Mitochondrial Medicine

Mitochondrial medicine is two fields sharing a word. One is the clinical genetics of primary mitochondrial disease: several hundred known nuclear and mitochondrial genes, a minimum adult prevalence of roughly 1 in 4,300, and children who die of Leigh syndrome. The other is the much larger claim that mitochondrial dysfunction underlies ageing, fatigue, neurodegeneration and metabolic illness, and that supplements or peptides can correct it. The first has excellent diagnostics and almost no drugs. The second has a large market.

Established The therapeutic record of this field is mostly negative, and saying so is the main service this brief performs. The flagship mitochondrial drug failed its phase 3 trial in mitochondrial myopathy and its randomised trial in Barth syndrome before being approved for Barth syndrome on the strength of an open-label extension in twelve patients. The best-known gene therapy for a mitochondrial optic neuropathy improved the untreated eye as much as the treated one. The one intervention that clearly works is not a treatment at all: it is a reproductive technique that prevents transmission.

Established That technique has now produced children. The United Kingdom’s regulator confirmed in 2025 that eight babies had been born following mitochondrial donation treatment at the licensed Newcastle centre, from a cohort of 22 women, with the clinical report published the same year. It is the only place in this brief where a heritable intervention in humans has been performed under statutory licence, followed up, and reported.

This brief covers inherited mtDNA disease, mitochondrial replacement, heteroplasmy editing, delivery, and the organ-specific disease claims. Germline editing of the nuclear genome belongs to Genetic Engineering; the ageing-as-damage argument belongs to Cellular Rejuvenation; enhancement framings belong to Biological Enhancement. What is owned here is the gap between a diagnostic capability that is genuinely world-class and a therapeutic capability that, for most patients, is still supportive care.

Established A note on sourcing. This brief was commissioned in September 2026 from the Institute’s research base. Reading-list entries without links are cited from the bibliographic record rather than re-fetched, and claims are dated no later than early 2026 unless carried by a linked source.

Established The genetics are unusual and they constrain everything downstream. Human mitochondrial DNA is a 16,569-base-pair circle carrying 37 genes, of which 13 encode subunits of the respiratory chain; the other 1,100 or so mitochondrial proteins are encoded in the nucleus and imported. Each cell carries hundreds to thousands of mtDNA copies, so a mutation is usually present as a fraction — heteroplasmy — rather than as a yes or no. Biochemical failure typically appears only above a threshold in the region of 60 to 90 per cent mutant load, which varies by mutation and tissue.

Established Inheritance is maternal, which is why prevention is reproductive. mtDNA is transmitted through the oocyte. A woman carrying a pathogenic mtDNA variant can transmit anything from none to all of it, because of a germline bottleneck that resamples a small number of genomes. This makes prediction poor for individual pregnancies and makes preimplantation genetic testing effective for some couples and useless for women at or near homoplasmy.

Established Prevalence is small and the diagnostic yield is now good. The standard population estimate from a large north-east England study is about 9.6 per 100,000 adults for mtDNA disease and roughly 12.5 per 100,000 for mtDNA and nuclear causes combined, or about 1 in 4,300. Genomic sequencing now yields a molecular diagnosis in something over half of suspected cases. The field can name the mutation in most patients it sees and can offer almost none of them a disease-modifying drug.

Established Mitochondrial replacement: the technique. Two methods are in clinical use: pronuclear transfer, moving the pronuclei from a fertilised affected egg into an enucleated donor zygote, and maternal spindle transfer, moving the spindle from an unfertilised affected oocyte into an enucleated donor oocyte. Both leave a small carryover of the patient’s mitochondria in the reconstructed embryo. The resulting child has nuclear DNA from two parents and mtDNA predominantly from a donor.

Established Mitochondrial replacement: the outcome data. The Newcastle programme reported in 2025 on 22 women treated by pronuclear transfer, with eight babies born and a further pregnancy ongoing. In several infants the maternal pathogenic variant was undetectable; in others it was detectable at levels reported in the region of 5 to 20 per cent — below disease thresholds, but not zero. Reported neonatal problems were judged unrelated or transient by the investigators. That is the entire licensed clinical record, and it is eight children.

Frontier Reversion is the technical risk the trial numbers are watched for. Human oocyte work published in 2016 showed that small carryover fractions can drift upward in derived stem-cell lines, in some cases to majority mutant load, apparently through replicative advantage of particular haplotypes. Whether the same drift happens in a growing child over decades is unknown, because the oldest children born after mitochondrial donation for disease prevention are still young. This is the specific reason long-term follow-up is not a formality.

Established Unlicensed and semi-licensed use came first. A live birth after maternal spindle transfer for Leigh syndrome was reported in 2017 following a procedure performed in Mexico by a New York clinic; clinics in Ukraine and Greece applied the techniques to infertility rather than mtDNA disease, including a reported pilot series in Greece. In the United States a legislative rider has since 2015 prevented the regulator from accepting applications for heritable modification, which effectively displaced this work to jurisdictions with less oversight rather than stopping it.

Established The drug record: the flagship failed twice and was approved once. Elamipretide, a mitochondria-targeted tetrapeptide, failed the MMPOWER-3 phase 3 trial in primary mitochondrial myopathy — roughly 218 patients, 24 weeks, no significant benefit on the six-minute walk distance or the fatigue score. In Barth syndrome the TAZPOWER crossover trial in twelve patients missed its primary endpoint at twelve weeks; improvements appeared in the open-label extension, an advisory committee split, and accelerated approval for Barth syndrome followed in 2025. A drug approved for an ultra-rare disease on an open-label extension after two failed randomised comparisons is a defensible regulatory judgement and a weak evidence base, and both descriptions are true at once.

Established The mitochondrial optic neuropathy programmes are the clearest cautionary case. Idebenone in Leber hereditary optic neuropathy missed its primary endpoint in the RHODOS trial of about 85 patients while showing secondary signals, and was approved in Europe on that basis. The gene therapy lenadogene nolparvovec, injected into one eye in the RESCUE and REVERSE phase 3 trials, produced visual improvement in the treated eye and a comparable improvement in the sham-injected fellow eye — an unexplained contralateral effect that destroyed the within-patient control and left the regulatory case unresolvable on those trials. A treatment can be real and still be untestable by the design used to test it.

Established The adjacent successes are nuclear, not mitochondrial. Omaveloxolone in Friedreich ataxia — a nuclear-gene disease of iron-sulfur cluster assembly with mitochondrial consequences — met its endpoint in the MOXIe part 2 trial, about 103 patients, with a difference of roughly 2.4 points on the modified Friedreich Ataxia Rating Scale (95 per cent CI about 0.5 to 4.3) and was approved in 2023. Vatiquinone in the same disease missed its primary endpoint. Deoxynucleoside therapy for thymidine kinase 2 deficiency has produced striking survival differences in retrospective cohorts without a randomised comparison. The pattern: where the defect is nuclear and the biochemistry is a substrate problem, drugs sometimes work.

Established Heteroplasmy editing works in cells and animals and has never been used in a patient. Because no mechanism imports guide RNA into mitochondria, CRISPR is unavailable there. What exists instead is protein-only editing: mitochondrially targeted TALE nucleases, which cut mutant genomes and let the cell degrade them, shifting heteroplasmy; the DddA-derived cytosine base editors reported in 2020, which make C-to-T changes in mtDNA without cutting; and TALE-linked deaminases reported in 2022 that extend this to A-to-G. Adeno-associated viral delivery of these editors has shifted heteroplasmy in mouse heart and muscle. No clinical trial of mtDNA editing has opened.

Frontier The off-target record for mitochondrial base editors is a live problem. A 2022 report found that a mitochondrial cytosine base editor produced substantial off-target mutations in the nuclear genome, and subsequent work has traced off-target editing in mtDNA itself. Nuclear editing’s own lessons — that double-strand-break repair generates large deletions and complex rearrangements at the target site — do not transfer directly, since these tools are mostly nuclease-free, but they set the standard of scrutiny this field has not yet met.

Speculative Mitochondrial transplantation is a clinical practice built on case series. Autologous mitochondria isolated from skeletal muscle and injected into ischaemic myocardium have been reported to improve ventricular function in small paediatric series without controls. Uptake mechanisms are not established, the delivered organelles are few relative to resident mitochondria, and no randomised trial has reported. It is nevertheless offered commercially in several countries.

Frontier The supplement layer is where the money is and the evidence is thinnest. Mitochondria-targeted antioxidants, NAD precursors and urolithin A have human trials measuring biomarkers — mitophagy gene expression, skeletal-muscle endurance, plasma metabolites — and a few report modest functional changes in middle-aged or older adults. None has shown a clinical outcome in mitochondrial disease. The marketing routinely uses the word mitochondrial in the sense of the first field while citing evidence from the second.

Frontier Can heteroplasmy be shifted safely in a living human tissue? Everything downstream depends on this. The animal results are real but modest, the delivered dose of editor protein is hard to control, and the tissues that matter most — brain, retina, skeletal muscle — are exactly the ones adeno-associated vectors reach least evenly. A first-in-human heteroplasmy-shifting trial would be the field’s largest step and nobody has announced one.

Frontier Does carryover drift in people? The in-vitro drift result is the strongest argument for caution about mitochondrial replacement, and the clinical cohort is eight children. Serial measurement in accessible tissues over decades is the only way to answer it, and the answer determines whether replacement is a cure or a deferral.

Frontier Is mtDNA mutation a cause of normal ageing or a passenger? Mice engineered with a proofreading-deficient mitochondrial polymerase accumulate mtDNA mutations and show premature ageing phenotypes, which is the strongest evidence for causality. Human somatic mtDNA mutation loads in most tissues are far lower than those mice carry. Cellular Rejuvenation treats the wider damage-versus-information dispute; the mitochondrial version of it is unresolved in the same way.

Speculative Allotopic expression — moving mitochondrial genes into the nucleus — remains attractive and unproven. The natural precedent is strong, since most mitochondrial proteins are already nuclear-encoded and imported. The engineering precedent is thin: hydrophobic respiratory-chain subunits import badly, and the clearest demonstrations of importing large engineered protein sets into a mitochondrial matrix come from plant work on nitrogenase rather than from human therapeutics.

Frontier Do cells exchange mitochondria in ways that could be exploited? Intercellular transfer of mitochondria through tunnelling nanotubes and vesicles has been observed in culture and in some in-vivo models, and is the mechanistic hope behind transplantation. Whether it happens at therapeutically relevant rates in human tissue is not established.

Established The inner membrane is the bottleneck. There is no established route for delivering nucleic acids into the mitochondrial matrix in vivo, which removes the entire RNA-guided toolkit and forces every editing approach through protein targeting. This single fact explains why mitochondrial genome medicine lags nuclear genome medicine by roughly a decade.

Frontier Measurement is harder than it looks. Heteroplasmy differs between blood, urine, muscle and brain in the same patient, so a blood measurement can be reassuring and wrong. Trials therefore need tissue sampling or imaging surrogates that do not yet exist, and follow-up of mitochondrial-donation children rests on what can be sampled non-invasively from a child.

Established Endpoints defeat trials in this disease more often than biology does. Primary mitochondrial myopathy trials have used six-minute walk distance and fatigue questionnaires in heterogeneous populations of a few hundred patients; the flagship phase 3 failure was a failure on those instruments. Until an endpoint exists that detects change in a disease with this much clinical variability, negative trials will be ambiguous between an inert drug and an insensitive ruler.

Established The patient numbers cannot support conventional development economics. Individual mtDNA syndromes have patient populations in the hundreds. That drives orphan pricing, accelerated approval on surrogate or open-label evidence, and a permanent temptation to treat regulatory approval as if it were evidence of efficacy.

Established It depends on protein-engineering platforms owned elsewhere. TALE scaffolds, deaminase domains and viral vectors all come from the nuclear gene-editing programme described in Genetic Engineering, including its safety literature on off-target effects and on editing in human embryos.

Established It depends on assisted reproduction as an industry. Mitochondrial replacement is an in-vitro fertilisation procedure with extra micromanipulation steps. It requires egg donors, embryology skill and licensed clinics, and it inherits the access, cost and consent problems of fertility medicine wholesale.

Frontier It depends on a natural-history dataset that barely exists. Disease-modifying trials need to know how quickly untreated patients decline, by genotype. Registries exist and are small, which is why trial sizes are guessed and effect sizes are surprises.

Speculative The ageing claims depend on biomarkers this field does not own. If mitochondrial interventions are to be assessed for age-related outcomes, they need the endpoint infrastructure discussed in Longevity Therapies, which does not exist for ageing either.

Frontier The decisive result is the long-term follow-up of the children born after mitochondrial donation: serial heteroplasmy measurement in accessible tissues, with growth, metabolic and neurodevelopmental outcomes, over decades. It is decisive because it tests the one claim the whole technique rests on — that carryover stays below threshold instead of drifting upward as it did in cultured cells — and because it cannot be simulated, substituted or accelerated. This experiment is already running, in a cohort of eight children and their families, under the licensing authority that permitted the treatment. Its weakness is its size: eight children cannot detect a one-in-twenty outcome.

Frontier Second: a first-in-human heteroplasmy-shifting trial in a single accessible tissue. The obvious candidate is a mitochondrial myopathy with a defined mutation and a muscle compartment that can be dosed and biopsied, with the primary endpoint being a measured shift in mutant load rather than a clinical score. It would separate the delivery question from the endpoint question that has defeated every drug trial in this disease.

Frontier Third: a properly controlled trial of mitochondrial transplantation. A sham-controlled or randomised comparison in a defined ischaemia-reperfusion setting, with pre-specified functional endpoints, would settle whether a practice already being sold does anything. It is cheap relative to what it would resolve.

Frontier Fourth: an endpoint-validation study for primary mitochondrial myopathy. Follow a genotyped natural-history cohort with six-minute walk distance, quantitative strength, wearable activity and imaging, and establish which instrument detects real change. Every future negative trial in this disease is uninterpretable until one does.

Speculative Fifth: large-animal in vivo base editing with tissue-by-tissue off-target sequencing. Before any human mtDNA editing trial, somebody has to show in an animal closer to human size that the editor reaches the target tissue, shifts heteroplasmy durably, and does not rewrite the nuclear genome.

Established Mitochondrial targeting is a solved-enough engineering problem at the protein level. Targeting sequences reliably carry engineered proteins into the matrix, and the editing tools now in use are built on that. The unsolved half is control: how much editor reaches which tissue, for how long, at what copy number.

Established Vector manufacture is the industrial constraint. Adeno-associated vector production for systemic dosing runs to very high particle counts per patient, with batch costs that make ultra-rare indications marginal. This is the same manufacturing wall the wider gene-therapy sector hit, and it is the reason several approved rare-disease therapies carry the highest list prices in medicine.

Frontier Micromanipulation for replacement is skill-limited, not equipment-limited. Pronuclear and spindle transfer need experienced embryologists, and outcomes in published series vary with operator and laboratory. Scaling the technique means training people, not buying machines, which caps throughput at a few centres per country.

Speculative Organelle-scale quality control is the missing instrument. Nothing in routine use measures mitochondrial function per cell in a living human tissue. Imaging and metabolic surrogates exist and are indirect, which is why trials fall back on walking distance.

Established Nuclear gene editing is the direct neighbour and the source of every tool here. Genetic Engineering covers approved editing therapies, the embryo-editing evidence on unintended rearrangements, and the international response to heritable modification — all of which frame how mitochondrial donation is governed.

Frontier Nucleic-acid medicine is adjacent and pointedly excluded from the matrix. The lipid-nanoparticle and antisense platforms covered by RNA Medicines have transformed nuclear-encoded disease and cannot currently address the mitochondrial genome at all, which is the cleanest illustration of the delivery bottleneck.

Frontier Geroscience is adjacent and should be kept at arm’s length. Longevity Therapies and Cellular Rejuvenation cover the interventions that claim mitochondrial mechanisms. Their standard of evidence — replicated lifespan or function data, blinded where possible — is the standard the mitochondrial supplement market has not met.

Established Reproductive medicine is adjacent in a way that shapes policy. Because replacement is an IVF procedure, it inherits the regulatory architecture of fertility treatment, including licensing, donor registries and the scrutiny that comes with unlicensed clinics selling unproven cell treatments.

Established The United Kingdom built the only working licensing route and it took a decade. Parliament amended the law in 2015 to permit mitochondrial donation; the regulator licensed the Newcastle centre in 2017 and approves patients case by case; the first outcomes were published in 2025. That is a ten-year cycle from statute to clinical report for a single technique in a single country.

Established Australia legislated a second route. A 2022 Act permits mitochondrial donation under a staged licensing scheme beginning with research and a clinical trial. The design copies the British approach: statute first, licensed pilot second, general availability conditional on results.

Established The United States foreclosed the question by appropriations rider. Since 2015 the regulator has been barred from acknowledging applications for clinical research in which a human embryo is intentionally modified to include a heritable genetic modification. The rider does not distinguish nuclear editing from mitochondrial donation, and the practical effect has been to export the procedure rather than to prevent it.

Frontier The follow-up register is the institutional gap. There is no international registry with a mandate to track children born after mitochondrial donation across jurisdictions, including those born from unlicensed procedures. Without one, the decisive experiment in this brief is being run as a set of national case series, and the unlicensed arm of it is being run with no data collection at all.

Established Approval standards for ultra-rare disease are being set here and will be cited elsewhere. An accelerated approval granted after two failed randomised trials, on open-label extension data in a dozen patients, is a precedent that other ultra-rare programmes will invoke. Whether that is compassionate flexibility or evidentiary erosion is a live argument, and it is being settled case by case rather than by rule.

Established Mitochondrial donation is germline modification, and the British debate conceded this rather than denying it. A daughter born after the procedure will transmit the donor’s mtDNA to her own children. The argument for permitting it was never that it is not heritable; it was that mtDNA does not encode the traits that make germline modification contentious, and that the alternative for these families is a child with a fatal disease.

Frontier The three-parent framing is journalistically durable and analytically poor. The donor contributes 37 genes out of roughly 20,000 and no nuclear inheritance. Families and regulators have largely rejected the parenthood framing; the same framing keeps reappearing because it is a better headline than heteroplasmy.

Established Egg donors bear the physical burden and receive the least attention. Every replacement cycle requires donor oocytes, obtained through hormonal stimulation and retrieval with real if small risks. Debates about the resulting children rarely account for the donors, and donor compensation frameworks differ sharply between the jurisdictions doing this work.

Frontier Reproductive tourism is the predictable consequence of divergent law. When one country licenses a technique, another bans it, and a third has no rule, patients travel. The procedures performed for infertility rather than disease prevention in less-regulated settings are the clearest example: the same technique, a different indication, weaker evidence, and no follow-up obligation.

Frontier Consent for a lifetime of surveillance is genuinely difficult. The children in the follow-up cohort did not consent to being research subjects, and the value of the data rises with the length of follow-up. This is an ethical structure shared with paediatric gene therapy and it has no clean answer, only a governance one: independent oversight, the child’s right to withdraw at majority, and honest reporting of who holds the data.

Frontier The disability critique applies here in an unusually concrete form. Preventing transmission of a fatal childhood mitochondrial disease is not the same act as selecting against a condition compatible with a full life, and mitochondrial disease spans both. The families in the licensed programme were selected for severe risk; the boundary is not obviously stable if the technique becomes routine or is marketed for infertility.

Frontier This field is where heritable human modification stopped being hypothetical. Whatever one concludes about the scale of the change, the precedent is established: a state authorised a heritable alteration, a clinic performed it, children exist, and follow-up is published. Every later argument about germline intervention now has a real case to reason from rather than a thought experiment.

Speculative If heteroplasmy editing ever works in vivo, the reproductive route becomes optional. Treating an affected child directly would remove the strongest ethical objection to the reproductive approach and shift the field from prevention to therapy. Nothing in the current evidence indicates when or whether that happens.

Speculative The larger civilizational claim runs through ageing, and it is the weakest link. If mitochondrial decline were a principal driver of age-related frailty, an intervention here would matter far beyond a few thousand families. The mutator-mouse evidence is suggestive, the human somatic mutation loads are much lower, and no mitochondrial intervention has produced a clinical outcome in an age-related disease.

Handwave The step where the argument works by assertion is the jump from organelle to organism. Showing that a tissue’s mitochondria function better after an intervention, and concluding that the person will live longer or better, skips every endpoint the rest of medicine insists on.

These horizons track what would have to be reported, not what could be announced; the constraint in this field is evidence accumulating in very small cohorts.

• 10 yr: Frontier The first mitochondrial-donation cohorts reach school age with published heteroplasmy trajectories; a second and third jurisdiction report licensed births; a first-in-human heteroplasmy-editing trial opens in a single tissue if the large-animal off-target data support it. Speculative An endpoint for mitochondrial myopathy validated well enough to make a negative trial mean something.
• 25 yr: Speculative Donation cohorts reach reproductive age, making the second-generation transmission question answerable for the first time; mtDNA editing either enters routine use for defined mutations or is abandoned for delivery reasons, and both outcomes are currently plausible.
• 50 yr: Speculative Primary mitochondrial disease is managed as a genotype-directed speciality with a handful of disease-modifying options, or it remains supportive care with better prevention. Which of those happens depends almost entirely on the delivery problem.
• 100 / 250+ yr: Handwave Claims that engineered mitochondrial genomes become a routine feature of human biology, or that mitochondrial rejuvenation extends healthy human lifespan, are assertions about an organelle biology nobody can currently manipulate in a living person.

• Depends on Genetic Engineering supplies the protein-editing scaffolds, the viral vectors and — importantly — the negative safety literature, including evidence that editing in human embryos produces unintended rearrangements and that nuclease-based repair can delete far more than intended. Cellular Rejuvenation supplies the ageing framework within which mitochondrial damage is either a cause or a consequence, a dispute this brief inherits and cannot settle. Biological Enhancement supplies the boundary argument that decides whether replacement for infertility is therapy or enhancement.
• Requires (not on this map) A delivery mechanism that puts nucleic acids into the mitochondrial matrix, without which the RNA-guided toolkit stays locked out and every editor must be a protein; a register that follows donation-born children across borders, including those born from unlicensed procedures, since the decisive experiment is otherwise a set of disconnected case series; a clinical endpoint sensitive enough that a failed myopathy trial means the drug failed rather than the ruler; vector manufacturing at a cost that an indication of a few hundred patients can bear; a pricing model for therapies whose entire market is three figures; and a licensing route in more than the two countries that have built one, because the alternative is not prohibition but displacement.
• Enables A working heteroplasmy-shifting therapy would enable treatment rather than prevention of maternally inherited disease, remove the need for donor oocytes in most cases, and give the ageing debate its first direct test of whether lowering somatic mtDNA mutation load changes anything in a person. A validated myopathy endpoint would enable the dozen stalled drug programmes in this space to produce interpretable results. And a functioning cross-border register would enable every future heritable-intervention debate to argue from data.
• Adjacent RNA Medicines is adjacent as the platform that cannot reach this compartment; Precision Medicine is adjacent as the diagnostic engine that finds these patients and then has little to offer them; Longevity Therapies is adjacent as the field whose evidentiary standards the mitochondrial supplement market should be held to.

Established “Three-parent babies.” The mitochondrial donor contributes 37 of roughly 20,000 genes and nothing nuclear. The phrase survives because it is vivid, and it distorts the actual questions, which are about heteroplasmy drift, donor burden and follow-up.

Established “Mitochondrial donation is gene editing.” No sequence is altered. Whole nuclear genomes are moved between eggs; the mtDNA that ends up in the child is a donor’s unedited genome plus a residue of the mother’s. It is heritable modification without editing, which is precisely why it fell into a legal category that had to be created for it.

Established “Mitochondrial donation cures mitochondrial disease.” It prevents transmission of mtDNA mutations in some families and does nothing for anyone already born, and it is irrelevant to the substantial fraction of mitochondrial disease caused by nuclear genes. For many carriers, preimplantation genetic testing already achieves the same goal more simply.

Frontier “Elamipretide is proof that mitochondrial drugs work.” It failed a phase 3 trial of roughly 218 patients in mitochondrial myopathy and missed its primary endpoint in a twelve-patient crossover in Barth syndrome before receiving accelerated approval for that disease. The approval is a judgement about an ultra-rare population with no alternatives, not a demonstration that the mechanism generalises.

Established “Gene therapy restored sight in LHON.” In the pivotal trials, eyes injected with the vector and eyes injected with sham both improved to a comparable degree. Either the vector crossed to the untreated eye, or improvement was not caused by the vector. The trials cannot distinguish these, which is why the programme stalled rather than being approved.

Frontier “Fatigue is mitochondrial dysfunction.” Fatigue is a symptom with many causes, and a small number of patients with it do have primary mitochondrial disease. Clinics that test unvalidated mitochondrial function panels and sell supplements on the result are using a real disease category to label an unexplained symptom.

Speculative “NAD precursors and mitochondrial antioxidants reverse mitochondrial ageing.” The human trials measure blood metabolites, gene-expression signatures and short-duration performance tests, mostly in small cohorts. Some of those move. No trial has shown a clinical outcome, and the field’s own replication facility standards, described in Longevity Therapies, have not been applied to most of these compounds.

Speculative “You can transplant mitochondria into people.” Isolated mitochondria have been injected into human myocardium in uncontrolled series with reported functional improvement. Uptake and persistence are not established, controls are absent, and the clinics selling intravenous mitochondrial infusions for fatigue and ageing are several steps beyond even that evidence.

Handwave “Fix the mitochondria and you fix ageing.” This requires that mitochondrial decline is upstream rather than downstream of ageing, that an intervention can reach every tissue, and that organelle-level improvement translates into organism-level outcome. Each step is contested, and the argument usually asserts all three.