12–18 October 2026

Take  Review

Steroid users' hearts were still underperforming after they stopped

Anabolic steroids get discussed in gyms in terms of what they do while you are on them. What happens after you stop is the part users hear least about.

A pooled analysis of cohort studies comparing non-medical users with non-users found excess deaths from all causes, and weaker heart pumping in people who had already quit. Nobody was randomised, so training habits, other drugs taken alongside, and who starts using at all are tangled into the numbers.

That tangle could inflate a mortality figure. It could equally produce the weaker heart pumping seen in former users: nobody measured their hearts before they started. ⁶

HARMSReported: the affected readings are global longitudinal strain and ejection fraction, two measures of how well the heart squeezes, and the review formally assessed its own risk of bias.

WHAT WOULD MOVE THISProspective follow-up after people stop, with imaging and events, showing heart function returning to normal.

6 Medical Sciences · systematic review and meta-analysis of comparative cohort studies

Move  Review

Training frequency mattered less than weekly volume for muscle growth, across 16 pooled trials

◆ Biomarkers are not outcomesMuscle size stands in for the strength and function outcomes a reader likely actually cares about.

A systematic review pooled 16 trials, 380 participants total, that held weekly training volume constant while varying how many times a week each muscle was trained, from once to three times. Once volume was equated, splitting it across more sessions had little independent effect on how much muscle grew, with GRADE-rated certainty behind the comparison.

Participants were not specifically older adults, so this should not be read as guidance for preserving muscle against age-related loss specifically -- it says something about hypertrophy training in general, not about sarcopenia prevention. The outcome measured was muscle size, not strength or physical function. ⁹

HARMSResistance-training harms are not reported in this meta-analysis; it compares outcomes between training schedules, not training against no training.

WHAT WOULD MOVE THISThe same comparison repeated in older adults specifically, measuring strength or function rather than muscle size.

9 systematic review and meta-analysis of volume-equated resistance-training trials, muscle size (surrogate) outcome · n = 380

Eat  Review

A belly-fat extract missed its target, and the collagen case still isn't made

◆ Funding is not validationThe collagen review names industry funding as one reason the positive results in that literature are hard to trust.

If you have been eyeing collagen powder for skin and joints, or a herbal capsule for belly fat, two reports this week are worth a minute.

A review of the randomised trials and meta-analyses on oral collagen found the evidence insufficient for its skin, body-composition and joint claims. It named industry funding among the reasons positive results there are hard to read. A trial of tamarind seed husk extract in 112 people missed its main target, visceral fat at twelve weeks, after earlier timepoints had looked significant.

When a product leads with a secondary result, the thing it set out to test usually did not come in. Visceral fat is a stand-in anyway: even a hit would not have told you about illness or lifespan. ^4,5

HARMSNot measured: neither the collagen review nor the tamarind trial carries adverse-event data.

WHAT WOULD MOVE THISA pre-registered trial hitting the primary endpoint it declared, published whether or not it worked.

4 International Journal of Disease Reversal and Prevention · narrative review of RCTs, systematic reviews, and meta-analyses  ·  5 Journal of Food Bioactives · randomised, double-blind, placebo-controlled trial (UMIN000057223) · n = 112

Measure  Not a study

A US service now sets GLP-1 plans by a blood aging score, retested every three months

◆ Biomarkers are not outcomesThe aging score stands in for health itself, and no study here shows that moving it changes how long or how well anyone lives.

If a clinic wants to run your treatment off a biological age score, settle one thing first: what would the number change? One US telehealth service now sets GLP-1 plans against a blood-based aging score, and retests every three months.

These scores are not empty. A biological age estimate read off CT scans was tested against mortality records and did carry information about who died. But this week's causal test found that scores built from chemical marks on DNA and from blood proteins did not raise the risk of fibrotic lung disease, while short telomeres did.

Predicting death across a group is a different claim from being a dial worth turning in one person. ^1–3

HARMSNot measured: no study here reports outcomes, good or bad, for anyone whose care was steered by an aging score.

WHAT WOULD MOVE THISA trial randomising people to care guided by an aging score against usual care, reporting illness, independence or death.

1 Longevity.Technology · Not a study  ·  2 two-sample Mendelian randomisation with pleiotropy dissection and replication  ·  3 Scientific Reports · retrospective validation of an imaging-derived biological age estimator against observational mortality outcomes

Measure◆◆◆◇◇  People, watched

A new blood-cell ratio was linked to immune aging and health trajectories, in 2,609 people

◆ Biomarkers are not outcomesThe ratio being associated with health trajectories does not mean shifting it would change them.

A single-cell analysis of blood from 2,609 people, ages 18 to 98, across eight international cohorts, measured the balance between two subtypes of a common immune cell. One subtype (marked GZMK+) rose with age in people who stayed healthier; the other (GZMB+) rose more in people with worse health trajectories. The ratio between them tracked better with health outcomes than age alone.

This record is Eric Topol's newsletter describing the peer-reviewed paper, not the paper itself. Nobody was assigned to anything here. The cohorts were watched, and the cell ratio was measured alongside whatever health outcomes people already had or went on to have. ⁷

HARMSA blood draw carries its own ordinary risk. No harms specific to this analysis are reported, because nothing was administered.

WHAT WOULD MOVE THISA clinical test built on this ratio, validated against a real outcome like illness or survival in a separate cohort, not just correlated with health status in the same one.

7 observational cohort, single-cell RNA-seq across 8 international cohorts · n = 2,609

Measure◆◆◆◇◇  People, watched

Wildfire and traffic pollution components were linked to millions of new dementia diagnoses in Medicare data

Researchers linked claims records from nearly 40 million Medicare beneficiaries to detailed estimates of the PM2.5 pollution components they breathed, from wildfire smoke and other combustion sources, over 14 years. The analysis found 3.5 million new Alzheimer's diagnoses and 8.4 million new dementia diagnoses in that population. It then asked which specific pollution components carried the strongest association.

This is observational: nobody's exposure was assigned or controlled, so the finding is an association across a huge population, not proof that any one component causes dementia. A companion analysis found a similar pattern for mortality. The transferable part is which chemical components matter, not wildfire specifically. India's own exposure comes mostly from different combustion sources, including crop burning, so the component-level finding may travel even where the wildfire source does not. ⁸

HARMSThe study measures an existing, involuntary exposure's association with disease; it is not evaluating a product or therapy with its own side effects.

WHAT WOULD MOVE THISA natural experiment, such as a region-level pollution-reduction policy, tested directly against dementia incidence.

8 medRxiv · observational cohort, national Medicare claims linked to environmental exposure data · n = 39,678,630