NeuroAge Therapeutics — Deep Analysis

Company: NeuroAge Therapeutics, Inc. (neuroagetx.com), San Francisco, CA
Prepared: August 2026
Scope: Page-by-page site audit, product-by-product science review, and evidence grading

0. Executive summary

What the company is. NeuroAge Therapeutics is a ~2022-founded, grant- and pre-seed-stage San Francisco company selling a direct-to-consumer, cash-pay, multi-modal "biological brain age" assessment. The product line spans brain MRI volumetrics, a proprietary 52-transcript blood RNA panel, whole-genome/genotype-based risk and pharmacogenomics, a resold Quest Diagnostics plasma amyloid/p-tau217 panel, and a cognitive testing-and-training layer called NeuroGames. Prices range from $9.99/month to $3,895.

Founder and CEO Dr. Christin Glorioso (MD/PhD) trained in the University of Pittsburgh / Carnegie Mellon Medical Scientist Training Program and did postdoctoral work at MIT in Leonard Guarente's lab. The company is an alumnus of Berkeley SkyDeck and has received grant funding from the Alzheimer's Drug Discovery Foundation (~$250K). Co-founder Dr. Priyanka Joshi serves as CSO.

The core finding. NeuroAge's science page is unusually rigorous for a direct-to-consumer company — roughly 40 real citations, accurately characterized, with no obvious misrepresentation. But there is a systematic gap between the strength of the literature they cite and the strength of evidence for the product they actually sell:

Overall assessment: intellectually serious, commercially premature. The founder is a genuine physician-scientist with real publications, the citations are honest, the biological thesis — that brain aging is an upstream, modifiable, measurable substrate for neurodegeneration — is scientifically live and arguably correct, and the site's self-assigned evidence tiers show good faith. But not one product has published performance characteristics for its intended use, the flagship differentiator is explicitly unvalidated in living humans, and the composite score the entire brand rests on has no published validation of any kind.

The most important structural observation: NeuroAge is openly developing therapeutics — cell-culture and animal studies on two compounds targeting "master regulators of brain aging," with the biomarker panel positioned as the companion monitoring tool. Read that way, the consumer diagnostics business is substantially a data-and-cash generation engine for a drug program, which explains the aggressive marketing of an unvalidated panel more charitably than a pure-diagnostics reading would.

1. Site audit — page by page

1.1 Site map

Three pages in the site navigation (/products, /partnering, /resources) returned 404 at time of audit — the site is under active construction.

1.2 Home (/)

Headline: "Measure your brain age. Become optimally sharp." Three trust badges recur across every page:

• "Validated on data from 500,000+ individuals"
• "Supported by dozens of peer-reviewed publications"
• "Funded by the Alzheimer's Drug Discovery Foundation"

Read on the badges. All three are technically defensible and rhetorically misleading — this is the site's central credibility-transfer mechanism.

• "Validated on 500,000+ individuals" refers to the published literature the components draw on (UK Biobank, MyCogHealth cohorts, postmortem brain banks) — not to a validation of the NeuroAge product on 500,000 people. NeuroAge has validated its composite on nobody.
• "Dozens of peer-reviewed publications" are overwhelmingly other research groups' papers about brain aging biology, not papers about NeuroAge's test.
• "Funded by the ADDF" is a legitimate but modest (~$250K) grant, presented with the weight of an institutional endorsement.

Acquisition framing is anxiety-first and worth flagging: "Detect changes before symptoms appear—you may already be losing memory, just not noticing yet." This is aimed squarely at the worried well.

HSA/FSA eligibility is flagged throughout — a deliberate price-softening device for a $1,400–$3,900 cash-pay purchase.

1.3 Our Science (/our-science) — the substantive page

Eleven sections, roughly 6,000 words, ~40 numbered citations with DOIs:

1. Premise — brain aging precedes symptoms by decades; dementia is not inevitable
2. Multi-modal beats single-modality (Wang 2022, Brain Communications)
3. Biological brain age beats chronological age (Glorioso 2019; Nguyen 2025; Liang 2022; Lancet Commission 2024; Chene 2025)
4. NeuroGames as a cognitive clock (Talboom 2021, 2019; Rentz 2023; Bilgel 2017; Bopp & Verhaeghen 2005; Edwards 2017)
5. MRI volumetrics (Qi 2020; Zhou 2022; Brown 2022; Wardlaw 2017; Vernooij 2007; Toh 2022)
6. RNA blood biomarkers (Glorioso 2019, 2010)
7. Genomics and Mendelian randomization (Andrews 2024; Liu 2024; Fan 2023)
8. Quest AD-Detect plasma biomarkers (Barthélemy 2024; Khalafi 2025; Racke 2024, 2025)
9. The ADDF clinical validation study (n=75, WCG IRB)
10. Reversibility evidence (Erickson 2011; Ten Brinke 2015; Ornish 2024)
11. How recommendations are generated — biomarker→intervention mapping with self-assigned evidence tiers

Assessment. This is the most credible science page I have encountered from a direct-to-consumer brain-health company. Citations are real, primary, and accurately characterized. The recommendation section self-grades its own evidence tiers (Strong / Moderate / Emerging) — a genuinely honest touch most competitors don't attempt, and one that argues for good faith rather than deception.

Two structural problems remain:

• Citation substitution. Nearly every citation supports a biological premise, not a product performance claim. The page builds a compelling argument that brain aging is real, measurable, and modifiable — which is true — and lets the reader infer that NeuroAge measures it accurately, which is unestablished.
• No performance characteristics anywhere. No sensitivity, specificity, test-retest reliability, mean absolute error on the brain age estimate, or confidence interval on any NeuroAge output appears on the site. For a product whose entire value proposition is a single number, this is the most consequential omission in the audit.

1.4 About Us (/about-us)

Founder: Dr. Christin Glorioso, MD/PhD — Pitt/CMU Medical Scientist Training Program, postdoc at MIT (Guarente lab), previously Head of AI at TeachAids and Chief Strategist at UCSF's Bakar Aging Research Institute. Origin story: her grandmother died of Alzheimer's during her childhood.

Team is small and junior-weighted: VP Engineering, one AI engineer, director of business development, programs/ops coordinator, clinical administrator, and three ML engineering interns. There is no VP of Clinical Affairs, no regulatory lead, no biostatistician, no medical director, and no genetic counselor listed.

Advisory board is genuinely strong:

• Dr. Matt Kaeberlein — CEO, Optispan; one of the most credible names in geroscience
• Dr. Lawrence Tanenbaum — board-certified radiologist, former CTO of RadNet Imaging; a serious radiology credential
• Dr. Ronjon Nag — Adjunct Professor in Genetics, Stanford Medicine; founder, R42
• Dr. Joshua Kuluva — neurologist and psychiatrist, President of Piedmont Neuroscience Center
• Dr. Salah Mahmoudi — CEO, Reneu Bio; formerly Pfizer and Alkahest

The gap between advisory board strength and operating team depth is the clearest single signal of company stage. This is an organization with excellent scientific taste and very little clinical or regulatory infrastructure.

1.5 ADDF Study (/addf-study)

Free brain MRI, cognitive assessment, at-home phlebotomy, and (for selected participants) an Oura ring, offered to adults 60+ with or without MCI/early Alzheimer's. n=75 — 50 cognitively healthy controls aged 25–95 and 25 participants with MCI or early AD. IRB approval through WCG. Powered to detect 5% score differences at p<0.05 based on variance in prior postmortem cohorts. Follow-up at 6-month intervals, with an extended lifestyle-intervention arm tracking 12 and 24 months.

Two stated aims: (1) test whether blood transcriptomic and methylation signatures predict biological brain age as measured by cognitive testing (digit span, reaction speed, verbal recall, MMSE, ADAS-CoG); (2) test whether the same peripheral multi-omics signatures predict brain age derived from structural MRI via BrainKey volumetrics.

This page functions simultaneously as a research protocol and a customer acquisition funnel. The banner "Age 60 or older? Receive a free brain health assessment by participating in our research study" appears site-wide. This is not improper — it is IRB-approved recruitment — but the study doubles as a zero-CAC channel into their highest-value demographic.

What this page reveals is the most important fact in the audit: the study exists to determine whether the blood panel works in living people at all. It is a single-arm pilot feasibility study, n=75, using the sponsor's own composite as the index test. NeuroAge is selling the $899 blood test today while running the study that determines whether it measures what they claim.

1.6 Legal (/legal)

A material compliance gap. The Legal page contains website terms of use, cookie policy, content licensing, hyperlinking guidelines, and privacy language — and no medical disclaimer, no FDA statement, no CLIA or laboratory-developed-test disclosure, no "not intended to diagnose, treat, cure, or prevent any disease" language, and no statement of test limitations.

For a company selling an Alzheimer's-associated blood biomarker, whole-genome sequencing returning pharmacogenomic results with direct prescribing consequences (CYP2C19, VKORC1, SLCO1B1, HLA-B*15:02), and brain MRI marketed to asymptomatic adults, the absence of consolidated medical and regulatory disclosure is a genuine exposure. A "not intended to diagnose, treat, or prevent disease" line appears buried in NeuroGames product copy, but nowhere centrally.

2. Product-by-product: what it is, the science, and how strong the science is

Grading scale:

• A — Multiple independent RCTs, regulatory-grade validation, or replicated meta-analytic evidence for this specific use
• B — Strong, replicated observational or mechanistic evidence; validated in independent cohorts, but not for this specific consumer application
• C — Plausible and supported, but limited, single-lab, indirect, or with unresolved generalization problems
• D — Preliminary, unvalidated, or the claim relies on evidence from a different product or population

2.1 NeuroGames — $9.99/month

What it is. Five validated cognitive tests (memory, processing speed, visuospatial ability, executive function, attention) completed online in ~30 minutes, plus a ~10-minute/day speed-of-processing training game. Outputs age- and sex-normed scores, a trend line across sessions, a brain age estimate that updates with continued play, and personalized recommendations.

The science cited.

• Talboom et al., 2021, npj Aging and Mechanisms of Disease (n>75,000, ages 18–85): reaction time and memory performance closely linked across sexes and the full age range; first-degree family history of AD, diabetes, stroke, and smoking each independently predicted slower reaction time and worse memory. Large, real, well-powered.
• Talboom et al., 2019, eLife: family history of AD associated with lower paired-associate learning performance beginning four decades before typical clinical onset; the effect was attenuated by higher education and absence of diabetes. Solid.
• Rentz et al., 2023, Alzheimer's & Dementia: face-name associative memory paradigm developed by Dr. Dorene Rentz at Harvard Medical School, validated in the ARMADA study; in a 257-participant cohort with amyloid biomarker data, achieved "acceptable discriminatory accuracy" separating amyloid-positive from amyloid-negative individuals. Real — but "acceptable" is a weak descriptor, likely implying AUC in the 0.6–0.7 range.
• Bilgel et al., 2017, Journal of Alzheimer's Disease (n>3,100, two independent longitudinal cohorts): trajectory modeling established that working memory span, assessed by digit span, was the first measure to show detectable decline — occurring primarily while individuals were still cognitively normal, ahead of verbal memory, language, and executive function. Strong, and genuinely important. This is the single best scientific justification for the product category.
• Bopp & Verhaeghen, 2005: meta-analysis of 123 studies; working memory span shows considerably larger age effects than simple storage, older adults retaining ~73–75% of younger adults' recall. Strong.
• Barnes et al., 2015, Alzheimer's & Dementia (n=7,815, NACC database): non-memory first symptoms far more prevalent in younger patients — 26% presenting before age 60 vs 6% after age 79. Solid; supports multi-domain rather than memory-only testing.
• Edwards et al., 2017 (ACTIVE trial): up to 29% lower rates of dementia diagnosis after structured speed-of-processing training, 10-year follow-up, n=2,802.

Evidence grade: assessment layer = B. Training layer / dementia-reduction claim = D.

Why the split. The measurement side is well-founded — these are validated paradigms drawn from real cohorts, and the Bilgel finding that digit span moves first is a legitimate rationale for early cognitive testing. The training and prevention side is where the argument breaks down:

1. The 29% figure is borrowed from a different product. ACTIVE's speed-of-processing arm used the Useful Field of View (UFOV) task, which was commercialized by a third party. NeuroGames is a different intervention with no efficacy trial of its own. Attributing a 29% dementia reduction to NeuroGames on the basis of a trial of someone else's training program is an unsupported inferential leap, and it is the kind of claim that has drawn FTC enforcement in this category.

2. The ACTIVE dementia finding is itself contested. Dementia was not a pre-specified primary outcome of ACTIVE; the trial's original headline result was that training improved the trained ability with minimal far transfer to everyday function. The dementia analysis was a long-follow-up secondary analysis using proxy-based ascertainment rather than clinical diagnosis, and the effect concentrated in participants who returned for booster sessions — a self-selected, adherent subgroup. The memory and reasoning training arms showed no significant effect at all.

3. "Brain age estimate that updates with ongoing play" conflates practice effects with biological change. If a score improves because the user got better at the game, the product is measuring task familiarity, not brain aging. Nothing on the site addresses practice-effect correction, which is a first-order methodological requirement for any repeat-administered cognitive assessment. This is arguably the most serious unaddressed technical flaw in the product line.

Bottom line: NeuroGames is the weakest product in the lineup and carries the most aggressive claim. It is also, at $9.99/month, the funnel entry point.

2.2 Brain MRI — $1,398 new scan / $999 for upload (listed as $399 elsewhere); +$1,299 full-body add-on

What it is. Non-contrast structural brain MRI through a partner imaging network across the US, processed by the BrainKey platform: volumetrics across 25+ brain regions (the Essentials page claims 100+), a brain age estimate against the age/sex normative distribution, 3D visualizations, and radiologist review for incidental findings on network scans. Existing scans can be uploaded if they include required sequences and original DICOM files.

BrainKey was founded by Dr. Owen Phillips, a Stanford-trained neuroscientist; it grew out of a Stanford postdoctoral project and received Y Combinator backing in 2019. Its models are described as trained on tens of thousands of scans from clinics across multiple countries.

The science cited.

• Hippocampal volume — the brain region most directly involved in forming new memories and among the earliest to atrophy in Alzheimer's disease. Grade A as a risk marker.
• White matter hyperintensities — Qi et al., 2020, Neuroscience and Biobehavioral Reviews: meta-analysis of 36 prospective studies, >19,000 participants. Higher baseline WMH burden associated with +14% all-cause dementia, +25% Alzheimer's disease, +73% vascular dementia. Zhou et al., 2022: WMH volume predicted a 49% increased risk of progression to dementia. Grade A.
• WMH reversibility — Brown et al., 2022, Neurology: WMH regression documented in up to one-third of participants across studies, with regression associated with improved clinical and cognitive outcomes. Wardlaw et al., 2017, Neurology: in 190 adults with serial MRI, WMH decreased in 37% within one year, with higher baseline blood pressure predicting which lesions resolved. Grade B — and genuinely underappreciated. This is a legitimately motivating finding for a consumer, because it makes the measure actionable rather than merely prognostic.
• Combined structural prediction — Gons et al., 2015: white matter and hippocampal volume together are the two strongest structural predictors of incident dementia in small-vessel-disease cohorts, outperforming diffusion tensor imaging. Grade B.
• Incidental findings — Vernooij et al., 2007, NEJM (population-based, n=2,000): 1.8% asymptomatic cerebral aneurysm, 1.6% benign primary brain tumor, nearly all without symptoms. Toh et al., 2022, Acta Neurochirurgica (meta-analysis, n=40,777): neoplastic incidental findings in ~1% of brain MRIs, rising with age. Grade A for prevalence; the interpretation is where it gets contentious.

Evidence grade: underlying markers = A. The derived "brain age" number = C. The screening-of-asymptomatic-adults use case = C/D.

Three problems.

1. Brain age models generalize poorly to individuals. The 2025 literature is blunt: brain age models show marked performance drops on external datasets, systematically over- or under-estimate depending on training-set age skew, produce estimates not centered on zero in cognitively normal subjects, and vary considerably between software packages analyzing the same scans. Recent work supports brain age gap as a group-level biomarker while explicitly stating that bias correction is required before reliable individual clinical use. Additional findings show weak, non-significant associations between baseline brain age predictions and subsequent decline in gray matter volume or memory in cognitively normal individuals — which directly undercuts the "track your brain age annually" premise. NeuroAge sells an individual number and a longitudinal trajectory.

2. BrainKey's regulatory status is not disclosed. Competing volumetric packages — NeuroQuant, VUNO Med-DeepBrain, CorticoMetrics THINQ, icometrix — hold FDA 510(k) clearances and advertise them. I found no evidence of a BrainKey 510(k), and NeuroAge makes no clearance claim anywhere on the site. A YC-backed Stanford spinout with large training data is a reasonable technical bet, but it is not equivalent to a cleared device, and the site does not draw that distinction for the buyer.

3. The incidental-findings pitch cuts both ways. NeuroAge frames incidental findings as a benefit — a testimonial claims a brain aneurysm was caught that another imaging provider missed, and the site correctly notes a 30–40% short-term mortality rate for ruptured aneurysms. That's real and it is the strongest emotional argument for the product. But the base rates cited are also the argument against population screening: at ~1.8% aneurysm prevalence in asymptomatic adults, most detected aneurysms are small, will never rupture, and generate surveillance imaging, sustained anxiety, and occasionally procedures with their own morbidity. The broader literature on DTC whole-body MRI finds 40–50% of scans produce at least one incidental finding; one 1,000-person asymptomatic series required workup in 180 people to identify 2 tumors — a positive predictive value of ~0.011. No professional society recommends brain MRI screening in asymptomatic adults. NeuroAge presents only the upside.

This is nonetheless the strongest product in the lineup, and NeuroAge says so themselves — accurately: "MRI volumetrics is currently the most predictive single component in the NeuroAge panel, consistent with its established role as the primary structural endpoint in Alzheimer's prevention trials." Non-contrast MRI is genuinely safe and repeatable, and hippocampal volume and WMH have the best predictive track record of any measure they offer.

2.3 Blood Biomarkers Test — 52-transcript RNA panel — $899

What it is. NeuroAge's proprietary and only truly differentiated asset. A 52-transcript blood RNA panel positioned as a brain-specific aging clock, developed in research cohorts with RNA sequencing from both brain tissue and blood collected from the same individuals. The transcripts map to five hallmarks of brain aging: metabolic dysfunction, neuroinflammation, synaptic decline, cellular senescence, and stress response. Explicitly positioned as not an Alzheimer's pathology test but a measure of upstream aging biology.

Site claim: "NeuroAge's RNA blood biomarker panel is proprietary, and no other company offers it."

The stated rationale for RNA over DNA is sound and well-argued: "While DNA is fixed at birth, RNA expression changes in response to lifestyle, environment, metabolic state, and interventions. This means the RNA panel can detect whether biological changes are actually occurring, not just whether a risk exists."

The science cited.

• Glorioso et al., 2019, Life Science Alliance — "Rate of brain aging and APOE ε4 are synergistic risk factors for Alzheimer's disease." Transcriptome analysis of 673 postmortem brain samples across four/five human cohorts, ages 25–97, free of neurological disease. Established a transcriptome-based gauge of molecular brain age that correlates well with DNA methylation clocks, and showed APOE ε4 accelerates molecular brain aging synergistically. The site's headline claim — being five or more years biologically younger is associated with ~6-fold lower risk of AD, Parkinson's, and age-related cognitive decline, even in ε4 carriers — traces to this work. This is a real, well-regarded paper in a legitimate journal.
• Glorioso, Oh, Douillard, Sibille, 2010 (Sibille lab, University of Pittsburgh) — PhD work characterizing the molecular aging signature across four distinct brain regions, establishing regional consistency and connection to neurological disease risk.
• The stated foundation for the blood proxy: blood transcriptomics from postmortem cohorts predicted brain biological age from 23 blood transcripts at p<10⁻⁶.
• Irmady et al., 2023, Nature Communications — precedent that postmortem Parkinson's brain transcriptome patterns are detectable in antemortem peripheral blood and correlate with clinical features.
• Supporting brain age gap literature: Nguyen et al., 2025 — each one-year increase in brain age gap raised AD risk 16.5% and all-cause mortality 12%; highest-risk quartile showed 2.8-fold increased AD risk. Liang et al., 2022, Frontiers in Neuroscience — each additional brain age year associated with 10% elevated AD risk.

Evidence grade: D (with a credible path to B).

This is the central finding of the entire audit. The most expensive recurring test, the one product no competitor offers, the single thing that makes NeuroAge a company rather than a reseller of Quest and BrainKey — has never been validated in living humans.

The evidence chain:

Postmortem brain tissue (n=673) → a molecular aging clock — ✅ well-established, published, credible
Postmortem paired blood → 23 transcripts predict brain age at p<10⁻⁶ — ✅ promising, but postmortem blood
Living people's blood → 52 transcripts predict their brain age — ❓ this is what the n=75 study is testing right now

NeuroAge states this openly on their own science page: the study's purpose is "to determine whether blood-based transcriptomic and methylation signatures can predict biological brain age as accurately in living people as they do in postmortem brain-blood paired cohorts." That is a direct admission that the answer is not yet known — and to their credit, it is stated plainly rather than obscured.

Additional technical concerns:

• Postmortem blood transcriptomes are not equivalent to living blood transcriptomes. Agonal state, hypoxia, post-mortem interval, and terminal illness all substantially reshape the transcriptome. The expansion from 23 transcripts to 52 is also unexplained anywhere on the site.
• Blood RNA is extremely labile. Whole-blood transcriptomes shift with acute infection, circadian timing, fasting state, recent exercise, stress, and medication. The COVID-era literature documents transcriptomic age shifting by up to 21 years during acute infection and reverting on recovery. NeuroAge markets this responsiveness as a feature — the panel "can detect whether biological changes are actually occurring" — but the same lability that makes it responsive to intervention makes any single measurement noisy. No pre-analytical protocol, test-retest reliability, or biological-variation data is published. For a test sold on annual or biennial repeat measurement, this is a load-bearing omission.
• "Brain-specific" is doing heavy lifting. Blood cells are not brain cells. The claim is that these 52 transcripts were selected for correlation with brain tissue aging — a defensible selection strategy, but not a demonstration of brain specificity in circulating blood.
• Bulk-tissue clocks have known resolution limits. The 2025 literature notes that clocks derived from bulk tissue lack cell-type-specific resolution and that prediction accuracy degrades substantially in elderly populations, where mean absolute error can exceed 15 years compared to 5–8 years in middle age. That degradation falls precisely on NeuroAge's target demographic.
• The pathway-level readout is explicitly a future capability. The site is honest about this: "As the NeuroAge dataset grows... the data will eventually make it possible to identify not just how fast a given brain is aging, but which specific processes are driving that acceleration." The $899 today buys an overall rate, not a pathway breakdown.

The upside case is real. If the study reads out positive with a meaningful effect size and replicates in an independent cohort, this becomes a genuinely valuable asset. The field lacks a modifiable, repeatable, upstream brain-aging biomarker, and NeuroAge's argument for why one is needed is correct and well-made: "Current blood-based monitoring biomarkers are amyloid-focused and would not be informative in clinical trials targeting other disease pathways." For a company developing non-amyloid therapeutics, owning the companion monitoring biomarker is strategically coherent. But n=75, single-arm, sponsor-run is a feasibility pilot — not validation.

2.4 Genetic Resilience Analysis — $249 (analysis) / +$550–799 (30× whole-genome sequencing)

What it is. APOE genotype plus a 41-variant Alzheimer's polygenic risk score; eight longevity/resilience variants (Klotho KL-VS, APOE ε2, FOXO3, CETP, IGF1R, IL-6, mTOR, PCSK9); six medically actionable pharmacogenomic genes (MTHFR, CYP2C19, CYP2C9, SLCO1B1, VKORC1, HLA-B*15:02); and polygenic risk for eight additional conditions (coronary artery disease, atrial fibrillation, osteoporosis, colorectal cancer, primary open-angle glaucoma, psoriasis, Crohn's disease, ovarian cancer).

Component-by-component grading:

Evidence grade: pharmacogenomics = A. APOE = A. Everything else = C/D.

The real issue is delivery, not validity. Returning APOE ε4 homozygous status — a 10–15× Alzheimer's risk — direct to a consumer, alongside HLA-B*15:02 and VKORC1 results with immediate prescribing implications, is genetic counseling territory. The site's only framing is "Inherited risk does not determine your future, but it can tell you where to pay closer attention." The Essentials Bundle includes one 30-minute virtual consultation; the standalone $249 genetic product does not appear to include any. No genetic counselor is listed on the team.

Two things NeuroAge gets right here and deserve credit for: the contextualization principle is sound and well-stated — "A lower overall NeuroAge score is associated with reduced neurological disease risk even in individuals with lower genetic resilience" — correctly conveying that genes are not destiny and current-state biomarkers carry more actionable information. And recommending confirmatory biochemical testing (homocysteine, folate, omega-3 index) rather than acting on genotype alone is genuinely good practice that much of the consumer genomics industry skips.

2.5 AD-Detect Blood Test — $799 (Quest Diagnostics)

What it is. A resold Quest Diagnostics panel: plasma Aβ42/40 ratio plus phosphorylated tau (p-tau217, and p-tau181 per the product page), combined into Quest's proprietary AD-Detect Likelihood Score.

The science cited.

• Barthélemy et al., 2024, Nature Medicine: plasma p-tau217 clinically equivalent or superior to FDA-approved CSF tests for classifying both amyloid and tau PET status. Grade A. Real and important.
• Khalafi et al., 2025, Alzheimer's & Dementia: meta-analysis of 30 studies; p-tau217 shows 82% sensitivity, 86% specificity for amyloid PET positivity across independent cohorts. Grade A.
• Racke et al., 2024, Journal of Investigative Medicine (AB255): low baseline Aβ42/40 increases risk of progression to dementia by ~70%. Grade B.
• Racke et al., 2025, Neurology Clinical Practice: real-world analysis of >4,000 patient specimens; 42% classified high likelihood for amyloid PET positivity, 15% indeterminate (falling to 10% with ApoE status). Grade B.
• Quest's combined AD-Detect Likelihood Score: 91% sensitivity / 91% specificity, validated in a well-characterized cohort from the 1Florida Alzheimer's Disease Research Center. Grade B — sponsor-validated, single cohort.

Evidence grade: the biomarkers themselves = A. This use case (asymptomatic consumers) = D.

This is simultaneously the most defensible assay and the least defensible application in the lineup. Plasma p-tau217 is arguably the most significant advance in Alzheimer's diagnostics in twenty years. The problem is entirely about who it is being sold to and how.

1. It is a laboratory-developed test. It is not FDA cleared or approved. Quest's own materials state this explicitly. NeuroAge's site does not disclose it anywhere.

2. The Alzheimer's Association's 2025 Clinical Practice Guideline on blood-based biomarkers explicitly excludes cognitively unimpaired individuals — "given the current lack of clinical relevance for blood-based biomarker use in this population" — and also excludes primary care settings. The guideline addresses specialists testing patients with objective cognitive impairment (MCI or dementia). NeuroAge markets this to asymptomatic adults purchasing online, outside specialty care. That is squarely outside guideline-recommended use.

3. Base rates destroy positive predictive value in this population. 91%/91% performance in an enriched memory-clinic cohort does not transfer to a 55-year-old with no symptoms, where amyloid positivity prevalence might be 10–15%. At those base rates a substantial fraction of positive results are false positives. The sensitivity and specificity figures quoted on the site materially overstate what an asymptomatic buyer should expect from their own result.

4. The direct-to-consumer version of this test drew unusually sharp professional criticism. When Quest launched DTC amyloid testing in 2023, Alzheimer's researchers and clinicians raised concerns about accuracy, false positives, patient comprehension, and unjustified anxiety — trade coverage described it as opening a "Pandora's box," and one report quoted neurologists calling the consumer-initiated offering "an absolute catastrophe." Quest subsequently added a clinical-involvement requirement. NeuroAge's product page states only that it "does not replace physician-led diagnosis."

NeuroAge's own framing is the best argument against buying it from them. They explicitly position the RNA panel as targeting "the upstream biological aging process that creates a permissive environment for that pathology to develop" — with AD-Detect as the downstream-pathology complement. If their thesis is correct, AD-Detect is the product least aligned with it, and it is also the one product a customer could obtain through their own physician, appropriately supervised, potentially covered by insurance.

2.6 Essentials Bundle — $1,398/year (from $1,746, 20% discount)

Includes: Brain MRI with 3D dashboard and preventative screening, Genetic Resilience analysis, NeuroGames cognitive assessment, full interpretation of scores, personalized recommendations, and one 30-minute virtual consultation.

Add-ons: genome sequencing +$799, RNA blood test +$899, full-body MRI +$1,299.

Dashboard reports: brain age estimate, brain health scores across three dimensions (structural, functional, genetic), MRI analysis across 100+ brain regions, cognitive trend tracking, and recommendations linked to specific brain areas.

Claims audit:

• "Dementia starts 20–30 years before the first symptom" — defensible and consistent with the biomarker cascade literature.
• "40% of dementia cases are potentially preventable" — the 2020 Lancet Commission figure. The 2024 update says 45%, and NeuroAge's own science page correctly uses 45% (and cites Chene 2025 for up to 65% with a fuller factor set). The bundle page is simply stale.
• "Trained on data from over 500,000 people" — the credibility-transfer issue again; this describes the source literature, not NeuroAge's validation.

The pricing structure deserves specific attention. The bundle is priced annually, which implies annual brain MRI in asymptomatic adults. Non-contrast MRI is safe to repeat — that's true and well-argued on the science page. But there is no evidence base supporting annual structural brain MRI as surveillance in healthy people, and the year-over-year change in hippocampal volume in a healthy adult is on the order of the measurement error of most volumetric pipelines. The recurring-revenue model is running ahead of the measurement science.

2.7 Coaching Program — $280 per 50-minute session (1, 4, or 12-session tiers)

One-on-one brain health coaching with Dr. Glorioso herself, organized around "the 9 Pillars of Healthy Brain Aging." Covers interpretation of MRI, genetic, blood biomarker, and cognitive results; prevention planning; and accountability.

Grade: N/A as science — this is a service. Two observations:

1. It does not scale. The CEO of the company is the delivery unit, capping this as a boutique revenue line and consuming the founder's time at the stage when it is most valuable elsewhere.
2. It is, in practice, the clinical interpretation layer the rest of the product line lacks. NeuroAge has effectively priced genetic and biomarker counseling as a premium add-on rather than building it into the products that generate results requiring interpretation. A customer buying the standalone $249 genetic panel receives APOE status with no counseling included; a customer who pays an additional $280 gets it.

2.8 NeuroAge Test Ultra — $3,895

"The full multi-modal workup" — all modalities combined. No dedicated product page was found; it is referenced from the routing quiz and homepage. Functionally this is the anchor price that makes the $1,398 Essentials Bundle read as moderate.

2.9 Summary scorecard

3. Cross-cutting assessment

3.1 What NeuroAge gets genuinely right

• The core thesis is scientifically live and probably correct. Brain aging as an upstream, modifiable, measurable substrate that creates permissive conditions for neurodegeneration is a legitimate and increasingly mainstream framing. Their argument that amyloid-focused biomarkers are uninformative for monitoring non-amyloid interventions is correct and well-made.
• Multi-modal beats single-modality. Correctly argued and correctly cited. Dementia genuinely is a multi-pathway disease.
• The reversibility evidence is the strongest and most motivating part of the site, and it is accurately presented: Erickson 2011 (PNAS, n=120 RCT, 2% hippocampal volume increase with one year of aerobic exercise vs decline in controls); Ten Brinke 2015 (BJSM, 5.6% left hippocampal increase in women with MCI); Ornish 2024 (Alzheimer's Research and Therapy, n=100 RCT, ~70% of intervention participants stabilized or improved vs ~70% of controls worsening, with a dose-response relationship across all four cognitive measures). These are real randomized trials, correctly characterized.
• Self-graded evidence tiers. Labeling their own recommendations Strong / Moderate / Emerging, and explicitly flagging Lion's Mane and EGCG as emerging, is more intellectual honesty than the category norm.
• Recommending confirmatory biochemical testing (homocysteine, folate, omega-3 index) rather than acting on genotype alone.
• Non-contrast MRI protocol. No gadolinium, no ionizing radiation — a genuinely safer choice than alternatives, and correctly explained.

3.2 The four material gaps

1. No published performance characteristics for anything. Not one sensitivity, specificity, AUC, test-retest coefficient, mean absolute error, or confidence interval for any NeuroAge-generated output appears on the site. This is the defining omission. A buyer cannot evaluate the product, and neither can a reviewer.

2. The flagship differentiator is unvalidated in living humans, and they are selling it anyway. The n=75 study running now is the validation. This is the clearest instance of commercialization preceding evidence.

3. No consolidated medical or regulatory disclosure. No FDA statement, no LDT/CLIA disclosure, no test limitations, no "not intended to diagnose" language on the Legal page. Given the content of the results being returned, this is a real exposure.

4. Claims run ahead of product-specific evidence. The 29% dementia-reduction figure attached to NeuroGames is the sharpest example, but the pattern is systemic: a large body of legitimate literature about brain aging is marshaled in a way that invites the reader to attribute that literature's strength to NeuroAge's products.

3.3 Stage-appropriate read

Much of the above is what an early, under-resourced company looks like rather than evidence of bad faith. The team has three ML interns and no regulatory lead; the missing disclaimers and undisclosed LDT status are as likely oversight as strategy. The founder's own publications are real, the citations are honest, and the study is IRB-approved and openly described as validation-in-progress.

The reasonable summary is: the science is better than the product, and the product is being sold as though the reverse were true.

3.4 What to watch

• The ADDF study readout. If blood transcriptomic signatures predict MRI- and cognition-derived brain age in living people with a meaningful effect size, the RNA panel becomes a genuinely valuable asset and the company's differentiation is real. If not, NeuroAge's offering collapses to reselling Quest Diagnostics and BrainKey with a dashboard on top.
• Whether independent replication follows. n=75, single-arm, sponsor-run is a pilot. External-cohort replication is what would move this from D to B.
• Regulatory posture. FDA's LDT enforcement direction, and FTC attention to dementia-prevention claims in the consumer wellness category, both bear directly on this business model.
• The therapeutics program. Two compounds targeting master regulators of brain aging are in cell-culture and animal studies. This is the actual long-term thesis; the diagnostics business is arguably infrastructure for it.
• Whether clinical infrastructure gets built. A medical director, regulatory lead, biostatistician, and genetic counselor would each address a specific gap identified above.

4. Caveats on this analysis

• Content was captured from the live site in August 2026 via automated page fetch and summarization. Prices for the MRI upload option were reported inconsistently across pages ($999 on the product page, $399 on the routing quiz), and one fetch of the pricing page returned a corrupted table; treat all figures as indicative rather than exact.
• /products, /partnering, and /resources returned 404 during the audit. The therapeutics pipeline and partnership model could not be assessed from dedicated pages and are reconstructed from /our-science and third-party coverage.
• Evidence grades reflect my judgment against each product's stated intended use, not a formal systematic review. I did not obtain full texts of all ~40 cited papers; citation accuracy was spot-checked on the load-bearing ones (Glorioso 2019, Edwards 2017, Barthélemy 2024, Vernooij 2007, and the 2025 Alzheimer's Association blood-biomarker guideline).
• Assessments of BrainKey's regulatory status reflect an absence of found evidence of FDA clearance, not confirmation that none exists.

Sources

NeuroAge site: Our Science · Home · NeuroGames · Brain MRI · Blood Biomarkers · Genetic Resilience · AD-Detect · Essentials Bundle · Coaching · ADDF Study · About Us · What test is right for you

Company background: Crunchbase · Longevity.Technology · Propel(x)

Key primary science: Glorioso et al. 2019, Life Science Alliance · Edwards et al. 2017 (ACTIVE) · Vernooij et al. 2007, NEJM · NCPT-independent brain-aging clock work, 2025

Clinical guidelines: Alzheimer's Association Clinical Practice Guideline on blood-based biomarkers, 2025 · AAIC 2025 guideline release

Brain age model limitations: Clinical validity comparison of brain age software, eBioMedicine 2025 · Bias and generalizability of brain age prediction models · Distribution bias in brain age research · Brain age gap as predictive biomarker, Communications Medicine 2025

AD-Detect and DTC biomarker criticism: Quest AD-Detect p-tau217 test directory · Alzforum: DTC Alzheimer's blood test opens Pandora's box · 360Dx: experts raise concerns · Being Patient

MRI screening in asymptomatic adults: Incidental findings on brain MRI, MedLink Neurology · DTC whole-body MRI screening outcomes · Brain incidental findings in healthy young adults (MRi-Share)