Medical Bulletin 15/September/2025 - Video

A recent study led by researchers from the Yong Loo Lin School of Medicine at the National University of Singapore (NUS Medicine) has found that artificial intelligence (AI) analysis of retinal images can accurately predict a person’s risk of developing cognitive decline and dementia. Published in the journal Alzheimer’s & Dementia, the study introduces a deep-learning biomarker called RetiPhenoAge, which estimates the biological age of the retina using standard eye scans.

Jointly spearheaded by Professor Cheng Ching-Yu, Director of the Center for Innovation and Precision Eye Health, and Professor Christopher Chen, Deputy Chair of the Healthy Longevity Translational Research Programme, the study analyzed data from over 500 participants recruited from memory clinics across Singapore. Researchers discovered that a higher retinal biological age, as indicated by RetiPhenoAge, was associated with a 25% to 40% increased risk of cognitive decline or dementia over five years.

To validate their findings, the team extended the analysis to a large cohort of over 33,000 participants from the UK Biobank. The results remained consistent, with elevated RetiPhenoAge correlating strongly with increased dementia risk over a 12-year follow-up. The researchers also linked retinal aging to neurodegenerative changes in the brain and age-related alterations in blood proteins, using MRI brain scans and blood biomarkers to reinforce the biological relevance of the retinal biomarker.

“With RetiPhenoAge, we are able to non-invasively estimate an individual's biological age, offering valuable insights for both cognitive health management and broader aging research,” said Prof Cheng. “This can help doctors identify people at risk of cognitive decline or dementia, before symptoms appear, enabling more targeted interventions.”

The study marks a major advance in digital biomarker development and highlights how AI can be harnessed with existing, non-invasive tools to combat dementia.

Reference: Ming Ann Sim et al, A deep‐learning retinal aging biomarker for cognitive decline and incident dementia, Alzheimer's & Dementia (2025). DOI: 10.1002/alz.14601

Meta-analysis Reveals AI Tools Fail to Reliably Predict Suicide Risk

Machine learning algorithms, often seen as promising tools for revolutionizing mental health care, may not be as effective as hoped when it comes to predicting suicidal behavior. A comprehensive new study published in PLOS Medicine reveals that the accuracy of these AI models is too low to be clinically useful for screening or prioritizing high-risk individuals.

Led by Matthew Spittal of the University of Melbourne, the research team conducted a systematic review and meta-analysis of 53 studies from around the world. These studies applied machine learning algorithms to vast datasets of over 35 million medical records, including nearly 250,000 cases of suicide or hospital-treated self-harm. The goal was to assess whether AI could outperform traditional risk assessment tools in identifying individuals most at risk of future suicide or self-harm.

While the algorithms demonstrated high specificity—meaning they were good at identifying people unlikely to self-harm—they showed only modest sensitivity, failing to correctly identify many individuals who later presented with suicidal behavior. “Specifically, the researchers found that these algorithms wrongly classified as low risk more than half of those who subsequently presented to health services for self-harm or died by suicide,” the study noted.

Even among those classified as high-risk, only 6% went on to die by suicide, and fewer than 20% returned for treatment after self-harm, suggesting a high rate of false positives. “We found that the predictive properties of these machine learning algorithms were poor and no better than traditional risk assessment scales,” the authors said. “The overall quality of the research in this area was poor, with most studies at either high or unclear risk of bias.”

The findings align with existing clinical practice guidelines, which already caution against using risk assessment scores to determine aftercare strategies.

Reference: Spittal MJ, Guo XA, Kang L, Kirtley OJ, Clapperton A, Hawton K, et al. (2025) Machine learning algorithms and their predictive accuracy for suicide and self-harm: Systematic review and meta-analysis. PLoS Med 22(9): e1004581. https://doi.org/10.1371/journal.pmed.1004581

GLP-1 Drugs Used for Weight Loss May Be Linked to Unplanned Pregnancies in Young Women

A new study from Flinders University, published in the Medical Journal of Australia, has raised alarms over the widespread use of GLP-1 receptor agonists such as Ozempic among women of reproductive age, many of whom are not using effective contraception despite known pregnancy-related risks. The research, which analyzed data from over 1.6 million Australian women aged 18 to 49 between 2011 and 2022, found that only 21% of the 18,010 women who were first prescribed these medications had recorded contraceptive use.

Originally developed to treat type 2 diabetes, GLP-1 receptor agonists have become increasingly popular for weight management due to their appetite-suppressing effects. The study found that most women receiving prescriptions did not have diabetes, with more than 6,000 women initiating treatment in 2022 alone and over 90% of them lacking a diabetes diagnosis.

The data also revealed that 2.2% of women became pregnant within six months of starting treatment, with higher pregnancy rates seen in women with diabetes and those in their early thirties without diabetes. Notably, women with polycystic ovary syndrome (PCOS) were twice as likely to conceive, likely due to improved fertility resulting from weight loss.

A previous animal study from the University of Amsterdam linked GLP-1 use during pregnancy to fetal growth issues and skeletal abnormalities, adding urgency to the current findings. While human data is still limited, the potential risks are enough to prompt stronger clinical guidelines.

Lead author Associate Professor Luke Grzeskowiak, from Flinders University’s College of Medicine and Public Health, emphasized the concerning trend: “We're seeing widespread use of these medications among women of childbearing age, but very little evidence that contraception is being considered as part of routine care. These medications can be incredibly helpful, but they're not risk-free, especially during pregnancy." “We need to ensure that reproductive health is part of every conversation when these drugs are prescribed to any women of childbearing age.”

Researchers are calling for clearer prescribing guidelines and further investigation into GLP-1 safety during pregnancy.

Reference: Kailash Thapaliya, Arianne Sweeting, Black I Kirsten, Amanda Poprzeczny, Danielle Mazza, Luke E Grzeskowiak. Incidence of GLP‐1 receptor agonist use by women of reproductive age attending general practices in Australia, 2011–2022: a retrospective open cohort study. Medical Journal of Australia, 2025; DOI: 10.5694/mja2.70026