Weekly Issue — 2025-07-27

Now there's a serious piece of evidence to put on the table. A new study published in The BMJ used UK primary care records linked to hospital admissions and the national death registry to compare people newly started on GLP-1 receptor agonists against people newly started on two other classes of diabetes drugs — DPP-4 inhibitors and SGLT-2 inhibitors. The design is the kind regulators actually trust: an active comparator, new user cohort study, which is research-speak for “we compared real patients on Drug A to real patients on Drug B who looked a lot like them, starting from day one.”

The headline finding, in plain English: no increased risk of suicidal ideation, self-harm, or suicide among GLP-1 users compared with patients on the other two drug classes. That's the answer to the question people have been asking. It's also — and this is the part worth lingering on — not the same thing as “GLP-1s are proven safe for your mood, forever, full stop.” We'll get to why.

Most of the scary GLP-1/mental-health stories you've seen were built on a thinner kind of evidence: spontaneous adverse-event reports, where a patient or doctor flags something that happened while on a drug. Those reports are valuable as smoke detectors — they're how the original concern was raised — but they can't tell you whether the drug caused the event, or whether people prone to depression are simply more likely to be on a weight-loss medication in the first place.

An active-comparator new-user design tries to fix that. Instead of comparing GLP-1 users to the general population (apples to a fruit salad), the BMJ team compared them to people starting a different diabetes drug at the same point in their disease — patients with similar weight, similar blood sugar, similar prescribing histories. Then they used propensity-score weighting to balance the remaining differences, and followed both groups for suicidal ideation, self-harm, and suicide as recorded in primary care notes, hospital admissions, and death records.

It's not a randomized trial. It can't be. But for a question this specific, in a class of drugs this widely prescribed, it's about as close as observational research gets.

Across two large cohorts of patients with type 2 diabetes followed through March 2021, the authors reported no statistically significant increase in the composite outcome of suicidality — that's suicidal ideation, self-harm, and suicide considered together — among people on GLP-1 receptor agonists versus people on DPP-4 inhibitors or SGLT-2 inhibitors. The individual outcomes, looked at separately, didn't light up either.

For a class of drugs that's been prescribed to millions, that's the kind of finding that quietly moves the needle for regulators. It aligns with what the European Medicines Agency and the FDA had previously signaled after their own reviews: no causal link established. The BMJ paper is the kind of well-powered real-world analysis those agencies lean on when they're deciding whether to keep a warning on a label or take one off.

Here's where I have to be the friend who reads the methods section so you don't have to. The BMJ cohort studied patients with type 2 diabetes. It did not study otherwise-healthy women in their forties using semaglutide or tirzepatide for weight management, perimenopausal metabolic shifts, or PCOS. Those are different populations, on potentially different doses, with different baseline mental-health profiles. The biological mechanism — GLP-1 acting on receptors in the brain — is presumably the same. The risk profile may not be identical.

The study also runs through March 2021, which predates the current tidal wave of off-label and weight-loss prescribing. And like every observational study, it can only measure what shows up in the records. Quieter symptoms — the low mood that doesn't make it into a chart note, the intrusive thought a patient never mentions — are harder to capture than a hospital admission for self-harm.

None of this undoes the finding. It just sets the lane. The signal regulators were chasing didn't show up in the data most likely to find it. That is genuinely reassuring. It is not a permission slip to ignore your own mood on a new medication.

If you're already on a GLP-1 and the headlines from 2023 made you uneasy, this is the kind of follow-up evidence worth knowing about. It doesn't erase the earlier concern; it contextualizes it. Drug-safety science is supposed to work like this — a signal goes up, the right studies get done, and the picture sharpens.

If you're weighing whether to start one, the mental-health question is now a smaller part of the conversation than it was. The bigger conversation — what GLP-1s do to muscle mass, what happens when you stop, who actually benefits long-term, what they cost — is the one to have with a clinician who knows your history. That's a conversation, not a verdict from a magazine. Including this one.

One large, well-built UK study isn't the end of this inquiry — more cohorts, in more populations, on longer follow-up, are coming. But it's a meaningful piece of the answer to a question that has hovered over the most prescribed drug class of the decade. For now, the most honest sentence is also the most useful one: the data we have don't show that GLP-1s raise suicide risk, and the study designed to find that signal didn't find it.

That's not a miracle-cure headline. It's better. It's the boring, careful kind of finding that lets the rest of the conversation — about benefits, trade-offs, and who these drugs are actually right for — finally happen on solid ground.

If you train hard and read the actual papers, you already know GLP-1 receptor agonists (semaglutide, liraglutide, tirzepatide and friends) do more than blunt hunger. They slow gastric emptying, sharpen insulin response, and bind receptors that live well outside the gut — including in the central nervous system. That receptor distribution is why the new evidence keeps showing up in unexpected specialties. The question isn't whether GLP-1s are 'just' weight-loss drugs. It's how broad the metabolic-neurologic footprint actually is — and where the data is solid versus where it's still a signal.

Let's be honest up front: the evidence here is moderate, not airtight. Most of what follows comes from retrospective cohorts, propensity-matched analyses, and systematic reviews of mixed-quality studies. That's enough to take seriously. It is not enough to declare anything proven.

The headline finding everyone has been chasing: do these peptides protect the aging brain? A large 2025 retrospective using the TriNetX database — over 307,000 SGLT2 inhibitor users matched against 348,000 GLP-1 receptor agonist users with type-2 diabetes — gave the cleanest head-to-head comparison yet. After propensity score matching produced 221,883 balanced pairs, SGLT2 inhibitors edged GLP-1 RAs on overall dementia incidence (2.7% vs. 3.6%; HR 0.92), with similar signals for Alzheimer's and vascular dementia.

Read that carefully. It's not that GLP-1s failed — both classes were associated with low absolute dementia rates in a population that historically carries elevated risk. The story is that SGLT2s came out modestly ahead in a direct comparison. If you're a lifter optimizing for the long game, the takeaway isn't 'switch drugs.' It's that the entire metabolic-drug landscape may matter more for cognition than we thought five years ago.

The mechanistic case got reinforced by a separate 2025 systematic review that pulled together studies on GLP-1 and neurogenesis. Exenatide, liraglutide, semaglutide and others increased neurogenesis in the dentate gyrus, hippocampus, olfactory bulb and medial striatum — the same regions implicated in mood and memory disorders. The authors also noted modulation of apoptotic pathways and upregulation of cell-survival signaling.

The caveat the review itself flags: most of that data is animal. Translational implications are plausible — particularly for conditions defined by neurogenesis defects — but plausible isn't proven. File this under 'mechanism that supports the epidemiology,' not 'clinical claim.'

This is the result that should make every neurologist look up. Idiopathic intracranial hypertension (IIH) — elevated pressure inside the skull, mostly in women with obesity — historically gets managed with weight loss, acetazolamide, and sometimes surgery. A 2025 propensity-matched analysis of 193 tirzepatide-exposed IIH patients vs. 193 standard-care controls reported a 68% reduction in papilledema risk, a 73.9% reduction in visual disturbance and blindness risk, and a 19.7% reduction in headache risk at 24 months, alongside a BMI drop of -1.147 kg/m².

Effect sizes that large get a healthy dose of skepticism. It's a retrospective cohort, not a randomized trial, and some of the benefit probably tracks the weight loss itself rather than a direct GLP-1/GIP effect on cerebrospinal fluid dynamics. But the magnitude is hard to ignore. For a condition with limited drug options, this is a real signal worth a proper RCT.

Two more 2025-era findings round out the picture. The first is a Cochrane systematic review evaluating GLP-1 receptor agonists in people with chronic kidney disease and diabetes — a population where roughly 40% of diabetics eventually face kidney failure and elevated cardiovascular risk. Cochrane reviews are the high-water mark for evidence synthesis; the fact that one exists for this question means the field has reached enough trial volume to ask serious questions about benefits and harms. That alone is a maturity marker for the drug class.

The second is for anyone who has ever had a knee scoped or is dragging a beat-up joint into their forties. A retrospective study of 34,696 type-2 diabetes patients undergoing total knee arthroplasty compared propensity-matched GLP-1 users to controls. Surgical site infections, prosthetic joint infections, all-cause revision, aseptic revision at one year — no significant differences. Controls actually had higher odds of extended hospital stays (OR 1.29). That doesn't make GLP-1s a performance enhancer for surgery, but it pushes back hard against the assumption that they create perioperative trouble.

What's actually happening here is that we built a class of drugs to fix glucose, watched them crush appetite, and are now discovering — paper by paper — that the receptor they target is everywhere. The brain. The kidney vasculature. CSF-producing tissues. The metabolic-inflammatory circuits that mediate surgical recovery. Of course a peptide that pulls that many levers shows up in unexpected places.

The discipline this calls for is the same discipline that separates lifters who actually progress from lifters who chase hype: read the methods, respect effect sizes, and don't confuse a signal with a sure thing. The 2025 GLP-1 literature is exciting because the signals are converging from multiple independent angles. It's not yet exciting enough to rewrite clinical practice. Watch the next 24 months — the RCTs that follow these cohort studies are where the verdicts get written.

Here's the setup. Researchers ran whole-exome sequencing — basically a deep read of the protein-coding parts of DNA — on 1,245 German long-lived individuals and compared them to 4,105 younger people from the same region. The big reveal: rare variants that showed up more often in the long-lived group weren't scattered randomly. A bunch of them clustered inside one specific cellular pathway called mTOR signaling.

If you've heard the word mTOR before, it's probably because of a drug called rapamycin. Rapamycin tamps mTOR down, and in lab animals — worms, flies, mice — doing that tends to stretch lifespan. The catch has always been that the human evidence is thin. Animal studies are exciting, but humans aren't giant mice. So finding human genetic fingerprints on the same pathway? That's a quiet, careful, kind of big deal.

Think of mTOR as the cell's growth manager. When food and signals are abundant, mTOR says: build stuff, grow, multiply. When resources are scarce, mTOR chills, and the cell switches into cleanup-and-recycle mode (a process called autophagy — basically taking out the trash). A lot of aging research suggests that cells which spend a little more time in cleanup mode tend to age more gracefully. Rapamycin nudges that dial. So do, apparently, some rare gene variants.

The German team flagged three rare single variants in mTOR-pathway genes — RPS6, FLCN, and SIK3 — that were enriched in the long-lived group. They also found that a gene called RWDD1 showed a meaningful burden of rare missense variants among the long-lived, making it a strong new candidate gene for longevity.

Longevity studies usually turn up common variants with small effects, or one-off curiosities that don't replicate. This one is different in a quietly useful way: the hits cluster around a pathway we already have a drug for. That's the part that made my eyebrows lift. The same mTOR system rapamycin acts on in mice is now showing human genetic signal in people who lived a really long time.

The authors are careful, and so should we be. The study identified novel exome-wide significant associations at both the single-variant and gene level, with a notable over-representation of mTOR genes. It also flagged other associations — PRAC2, SLC16A6, FOCAD, IHH, MESD, HOXA4, and DNAJB13 — that are interesting but need more work. None of this means scientists have cracked the code. It means they've found a clue worth chasing.

Here's where the smart-friend version of me has to pump the brakes. The same paper also found an enrichment of protein-truncating variants in two genes — ASXL1 and TET2 — among the long-lived. But the authors think this is likely a side effect of something called clonal hematopoiesis: as people age, certain blood-cell mutations naturally accumulate. Translation: not every signal in a longevity study is actually about longevity. Some of it is just what older blood looks like. That's a good reminder that genetics is messy and context matters.

It's also worth saying: this was one population (German), one study design (rare variants in coding regions), and a moderate-strength set of findings that will need replication in other groups. Living past 95 is influenced by genes, sure — but also food, environment, social ties, luck. Genetics is one slice of a much bigger pie.

If you're a beginner to longevity science, here's the takeaway worth holding onto: mTOR is one of the most-studied levers in aging research, and the evidence base just got a little more human. That doesn't mean anyone should run out and start experimenting with rapamycin — it's a prescription drug with real side effects, used mainly to prevent organ-transplant rejection, and its use for healthy aging is still being studied. If you're curious about it, that's a conversation for a clinician, not a magazine article.

What this study does suggest is that the broader strategy aging researchers have been excited about — dialing mTOR activity to encourage cellular cleanup — has more grounding in human biology than we could prove before. It's a moderate piece of evidence pointing in a hopeful direction. The kind of finding that doesn't change your life today, but might shape the medicines and recommendations of tomorrow.

Bottom line, friend: a careful German study just added a real human data point to one of the most promising stories in aging science. It's not a miracle cure, and it won't change what's in your medicine cabinet tomorrow. But it's the kind of slow, steady clue that moves a whole field forward — and if you're new to longevity research, it's a great moment to start paying attention.