Analysis of Costs for Imaging-Assisted Pharmaceutical Intervention in Alzheimer’s Disease with Lecanemab: Snapshot of the First 3 Years

INTRODUCTION

Alzheimer’s disease (AD) is an age-related chronic illness characterized by a progressive decline in learning, memory, communication, and overall cognitive function. AD is the most common neurodegenerative dementia spectrum disorder (DSD) and accounts for nearly two-thirds of all dementia diagnoses worldwide [1, 2]. There is no curative therapy for AD and continued neurodegeneration is ultimately fatal to the patient.

Considering the lack of treatment options for patients with AD, in July 2021 the Food and Drug Administration (FDA) granted accelerated approval for aducanumab (brand name: Aduhelmcopyright), a first of its kind pharmaceutical for the treatment of AD that may slow disease progression. Researchers demonstrated that aducanumab functions by clearing abnormal accumulations of insoluble amyloid-β proteins from the brain [3, 4]. This protein was chosen as the target of this pharmaceutical due to its widely believed role as an initiating factor in the amyloid cascade hypothesis, which purports that AD is caused by a cascade of events which lead to cell death and eventually cause dementia [5]. It does, however, remain to be seen if the drug stops or slows down the progression of dementia and its recent accelerated approval remains controversial due to ongoing concerns regarding exorbitant cost, safety and efficacy [6].

In April 2022, Centers for Medicare and Medicaid Services (CMS) released a coverage policy for aducanumab which limited its coverage only to those who are enrolled in clinical trials, citing potential significant risks of serious side effects [6]. This unusual decision on behalf of CMS, who have typically approved coverage for FDA-approved drugs in the past, appears to be the result of an abundance of caution because millions of people are expected to be treated with aducanumab and other similar Alzheimer-modifying drugs in the coming years. This newfound availability of a disease-modifying therapy will likely change the approach of medicine in managing age- and disease-related cognitive decline because previously only modest symptomatic treatment was available (e.g., acetylcholinesterase inhibitors) and many patients did not seek medical attention upon experiencing subjective cognitive decline [7]. Therefore, it is quite likely that new medications will spur an increase in demand from memory/cognitive specialists and from primary care physicians.

While many anti-amyloid medications have undergone clinical trials and failed to receive approval, further pharmaceuticals are currently under development and at various stages of approval. For instance, donanemab is currently in phase III trials and results are expected to be released at the end of 2023. More promising is the recent full approval of the anti-amyloid pharmaceutical called lecanemab (brand name: Leqembicopyright). The very recently completed double-blind phase III clinical trial involved 1,795 amyloid-positive patients. Results over 18 months showed a 27% reduction of cognitive decline and decreased amyloid burden (difference: –59.1 centiloids) in patients treated with lecanemab compared to placebo group [8]. Like aducanumab, lecanemab carries a significant price (approximately $26,500 USD/year) and carries the risk of very severe side-effects [8]. Questions of affordability, therapeutic efficacy, and patient safety have been raised as the FDA did not hold any public advisory meetings regarding the approval of lecanemab.

Like aducanumab, the FDA indicated that lecanemab should only be used with patients at the earlier stages of AD progression; however, while dementia diagnosis is a somewhat routine clinical task, AD-specific dementia diagnoses are more complex. Current clinical practices combine medical history and neuropsychological examination as the recommended means for the diagnosis of dementia [9]. Large postmortem studies have shown that this method of AD diagnosis is moderately sensitive (70.9% –87.3%) but only modestly specific (44.3% –70.8%) at best and suffers from numerous issues of replication between facilities, even for the oft-touted gold standard of autopsy [10–12]. AD diagnosis is further complicated by a potentially prodromal state known as mild cognitive impairment (MCI). Defined as a group by greater than expected cognitive decline (measured by neuropsychological testing) compared to age and education matched healthy controls, patients with MCI are at specific risk for developing dementia; however, the MCI patients who indeed progress to dementia (pMCI) only represent a minority of patients with MCI and most remain cognitively stable (sMCI) for years, though they may progress to dementia later or even revert to healthy cognition [13–16].

While dementia diagnosis does allow for access to specialized therapies, care, and insurance coverage, diagnosis at the stage where the impact of AD on cognition may be classified as dementia does not allow sufficient time for the full effect of disease-modifying therapies to be realized before patient death. Under a paradigm of early detection for pharmaceutical intervention, disease-modifying therapies and preventative measures would be more effective when used as early as possible in AD progression; however, because identification of AD is already a difficult diagnostic task, predicting cognitive decline at the MCI stage is prohibitively difficult for physicians using current practices. Due to the high cost and risk of extremely serious side effects of anti-AD drugs, it would be unethical, exorbitantly costly, and inefficient to prophylactically treat all patients with MCI because the vast majority would not require these therapies then or perhaps ever. There is, therefore, a significant and pressing clinical need for the accurate identification and discrimination of pMCI and sMCI neuropathologies. We present an analysis of the costs and savings associated with using neuroimaging methods to identify patients with incipient dementia caused by prodromal AD as candidates for pharmaceutical intervention with lecanemab and other future anti-amyloid medications, specifically with a focus on reducing the number of patients with sMCI erroneously identified as candidates for these pharmaceuticals.

Diagnostic neuroimaging procedures have emerged as a valuable tool for the in vivo identification and differential diagnosis of DSDs. Positron emission tomography (PET) has been invaluable to the neurodegenerative research community in elucidating the mechanisms and etiology of AD within the brain [17]. PET images are based on molecules called tracers which are administered in sub-pharmacological dosages to a subject for the imaging procedure. The principle upon which PET studies are predicated is that the distribution and concentration of a tracer is indicative of regional abundances of the target for which that tracer was designed.

PET imaging techniques, and in particular imaging with the glucose analog fluorodeoxyglucose (FDG-PET), are especially valuable in the early detection of AD. Distinct changes in the resting rate of glucose metabolism have been found which precede and coincide with the clinically evaluable manifestation of dementia-related cognitive symptoms [13, 15]. PET studies have also shown immense success in the development of the widespread AT(N) biomarker system (A: amyloid, T: tau, N: neurodegeneration) [5]. Amyloid-PET has also seen widespread application in clinical trials of anti-amyloid pharmaceuticals and the differential diagnosis of DSDs [8, 18]. Tau-PET still currently struggles with non-trivial issues like off-target binding or binding to multiple isoforms of tau proteins, but design and validation of second-generation tau tracers is well underway within the radiochemical research community [19–23]. Furthermore, tau-PET has been demonstrated to correlate exceptionally well to the already well-defined and commonplace Braak staging system for the spread of neurofibrillary tau tangles used at autopsy [24]. In fact, even with outstanding issues, tau-PET may outperform amyloid-PET and volumetric magnetic resonance (MR) imaging for prognosticating cognitive decline when compared on a one-to-one basis [25]. Assays of the patient’s cerebrospinal fluid (CSF) may also be used as biomarkers of amyloid and tau proteinopathies [5]. Despite these being frequently touted as less costly than amyloid PET or tau PET, the collection of CSF is nonetheless an invasive and potentially injurious procedure that also opens the possibility of meningeal infections, though these procedures are becoming commonplace in the practice of cognitive neurology.

Despite the great success of amyloid-PET and tau-PET in the research context, only the use of FDG-PET is considered clinical norm, though use of amyloid tracers is rapidly becoming more common since amyloid-PET and MR imaging are often requirements for determining eligibility in trials of anti-AD medications and monitoring their effects [8, 26]. FDG-PET is a commonly available and relatively widespread imaging modality, but even so neuroimaging studies have not yet become the clinical norm for dementia work-up [9]. Aside from niche usages such as eligibility for clinical trials, when neuroimaging studies are ordered, they are most often used for the differential diagnosis of DSDs. They are then evaluated by a specialist physician’s subjective impression of the spatial distribution of tracer uptake throughout the brain rather than being subject to objective data-driven analysis. Due to the rather modest sensitivity of traditional AD diagnosis and impracticality of the prognostication of cognitive decline, current practices are insufficient to implement the model of early detection for pharmaceutical intervention that must exist for the ethical and cost-efficient use of currently available anti-AD medications. Their great risk and cost necessitate correspondingly stringent specificity and sensitivity since rates of dementia progression are relatively small for patients with clinically verified MCI—approximately 0.7% monthly (or 22.3% overall) of all subjects over a 3-year period [27].

The in vivo quantification of AD-specific proteinopathies and abnormal resting cerebral glucose metabolism have repeatedly proven themselves in the neurodegenerative research literature as valuable methods of identifying clinical and prodromal AD in a variety of contexts with multiple distinct neuroimaging modalities, especially when they have been combined with machine intelligence [28]. The combination of machine intelligence with well-validated quantitative methods have yielded tools capable of differentiating subjects with AD from stable healthy controls with accuracies more than 95% and discriminating between pMCI and sMCI subjects with accuracy approaching 85% [28]. Both of these are well in excess of the performance of specialist human physicians, however in the following presentation we will make no assumption about the integration of machine intelligence into clinical decision-making, but rather leave an open door towards accepting any combination of machine intelligence, biomarker-based investigation, data-driven investigation or other clinical procedure/calculation which accurately predicts cognitive decline or stability with neuroimaging-based data.

MATERIALS AND METHODS

Since the scope of the current article is a snapshot of the approximate costs anti-AD treatment (e.g., lecanemab) of the first 3 years of broad availability to the general population, several assumptions are warranted in the discussion to follow.

There are approximately 5 million people living with MCI due to AD in the United States as of 2023 [29]. The following scenarios presented will assume that 1 million people would be willing take on the financial burden and risk the potential side effects of anti-AD pharmaceutical treatment over a 3-year period and present the relevant cost for different treatment scenarios related to the sensitivity and specificity of different combination of neuroimaging modalities for the identification of patients with prodromal AD. Reasonable direct and indirect costs incurred to both the patient and society respectively (irrespective of insurance coverage) because of their cognitive status are assumed to be $691 and $49 monthly for MCI, and $1,049 and $1,483 monthly for very mild and mild dementia with AD [27, 30]. The annual cost of lecanemab is currently estimated to be $26,500 [31]. All dollar values are in 2020 USD.

We also assume that all patients with MCI and AD survive throughout the 3-year period and that treatment with lecanemab has 90% efficacy [30]. Reasonable costs for diagnosis of MCI are assumed and used as a baseline estimate [32]. Costs for MCI diagnosis were reported in 2015 Euros (EUR) but were adjusted for purchasing power parity and average exchange rate in 2015 and then adjusted for inflation from 2015 to 2020 to increase comparability to other costs. The total amount assumed per subject is $697. In this figure we assume visitation to a memory clinic consists of a single appointment, multiple forms of blood testing, neuropsychological examination and consideration of medical history.

Reasonable costs from experience with our local imaging programs have assumed prices of $2,000 for FDG-PET, $4,000 for amyloid PET and $500 for MR imaging sessions. We will present cases for several scenarios of cost for the pharmaceuticals and for the diagnostic neuroimaging studies. Summaries of all results are presented in Tables 1–5.

RESULTS

DISCUSSION

In the present work, we have discussed the potential costs of the first 3 years of mass prescription of lecanemab to patients with MCI per 1,000,000 people under a variety of scenarios. We specifically examined the status quo scenario, a prophylactic scenario, and multiple scenarios of imaging-assisted prognostication with different combinations of neuroimaging modalities at varying levels of sensitivity and specificity, and then compared costs of pharmaceutical intervention under a variety of pricing scenarios.

The status quo scenario of no treatment is the current state of AD therapeutics outside of clinical trials of comparatively small size when compared to the population with MCI. Not surprisingly there is a substantial economic cost involved in the current situation to the individual and to society at large, in addition to unaccounted costs such as the suffering of the patient, their families and friends, and the impact of reduced quality of life (Table 3, Scenario 0). In a prophylactic scenario where all patients with MCI are given treatment with anti-AD pharmaceuticals, we see an exorbitant cost for treatment compounded by the unethical delivery of an unnecessary treatment to approximately 77.7% of the patient population; however, we do also see a substantial drop in the direct and societal costs of care (Table 3, Scenario 1). As we introduce neuroimaging and the use of neuroimaging-driven cognitive prognostication software to complement human specialist physicians (Table 3, Scenarios 2–5), we see this drop in cost remains but does not change significantly compared to the prophylactic scenario. It is noteworthy however that even at very high levels of sensitivity/specificity (Table 3, Scenario 3), the cost of diagnostic neuroimaging and anti-AD is substantially larger than the savings and in fact increases the expense by more than 50% compared to scenario 0. This cost is driven primarily by the price of treatment. With the status quo pricing of lecanemab at $26,500 annually, anti-AD therapies appear to be extremely expensive compared to the cost of no treatment. In this pricing scenario, the costs per positive outcome (preventing conversion from MCI to dementia) are approximately $509,000, $296,000, $258,000, $246,000, and $259,000 respectively for Scenarios 1–5 over a 3-year period. The large differences in cost are primarily accounted for by differences in the specificity of the method of identification of those for whom treatment would be inappropriate since they have no incipient dementia. Because there is a much larger number of patients who will not develop dementia than those who will over our hypothetical 3-year period, it is important to ensure that as few patients as possible as are classified as false positives to reduce the cost of medication and avoid unnecessary treatment.

We see with status quo pricing that none of the available scenarios made treatment economical compared to the cost of no treatment. This observation spurred an exploration of pricing scenarios for anti-AD medications. We presented price adjustment scenarios for both small and large decreases in the cost of anti-AD medications of 20% and 65% respectively. Under either pricing adjustment scenario, we see a substantial decrease in the overall cost of pharmaceutical intervention across the board, but it is only with the larger reduction in price that any diagnostic scenario becomes cost-competitive, or cost-saving, compared to no treatment. We see that even with the highest sensitivity/specificity assumed from the combination of amyloid-PET, MR, and FDG imaging, the cost of intervention is approximately $500,000,000 greater than of no treatment. The use of three imaging modalities is, however, the most expensive imaging scenario presented, and even at lower sensitivity/specificity the substantial change in price from using only FDG (or FDG and MR, which is likely a more realistic scenario) pushes the total cost under that of no treatment and introduces a cost-savings of approximately $3,000,000,000. The costs per positive outcome under such a large price adjustment scenario fall to approximately $252,000, $197,000, $190,000, $177,000, and $181,000 for Scenarios 1–5 (Table 5).

The argument in favor of pricing scenarios is not only motivated by curiosity of where cost of treatment versus status quo becomes economical, but may also be realizable under economies of scale, refining production processes, increases in supply and demand leading to competition between manufacturers and the possibility of single-payer negotiation on a price for anti-AD medications. Furthermore, our assessment assumes only 1 million of the overall population of approximately 5 million people with MCI in the US alone exists as a target group. Demographics worldwide are becoming increasingly skewed towards older adults and represent additional markets which may further reduce cost overall by encouraging global competition and innovation, especially for people in high-income countries with large populations of older adults outside the United States (e.g., countries within the European Union, Canada, South Korea, Japan).

We emphasize again that the present article does not consider the economic value of quality-adjusted life-years or the impact on the productivity and livelihood of unpaid and informal caregivers who are most often family and friends. Nor did we account for the increasing cost of care to the individual or society as dementia increases in severity over time, but rather opted to use estimates for costs associated with very mild and mild dementia since both available anti-AD medications (aducanumab and lecanemab) are only approved for use in patients with MCI or very early dementia due to AD. We further assumed that the cost of MCI diagnostics serve as representative figure from which adjustments may be made; however, this accounts for an extreme minority of the total cost of administering anti-AD pharmaceuticals on such a large-scale, which is dominated by the cost of the drugs themselves. Furthermore, visitation to a memory clinic may not be practical in the foreseeable future due to the relatively small size of this medical specialty. It may become more commonplace for primary care physicians to manage cognitive decline and dementia for those who cannot access or afford the cost of a specialty physician. We also did not account for the cost of treating the side effects of anti-AD medications or the cost of longitudinal diagnostic testing. Considerations such as these and others are beyond the scope of the snapshot of the roll-out for mass adoption of these pharmaceuticals over their first 3 years of availability beyond those enrolled in clinical trials. It is certain that a more comprehensive cost-estimation modeling would show even greater cost-savings compared to the prophylactic scenario and perhaps even break-even with the current status quo of no treatment with some form of more modest cost adjustment scenario.

We also did not consider the use of CSF assays or tau-PET. CSF-based diagnostics are much cheaper than PET studies (approximately the cost of an MR study) and provide information about instantaneous amyloid and tau injury (A and T biomarkers) that is complementary to cerebral glucose metabolism or total brain volume (FDG- and MR- based N biomarkers); however, they are extremely invasive to the patient because they involve puncturing into the subarachnoid space of the vertebral column at the lumbar cistern. These also suffer from some issues of concordance with PET studies and there is significant natural variability in the general population [34, 35]. Tau-PET is a promising technology for the detection and prognostication of cognitive decline, but its use is primarily limited to trials of second-generation tracers that have yet to be approved for clinical use in humans [28]. While first-generation tracers have been approved for use, they suffer from significant off-target binding and are non-specific with respect to isoforms of tau proteins [28]. Regarding imaging, we must explicitly state that the only reasonable scenario with cost-savings compared to that of no treatment is the use of FDG and MR in the case of a large price adjustment, and that we have assumed a sensitivity and specificity that is currently at or beyond the performance of the best available techniques using both FDG and MR, even those based on data-driven analysis (e.g., machine intelligence-based methods) combined with clinical judgement [28]. While this is obviously far from the ideal situation, it is indeed a call for the refinement of these methods, or development of new techniques based on machine intelligence, biomarkers, or newfound clinical measures, spurred by the overwhelming economic argument of saving literally tens of billions of dollars every 3 years.

We stress the important point that we have put forth a simulation study with set hypothetical parameters. The chosen parameters for the price of neuroimaging are based on reasonable expected costs estimated through experience with our institutional imaging program and in consultation with a cognitive neurologist (C.S.). All other costs are taken from examples found in the literature and have been adjusted to 2020 USD to help in comparison [27, 30–32]. It is likely that pricing will vary according to jurisdiction and with the availability of public/private insurance schemes; however, we strove to make our work representative of costs for patients in the US since this is the only jurisdiction where lecanemab is soon to be available to the public. It should also be noted that prices may vary even within one nation due to factors such as accessibility of cognitive/clinical neurologists versus primary care physicians, cost of living in urban versus rural areas, private versus public hospitals and the extent of coverage available with insurance (public or private). These may also in turn be affected by socioeconomic factors that make estimating hypothetical costs associated with the newfound availability of a drug difficult to estimate in a nation as diverse as the US. Further study and more sophisticated analysis which consider the multitude of sociological and economic factors that may affect access to care and cost of care are necessary to further our understanding of the benefits, side-effects, changes to quality of life, changes in physician’s attitudes toward treating dementia and any societal ripple effects associated with public availability of lecanemab and neuroimaging-based identification of candidates for pharmaceutical intervention in the treatment of AD.