A hospital stay can put an older relative’s independence at stake. Booklets, videos and staff training were meant to help people keep moving, but at 30 days, their mobility scores were much the same as with usual care.
Moving little in the hospital goes along with losing independence. Earlier programs that increased movement needed resources that everyday hospital care often lacks.
The trial included 383 adults aged 60 or older who lived in the community, across three Swiss hospitals. The 12 wards received either usual care or patient booklets and videos, staff training and posters. The median hospital stay was five days.
At 30 days, the adjusted gap between the program and usual care was 0.67 points on a 0 to 120 scale measuring how far and often people moved. The result left room for worse or better mobility with the program. Other clinical outcomes showed no consistent benefit. No serious unwanted effects were considered potentially related to the program.
Missing results could affect that comparison: 327 of 383 participants had a mobility score at 30 days.
The challenge is to turn a practical hospital program into better mobility after discharge.
🤖 Made with AI
1
1
Sources and notes
Multilevel Mobility Intervention and Life-Space Mobility in Older Hospitalized Medical Patients. Published in JAMA Network Open. DOI: 10.1001/jamanetworkopen.2026.43903
agingresearch.news/s/ewgvjn
This partially blinded, cluster randomized superiority trial assigned 12 general internal medicine wards at three Swiss hospitals to the program or standard care. Participants were community-dwelling adults aged 60 or older with planned hospital stays of at least 3 days. Of 1334 eligible patients, 383 participated: 183 in the program group and 200 in standard care. Median age was 75 years (IQR, 68 to 81); 234 participants were male (61.1%). Median stay was 5 days (IQR, 3 to 7). Follow-up lasted 180 days.
The program combined patient booklets and videos, staff training and ward posters. The primary outcome measured how far and often people moved, from their bedroom to beyond their town, at 30 days. Scores ranged from 0 to 120, with higher scores indicating greater mobility.
The adjusted difference was 0.67 points (95% CI, −4.35 to 5.68), leaving room for worse or better mobility with the program. The outcome was available for 327 participants (85.4%). Secondary outcomes included physical activity during hospitalization, new nursing home placements and quality of life; there were no consistent differences across secondary outcomes. No serious unwanted effects were considered potentially related to the program.
- Missing 30-day results could affect the comparison.
- The findings do not address younger patients or people already living in nursing homes.
5
Picture two people at 75, both thinking clearly. With more of an Alzheimer’s-related protein in the blood, estimated risk of developing thinking problems over five years was 17%, versus 7% with less.
Alzheimer’s-related changes can be detected years before symptoms appear. The question is what those early signs mean while someone is still thinking clearly.
The study included 1,644 adults in Minnesota and 534 in a second group, all aged 60 to 90 and initially thinking clearly. Blood tests measured p-tau217, a form of tau protein associated with Alzheimer’s disease. Follow-up assessments helped model risks over five and 10 years.
At age 75, estimated 10-year risk of mild thinking problems was 35% with higher blood levels versus 15% with lower levels. The estimates used the Minnesota group’s average sex and status for an inherited DNA variant.
The study did not test these estimates in younger adults or people who already had thinking problems.
Blood protein levels and age together could help put a number on the chance of future thinking problems.
🤖 Made with AI

ALT Can blood flag future thinking problems? A higher blood marker meant more risk of thinking problems. Lower blood marker: 7%; Higher blood marker: 17%. Estimated five-year risk at age 75. Sex and genetic status set to group averages. 1,644 and 534 unimpaired adults · Observational study. Peer reviewed.
1
5
Sources and notes
Absolute Risk of Progression to Cognitive Impairment Based on Plasma Phosphorylated Tau 217, Age, Sex, and APOE ε4. JAMA Neurology. Journal article, not a preprint. DOI: 10.1001/jamaneurol.2026.4292
agingresearch.news/s/havge8
Two observational cohorts: Mayo Clinic Study of Aging (MCSA), 1,644 participants in Olmsted County, Minnesota; Alzheimer’s Disease Neuroimaging Initiative (ADNI), 534 participants across multiple centers. All were aged 60 to 90 and cognitively unimpaired at baseline. Data covered October 2008 to September 2025 in MCSA and September 2005 to January 2026 in ADNI.
Models estimated progression to mild cognitive impairment (MCI) and dementia over 5 and 10 years using age, sex, APOE ε4 (an inherited DNA variant), and p-tau217 (a form of tau protein), or its ratio to amyloid β 42, another protein.
At age 75, setting sex and APOE status to MCSA means (0.511 and 0.268), p-tau217 of 0.072 versus 0.324 pg/mL gave 5-year MCI risks of 7% (95% CI, 5%-8%) versus 17% (13%-20%); 10-year risks were 15% (12%-18%) versus 35% (29%-41%). P < .001 for both.
At 0.324 pg/mL, ages 65 versus 85 had 5-year MCI risks of 6% (95% CI, 4%-8%) versus 37% (32%-43%) and 10-year risks of 17% (12%-21%) versus 66% (57%-74%). P < .001 for both.
Comparing 0.324 with 0.072 pg/mL (90th versus 10th percentile), progression hazard ratios were, in order, women without ε4, women with ε4, men without ε4, and men with ε4:
- MCSA: 2.66 (95% CI, 2.02-3.51); 3.02 (2.21-4.12); 2.62 (1.99-3.46); 2.79 (2.04-3.81).
- ADNI: 2.67 (95% CI, 1.86-3.85); 2.83 (1.89-4.22); 2.08 (1.48-2.92); 3.99 (2.59-6.13).
All intervals above are 95% confidence intervals. Results were similar using p-tau217/amyloid β 42.
- These modeled risks concern adults aged 60 to 90 who began without cognitive impairment.
- They do not describe younger adults or people already impaired.
14
What happened to older adults’ thinking as their social lives shrank in the pandemic? Among 1,679 adults, those whose participation dropped declined more than peers who stayed involved, resembling people who stayed isolated.
Earlier work tied persistent isolation and loneliness to poorer memory. This paper asks what happened when older adults’ community participation changed during the pandemic.
The analysis included 1,679 older adults for thinking scores and 2,457 for depressive symptoms. It tracked community participation from 2016 to 2020 and changes in both outcomes from 2016 to 2022, comparing people whose participation fell, rose or stayed consistent.
Those whose participation dropped declined more in thinking than consistently active peers, resembling people who stayed isolated. People who started participating held up comparably to consistently active peers. The drop did not predict worse long-term depressive symptoms. The paper does not give the size of the cognitive difference.
Health changes could explain both reduced participation and poorer thinking, despite the analysis accounting for physical health constraints.
Testing support for social activities would help answer whether participation itself protects thinking skills in later life.
🤖 Made with AI
1
3
Sources and notes
The Long-term cognitive and mental health impacts of pandemic-induced social isolation: a trajectory analysis. The Journals of Gerontology, Series B. Journal article, not a preprint. DOI: 10.1093/geronb/gbag220
agingresearch.news/s/4i3w2v
A nationally representative longitudinal analysis of older adults included 1,679 for cognition, meaning thinking skills, and 2,457 for depressive symptoms. Community participation trajectories from 2016–2020 were consistently isolated, pandemic drop-off, consistently engaged and pandemic initiators. Changes in cognition and depressive symptoms were assessed from 2016 to 2022. Models used survey weights to reflect the population and adjusted for starting scores, social and economic disadvantages, and physical health constraints. The paper does not give counts for individual participation groups.
Compared with consistently engaged peers, adults whose participation dropped had greater cognitive decline by 2022, resembling consistently isolated adults. People who began participating preserved cognition comparably to consistently engaged peers. Drop-off did not predict worse long-term depressive symptoms. Childhood adversity independently predicted poorer mental health.
Models using two-category measures detected no change in the relationships between participation and outcomes according to childhood adversity. Checks using measures across a range suggested a possible interaction between starting social participation later in life and cumulative early-life trauma, favoring cognitive preservation. Effect sizes, confidence intervals and p-values are not given.
Limit: Health changes could have contributed to both reduced community participation and cognitive decline despite adjustment for physical health constraints.
3
Keeping diabetes out of later life is a reason to care about what's on your plate. People who ate closer to a plant-rich diet had less diabetes, with the lowest risk at 84 points out of 130.
Vegetables, fruit, nuts and beans take priority over animal foods in the Planetary Health Diet. Earlier studies tied it to lower death rates and less heart and blood vessel disease.
More than 100,000 UK adults, averaging around 55 years old, completed at least two questionnaires about the previous day's food and drink. Diet was scored from zero to 130; health records tracked disease over a median 12.5 years.
Each five-point higher score went with 3% lower stroke risk and 3% lower type 2 diabetes risk, compared with a score five points lower. Diabetes risk was lowest at 84 points. There was no link with heart disease caused by reduced blood supply, or with having two or more of the diseases studied.
The catch is other health advantages: people with higher scores were more educated and less often overweight or obese. Adjustments covered these differences but could miss others; extra adjustment for medication use left the findings much the same.
The next test is whether changing the diet actually prevents a first illness.
🤖 Made with AI
1
4
Sources and notes
Association of Planetary Health Diet with decreased risk of cardiometabolic diseases, multimorbidity, and its progression: a prospective cohort study. Annals of Medicine. Not a preprint. DOI: 10.1080/07853890.2026.2718534
agingresearch.news/s/2pwseu
Prospective UK Biobank study: 108,596 to 112,323 adults across disease analyses; 105,643 in progression analyses. Diet-score quarters were compared, with the lowest as reference. Participants completed at least two dietary recalls; median follow-up was 12.5 years. Each disease analysis excluded existing cases. Adjustments covered age, sex, ethnicity, education, employment, deprivation, body mass index, smoking, alcohol, activity and energy intake.
Per five-point score increase, stroke risk was 3% lower (hazard ratio 0.973, 95% confidence interval 0.951–0.997; p < 0.05 after adjustment for multiple comparisons); diabetes risk was 3% lower (0.968, 0.952–0.984; p < 0.001). Hazard ratios compare rates during follow-up. Diabetes hazard ratios for successive quarters above the lowest were 0.875 (0.805–0.951), 0.835 (0.773–0.902) and 0.881 (0.805–0.965), all with 95% confidence intervals; adjusted p < 0.01. Risk was lowest around 84 points (nonlinearity p = 0.04; trend p < 0.0001). No link emerged for heart disease caused by reduced blood supply or multiple diseases.
Highest versus lowest quarter: first-disease hazard ratio 0.89 (95% confidence interval 0.84–0.94); death before first disease 0.81 (0.73–0.89), both adjusted p < 0.001. Other disease transitions could reflect chance.
Findings stayed similar across demographic subgroups and after excluding disease during dietary assessment or the first two years, accounting for competing deaths, or adjusting for medications or menopause. Progression findings also stayed similar after excluding same-day transitions or using disease-specific deaths. E-values, measures of sensitivity to unmeasured confounding, ranged from 1.11–1.67 for disease outcomes and 1.11–1.77 for progression.
- Other health differences can remain after adjustment; diet was not assigned.
- Treatment after first disease was unaccounted for.
- Selection into the progression analysis could bias results.
- Recalled diets can be misreported; soybean intake may be missed.
- Participants were mainly White and recruited at ages 40–69, restricting generalization.
- Estimated ordering of same-day diagnoses can be inaccurate; excluding these cases left results similar.
6
For anyone worried about developing diabetes, a blood sample holds more than a sugar reading. A model using blood proteins ranked future diabetes risk better than a blood-fat measure alone in about 20,000 people.
The body's response to insulin, the hormone that regulates blood sugar, starts to falter early on the path to diabetes. Blood proteins can reflect that trouble across several organs.
The study began with 161 men and women with or without type 2 diabetes. Insulin infusions, with blood sugar held steady, measured how well their bodies responded. The model combined 13 blood proteins with sex, a longer-term blood sugar measure and a blood-fat ratio.
In about 20,000 UK Biobank participants, its score for ranking future diabetes risk was 0.86, versus 0.74 for the blood-fat ratio alone. A higher score means better ranking. But the longer-term blood sugar measure alone scored 0.88: the protein model performed similarly to that familiar test.
The initial insulin-testing groups were modest in size and mostly of European ancestry, leaving performance in other populations unsettled.
Its promise is broader risk prediction: the model also ranked fatty liver disease risk better than the blood sugar measure.
🤖 Made with AI
1
14
Sources and notes
Plasma proteomics identifies insulin resistance signatures predictive of cardiometabolic disease risk. Nature Metabolism; not a preprint. DOI: 10.1038/s42255-026-01662-1
agingresearch.news/s/he8g7b
Human protein-profiling and prospective prediction study: 161 adults with or without type 2 diabetes, in discovery (75), validation (43) and exercise (43 men) groups. Exercise lasted eight weeks. Validation matched sex, body mass index, age and smoking status. Predictions were tested in about 20,000 UK Biobank participants and 1,700 independent participants, with median follow-up of 18.1 years in the latter group.
Protein associations adjusted for sex; adding age and body fat reduced the associated set from 488 to 62 (adjusted P < 0.05). Model candidates adjusted for body mass index, age and sex.
The 13-protein model also included sex, longer-term blood sugar and a blood-fat ratio. Diabetes ranking scores (concordance indices; 95% confidence intervals): model 0.86 (0.84–0.89), blood-fat ratio 0.74 (0.71–0.77), longer-term blood sugar 0.88 (0.85–0.91). Fatty liver: model 0.78 (0.71–0.85), blood sugar 0.58 (0.48–0.69).
Insulin-sensitivity prediction: validation correlation R = 0.73 versus 0.49 with clinical features alone; cross-validation R = 0.67.
Exercise improved insulin sensitivity by 30–40% and aerobic capacity by 8–15%; 22 proteins changed (adjusted P < 0.05). Responses differed across metabolic groups.
Independent follow-up hazard ratios (relative rates; 95% confidence intervals): diabetes 0.49 (0.39–0.61), obesity 0.61 (0.48–0.76), chronic kidney disease 0.5 (0.41–0.61), sleep apnea 0.68 (0.53–0.87), death 0.56 (0.45–0.71), stroke 0.66 (0.5–0.87), chronic obstructive pulmonary disease 0.63 (0.5–0.8), lung cancer 0.62 (0.46–0.84). Higher predicted sensitivity meant lower risk. A glucose-challenge index was the strongest single diabetes predictor: 0.21 (0.12–0.36). For most other outcomes, the model matched or exceeded fasting measures.
- The insulin-testing groups were modest in size and mainly of European ancestry, constraining generalizability.
- Organ assignments used gene-activity databases rather than protein databases.
- Model inputs were restricted to proteins available on one testing platform.
- Medication effects on blood proteins were not systematically studied.
- Exercise subgroup analyses were exploratory, with small groups and unadjusted P values.
- Broader populations and repeated sampling are needed to assess prediction stability over time.
13
Feeling lonely came before more trouble remembering words and getting around. Those problems also came before loneliness, but the pattern was weaker in that direction across 12 years of older adults’ lives.
Earlier findings on loneliness and later changes in thinking have been mixed. An earlier study using smartphone assessments found limited links between the two.
The new study covered 22,097 adults aged 50 and older in 13 European countries across six surveys over 12 years. Loneliness meant feeling left out, isolated or short of companionship. Memory was tested by recalling words; physical limitations included difficulty walking and climbing stairs.
The links were stronger from loneliness to later memory problems, depressive symptoms and physical limitations than in reverse. More long-term illnesses predicted later loneliness; loneliness predicted more illnesses only between the first two surveys.
Personality was not accounted for and could influence both loneliness and health. Adjusting for social relationships, health behaviors and other measured factors reduced some links, but generally preserved their direction.
The next test is whether easing loneliness helps preserve memory, mood and mobility.
🤖 Made with AI
1
2
Sources and notes
Does loneliness cause poor health or vice versa? A longitudinal analysis across European countries. European Journal of Ageing. Journal article, not a preprint. DOI: 10.1007/s10433-026-00946-5
agingresearch.news/s/hg5twg
This observational analysis included 22,097 adults aged 50 and older in 13 European countries. Six survey waves covered 12 years. Participants had to complete at least three consecutive waves. People reporting Alzheimer’s disease or dementia at baseline, or missing loneliness or health measurements, were excluded. Average age was 64; 57% were women, and 10.92% met the loneliness threshold at baseline.
Cross-lagged models compared earlier loneliness with later health and earlier health with later loneliness. Adjustments included sociodemographic characteristics, social relationships, health behaviors, stressful lifetime events and country. There were no assigned treatment groups.
Greater loneliness predicted later poorer word recall, more depressive symptoms and more physical limitations. Each also predicted later loneliness. The standardized associations were larger from loneliness to these health outcomes at every time point.
More chronic illnesses predicted later loneliness across all subsequent waves. Loneliness predicted more chronic illnesses only between the first two waves; later estimates could be chance. Comparing unadjusted and adjusted models generally preserved the direction of the relationships, although several estimates decreased after adjustment.
- Personality traits were not included and could influence both loneliness and health.
- The survey excluded people living in institutions, restricting its relevance to adults in long-term care.
- The models could not fully separate lasting differences between people from changes within the same person.
- Requiring consecutive participation selected a group whose excluded peers were older and reported slightly poorer health at baseline. The authors say this could underestimate the loneliness–health relationship in the broader older population.
- Adjusting for health behaviors may also account for pathways through which loneliness might affect health.
2
Imagine heart or blood vessel disease, diabetes and fatty liver piling up in the same person. Among 375,790 people who started free of them, those with better blood sugar handling scores were less likely to develop that mix.
The buildup of these illnesses matters for health in later life. The liver condition is a buildup of fat tied to metabolic problems. Insulin is a hormone that controls blood sugar.
The study tracked 375,790 UK Biobank participants initially free of cardiovascular disease, type 2 diabetes and this liver condition for an average 13.74 years. During that time, 8,984 developed combinations of these diseases. Of nine metabolic scores, an estimate of how well the body responds to insulin best distinguished future cases from non-cases.
Each increase of one standard deviation—a measure of variation among participants—in that score went with a 66% lower relative risk of developing multiple diseases. Higher scores also went with less progression from no disease to one disease, then to several.
Everyone began free of these diseases, so the ranking may differ for people who already have one.
The finding puts an estimate of insulin response first among these scores for spotting the risk of illnesses accumulating in later life.
🤖 Made with AI

ALT Who develops several diseases? An insulin response score best flagged future risk. Lower relative risk of multiple diseases: 66%. Per standard deviation higher insulin response score. Standard deviation: variation among participants. 375,790 initially disease-free people · Prospective cohort study. Peer reviewed.
1
10
Sources and notes
Associations of nine metabolic-related indices with the incidence, progression, and mortality outcomes of cardiovascular–liver–metabolic multimorbidity: evidence from a large prospective cohort. Frontiers in Nutrition. DOI: 10.3389/fnut.2026.1988246
agingresearch.news/s/9q8jjm
This prospective UK Biobank study included 375,790 participants initially free of cardiovascular disease, type 2 diabetes and metabolic dysfunction-associated steatotic liver disease: liver fat accumulation tied to metabolic problems. Mean follow-up was 13.74 years. The comparison was between nine metabolic indices. Analyses examined disease onset, discrimination between future cases and non-cases, and transitions between disease states. Mortality risks with increasing disease burden were also examined.
During follow-up, 8,984 participants developed combinations of these diseases. All nine indices were associated with disease onset and progression. Estimated glucose disposal rate, an estimate of the body's response to insulin, best distinguished future cases from non-cases. Its AUC, a measure of that discrimination, was 0.811 (95% confidence interval 0.807–0.815), outperforming the other eight indices.
Each one-standard-deviation increase—a measure of variation among participants—went with a 66% lower relative risk of multiple diseases: hazard ratio 0.34 (95% confidence interval 0.33–0.35). A hazard ratio compares the rate at which an outcome occurs during follow-up. The corresponding reductions were 65% for developing two of the diseases and 70% for developing all three. Higher scores also went with lower risks of progressing from no disease to one disease and then to multiple diseases.
- Other health differences could have contributed to the relationships between the scores and later disease.
- The findings may not apply to people who already had one of these diseases at enrollment.
6
Why do older adults with less education get fewer years of independence? Low activity accounted for more of that gap than smoking, drinking or obesity in a group of 754 older adults.
Social disadvantage goes with shorter lives and fewer years of independence. Here, disability meant needing help with essential daily activities, so the question was about both living longer and managing daily life.
The study covered 754 adults aged 70 or older in South Central Connecticut, using records from 1998–2024. Disability was checked monthly. Models estimated lifespan and years without disability at age 70, then simulated each lifestyle risk being absent.
For education-related gaps, low activity accounted for 35.4% of men's gap in years without disability and 67.4% of women's. For total life expectancy, the figures were 54.5% and 93.4%, respectively. Smoking, drinking and obesity contributed little to these social gaps.
Poor health could explain part of the pattern: it can make people less active and shorten their lives, and the simulations could not rule that out.
The practical question is whether making activity easier in later life can narrow these gaps.
🤖 Made with AI
1
8
Sources and notes
Contributions of lifestyle risk factors to socio-economic disparities in active and total life expectancy among older adults: A longitudinal cohort study. The Journals of Gerontology, Series A. DOI: 10.1093/gerona/glag260.
agingresearch.news/s/8c3zqk
This longitudinal cohort study used 1998–2024 data from 754 adults aged 70 or older in South Central Connecticut. Education, income, neighborhood disadvantage and alcohol consumption were assessed at baseline only. Smoking, low physical activity and obesity were updated at 18-month intervals. Disability was assessed monthly and meant dependence in essential daily activities.
Models estimated active life expectancy, meaning years without disability, and total life expectancy at age 70 separately for men and women. Simulations set lifestyle risks to absent, separately and together, to estimate their contributions to socioeconomic gaps.
Low activity made the largest modeled contribution to men's gaps by education and neighborhood disadvantage, and women's gaps by education, income and neighborhood disadvantage. Its estimated shares, with years without disability first and total life expectancy second, were:
- Men, education: 35.4% and 54.5%.
- Men, neighborhood disadvantage: 49.4% and 53.4%.
- Women, education: 67.4% and 93.4%.
- Women, income: 39.9% and 55.3%.
- Women, neighborhood disadvantage: 23.8% and 20.1%.
Alcohol consumption, smoking and obesity made negligible contributions to these gaps. These estimates concern differences between socioeconomic groups, rather than the overall health effects of each risk.
- Poorer health could reduce activity and affect survival; the simulations could not rule this out.
- The estimates may not apply to adults under 70 or populations outside South Central Connecticut.
For families thinking about health in later life, this pattern reaches widely: people with higher blood sugar and fat scores had higher rates of 283 diseases as the years went by.
Blood sugar and blood fat are combined in a score used to flag how poorly the body responds to insulin. Called the triglyceride-glucose index, it marks insulin resistance, when the body responds less well to insulin, without measuring it directly.
The study tracked new diagnoses in 361,175 adults, averaging 56.5 years old, for a median 13.5 years. It tested 1,127 diseases, accounting for other factors and the many comparisons.
Higher scores went with higher rates of 283 diseases. An exploratory statistical analysis attributed part of 217 of these links to chronic, low-level inflammation, though its estimated contribution was often small.
Underlying health or other differences could explain both higher scores and later illness. For urinary stones, stricter diagnosis rules and extra adjustments left the direction of the results unchanged.
Together, blood sugar and blood fat may mark a broader pattern of illness as people age.
🤖 Made with AI
2
10
Sources and notes
Triglyceride-glucose index and multiple diseases with systemic inflammation as a potential mediator in a phenome-wide association study. Scientific Reports. Published journal paper, not a preprint. DOI: 10.1038/s41598-026-74168-1
agingresearch.news/s/vw49nh
This prospective UK Biobank cohort included 361,175 participants, mean age 56.5 years; 53.2% were female. Median follow-up was 13.5 years. The study examined the triglyceride-glucose index against new diagnoses of 1,127 diseases. Cox proportional hazards models adjusted for multiple covariates, and P values were corrected for multiple comparisons.
Higher index values were associated with 283 diseases, or 25.1% of those tested, after adjustment and correction for multiple comparisons. Reported hazard ratios reached 4.61.
An exploratory mediation analysis estimated how much of each positive association might operate through a low-grade chronic inflammation score. Indirect effects passed the statistical threshold for 217/283 associations (76.7%). Estimated mediation exceeded 30% for 42 diseases, was 10–30% for 136, and was 0–10% for 39. The estimated contribution was small for many outcomes.
Targeted sensitivity analyses for anemia and urinary calculus, meaning urinary stones, kept the same direction under stricter outcome definitions and additional outcome-specific covariate adjustment.
- The index is a surrogate marker of insulin resistance; it does not measure resistance directly.
- Underlying health or other differences could contribute to both index values and later disease.
- Findings in UK Biobank adults may not apply to other populations.
- The mediation analyses were exploratory, rather than evidence of an established mechanism.
1
7
When shoulder arthritis makes surgery a worry, a drug that helps avoid it would matter. The drugs studied here fell short: replacements were more common among users, at 2.25% versus 1.77% among non-users.
Reports of fewer hip and knee replacements had raised hopes for these drugs in arthritis. They are GLP-1 receptor agonists, drugs that activate a hormone receptor.
Medical records covered 24,669 people with shoulder arthritis in each group, matched for other medical conditions. One group had prescriptions before diagnosis; the other did not. Average follow-up was 731 days for users and 1,220 days for non-users.
Shoulder replacements occurred in 2.25% of users versus 1.77% of non-users. Users also had lower levels of two markers of inflammation, but that finding came alongside more replacements.
The groups could still differ in how severe their arthritis was or how surgery decisions were made. Matching other medical conditions could not settle that.
The shoulder results give no sign of the protection reported for hips and knees.
🤖 Made with AI
1
10
Sources and notes
Do Glucagon Like Peptide-1 Receptor Agonists Modify Progression of Primary Glenohumeral Osteoarthritis? A Propensity Score-Matched Evaluation of Shoulder Arthroplasty Incidence. The Archives of Bone and Joint Surgery. Published journal paper, not a preprint. DOI: 10.22038/abjs.2026.95223.4279
agingresearch.news/s/hkrwn9
This retrospective database study compared people with primary shoulder osteoarthritis by whether they had prescriptions for GLP-1 receptor agonists before diagnosis. People with other shoulder conditions were excluded. Groups were propensity-score matched 1:1 to balance relevant medical conditions, leaving 24,669 people per group.
All comparisons below give users first, then non-users. Mean follow-up was 731 ± 694 versus 1,220 ± 1,006 days (P < 0.001).
Shoulder replacement incidence was 2.25% versus 1.77% (P < 0.001). Estimated survival without shoulder replacement was 95.50% versus 97.23% (log-rank P < 0.001); the paper does not give the time point for these estimates.
Major joint injections were recorded in 18.93% versus 16.87% (P < 0.001), although the mean number of injections per patient was lower among users: 2.48 ± 2.86 versus 2.74 ± 3.23 (P < 0.001).
Mean C-reactive protein and erythrocyte sedimentation rate, two markers of inflammation, were lower among users (both P < 0.001). Mean cortisol, a hormone, was similar between groups.
- Matching medical conditions could not rule out differences in arthritis severity or decisions about surgery.
- Follow-up was shorter among drug users.
The authors concluded that GLP-1 receptor agonist use showed no protection against progression to shoulder replacement, contrasting with findings reported for hip and knee osteoarthritis.
11









