Alzheimer brain organoids have a consent problem before they have a manufacturing one

What the work claims

This is a review, not a primary result, and it should be weighted as one: a survey of the literature from 2018 to April 2026 on human induced pluripotent stem cell (iPSC) derived brain organoids used to model Alzheimer's disease (AD).¹ Its technical thesis is unremarkable and well supported. Organoids grown from patient somatic cells recapitulate core AD pathology, including extracellular amyloid-beta (Abeta) plaques, intracellular tau tangles, neuroinflammation and blood-brain barrier dysfunction, and they preserve the donor's genetic background well enough to reproduce genotype-to-phenotype links for the major AD genes: APP, PSEN1, PSEN2 and the APOE epsilon-4 risk allele. The review presents these as a platform for patient-specific "avatar" models and high-throughput drug screening. It is candid about the limits: no functional vascularization, batch-to-batch variability, and insufficient standardization, which together keep the models fetal-like and prevent them from capturing chronic, age-dependent disease.

What makes the review worth a governance reading is not that technical core but the two problems it treats as ethics appendices. Read as questions about platform access and oversight, they are not appendices at all. They are the constraints that decide who can industrialize AD organoids, and whether the industry that results rests on solid ground.

How it works: two constraints hiding in the ethics section

The first constraint is consent, and it is specific to this disease in a way that is easy to miss. An AD organoid is only disease-relevant if it carries a disease genotype, which means the valuable donors are people with familial or sporadic AD. Those are, by definition, people whose cognition is impaired. The review cites a cross-sectional study using the MacArthur Competence Assessment Tool for Treatment (MacCAT-T), a validated instrument that scores the four legally recognized components of decisional capacity: understanding, appreciation, reasoning, and expression of a choice. In 102 patients with mild-to-moderate AD, all four were impaired, and impaired non-linearly rather than in lockstep, so a donor can look competent on one axis and fail another.² The review reports this as an obstacle to obtaining valid consent from exactly the population the platform depends on.

The second constraint is moral status, and here the review assembles a set of claims that need careful handling. It reports that current expert agreements hold that brain organoids lack the complexity to experience pain or consciousness. It then reports two observations offered as reasons for unease: that organoids can produce electroencephalogram (EEG) activity said to resemble that of pre-term infants, and that they can reach myelination levels at ten weeks of culture comparable to a twenty-week fetus, which is framed as raising a question of "developmental distress."³ The review also notes the terminological hazard of calling these constructs "mini-brains," which inflates public perception of what they are.

These are the mechanisms. A supply chain gated on cognitively impaired donors, and a safety assurance resting on the claim that the tissue cannot suffer. Both are load-bearing for any business built on the platform.

Where a skeptic should push

Because this is a review, the first discipline is to separate what it demonstrates from what it aggregates. Every ethics claim above is the review's report of a secondary source, not its own measurement, and I did not retrieve those underlying studies. The MacCAT-T figure of 102 patients, the pre-term EEG comparison and the ten-week myelination figure should each be read as "the review cites" and not as settled fact. That caveat matters most for the moral-status claims, which are the softest. One internal tension is worth not inheriting: the review lists blood-brain barrier dysfunction among the phenotypes these models capture while also naming the absence of functional vascularization as a core limitation, and an avascular organoid can only model barrier biology weakly, so that particular capability claim should be read with caution.

The load-bearing move to resist is the review's implied slider from developmental resemblance to the capacity for distress. Faster myelination and EEG that is statistically busier are not evidence of the architecture that pain requires. Nociception needs peripheral sensory neurons and afferent pathways the organoid does not have; the felt quality of suffering, on every mainstream account, needs thalamocortical arousal circuitry and a brainstem the organoid also lacks. A cortical organoid is a disconnected fragment with no body to be hurt and no arousal system to register it. Myelination speed is a timer on oligodendrocyte maturation, not a dial on sentience, and oscillatory EEG complexity is what any sufficiently large, spontaneously active neural network produces. Treating "more developed" as "closer to suffering" is the exact error the field's own consensus warns against, and the review reproduces it even while citing that consensus. It is worth adding that the pre-term EEG comparison is itself contested in the primary literature, and restating it even in order to reject it risks lending it more weight than it has earned.

The consent claim is sturdier, but it too has a defensible bound. Impaired capacity is not the same as no capacity, and every jurisdiction has machinery for exactly this situation: surrogate or proxy consent, assent from the patient plus authorization from a legally authorized representative, and advance directives given while capacity was intact. The MacCAT-T finding does not mean AD donors cannot contribute tissue; it means the consent has to move to a surrogate, and that the resulting line carries a heavier provenance burden than a line from a healthy adult. The correct claim is not "consent is impossible" but "consent is displaced and encumbered," which is a real and permanent friction on the supply chain rather than a wall.

Access, consent, and the moral-status placeholder

For a title that tracks platform access, vendor capability, and the governance of computing on living neural tissue, the non-obvious implication is what kind of upstream constraint this industry actually inherits. Almost all vendor effort, and almost all of this review's own technical discussion, aims at the manufacturing layer: better vascularization, lower batch variability, more standardization. Those improvements make each organoid better. None of them touches the input the whole enterprise depends on, but that input constraint is not volume. An iPSC line is immortal and renewable, so a platform needs only a handful of consented donors per genotype rather than a running stream of tissue, and monogenic familial-AD mutations can often be sourced from young, presymptomatic carriers who are fully competent to consent. What manufacturing scale cannot launder is legitimacy and representativeness. The common form of the disease is late-onset sporadic AD, and modelling it faithfully means recruiting exactly the older, cognitively impaired donors whose consent must run through surrogates or advance directives. So the binding upstream question is not how much tissue can be sourced but whether the most representative donor cohorts are being consented in a way that holds up, and no amount of bioreactor engineering answers it. The opportunity is for whoever builds provenance-clean, legitimately and representatively consented donor cohorts early: the durable asset in this field is not the printer or the protocol, it is the defensibly consented, representative cell bank, and it is scarce for a reason that lives at the consent layer rather than the factory floor.

The genuine threat is a value-capture asymmetry that the disease makes unusually sharp. That donors generally hold no property interest in a line derived from their cells, and that broad consent can lawfully cover unspecified commercial downstream use, is the settled status quo of cell-line law rather than a novel hole; a healthy donor faces the same thin ongoing say. What is specific to AD is that the asymmetry bites hardest exactly where agency is weakest. The review flags that donors want continued involvement, yet a sporadic-AD donor who consented through a surrogate under cognitive impairment is, more than almost any other donor, unable to exercise or renegotiate that involvement as an immortal line is screened against thousands of compounds for decades. Put in access terms: the platform converts a one-time, surrogate-mediated consent into a perpetual revenue-bearing asset held by the party least able to be checked by the person it came from. This is not a reason to stop; it is a reason to build benefit-sharing and dynamic-consent structures before large collections of legacy lines are locked in, because the population that would need to renegotiate is precisely the one losing the capacity to do so.

The governance implication of the moral-status material is narrower than the review's framing but more actionable. Strip out the developmental-slider reasoning and what remains is an epistemic gap. The assurance that these tissues cannot suffer is explicitly evidential and revisable: it holds "on current evidence," and the same assurance covers the whole field, not AD organoids specifically. Yet no body is charged with watching for the evidence that would revise it, and no automatic trigger is keyed to the very trends this review celebrates, longer culture and greater functional complexity, that would prompt a re-examination. The threat is not that today's AD organoid suffers; the honest reading is that it almost certainly does not. The threat is that the field is scaling the exact variables the carve-out is conditioned on while leaving the carve-out unmonitored, so the day the assurance stops being true, nobody is positioned to notice. That is a governance design defect, and it is fixable cheaply now and expensively later.

The bottom line

As a technical survey this review is competent and unsurprising. As a document about access and governance it is more useful than it means to be, because it locates two constraints that vendor roadmaps ignore. The consent issue is real but narrower than a supply bottleneck: the cited MacCAT-T data show that decisional capacity in mild-to-moderate AD is impaired unevenly across the four legal components, which turns the sourcing of representative sporadic-AD donors into a legitimacy question that manufacturing scale cannot settle. The moral-status concern is a hypothesis, and a weak one on today's evidence, but the governance gap it points to is real regardless of whether the tissue ever crosses a morally relevant threshold. What would change the picture: a demonstrated route to legitimately and representatively consented AD cohorts, through surrogate and dynamic consent, would neutralize the access question; a standing, evidence-triggered review body tied to culture duration and integration would close the governance gap. Absent both, the field is building an industry on an input whose legitimacy it has not secured and a safety claim nobody is watching.