Rare cancers account for approximately one quarter of cancer diagnoses in Europe, yet prevailing clinical trial governance frameworks were developed for common malignancies and conventional randomised designs. The 2024 WHO Guidance for best practices for clinical trials and ICH E6(R3) reframe trial quality around scientific validity, ethical proportionality, feasibility, equity, and meaningful patient engagement — principles that are structurally tested by rare cancer research. This narrative review examines how these principles are operationalised in representative sarcoma trials, where biological heterogeneity, low incidence, and international collaboration expose the limits of standard approaches. Drawing on examples from Bayesian designs, umbrella and platform trials, registry-based approaches, decentralised participation, and structured advocacy involvement, we identify recurring governance features associated with successful implementation in rare cancer settings. Patient-centred approaches — particularly early integration of patient-reported outcomes (PROs), risk-proportional oversight, and attention to participant burden — emerge as key determinants of feasibility, recruitment, retention, and interpretability. In rare cancers, patient-centred, fit-for-purpose governance is not an ethical adjunct but a structural determinant of operational validity. Lessons from sarcoma trials can be extended beyond rare cancers to molecularly stratified subsets of common malignancies. Funding This work received no specific funding from public, commercial, or not-for-profit organisations.
Background The first randomisation results from the PACT-21/CASSANDRA trial demonstrated that cisplatin, nab-paclitaxel, capecitabine, and gemcitabine (PAXG) is a standard neoadjuvant treatment option for resectable/borderline resectable (R/BR) pancreatic adenocarcinoma (PDAC). Here, we report the results of the second randomisation comparing long versus short pre-operative chemotherapy. Methods PACT-21/CASSANDRA was a 2 × 2 factorial phase 3 trial involving 17 academic hospitals across 10 regions in Italy. Eligible patients were aged 18–75 years with pathologically confirmed R/BR PDAC and were first randomly assigned to either PAXG or fluorouracil, leucovorin, irinotecan, oxaliplatin (mFOLFIRINOX). Patients without progression or limiting toxicity after 4 months were randomised to receive 2 further months of the same chemotherapy either before (long) or after (short) surgery. The trial was designed under a two-sided superiority hypothesis (H0: HR long versus short = 1.0; H1: HR ≠ 1.0). Randomisation lists were stratified according to previous chemotherapy. The primary endpoint was event-free survival (EFS) in the intention-to-treat (ITT) population. Secondary endpoints were radiological, CA19.9 and complete pathological response, R0 and N0 resections, chemotherapy dose-density, and overall survival (OS). Analyses were performed using SAS version 9.4. This trial is registered with ClinicalTrials.gov (NCT04793932) and EudraCT (2020-003080-26 and 2024-519031-42-00). Findings Between Feb 23, 2021, and Sept 03, 2024, 86 patients were assigned to long and 79 to short preoperative chemotherapy. No difference in EFS was observed between the two groups in the intention-to-treat population (unadjusted HR 0.98; 95% CI; 0.68–1.41; P = 0.90). CA19-9 response (58 [97%] of 60 versus 41 [84%] of 49 patients; P = 0.02), pathological complete response (4 [5%] of 86 versus 0 of 79 patients; P = 0.05), N0 resection rate (39 [45%] of 86 versus 21 [27%] of 79 patients; P = 0.01), and chemotherapy dose-density were improved by longer chemotherapy. OS data are not yet mature. Interpretation Long and short preoperative chemotherapy obtains similar EFS in R/BR PDAC. Future research should focus on more effective treatment regimens and explore the optimal post-operative therapy. Funding MyEverest and Codice Viola (patients’ associations).
Background The optimal duration of antifungal therapy for candidemia in intensive care unit (ICU) patients remains uncertain. We aimed to evaluate whether shorter antifungal therapy was not associated with increased mortality using a target trial emulation framework. Methods We conducted a multicenter cohort study including adult ICU patients with candidemia from 16 ICUs in France (January 1, 2015 to January 1, 2023), emulating a target trial comparing short-course (<10 days) versus long-course (≥10 days) antifungal therapy. Patients treated for more than 21 days were excluded to minimize indication bias related to complicated candidemia. Patients who died within five days were excluded to ensure eligibility for both strategies. To account for time-dependent bias and confounding, we applied a cloning–censoring–weighting approach with inverse probability weighting. The primary outcome was all-cause mortality at day 90 after candidemia onset. Findings Among 237 eligible patients with uncomplicated candidemia from the 492 enrolled in the CandidICU cohort, 48 (20.3%) received short-course and 189 (79.7%) long-course therapy. Overall 90-day mortality was 47.7% (113/237). Median age was 62 years [IQR 51–70], median SAPS II was 53 [40–68], and Candida albicans accounted for 66.7% of cases. After weighting, baseline characteristics were well balanced. Short-course therapy was not significantly associated with 90-day mortality (weighted HR 1.14, 95% CI 0.82–1.58). No differences were observed in secondary infections or organ support between groups. Interpretation In this multicenter ICU cohort, among patients who survive the initial 5 days after candidemia onset and do not have obvious indication for prolonged therapy, short-course antifungal therapy was not associated with increased 90-day mortality after adjustment for time-dependent bias and confounders, although the confidence interval is wide and compatible with both clinically important benefit and harm. These findings are hypothesis-generating and should not alter current clinical practice. Randomized controlled trials are needed to definitively determine optimal treatment duration in critically ill patients. Funding No funding was received for this work.
Background Pharmacologic neuroprotection for out-of-hospital cardiac arrest (OHCA) remains elusive. We evaluated whether early hydrocortisone, ascorbic acid, and thiamine reduced neuronal injury in comatose OHCA survivors. Methods In this multicentre, placebo-controlled trial across four South Korean hospitals, comatose adult OHCA survivors undergoing targeted temperature management were randomly assigned (1:1) to intravenous hydrocortisone (100 mg), ascorbic acid (50 mg/kg), and thiamine (200 mg) every 12 h for 3 days, or placebo. Randomisation was computer-generated. Bedside clinicians were aware of treatment allocation, whereas laboratory personnel and independent outcome assessors were masked. The primary outcome was peak neuron-specific enolase (NSE) concentration, defined as the highest available value measured at 48 h or 72 h after return of spontaneous circulation. Efficacy was analysed using a modified intention-to-treat (mITT) principle, with multiple imputation for the full intention-to-treat (ITT) population. The completed trial was registered with ClinicalTrials.gov on June 4, 2021 (NCT04921189). Findings Between Dec 31, 2021, and Dec 7, 2024, 160 participants (123 male, 37 female) were assigned (80 per group); 147 were included in the mITT analysis (74 treatment, 73 placebo). Peak NSE levels were 47·4 ng/mL (interquartile range [IQR] 22·4–262·6) in the treatment group and 53·3 ng/mL (IQR 26·6–244·2) in the placebo group (median difference −5·9 ng/mL; 95% confidence interval [CI] −34·5–40·8; p = 0·54). ITT sensitivity findings were consistent (geometric mean ratio 0·94; 95% CI 0·61–1·46; p = 0·80). At 180 days, survival (63·5% vs 69·9%) and good neurological outcome (54·1% vs 56·2%) were similar. Serious adverse events were comparable in the ITT population (52·5% vs 50·0%), with none deemed related to the study medication. Interpretation Early administration of hydrocortisone, ascorbic acid, and thiamine did not reduce neuronal injury or improve 6-month clinical outcomes. The findings do not support empiric neuroprotective use in comatose OHCA survivors. Funding National Research Foundation of Korea and Korea Health Technology R&D Project.
Background Preschool children with asthma are more susceptible to develop acute life-threatening respiratory distress leading to hospital admissions compared to school children and treatment options are limited for this age group. We evaluated the safety and efficacy of tiotropium as add-on therapy to inhaled corticosteroids (ICS) in preschool children with high-risk, partly controlled or uncontrolled asthma. Methods TIPP was a phase III, prospective, multicentre, randomised, double-blind, placebo-controlled, parallel-group (investigator-initiated) trial conducted at 13 German centres (12 hospitals, one specialised medical practice). Children aged 1–5 years with physician-diagnosed asthma, partly controlled or uncontrolled symptoms despite ICS, and a history of severe exacerbations were randomly assigned to receive once-daily tiotropium (two puffs of 1.25 μg) via Respimat® inhaler or placebo as add-on to ICS therapy over 52 weeks. The primary outcome was time to first severe asthma exacerbation requiring hospitalisation and/or systemic corticosteroid treatment. All randomised participants were included in the intention-to-treat analysis and analysed by treatment received for safety. No primary or safety data were missing; therefore, no imputation was required. The study was registered in the “EU Clinical Trials Registry” (EudraCT 2021-000190-81). Findings Between February 2022 and March 2025, 100 of the planned 204 children were enrolled, of whom 86 were randomised to tiotropium + ICS (n = 44) or placebo + ICS (n = 42). Mean age was 3.3 years (Standard deviation (SD) 1.3). The primary outcome time to first severe asthma exacerbation did not differ between groups (Hazard ratio (HR): 0.99, 95% confidence interval (CI): 0.51–1.92; p = 0.98). At least one adverse event was reported in 41 (93%) of 44 children in the tiotropium + ICS group and 42 (100%) of 42 children in the placebo + ICS group. 397 adverse events were reported with tiotropium + ICS group and 450 with placebo + ICS group. Tiotropium was well tolerated, and no new safety signal was identified. Interpretation Tiotropium did not reduce the risk of a severe asthma exacerbation, but was well tolerated as add-on therapy to ICS in preschool children with high-risk asthma. Funding German Federal Ministry of Education and Research (01KG2030); study medication provided by Boehringer Ingelheim Pharma GmbH & Co. KG (0205-0546).
Background Radiation pneumonitis (RP) remains a significant treatment-related toxicity in patients with unresectable, locally advanced non-small cell lung cancer (NSCLC) undergoing chemoradiotherapy (CRT). Most existing predictive models rely on static baseline demographic or dosimetry variables and lack real-time clinical applicability. We developed a novel predictive framework that integrates longitudinal symptom data extracted from clinical notes using natural language processing (NLP) with clinical and dosimetry features to improve early RP prediction. Methods We retrospectively identified 227 patients with locally advanced NSCLC treated with definitive CRT at a high-volume cancer center in the United States. We included all patients older than 18 years who were diagnosed between Jan 1, 2006, and Dec 31, 2022 with histologically or cytologically confirmed unresectable Stage 2 or 3 NSCLC and treated with conformal radiotherapy to a minimum dose of ≥45 Gy with or without chemotherapy. Of these, 31 RP events were identified through manual adjudication using radiologic criteria and chart review. NLP was used to extract the temporal relationship of 16 pre-specified symptoms with treatment from over 100,000 clinical notes spanning pre- and during-treatment intervals. We trained and validated machine learning models on combinations of baseline clinical data, radiation dosimetry, and NLP-derived symptom features. Model performance was evaluated using a nested cross-validation framework, with an outer cross-validation loop reserved for performance assessment and an inner cross-validation loop used for model training and integration, and summarized using area under the receiver operating characteristic curve (AUC) and partial AUC (pAUC) at high specificity thresholds. Clinical utility was evaluated using decision curve analysis (DCA). Findings The best-performing model incorporated longitudinal NLP features and achieved a median AUC of 0.759 (90% confidence interval 0.753–0.766), significantly outperforming baseline models using only dosimetry (AUC 0.613) or clinical variables (AUC 0.635). NLP-based features such as cough trajectory, shortness of breath, and wheezing were among the most important predictors. Inclusion of NLP-derived symptom data improved early identification of high-risk patients, particularly in the clinically relevant high-specificity range (pAUC 0.021 vs. 0.010 for dosimetry alone). DCA showed that the calibrated MLP model provided greater net benefit than default strategies of treating all or no patients across clinically relevant threshold possibilities. Interpretation In this early work, NLP-based extraction of longitudinal symptoms from routine clinical documentation meaningfully enhances RP prediction in patients undergoing CRT for NSCLC. This approach leverages existing electronic health record infrastructure to deliver real-time, scalable, and interpretable risk estimates, offering a pathway toward potential early intervention and personalized toxicity management. The model and DCA requires external and prospective validation before clinical deployment; as such, future work should focus on this validation and integration into clinical decision support systems. Funding AstraZeneca.
Background The 2026 Bundibugyo virus disease (BVD) outbreak in DRC and Uganda underscores that filoviruses remain an unpredictable global health threat, particularly in resource-limited settings. The sporadic occurrence of outbreaks has constrained evidence-generation. We implemented a Rapid Research Needs Appraisal (RRNA) to rapidly and systematically synthesise published research evidence and identify gaps in the evidence base to support management and research strategies. Methods We searched Ovid Medline, Embase and Scopus for primary research studies published between Jan 1, 1999, and Feb 6, 2026 by combining keywords and thesaurus headings for Bundibugyo and humans. This search was supplemented by an updated search in PubMed and MedRxiv up to June 25, 2026. Searches were conducted in English, but without language restrictions. Eligible studies including humans or human samples focused on at least one of the RRNA domains (clinical characteristics, immune response, transmission, risk factors, medical countermeasures and associated social and behavioural factors). Two reviewers screened records for inclusion, one reviewer extracted data with a second reviewer checking a proportion of the data. Data were analysed descriptively. Findings 29 primary studies were eligible for inclusion, accounting for 1593 probable or confirmed BVD cases. Of these 29 studies, most were observational (n = 28, 96.6%) and were most commonly conducted in Uganda (n = 15, 51.7%) and the DRC (n = 10, 34.5%). Only one interventional study was included, a non-randomized interventional study exploring EBOV, SUDV and BDBV cross-reactivity following EBOV vaccination. Among studies that reported age, most (58.6%, 17/29) included adults (18+ years), eight (27.6%) included children (0–18 years), and only two (6.9%) included pregnant women. Fever, vomiting, diarrhoea, headache, fatigue, dysphagia, anorexia, dyspnoea and abdominal pain were the most commonly reported symptoms. Haemorrhagic manifestations ranged from 10.4 to 54.0%, with a meta-analysis estimating a CFR of 32.8% (CI 25.8–40.2) based on past outbreaks. Evidence on cross-reactivity to previous Ebola exposure following natural infection of vaccination was sparse and conflicting. One study identified low levels of cross-reactivity in EBOV vaccinated participants, another among EBOV survivors, whereas a third did not detect cross-reactivity following EBOV vaccination. Long-term sequelae, persistent BVD survivor stigma and limited community awareness were also reported. Interpretation Whilst acknowledging the limitations of such a rapid appraisal, our findings show that urgent, coordinated investment is needed in countermeasure trials inclusive of pregnant women and children, supported by observational studies community engagement and co-production. Funding The National Institute for Health Research (NIHR) and the Wellcome Trust.
Background Long-acting HIV pre-exposure prophylaxis (PrEP) may overcome adherence and discontinuation challenges facing oral PrEP, which remain unsystematically reviewed. This study aimed to synthesize the evidence on long-acting PrEP adherence and discontinuation. Methods We conducted a systematic review and meta-analysis of studies published from database inception to July 1, 2026 in PubMed, Web of Science, EMBASE, and Cochrane Library. Eligible studies reported adherence (on-time doses receipt at time points, e.g., at 2 months) or discontinuation (stopping doses or loss to follow-up within a time period, e.g., within 6 months) of long-acting PrEP, including long-acting injectable (LAI) and intravaginal ring (IVR) PrEP. Non-longitudinal designs and non-English publications were excluded. We used random-effects meta-analyses to pool estimates. Subgroup analyses were conducted by population, region, and study design. Findings We identified 46 eligible studies (28 LAI and 18 IVR) including 22,945 individuals (14,190 and 8755). Adherence to six-monthly and two-monthly LAI-PrEP at 12 months was 93.0% (95% CI: 91.6–94.3) and 88.1% (84.7–90.8%); discontinuation for the six-monthly and two-monthly regimens within 12 months was 8.2% (4.9–13.4) and 16.0% (10.8–23.1). Discontinuation for two-monthly LAI-PrEP was higher in implementation studies than in clinical trials (14.6% vs. 1.8% within 6 months; 32.0% vs. 10.3% within 12 months, P < 0.0001). IVR-PrEP adherence was 66.1% (42.2–83.9%) at 6 months and discontinuation was 21.4% (12.6–33.8%) within 12 months. Compared with clinical trials, implementation studies showed lower adherence (32.0% vs. 73.8%) and higher discontinuation (41.1% vs. 7.7%) at 6 months (P < 0.0001). Interpretation Adherence to LAI-PrEP remained high through 12 months, with the best adherence being observed in clinical trials, which serve as upper-bound benchmarks. Compared to LAI-PrEP, IVR-PrEP had lower adherence and higher discontinuation. Efforts facilitating longer-term adherence warrant research. Funding This study was supported by the National Key R&D Program of China.
Artificial intelligence (AI) and machine learning (ML) are increasingly applied across the heart and lung transplant continuum, from donor organ evaluation and waitlist prioritisation through peri-operative decision support to post-transplant rejection surveillance and personalised immunosuppression. Deep learning for endomyocardial biopsy interpretation has achieved an area under the curve (AUC) of 0.962, with performance comparable to expert pathologists, while ML-augmented ex-vivo lung perfusion platforms have demonstrated the potential to expand organ utilisation in real-world implementation studies. Noninvasive donor-derived cell-free DNA assays and electrocardiogram-based deep learning models are beginning to reduce dependence on invasive surveillance biopsies. Across published studies, models predicting 1-year post-transplant mortality have shown only moderate discrimination. Most published models remain retrospectively validated on single-centre datasets with limited external validation, inconsistent calibration reporting, and unresolved concerns about algorithmic bias and global health equity. This scoping review synthesises current evidence, critically appraises model performance, and identifies the regulatory, ethical, and methodological steps required for responsible clinical integration. Future research should prioritise prospective, multicentre external validation, transparent calibration reporting, and formal evaluation of clinical utility before these tools are adopted in routine transplant care.
Background Research suggests that hearing loss in older adults is associated with elevated neuroinflammation and neurodegeneration, both of which are risk factors for dementia. In a secondary analysis of the Aging and Cognitive Health Evaluation in Elders (ACHIEVE) randomised controlled trial (ClinicalTrials.gov identifier: NCT03243422), we examined if hearing intervention slowed 3-year change in blood-based biomarkers of neuroinflammation (glial fibrillary acidic protein [GFAP]) and neurodegeneration (neurofilament light chain [NfL]). Methods ACHIEVE, a clinical trial in the United States conducted between January 4, 2018 and November 30, 2022, tested the effects of hearing intervention (audiological counselling, provision of hearing aids) versus health education control (counselling on chronic disease prevention) on 3-year cognitive decline among older adults without substantial cognitive impairment and with untreated hearing loss. Plasma was collected at baseline and three years later among ACHIEVE participants recruited from the Atherosclerosis Risk in Communities cohort. Covariate-adjusted linear mixed effects models estimated intention-to-treat effects on 3-year change in inverse-normal transformed values of GFAP and NfL. Findings Participants in the analytic sample (n = 164) were mean (SD) age 78.1 (2.9) years, 64.0% female, and 73.2% White. Over three years, hearing intervention was associated with a slower rise in GFAP (intervention: −0.026; 95% CI: −0.144, 0.093, control: 0.149; 95% CI: 0.039, 0.260; difference: −0.175; 95% CI: −0.322, −0.028; p-difference: 0.019) and NfL (intervention: 0.171; 95% CI: 0.022, 0.319, control: 0.330; 95 %CI: 0.185, 0.475, difference: −0.159; 95 %CI: −0.348, 0.029; p-difference: 0.098). Interpretation Randomisation to hearing intervention was associated with a slower three-year rise in a blood-based biomarker of neuroinflammation. Comparable estimates were observed for neurodegeneration, although the effect was not statistically significant. Findings suggest that hearing intervention may reduce the rate of change in non-specific blood-based biomarkers associated with brain health and dementia progression. Funding US National Institutes of Health.
Background Important drivers of antibiotic resistance in low- and middle-income countries (LMICs) include antibiotic overuse and scarcity of microbiological diagnostics. Procalcitonin point-of-care testing is a potential tool to guide antibiotic de-escalation in patients admitted with suspected bacterial sepsis. Methods In an open-label randomised controlled clinical trial in Chittagong Medical College Hospital in Bangladesh, adult patients with suspected bacterial sepsis were randomised to procalcitonin (PCT)-guided de-escalation of antibiotic therapy or standard-of-care. In the intervention group, serum procalcitonin was assessed daily and clinicians received a non-binding advice to stop antibiotics after 72 h if procalcitonin dropped below pre-defined threshold concentrations (0.5 ng/mL or 80% from the peak value if baseline value > 2 ng/mL). The primary endpoint was length of antibiotic therapy analysed by intention-to-treat. The study is registered with ClinicalTrials.gov (NCT05955612) and was completed in July 2024. Findings Between 26 July 2023 and 20 June 2024, 532 of the 1002 patients screened for eligibility were randomly assigned to procalcitonin-guided antibiotic therapy (n = 266) or standard-of-care (n = 266). Follow-up was completed on 30 July 2024. Median (interquartile range [IQR]) length of antibiotic therapy was 4.7 (2.8–8.0) days with procalcitonin-guided therapy compared to 9.0 (5.8–12.0) days in the control arm (p < 0.0001). Mortality was similar between the two groups. In the procalcitonin-guided arm, 23 of 265 participants died (8.7%; 95% confidence interval [CI], 5.6%–12.7%), compared with 21 of 265 participants (7.9%; 95% CI, 5.0%–11.9%) in the control arm. The absolute risk difference was 0.8 percentage points (95% CI, −3.9 to 5.5 percentage points) (p = 0.875). Following review, the Data Safety Monitoring Board did not identify any deaths as attributable to antibiotic discontinuation. The incidence of recurrent infections was 8/265 (3.0%) in procalcitonin-guided antibiotic therapy and 10/265 (3.8%) in the control arm (p = 0.811). Interpretation In this single-centre trial in Bangladesh, procalcitonin-guided de-escalation reduced the duration of antibiotic therapy in adults with suspected bacterial sepsis. Mortality and recurrent infection were similar between the groups, although the trial was not powered to assess these outcomes formally. This promising approach of point-of-care procalcitonin-guided antibiotic discontinuation should be evaluated further in comparable resource-limited settings. Funding Wellcome Trust of Great Britain.
Background Whether correction of iron deficiency improves functional outcomes after kidney transplantation is unknown. We evaluated intravenous ferric carboxymaltose for improving exercise capacity in iron-deficient kidney transplant recipients. Methods In this multicentre, double-blind, randomised, placebo-controlled trial, adult kidney transplant recipients with iron deficiency (ferritin <100 μg/L or 100–299 μg/L with transferrin saturation <20%) at least six months post-transplantation were recruited at two Dutch university hospitals between 7 October 2019 and 16 February 2024. Participants were assigned (1:1) to four intravenous doses of 500 mg ferric carboxymaltose or placebo at six-week intervals. The primary outcome was the between-group difference in 24-week change in 6-min walk test distance. Secondary outcomes included cardiac, kidney, muscle and cognitive function and quality of life. Efficacy was analysed by assigned treatment in 148 participants, with a per-protocol sensitivity analysis; safety analyses included 154 dosed participants. The trial was registered at ClinicalTrials.gov (NCT03769441). Findings Of 155 participants randomised, seven were withdrawn under a prespecified COVID-19 amendment, leaving 148 participants for analysis (ferric carboxymaltose, n = 75; placebo, n = 73). Baseline 6-min walk test distance was 515 ± 91 m and 510 ± 114 m, respectively. Ferric carboxymaltose did not improve 6-min walk test distance versus placebo (adjusted mean difference +9 m [95% CI −10 to +28]). Ferric carboxymaltose corrected iron deficiency and increased haemoglobin, but did not improve cardiac function, muscle mass or strength, cognitive performance, or quality of life. Exploratory between-group differences occurred in estimated glomerular filtration rate and fibroblast growth factor 23. Infections and other clinically relevant adverse events did not differ between groups. Interpretation Ferric carboxymaltose did not improve exercise capacity or most secondary outcomes, arguing against routine supplementation to improve functional capacity. Future studies should investigate observed differences in kidney function and fibroblast growth factor 23 concentrations and long-term clinical outcomes. Funding Dutch Kidney Foundation, Dutch Heart Foundation, CSL Vifor.
Background The epidemiology of pediatric yeast bloodstream infections remains poorly characterized in high-income settings despite global shifts over the past two decades. As fungal species distribution varies geographically, national surveillance is essential. In France, recent reports have focused predominantly on adults, leaving a significant evidence gap for children. Methods We analyzed data from the YEASTS program, a prospective multicentre hospital-based surveillance study conducted in 27 hospitals in the Paris area, France, between Oct 1, 2002, and Dec 31, 2022. All incident episodes of yeast bloodstream infection diagnosed during this period, were eligible. For the primary analysis, we included children and adolescents aged 6 months to 16 years; neonates and recurrent episodes were excluded. Adult cases (>16 years) recorded during the same period were included for comparison. Species identification and antifungal susceptibility testing, using EUCAST standardized method, were performed at the National Reference Center of invasive Mycoses and Antifungals (NRCMA). The primary objective was to describe the epidemiology of yeast bloodstream infections in children and adolescents and compare findings with those observed in adults (>16 years). The main outcomes assessed were species distribution, antifungal susceptibility patterns, treatment strategies, and 30-day all-cause mortality. Findings Among 346 pediatric episodes, hematological malignancy was the main underlying condition (20.5%). Candida albicans predominated (42%), although non-albicans Candida species collectively accounted for a larger proportion of episodes, notably Candida parapsilosis (23.5%) and Candida tropicalis (10%). Species distribution varied by age: younger children had more C. parapsilosis, older more Pichia kudriavzevii (syn. Candida krusei) and Nakaseomyces glabratus (syn. Candida glabrata). Trichosporon asahii represented 2%. Resistance rates remained low (4% to caspofungin, 8.1% fluconazole) including 4 echinocandin-resistant isolates that harbored S645P mutations in the Fks coding region. 30-day mortality was 13.5%, independently associated with C. tropicalis, N. glabratus, T. asahii, female sex, and ICU admission. Compared with 5889 adult cases, pediatric patients had a distinct species distribution, more frequent prior antifungal exposure, greater use of liposomal amphotericin B, and significantly lower adjusted 30-day mortality. Interpretation Pediatric yeast bloodstream infections exhibit distinct epidemiological characteristics, species distribution, antifungal exposure patterns, and outcomes compared with adult infections. Future research should focus on prospective age-stratified studies to better define optimal antifungal treatment strategies, evaluate the clinical significance of emerging and resistant yeast species, and identify factors associated with poor outcomes in pediatric patients. Funding Santé Publique France and Institut Pasteur.
Background Camera-based methods have shown potential for respiratory rate measurement in neonatal intensive care. However existing work is tested on short recordings, which do not represent the range of environments present in the neonatal intensive care unit (NICU) or assess challenging conditions such as movement or poor visibility. This study evaluates a fully automatic respiratory rate measurement system over continuous 24-h periods in the unmodified NICU environment. Methods In this single-centre cross-sectional study, 20 babies in the NICU in Addenbrooke's Hospital, Cambridge, UK were recorded using a colour + depth camera, each for around 24 h. Recordings took place between 11 July 2022 and 11 November 2025. Inclusion criteria were: pre-term and nursed in closed incubators; no specific exclusion criteria applied. Existing methods were used to automatically locate the baby, and we used a novel clustering method to extract a respiratory signal and detect the patient's breaths. The primary outcome was the measurement of respiratory rate using the camera, assessed by comparison with data from the clinical standard patient monitor, recorded concurrently. Further post-hoc analysis was conducted on clinical and demographic subsets. This trial is registered on clinicaltrials.gov, ID NCT04831242. Findings After excluding intervals where the camera or baby was removed from the incubator, 380 h of valid recording from 20 participants remained. During non-intervention periods (338 h, 77.6% of all, 88.7% of valid), we configured the patient monitor signal quality indices to 57.9% valid measurement, a likely overestimate; the camera achieved 57.8% valid measurement with 1.44 breaths/minute agreement (mean error 0.42) averaged over all periods where both camera and monitor had valid data (136 h, 40.4% of non-intervention, 31.4% of all). Covering affects the valid time (63.6% when uncovered, 51.3% when covered), as do intervals where the baby is more active (61.4% when inactive, 53.5% when active). The error and valid time were worse when the monitor respiratory rate was below 35 breaths/minute (31.0% vs 57.8% valid, 1.98 vs 1.44 error) and more so below 25 breaths/minute (25.6% valid, 7.57 error), though data in this range was limited (0.2% of non-intervention) and the patient monitor may be unreliable. Interpretation Respiratory rate measurement using cameras has potential for continuous use in the NICU, beyond previously studied short recordings, achieving mean absolute error (MAE) averaged across all data below 2 breaths/minute while providing valid data 57.8% of the time. The relationship between covering/activity and valid measurement time shows the importance of long recordings in the real-world NICU. The MAE and valid time was worse for babies under 1000 g, requiring continuous positive airway pressure, or when the respiratory rate is below 35 breaths/minute. Further work should focus on bradypnoea/apnoea detection, the aforementioned demographic groups, and results should be confirmed in a cohort with greater ethnic diversity. Funding Rosetrees Trust, Stoneygate Trust, Isaac Newton Trust, EPSRC Impact Acceleration Account.
Background Dengue is a growing global health threat. Our initial evaluation of the TAK-003 dengue vaccine in adolescents during the 2024 outbreak in São Paulo State, Brazil, showed short-term protection but suggested possible effectiveness waning after 90-days of the first dose, and had limited power to estimate second-dose effectiveness against hospitalisation. We evaluated 1-year vaccine effectiveness against symptomatic, virologically confirmed dengue and dengue hospitalisation in adolescents during 2024–2025, when DENV-1 and DENV-2 predominated in 2024 followed by increasing DENV-3 transmission in 2025. Methods We did a test-negative case–control study among adolescents aged 10–15 years in São Paulo State between Feb 20, 2024, and Dec 31, 2025. We extracted individual-level information on demographic characteristics, comorbidities, dengue testing, dengue vaccination, hospitalisation, and death during the study period from the national surveillance databases for dengue (SINAN-Dengue; compulsory notification) and the São Paulo State Secretary of Health vaccination registry (Rede Nacional de Dados em Saúde) on Feb 4, 2026. Cases were defined as acute febrile illness episodes with virologically confirmed dengue, as detected by NS1 antigen or RT-PCR testing within 5 days of symptom onset. Controls were test-negative acute febrile illness episodes. Surveillance data were linked to the state vaccination registry. Vaccine effectiveness after one or two TAK-003 doses was estimated with mixed-effects logistic regression. Findings We analysed 166,390 tests: 65,365 from cases and 101,025 from controls. Vaccine effectiveness against symptomatic dengue was 59.2% (95% CI 56.1–62.0) after one dose and 73.1% (69.1–76.6) after two doses. Against dengue hospitalisation, effectiveness was 74.6% (62.7–82.7) after one dose and 87.9% (73.0–94.6) after two doses. One-dose protection against symptomatic disease showed no evidence of waning over 1-year follow-up: effectiveness was 73.8% (64.7–80.6) at 270–364 days and 76.8% (59.3–86.8) at 365 days or more. Two-dose protection remained stable to 1 year (74.6%, 95% CI 59.0–84.2, at 270–365 days). Findings were robust in sensitivity analyses and comparable during increased DENV-3 circulation. Interpretation Our findings show that TAK-003 demonstrated sustained 1-year protection against symptomatic dengue and dengue hospitalisation in adolescents during a large outbreak. Notably, a single dose showed sustained effectiveness without evidence of waning over 1 year, generating an important hypothesis regarding dose reduction that warrants continued long-term observational evaluation. A second dose provided optimal protection, particularly against hospitalisation. These findings reinforce the importance of dengue vaccination as a public health measure. Funding National Council for Scientific and Technological Development.
Background Attention-deficit/hyperactivity disorder (ADHD) may compromise adherence to pharmacotherapy in adults with type 2 diabetes, but the relationship is unknown. We investigated the association of ADHD and ADHD medication with antidiabetic medication adherence in multinational cohort data. Methods Using a common analytical protocol and population-based databases from seven countries, we identified adults initiating non-insulin antidiabetic medication between 2010 and 2020. ADHD was identified by diagnosis or ADHD medication use. We used the proportion of days covered (PDC) to measure antidiabetic medication adherence and defined the primary outcome, poor adherence, as PDC <80% over 1-, 2-, and 5-year follow-up periods. Time-to-first-discontinuation was the secondary outcome. Country-specific estimates were obtained using logistic and Cox regression and pooled using random-effects meta-analyses. Findings Among 3,643,197 individuals included, 1,871,870 (51.4%) were female and 88,339 (2.4%) had ADHD (76,973 [87.1%] received ADHD medication). Overall, ADHD was not associated with poor adherence to antidiabetic medication (1-, 2-, and 5-year pooled odds ratios [OR]: 1.01 [95% CI 0.90–1.14], 1.03 [0.91–1.17], and 1.12 [0.95–1.31]); country-specific estimates ranged from 0.79 to 1.32, 0.80 to 1.35, and 0.80 to 1.47 at 1, 2, and 5 years, respectively. Among males, ADHD was associated with poorer adherence (1.09 [95% CI 1.00–1.20], 1.12 [1.00–1.24], and 1.23 [1.11–1.37] at 1, 2, and 5 years). Similar results were found for discontinuation. Among individuals with ADHD, ADHD medication use was consistently associated with better adherence over 1-, 2-, and 5-year follow-up periods (e.g., 5-year pooled OR 0.45 [0.30–0.67]); associations were directionally consistent across countries. Interpretation Among adults with type 2 diabetes, ADHD may not necessarily compromise antidiabetic medication adherence when appropriately managed in routine care. ADHD pharmacological treatment may support adherence among adults with comorbid ADHD and type 2 diabetes, although its causal effect requires further evaluation. Funding EU Horizon 2020 Research and Innovation Programme.
Background Evidence regarding outcomes in patients with hormone receptor (HR)-positive/HER2-negative breast cancer (BC) with a second consecutive ipsilateral locoregional recurrence (LRR) remains limited. Methods Single-center, cohort study including consecutive patients with HR-positive/HER2-negative BC who underwent surgery (study baseline) for a second consecutive, ipsilateral LRR, at the European Institute of Oncology (Milan) from January 2001 to September 2025. Endpoints included invasive breast cancer-free survival (IBCFS) and distant recurrence-free survival (DRFS). Findings 111 patients were included: median age at study baseline was 61 years. The second LRR occurred in the breast in 44% of patients, in locoregional lymph nodes in 23%, on the mastectomy scar or chest wall in 31%. Endocrine-resistant recurrence was observed in 69% of patients. Overall, 14% of patients received adjuvant chemotherapy and 5% a CDK4/6-inhibitor. At a median follow-up of 6.3 years (95% CI: 4.7–7.9), 57 IBCFS events and 40 DRFS events were reported. Median IBCFS was 6.8 years (95% CI: 4.8–7.6) and median DRFS was 9.6 years (95% CI: 6.9–Not Reached [NR]). Median IBCFS was longer in endocrine-sensitive compared to endocrine-resistant LRR (8.6 years [95% CI: 4.6–NR] vs. 5.3 years [95% CI: 4.0–6.9]), and in intraparenchymal LRR (6.8 years [95% CI: 4.8–NR]) compared to nodal (6.4 years [95% CI: 1.4–NR]) or chest wall (5.1 years [95% CI: 4.0–NR]) recurrence. Interpretation Among patients experiencing a second consecutive LRR, approximately 50% develop a subsequent IBCFS event within 7 years, supporting consideration of treatment intensification, particularly in endocrine-resistant disease or recurrence involving the chest wall or regional lymph nodes. Funding None.
Artificial intelligence (AI) systems have quietly crossed an uncomfortable threshold: on a growing range of diagnostic tasks, they often match, and in some well-defined tasks even exceed, human performance. Evidence from large language models such as Med-PaLM and GPT-4, and deep learning systems in pathology, radiology, and hematology, shows high accuracy and reproducibility, challenging traditional human benchmarks, although still being subject to substantial heterogeneity in the studies cited. Yet most evaluation frameworks continue to treat human consensus on curated, retrospectively labeled datasets as ground truth, despite well-documented record of cognitive bias, interobserver variability, and diagnostic errors in clinical practice. This creates a striking asymmetry: AI is scrutinized against a putative gold standard whose gold content has never been formally tested—a benchmark illusion. Acknowledging AI's own distinctive failure modes and generalizability problems, this Viewpoint argues that such limitations do not justify clinging to an untested reference. It calls for explicit modeling of ground truth uncertainty, outcome-based validation, and, where appropriate, using AI as an additional reference layer for human performance. The relevant benchmark is no longer clinicians alone, but the human–AI ensemble evaluated within real diagnostic ecosystems and against patient-relevant outcomes. Funding None.
Background Patients with recurrent or metastatic nasopharyngeal carcinoma (R/M NPC) who progress after immune checkpoint inhibitor (ICI) therapy have limited treatment options and poor outcomes. We evaluated the efficacy and safety of liposomal irinotecan plus nimotuzumab in patients with epidermal growth factor receptor (EGFR)-positive, ICI-refractory R/M NPC. Methods In this single-arm, open-label, phase II trial (NCT06414577) conducted in China, patients with EGFR-positive, ICI-refractory R/M NPC were enrolled between May 27, 2024, and January 9, 2025. Patients received intravenous liposomal irinotecan (70 mg/m2) plus nimotuzumab (400 mg) every 2 weeks for up to eight cycles. The primary endpoint was objective response rate (ORR) per Response Evaluation Criteria in Solid Tumors version 1.1. Findings Thirty-six heavily pretreated patients were enrolled, of whom 12 (33.3%) had received three or more prior lines of therapy for advanced disease. In the intention-to-treat population, 13 of 36 patients achieved an objective response (ORR 36.1%; 95% CI 20.8–53.8), and 22 of 36 patients achieved disease control (61.1%). Median progression-free survival was 3.0 months (95% CI, 2.2–4.1), and median duration of response was 2.3 months (95% CI 1.5–3.1). After a median follow-up of 22.0 months, the median overall survival was 14.1 months (95% CI, 8.7-not estimable). Grade 3 or higher treatment-related adverse events occurred in eight of 36 patients (22.2%), and no treatment-related deaths were observed. In a post-hoc exploratory analysis, an early decline of at least 50% in plasma Epstein–Barr virus DNA was associated with a higher ORR. Interpretation Liposomal irinotecan plus nimotuzumab showed preliminary antitumor activity and manageable safety in heavily pretreated patients with ICI-refractory R/M NPC. However, response durability was limited, and further comparative studies are needed to evaluate its clinical value and identify patients most likely to derive durable benefit. Funding Supported by grants from the National Natural Science Foundation of China, Science and Technology Program of Guangdong Esophageal Cancer Institute, and China Postdoctoral Science Foundation.
Background:Circadian disruption may affect metabolism through altered timing of sleep, activity, light exposure, hormonal rhythms, and food intake. However, its relationship with metabolic dysfunction-associated steatotic liver disease (MASLD), including fibrosis and cirrhosis remains unclear. To evaluate associations between circadian and behavioural timing exposures and MASLD in observational studies. Methods:PubMed, Embase, and Scopus were searched from inception to 14 July 2026. Eligible studies examined naturally occurring timing-related exposures in adults and reported steatosis, fibrosis, cirrhosis, or quantitative liver-fat outcomes. Findings were synthesized narratively because heterogeneity precluded pooling. Risk of bias was assessed using AXIS or the Newcastle-Ottawa Scale, and certainty using an adapted domain-level GRADE approach (PROSPERO Registration number: CRD420261381636). Findings:Twenty-four independent MASLD-related evidence sources were included in the primary synthesis; two additional mixed-aetiology cirrhosis studies were considered contextually. Shift work showed the most consistent associations and most longitudinal evidence. Chronotype was generally associated with disease occurrence but inconsistently with advanced outcomes: intermediate or late chronotype was associated with fibrosis in one clinical cohort, whereas morning chronotype was not independently associated with incident cirrhosis in a prospective cohort. Later sleep timing was associated with prevalent MASLD and, in one NHANES analysis, fibrosis. Evidence for objective rest-activity rhythms, light exposure, meal timing, and melatonin phase remained limited or heterogeneous. Mixed-aetiology cirrhosis studies suggested that advanced liver disease may itself disrupt circadian physiology. Interpretation:Circadian and behavioural timing exposures showed trends associated with MASLD-related outcomes, with the most consistent evidence for shift work. Residual confounding, cross-sectional designs, self-reported exposures, heterogeneous outcomes, and reverse causality preclude causal conclusions. Prospective studies using objective circadian phenotyping and standardized fibrosis outcomes are needed. Funding:This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.
