The recurring mistake in this category is reading a rising number as a result. NAD+ goes up on a blood test, so the protocol worked. Muscle ATP goes up on a scan, so the protocol worked. That reasoning has been tested directly in humans more than once, and it keeps failing. The number moves and the person doesn’t change.
That gap is the whole story of longevity peptides right now, and it is more interesting than the marketing.
What mitochondrial dysfunction means
Mitochondria aren’t simply batteries. The inner membrane is folded into cristae, and the folding matters as much as the chemistry, because the machinery that makes ATP sits along those folds. A phospholipid called cardiolipin lives almost exclusively in that inner membrane and is required for it to work properly.
What cardiolipin loss and peroxidation are established to do is impair electron transport chain supercomplex assembly, disrupt cytochrome c binding, and degrade carrier-protein function. Proton leak in aged tissue is a related but distinct problem, attributed mostly to adenine nucleotide translocase, uncoupling proteins and general membrane peroxidation rather than to cristae losing their shape. The popular image of folds sagging and protons spilling through is a simplification that runs ahead of what has been shown.
So mitochondrial dysfunction in aged tissue means at least three separable failures: less NAD+ to run the reactions, a structurally compromised membrane that wastes fuel it does process, and a weakened signalling output where mitochondria stop communicating properly with the rest of the cell. A compound addressing one has no obligation to address the others, and the research is now specific enough to say which is which.
MOTS-c and the exercise interaction
MOTS-c is a 16-amino-acid peptide encoded inside mitochondrial DNA. The common description has mitochondria releasing it into circulation to act on distant skeletal muscle, which overstates what has been shown. Skeletal muscle expresses MOTS-c itself, and the headline human observation is an intramuscular rise of about 11.9-fold after high-intensity cycling (n=10) rather than a demonstration that circulating peptide acts at a distance. The link to AMPK is also indirect, running through folate and methionine cycle inhibition and AICAR accumulation, and a 2024 iScience paper proposes CK2 as a different proximal target. The mechanism is not settled.
The animal work is genuinely strong. Reynolds and colleagues (Nature Communications 2021) gave 22-month-old C57BL/6 mice 15 mg/kg per day by intraperitoneal injection for two weeks. Treadmill running time roughly doubled and distance rose about 2.16-fold. Seventeen percent of treated old mice reached the highest-speed stage of the test, against none of the untreated animals. The same animals were then transitioned to intermittent dosing at 15 mg/kg three times weekly after eight weeks of daily injections, reaching 23.5 months of age, and improved on grip strength, stride length and 60-second walking capacity. That matters for interpretation: the intermittent results sit on top of two months of prior daily high-dose exposure rather than representing a clean late-life start.
Lifespan is quieter than it is usually reported. Median lifespan rose 6.4% and maximum 7.0%, and the frequently quoted p=0.05 attaches only to a lifespan curve truncated at 31.8 months. The overall log-rank test was p=0.23. The authors wrote that larger cohorts are needed. That is a trend, not a finding, and the p-value usually attached to it makes it look closer to significance than the reported overall test was.
The strongest human evidence for MOTS-c is genetic rather than interventional. Zempo, Yen and colleagues pooled 27,527 people across three East Asian cohorts and examined the m.1382A>C variant, which produces K14Q MOTS-c. A 2024 iScience paper later showed K14Q has reduced binding to CK2 and doesn’t activate it, which offers a mechanism for the loss of function.
Men carrying the variant had significantly higher type 2 diabetes prevalence in the meta-analysis across all three cohorts. Women didn’t. The gene-by-exercise interaction is a narrower result than it is usually reported to be: the male effect was confined to the lowest tertile of physical activity within J-MICC alone, because that was the single cohort with activity data.
Read carefully, that interaction says exercise appears to offset a loss-of-function variant. It does not establish what exogenous MOTS-c does in someone carrying the normal allele, and the popular reading that training substitutes for the peptide is a step beyond what the design supports.
Humanin and the one longevity test it was given
Humanin is a survival signal. Mitochondria release it under stress and it suppresses apoptosis, which is why it appears protecting neurons and cardiac cells in the disease literature.
The correlational human data looks favourable. Yen and colleagues (Aging 2020) found circulating humanin significantly higher in centenarian offspring than in age-matched controls, though at n=18 against n=19 that is a small comparison. Humanin fell roughly 40% over the first 18 months of mouse life. Naked mole-rats showed only a non-significant decline (p=0.08) from basal levels around four-fold above young mice, though the authors caveat that cross-species comparison directly: naked mole-rat humanin is only 75% identical to the human sequence, so they express confidence in the age-related relative changes rather than in absolute cross-species levels.
The same paper ran the intervention. One hundred mice, 18-month-old females, HNG at 4 mg/kg intraperitoneally twice weekly, continued for 14 months from mid-life. No significant lifespan difference against control.
Healthspan markers did move, with visceral fat down, lean mass up and IGF-1 significantly reduced, though those measurements were micro-CT at n=5 per group. In C. elegans, humanin overexpression raised mean lifespan about 7%, from 17.7 to 19.0 days, and that effect disappeared in daf-16/FOXO mutants.
Higher humanin in long-lived people is equally consistent with humanin being a readout of healthy mitochondria rather than a cause of longevity. The causal version was tested in mammals and returned null. Worth holding alongside that: the study used about 50 animals per arm with no reported power calculation, which is thin for a mouse lifespan experiment and cannot exclude an effect the size of the MOTS-c trend above. The null and the trend deserve the same evidentiary standard.
NAD+ precursors keep winning the biomarker and losing the outcome
Raising NAD+ works and is among the more reproducible things in this field. Conze’s dose-ranging work put whole-blood NAD+ up 22% at 100 mg of nicotinamide riboside, 51% at 300 mg and 142% at 1000 mg.
Dollerup and colleagues (American Journal of Clinical Nutrition 2018) then ran 40 obese sedentary men aged 40 to 70 on 1000 mg of NR twice daily for 12 weeks, with insulin sensitivity measured by hyperinsulinaemic-euglycaemic clamp. Nothing moved: not insulin sensitivity, endogenous glucose production, glucose disposal or oxidation, resting energy expenditure, lipolysis, lipid oxidation or body composition.
Elhassan and colleagues went into the tissue (Cell Reports 2019), a crossover trial in 12 aged men taking 1 g/day of NR for 21 days with muscle biopsies and targeted metabolomics. The muscle NAD+ metabolome rose. Mitochondrial bioenergetics were unchanged. Whole-body and muscle metabolism were unchanged. The RNA-seq showed NR-mediated downregulation of energy-metabolism and mitochondrial gene pathways alongside an anti-inflammatory signature. The precursor reached the target tissue, raised the target molecule, and the transcriptional response in mitochondrial pathways went down rather than up.
NMN’s flagship trial warrants the same scrutiny. Yoshino and colleagues (Science 2021) studied 25 postmenopausal women with prediabetes, 13 on 250 mg/day NMN and 12 on placebo for 10 weeks. Muscle insulin sensitivity rose about 25%, with increased AKT/mTOR phosphorylation. Every other measured endpoint was null, including body composition, intra-abdominal fat, intrahepatic triglyceride, blood pressure, glucose, insulin, free fatty acids, lipids, adiponectin and leptin.
A published Comment in Science (Brenner 2021, abj1696) noted that baseline liver fat was 6.3 ± 1.2% in the NMN arm against 14.8 ± 2.0% in placebo, an imbalance larger in effect size than the treatment effect the paper reported. Two things belong with that. The Comment’s author is the chief scientific officer of ChromaDex, the commercial sponsor of nicotinamide riboside and therefore of a competing product, and a co-author on NR work cited approvingly elsewhere in this piece. And there is a published Response (Science 2021, abj7375) noting that muscle insulin sensitivity was identical between arms at baseline. A chance baseline imbalance in a 25-person trial is expected rather than evidence of procedural fault. The imbalance is a real reason to hold the result loosely. It is not proof the randomisation failed.
Epitalon’s famous 12 percent
The repeated line is that epitalon extends lifespan around 12%. The primary source says something narrower. Anisimov, Khavinson and colleagues (Biogerontology 2003) used female SHR mice, 54 per group, dosed at 1.0 µg per mouse subcutaneously on five consecutive days each month from three months of age until natural death.
The paper states that epitalon didn’t influence mean lifespan, body weight or food intake. The 12.3% figure is maximum lifespan. The 13.3% figure is the lifespan of the last 10% of survivors. Total spontaneous tumour incidence was unchanged, although leukaemia was inhibited 6.0-fold and bone marrow chromosome aberrations dropped 17.1%.
The companion rat study by Vinogradova in 2007 found no lifespan change under a standard light and dark cycle. Benefit appeared only in animals under constant or northern natural illumination, meaning animals whose circadian rhythm had been deliberately disrupted. That reads less like a general geroprotector and more like a possible circadian-rescue agent in stressed animals.
On dose, the mouse protocol works out to roughly 30 to 40 µg/kg. A common consumer figure of 10 mg/day in a 70 kg adult is about 143 µg/kg, so roughly four times the mouse dose per kilogram rather than the thousandfold gap sometimes claimed. The very large ratio that circulates compares absolute dose per animal, which is the comparison per-kilogram scaling exists to correct. The animal data still doesn’t establish a human dose, but the gap is a factor of a few rather than orders of magnitude.
A 2025 in-vitro study from Brunel University London, the first substantive non-Russian primary work on the telomere claim, found dose-dependent telomere lengthening in normal cells via hTERT and telomerase upregulation. In the breast cancer lines 21NT and BT474, telomere lengthening also occurred, but through ALT activation, an effect the authors described as specific to cancer cells. That paper carries a published correction to three of its figures, dated November 2025.
SS-31 has the best mechanism and still splits biomarker from function
SS-31, also called elamipretide, binds cardiolipin directly and acts as a structural stabiliser of the inner mitochondrial membrane. Of the compounds here it has the cleanest mechanism and the most human data.
Roshanravan and colleagues (PLOS One 2021) randomised 39 healthy adults aged 60 to 85 from 120 screened, pre-selected for genuinely poor mitochondrial function (ATPmax below 0.7 mM/sec and P/O below 1.9). A single two-hour IV infusion at 0.25 mg/kg/hr raised in-vivo skeletal muscle ATPmax 27% from baseline against 12% on placebo. Change in ATPmax was the pre-specified primary endpoint and it was met, at p=0.045 for percent change, with absolute change at p=0.055. Complete ATPmax data was available for 36 participants, 18 per group, after technical data loss, which the authors flag as a power limitation. Resting P/O coupling didn’t change. The fatigue-resistance measure, force-time integral in the first dorsal interosseous, didn’t improve against placebo, and it was a secondary endpoint the authors say the study was underpowered to test. Only a post-hoc contraction-count analysis reached significance.
The larger test was MMPOWER-3 (Karaa et al., Neurology 2023;101(3):e238-e252, PMID 37268435), 218 adults with primary mitochondrial myopathy on 40 mg/day subcutaneous for 24 weeks. It missed both co-primary endpoints. The six-minute walk difference against placebo was −3.2 m (95% CI −18.7 to 12.3, p=0.69). The fatigue score is reported inconsistently within the paper itself, appearing in the abstract at p=0.37 and in the Results section, for the identical −0.07 difference, at p=0.81. A subgroup with nuclear-DNA variants did improve on walk distance while the mtDNA subgroup showed nothing, and that analysis was exploratory and labelled post hoc in the paper’s own figure legend and discussion, not pre-specified.
Elamipretide received FDA accelerated approval on 19 September 2025 for Barth syndrome. The route is worth knowing: the TAZPOWER crossover trial had 12 patients and its randomised phase showed no knee-extensor strength gain. Approval rested on the open-label extension, a median +63 newtons at week 168 in 8 patients.
What the published protocols used
The parameters below record what researchers administered. They aren’t a plan for anyone.
NR was given orally across human trials spanning 100 to 2000 mg/day, including 1000 mg/day for 21 days in aged men (Elhassan), 1000 mg/day for 6 weeks (Martens 2018, n=24) and 2000 mg/day for 12 weeks (Dollerup). Safety has been demonstrated to roughly 20 weeks.
NMN was given orally at 250 to 1250 mg/day across published randomised trials, with the pivotal insulin-sensitivity trial at 250 mg/day for 10 weeks. Human absorption is contested, since oral NMN may be degraded to nicotinamide before intact uptake.
MOTS-c has no established human dose and no published human results. Rodent work used 5 or 15 mg/kg/day intraperitoneally for two weeks, or 15 mg/kg three times weekly in the extended protocol. Human dosing in this family so far belongs to CB4211, a different molecule and a MOTS-c analogue, run at 25 mg once daily for four weeks in 20 people.
Humanin and HNG have no established human dose. The rodent longevity protocol was 4 mg/kg intraperitoneally twice weekly for 14 months.
Epitalon rodent protocols used 1.0 µg per mouse subcutaneously on five consecutive days per month, and 0.1 µg/day five times weekly in rats. The Russian human literature used epithalamin, a bovine pineal extract given in intramuscular courses, rather than synthetic AEDG, and the course dosing in those reports isn’t recoverable from an accessible primary source.
SS-31 human protocols were 40 mg/day subcutaneous for 24 weeks in MMPOWER-3, and a single two-hour infusion at 0.25 mg/kg/hr in the ageing study. Rodent work used 3 mg/kg/day by subcutaneous osmotic minipump for 8 weeks.
The Roshanravan design contains the one detail worth borrowing conceptually: participants were screened for measurably impaired mitochondrial function before entry, making it the only study here that defined its population by the thing the drug targets.
Where this evidence is genuinely thin
MOTS-c has no published human results and no established dose. It is no longer true that it has no human interventional trial: NCT07505745, a phase 2a randomised double-blind placebo-controlled study run by Hudson Biotech, began on 2 February 2026 and is recruiting 120 adults with prediabetes and overweight or obesity for daily subcutaneous MOTS-c over 12 weeks. Until it reports, the human literature remains observational or genetic. CB4211 was reported through a company press release and a conference late-breaker, and that program was subsequently wound down.
Humanin has no human interventional trial, plus a null result in the one mammalian longevity experiment it was given. Its expanding rodent literature sits in disease models, mostly small and mostly single-lab.
Epitalon is functionally a single-institution literature, roughly 25 years of work from the St. Petersburg Institute of Bioregulation and Gerontology with essentially no independent Western in-vivo replication. No double-blind randomised trial anywhere measures human telomere length before and after epitalon. The frequently cited human mortality numbers come from epithalamin, in small non-blinded studies with 39 treated against 40 control.
For NAD+ precursors, no trial has measured a hard ageing outcome. No mortality, no incident disease, no disability, no maintained function over years. Brain, liver and heart NAD+ in humans are essentially unmeasured. The dedicated NR blood pressure trial designed to test the 2018 subgroup signal (NCT03821623, 94 participants) has a protocol published in 2022, and no published results are locatable. That is unverified rather than negative.
For SS-31, the longevity evidence is entirely aged rodent. There is no chronic-dosing study in healthy older adults and no long-term safety data outside disease populations.
Head-to-head and combination work is nearly absent, with one exception worth knowing. Whitson and colleagues (Aging Cell 2020) compared SS-31 and NMN directly in aged mouse hearts and also tested them together, reporting that the combination was synergistic and best recapitulated the young state. That is a single rodent study, and it is more than the rest of this category has.
What to read next
Pull Elhassan 2019 (Cell Reports 28:1717-1728) and read the RNA-seq section rather than the abstract, since the direction of the transcriptional change is the part nobody quotes. Read Dollerup 2018 alongside it. Read Anisimov 2003 in Biogerontology directly and locate the sentence about mean lifespan. Read the Science Comment on Yoshino 2021 next to the paper and next to the published Response, and note who wrote each. For SS-31, read the MMPOWER-3 primary publication in Neurology before any coverage of the Barth approval.
The habit worth building is checking whether a cited result was a pre-specified endpoint or a post-hoc analysis, and whether a lifespan claim refers to mean or maximum. Those two questions reclassify most of what circulates about this category.
This article is a review of published research. It is not medical advice, and nothing here is a recommendation to use any compound. Several compounds discussed have no published human data and are sold for laboratory research use only.
