I thought I would have a look at that paper. I am not a fan of metformin I think the papers that suggested a health benefit for non-diabetics were generally subject to selection bias and there is good evidence against it.
Hence I have found the link downloaded the paper and had a chatGPT(5.5paid) analysis:
Looking at the analysis my basic first point is a question as to how replicable this study will end up as being.
https://www.cell.com/cell/fulltext/S0092-8674(24)00914-0
Paper
Yang Y. et al. âMetformin decelerates aging clock in male monkeys.â Cell 187 (2024): 6358â6378.
Overall summary
This study tested whether long-term metformin treatment can slow biological aging in healthy, non-diabetic male cynomolgus monkeys.
The investigators treated six older monkeys, initially 13â16 years old, with metformin 20 mg/kg daily for 1,200 daysâabout 3.3 years. Five untreated older monkeys completed the study after one control animal died of kidney failure. Separate young and middle-aged groups were used to define normal age-related trajectories.
The study combined:
- behavioural cognitive testing;
- MRI and CT imaging;
- histology across multiple organs;
- bulk RNA sequencing of 79 tissues;
- DNA-methylation profiling;
- plasma proteomics and metabolomics;
- single-nucleus RNA sequencing of liver and frontal cortex;
- experiments in human stem-cell-derived neurons.
The central conclusion is that metformin shifted numerous molecular, histological and functional measurements in an apparently younger direction, with particularly prominent effects in the brain and liver.
Main findings
1. Cognitive performance and brain structure
Metformin-treated monkeys performed better than untreated old monkeys in tests of:
- delayed memory;
- object discrimination learning;
- reversal learning or cognitive flexibility.
MRI showed preservation of cortical thickness, particularly in frontal regions involved in executive function, working memory and reversal learning. Histology also suggested greater frontal cortical thickness.
Nine of 88 mapped brain regions showed significant preservation or apparent restoration of cortical thickness, with most located in the frontal lobe. The paperâs Figure 1, on pages 4â5, links these structural findings to the behavioural tests.
The authors therefore argue that metformin protects both the structure and function of the aging primate brain.
2. Broad transcriptomic effects across tissues
RNA sequencing was performed across 79 tissues from 11 organ systems. Aging was associated with:
- increased inflammatory and innate-immune gene expression;
- reduced extracellular-matrix maintenance, development and regenerative pathways;
- increased apoptosis, fibrosis and oxidative-stress signatures;
- reduced lipid metabolism, Wnt signalling and DNA-repair pathways.
Metformin shifted many of these age-associated patterns towards those of younger monkeys. The strongest transcriptomic ârescue scoresâ were observed in tissues including:
- frontal cortex;
- liver;
- kidney;
- lung;
- skin;
- skeletal muscle.
This pan-tissue atlas is one of the studyâs most substantial contributions.
3. Histological markers of aging
Across several organs, metformin-treated monkeys showed lower levels of:
- p21-positive putatively senescent cells;
- fibrosis;
- inflammatory-cell infiltration;
- TNF-ι, IL-1β, S100A8 and other inflammatory or SASP-associated markers;
- lipid peroxidation marked by 4-HNE;
- loss of H3K9me3;
- expression of endogenous-retroviral proteins.
The treatment also appeared to preserve fast type II skeletal-muscle fibres.
These observations are consistent with reduced inflammation, senescence-associated signalling and tissue degeneration, although many are surrogate markers rather than direct demonstrations of rejuvenated function.
4. Biological-age clocks
The investigators constructed monkey aging clocks from:
- DNA methylation;
- tissue transcriptomes;
- plasma proteins;
- metabolites;
- single-nucleus transcriptomes.
Reported age shifts included approximately:
-
â6.4 years for plasma protein age;
-
â6.1 years for frontal-lobe DNA-methylation age;
- roughly â4 to â5 years in several liver, kidney, lung and tendon measurements;
-
â5.9 years for integrated frontal-lobe single-nucleus transcriptomic age;
-
â4.3 years for integrated liver single-nucleus transcriptomic age.
The often-repeated claim that metformin produced a âsix-year regressionâ therefore refers to model-derived biological-age estimates, not to a demonstrated six-year increase in lifespan or healthspan.
5. Liver effects
Single-nucleus RNA sequencing suggested particularly strong effects in:
- hepatocytes;
- Kupffer cells;
- T cells.
Metformin partly reversed age-associated changes involving:
- lipid and amino-acid metabolism;
- inflammatory signalling;
- TGF-β signalling;
- fibrosis.
Histology showed less lipid-droplet accumulation, fibrosis and inflammatory-marker expression.
6. Neuronal effects and Nrf2
The paper proposes that neuronal protection is mediated partly through Nrf2, a transcription factor controlling antioxidant and cellular-stress responses.
In cultured human stem-cell-derived neurons, metformin:
- increased phosphorylated nuclear Nrf2;
- raised expression of Nrf2 targets such as HO-1, NQO1, SOD3, GPX1 and GPX2;
- lowered ROS and 4-HNE;
- reduced senescence-associated β-galactosidase, protein aggregates, amyloid-β and IL-6;
- preserved lamin B2.
Nrf2 knockdown substantially weakened the protective effects of metformin. Conversely, an activating Nrf2 E82G variant produced stronger protection than metformin, and metformin added little further benefit.
This supports the conclusion that Nrf2 is required for at least part of the observed in-vitro neuronal effect. It does not prove that Nrf2 is the dominant mechanism in the whole animal.
What is genuinely novel?
1. Long-duration intervention in healthy aging primates
The strongest novelty is the use of a prolonged pharmacological intervention in healthy, aging non-human primates rather than rodents or diabetic animals.
A 40-month trial is unusually long and translationally relevant. Cynomolgus monkeys are closer to humans than standard laboratory models in brain organisation, metabolism, lifespan and age-associated pathology.
2. System-wide, multi-tissue assessment
The study is much broader than a conventional metformin experiment. Profiling 79 tissues allowed the authors to examine whether treatment effects were:
- systemic;
- organ-specific;
- cell-type-specific;
- shared across different molecular layers.
The combination of imaging, behavioural testing, pathology, bulk transcriptomics, single-cell data, methylation, proteomics and metabolomics is an important methodological advance.
3. Cell-type-specific primate aging clocks
The creation of single-nucleus transcriptomic aging estimates for liver and frontal-cortex cell types is novel. It allowed the authors to propose that particular populationsâsuch as excitatory neurons, microglia, hepatocytes and Kupffer cellsâwere more responsive than others.
4. Direct evidence of primate neuroprotection
Previous metformin geroscience work relied heavily on rodents, epidemiology or metabolic disease populations. Here, metformin was associated with:
- better cognition;
- preservation of cortical thickness;
- reduced neuronal pathology;
- improved myelin measures;
- younger neuronal transcriptomic profiles.
The convergence of behavioural, structural and molecular observations is more convincing than any one of these outcomes alone.
5. Nrf2 as a mechanistic link
Metformin is conventionally discussed in relation to:
- mitochondrial complex I inhibition;
- changes in cellular energy state;
- AMPK activation;
- mTOR suppression;
- insulin and glucose metabolism.
The identification of an Nrf2-dependent, cell-autonomous neuronal effect adds a potentially important mechanism. The knockdown and activating-mutation experiments make this component stronger than a purely correlative pathway analysis.
Critical appraisal
Strengths
Broad convergence of measurements
A major strength is that the result does not depend on one clock or biomarker. Cognitive tests, MRI, histology, transcriptomics and inflammatory measurements generally point in the same direction.
Clinically relevant dose and duration
The monkey dose was chosen to approximate human therapeutic exposure, making the work more relevant than studies using extremely high experimental concentrations.
Healthy rather than diabetic animals
Because the monkeys were not diabetic and metformin did not materially lower glucose or body weight, the findings cannot easily be attributed only to correction of hyperglycaemia or obesity.
Mechanistic intervention
The Nrf2 knockdown and gain-of-function experiments go beyond association and provide evidence of mechanistic necessity and sufficiency within the cultured-neuron model.
Major limitations
1. Very small number of animals
The central comparison involved approximately:
- six metformin-treated old monkeys;
- initially six old controls, with only five completing the study;
- six young controls;
- only three middle-aged controls.
This is a very small study for an intervention generating thousands of molecular measurements. Individual-animal differences can exert disproportionate effects, particularly for cognition, imaging and aging-clock estimates.
The large number of cells, tissue samples or molecular features must not be confused with a large number of independent experimental subjects. The monkey, not each cell or tissue sample, is the primary experimental unit.
2. Male-only study
All animals were male. Metformin responses may differ by sex because of differences in:
- sex hormones;
- body composition;
- hepatic metabolism;
- mitochondrial biology;
- immune aging;
- pharmacokinetics.
The results cannot automatically be generalized to female monkeys or women.
3. Cross-sectional reference groups
The young and middle-aged animals were distinct cohorts, rather than the treated monkeys being followed from youth through old age.
Therefore, the aging trajectories used to construct the clocks combine:
- true age-related change;
- cohort effects;
- individual differences;
- possible differences in early environment or life history.
A longitudinal untreated cohort with serial tissue sampling would be stronger, although clearly difficult in primates.
4. Limited evidence of actual healthspan or lifespan extension
The paper demonstrates changes in intermediate outcomes, but not:
- reduced mortality;
- longer lifespan;
- delayed onset of diagnosed disease;
- reduced frailty;
- sustained benefit after metformin withdrawal.
Only a limited set of functional outcomes was assessed, principally cognition. Most other conclusions rest on molecular or histological proxies.
Thus, âslowing agingâ is a plausible interpretation, but âextending healthspan or lifespanâ remains unproven.
5. Biological-age regression may be overstated
The reported four-to-six-year rejuvenation figures sound more precise than the data justify.
These values depend on:
- the selected molecular features;
- model structure;
- a small training population;
- calibration against chronological age;
- extrapolation from cross-sectional age groups;
- the assumption that movement towards a younger molecular profile is beneficial.
An aging clock can show a younger estimate because a drug directly changes its constituent biomarkers, even when the drug has not altered the underlying rate of organismal aging.
Consequently, it is safer to say that metformin shifted the clocks towards younger reference profiles, rather than literally making the animals six biological years younger.
6. Risk of model overfitting and non-independence
ElasticNet regularisation and leave-one-out validation help, but the overall dataset is still small relative to the number of possible transcriptomic, methylomic and proteomic predictors.
Some clock evaluations may also be partly circular:
- features are selected because they vary with age;
- metformin reverses some of those features;
- the same or closely related features are then used to conclude that age has been reversed.
External validation in a completely independent monkey cohort would be needed to establish that these clocks predict meaningful future outcomes.
7. Multiple-comparison burden
The study examined:
- 79 tissues;
- many cell types;
- thousands of genes;
- numerous pathways;
- multiple histological markers;
- dozens of brain regions;
- several aging clocks.
Although adjusted statistics were used for many omics analyses, such a vast analytical search space increases the risk of selective emphasis on favourable findings. Some reported regional or histological results rely on nominal (p<0.05), where more stringent correction would be appropriate.
8. Possible pseudoreplication
Single-nucleus analyses generate many metacells or cellular observations from very few monkeys. Treating these as fully independent observations would artificially narrow confidence intervals.
The paper appears to aggregate cells in parts of its analysis, but the effective sample size remains the number of animals. Stronger hierarchical models explicitly nesting cells within animal would make the inference more secure.
9. Nrf2 mechanism is only partially established in vivo
The Nrf2 experiments are strongest in cultured neurons. In the monkeys, Nrf2 activation correlates with metformin treatment, but the investigators did not block Nrf2 in vivo.
Therefore, the evidence supports:
Metformin can protect cultured neurons through an Nrf2-dependent mechanism.
It does not fully establish:
Nrf2 is responsible for the systemic or brain-wide anti-aging effects of metformin in primates.
AMPK, mitochondrial complex I, mTOR, insulin signalling, autophagy, inflammatory regulation and altered one-carbon metabolism may all contribute.
10. Culture concentration needs careful interpretation
The neuronal experiments used 5 ÎźM metformin, which is described as a low concentration and is more physiologically plausible than the millimolar concentrations often used in cell culture.
Nevertheless, extracellular culture concentration does not directly establish the concentration reached in particular human neuronal compartments during ordinary oral dosing. Cellular uptake depends on transporter expression and tissue pharmacokinetics.
11. Cognitive testing could be vulnerable to expectancy or handling effects
The article states that older monkeys were randomly assigned, but the paper would be more persuasive with clearer reporting of:
- blinding of behavioural assessors;
- blinding during MRI segmentation;
- blinding of histological quantification;
- pre-specified primary cognitive endpoints;
- handling and training equivalence.
In a small behavioural study, even subtle differences in testing or animal-handler interaction can matter.
12. Mortality imbalance cannot be interpreted
One control monkey died from kidney failure, while none of the treated animals died. The numbers are far too small to infer a survival benefit, and excluding the deceased animal from end-stage tissue analyses could introduce survivorship bias.
13. Safety assessment is underpowered
The study reports no major abnormalities in weight, glucose, blood counts or urine parameters. However, six treated animals are insufficient to detect uncommon adverse effects.
Human concerns such as:
- vitamin B12 depletion;
- gastrointestinal intolerance;
- lactic acidosis in susceptible individuals;
- effects in renal impairment;
- loss of exercise adaptation in some contexts;
cannot be resolved by this experiment.
Interpretation of the Nrf2 finding
The Nrf2 result fits a coherent model:
- neuronal aging lowers effective Nrf2 activity;
- antioxidant and detoxification genes decline;
- ROS and lipid peroxidation rise;
- protein aggregation, nuclear-envelope disruption and inflammatory signalling increase;
- metformin restores Nrf2 activity and partially reverses these changes.
However, Nrf2 activation is not necessarily synonymous with reversal of aging. Nrf2 is fundamentally a stress-response pathway. A younger molecular profile could reflect improved damage resistance rather than a resetting of the underlying aging process.
The distinction is important:
-
Geroprotection: reduces damage or preserves function while treatment continues.
-
Rejuvenation: reverses accumulated age-related damage or permanently restores a younger state.
-
Slower aging rate: changes the future slope of deterioration.
This experiment supports geroprotection most strongly. It provides suggestive, but not definitive, evidence for rejuvenation or a lower intrinsic rate of aging.
Bottom-line assessment
This is an ambitious and important proof-of-concept study. It provides some of the strongest experimental evidence so far that clinically relevant long-term metformin exposure can produce broad geroprotective effects in a healthy non-human primate.
Its most convincing findings are:
- preservation of frontal cortical structure;
- improved cognitive test performance;
- broad reduction in inflammatory and degenerative tissue signatures;
- consistent shifts towards younger molecular profiles;
- an Nrf2-dependent protective effect in cultured human neurons.
Its weakest aspect is the quantitative claim of several years of biological-age reversal. Those figures derive from newly developed clocks trained on very small cohorts and should be viewed as descriptive molecular shifts, not literal measures of rejuvenation.
A fair conclusion is:
Long-term metformin treatment appears to preserve several aspects of brain and systemic function in aging male monkeys and shifts many molecular markers towards younger states. The study supports metformin as a candidate geroprotective drug, but it does not yet prove lifespan extension, durable rejuvenation, or equivalent benefits in humans.