. Doing so can help avoid unnecessary work-up and/or premature discontinuation of the treatment.

37 Knowledge of genetic factors that influence pharmacodynamics of responses to antihypertensive drugs may provide insights into molecular mechanisms contributing to elevation of BP and response to treatment. The T-allele of the C825T polymorphism of the gene coding for the G-protein β 3 subunit ( GNB3 ) has been associated with increased Na + -H + exchange and low plasma renin in individuals with EHT. Moreover, diminished counterregulatory activity of the renin-angiotensin system has been associated with greater BP declines in response to negative sodium balance induced by diuretic therapy. To assess the possible influence of GNB3 genotype on BP response to diuretic therapy, we measured the C825T polymorphism in community-based samples of 204 blacks (139 women, 65 men) and 192 whites (77 women, 115 men) with EHT (mean age=48±7 years) who underwent monotherapy with hydrochlorothiazide (HCTZ) for 4 weeks. In univariate analyses, the TT genotype was significantly associated with lower plasma renin activity ( p <0.001) and higher plasma aldosterone concentration ( p <0.019). In addition, mean declines in systolic blood pressure (SBP) and diastolic blood pressure (DBP) were 6±2 mmHg ( p <0.001) and 5±1 mmHg ( p <0.001) greater, respectively, in TT than in CC homozygotes. Responses in heterozygotes were intermediate between homozygote groups. Other univariate predictors of greater declines in both SBP and DBP were higher pretreatment BP ( p <0.001), black race ( p <0.001), female gender ( p <0.005), older age ( p ≤0.064), lower waist/hip ratio (p≤0.087), and lower plasma renin activity ( p <0.001). Even after considering effects of these other predictors, the TT genotype remained associated with greater declines in SBP ( p <0.058) and DBP ( p <0.001). Thus, the C825T polymorphism of GNB3 may help identify individuals with EHT who are more responsive to diuretic therapy and may have Na + -dependent BP elevation.

Background: Compared to those with natural nocturnal blood pressure (BP) decreases, those lacking nocturnal BP decreases (“non-dippers”) are at greater risk of cardiovascular disease (CVD). Substantial evidence also supports a role of chronic systemic inflammation in the pathogenesis of CVD. The purpose of this study was to examine associations of markers of inflammation with two common definitions of BP dipping status using 24-hour ambulatory blood pressure (AMBP) readings. Methods: AMBP monitoring and assays of C-reactive protein (CRP), interleukin-6 (IL-6), e-selectin, tissue inhibitor of metalloproteinase-2 (TIMP-2), and a selection of thirteen other inflammatory-pathway biomarkers were performed in 722 participants (mean age 64 [range 42-83] years, 36% men, 39% black, 83% hypertensive) in the Genetic Epidemiology Network of Arteriopathy (GENOA) Study (2001-2004). Non-dippers were classified using two common definitions: (1) nocturnal systolic and diastolic BP both decreased by <10% compared to daytime values [“dual non-dippers”, n=161 (22%)], and (2) nocturnal SBP decreased by <10% [“systolic non-dippers”, n=380 (53%)]. Logistic regressions, fit with generalized estimating equations to account for sibship clustering, yielded odds ratios (OR) of dipping status for log-transformed inflammatory biomarkers, adjusted for age, race, and sex. Results: IL-6 (range 0-37 pg/ml) was associated with dual non-dipping status (OR=1.55 [95%CI: 1.03, 2.32] p=0.035), translating into a 55% increase in the likelihood of being a dual non-dipper for every approximate doubling of IL-6. IL-6 (OR=1.63 [1.13, 2.35] p=0.009) and CRP (OR=1.35 [1.08,1.68] p=0.008) overall, as well as e-selectin (OR=0.49 [0.28, 0.87] p=0.014) and TIMP-2 (OR=0.54 [0.30,0.98] p=0.041) in hypertensive subjects were associated with systolic non-dipping status. Conclusions: These results suggest that non-dipping classification should be approached with prudence, as the two different criteria may influence conclusions about predictive marker associations. In addition, markers involved in direct versus inhibitory inflammation pathways may operate differentially across patient subgroups. Our findings support the link between inflammation and non-dipping circadian BP patterns and provide novel information on markers and definitions which may be useful for early detection and subsequent prevention of CVD.

Objective: Vasopressin increases water permeability and activity of the epithelial sodium channel ENaC in the distal nephron of the kidney, favoring increased water and sodium retention. We hypothesized that genetically mediated differences in circulating vasopressin concentrations could contribute to features of hypertension that vary between non-Hispanic blacks and whites. To test this hypothesis we measured copeptin, a fragment of pre-pro arginine vasopressin with a prolonged half-life that is more easily quantified. Methods: The Genetic Epidemiology of Responses to Antihypertensives (GERA) I study was designed to identify predictors of response to diuretic therapy in non-Hispanic black and non-Hispanic white adults with essential hypertension. Blood samples were obtained after a 4 week washout of anti-hypertensive medication while on a sodium intake of 2 mmol/kg. Copeptin was measured from thawed EDTA plasma samples using the copeptin proAVP homogeneous automated immunofluorescent assay on the BRAHMS Kryptor Compact PLUS (Thermo Fisher). Results: Copeptin was significantly higher in blacks (median: 7.4 pmol/L, IQR: 4.4-10.2 pmol/L) than in whites (median: 5.4 pmol/L, IQR: 3.6-8.0 pmol/L) (p <0.001), while renin was lower (Table 1). Racial differences remained significant (p = 0.001) in multivariate models that included 24 hour urine sodium excretion and calculated plasma osmolality. Conclusions: Higher copeptin (vasopressin) concentrations in non-Hispanic blacks compared to whites may contribute to the distinct differences in the hypertensive phenotypes between these racial groups. The underlying mechanism(s) require further study.