Fleti, Farah

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.

Abstract Background Partial breast irradiation (PBI) is an effective treatment for early-stage breast cancer. However, evidence comparing different PBI modalities is limited. Methods This trial emulation uses electronic health records from a multi-state large tertiary health system. Three PBI modalities were compared: proton radiation therapy, photon radiation therapy, and applicator-based brachytherapy. Inverse probability weighting (IPW) was used to balance confounders. The primary outcome was ipsilateral breast recurrence (IBR). Results Between 2013 and 2023, 1041 women with early-stage breast cancer were treated with PBI. Of these, 1026 patients were included in the analyses. The median age was 66 years; 22.51% had ductal carcinoma in situ (DCIS), and 97.56% were estrogen receptor positive. With median follow-up of 38.2 months, the rate of IBR in the overall group was 3.05% (95% confidence interval [CI] = 1.99% to 4.65%) at 3 years. After IPW, compared with photon PBI, the hazard ratio for IBR for proton PBI was 2.11 (95% CI = 0.70 to 6.33) and for brachytherapy, 3.86 (95% CI = 1.39 to 10.69). Compared with proton PBI, the hazard ratio for IBR for brachytherapy was 1.83 (95% CI = 0.88 to 3.81). IBR risk was similar across PBI modalities for patients with tumor size ≤ 10 mm and in patients treated for DCIS. Among patients with tumor size >10 mm, the hazard ratio for IBR for brachytherapy was 7.64 (95% CI = 1.64 to 35.58) and 3.59 (95% CI = 1.22 to 10.56) relative to photons and protons, respectively. Conclusion PBI with applicator-based brachytherapy was associated with higher IBR than photon or proton PBI in patients with tumor size > 10 mm, suggesting the need for more careful PBI patient selection for this modality.

Background and Aims: With rapid changes in the management landscape of chronic hepatitis B (CHB), this technical systematic review addresses 4 critical Population, Intervention, Comparator, Outcome (PICO) questions to provide guidance to the formulation of recommendations to the 2025 AASLD practice guideline for management of CHB. Approach: The review was reported in accordance with PRISMA guidelines. Outcomes were evaluated across 4 key PICOs: (1) antiviral therapy for prevention of horizontal HBV transmission in high-risk groups, (2) antiviral therapy versus observation for persons in the immune-tolerant phase, (3) discontinuation versus continuation of nucleos(t)ide analogue therapy in HBeAg-negative individuals with undetectable HBV DNA, and (4) hepatocellular carcinoma (HCC) surveillance in non-cirrhotic individuals with HBsAg clearance or co-infections with HCV, HDV, or HIV. Results: For PICO 1, limited evidence suggests antiviral therapy may reduce horizontal transmission risk, though with low certainty. PICO 2 analyses reveal uncertain benefits of treating persons in the immune-tolerant phase, with very low certainty due to heterogeneity and bias. PICO 3 analyses demonstrate that discontinuing antiviral therapy in persons who are HBeAg negative with undetectable HBV DNA increases HBsAg loss rates (OR: 12.65, 95% CI: 1.58–101.51) but carries moderate risks of virologic relapse (OR: 47.17, 95% CI: 2.79–797.35) and clinical flares. PICO 4 analyses on several cohort studies showed that in patients co-infected with HCV, HBV, or HIV, the annual incidence of HCC was higher than the level where screening becomes cost-effective, suggesting that regular liver cancer screening could be beneficial for these patients. Conclusions: Despite low certainty, the findings support shared decision-making in high-risk horizontal transmission scenarios or in treating individuals in the immune tolerance phase, caution in discontinuing antiviral therapy in virologically suppressed individuals without HBsAg loss, and tailored HCC surveillance for those with co-infection or cirrhosis.

Importance Suicide is a leading cause of death among young people and an escalating public health crisis. Objective To assess the effectiveness and harms of available treatments for suicidal thoughts and behaviors among youths at heightened risk for suicide. Evidence Review A systematic review was conducted searching databases including MEDLINE, Embase, Cochrane Central Register of Controlled Trials, Cochrane Database of Systematic Reviews, Scopus, and various gray literature sources from January 1, 2000, to September 26, 2024. We included randomized clinical trials, comparative observational studies, and before-after studies of psychosocial interventions, pharmacologic interventions, neurotherapeutics, emerging therapies, and combination therapies. Eligible patients were youths (aged 5-24 years) with heightened risk for suicide, adolescents from racial and ethnic minority groups known to be at increased risk of suicide, or those exposed to crime or violence. Pairs of independent reviewers selected and appraised studies. Findings This review included 65 studies (33 randomized clinical trials, 13 comparative observational studies, and 19 before-after studies) reporting on 14 534 patients (median age, 15.1 years; 75.1% female patients). Psychosocial interventions comprised psychotherapy interventions (33 studies; cognitive behavioral therapy, dialectical behavioral therapy, collaborative assessment and management of suicidality, dynamic deconstructive psychotherapy, attachment-based family therapy, and family-focused therapy), acute (ie, 1-4 sessions or contacts) psychosocial interventions (19 studies; safety planning, family-based crisis management, motivational interviewing crisis interventions, continuity of care after crisis, and brief adjunctive treatments), and school- or community-based psychosocial interventions (13 studies; social network interventions, school-based skills interventions, suicide awareness or gatekeeper programs, and community-based, culturally tailored adjunct programs). Dialectical behavior therapy showed moderate strength of evidence for reducing suicidal ideation. Other categories of psychosocial treatments showed insufficient to low strength of evidence for reducing suicidal outcomes. None of the studies evaluated adverse events. The evidence base on pharmacologic treatment for youths at risk of suicide was largely nonexistent. Conclusions and Relevance This systematic review found that the current evidence on available interventions targeting youths at heightened risk of suicide is uncertain. Medication, neurotherapeutics, and emerging therapies remain unstudied among this population. The limited evidence base calls for the development of novel, developmentally and trauma-informed treatments, as well as multilevel interventions to target the increasing suicide risk among youths.

SummaryAmong 210 patients with myelodysplastic syndromes (MDSs) with del(5q), molecular information was available at diagnosis or at least 3 months before leukaemic transformation in 146 cases. Multivariate analysis identified therapy‐related setting (p = 0.02; HR 2.3) and TP53 variant allele frequency (VAF) ≥22% (p < 0.01; HR 2.8), but not SF3B1 mutation (p = 0.65), as independent risk factors for survival. Median survival was 11.7 versus 4 years (5/10‐year survival 73%/52% vs. 42%/14%) in the absence (N = 112) versus presence (N = 34) of ≥1 risk factors; leukaemia‐free survival was affected by TP53 VAF ≥22% (p < 0.01). Such information might inform treatment decision‐making in MDS‐del(5q) regarding allogeneic stem cell transplant.

Background: According to the international consensus classification (ICC), diagnosis of myelodysplastic syndromes (MDS) with del(5q) requires presence of i) del5q, alone or associated with no more than one other cytogenetic abnormality, exclusive of -7/del(7q), ii) &lt;5% bone marrow (BM) and &lt;2% circulating blasts, and iii) the absence of “multi-hit” TP53 ( Arber et al. Blood 2022;140: 1200). We have recently reported on TP53 variant allele frequency (VAF) ≥22% being the sole genetic predictor of inferior survival in MDS-del(5q) (Fleti. et al. AJH 2023;98:E76). In the current study, we examined the additional prognostic impact of therapy-related setting, in the context of TP53 VAF and other risk factors. Methods : The current study included patients with MDS-del(5q) who were seen at the Mayo Clinic or Moffitt Cancer Center, USA. Diagnosis of MDS-del(5q) and therapy-related qualification was according to ICC criteria ( Arber et al. Blood 2022;140: 1200), with the exception that patients with biallelic or multi-hit TP53 were included to allow unbiased analysis of impact on prognosis. Patient selection was based on availability of mutation information from next-generation sequencing (NGS). The latter was assessed either at diagnosis or at different time points during the disease course, but at least 3 months before leukemic transformation. Otherwise, all analyses, including estimation of overall and leukemia-free survival, were based on variables collected at time of initial diagnosis of MDS-del(5q). TP53 mutation VAF threshold of 22% was considered based on our previously published report cited above. Age cutoff at 67 years was determined by ROC analysis. Conventional statistical methods were applied using JMP Pro 16.0.0 software (SAS Institute, Cary, NC, USA). Results: A total of 210 patients (median age 72 years; 65% females), meeting the above-stipulated criteria for MDS-del(5q), were considered: 172 primary and 38 therapy-related. Table 1 outlines clinical and laboratory characteristics obtained at the time of diagnosis: 42% were transfusion-dependent, 69% had hemoglobin level &lt;10 g/dL, 30% platelet count &lt;150 x 10(9)/L, and 15% carried additional cytogenetic abnormality. NGS data performed either at diagnosis or at least 3 months prior to leukemic transformation was available in 146 patients; the most frequent mutations were TP53 (21%), SF3B1 (16%), DNMT3A (12%), TET2 (12%), ASXL1 (10%), and JAK2 (7%) (Table 1). At a median follow-up of 3.9 years (range 0.002-18.5) for living patients, 95 (45%) deaths and 34 (16%) leukemic transformations were recorded; treatment undertaken included lenalidomide in 162 (77%) cases. NGS data-exclusive multivariable analysis that included all 210 patients identified therapy-related setting (p&lt;0.01; HR 2.9; Figure 1a), age &gt;67 years (p=0.01; HR 1.8), transfusion-dependency (p=0.04; HR 1.5), and platelet count &lt;150 x 10(9)/L (p=0.03; HR 1.6), as independent predictors of inferior survival. The significant effect of age and platelet count was sustained when analysis was limited to patients with primary MDS-del(5q) but lost when applied to therapy-related disease. NGS data-inclusive multivariable analysis, which was applicable to the subset of 146 patients, identified, as risk factors, therapy-related setting (p&lt;0.01; HR 2.3) and TP53 VAF ≥22% (p&lt;0.01; HR 2.4; Figure 1b) while age &gt;67 years sustained borderline significance (p=0.07; HR 1.8). Figure 1c depicts survival data based on the presence or absence of TP53 VAF ≥22% and therapy-related qualification; low-risk with no risk factors (N=112; median survival 11.7 years; 5/10-year survival 73%/52%) and high-risk with one or both risk factors (N=34; median survival 4 years; 5/10-year survival 42%/14%); leukemia-free survival was affected by TP53 VAF ≥22% (p&lt;0.01) but not therapy-related setting (p=0.13). Conclusions : The current study highlights TP53 VAF and therapy-related setting as the most prominent risk factors for survival in MDS-del(5q) and proposes a simple risk model that distinguishes patients with a projected 10-year survival that exceeds 50%, in the absence of both risk factors, and plummets to 14%, in the presence of one of the two risk factors. Accordingly, allogeneic stem cell transplant might be deferred in the former instance while it should be pursued sooner than later, in the latter instance.

Abstract We surveyed the performance of ponatinib, as salvage therapy, in a real-world setting of chronic phase chronic myeloid leukemia (CML-CP). Among 55 consecutive patients (median age 49 years) with relapsed/refractory CML-CP, 35 (64%) had failed ≥ 3 tyrosine kinase inhibitors (TKIs), 35 (64%) were pre-treated with nilotinib, and 14 (28%) harbored ABL1T315I. At start of ponatinib (median dose 30 mg/day), 40 patients were already in complete hematologic (CHR), 4 in complete cytogenetic (CCyR), 3 in major molecular (MMR) remission, while 8 had not achieved CHR (NR). Ponatinib improved the depth of response in 13 (33%), 3 (75%), 2 (66%), and 4 (50%) patients with CHR, CCyR, MMR, and NR, respectively (p = 0.02). At a median follow-up of 42 months, 13 (23%) deaths, 5 (9%) blast transformations, and 25 (45%) allogeneic transplants were recorded. Five/10-year post-ponatinib survival was 77%/58% with no significant difference when patients were stratified by allogeneic transplant (p = 0.94), ponatinib-induced deeper response (p = 0.28), or a post-ponatinib ≥ CCyR vs CHR remission state (p = 0.25). ABL1T315I was detrimental to survival (p = 0.04) but did not appear to affect response. Prior exposure to nilotinib was associated with higher risk of arterial occlusive events (AOEs; 11% vs 0%; age-adjusted p = 0.04). Ponatinib starting/maintenance dose (45 vs 15 mg/day) did not influence either treatment response or AOEs. Our observations support the use of a lower starting/maintenance dose for ponatinib in relapsed/refractory CML-CP but a survival advantage for deeper responses was not apparent and treatment might not overcome the detrimental impact of ABL1T315I on survival. The association between prior exposure to nilotinib and a higher risk of post-ponatinib AOEs requires further validation.