Elizabeth Phillips

Large language models (LLMs) have emerged as transformative technologies, revolutionising natural language understanding and generation across various domains, including medicine. In this study, we investigated the capabilities, limitations, and generalisability of Generative Pre-trained Transformer (GPT) models in analysing unstructured patient notes from large healthcare datasets to identify immune-related adverse events (irAEs) associated with the use of immune checkpoint inhibitor (ICI) therapy.
We evaluated the performance of GPT-3.5, GPT-4, and GPT-4o models on manually annotated datasets of patients receiving ICI therapy, sampled from two electronic health record (EHR) systems and seven clinical trials. A zero-shot prompt was designed to exhaustively identify irAEs at both the patient level (main analysis) and the note level (secondary analysis). The LLM-based system followed a multi-label classification approach to identify any combination of irAEs associated with individual patients or clinical notes. System evaluation was conducted for each available irAE as well as for broader categories of irAEs classified at the organ level.
Our analysis included 442 patients across three institutions. The most common irAEs manually identified in the patient datasets included pneumonitis (N = 64), colitis (N = 56), rash (N = 32), and hepatitis (N = 28). The GPT models demonstrated generalisable abilities in identifying irAEs across EHRs and clinical trial reports. Overall, the models achieved relatively high sensitivity and specificity but only moderate positive predictive values, reflecting a potential bias towards overpredicting irAE outcomes. GPT-4o achieved the highest F1 and micro-averaged F1 scores for both patient-level and note-level evaluations. Highest performance was observed in the haematological (F1 range = 1.0–1.0), gastrointestinal (F1 range = 0.81–0.85), and musculoskeletal and rheumatologic (F1 range = 0.67–1.0) irAE categories. Error analysis uncovered substantial limitations of GPT models in handling textual causation, where adverse events should not only be accurately identified in clinical text but also causally linked to immune checkpoint inhibitors.
This study demonstrated that GPT models can automate the detection of immune related adverse events in varied healthcare datasets, reducing the burden on physicians and other healthcare professionals by limiting the need for manual review. This capability will accelerate the generation of safety insights across large healthcare datasets and facilitate the characterisation of patient-level drivers of toxicities, thus enhancing safety monitoring and ultimately improving patient care.
National Institutes of Health, Roche, National Health and Medical Research Council of Australia, Stevens-Johnson Syndrome Foundation, Angela Anderson Research Fund, Larry L Hillblom Foundation and UCSF Research Allocation Program.

Immune checkpoint inhibitors are used in a wide range of cancers, offering durable responses for a substantial subset of patients. However, immune-related adverse events, the most clinically consequential checkpoint inhibitor–associated adverse reactions, pose a key challenge in practice, affecting virtually any organ system, resulting in treatment interruption, morbidity, or mortality. Patient education, early recognition, and effective management are essential to limit complications and maintain continuity of immunotherapy. Achieving this requires well-informed multidisciplinary teams who can identify, evaluate, and manage immune-related adverse events promptly. This review summarizes the most clinically significant immune-related adverse events and highlights the key principles of multidisciplinary diagnosis and management most relevant to the practicing allergist-immunologist to optimize patient outcomes.

Background
Co-trimoxazole is a leading global cause of severe cutaneous adverse drug reactions (SCAR) including Stevens-Johnson syndrome/toxic epidermal necrolysis (SJS/TEN) and drug reaction with eosinophilia and systemic symptoms (DRESS). Co-trimoxazole-induced SCAR are associated with HLA class I alleles including HLA-B*13:01 and HLA-B*38:02 in Southeast Asian (SEA) populations. However, the global generalizability of these associations is unknown but critical for population-appropriate risk stratification and diagnosis.

Objective
To determine HLA risk factors associated with co-trimoxazole-induced SJS/TEN and DRESS in populations from the United States (US) and South Africa (SA).

Methods
We performed high-resolution HLA typing on dermatologist-adjudicated co-trimoxazole-induced SCAR patients in the US (n=63) and SA (n=26) compared to population controls. Peptide binding and docking analyses were performed using MHCcluster2.0 and CB-Dock2.

Results
In a multiple logistic regression model, HLA-B*44:03 (Pc<0.001, OR: 4.08), HLA-B*38:01 (Pc<0.001, OR: 5.66), and HLA-C*04:01 (Pc=0.003, OR: 2.50) were independently associated with co-trimoxazole-induced SJS/TEN in the US. HLA-B* 44:03 was also associated with co-trimoxazole-induced DRESS in SA (Pc=0.019, OR: 10.69). Distinct HLA-B variants with shared peptide binding specificities (SPBS) and HLA-C*04:01 identified 94% and 78% of co-trimoxazole-induced SJS/TEN and DRESS in the US, respectively. The SEA risk allele HLA-B*13:01, with SPBS to HLA-B*44:03, was identified in just 1/63 US SCAR patient.

Conclusion
HLA alleles with SPBS to SEA-related risk alleles including HLA-B*44:03 (SPBS with HLA-B*13:01) and HLA-B*38:01 (SPBS with HLA-B*38:02) but also HLA-C*04:01 predisposed to co-trimoxazole-induced SCAR in the US and SA. These findings provide biological plausibility and strategies for global risk prediction and diagnosis of co-trimoxazole-induced SCAR.

Inaccurate penicillin allergy labels (PALs) affect antimicrobial stewardship, health outcomes, and costs. More than 95% of PALs can be de-labeled when tested, but this rarely happens.
We sought to determine whether education and an electronic health record (EHR) tool kit to identify low-risk PALs would facilitate inpatient penicillin allergy de-labeling by pharmacists.
Pragmatic Removal of Penicillin Allergy Electronic Health Record Labels was a stepped-wedge, nonblinded, randomized, controlled, pragmatic clinical trial including 12 inpatient medical units. From November 2020 to November 2021, units entered intervention at 1-month stepped intervals in random order. Medically stable, nonpregnant adults with an her-documented PAL who were hospitalized on intervention units for 24 hours were included. The intervention included pharmacist and nursing education, a patient list for systematic identification of PALs, presentation of a risk-assessment tool, and an oral amoxicillin challenge order set. The primary outcome was removal of PAL by hospital discharge. Secondary outcomes included safety and implementation measures, longer-term continued label removal, and antibiotic use.
On 12 randomized units, 2,052 patients with a PAL presented during the 1-year trial. More intervention-exposed patients had EHR-documented penicillin allergy removal compared with controls (45 of 1,018 [4.4%] vs 31 of 1,034 [3%], respectively; adjusted odds ratio = 2.05; 95% CI, 1.08-3.91). More intervention-exposed patients received an EHR-documented penicillin allergy risk assessment than did controls (86 of 1,018 [8.4%] vs 27 of 1,034 [2.6%]; adjusted odds ratio = 6.42; 95% CI, 3.08-13.38). Moreover, 27 of 1,018 intervention patients (2.7%) received amoxicillin challenge compared with 21 of 1,034 control patients (2.0%). All amoxicillin challenges were tolerated.
Education and deployment of an inpatient EHR tool kit increased the rate of inpatient removal of inaccurate penicillin allergies before discharge.

Consensus guidelines outline the recommended management of delayed drug hypersensitivity reactions (DHRs), but clinical practices are largely expert-based and may vary significantly across populations and regions. This study evaluated the approach and management of health care specialists globally regarding DHRs, including severe cutaneous adverse reactions.
To assess the current practices, knowledge, and availability of allergy testing for DHRs, including severe cutaneous adverse reactions, among relevant health care professionals globally.
The Global AppRoAch for Severe Cutaneous Adverse Reactions Survey study is a prospective, web-based survey conducted using the Research Electronic Data Capture platform. The 10- to 15-minute survey was directed to 14 allergy/immunology and dermatology associations through international mailing lists across 48 countries.
Among the 8561 members contacted, answers from 130 health care providers practicing in all 6 continents were analyzed. Most respondents identified as White (88 of 130 [68%]) and female (79 of 130 [61%]), representing a range of age groups and clinical expertise. Most practiced in dermatology (69 of 130 [53%]), followed by allergy/immunology (46 of 130 [35.4%]), and were based in an urban setting (127 of 130 [98%]). Among the 45 respondents (34.6%) who managed severe cutaneous adverse reactions, 27 (60%) had access to multidisciplinary care. Practices varied widely by clinical phenotype; for example, for maculopapular exanthema, 63 respondents indicated treating through the reaction (63 of 130 [49%]), whereas 16 (16 of 130 [12%]) offered desensitization. Under patient education and follow-up, only 60 (60 of 130 [46.2%]) participants organized a follow-up with the implicated specialist.
Significant variability in the management of DHRs was observed among the respondents. The results highlight the need for evidence-based protocols to standardize guidelines for managing DHRs.

Importance Stevens-Johnson syndrome and toxic epidermal necrolysis (SJS/TEN) survivors experience substantial long-term sequelae. Research on physical symptoms experienced during acute hospitalization is well documented, but limited studies have been completed on the long-term biopsychosocial effects of SJS/TEN, particularly from the patient’s perspective.

Objective To increase the understanding of the long-term complications of SJS/TEN.

Design, Setting, and Participants This qualitative investigation was completed from within a community-based study, the SJS Survivors Study, using a semistructured, in-depth interview guide to query participants about their SJS/TEN experience postdischarge from the hospital. Interviews took place by phone from July 2021 through August 2023. This study included adults who experienced SJS/TEN within the United States.

Main Outcome and Measures A biopsychosocial theory-based framework and hierarchical coding system were utilized to understand the long-term life impacts of survivors of SJS/TEN.

Results The 29 participants, aged 26 to 76 years, were 66% female and 69% White and had experienced SJS/TEN from a wide range of drugs. Patients experienced support while in the hospital, but once discharged, felt isolated and without support to understand the potential sustained impacts of SJS/TEN in their lives and the lives of their family members. Patients experienced ongoing biological symptoms, such as skin issues, debilitating visual impairment, blindness, and lack of functional autonomy. Psychological impacts included symptoms of anxiety, obsessive thinking, flashbacks, and depression. Socially, some survivors expressed a sense of abandonment and described negative impacts on their careers. Survivors also expressed frustration and isolation with having to navigate posthospital care alone. There was a lack of preemptive discharge education and SJS/TEN-specific planning. Lack of physician knowledge about SJS/TEN was particularly noted and survivors turned to the internet for guidance instead of receiving direction from their physicians. Medical distrust among survivors was frequently noted.

Conclusions and Relevance The findings highlight the need for postdischarge care coordination among patients and their primary physicians, including mental health support. This care coordination should be arranged prior to discharge to ensure the availability of adequate support and optimal health outcomes. It is essential that clinicians and researchers prioritize the understanding of long-term sequelae of SJS/TEN and improve current discharge education and protocols for patients and their families.

Cytomegalovirus (CMV) establishes lifelong latency and is linked to immunosenescence in older and immunocompromised individuals. We hypothesize that CMV drives systemic and tissue-specific immune changes that may contribute to cardiovascular disease (CVD). Thoracic aorta, blood and perivascular mediastinal adipose tissue from cardiac surgery patients (n = 11) were processed within 30–60 min of excision. CMV IgG titres were quantified through ELISA to determine CMV status: CMV(−) (n = 4) and CMV(+) (n = 7). Immune profiling was performed using flow cytometry and single-cell RNA sequencing. Analyses included MiloR and differential gene expression. Participants (mean age 69.7 ± 8.4 years) were 80% male and 70% Caucasian. CMV(−) and CMV(+) participants had mean IgG titres of 0.038 and 13.55 IU ml−1, respectively. CD8+ T-cells expressing CD57+, GPR56+ and CX3CR1+ (CGC) were increased in the blood of CMV(+) participants. In the aorta of CMV(+) participants, CD8+ T cells and CD4+ T cells had decreased HLA-C expression and suppressed interferon-α pathways. In contrast, the TNF-α signalling pathway was increased. CMV infection shapes immune responses and in this pilot, we observed suppression of interferon-α signalling and increased TNF-α-associated pathways in the aorta. Larger studies are needed to define how CMV-driven immune remodelling contributes to CVD.

This article is part of the discussion meeting issue ‘The indirect effects of cytomegalovirus infection: mechanisms and consequences.’

The COVID-19 pandemic highlighted the critical need for vaccine strategies capable of addressing emerging viral threats. Betacoronaviruses, including severe acute respiratory syndrome coronavirus (SARS-CoV), Middle East respiratory syndrome (MERS), and SARS-CoV-2, present significant pandemic risks due to their zoonotic potential and genetic diversity. T cell-mediated immunity has demonstrated durable responses and strong cross-reactivity, offering a promising avenue for achieving broad immunity within a viral family. In this study, we combined comprehensive epitope mapping with sequence conservation analyses to identify conserved T cell epitope regions (CTERs), which constitute 12% of the complete SARS-CoV-2 proteome. We showed that SARS-CoV-2 CTER-specific T cells cross-reactively recognize sequences from multiple Betacoronavirus subgenera. Importantly, incorporating CTERs from non-spike proteins significantly enhanced T cell cross-reactivity potential and human leukocyte antigen (HLA) coverage compared with T cells targeting only spike proteins. Our findings lay the groundwork for a multi-antigen vaccine strategy that includes non-spike proteins to expand cross-reactive immunity across a broader spectrum of Betacoronaviruses.

Amyotrophic lateral sclerosis (ALS) is a neurodegenerative disease characterized by a progressive loss of motor neurons. Neuroinflammation is apparent in affected tissues, including increased T cell infiltration and activation of microglia, particularly in the spinal cord
. Autoimmune responses are thought to have a key role in ALS pathology, and it is hypothesized that T cells contribute to the rapid loss of neurons during disease progression
. However, until now there has been no reported target for such an autoimmune response. Here we show that ALS is associated with recognition of the C9orf72 antigen, and we map the specific epitopes that are recognized. We show that these responses are mediated by CD4
T cells that preferentially release IL-5 and IL-10, and that IL-10-mediated T cell responses are significantly greater in donors who have a longer predicted survival time. Our results reinforce the previous hypothesis that neuroinflammation has an important role in ALS disease progression, possibly because of a disrupted balance of inflammatory and counter-inflammatory T cell responses
. These findings highlight the potential of therapeutic strategies aimed at enhancing regulatory T cells
, and identify a key target for antigen-specific T cell responses that could enable precision therapeutics in ALS.