Nightmares are a hallmark of post-traumatic stress disorder (PTSD), yet pharmacological treatments remain limited. Here we conducted a multicenter, double-blind, randomized, placebo-controlled trial to evaluate the efficacy and safety of dronabinol (BX-1; Δ9-tetrahydrocannabinol (THC)) for PTSD-related nightmares. Adults with PTSD and recurrent nightmares were randomly assigned to receive dronabinol (2.5-15 mg) or placebo, once daily before bedtime, for 10 weeks. The primary endpoint was change from baseline to week 10 in the Clinician-Administered PTSD Scale for DSM-IV (CAPS-IV) B2 item assessing nightmare frequency and intensity. A total of 171 patients underwent randomization (mean age, 37.9 ± 12.8 years; 79.3% female), with 87 assigned to dronabinol and 84 to placebo. At week 10, reduction in CAPS-IV B2 scores was significantly greater with dronabinol than with placebo (between-group difference -1.50; 95% confidence interval -2.28 to -0.71; Cohen's d = 0.65; P < 0.001). Adverse events occurred in 78 patients (89.7%) receiving dronabinol and 64 (77.1%) receiving placebo. Treatment discontinuation due to adverse events occurred in five (5.7%) and six (7.2%) patients, respectively. Serious adverse events occurred in seven patients (8.0%) receiving dronabinol and in none receiving placebo. This trial provides evidence that dronabinol reduces the frequency and intensity of PTSD-related nightmares, although long-term efficacy and safety require further evaluation. ClinicalTrials.gov: NCT04448808 .

Artificial intelligence (AI) is increasingly permeating healthcare, from serving as a physician assistant to powering consumer applications. The opacity of AI algorithms makes the ability of humans to interact with AI algorithms challenging. To overcome this limitation, explainable AI (XAI) provides insight into AI decision-making, but evidence suggests that XAI can paradoxically induce bias in the human decision-making process. Here we present results from two large-scale experiments, involving 623 lay people and 153 primary care physicians (PCPs), respectively, in which a fairness-based AI model for dermatological diagnoses and different XAI-based explanations were combined to examine how XAI assistance, particularly multimodal large language models (LLMs), influences diagnostic performance. With fairness-constrained model training, assistance from an AI model that achieved balanced performance across skin tones improved final diagnostic accuracy and reduced skin-tone-related performance disparities among both lay people and PCPs. In this setting, LLM explanations yielded divergent effects: lay users showed higher automation bias-accuracy was boosted when the diagnoses provided by the AI model were correct but was reduced when the model erred-whereas experienced PCPs remained resilient, benefiting irrespective of the AI model's accuracy. In addition, presenting the AI model's diagnosis before human decision-making may lead to stronger anchoring bias. These findings highlight XAI's varying impacts based on human expertise and the timing of when the AI-based prediction is provided, underscoring the concept that LLMs can act as a 'double-edged sword' in medical AI and informing future human-AI collaborative system design.

Antibody-drug conjugates (ADCs)-composed of a monoclonal antibody linked to a payload-were designed to deliver high concentrations of cytotoxic agents to cancer cells. Since their inception, a deeper understanding of the mechanisms of action and resistance to ADCs in patients has generated a wealth of advances in the chemistry of ADC constructs, together with rational therapeutic combinations. However, the pace of innovation now exceeds clinical trial capacity. Also, any single modification or combination is unlikely to generate clinically meaningful benefit on its own. In this context, there is a need to integrate multiple chemistry advances into individual ADCs and to develop frameworks, infrastructures and tools to accelerate and de-risk the preclinical and early clinical development of these agents. In addition, the development of multidimensional molecular tools to predict ADC sensitivity, together with the optimal use of new ADCs in early-stage cancers, should contribute to improved outcomes for patients. We discuss these opportunities and challenges and predict that in the longer term, the development of diversified ADC libraries incorporating distinct constructs and drug-to-antibody ratios will enable personalized treatment strategies aligned with individual tumor biology.

Recent rapid progress in the field of computational pathology has been enabled by foundation models. These models are beginning to move beyond encoding image patches toward whole-slide understanding, but their clinical utility remains limited. Here we present PRISM2, a multimodal slide-level foundation model trained on 2.3 million whole-slide images and 14 million question-answer pairs derived from 700,000 pathology reports. Through clinical dialogue supervision, PRISM2 aligns histomorphology with diagnostic reasoning, yielding representations that support both prompt-based inference and transferable embeddings for downstream tasks. With prompt-based inference, PRISM2 achieves or exceeds (P < 0.05) the balanced accuracy of clinical-grade products calibrated for cancer detection in the prostate, breast and breast lymph node. Additionally, across comprehensive diagnostic, biomarker and survival benchmarks, PRISM2 embeddings never statistically underperform previous foundation models via linear probing (P < 0.05). Furthermore, task-specific fine-tuning on survival prediction outperforms training from scratch on the same large survival dataset. PRISM2 demonstrates how language-supervised pretraining provides a scalable, clinically grounded signal for generalizable pathology representations, bridging human diagnostic reasoning and foundation model performance.

Combining multidisciplinary care plans into integrated care pathways (ICPs) is scalable, generalizable and effective for several long-term conditions, but not evaluated in Long COVID (LC). Our phase 3, cluster-randomized, multicenter clinical trial investigated the effectiveness of ICP interventions for LC. Recruitment of adults ≥18 years with LC was conducted in 6 National Health Service LC clinics in England with specialist ICPs. The intervention arms were (i) multi-organ magnetic resonance imaging (MRI; Coverscan), (ii) digital rehabilitation (Living with COVID Recovery), (iii) both multi-organ MRI and digital rehabilitation, and (iv) neither multi-organ MRI nor digital rehabilitation (usual care); cluster randomizing at primary care network (PCN) level to deliver interventions as 'standard of care' in that area. The primary endpoint was mean Fatigue Assessment Scale (FAS) at 12 weeks. Secondary endpoints included FAS at 24 weeks and EQ-5D-5L visual assessment scale at 12 and 24 weeks. A total of 1,152 participants from allocated PCNs consented to data collection, and 122 PCN clusters were allocated to the multi-organ MRI (33 PCNs), digital rehabilitation (32 PCNs), 'both' (27 PCNs) and 'neither' (30 PCNs) arms. Baseline FAS was equivalent across groups (mean 35.8, s.d. 8.21, 66.4% female). Overall, the primary outcome was achieved in all arms, with scores improving by 4.5 points to 31.3 (s.d. 9.31) at 12 weeks. Compared with usual care, effects on 12-week FAS were -0.18 (95% confidence interval -0.72, 1.09; P = 0.69) with multi-organ MRI; -0.53 (-1.42, 0.36; P = 0.25) with digital rehabilitation; and -0.17 (-1.06, 0.72; P = 0.71) with their interaction. Absolute differences for other interventions from usual care were modest. There were no serious adverse events related to ICP interventions. Multidisciplinary holistic clinical care is associated with fatigue reduction in LC, irrespective of multi-organ MRI. Digital rehabilitation could be useful in longer-term LC management. Large, multi-site trials of optimal ICP interventions are feasible, but further trials are required. ISRCTN identifier: ISRCTN10665760 .

Group B Streptococcus (GBS) disease is a major cause of infant morbidity and mortality. Capsular polysaccharide conjugate GBS vaccines have been evaluated for decades; however, none is licensed. This phase 1/2 trial evaluated an investigational maternal hexavalent polysaccharide-protein conjugate GBS vaccine (GBS6) in non-pregnant females and pregnant participants and their infants. In stage 1 (conducted in South Africa), healthy non-pregnant participants received GBS6, GBS6 + aluminum phosphate (AlPO4) or placebo; a subset received GBS6 + AlPO4 booster approximately 2 years later. Stage 2 (conducted in South Africa) maternal dose-finding data were reported previously and are not presented here. In stage 3, healthy pregnant participants from South Africa, the United States and the United Kingdom received GBS6 or placebo (24-36 weeks' gestation). In stage 1, 66 non-pregnant participants received GBS6, GBS6 + AlPO4 or placebo (n = 22 each); 26 also received GBS6 + AlPO4 booster. In stage 3, 216 pregnant participants were included (GBS6, n = 108; placebo, n = 108); 209 infants were born to maternal participants (GBS6, n = 104; placebo, n = 105). Primary objectives described GBS6 safety/tolerability profiles. Nearly all reactogenicity events were mild or moderate. Among non-pregnant participants, no serious adverse events (SAEs) or adverse event (AE)-associated withdrawals were reported after vaccination; medically attended adverse events (MAEs) were reported in similar percentages across primary vaccination groups (GBS6 groups = 41-45%; placebo = 41%). Rates of AEs (GBS6 = 75/108 (69%); placebo = 70/108 (65%)), SAEs (GBS6 = 21/108 (19%); placebo = 23/108 (21%)) and delivery outcome frequencies were similar in GBS6 and placebo maternal participants. Most infants were born full-term. Infant AE (GBS6 = 87/104 (84%); placebo = 86/105 (82%)), SAE (GBS6 = 45/104 (43%); placebo = 46/105 (44%)) and MAE (GBS6 = 65/104 (63%); placebo = 66/105 (63%)) frequencies were similar across groups. Two neonatal sepsis-associated deaths occurred (GBS6 = 1; placebo = 1); neither was considered vaccine related. Secondary objectives described immune responses. GBS6 elicited robust immune responses in non-pregnant and pregnant individuals; additionally, serotype-specific quantitative and functional antibody responses were observed in infants whose parent received GBS6. These data support continuing clinical investigation of maternal GBS6 vaccination. ClinicalTrials.gov identifier: NCT03765073 .

Depression is a major contributor to the global burden of diseases with substantial socioeconomic consequences. Here, to estimate the economic effects of depression, we developed a macroeconomic model using data from the World Bank and Global Burden of Disease Study 2021. The model estimated the impact of depression on labor force participation, educational attainment and work experience, adjusted for age and sex. The projected global economic burden from 2025 to 2050 is estimated at US$12 trillion (2024 US dollars), representing 0.460% of annual global gross domestic product. When suicide-related deaths attributable to depression are accounted for, based on an assumed 60% attribution rate representing the upper bound of estimates in the literature, the overall economic burden increases to US$14 trillion. The highest economic burdens are observed in the United States and China. Relative to regional gross domestic product, the impact is greatest in North America. Reduced labor force participation and productivity were identified as the predominant drivers of this cost. The macroeconomic burden of depression is substantial and inequitably distributed globally. These findings underscore the importance of addressing depression as a public health and economic concern. Reducing the health burden of depression could yield economic returns by supporting productivity and capital accumulation.

Fibromyalgia is a common and debilitating chronic pain syndrome of poorly understood etiology. Here we conduct a multi-ancestry genome-wide association study meta-analysis across 2,563,755 individuals (54,629 cases and 2,509,126 controls) from 11 cohorts, identifying 26 risk loci for fibromyalgia. The strongest association was with a coding variant in HTT, the causal gene for Huntington's disease. Gene prioritization implicated the HTT regulator GPR52, as well as diverse genes with neural roles, including DCC, DRD2/NCAM1, MDGA2 and CELF4. Fibromyalgia heritability was exclusively enriched within brain tissues and neural cell types. Fibromyalgia showed strong, positive genetic correlation with a wide range of chronic pain, psychiatric and somatic disorders, including genetic correlations above 0.7 with low back pain, post-traumatic stress disorder and irritable bowel syndrome. Despite large sex differences in fibromyalgia prevalence, the genetic architecture of fibromyalgia was nearly identical between males and females. This study provides robust genetic evidence defining fibromyalgia as a central nervous system disorder, thereby establishing a biological framework for its complex pathophysiology and extensive clinical comorbidities.