Blog

Semaglutide or tirzepatide treatment among patients with diabetes was associated with an increased risk of optic nerve disorders, such as nonarteritic anterior ischemic optic neuropathy (NAION), according to a study published in JAMA Network Open.¹

“Glucagon-like peptide-1 receptor agonist (GLP-1RA) medications have transformed the treatment of type 2 diabetes (T2D) and obesity, with associated reductions in cardiovascular and nephrological complications,” wrote authors of the study. “Semaglutide and tirzepatide are second-generation GLP-1RA medications approved by the FDA for the treatment of diabetes and obesity.”

GLP-1s and their ability to transform diabetes care and obesity management have been increasingly noticeable as of late. According to a KFF Health tracking poll, at least 12% of US adults said they’ve taken a GLP-1RA at some point in time. For the sheer recognition of these medications amongst the public, 32% of US adults said they’ve heard “a lot” about GLP-1RAs, which is a 13% increase from 2023 to 2024.²

READ MORE: Initiation of Semaglutide Does Not Decrease Health Care Spending

Despite the notable rise in GLP-1 use in recent history, this drug class is often associated with adverse effects typically related to gastrointestinal outcomes, such as nausea, vomiting, diarrhea, gastroparesis, and constipation.^1,3 Recent studies, however, have shown associations between NAION and semaglutide use, leading to concerns of adverse outcomes among diabetes patients outside of the digestive system.

“Several retrospective studies from 2024 and 2025 reported a potential association of semaglutide with NAION in patients with diabetes and patients with obesity,” they continued.¹ “However, a meta-analysis of randomized clinical trials did not detect an association of GLP1-RA therapy, including semaglutide, with NAION.”

With conflicting evidence reported in the past few years, researchers wanted to better determine the association between NAION and GLP-1 use—namely semaglutide and tirzepatide. They hoped the significant reach of study participants within this new cohort would better inform pharmacist and provider decisions when counseling patients with diabetes, obesity, or a general need for GLP-1RAs.

“It remains unknown if semaglutide or tirzepatide is associated with other optic nerve and visual pathway disorders,” wrote the authors.¹ “This study leveraged a nationwide, multicenter database of electronic health records (EHRs) of more than 118 million US patients to conduct rigorous target trial emulation in patients with T2D to examine associations of semaglutide or tirzepatide with optic nerve and visual pathway disorders, including NAION.”

In this cohort study, researchers included EHR patient data from December 2017 to January 2023. Inclusion criteria called strictly for study participants with T2D, no previous diagnoses of eye disorders, and prescriptions for either semaglutide, tirzepatide, or another antidiabetic drug.

Using target trial emulation framework, researchers conducted their study to determine the causal effects of GLP-1s on outcomes like NAION.⁴ With this specific study design, they separated participants into 2 groups: those prescribed the GLP-1RAs semaglutide or tirzepatide and those prescribed alternative antidiabetic medications.¹

Among patients reporting use of semaglutide, tirzepatide, or other antidiabetic medications, the main outcomes researchers explored were any first-time diagnoses of disorders of optic nerve and visual pathways. Some of these disorders included optic neuritis, NAION, optic atrophy, and more.

The final analysis included a total of 159,398 patients, with an even split of 79,699 participants in the GLP-1RA group (mean age, 56.8 years; 51.7% women) and the other antidiabetic medications group (mean age, 56.2 years; 52.6% women).¹

Including all disorders explored in this study, the main associations between diabetes treatment and optic nerve disorders were only notable for NAION and “other” optic nerve disorders outside of those included in the study. However, among these associations, patients treated with semaglutide or tirzepatide were more likely to develop optic nerve disorders compared with patients on other antidiabetics.

Indeed, a total of 35 patients (0.04%) on semaglutide or tirzepatide reported NAION diagnoses compared with just 19 (0.02%) from the antidiabetic medication group. Furthermore, 93 participants (0.12%) reported other various optic nerve disorders compared with just 54 patients in the antidiabetic group.

“In a population of patients with T2D who had no prior diagnosis of eye diseases, this cohort study found that semaglutide or tirzepatide compared with other antidiabetic medications was associated with a differential risk of optic nerve and visual pathways, including increased risk of NAION and other optic nerve disorders, but not optic neuritis, papilledema, optic atrophy, or optic disc orders,” they continued.¹

With the overall risk of developing optic nerve disorder being minimal, the researchers findings are still significant enough to inform future use cases of any antidiabetic drug. Despite aligning with previous findings from 2024 and 2025, researchers believe further investigation is needed to replicate and confirm findings from the current study.

“Future studies are needed to replicate these findings, explore underlying mechanisms, identify individuals at increased risk for these potential complications, and examine other eye disorders,” concluded the authors.¹

READ MORE: Diabetes Resource Center

Don’t get left behind: Sign up today for our free Drug Topics newsletter and get the latest drug information, industry trends, and patient care tips delivered straight to your inbox.

References

1. Wang L, Volkow ND, Kaelber DC, Xu R. Semaglutide or tirzepatide and optic nerve and visual pathway disorders in type 2 diabetes. JAMA Netw Open. 2025;8(8):e2526327. doi:10.1001/jamanetworkopen.2025.26327

2. Montero A, Sparks G, Presiado M, et al. KFF Health tracking poll May 2024: the public’s use and views of GLP-1 drugs. KFF. May 10, 2024. Accessed August 12, 2025. https://www.kff.org/health-costs/poll-finding/kff-health-tracking-poll-may-2024-the-publics-use-and-views-of-glp-1-drugs/

3. Catanese L. GLP-1 diabetes and weight-loss drug side effects: “Ozempic face” and more. Harvard Health Publishing. February 5, 2024. Accessed August 12, 2025. https://www.health.harvard.edu/staying-healthy/glp-1-diabetes-and-weight-loss-drug-side-effects-ozempic-face-and-more

4. Fu EL. Target trial emulation to improve causal inference from observational data: what, why, and how? J Am Soc Nephrol. 2023 Aug 1;34(8):1305-1314. doi: 10.1681/ASN.0000000000000152.

Characterization and relatedness of the isolates

The susceptibility profiles of six clinical A. pittii isolates are presented in Table 1. All isolates were susceptible to aminoglycosides, tetracyclines and fluoroquinolones. All isolates were susceptible to ceftolazone/tazobactam, colistin, ampicillin/sulbactam, trimethoprim/sulfamethoxazole and nitrofurantoin; and resistant to ampicillin, first generation cephalosporin cefazolin, 3rd generation cephalosporins ceftriaxone and cefotaxime and aztreonam. Two isolates (AP290R and AP5092) were intermediate to piperacillin/tazobactam and resistant amoxicillin/clavulanic acid; only one isolate (AP290R) was resistant to all carbapenems tested and had elevated levels of ceftazidime (8 ug/ml).

The six A. pittii isolates and reference strain ATCC19606 were analyzed by FT-IR to determine phenotypic similarity. No clusters (indicating similar isolates) were evident (Fig. 1A). To determine whether FT-IR could distinguish between A. pittii and A. baumannii, we added 11 A. baumannii strains, belonging to different sequence types (STs), to the FT-IR analysis. Still no clusters were identified, and A. pittii isolates were not always more similar to each other than to A. baumannii isolates (Fig. 1B).

All 6 A. pittii strains were sequenced and their genomes were analyzed. The isolates belonged to different STs (Table 2) with different KL types. We constructed a phylogenetic tree based on pangenome analysis (Supplementary Fig. 1). The tree revealed distinct separation, with consistent and even distances observed between each isolate. Next, we conducted core genome alignment of the six A. pittii isolates compared to 64 publicly available complete genomes of A. pittii (Supplementary Table 1) from different countries and different years. The genomes represent a temporal range spanning from 2011 to 2024, allowing for examination of potential genomic changes over nearly three decades. This analysis placed AP5092 and MML4 (isolated in 2023, Hong Kong) on one branch close to each other. Other four isolates were located on a second branch, with AP5047 and AP290R most closely related to each other (Fig. 2). Isolate AP290R most close to isolate HCG18 isolated in Mexico in 2023.

Antibiotic resistance genes

Genomic analysis revealed the presence of several antimicrobial resistance determinants among the isolates (Table 3).

All isolates carried at least two genes conferring β-lactam resistance: a variant of ampC cephalosporinase (ADC) and a variant of OXA-type β-lactamase. Notably, isolate AP290R carried three bla_OXA genes – bla_OXA−272, bla_OXA−255 and bla_OXA−72 carbapenemases. Correspondingly, this was the only isolate to display carbapenem resistance.

In addition, all six study isolates harbored genes conferring quinolone resistance (adeF and abaQ), efflux pump components (abeS, adeF and adaQ), and a gene involved in colistin resistance (lpsB). Three isolates (AP4773, AP4968, and AP5092) contained a single putative aminoglycoside resistance gene, ant(3’’)-IId, with relatively low sequence homology (69%) to the closest match in the Comprehensive Antibiotic Resistance Database (CARD).

Virulence studies

In vitro phenotype

We next evaluated characteristics of the A. pittii isolates which are classically associated with bacterial virulence – growth, motility and biofilm formation. A. baumannii reference strain ATCC19606 served as a reference strain. The growth of all isolates was similar to that of the control strain (p > 0.05; Supplementary Fig. 2). Three A. pittii isolates demonstrated significantly higher motility than the control stain (p < 0.0001) (Fig. 3A), and four isolates produced significantly less biofilm (p < 0.001) (Fig. 3B).

All A. pittii isolates were serum sensitive (Fig. 3C), displaying either death (4/6 isolates) or reduced growth (2/6 isolates) after 4 h exposure to NHS. Heat inactivation of the complement system reduced the serum sensitivity of 5/6 isolates (except for AP5091), with one isolate (AP5092) displaying full resistance to the inactivated serum.

In vivo virulence

We next evaluated the virulence of the A. pittii isolates in vivo, using killing assay in Z. morio larvae. In this model, AP290R exhibited high virulence potential (Fig. 4), killing 70% of infected Z. morio larvae within 24 h. The other A. pittii strains were less lethal, with a 10%−20% lethality rate 24 h post infection and a 25%−35% lethality rate 7 days post infection.

Virulence factors

AP290R possessed the largest number of virulence genes, and AP4773 and AP5092 – the smallest. Nine virulence genes (cpaA, entE, gspDE1E2FINO, bauCDF, pilBCFGHMTU2D, pbpG, lpxABCL, barB, lbsB) were present in all 6 isolates (Fig. 5). These genes are related to various virulence functions, including biofilm formation, adherence, coagulation (cpaA), siderophore biosynthesis (entE), motility, and others. Several differences in genomic content were evident. The cluster of three genes – pilA, pilQ (surface motility) and basI (siderophore biosynthesis) – was missing from AP290R, AP4773, AP4968 and AP5091. However, only pilA was missing in isolate AP5092, and only basI was missing in isolate AP5047. The iron acquisition gene barA was missing in AP5047.

Plasmid content

Whole genome analysis revealed that two A. pittii isolates carried plasmids. Properties of these plasmids are described in Table 4 and the specific ORFs – in Supplementary Table 3. Plasmid p290R (Fig. 6A) carried ten hypothetical proteins and 4 genes of known function: ydhP (associated with glycosidase and hydrolase activity), azoR1 (azoreductaze), gcvA (regulates the glycine cleavage system, transcriptional activator), and most significantly – the carbapenemase bla_OXA−72 gene. The much larger plasmid p5092 (Fig. 6C) carried 119 genes, among them genes involved in different metabolic pathways, transporters and transcriptional regulators, and several insertion sequences (ISAba46, ISAba22, ISAba23, ISAha3, ISAcsp3, ISAba26 and IS1301), and 79 hypothetical proteins.

BLASTn analysis revealed a significant number of comparable plasmids within the public database, all reported in Acinetobacter spp. From this pool, we selected the ten most closely related plasmids for each of the plasmids (pAP5092 and pAP290R) and constructed a phylogenetic tree (Fig. 6B, D). The plasmid closest to pAP290R was pSU8507_OXA-2 (accession number LC777725.1) isolated from A. pittii in Japan in 2023, followed by 5 closely related plasmids from the US, China and Spain isolated in 2017–2021 (Fig. 6B). The plasmid closest to pAP5092 found in the database was pML4_1 (accession number CP118934.1) isolated from A. pittii in Hong Kong in 2023 (Fig. 6D). Additional information regarding accession numbers, year and country of isolation can be found in Supplementary Table 2.

Plasmid pAP290R contained the bla_OXA−72 gene (an OXA-24 family carbapenemase). To test the inter-species transferability of this plasmid, it was extracted from A. pittii AP290R and electroporated it into AB2142, a carbapenem-susceptible A. baumannii strain. Plasmid pAP290R conferred resistance to beta-lactams and carbapenems (Supplementary Tables 4 and Supplementary Fig. 3). However, resistance to ceftazidime in AP290R seems to be intrinsic; it was not transferred to AB2142 by the pAP290R plasmid.

Crucially, the plasmid also significantly increased the virulence of AB2142 (Fig. 7), suggesting that it contributes to the virulence of AP290R, raising the possibility that some of the hypothetical genes it carries are in fact virulence factors.

The diagnosis of chronic obstructive pulmonary disease, or COPD, is improved by incorporating CT lung imaging and respiratory symptoms. This fresh, multidimensional approach better identifies patients who are at risk of poor respiratory outcomes, while ruling out those who have airflow obstruction without respiratory symptoms or structural lung disease, Surya Bhatt, M.D., and colleagues report in the Journal of the American Medical Association.

This new COPD diagnostic schema, which includes chest imaging, respiratory symptoms and spirometry, identified additional individuals at risk of poor respiratory outcomes.”

Surya Bhatt, M.D., professor, Department of Medicine, University of Alabama at Birmingham Marnix E. Heersink School of Medicine and director, Center for Lung Analytics and Imaging Research

The schema includes airflow obstruction on spirometry as the major criterion and minor criteria based on chest imaging and respiratory symptoms. Imaging includes visual signs of emphysema and airway wall thickening on computed tomography, and symptom-based criteria consider difficulty breathing, reduced quality of life and the presence of chronic bronchitis. Under the new framework, a patient must have airflow obstruction and at least one minor criterion or, in the absence of airflow obstruction or if lung function tests are not available, at least three of five minor criteria.

Bhatt and colleagues found that, among 9,416 participants enrolled in a multicenter cohort, those newly diagnosed with COPD by the schema had greater all-cause and respiratory-specific mortality, more frequent exacerbations, and faster lung function decline compared with people classified as not having COPD based on the new classification. This new diagnostic schema included additional individuals with high respiratory morbidity and excluded some with airflow obstruction who had no symptoms or evidence of structural lung disease.

Before this study, clinicians had increasingly recognized that lung function tests did not capture all aspects of the complex heterogeneous disease COPD, a leading cause of disability and death. Some 392 million people globally, and 16 million in the United States, are estimated to have COPD.

“This new diagnostic schema will likely change the way we diagnose COPD and enable its diagnosis, even in the absence of overt airflow obstruction on spirometry,” Bhatt said. “Whether treating individuals newly diagnosed this way will result in improved outcomes remains to be tested, but practitioners have already been using imaging and symptoms to diagnose COPD. This new schema sets some parameters to operationalize this.”

In an editorial in JAMA, Francesca Polverino, M.D., Ph.D., called the study a milestone in COPD diagnosis. “COPD classification has remained overly dependent on airflow limitation as the main diagnostic criterion,” Polverino wrote. “What truly sets this reclassification apart is its groundbreaking assertion that airflow obstruction is no longer a requirement for a COPD diagnosis.”

In the system proposed by Bhatt and colleagues, airflow obstruction remains the major criterion for COPD. “However, what makes this model significantly more inclusive and reflective of clinical COPD diversity are the minor criteria, which are split between imaging and symptom-based factors,” said Polverino, a professor at the Baylor College of Medicine, Houston, Texas.

Bhatt is corresponding author of the study, “A new multidimensional diagnostic approach for chronic obstructive pulmonary disease,” and Edwin K. Silverman, M.D., Ph.D., Brigham and Women’s Hospital, Boston, Massachusetts, and James D. Crapo, M.D., National Jewish Health, Denver, Colorado, are co-senior authors. Other UAB authors include Mark Dransfield and Sandeep Bodduluri, UAB Department of Medicine Division of Pulmonary, Allergy and Critical Care Medicine; and Arie Nakhmani, UAB School of Engineering Department of Electrical and Computer Engineering.

The study includes 52 authors from 24 universities and institutions in the United States and Canada.

At UAB, Bhatt works in the Division of Pulmonary, Allergy and Critical Care Medicine.

Source:

University of Alabama at Birmingham

Journal reference:

Bhatt, S. P., et al. (2025). A Multidimensional Diagnostic Approach for Chronic Obstructive Pulmonary Disease. JAMA. doi.org/10.1001/jama.2025.7358.

When it comes to matters of life and death, could human survival depend on our mastery of game strategy?

For biologists, evolutionary game theory is a way of studying how different traits or behaviors evolve in a population over time, impacting the probability of a population’s survival in relation to the strategies present in other competing populations. This creates a constantly shifting balance – comparable to a dynamic and ever-changing version of rock-paper-scissors – in which strategy is a matter of timing and knowing your opponent.

A research team led by Paul Newton, professor of aerospace and mechanical engineering, mathematics, and quantitative and computational biology at USC Viterbi School of Engineering, has published a new paper in PNAS demonstrating how principles of game theory can be applied to advance cancer therapy.

The authors of the paper have developed a mathematical model that taps into the dynamics of the cancer-immunity cycle, predicting the competition of cancer cells, healthy cells, and immune system cells (T-cells). The insights developed from the model have the potential to allow medical practitioners to effectively “game” the cycle – synchronizing treatment schedules based on the battles taking place in the human body.

“You can think of a tumor as an ecosystem consisting of cancer cells competing with healthy cells,” said Newton. “Chemotherapy and immunotherapy are essentially attempts to steer the evolution of the tumor in a beneficial way. But that’s not how oncologists have typically framed approaches to treatment.”

The new paradigm proposed by the paper will seem shocking to some – after all, we’re used to thinking in terms of eliminating cancer cells, not trying to get the tumor on our side. The trouble is, the elimination method rarely works; the cells that are most sensitive to the chemotherapy will be killed off, while those that have developed resistance via mutations will survive. Those resistant cells regrow, leading to cancer recurrence.

“The motivation of our model is to develop an evolutionary game theory model which incorporates this selection dynamics of these competing cell populations,” said Newton. “Of course, we want to reduce the size of the tumor by killing some of the sensitive cells – but if you kill all of them, then the resistant cells are going to take over.”

As war games go, this is among the more complex. The immune system is an ally that requires careful management, as T cell populations start to attack the cancer cells and shape the rise and decline of different subpopulations of cells. In Newton’s framework, cancer cells act as defectors in a population dynamics game, while healthy cells act as cooperators, and the immune system serves as a dynamic regulator that modulates the rules of the game through feedback.

“Our thinking about the cyclical process of how cancer cells interact with T cells was influenced by Daniel S. Chen and Ira Mellman’s influential paper ‘Oncology meets immunology: the cancer-immunity cycle,’” Newton explained. “We set out to ask a series of important questions that build upon this foundation. What are the benefits of synchronizing chemotherapy and immunotherapy schedules with the cycle, as predicted by mathematical modelling? And could this strategic timing of treatment enable lower doses with the same – or greater – positive impact as standard doses?”

The work conducted by Newton’s team represents one of the first comprehensive mathematical models to treat the cancer-immunity cycle as a dynamic, game-theoretic system. If validated in patient populations, the findings could reshape how oncologists schedule combination therapies – not just based on standard cycles or tolerance, but on personalized biological rhythms.

Measuring the exact period of a patient’s cancer-immunity cycle remains a challenge. But advances in real-time immune-monitoring – via circulating tumor DNA, immune profiling, or imaging – may soon make it possible. Newton’s team envisions future clinical protocols that use sparse data collection and statistical inference to approximate the cycle and adjust therapy in real time. “We’re not just fighting cancer we’re negotiating with it,” said Newton. “And timing is everything.”

Published on August 12th, 2025

Last updated on August 12th, 2025