Cost-effectiveness of Aducanumab and Donanemab for Early Alzheimer Disease in the US

Key Points

Question

Would aducanumab and donanemab, 2 novel anti–amyloid monoclonal antibodies, be cost-effective for treating early Alzheimer disease in the US?

Findings

In this decision analytic modeling study based on clinical trial data, neither aducanumab nor donanemab was cost-effective at their expected prices of more than $25 000/y. Aducanumab became cost-effective when priced below $3000/y, whereas owing to its possibly greater efficacy (based on phase 2 trial data) and limited-duration dosing, donanemab was cost-effective when priced around $20 000/y.

Meaning

These findings suggest that anti–amyloid monoclonal antibodies are currently unlikely to be cost-effective, but future treatments with improved efficacy and limited-duration dosing could provide good health economic value.

This decision analytical model study evaluates the cost-effectiveness of aducanumab and donanemab for treatment of early Alzheimer disease in the US and estimates prices at which these agents would become cost-effective.

Abstract

Importance

Several anti–amyloid monoclonal antibodies have been developed for slowing the progression of Alzheimer disease (AD). Among the furthest developed are aducanumab, which received accelerated approval from the US Food and Drug Administration in 2021, and donanemab, which is currently undergoing phase 3 trials. The cost-effectiveness of these treatments has not been established.

Objectives

To estimate the cost-effectiveness of aducanumab and donanemab relative to standard care for early AD in the US.

Design, Setting, and Participants

A decision analytic model was used to estimate the lifetime health and economic outcomes of adults with early AD, from US healthcare sector and societal perspectives. Simulated patients had a mean (SD) age of 75.2 (5.5) years; 65% had mild cognitive impairment and 35% had mild dementia. Analyses were conducted from April 6, 2021, to January 20, 2022.

Interventions

Standard care, aducanumab (selected inputs including disease progression hazard ratio [HR] of 0.89 [95% CI, 0.63-1.15], annual price of $28 000, and twice-yearly monitoring with magnetic resonance imaging [MRI] of the brain), or donanemab (selected inputs including disease progression HR of 0.68 [95% CI, 0.44-0.99], annual price of $28 000, and twice-yearly monitoring with MRI of the brain and amyloid positron emission tomography [PET] monitoring). Donanemab was switched to placebo after substantial amyloid reduction on PET imaging, which occurred in 27% of patients at 6 months and 55% of patients at 12 months.

Main Outcomes and Measures

Quality-adjusted life-years (QALYs); costs, in 2020 US dollars; incremental cost-effectiveness ratios (ICERs); and value-based prices, defined as the maximum price at which a treatment would be cost-effective given a cost-effectiveness threshold of ICER of $150 000/QALY.

Results

Lifetime QALYs increased by 0.133 with aducanumab and 0.408 with donanemab. Total health care sector and societal costs increased by $130 100 and $127 800, respectively, with aducanumab and by $78 700 and $71 600, respectively, with donanemab, driven largely by drug costs ($119 000 for aducanumab and $44 600 for donanemab). Health care sector and societal ICERs relative to standard care were $981 000/QALY and $964 000/QALY, respectively, for aducanumab and $193 000/QALY and $176 000/QALY, respectively, for donanemab. In sensitivity analysis, aducanumab’s value-based price remained less than $50 000/y, even when assuming a 90% reduction in disease progression. Donanemab’s value-based price surpassed $50 000/y once its efficacy exceeded 50%.

Conclusions and Relevance

These findings suggest that at current expected prices, neither aducanumab nor donanemab would be cost-effective for early AD in the US. Donanemab’s dosing scheme, in which patients suspend treatment on achieving substantial amyloid reductions, may provide a rubric by which sufficiently effective anti–amyloid antibody treatments could be cost-effective even when priced comparably to other biologics.

Introduction

In June 2021, the US Food and Drug Administration granted accelerated approval to aducanumab, a monoclonal antibody against β-amyloid, for treatment of Alzheimer disease (AD).¹ This sparked controversy surrounding the approval process,^1,2 the dearth of evidence for the drug’s clinical efficacy,^3,4,5 its potential for adverse effects,^6,7 and especially its cost. With the drug’s initial price of $56 000/y, even modest uptake was projected to double or triple Medicare Part B drug spending^8,9; partly owing to these financial concerns, initial use of aducanumab has been limited, leading to its price being halved (to $28 200/y) in December 2021.¹⁰

Along with aducanumab, multiple additional anti–amyloid antibody treatments are progressing through the drug development pipeline, including donanemab, lecanemab, and gantenerumab.^11,12 The furthest developed of these is donanemab, which showed efficacy on some clinical end points in a phase 2 trial¹³ and began submission for accelerated approval in late 2021.¹⁴

The need for effective AD treatments is substantial. There are 6 million people with dementia due to AD in the US, a number that will double by 2050.¹⁵ Dementia increases mortality rates 2- to 3-fold and is responsible for 10% to 30% of US deaths.^16,17 US health care expenditures for people with AD were $355 billion in 2021 and could exceed $1 trillion by 2050.¹⁸ Thus, considering the magnitude of AD’s effects, even an expensive treatment could yield immense health and economic gains; but if its costs outweigh its clinical benefits, it could equally well exacerbate AD’s societal burden.

To address this challenge, we developed a decision-analytic model of AD treatment, incorporating published data on AD natural history, its health care and societal costs, and the efficacy and adverse effects of anti–amyloid antibody treatments. We used this model to evaluate the cost-effectiveness of aducanumab and donanemab for early AD in the US, and to estimate prices at which these agents would become cost-effective.

Methods

Overview

This decision analytic model study adhered to the 2013 Consolidated Health Economic Evaluation Reporting Standards (CHEERS). Analyses were conducted from April 6, 2021, to January 20, 2022. We developed a novel state-transition model of AD treatment in Excel 365 (Microsoft Corporation) (Figure 1). We used this model to simulate, over a lifetime horizon, the clinical and economic consequences of 3 treatment strategies for early AD: (1) standard care; (2) aducanumab infusions, with the dose titrated to 10 mg/kg every 4 weeks³; and (3) donanemab infusions, with the dose titrated to 1400 mg every 4 weeks.¹³

Figure 1. Model Structure.

Model states are represented by boxes; transitions between states are represented by arrows. Patients in each clinical stage have a monthly probability of transition to the next stage, denoted by r_n; when receiving anti–amyloid antibody treatment, this probability is increased or decreased according to the efficacy of the treatment, denoted by H. Patients in all states are subject to an age- and stage-specific mortality probability; for clarity, these arrows are omitted from the diagram. Every 12 months, all surviving patients transition to the next age category, as represented by the horizontal arrows. MCI indicates mild cognitive impairment.

Model outcomes included survival; quality-adjusted life-years (QALYs), a measure combining quality of life and longevity¹⁹; and health care sector and societal costs (eTable 1 in the Supplement). All costs were inflated to consistent 2020 US dollars using the Personal Consumption Expenditure index²⁰; future costs and QALYs were discounted 3% annually to reflect their present value.²¹

We calculated the incremental cost-effectiveness ratio (ICER) of each treatment relative to standard care as the ratio of its incremental cost to its incremental QALYs.¹⁹ Because no published studies have directly compared aducanumab and donanemab, and because the drugs’ trials used differing enrollment criteria,^3,13 we did not calculate ICERs comparing the two directly. We designated treatments with ICER of $150 000/QALY or less as cost-effective and explored alternative cost-effectiveness thresholds in sensitivity analyses.^22,23 Finally, we calculated each drug’s value-based price, defined as the maximum price at which it would be cost-effective.²⁴

Model Description

The model’s structure is shown in Figure 1. A simulated cohort is categorized by age and AD clinical stage. Stages are derived from a validated clinical rating scale (eg, Clinical Dementia Rating Scale [CDR])^25,26,27 and include mild cognitive impairment (MCI), mild dementia, moderate dementia, and severe dementia due to AD.

To capture the short-term dynamics of treatment discontinuation,^3,13 the model uses a 1-month cycle length. With each cycle, patients have a stage-specific probability of progression to the next disease stage; every 12 cycles, all patients progress to the next age. Each stage has a unique health care cost, additional societal cost, utility, and mortality hazard ratio (HR).

Anti-amyloid treatments may slow or accelerate disease progression according to a specified HR. Treated patients have a monthly probability of adverse events, with an associated cost and utility decrement. Finally, treated patients incur monthly costs of anti-amyloid drugs, infusion supplies and services, and treatment monitoring imaging and a 1-time cost of screening workup to identify eligible patients.

Input Parameters

Input values, sensitivity analysis ranges, and data sources^{3,10,13,16,17,21,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41} are shown in Table 1. Parameters are described in detail as follows.

Table 1. Model Input Data.

Parameter	Base case value	Probabilistic sensitivity analysis range	Distribution	Source
General
Annual discount rate, %	3	NA	NA	Sanders et al,21 2016
Time horizon	Lifetime	NA	NA	Sanders et al,21 2016
Population characteristics
Initial age, mean (SD), y	75.2 (5.5)	NA	Normal	Mintun et al,13 2021
Initial dementia stage, proportion of cohort
MCI	0.65	0.60-0.71	β	Mintun et al,13 2021
Mild	0.35	0.29-0.40	β	Mintun et al,13 2021
Natural history
Monthly probability of progression from stage
MCI	0.007	0.005-0.009	Normal	Mitchell et al,42 2009
Mild	0.016	0.013-0.020	Normal	Spackman et al,31 2012
Moderate	0.026	0.021-0.032	Normal	Spackman et al,31 2012
Mortality
Mortality without AD, monthly probability	0.0008-0.0193a	NA	NA	James et al,16 2014; Arias,29 2019; Matthews et al,30 2019
Mortality with AD, HR
MCI	1.61	1.49-1.74	Log-normal	Stokes et al,17 2020
Mild	2.23	1.77-2.82	Log-normal	Villarego et al,28 2011
Moderate	3.10	2.47-3.89	Log-normal	Villarego et al,28 2011
Severe	4.98	3.85-6.44	Log-normal	Villarego et al,28 2011
Utility with AD
MCI	0.73	0.69-0.77	Normal	Newmann et al,32 2000
Mild	0.69	0.67-0.71	Normal	Newmann et al,32 2000
Moderate	0.53	0.51-0.55	Normal	Newmann et al,32 2000
Severe	0.34	0.31-0.36	Normal	Newmann et al,32 2000
Treatment effects
Disease progression, HR vs placebo
Aducanumab	0.89	0.63-1.15	Normal	Budd Haeberlein et al,3 2019
Donanemab	0.68	0.44-0.99	Log-normal	Mintun et al,13 2021
ARIA probability
Aducanumab	0.41	0.38-0.44	β	Budd Haeberlein et al,3 2019
Donanemab	0.39	0.31-0.47	β	Mintun et al,13 2021
Utility decrement with symptomatic ARIA	−0.065	NA	NA	Pitkala et al,39 2008
Costs, 2020 $US
Monthly background health care sector
MCI	691	0-1382	Normal	Michaud et al,27 2017
Mild	1049	628-1469	Normal	Michaud et al,27 2017
Moderate	1274	780-1768	Normal	Michaud et al,27 2017
Severe	3764	2324-5203	Normal	Michaud et al,27 2017
Monthly background additional societal
MCI	49	2-96	Normal	Michaud et al,27 2017
Mild	1483	966-1999	Normal	Michaud et al,27 2017
Moderate	1932	1337-2527	Normal	Michaud et al,27 2017
Severe	2153	1128-3177	Normal	Michaud et al,27 2017
Annual drug
Aducanumab, year 1	20 500	NA	NA	Biogen Investor Relations,10 2021; Biogen,33 2021
Aducanumab, years 2 or later	28 200	NA	NA	Biogen Investor Relations,10 2021; Biogen,33 2021
Donanemab	28 200	NA	NA	Biogen Investor Relations,10 2021; Biogen,33 2021
Monthly infusions and monitoring
Aducanumab, months 1-6	112	57-167b	Normal	CMS,34 2021; Biogen,36 2018
Aducanumab, months 7-12	87	45-130b	Normal	CMS,34 2021; Biogen,36 2018
Aducanumab, months 13 and later	80	41-120b	Normal	CMS,34 2021; Biogen,36 2018
Donanemab, months 1-6	599	305-892b	Normal	Mintun et al,13 2021; CMS,34 2021
Donanemab, months 7-12	302	154-450b	Normal	Mintun et al,13 2021; CMS,34 2021
Donanemab, months 13 and later	287	146-427b	Normal	Mintun et al,13 2021 CMS,34 2021
Screening per eligible patient
Aducanumab	6957	3548-10366b	Normal	CMS,34 2021; Krudys,37 2021
Donanemab	17 096	8719-25 474b	Normal	Mintun et al,13 2021; CMS,34 2021
ARIA-related
Follow-up imaging	1060	NA	NA	CMS,34 2021; Cummings et al,352021; Chalkias et al,38 2021
Hospitalization	35 934	NA	NA	Vahidy et al,40 2019; AHRQ,41 2021
Outpatient visits	128	NA	NA	CMS,34 2021

Abbreviations: AD, Alzheimer disease; AHRQ, Agency for Healthcare Research and Quality; ARIA, amyloid-related imaging abnormalities; CMS, Centers for Medicare & Medicaid Services; HR, hazard ratio; MCI, mild cognitive impairment; NA, not applicable.

^a Mortality probability without AD is age-dependent; the range of mortality probabilities from ages 60 to 99 years is shown.

^b For cost inputs where the data source did not enable calculation of a 95% CI, a normal distribution is used in uncertainty analysis with SD set to one-quarter of the mean.

Population Characteristics

The mean (SD) age of the simulated population was 75.2 (5.5) years, drawn from donanemab’s phase 2 trial.¹³ At baseline, 65% of patients had MCI and 35% mild dementia due to AD, derived by applying disease stage cut points²⁵ to the trial’s baseline CDR Sum of Boxes (CDR-SB) score distribution.¹³

Natural History

We used a mortality HR of 2.23 (95% CI, 1.77-2.82) for mild dementia, 3.10 (95% CI, 2.47-3.89) for moderate dementia, and 4.98 (95% CI, 3.85-6.44) for severe dementia from a European observational study.²⁸ For MCI, we used an HR of 1.61 (95% CI, 1.49-1.74), from individuals with cognitive impairment without dementia in a US prospective cohort.¹⁷ We calculated age-specific mortality probabilities for individuals without dementia (to which the above HRs are applied) using US Centers for Disease Control and Prevention life tables²⁹ along with AD prevalence rates of 3.6% at ages 65 to 74 years, 13.6% at ages 75 to 84 years, and 34.6% at 85 years or older³⁰ and an overall mortality HR of 3.13 (95% CI, 2.74-3.58) for patients with AD clinical syndrome.¹⁶

Disease progression probabilities were derived from a national multicenter study³¹ and a meta-analysis of MCI progression in specialist settings.⁴² For consistency with our model, we dichotomized probabilities into progression and nonprogression rather than allowing transitions to less severe stages or across multiple stages.³¹ This yielded monthly progression probabilities of 0.007 for MCI, 0.016 for mild dementia, and 0.026 for moderate dementia due to AD.

Quality of Life

We used utility estimates of 0.73 for patients with MCI, 0.69 for patients with mild dementia, 0.53 for patients with moderate dementia, and 0.34 for patients with severe dementia, from a US cross-sectional study of patients with AD clinical syndrome and their caregivers.³² Based on data from the aforementioned study and a European multicenter analysis suggesting that current measures of caregivers’ health-related quality of life do not vary markedly with dementia severity,^32,43 we did not model caregiver utility outcomes.

Background Care Costs

We derived background health care sector costs (including outpatient care, inpatient care, home care, medications, and long-term care) and societal costs (including the above plus unpaid caregiving) from a 3-year survey-based study of patients with AD clinical syndrome in Michigan.²⁷ As disease stage increased from MCI to severe dementia, monthly health care sector costs increased from $690 to $3760, and additional societal costs increased from $50 to $2150.

Treatment Efficacy

We estimated the efficacy of aducanumab and donanemab using primary outcome data from their phase 3 and phase 2 trials.^3,13 To parameterize our model, we assumed a treatment’s relative reduction in disease progression would equal its relative improvement in mean score on a cognitive and functional scale; this assumption is discussed further in eMethods 1 and eTable 2 in the Supplement.

For aducanumab, two phase 2 trials have been reported,³ with a primary outcome of CDR-SB score. In ENGAGE (N = 1647), high-dose aducanumab accelerated decline in the CDR-SB score by 0.03 (95% CI, −0.26 to 0.33) vs placebo; in EMERGE (N = 1638), high-dose aducanumab slowed decline in the CDR-SB score by −0.40 (95% CI, −0.71 to −0.10) vs placebo.³ Random-effects restricted maximum-likelihood meta-analysis (eFigure 1 in the Supplement), followed by conversion to HR vs placebo, yielded a disease progression HR of 0.89 (95% CI, 0.63-1.15). For a more optimistic sensitivity analysis, we estimated efficacy for a post hoc patient subgroup who followed updated trial protocols, which enabled greater cumulative dosing (eFigure 2 in the Supplement)³; this yielded a disease progression HR of 0.71 (95% CI, 0.50-0.92).

For donanemab, one phase 2 trial (N = 257) has been reported,¹³ with a primary outcome of Integrated AD Rating Scale score, a composite comprising elements of 2 preexisting functional and cognitive scales. Relative to placebo, donanemab slowed the decline in Integrated AD Rating Scale score by 32% (95% CI, 1%-56%), corresponding to a disease progression HR of 0.68 (95% CI, 0.44-0.99).

Treatment Cost

Aducanumab’s base-case price was $20 500 for the first year (accounting for dose titration) and $28 200/y thereafter.^10,33 In the absence of data on donanemab’s projected price, we assumed a base-case price of $28 200/y to ensure comparability.

We applied $72 in materials and services costs per infusion (Current Procedural Terminology [CPT] code 96365).³⁴ Patients taking aducanumab received twice-yearly magnetic resonance imaging (MRI) of the brain,^3,35 costing $353 per study (CPT code 70553).³⁴ Patients taking donanemab received twice-yearly brain MRI and twice-yearly amyloid positron emission tomography (PET),¹³ with per-study costs of $2950 for florbetapir tracer (CPT code A9586) and $1507 for the imaging (CPT code 78814).³⁴

For aducanumab, 38% of patients in phase 3 trials withdrew or discontinued treatment; we simulated linear discontinuation through 12 months, then stable treatment rates thereafter.³ For donanemab, 66% of patients withdrew or discontinued treatment; treatment was suspended after substantial reduction of amyloid levels in 27% of remaining patients at 6 months and 55% at 12 months.¹³ We assumed patients suspending treatment after reduction of amyloid levels would incur ongoing monitoring costs; patients withdrawing from treatment incurred no further treatment-specific costs. We assumed patients with severe dementia would discontinue treatment given the minimal expected benefit.⁴⁴

Full details on screening costs are provided in eMethods 2 in the Supplement. Briefly, for every patient eligible for aducanumab’s phase 3 trials, 0.41 were excluded by amyloid PET findings and 0.03 by MRI findings.^36,37 For every patient eligible for donanemab’s phase 2 trial, 1.51 were excluded by tau PET findings, 0.04 by amyloid PET findings, and 0.03 by MRI findings.¹³ Combining these estimates of the number of patients undergoing imaging workup with the unit costs described above yielded total screening costs per eligible patient of $6957 for aducanumab and $17 096 for donanemab.

Adverse Events

Amyloid-related imaging abnormality (ARIA) probabilities were 41% (24% of which were symptomatic) in aducanumab’s clinical trials³ and 39% (16% of which were symptomatic) in donanemab’s clinical trial.¹³ Two participants (4%) with ARIA in donanemab’s trial were hospitalized for confusion¹³; because hospitalization data were not reported for aducanumab, we assumed an equivalent rate for aducanumab-treated patients. We assumed 50% of ARIA cases would occur by month 3, 40% in months 4 to 12, and 10% in months 13 to 24.^3,13,45

All patients with ARIA incurred costs of 1 additional 30-minute physician visit (CPT code 99214 [$128]) and monthly MRIs until resolution,³⁵ with a mean duration of 3 months.^38,45 Because the most common symptoms of ARIA are confusion and altered mental status,^35,45 we used delirium as an analogue for ARIA’s consequences. For symptomatic ARIA, we applied a utility decrement of 0.065, lasting 3 months, drawn from a study on the quality-of-life consequences of delirium.³⁹ We assumed ARIA-related hospitalizations would have a mean duration of 11.6 days,⁴⁰ costing $3092/d.⁴¹

Sensitivity Analysis

To test our findings’ robustness to uncertainty and alternative assumptions, we performed multiple sensitivity analyses. In univariate sensitivity analysis, we examined the impact of individual model parameter values on cost-effectiveness outcomes. In probabilistic sensitivity analysis, we ran the model 1000 times while drawing parameter values at random from their uncertainty distributions, to estimate the overall uncertainty in model outcomes attributable to parameter uncertainty.¹⁹ Finally, we examined the following specific alternative scenarios:

1. a more favorable aducanumab disease progression HR of 0.71 (95% CI, 0.50-0.92), from the subgroup analysis of aducanumab’s clinical trials described above³;

2. faster disease progression probabilities for untreated patients (0.020/mo for MCI, 0.037/mo for mild dementia, and 0.028/mo for moderate dementia due to AD), from an observational study of amyloid-positive patients⁴⁶ (notably, these progression rates are markedly higher than in other analyses using a range of AD biomarkers, suggesting that they represent the upper range of progression rates that could be observed in biomarker-defined high-risk populations^47,48,49);

3. no screening costs, reflecting a scenario under which no nonroutine studies are needed to identify treatment-eligible patients;

4. a maximally optimistic scenario combining the above 3 alternative assumptions;

5. alternative background costs from a multinational study⁵⁰ (with progression from MCI to severe dementia, monthly health care sector costs ranged from $690 to 4920 and additional societal costs ranged from $50 to $1470); and

6. discontinuation of aducanumab and donanemab infusions in patients who progress to moderate dementia due to AD.

Results

Model Validation

To assess our model’s external validity, we compared simulated outcomes with independent published estimates. A survival analysis of patients with newly diagnosed AD clinical syndrome in Washington found 25% mortality at 2.7 years, 50% at 4.9 years, and 75% at 8.1 years.⁵¹ Simulating an equivalent population, our model projected corresponding survival times of 2.1, 4.4, and 7.8 years. Considering international data, a meta-analysis found mean (SD) survival from AD diagnosis of 5.8 (2.0) years,⁵² compared with 5.3 years in our model.

Economic outcome estimates are more heterogeneous, likely owing to variation in payers, informal caregiving costs, and out-of-pocket expenditures; we thus present a range of estimates rather than a single benchmark.^53,54,55 A review of Medicare managed care enrollees with dementia yielded annual expenditure estimates for prevalent cases ranging from $9252 to $28 698.⁵⁶ An analysis combining claims and survey data from people with dementia in the last 7 years of life yielded annual health care sector costs (including out-of-pocket costs) of $36 149; incorporating unpaid caregiving increased this to $54 605.⁵³ Finally, the Alzheimer’s Association estimated annual per-patient expenditures from all sources to be $52 481.¹⁸ For comparison, our model projected annual health care sector and societal costs of $24 217 and $44 084, respectively.

Base Case

Relative to standard care, aducanumab increased lifetime QALYs by 0.133; donanemab increased QALYs by 0.408 (Table 2). Total health care sector and societal costs increased by $130 100 and $127 800, respectively, with aducanumab, and by $78 700 and $71 600, respectively, with donanemab, driven largely by drug costs ($119 000 for aducanumab and $44 600 for donanemab). Drug costs were nearly equivalent through 1 year ($15 900 for aducanumab and $16 400 for donanemab) before diverging thereafter (Figure 2). Health care sector and societal ICERs were $981 000/QALY and $964 000/QALY, respectively, for aducanumab and $193 000/QALY and $176 000/QALY, respectively, for donanemab.

Table 2. Base Case Results.

	Standard care	Difference vs standard care
		Aducanumab	Donanemab
QALYs	4.948	0.133	0.408
Costs, 2020 $US
Total health care sector	118 000	130 100	78 700
Background care	118 000	−3800	−11 300
Screening workup	NA	7000	17 100
Drug	NA	119 000	44 600
Monitoring and infusions	NA	6900	27 300
Adverse effects	NA	1000	1000
Total societal	205 200	127 800	71 600
Background care	205 200	−6000	−18 400
Screening workup	NA	7000	17 100
Drug	NA	119 100	44 600
Monitoring and infusions	NA	6900	27 300
Adverse effects	NA	1000	1000
ICER, $US/QALY
Health care sector	NA	981 000	193 000
Societal	NA	964 000	176 000
Value-based price, 2020 $US
Health care sector	NA	2000	17 000
Societal	NA	3000	22 000

Abbreviations: ICER, incremental cost-effectiveness ratio; NA, not applicable; QALY, quality-adjusted life-year.

Figure 2. Treatment-Related Costs Over Time.

Vertical bars indicate the cost, in 2020 US dollars, of various components of anti–amyloid antibody treatment during 5 years from the start of treatment.

Value-Based Price Thresholds

In the base case, aducanumab’s value-based price was $2000/y from a health care sector perspective and $3000/y from a societal perspective (Table 2). Corresponding value-based price estimates for donanemab were $17 000/y and $22 000/y.

Figure 3 shows value-based price estimates for aducanumab and donanemab across a range of hypothetical efficacy values from a health care sector perspective. Aducanumab’s value-based price exceeded $25 000/y when its disease progression HR surpassed 0.50; even with a disease progression HR of 0.10, its value-based price remained below $50 000/y. Donanemab’s value-based price exceeded $25 000/y when its disease progression HR surpassed 0.70 and exceeded $50 000/y once its disease progression HR surpassed 0.50. Results were similar from a societal perspective (eFigure 3 in the Supplement).

Figure 3. Value-Based Price Estimates With Varying Treatment Efficacy, Health Care Sector Perspective.

Value-based price estimates for aducanumab and donanemab, in 2020 US dollars, are shown at varying values of treatment efficacy expressed in terms of disease progression hazard ratio (HR) relative to placebo. Solid lines show value-based price estimates for aducanumab and donanemab using base-case inputs (except for efficacy); shaded areas indicate 95% CIs derived from probabilistic sensitivity analysis around all parameters except efficacy. Diamonds indicate the value-based prices of aducanumab and donanemab at their base-case efficacy values.

Sensitivity Analysis

In univariate sensitivity analyses (eFigures 4-7 in the Supplement), aducanumab’s ICER remained greater than $150 000/QALY with all input variations. Donanemab’s ICER decreased to less than $150 000/QALY when its disease progression HR was set to the lower limit of its 95% CI, the population’s initial mean (SD) age was reduced to 65 years, or disease progression probabilities were increased to the upper limit of their 95% CIs (societal perspective only).

In probabilistic sensitivity analysis (eFigure 8 in the Supplement), aducanumab reduced QALYs vs standard care in 20.1% of model runs; donanemab reduced QALYs in 1.7% of runs. Aducanumab had a less than 5% likelihood of being cost-effective at cost-effectiveness thresholds up to $310 000/QALY. Donanemab had a 34% to 42% likelihood of being cost-effective (depending on perspective) at a $150 000/QALY threshold and a greater than 50% likelihood of thresholds above $170 000/QALY to $190 000/QALY.

In scenario sensitivity analyses (eTable 3 in the Supplement), aducanumab was not cost-effective under any scenario examined. Donanemab was cost-effective under several scenarios, with its value-based price exceeding $50 000/y only under the maximally optimistic scenario.

Discussion

In this model-based economic evaluation, we found that neither aducanumab nor donanemab is likely to be cost-effective by US standards at their expected prices of more than $25 000/y. To become cost-effective, aducanumab’s price would need to decrease to less than $3000/y; donanemab, in contrast, could be cost-effective at a price of $20 000/y.

Our findings are consistent with results from the Institute for Clinical and Economic Review,⁴⁴ which estimated an ICER for aducanumab of greater than $1 000 000/QALY based on its initially reported price. To our knowledge, no prior studies have evaluated donanemab’s cost-effectiveness, although 2 modeling analyses found that a hypothetical treatment with 20% to 30% efficacy and an annual cost of $10 000 to $20 000 could be cost-effective.^57,58 These studies did not capture many specifics of current anti-amyloid therapies, but their findings are roughly consistent with our estimates of donanemab’s cost-effectiveness.

Although our results suggest that donanemab could offer better health-economic value than aducanumab, there is a critical caveat: donanemab’s efficacy estimates were drawn from a small phase 2 trial,¹³ and hence should be considered preliminary until confirmed in phase 3 trials. In addition, our cost-effectiveness findings may well reflect differences in the drugs’ trial protocols more than intrinsic pharmacological differences. Specifically, donanemab-treated patients were required to have PET evidence of intermediate tau burden (vs positive amyloid PET findings only), were monitored with amyloid PET scans (vs brain MRI only), and suspended treatment once PET scans showed sufficient amyloid clearance (vs continuing indefinitely).^3,13

This last factor proved critical to the cost-effectiveness of anti-amyloid drugs. We found that an indefinitely dosed treatment would not warrant a price of greater than $50 000/y, as was initially proposed for aducanumab, even if it slowed disease progression by 90%. In contrast, a limited-duration treatment akin to donanemab could provide good health economic value at $50 000/y if it slowed progression by a more achievable 50%.

As additional anti-amyloid treatments are developed, our findings can help contextualize trial results and guide cost-effective drug utilization. As an example, lecanemab is another anti–amyloid monoclonal antibody undergoing phase 3 trials; in a phase 2 trial with indefinite dosing, it slowed disease progression by 26% to 56%, depending on outcome.⁵⁹ If these efficacy findings are borne out, our results suggest lecanemab might warrant a value-based price of $10 000/y to $25 000/y. However, if similar efficacy could be achieved with limited-duration dosing, its value-based price would rise to $15 000/y to $60 000/y, approaching typical biologic drug pricing in the US.^33,60,61

Considering the benefits of limited-duration dosing for the health care system (making impactful AD treatments economically viable) and drug manufacturers (enabling justifiably high drug prices), future research should evaluate this approach more thoroughly. In the short term, we would suggest tailoring aducanumab’s confirmatory and postmarketing surveillance studies to assess its efficacy after treatment suspension.⁹ For drugs earlier in the pipeline, assessing the stability of clinical and biomarker effects off treatment would support development of limited-duration dosing schemes.^62,63

Limitations

Our study has limitations with regard to the underlying data used, the scope of the analysis, and its modeling assumptions. Most importantly, aducanumab’s clinical efficacy has not been definitively demonstrated: critics have noted that the drug failed to meet its primary end point in 1 of 2 phase 3 trials, and attempts to rectify this discrepancy relied on post hoc subgroup analyses.^2,3,4 To address this, we derived our base-case efficacy estimate from all patients in the trials’ high-dose arms and used the more optimistic subgroup analysis in sensitivity analysis. In addition, outcome projections for both donanemab and aducanumab relied on extrapolation beyond their trials’ horizons; long-term data on the durability of clinical effects, as well as the duration of treatment used in clinical practice, are currently lacking.^3,13,25,26

Next, our analysis has limitations in its scope and generalizability. Several clinical trials are currently under way evaluating anti–amyloid antibodies for preclinical AD (cognitively unimpaired individuals with biomarker evidence of AD).^64,65 The results of the present study are not generalizable to this novel indication owing to differences in eligibility criteria, dosing schemes, disease progression rates, and patient characteristics (including age, medical comorbidities, and employment status); dedicated studies will be needed to assess the cost-effectiveness of anti-amyloid treatment in preclinical AD. More broadly, we recognize that incremental cost-effectiveness is not the only means of defining a treatment’s value; factors such as scientific advances spurred by novel treatments⁶⁶ or the hope offered to patients with terminal illness⁶⁷ were not incorporated into our analysis, but could reasonably inform the decisions of patients, clinicians, and policy makers.

Finally, our model’s structure elides many complexities of AD care, including sociodemographic differences in outcomes,¹⁸ concomitant use of traditional symptomatic treatments such as acetylcholinesterase inhibitors and memantine,¹³ and impacts of AD on caregivers’ well-being that may not be captured by standard health-related quality of life scales.¹⁸ Despite these simplifying assumptions, the model is well-validated by external data^51,52,53; however, if any of these factors proves to substantially moderate anti-amyloid treatment efficacy, these simplifications should be reconsidered.

Conclusions

Our findings suggest that at their current expected prices, neither aducanumab nor donanemab would be cost-effective for the treatment of early AD in the US. Although aducanumab’s price would need to fall to less than $3000/y to become cost-effective, donanemab—if its efficacy is confirmed in phase 3 trials—could be cost-effective when priced at $20 000/y. The limited-duration dosing scheme used with donanemab is critical to its greater health-economic value; this approach may provide a rubric by which sufficiently effective anti-amyloid drugs could be economically viable in the US health care system, even when priced comparably to other biologics.

Supplement.

eMethods 1. Treatment Efficacy Derivation

eMethods 2. Screening Cost Derivation

eTable 1. Impact Inventory

eTable 2. Estimated Disease Progression Probabilities With Donanemab vs Placebo

eTable 3. Scenario Sensitivity Analyses

eFigure 1. Meta-analysis of Aducanumab’s Effects on CDR-SB Relative to Placebo

eFigure 2. Meta-analysis of Aducanumab’s Effects on CDR-SB Relative to Placebo, Post Protocol V4 Population

eFigure 3. Value-Based Price Estimates With Varying Treatment Efficacy, Societal Perspective

eFigure 4. Univariate Sensitivity Analyses, Aducanumab, Health Care Sector Perspective

eFigure 5. Univariate Sensitivity Analyses, Aducanumab, Societal Perspective

eFigure 6. Univariate Sensitivity Analyses, Donanemab, Health Care Sector Perspective

eFigure 7. Univariate Sensitivity Analyses, Donanemab, Societal Perspective

eFigure 8. Probabilistic Sensitivity Analysis