Checkpoint inhibitors (CPIs) targeting programmed death 1 (PD-1)/programmed death ligand 1 (PD-L1) and cytotoxic T lymphocyte antigen 4 (CTLA-4) have revolutionized cancer treatment but can trigger autoimmune complications, including CPI-induced diabetes mellitus (CPI-DM), which occurs preferentially with PD-1 blockade. We found evidence of pancreatic inflammation in patients with CPI-DM with shrinkage of pancreases, increased pancreatic enzymes, and in a case from a patient who died with CPI-DM, peri-islet lymphocytic infiltration. In the NOD mouse model, anti–PD-L1 but not anti–CTLA-4 induced diabetes rapidly. RNA sequencing revealed that cytolytic IFN-γ ⁺ CD8 ⁺ T cells infiltrated islets with anti–PD-L1. Changes in β cells were predominantly driven by IFN-γ and TNF-α and included induction of a potentially novel β cell population with transcriptional changes suggesting dedifferentiation. IFN-γ increased checkpoint ligand expression and activated apoptosis pathways in human β cells in vitro. Treatment with anti–IFN-γ and anti–TNF-α prevented CPI-DM in anti–PD-L1–treated NOD mice. CPIs targeting the PD-1/PD-L1 pathway resulted in transcriptional changes in β cells and immune infiltrates that may lead to the development of diabetes. Inhibition of inflammatory cytokines can prevent CPI-DM, suggesting a strategy for clinical application to prevent this complication.

Background: Persons with disabilities comprise more than one billion people in the world, yet they are one of the most discriminated groups and face significant health disparities. Particularly in developing countries, which contain 80% of the entire population with disabilities, these individuals experience major barriers in accessing sexual and reproductive health (SRH) services. Education is an important factor that greatly affects individuals’ SRH service utilization. Hence, we sought to investigate the relationship between education and SRH service utilization for persons with disabilities in Bangladesh. Methods: Using an explanatory sequential mixed-methods design, a total of 5000 persons with disabilities were surveyed for the quantitative component and 15 mini-ethnographic case studies were conducted with persons with disabilities for the qualitative component. Chi-squared tests and logistic regression analyses were performed on the survey data, while the qualitative interviews were coded and their SRH themes synthesized accordingly. Results: Our quantitative findings show that education statistically significantly increases persons with disabilities’ SRH service utilization of antenatal care, delivery care, postnatal care, and family planning (P < 0.05). Interestingly, for persons with disabilities, primary education shows increased adjusted odds of family planning use but is likely not enough to increase antenatal care, delivery care, or postnatal care use; secondary or post-secondary education may be required to improve utilization of these latter services. Qualitative findings support the association between higher education levels and greater SRH service use. Persons with disabilities of lower educational attainment held misinformation and distrust in SRH services and experienced mistreatment by SRH healthcare providers, discouraging them from seeking future SRH services. Conclusions: We report that higher formal education level is associated with greater SRH service use for persons with disabilities in Bangladesh. Formally educating persons with disabilities expands their SRH knowledge and familiarity with SRH services, as well as leads to more economic opportunities so they can afford SRH services. Increasing formal education levels for persons with disabilities, paired with integrating comprehensive sexuality education (CSE) in their schools, will likely help close the gap in SRH health disparities for this vulnerable population.

Eye diseases such as age-related macular degeneration (AMD) are major causes of irreversible vision loss. Early and accurate detection of these diseases is essential for effective management. Optical coherence tomography (OCT) imaging provides clinicians with in vivo, cross-sectional views of the retina, enabling the identification of key pathological features. However, manual interpretation of OCT scans is labor-intensive and prone to variability, often leading to diagnostic inconsistencies. To address this, we leveraged the RETFound model, a foundation model pretrained on 1.6 million unlabeled retinal OCT images, to automate the classification of key disease signatures on OCT. We finetuned RETFound and compared its performance with the widely used ResNet-50 model, using single-task and multitask modes. The dataset included 1770 labeled B-scans with various disease features, including subretinal fluid (SRF), intraretinal fluid (IRF), drusen, and pigment epithelial detachment (PED). The performance was evaluated using accuracy and AUC-ROC values, which ranged across models from 0.75 to 0.77 and 0.75 to 0.80, respectively. RETFound models display comparable specificity and sensitivity to ResNet-50 models overall, making it also a promising tool for retinal disease diagnosis. These findings suggest that RETFound may offer improved diagnostic accuracy and interpretability for specific tasks, potentially aiding clinicians in more efficient and reliable OCT image analysis.

Develop a machine learning model that predicts early-onset AMD based on early patient comorbidities DESIGN: Retrospective case-control study. This study used 2 datasets: a tertiary eye care center dataset (930 AMD patients, 392 early-onset, 538 late-onset) and the All of Us dataset for validation (560-580 AMD patients early-onset, 560-580 matched controls without AMD). Health records of AMD patients from a tertiary hospital (1999-2023) and the All of Us Research Program, a nationwide longitudinal cohort of U.S. adults (data collected 6/2016-2/2025) were collected. Unsupervised clustering using Uniform Manifold Approximation and Projection (UMAP) was performed on the tertiary eye care center dataset to identify distinct patient subgroups. Supervised machine learning models, including gradient-boosted decision trees (GBDT), logistic regression, and random forests, were trained to predict early-onset AMD. Model performance was evaluated using the area under the receiver operating characteristic curve (AUC), while feature importance was assessed using the Gini index for tree-based models and coefficient magnitudes for logistic regression. Additionally, odds ratios for comorbidities associated with early-onset AMD were estimated using 2 × 2 contingency tables in the validation dataset (All of Us). Clustering identified 2 distinct patient groups, with 1 (93.08% early-onset AMD dominated) showing high rates of inflammatory joint disorders, hypertension, and dyslipidemia. Multiple supervised machine learning models achieved ∼76% accuracy in predicting early-onset AMD from early comorbidities (diagnosed before age of 55). Like clustering, analysis of supervised machine learning models identified inflammatory joint disorders, hypertension, and dyslipidemia as the most predictive features. These comorbidities were validated using the All of Us dataset, with essential hypertension (OR: 5.80, 95% CI: [4.25-7.93], P < .05), rheumatoid arthritis (OR: 4.61, 95% CI: [1.55-13.72], P < .05), and hyperlipidemia (OR: 5.16, 95% CI: [3.76-7.08], P < .05) showing strong associations with early-onset AMD. Screening hypertension, dyslipidemia, and inflammatory joint disorders in ophthalmologic evaluations may enable earlier AMD detection and intervention.

To discover novel systemic associations that may lead to idiopathic epiretinal membrane (iERM) using interpretable machine learning models. Large data retrospective case-control study. All of Us Dataset, including a total of 10 380 patients: 2015 iERM patients with 2015 1:1 matched controls, 3175 secondary epiretinal membrane (sERM) patients with 3175 1:1 matched controls. Electronic health records of epiretinal membrane (ERM) patients from the All of Us Research Program, a nationwide longitudinal cohort of US adults (data from 6/2016 to 2/2025) were collected. Unsupervised clustering using principal component analysis was performed on the data set to identify distinct patient subgroups. Supervised machine learning models, including gradient-boosted decision trees and logistic regression, were trained to predict iERM. Model performance was evaluated using the area under the receiver operating characteristic curve (AUC), while feature importance was assessed using the Gini index for tree-based models and coefficient magnitudes for logistic regression. Additionally, odds ratios for comorbidities associated with both iERM and sERM were estimated using 2 × 2 contingency tables. Unsupervised clustering of iERM patients revealed 4 distinct subgroups characterized by unique systemic comorbidity profiles, including cardiometabolic, dermatologic, and joint disorder pathways. Clusters demonstrated significant associations with systemic conditions such as hypertension, hyperlipidemia, type 2 diabetes, inflammatory skin conditions, osteoarthritis, and anemia. Supervised models, including logistic regression and gradient-boosted decision trees, achieved AUC values exceeding 0.679 on a testing set. Key predictors of iERM included knee osteoarthritis, hyperlipidemia, essential hypertension, and sensorineural hearing loss, each demonstrating high coefficient magnitudes, Gini importance, and statistically significant odds ratios. This study challenges conventional distinctions between iERM and sERM, proposing systemic comorbidities as associations to ERM development. The observed associations with cardiometabolic dysfunction, chronic inflammation, and joint and dermatologic disorders suggest that systemic mechanisms may significantly influence ERM pathogenesis. Future studies are necessary to establish causality and explore targeted therapeutic approaches, potentially incorporating anti-inflammatory treatments or cardiovascular risk management to prevent ERM formation. These findings highlight opportunities for personalized risk assessment and preventative interventions based on systemic comorbidity profiles. Proprietary or commercial disclosure may be found in the Footnotes and Disclosures at the end of this article.