Wang, Zhen

Objectives This study aims to compare the reliability and accuracy of three large language models (LLMs) (Claude, Gemini and GPT) in assessing the risk of bias of nonrandomised studies using the ROBINS-I tool. Methods and analysis We conducted a secondary analysis of 171 nonrandomised studies previously assessed with Risk Of Bias In Non-randomized Studies of Interventions (ROBINS-I) tool by two independent human review teams. Only studies with concordant human domain-level ratings were included. Each study was independently assessed twice by Claude, Gemini and Generative Pre-trained Transformer (GPT) using agent-based structured implementations of the ROBINS-I tool. Reliability (agreement between two runs of the same LLM) was evaluated using percent agreement and Gwet’s AC1. Accuracy (agreement with human reviewers) was assessed only for studies with consistent LLM ratings, using the same metrics. Results Claude demonstrated high reliability across all domains (79.5–98.0% agreement, AC1=0.729–0.975). Gemini showed moderate-to-high reliability (agreement 76.7–100%, AC1=0.680–1.0). GPT exhibited lower reliability overall, though domain-level agreement ranged from 70.9–95.6% (AC1=0.596–0.944). In terms of accuracy, Claude showed overall poor agreement with human reviewers (14.4–68.5% agreement; low AC1 values). Gemini demonstrated moderate-to-high accuracy in several domains, including deviations from intended interventions (79.6%, AC1=0.848) and measurement of outcomes (73.9%, AC1=0.702), with the highest overall agreement (40.0%, AC1=0.672). GPT showed variable accuracy, with the highest in measurement of outcomes (62.8%, AC1=0.571) and classification of interventions (57.8%, AC1=0.498), but poor performance in selection (14.3%, AC1 = −0.041) and overall agreement (23.0%, AC1=0.267). Conclusions Claude was internally consistent but poorly aligned with human reviewers. Gemini achieved both high reliability and moderate-to-high accuracy, whereas GPT had lower reliability and mixed accuracy. Current off-the-shelf LLMs cannot reliably perform ROBINS-I risk of bias assessments.

Background: Spontaneous coronary artery dissection (SCAD) is defined as a non-traumatic separation of the epicardial coronary artery walls that creates a false lumen. SCAD poses a difficult challenge in management, as decisions regarding revascularization and medical management seem to be tailored to the individual patient. We evaluated and compared outcomes based on cardiogenic shock in patients with SCAD utilizing Nationwide Readmissions Database (NRD) between January 1, 2016, to December 30, 2020. Methods: We utilized the NRD 2016–2019 to carry out this study. We evaluated demographics (e.g., age, gender), conventional risk factors, comorbidities present on the index admission, and in-hospital outcomes using their specific ICD-10-CM codes. The primary outcomes were In-hospital mortality and 30-day readmission, and the secondary outcome was to compare the complications in SCAD patient with cardiogenic shock (CS) compared to those without CS. Results: We analyzed 2473 individuals with SCAD, 2199 of these individuals did not have cardiogenic shock whereas 274 of these individuals did have cardiogenic shock. When comparing SCAD with cardiogenic shock to SCAD without cardiogenic shock, there was a statistically significant increased odds ratio (OR) for death (propensity matched OR 24.93 (7.49–83.05), use of mechanical circulatory support (propensity matched OR 15.30 (6.87–34.04), ventricular tachycardia (propensity matched OR 4.45 (1.92–10.34), utilization of blood transfusions (propensity matched OR 3.82 (1.86–7.87), acute kidney injury (propensity matched OR 4.02 (1.45–11.13), need for mechanical ventilation (propensity matched OR 8.87 (3.53–22.31), and respiratory failure (propensity matched OR 4.95 (1.83–13.41)))))))). There was no statistically significant difference in 30-day readmission rates between the two groups. Conclusions: SCAD is a unique condition that can lead to many complications. In our analysis, we showed that SCAD associated with cardiogenic shock compared to SCAD not associated with cardiogenic shock results in greater odds of complications including death, use of mechanical circulatory support, need for blood transfusions, ventricular tachycardia, acute kidney injury, use of mechanical ventilation, and respiratory failure. SCAD with cardiogenic shock represents a significantly critical clinical scenario that requires a multi-disciplinary approach to prevent the many potential complications associated with this disease process.

A growing body of research has begun to link exposure to environmental contaminants, such as heavy metals, with a variety of negative health outcomes. In this paper, we sought to review the current research describing the impact of certain common contaminant metals on cardiovascular (CV) health. We reviewed ten metals: lead, barium, nickel, chromium, cadmium, arsenic, mercury, selenium, zinc, and copper. After a literature review, we briefly summarized the routes of environmental exposure, pathophysiological mechanisms, CV health impacts, and exposure prevention and/or mitigation strategies for each metal. The resulting article discloses a broad spectrum of pathological significance, from relatively benign substances with little to no described effects on CV health, such as chromium and selenium, to substances with a wide-ranging and relatively severe spectrum of CV pathologies, such as arsenic, cadmium, and lead. It is our hope that this article will provide clinicians with a practical overview of the impact of these common environmental contaminants on CV health as well as highlight areas that require further investigation to better understand how these metals impact the incidence and progression of CV diseases.

AbstractCorticosteroid dosing in the range of 0.5–2 mg/kg/day of methylprednisolone equivalents has become a standard part of the management of intensive care unit (ICU) patients with COVID‐19 pneumonia based on positive results of randomized trials and a meta‐analysis. Alongside such conventional dosing, administration of 1 gm of methylprednisolone daily (pulse dosing) has also been reported in the literature with claims of favorable outcomes. Comparisons between such disparate approaches to corticosteroids for Coronavirus disease 2019 (COVID‐19) pneumonia are lacking. In this retrospective study of patients admitted to the ICU with COVID‐19 pneumonia, we compared patients treated with 0.5–2 mg/kg/day in methylprednisolone equivalents (high‐dose corticosteroids) and patients treated with 1 gm of methylprednisolone (pulse‐dose corticosteroids) to those who did not receive any corticosteroids. The endpoints of interest were hospital mortality, ICU‐free days at Day 28, and complications potentially attributable to corticosteroids. Pulse‐dose corticosteroid therapy was associated with a significant increase in ICU‐free days at Day 28 compared to no receipt: adjusted relative risk (aRR): 1.45 (95% confidence interval [CI]: 1.05–2.02; p = 0.03) and compared with high‐dose corticosteroid administration (p = 0.003). Nonetheless, receipt of high‐dose corticosteroids—but not of pulse‐dose corticosteroids—significantly reduced the odds of hospital mortality compared to no receipt: adjusted Odds ratio (aOR) 0.31 (95% CI: 0.12–0.77; p = 0.01). High‐dose corticosteroids reduced mortality compared to pulse‐dose corticosteroids (p = 0.04). Pulse‐dose corticosteroids—but not high‐dose corticosteroids—significantly increased the odds of acute kidney injury requiring renal replacement therapy compared to no receipt: aOR 3.53 (95% CI: 1.27–9.82; p = 0.02). The odds of this complication were also significantly higher in the pulse‐dose group when compared to the high‐dose group (p = 0.05 for the comparison). In this single‐center study, pulse‐dose corticosteroid therapy for COVID‐19 pneumonia in the ICU was associated with an increase in ICU‐free days but failed to impact hospital mortality, perhaps because of its association with development of severe renal failure. In line with existing trial data, the effect of high‐dose corticosteroids on mortality was favorable.

Objective: Multiple large population based observational studies have evaluated the association of various ECG changes with long term adverse outcomes demonstrating conflicting results regarding the significance of right bundle branch block (RBBB). Therefore, we conducted a systematic review and meta-analysis to determine the prognostic significance of RBBB in healthy population. Methods: This systematic review follows an a priori established protocol and is reported following the PRISMA statement (Preferred Reporting Items for Systematic Reviews and Meta-Analyses). We systematically searched EMBASE, Web of Science and Scopus through January 2014. We also consulted with content experts to recommend studies for screening. Reviewers working independently and in duplicate screened all eligible abstracts, selecting randomized trials and prospective cohort studies that described all cause mortality or cardiovascular death in patients with RBBB. We excluded studies that reported unadjusted outcome or unadjusted event rates. For statistical analysis, hazard ratios were considered to approximate risk ratios. The main outcomes of the study are all-cause mortality and cardiovascular mortality. Results: Based on the eligibility, four large cohorts, with total patient count > 100,000 were included in the meta-analysis. Risk of bias was assessed using the Newcastle-Ottawa scale and included studies were deemed to be of high quality. As demonstrated in Figure 1, RBBB was not associated increase in all-cause mortality (RR 1.05; 95% CI 0.89-1.25) or cardiovascular mortality (RR 1.09; 95% CI 0.70-1.69). Conclusion: In asymptomatic healthy patients, RBBB is not associated with poor outcome.

Background The association between the 9p21-3 locus and coronary heart disease (CHD) is well known. We performed a systematic review of the literature to assess whether this locus was associated with severity of coronary atherosclerosis. Data Sources Multiple relevant electronic databases were searched through September 2012 for English language studies with comparative design. Data extraction We tested the association of the 9p21-3 locus: 1) measures of severity of coronary atherosclerosis [number of diseased vessels, Gensini Score, Duke CHD Prognostic Index (Dc.PI)]. 2) angiographic outcomes [change in minimum lumen diameter ([[Unable to Display Character: ∆]]MLD) and number of new lesions at follow-up] and 3) key clinical outcomes (all-cause mortality, recurrent myocardial infarction and need for coronary revascularization). Genotypes were classified either as homozygous low risk (Lo) or homozygous high risk (Hi). Data Synthesis Relative risks (RR) were pooled using the random effects model. Seventeen distinct data sets that included 15,826 participants were analyzed. Ten studies (n=9807) reported number of diseased vessels at baseline. Hi genotype was associated with a 33.7% increased risk of triple vessel disease (95% CI 1.08-1.65; p=0.01). The Hi genotype was associated 0.27 higher baseline Gensini Score (95% CI 0.04-0.48; p=0.01) in 3787 individuals. The locus was not associated with Dc.PI [RR=0.21 (95% CI -0.14-0.55; p=0.25)]. The high risk genotype did not predict [[Unable to Display Character: ∆]]MLD or number of new lesions at follow-up. There was no difference between Hi and Lo genotypes in terms of all-cause mortality [RR=1.13 (95%CI 0.90-1.41;p=0.28)], recurrent myocardial infarction (95%CI 0.92-1.40; p=0.24) or frequency of coronary revascularization [RR=1.101 95% CI (0.78-1.57); p=0.56]. Conclusion An association between 9p21-3 and greater baseline severity of coronary atherosclerosis was seen in this meta-analysis. However no association was seen between 9p21-3 and adverse angiographic and clinical outcomes which may be due to diminishing genetic risk following dietary modification and statin therapy.

2054 Background: Systematic reviews that summarize the toxicity of Immune checkpoint inhibitors (ICIs) become outdated very soon after publication. Therefore, we reported results of a toxicity meta-analysis at 2019 ASCO meeting and informed the intent to create a living systematic review (LSR). LSRs combine human and machine effort and support rapid evidence synthesis and living clinical practice guidelines. Now, we report our experience maintaining a LSR on toxicity of ICIs. Methods: Steps include quarterly literature searches to identify new clinical trials reporting ICI-associated adverse events (AEs), AI-enabled screening of new citations which meet the inclusion criteria, automated cumulative meta-analysis and an online reporting platform. Standard data formats and protocols were designed for inputting text, tables and graphics. Software was written to interpret these data and output the information in the appropriate format, such as a forest plot and summary tables. Finally, a dynamic interface that enables user inputs and displays the associated output was designed. Results: The LSR is continuously updated incorporating toxicity data from new clinical trials as it becomes available. We have screened 8000 relevant citations and summarized the odds of Grade 3 or higher AEs and AEs of special interest in patient receiving ICIs. The results are updated on quarterly basis and are available online. The results are updated on quarterly basis and will be available on a website at the time of publication. Prototype with dummy data is available at this link . This interface can also be manipulated via user input to organize and sort data tables and forest plots by type of cancer, name or mechanism (PD-1 or PD-L1) of ICI agent, single agent or combination, type of control arm, line of treatment and several other clinically relevant filters. For example, a user can instantaneously generate a meta-analysis summarizing the risk of colitis or pneumonitis in metastatic lung cancer trials with pembrolizmuab. Conclusions: This LSR engine can prospectively synthesize toxicity data from ICI trials in an efficient manner providing accurate and timely information for advanced clinical decision support at point-of-care. Efforts are ongoing to improve efficiency of screening, improve AI-enabled processes for automated screening and data abstraction, and test across multiple clinical questions.

Background: Four large RCTs (ASSURE, S-TARC, PROTECT, ATLAS) tested adjuvant VEGF-TKI therapy in high risk RCC. The results were variable for efficacy and there were concerns for increased toxicity and decline in quality of life (QoL). We performed an updated meta-analysis including results of ATLAS trial to asses a risk-benefit for adjuvant post-operative treatments in high risk RCC patients by assessing reported disease-free survival (DFS), overall survival (OS), and toxicity endpoints. Methods: Literature search was done using Medline, CENTRAL, and Embase. The DerSimonian and Laird random effects model was used to pool estimates for DFS, OS, and common side effects across the 4 trials. A subgroup analysis was performed for sunitinib alone because of its FDA approval. Heterogeneity was assessed with Cochrane Q statistic and was quantified with I2 test. Risk for bias was assessed using the Cochrane Collaboration’s tool. Results: The 4 RCTs included 4,820 patients. Adjuvant therapy with TKIs yielded no significant improvement in DFS or OS as compared to placebo (DFS HR=0.916; 95% CI, 0.832–1.009 and OS HR=1.09; 95% CI, 0.886–1.150). Separate analysis of DFS in sunitinib vs placebo did not show any benefit (2 studies, N=1,909; HR=0.90; 95% CI, 0.67–1.19). Use of TKIs was associated with significantly increased risk of drug toxicity. Increased risk of grade 3 or 4 adverse events (RR=5.110; 95% CI, 3.765–6.935), diarrhea (RR=10.725; 95% CI, 4.672–24.622), fatigue (RR=3.310; 95% CI, 1.879–5.829), hypertension (RR=4.274; 95% CI, 3.452–5.292) and palmar/plantar dysesthesia (RR=20.53; 95% CI, 9.006–46.801) was observed. There was no statistically significant heterogeneity amongst included trials. QoL endpoints were inconsistently reported. Risk of bias was low. Conclusions: This pooled analysis provides further evidence that there is no OS or DFS benefit associated with adjuvant TKI treatment. There was a significantly increased risk of grade 3 or 4 toxicity in greater than half of the patient population leading to decline in QoL during TKI therapy. Carefully selected very high-risk patients who can tolerate these agents without dose modifications may benefit from adjuvant TKI approach.