Eight per cent eligible: a US model prices the anti-amyloid pathway against dementia care

Every US memory service is now being asked the same two questions in the same week: should we start infusing, and should we build a care team. This paper prices both against each other in one model, over a lifetime horizon, in the same currency. Its answer is uncomfortable for anyone selling either side as the obvious buy.

What the model priced, and in what units

Watch the units, because this is where dementia economics quietly overstates itself. Lecanemab's effect does not enter this model as points on a cognitive scale. It enters as a hazard ratio of 0.69 (95% CI 0.57 to 0.83) for moving from mild cognitive impairment or mild dementia to a more severe stage, and the model turns that benefit off at 18 months because placebo-controlled evidence does not run past that point. The output is then years: 0.17 life years, 0.17 QALYs, 0.18 community years. Converted to plain units, 0.17 of a year is about two months of quality-adjusted life across an entire remaining lifetime, and 0.18 of a year is about two months longer outside a nursing or residential setting. Collaborative care returns about three months of quality-adjusted life and about four months in the community.

The paper sets no threshold for how large a QALY gain must be before it counts as clinically meaningful, and it names no willingness-to-pay figure either. It says only that the combined-care ratio "exceeded cost-effectiveness thresholds" without stating which. What it does state, and what is the single most decision-relevant number in the study, is the price of the increment: adding lecanemab on top of collaborative care bought a further 0.16 QALYs and cost $248,000 per QALY from the healthcare perspective and $218,000 from the societal perspective. So the honest reading is: no established threshold is applied here, and the number to carry is not "significant" or "modest" but "about two months, at $225,500 per quality-adjusted year against usual care, and $248,000 per quality-adjusted year against a care team you could build instead".

The price gap is easier to grasp. Lecanemab enters at a wholesale acquisition cost of $26,500 per year, plus administration, four MRI scans a year for monitoring, and three more scans for the 21.5% assumed to develop amyloid-related imaging abnormalities. Symptomatic abnormalities were assumed in 3.5% of everyone treated in the first year, not in 3.5% of that 21.5%. Collaborative care enters at $2,150 per year. The paper also notes that published value-based price estimates for lecanemab range from $5,100 to $8,500 per year against that $26,500 market price.

Eight per cent, and the price of reaching them

The national arithmetic is decided by who is allowed in the door. Anti-amyloid therapy requires biomarker-confirmed amyloid positivity, an early stage, and the absence of a long list of common comorbid contraindications; the authors' estimate is that this leaves 8% of the mild cognitive impairment and mild dementia population, or 1.03 million people. Collaborative care has a gate of its own, a different one: it was scaled to the 92% of people with Alzheimer's dementia who live in the community, and people at the mild cognitive impairment stage are not eligible for it at all. The two headline counts therefore sit on different denominators and are not two slices of one population: 1.03 million against 6.16 million. That is why a drug with a real per-person effect yields about 175,000 QALYs nationally at $39.5 billion, while a care model built on softer inputs yields 1.5 million QALYs and returns $300 billion.

That national figure deserves a paragraph of arithmetic, because the paper prints it twice and differently: 180,000 QALYs in the abstract, 170,000 in the Results. This is not an unexplained slip. The per-person gain is pinned down three independent ways: $38,500 divided by the healthcare ICER of $225,500 is 0.1707, $34,500 divided by the societal ICER of $203,000 is 0.1700, and Table 2 gives 3.46 QALYs under lecanemab against 3.29 under usual care, a difference of 0.17. Multiplied by 1.03 million people, all three land near 175,000, exactly on the boundary where rounding to two significant figures can go up to 180,000 or down to 170,000. One loose end remains on the cost side: 1.03 million people at $38,500 each is $39.655 billion, not the $39.5 billion printed, and back-solving the population from $39.5 billion gives about 1.026 million people, which still yields roughly 175,000 QALYs. Carry 175,000, and read the two published figures as one number.

The authors are candid that they assumed everyone eligible actually receives the intervention. They point out that in practice the minority who reach infusion are likely to be more affluent and health-literate, and that racial minorities are typically diagnosed later, when the drug is no longer indicated. Their conclusion follows: funding the molecule without funding the care pathway would widen existing disparities rather than close them.

The funding line, the disclosure line, and what neither says

These are two separate statements in the paper and they carry different information. The funding line sits inside the acknowledgments: the study was funded by the National Institute on Aging (grant R01 AG074710) and the Alzheimer's Association. The conflict-of-interest statement is a different paragraph. It declares that the funding sources had no role in design, conduct, analysis or the decision to submit; that Katherine L. Possin received grants and other research funding from the NIA, the National Institute of Neurological Disorders and Stroke, the Global Brain Health Institute and the John Douglas French Alzheimer's Foundation outside the submitted work; that James G. Kahn received partial funding from the Global Brain Health Institute; and that Kelly J. Atkins reported no disclosures. No pharmaceutical company appears in either statement.

What neither statement names is visible in the methods. The utility gain of 0.06 that makes collaborative care the dominant strategy is drawn from the pivotal UCSF Care Ecosystem trial, cited as reference 21 and attributed in Table 1 to "Possin", the senior author of this paper. The authors are, in effect, pricing their own intervention against a manufacturer's drug. That is not grounds for discarding the numbers, but it is grounds for reading the sensitivity analyses, which the authors do report: the lower bound they tested assumed no savings at all from collaborative care once programme costs were counted, and collaborative care remained dominant there too.

For your practice

When a family asks whether to pursue lecanemab, the useful frame is not efficacy versus futility but eligibility, monitoring burden and magnitude. Roughly one in twelve people at the mild cognitive impairment or mild dementia stage will clear the criteria; the modelled gain is on the order of two months of quality-adjusted life; the commitment is biweekly infusions with quarterly MRI monitoring. None of that makes the drug worthless: in this model the arms containing lecanemab are the only ones that gain life years at all, and collaborative care on its own gains none.

What it does mean is that the care-side offer is not a consolation prize handed to the ineligible. In this model it is where the larger measured gain sits, it is available across moderate and severe stages where no disease-modifying therapy is indicated, and it is the only arm that pays for itself. The combined pathway produced the most QALYs of all, 0.42 above usual care, which is the argument for integrated dementia centres rather than infusion suites bolted onto nothing.

Keep the scepticism symmetrical. The patient quality-of-life input behind the care arm had a confidence interval that included zero, and the long-term-care delay was borrowed from a different programme than the one that supplied the utility figure. This paper prices two options honestly; it does not prove either one.