Elizabeth Phillips

HLA class I alleles confer a striking risk for T cell–mediated drug hypersensitivity, yet positive predictive values are low—typically under 10% and as low as 0.12% for some drug–HLA pairs. We propose that persistent, human-adapted pathogens—notably herpesviruses—focus postnatal immune memory on conserved epitopes in the tissue niches where viral control occurs (the Keystone Epitope Theory). The phylogenetic basis for this proposal is that herpesviruses and their vertebrate hosts have co-adapted over hundreds of millions of years, and that this co-evolutionary relationship is replayed ontogenetically as each individual acquires these infections and builds tissue-specific immune memory. When a drug-altered self-peptide approximates the geometry of such a target and is presented by the same risk HLA in the same niche at sufficient density, pre-existing tissue-resident memory T cells (TRM) may be recruited, breaching local regulatory equilibria and driving immunopathology. We synthesize three strands of evidence: (i) heterologous immunity, in which virus-imprinted TRM cross-recognize drug-modified self; (ii) antigen presentation in the same tissue where antiviral memory already resides, which helps explain why injury is tissue-restricted; and (iii) public and private TCR solutions that bridge viral and self-targets. Beginning with T cell–mediated drug hypersensitivity as an empirical anchor, we extend this framework to EBV-associated multiple sclerosis and transplant rejection and conclude with proposed experimental validation strategies that may be applicable more broadly to T cell–mediated hypersensitivity and autoimmunity. From keystone protection to clinical misdirection. Postnatal immune focusing concentrates durable tissue-resident memory on conserved peptide–HLA targets in specific niches. Modified self-peptides (drug-altered or post-translationally modified) that approximate this geometry, presented by the same risk HLA in the same tissue, can recruit entrenched antiviral programs and, when mimicking ligand burden is sufficient, exceed local inhibitory set-points. The clinical result is high NPV but low PPV for HLA risk alleles. Abbreviations: TRM, tissue-resident memory T cell; PTM, post-translational modification.

Importance Antiepileptic drugs (AEDs) are frequently implicated in Stevens-Johnson syndrome (SJS) and toxic epidermal necrolysis (TEN), which are the most severe type of drug hypersensitivity reaction, with a mortality rate up to 50%. However, the overall proportion of these cases attributed to AEDs has not been systematically reviewed. Objective To evaluate the proportion of AED-associated SJS/TEN. Data Sources The MEDLINE and Embase databases were searched from inception through March 17, 2026, with results limited to English-language publications and human participants. Study Selection All experimental and observational studies that reported patient-level triggers of SJS/TEN were included. Data Extraction and Synthesis Two reviewers independently screened studies and extracted data using a predefined template. To estimate the proportion of AED-associated SJS/TEN, a random-effects meta-analysis was performed, calculating the pooled proportions with 95% CIs. Subgroup analyses by age group (children vs adults) and geographic region were performed to explore heterogeneity. Main Outcomes and Measures To estimate the proportion of AED-associated SJS/TEN. Results This systematic review and meta-analysis included 50 studies with 4403 patients that reported patient-level triggers of SJS/TEN. Single-drug triggers accounted for 88% (95% CI, 83-92) of cases. The pooled proportion of SJS/TEN cases attributed to AEDs was 23% (95% CI, 19-26). Nearly all AED-associated cases were due to aromatic AEDs (96%; 95% CI, 92-99), most commonly carbamazepine (39%; 95% CI, 29-50), followed by phenytoin (19%; 95% CI, 11-28), and lamotrigine (15%; 95% CI, 8-22). Nonaromatic AEDs were rarely implicated (2%; 95% CI, 0-4). Significant geographic variation was observed in the proportion of AED-associated SJS/TEN, ranging from 42% (95% CI, 26-59) in West Asia to 10% (95% CI, 0-33) in Africa (P < .001). The proportion of AED-associated SJS/TEN cases was significantly higher in children (35%; 95% CI, 22-48) than in adults (23%; 95% CI, 20-27). Conclusions and Relevance In this systematic review and meta-analysis of observational studies, AEDs accounted for nearly one-quarter of SJS/TEN cases reported worldwide, predominantly involving aromatic AEDs, such as carbamazepine, phenytoin, and lamotrigine. These findings suggest aromatic AEDs as leading contributors to SJS/TEN and support the preferential use of nonaromatic AEDs when clinically appropriate.

Background
Inaccurate multiple antibiotic allergy labels to first-line agents (penicillins, cephalosporins, and sulfonamides) restrict optimal therapy and often require sequential evaluations across multiple visits, delaying delabeling and increasing loss to follow-up.

Objective
To conduct a comparative effectiveness study of the Multiple Antibiotic Allergy Evaluation Strategy (MAAES), comparing its effectiveness, safety, and patient-reported outcomes (PRO) to traditional sequential evaluation (non-MAAES).

Methods
This retrospective cohort study (2014–2024) compared same-day multi-antibiotic testing (MAAES) with non-MAAES in patients with ≥2 low-risk labels. Multivariable regression adjusted for baseline differences and testing era. Inverse probability of treatment weighting (IPTW) and a restricted post-2020 subcohort analysis addressed selection bias and temporal shifts. Outcomes included delabeling efficacy, time to complete delabeling, loss to follow-up, relabeling, adverse events (AEs), and PRO.

Results
Baseline differences were present between the MAAES and non-MAAES groups, including penicillin anaphylaxis (1.4%vs.6.6%, p=0.002), angioedema (12.4%vs.19.1%, p=0.049), maculopapular rash (13.1%vs.24.3%, p=0.002), and unknown sulfonamide history (9.5%vs.3.7%, p=0.03). MAAES removed 97.4% of labels vs 79.6% for non-MAAES (adjusted odds ratio(aOR) 9.41, 95%CI 5.22-16.95, p<0.001). MAAES achieved faster complete delabeling (log-rank χ2=42.4, p<0.001), with 84.9%vs.21.1% delabeled at the initial visit (restricted mean survival time difference: 2.1 months). Results remained robust in post-2020 and IPTW models. Loss to follow-up (1.3%vs.17.2%; aOR 0.06, 95%CI 0.03–0.13; p<0.001) and AEs (<1%vs.2.8%; aOR 0.26, 95%CI 0.09–0.78; p=0.02) were significantly lower with MAAES. Favorable PRO was sustained, with no significant differences between groups.

Conclusions
A consolidated multi-drug evaluation strategy substantially improves effectiveness of first-line antibiotic allergy delabeling while maintaining excellent safety profiles and PRO.

Paramyxoviruses comprise a diverse family of viruses that threaten global human health through direct infection and zoonotic transmission. Understanding adaptive immune responses to these viruses is critical for characterizing host-pathogen interactions and evaluating vaccine performance. Here, we systematically map human CD4+ T cell epitopes across Nipah and measles viruses, two prototypic members of the Paramyxoviridae family. We identify broad epitope repertoires, including 186 Nipah and 288 measles epitopes recognized in multiple donors. Epitopes are characterized for HLA binding and inferred restrictions, and broader HLA binding correlates with immunodominance. We observe overlapping T cell targets between viruses, with N and F proteins immunodominant in both and L additionally dominant in Nipah. We define conserved T cell epitope regions (CTERs) in Nipah virus that encompass 17% of the proteome, capture over 50% of T cell responses, show high conservation across different Henipaviruses, and elicit broadly cross-reactivity, supporting broad population coverage.
[Display omitted]
•Systematic mapping reveals broad CD4+ T cell epitopes in Nipah and measles•HLA binding promiscuity correlates with epitope immunodominance•Conserved T cell epitope regions (CTERs) capture over 50% of Nipah T cell responses•CTERs elicit cross-reactive T cells across diverse Paramyxoviruses
Tarke et al. systematically map CD4+ T cell epitopes in measles and Nipah viruses, revealing broad and overlapping immune targets. They define conserved T cell epitope regions (CTERs) that drive cross-reactive T cell responses across Paramyxoviruses, providing a framework for vaccine strategies that enhance population coverage and pandemic preparedness.

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.

A label of betalactam (BL) allergy is estimated in around 10% of the population in their medical records. Second-line choices carry significant negative consequences, including reduced efficacy, effectiveness, and safety. This study aimed to develop a new highly specific score constructed by selecting variables assisted by artificial intelligence to identify low-risk BL-allergic patients.
In this study, derivation and validation of the BL-predictor score were performed on a retrospective cohort of 2207 patients who underwent penicillin allergy testing at Málaga University Hospital (Spain). The development of the BL-predictor encompassed expert drafting and a two-step variable selection process consisting of univariate analysis and variable filtering, followed by stepwise logistic regression with resampling. To assess the efficiency, a multicentric retrospective external validation was performed in 4261 patients from six populations: Salamanca and Madrid, Spain; Nashville, United States of America; Verona, Italy; Paris, France; and Copenhagen, Denmark.
The definitive questionnaire consisted of eight items and risk points were computed from the logistic regression model as follows: +1 for reactions after first dose or in less than 1 h (ITEM-1), +2 for anaphylaxis (ITEM-2); +1 for previous reaction with the culprit (ITEM-3); -1 for resolution in > 24 h (ITEM-4); +2 for spontaneous resolution (ITEM-5); -2 for unknown symptoms (ITEM-6); -2 for reaction occurred > 5 years (ITEM-7), and -1 for another reported drug allergy (ITEM-8). After establishing a threshold of ≤ 0 points to classify individuals with low risk, internal validation showed a specificity of 86% and a negative predictive value (NPV) of 83%. Overall multicenter external validation showed a specificity of 93%, which implies a 25% increase in specificity compared to the previously published BL decision tool.
This score would simplify diagnostic procedures in low-risk patients, enabling rapid delabeling, potentially in non-specialty settings, and reducing diagnostic costs and the negative consequences associated with incorrect antibiotic allergy labels.

Globally, approximately 100,000 patients receive hematopoietic stem cell transplants (HCTs) annually, with increasing use for curative intent for many diseases (eg, leukemia, lymphoma).1 There is increasing recognition that antibiotic utilization may have a unique impact on this patient population by affecting both infection- and transplant-related outcomes.2,3 From an infection perspective, utilizing appropriate antibiotics is critically important: among patients with HCTs, there is a growing incidence of multidrug-resistant (MDR) infections, and potentially treatable infection is the primary cause of death in up to 20% of patients following allogenic HCT.3 Simultaneously, a growing body of literature suggests that decreased gastrointestinal microbiome diversity (which is associated with broad-spectrum antibiotic utilization) may worsen therapy response, acute graft-versus-host disease outcomes, and overall survival among patients with HCTs.2 Social and structural determinants of health have also proven to be important determinants of health outcomes among patients with cancer generally.

Mammarenaviruses are classified into Old World and New World viruses (Old World arenaviruses [OWAs] and New World arenaviruses [NWAs]). Characterization of antigens recognized by human T cells is essential for identifying immunodominant targets, informing vaccine design, and performing immunological assessments. Here, we select the Lassa virus (LASV) as a prototype OWA to map the human CD4 T cell epitope repertoire. We then calculate conservation in different LASV lineages and other representative OWAs and NWAs and define conserved T cell epitope regions (CTERs) using Lassa as the OWA prototype and Junin as the NWA prototype. We show that these CTERs are able to broadly cross-recognize other OWA or NWA sequences. We validate our findings in humans immunized with an experimental glycoprotein complex (GPC) LASV vaccine or infected with lymphocytic choriomeningitis mammarenavirus (LCMV), as well as in the mouse model immunized with a stabilized GPC vaccine candidate. Our results on mammarenavirus-specific T cell immunity contribute to guiding the development of next-generation mammarenavirus vaccines.

Persistent human-adapted microbes can act as immunological "keystones," organizing host defense across tissues and shaping vulnerability under immune perturbation. More generally, tissue immunity is calibrated by persistent niche-resident organisms that tune compartment-specific thresholds of cytotoxicity and peripheral tolerance; keystone organisms represent the apex subset with multi-niche scope. Here we operationalize keystone organisms as pathogens whose containment requires coordinated engagement of multiple immune arms and whose residence is structured across anatomical niches. Using 18 curated immunological and evolutionary traits across 43 organisms, unsupervised analyses resolved four reproducible archetypes and identified a compact keystone set dominated by persistent herpesviruses and Mycobacterium tuberculosis. We then translated the clinical literature into a pathogen×immune-perturbation×niche tensor capturing where and when each organism emerges under defined immune deficits. We quantified "diagnostic breadth" with two complementary summaries: immune breadth (diversity of perturbations associated with emergence) and niche breadth (diversity of anatomical sites). Clinical emergence patterns perfectly separated trait-defined keystones from all other organisms and highlighted expanded niche breadth as the primary discriminator, whereas immune breadth showed no significant group separation. Finally, a mechanistic model integrating barrier disruption, latent reservoir activation, and tissue-resident immune control predicted clinical emergence from first principles-without fitting parameters to individual pathogens-and ranked true emergences 2.9-fold above chance among its highest-confidence predictions. Together, these results link evolutionary adaptation to clinically readable patterns of reactivation, motivate archetype-aware surveillance under immunosuppression, and provide a framework for immunogen design that prioritizes conserved, functionally constrained targets. Because the clinical tensor is literature-curated and sparse, "perfect separation" refers to keystone-vs-other discrimination within this dataset and is not a claim of universal out-of-sample performance.