Longevity

The honest state of longevity science

Longevity research has two adjacent worlds. One is a small, slow, legitimate field where pre-registered human trials are starting to read out and a handful of interventions have plausible — but unproven — effects on healthspan. The other is the longevity-podcast-industrial-complex, where mouse data is laundered into human protocols, biomarker movement is marketed as life extension, and supplement stacks are sold with a confidence that the underlying trials do not support.

The two worlds use the same vocabulary. That's the problem. When somebody says "this raises NAD+," the question worth asking is: where? Blood? Muscle? Brain? Measured how? Versus what control? In a randomized trial or a before-and-after with twelve volunteers? The marketing and the science sound identical until you press on the noun.

My working frame: in humans, we have one intervention with unambiguous large effects on all-cause mortality (exercise, specifically cardiorespiratory fitness), a handful of pharmacologic candidates with suggestive but incomplete evidence (rapamycin, metformin, GLP-1 receptor agonists for cardio-renal endpoints), and a much larger universe of compounds where the mouse story is clean and the human story is either missing or messy. The longevity-podcast circuit treats all three tiers as equally actionable. They are not.

Rapamycin — the human data, not the mouse data

Rapamycin (sirolimus) inhibits mTOR (mechanistic Target of Rapamycin), the central nutrient-sensing kinase that gates protein synthesis, autophagy, and cellular growth. In the NIA Interventions Testing Program, rapamycin is the single most reliable life-extending compound across mouse strains and sexes — replicated across multiple labs, which almost nothing else in the geroscience portfolio has done.

The mouse data is real. It is also mouse data. Translating it requires humility the field has not consistently shown.

The most useful human signal so far comes from intermittent low-dose rapamycin analogs (rapalogs) given before influenza vaccination in older adults — a series of trials led by Joan Mannick and collaborators, where the rapalog improved vaccine response and reduced infection incidence in the year after dosing. That is a human readout on an immune-aging endpoint, not a mortality endpoint, but it is a real signal.

The PEARL trial (Participatory Evaluation of Aging with Rapamycin for Longevity, 2023) was the first randomized, placebo-controlled trial of intermittent rapamycin in non-clinical older adults. It found acceptable safety at the doses tested and some signal on lean mass and pain in subgroups; it did not deliver a clean win on its primary composite. Interpret accordingly: promising on safety, ambiguous on efficacy at the dose and duration tested. Matt Kaeberlein, who has been the most rigorous public voice on rapamycin geroscience, reads the trial in roughly those terms.

The immune trade-off

Daily high-dose rapamycin (as used in transplant) causes immunosuppression, mucositis, and metabolic side effects. The longevity hypothesis rests on intermittent, lower-dose schedules — typically 5–8 mg weekly in community use — under the theory that pulsatile mTORC1 inhibition reproduces the geroscience effect without persistent mTORC2 suppression. That theory is plausible. It is not yet validated by a powered human outcome trial.

Rapamycin is the strongest geroscience candidate in mice and the most honest test of the field's intellectual discipline in humans. Both things can be true.

NAD+ precursors — NR, NMN, and what actually moves tissue NAD+

NAD+ (Nicotinamide Adenine Dinucleotide) is a coenzyme central to mitochondrial function and to the sirtuin enzyme family. Tissue NAD+ levels decline with age. From there, the supplement industry leaps to a conclusion the data does not support: that raising blood NAD+ via oral precursors reverses age-related decline in humans.

Two precursors dominate the market: nicotinamide riboside (NR) and nicotinamide mononucleotide (NMN). Both reliably raise whole-blood NAD+ in humans. Whether they raise NAD+ in the tissues that matter — muscle, brain, liver — is much less clear. Most controlled human trials of NR or NMN show modest or null effects on the downstream outcomes that would actually matter (insulin sensitivity, VO2max, grip strength, cognition), with the strongest signals appearing in older or metabolically compromised subjects.

David Sinclair's NMN papers established the mechanistic case and deserve credit for opening the field. They have also been the basis for marketing claims well beyond what those papers actually demonstrated in humans. Both can be true: the science is interesting, and the consumer narrative around it has run far ahead of the data.

NMN's regulatory status is also messier than the marketing suggests. The U.S. FDA has taken the position that NMN is excluded from the dietary supplement definition because it was first authorized for investigation as a drug — which has affected supply, labeling, and retail availability in ways most consumers are unaware of.

Senolytics — dasatinib + quercetin, fisetin

Senolytics are compounds that selectively kill senescent cells — non-dividing cells that accumulate with age and secrete pro-inflammatory signals (the senescence-associated secretory phenotype, or SASP). The rationale is strong: clear senescent cells, reduce chronic inflammation, delay age-related decline.

The most-studied human protocol is intermittent dasatinib (a tyrosine kinase inhibitor used in chronic myeloid leukemia) combined with quercetin (a flavonoid). A small open-label trial in patients with idiopathic pulmonary fibrosis showed improvements in physical function markers; a separate small trial in diabetic kidney disease reported reductions in senescent-cell burden in adipose tissue. These are pilot studies. They are not powered for clinical outcomes, and the dosing — usually three consecutive days, then a pause — is more a starting guess than a validated regimen.

Fisetin — another flavonoid — is in active senolytic trials at the Mayo Clinic in older adults and in frailty cohorts. Mouse data is encouraging. Human readouts are still emerging and have so far been more equivocal than the press releases would suggest.

Honest expectations: the mechanism is plausible, the dosing is unsettled (intermittent pulses are the dominant pattern, but how intermittent and at what dose is genuinely unknown), and the human evidence base is small enough that anyone who tells you they have an evidence-based senolytic protocol is overstating the literature.

Biological age tests — what's signal, what's noise

DNA methylation clocks estimate biological age from patterns of methylation at specific CpG sites. The first generation — Horvath's pan-tissue clock (2013) and Hannum's blood clock — were trained on chronological age and predicted it well. They were not, however, trained on health outcomes, and their utility for measuring intervention effects is limited.

Second-generation clocks — PhenoAge (Levine 2018) and GrimAge (Lu 2019) — were trained on clinical biomarkers and mortality respectively. They predict morbidity and mortality more strongly than the first-generation clocks. They are also more sensitive to acute biological state — illness, recent training load, stress — which is both their strength and their problem for individual use.

DunedinPACE (Belsky et al., 2022) is a different instrument: it estimates the rate of biological aging rather than a static age. It is currently the clock I find most defensible for tracking interventions, because it is designed to detect change rather than just to predict the present.

Consumer offerings like TruDiagnostic package these clocks for direct-to-consumer use. The lab analytics are real. The interpretation offered alongside them is often more confident than the underlying science. A single point estimate of "biological age" should not anchor major decisions; trends across multiple measurements, with attention to confounders, are the only useful read.

mTOR, autophagy, and the fasting story

Time-restricted feeding (TRF) and intermittent fasting (IF) compress eating into a daily window or impose periodic longer fasts. The molecular case rests on two pathways: down-regulation of mTOR and up-regulation of autophagy (the cellular recycling system that clears damaged proteins and organelles). De Cabo and Mattson's review (NEJM 2019) is the cleanest summary of the metabolic-switching hypothesis.

What the human trials actually show, when fasting is compared to equivalent caloric intake rather than ad-lib eating: modest improvements in glycemic control and lipids, weight loss largely attributable to the caloric deficit rather than the timing, and inconsistent effects on inflammatory markers. The TREAT randomized trial (Lowe et al. 2020) found 16:8 TRF performed no better than three meals a day on weight in non-obese adults — and showed a concerning lean-mass loss in the TRF arm.

Protein cycling is the more interesting question. Periodic days of lower protein intake may pulse autophagy without persistent sarcopenia, but the human evidence is mechanistic and small. The fasting-mimicking diet work from Valter Longo's lab is the most-cited human protocol; the trials are real but underpowered for the long-term outcomes that would justify the marketing.

Fasting is a useful tool for caloric control and metabolic flexibility. It is not, on current human evidence, a longevity intervention in its own right.

Metformin and the TAME trial

Metformin is a 70-year-old first-line diabetes drug that works primarily by inhibiting hepatic glucose output via complex I of the mitochondrial electron transport chain. The longevity case rests on observational data (diabetics on metformin appear to have lower all-cause mortality than non-diabetic controls in some cohorts), on mechanistic overlap with caloric restriction signaling, and on a plausible AMPK / mTOR story.

The TAME trial (Targeting Aging with Metformin, Nir Barzilai et al.) is the registered effort to test this directly: a large randomized trial in older adults with metformin versus placebo, using a composite morbidity endpoint rather than mortality alone. TAME has been ongoing for years; funding and recruitment have been the bottlenecks. A successful readout would be a landmark for geroscience-as-clinical-discipline regardless of the direction of effect.

The complication: a 2019 study (Konopka et al.) reported that metformin blunted the mitochondrial adaptations to aerobic exercise training in older adults. If you exercise seriously, that's a real cost to weigh against the hypothesized longevity benefit. The interaction has not been fully resolved, and it makes me cautious about recommending metformin for healthy, exercising adults outside of a trial context.

The boring foundations that win

Here is the part the longevity-podcast circuit underweights: the single largest mortality-prediction signal we have in healthy adults is cardiorespiratory fitness. Mandsager et al. (JAMA Network Open 2018) followed 122,000 patients through exercise treadmill testing and found that the hazard ratio for all-cause mortality between elite and low-fitness groups was roughly five-fold — larger than smoking, larger than diabetes, larger than coronary artery disease in that cohort. No drug or supplement in the longevity literature comes close to that effect size in humans.

The boring stack, in rough order of effect size:

• VO2max and zone 2. Build a base with 3–4 hours weekly of zone-2 aerobic work (conversational pace) and 1–2 weekly high-intensity intervals to push the ceiling. This is the single most evidence-backed longevity intervention you can do.
• Resistance training. Twice weekly minimum. Skeletal muscle is among the largest endocrine organs in the body and the only longevity-relevant tissue that responds directly to mechanical load.
• Sleep. Seven to nine hours, consistent timing. The cardiovascular and metabolic costs of chronic short sleep rival the costs of most pharmacologic risk factors.
• Protein. 1.2–1.6 g/kg reference body weight, higher in older adults. Sarcopenia is one of the few aging phenotypes with a clear and modifiable nutritional lever.
• Social connection. The mortality effect of loneliness in long-running cohorts is comparable to that of smoking 15 cigarettes a day. Hard to operationalize as a protocol; impossible to ignore as a finding.

I am not telling you the supplement aisle is empty. I am telling you that the marginal hour of zone 2 will out-perform the marginal capsule of NMN in any honest comparison the literature currently supports. Build the base first. The pharmacology is a rounding error on a body that doesn't move.