Health Outcomes Research of Novel Disease Modifying Medications in Alzheimerâ•Žs Disease and Cost Burden of Early Onset Dementia

The main etiologies of dementia, a neurodegenerative disease, consist of: Alzheimer’s Disease (AD), Vascular Dementia (VD), Frontotemporal Lobar Dementia (FTD), and Lewy Body Dementia (LBD). AD the most common form of dementia is the sixth leading cause of death in the US, where currently 5.3 million Americans are diagnosed with Late-Onset and 95% of cases are 65 years and older. Early-Onset represents the remaining 5% of cases where ages at diagnosis is younger than 65 years. AD is characterized by a progressive loss of neurons with impact on patient cognition, function, and behavior. The 2015 Alzheimer’s Association Report estimated direct and indirect costs of AD and other dementias will reach $226 billion with an expected five-fold increase to $1.1 trillion by the year 2050. With no treatment available that stops, or slows down progression of the disease places the cost estimates of AD and dementia among the most expensive chronic diseases. The next generation of AD medications being investigated will target progression of the disease. Disease-modifying medications (DMMs) are being developed with a mechanism of action directed towards the main hallmarks found in AD patients: the amyloid-beta (Aβ) plaques, and the tau tangles. Tolfenamic acid, a non-steroidal anti-inflammatory (NSAID) drug, is being repurposed in the US as a DMM for AD treatment; human clinical trials still pending. Aducanumab, a monoclonal antibody, binds Aβ and increases its clearance; Phase III human clinical trials are in progress. DMMs are anticipated to improve cognition, function and behavior. The objectives, hypotheses, methods and results of this dissertation follow the manuscript format, and are three fold: Manuscript 1: The objective was to estimate cost-effectiveness of novel disease-modifying medication (DMM) compared to standard medication currently used in the treatment of Alzheimer’s disease. The hypothesis was that the DMM option will show a favorable cost-effectiveness when compared to standard care. Using a Markov Model with a study population comprised of a hypothetical 1000 patients, 65 years and older, we evaluated quality life years (QALYs) gained by the new DMM and an appropriate price to develop a cost-effectiveness framework for the new product. In the Markov model we were able to determine an increase in QALYs when compared to standard of care with a cost value for DMM much higher than current standard care while still showing cost-effectiveness as a new treatment option. Manuscript 2: The objective was to determine affordability to payer’s budget i.e. insurance or hospital upon the introduction of the new cost-effective diseasemodifying medication (DMM) class in treatment of Alzheimer’s disease. The hypothesis was that the introduction of DMM will have minimal budgetary changes to direct costs incurred by payers. Using a 1-year budget impact analysis, a prospective short-term analysis was conducted using Optum ClinformaticsTM Data Mart (January 2010-Decemeber 2012), a large national insurer database with administrative health claims information, with a study population of patients 65 years and older. Two scenarios are to be compared: current mix treatment costs of medications used in Alzheimer’s versus a new mix treatment cost that included the addition of DMM to current mix treatment. The difference in total payer cost of the two scenarios represents the budget impact of the new therapy implementation, allowing us to predicate future cost of new treatment mix. The study estimated a total per-memberper-month (PMPM) treatment cost preand postintroduction of DMM that would be affordable to payer’s and recommended to be added to formulary. Manuscript 3: The objective was to describe prevalence, incidence, and direct total cost predictors associated with Early-Onset Dementia (EOD) and its etiologies. The hypothesis was that Alzheimer’s disease would be main predicator of overall EOD direct cost. We conducted a retrospective cohort study using Optum ClinformaticsTM Data Mart (January 2010-Decemeber 2012), a large national insurer database with administrative health claims information, with a study population of patients 21-64 years and older. Total cost components include: physician visits, hospital visits, nursing home care, and prescription drugs associated with EOD treatment. Using a Generalized Linear Model (GLM) to assess the relationships between total cost and the covariates of interest, we identified age, geographical regions, EOD subtypes, and comorbidities as total cost predictors of EOD.

The cost-effectiveness of novel DMM in Alzheimer's patients will depend on medication pricing and prolonging time spent in the community against early entry to nursing home care.

Introduction
Alzheimer's disease (AD) affects 5.3 million Americans with 95% of patients being 65 years or older, two-thirds being female, and with early symptoms showing 5 to 10 years prior to official diagnosis. [1][2][3] The prevalence of AD is projected to increase by 40% in the next 10 years reaching an estimated 14 million patients by the year 2050. [1][2][3] The incidence of AD increases with age and doubles every 5 years after the age of 65. [1][2][3][4] In 2015 a projected 700,000 people aged 65 or older will have AD as cause of death, and although deaths from other major chronic diseases (e.g. HIV) have experienced drastic declines it was reported that between 2000-2012 there was a 68% increase in cause of death attributed to AD. 5,6 The most common cause of death in patients with AD is pneumonia. 5,6 The 2015 Alzheimer's Association report estimated direct and indirect costs of AD and other dementias will reach $226 billion with an expected five-fold increase to $1.1 trillion by the year 2050. These estimates place AD and dementia as one of the most expensive chronic diseases. Current AD medication cost to payers range between $2500 and $3500 annually. 7,8 AD progression is characterized by the accumulation of amyloid-beta (Aβ) plaques and tau tangles leading to neuronal death and loss. [9][10][11] The amyloid precursor protein (APP) is cleaved via enzymes into the Aβ peptides which then aggregate, deposit into plaques, and initiate the pathology of AD. [9][10][11] In addition to the Aβ plaques tau protein is hyperphosphorylated resulting in aggregation and tau tangle formations initiating AD pathology. [12][13][14] These processes are the leading causes of decline in cognitive function, behavior, and performance of daily activities in patients with AD. [12][13][14] The mechanism of action of current AD medications do not target either of the previous mentioned triggers of AD development and progression. 8,15,16 The next generation of AD medications being investigated will target progression of the disease. [15][16][17] Disease-modifying medications (DMMs) are being developed with a mechanism of action directed towards the main hallmarks found in AD patients: the amyloid-beta (Aβ) plaques, and the tau tangles. 15,16 Tolfenamic acid, a non-steroidal anti-inflammatory (NSAID) drug, is being repurposed in the US as a DMM for AD treatment; human clinical trials still pending. 15,16 Monoclonal antibody drug class, e.g. Aducanumab along with other anti-Aβ trial drugs, target Aβ and increases its clearance; Phase III human clinical trials are in progress for some of these new drug therapies. [17][18][19] Majority of these new therapies will be targeting the early mild stages of AD, anticipating improvement in cognition, function and behavior. [17][18][19] Early detection of AD and providing DMM therapy has potential of limiting AD progression, spending more time in the earlier less severe and costly stages of disease and increasing time in the community. 20,21 Predicting the efficacy of novel DMM for AD will exceed current treatment options, certainly these medications will be priced at a premium. We assessed the value of DMM by conducting a cost-effectiveness analysis. We used Markov modeling to simulate the anticipated states of disease progression, and model patients with AD transitioning through the different health states predicting clinical outcome and cost of DMM therapy.

Methods
This study design was a Markov model utilizing published clinical data and registry databases as primary data source for: cost, utility values, transition probabilities, and medication efficacy converted to relative risk reduction. The model began with a hypothetical 1000 patient population distribution assigned to the initial MCI health state and simulates patient movement through the progressive states using transition probabilities. [20][21][22]34 The transition probabilities defined the rate of progression through the health states; for example a patient in moderate AD can remain in that same state, move to severe state, or enter the absorbing state of death which they do not exit. [20][21][22]34 No movement backwards to a less severe state was permitted.
The simulated patient cohort was assigned transition probabilities based on

Results
Diagram representation of the Markov model ( In the base case ( One-way sensitivity analyses were conducted for our model parameters to determine the variables with most impact on our results. Key input parameters were varied one input at a time while holding other constant at their base-estimates.
Sensitivity analyses was conducted on costs, probabilities, and utilities.
A tornado diagram ( Figure 6 Table   2. Using the WTP of $100,000 our DMM produced an ICER of $89,812/QALY with 92% of the iterations showing cost-effectiveness when compared to 8% in SC. ( Figure   4). The cost-effectiveness acceptability (CEA) curve also illustrates the probability that DMM therapy will be cost-effective at varying WTP thresholds for a patient ( Figure 5). increase, respectively.

Discussion
Our study demonstrated that a disease-modifying medication (DMM) delaying the cognitive decline, increasing time spent in the less severe and costly health states of Alzheimer's disease (AD) can possibly be cost-effective when compared to current medication therapies. This study can provide a framework or working design that estimates the assumed and possible effect of these newly proposed novel DMM that are currently in clinical trial pipelines. The study used published data to build the decision, but with deficiencies in direct evidence on DMM effects or possible costs associated with the new therapy. Currently there are several anti-amyloid beta monoclonal antibodies in the pipeline of companies, with some of the more recent trials reaching Phase III but unfortunately not being able to show progress. [17][18][19]36,44 Aducanumab, a monoclonal antibody, is currently the most promising of the pipeline products showing credible progress in Phase III trials in the early less several AD health states 18,19 i.e. MCI and Mild AD. Given that no such DMM is on the market yet we had to make an assumption of expected price, using sensitivity analysis to determine with all our input parameters and WTP level that a price as high as $23,189 per year will maintain cost-effectiveness favorability of new therapy. In similar manner, we estimated that a relative risk reduction of 0.546 would maintain cost- An aspect of our study to consider is that using DMM patients are expected to spend more time in the less severe states spending more time in the community setting, and if considering a societal prospective then we would predict a delay in nursing home placement and its cost but at the expense of increased unpaid family caregivers to aid with the patients. 22 Our model focuses primarily of direct cost of care, when the biggest factor contributing to the rising cost of AD are the indirect costs; costs associated with unpaid family caregivers, loss of productivity associated with both patients and caregivers. 22,45 When considering our study using a payer e.g. Insurance perspective we should be cautious when trying to generalize it to an insurer or a managed care plan. Given the uncertainty in relative risk reduction proposed for these DMM still in clinical trials, the utility values based on current AD status which we would expect to change with the slowing down of cognitive decline, and the cost information based on several older studies we are restrained when interpreting our results.

Limitations
This study provides a look into complexity of trying to model a chronic condition such as AD. All the assumptions, inputs, and sensitivity analysis are to be considered carefully given the lack of current DMM efficacy, cost, or impact on utility values. Our basic assumption was that DMM will work particularly well in the early less severe stages of AD, despite the several setbacks in clinical trials during the past years but with still compounds being developed in the pipeline. The direct costs our study used are dependent on the U.S. health system and might not be generalizable to other countries given the wide range of different health insurance systems available.
The utility values (QALYs) were not validated in any trial using DMM so we have no information on how the DMM will affect utility values or how they will change as disease progresses. Side effects and their costs are usually part of cost-effectiveness calculations, however with DMM still in clinical trial phases we yet to have access to such information and being a simulation no such information was added to our model.

Conclusion
Our model results, using available data, showed that future disease-modifying medications in Alzheimer's disease may be cost-effective when adopting a willingness-to-pay of $100,000, and adding additional quality-of life year gained to patients. We conclude that at cost of $23,000 or less and relative risk reduction of 0

. Cost-Effectiveness (CE) Acceptability Curve Represents Probability that a Treatment will be CE (percentage iterations (y-axis) for which treatment was CE) at varying Willingness-to-Pay Thresholds (x-axis). The amount, in dollars, payer willing to pay to pay to achieve an additional quality-adjusted life year
DMM -Disease-Modifying Medication (blue) SC -Standard of Care (orange)  Using administrative claims database to estimate AD medication cost and utilization, the introduction of novel DMM therapy will have a substantial effect on both the total and PMPM cost for AD medication budget. Given the importance of affordability of new treatments to decision and policy-makers greater attention and planning needs to be afforded to the expected sizable change predicted to occur with introduction of new novel DMM therapy.

Introduction
Currently there is no cure for Alzheimer's disease (AD), guidelines by both the American College of Physicians and the American Academy of Family Physicians divide treatment for AD into 2 medication categories 1,2 : Acetylcholinesterase Inhibitors (AChEIs) e.g. Donepezil and N-methyl-D-aspartate (NMDA) e.g.
Memantine. The AChEIs prevent the breakdown of acetylcholine, a neurotransmitter required for neuronal function, and help increase levels of the declining neurotransmitter due to neuronal loss. [1][2][3] The NMDAs reduce levels of glutamate receptor activation and decrease neuronal dysfunction. [1][2][3] The use of these therapeutic agents as monotherapies or in combination during various stages of AD improve cognition and daily functioning scale score slightly, but do not slow down progression, decline in cognition, or cure AD. [3][4][5] AD progression is characterized by the accumulation of amyloid-beta (Aβ) plaques and tau tangles leading to neuronal death and loss. 6,7 These hallmarks, Aβ and tau tangles, are the targets of new investigational disease-modifying medication (DMM), currently undergoing clinical trials, trying to slow AD progression and reduce cognitive decline. 4,5,8,9 Monoclonal antibody drug class, e.g. Aducanumab along with other anti-Aβ trial drugs, target Aβ and increases its clearance; Phase III human clinical trials are in progress for several of these new drug therapies. 5,9,10 Majority of these new therapies will be targeting the early mild stages of AD, anticipating improvement in cognition, function and behavior. 5,9,10 With the efficacy of proposed DMM monoclonal anti-body therapy expected to target the underlining causes of AD, these DMM are expected to be priced at a premium. Current AD medications, AChEIs and NMDAs, cost to payers range between $2500 and above $3500 annually. With limited resources, decision makers have to determine the future budgetary impact with the addition of new therapies, and judge competing treatments on both clinical and cost effectiveness.

Methods
This study design was a 1-year retrospective before and after budget impact which are mainly of the monoclonal drug class, are usually 5 times higher than current medication treatments. 12 Average cost of monoclonal antibodies currently on the market, in other disease states, were also considered to aid in the assumption as to the cost of these proposed new therapies in AD. Using this rationale the study assumed a cost of $24,000/year or $2,000 for a monthly prescription. We can assume a single prescription a month for 12-months of the new DMM therapy to determine the annual total cost with a per prescription cost of $2,000 that was previously assumed.

Results
The    Figure 2 summarizes the PMPM findings of our one-way analyses using a tornado diagram, ranking the parameters from most to least influence on our model and overall outcome of total cost and PMPM. When the price of DMM was increased by 50% the total cost increased by 42% (or by $0.068 PMPM). When percentage of patients using AD medications was increased by 50% total cost increased by 39% (or by $0.0637 PMPM). When new cases utilizing DMM was increased by 50% the total cost increased by 21% (or by $0.034 PMPM). When percentage of patients switching to DMM was increased by 50% the total cost increased by 19% (or by $0.0304 PMPM). When percentage of patients adding DMM was increased by 50% the total cost increased by 10.5% (or by $0.017 PMPM).

Discussion
Alzheimer's disease (AD) has become one of the most financially taxing chronic diseases on individuals, the health care system, and society. 15 have an increased, decreased, or no effect on budgets once DMM is introduced. 20-22 Most health plans medication benefits range between $25 to $35 PMPM. 20 But, an acceptable increase in PMPM for the implementation of a new therapy or drug to be considered affordable for health plans is between 0.5%-1%, per the Institute for Clinical and Economic Review (ICER) framework. In our study the increase in overall cost was significant but when considering the PMPM impact on medication benefits, will be within the acceptable 0.5%-1% range when using a $35 PMPM bench mark. In terms of the substantial increase in overall cost, given the disease-modifying and slowing down of disease progression properties associated with these new medications can lead us to assume or expect; that an increase in the drug expenditure associated with a DMM therapy may be offset by some of the savings resulting from effect of drug in delaying institutionalized high costs. i.e. nursing homes.
In determining our model we decided to focus on AD prescription medication cost and assumptions on the switching, and adding of DMM to future AD population.
The study's focus was not to determine or estimate any off-setting costs or measure the impact of other cost sources such as physician visits, hospital visits, or long-term facility care. DMM therapy is expected to improve patients' cognition, function, and behavior but as to the implications that might have on utilization cost of other medical services remains to be determined once the new therapy reaches the market. 4,5,19,23,24 Budget impact models are essential methods used by health plans, and Further, we know that DMM specifically target patients in the early less severe stages of AD and an assessment of disease severity could not be obtained from this database.
This might be essential in determining a more accurate representation on who would be more eligible to use the new therapy. Being a commercial based health care plan our results might not be generalized to government or other non-private health care plans.

Conclusion
This study, to the best of our knowledge, is first to predict and quantify a budget impact for a proposed disease-modifying medication for the treatment of

Abstract
Our purpose was to estimate direct cost of care for patients newly diagnosed with Early-Onset Dementia (EOD), age 21-64 years old, and evaluating predictors on total cost using a Generalized Linear Model (GLM

Introduction
Dementia has become a public health concern given the U.S. aging population, and although often associated with old age few studies estimate 220,000 to 640,000 of and personal responsibilities when the disease strikes them. [3][4][5] The burden of disease is both personal and economic, impacting the patients, their families, communities, and the health care system. The literature is scarce regarding the prevalence and incidence of EOD, and despite the high financial implication of dementia most of the analysis of cost burden have paid little attention to EOD compared to dementia with age of onset 65 or older. [3][4][5][6] This study aimed to provide information describing the direct cost of care associated with EOD and how patient clinical and demographic characteristics impact that total cost.

Methods
The primary data source was the Optum Clinformatics™ Data Mart, a large national insurer database with administrative health claims information for approximately 19 million patients, collected from January 2010 to December 2012.
Available information about these patients with private insurance included: member demographics, medical claims, pharmacy claims, lab results, and inpatient claims.
The study population was comprised of patients with only a single subtype Patients that recorded a diagnosis of dementia that was less than 365 days were captured. This helped us make the assumption that the patients did not come into the cohort with the diagnosis previously.
Two analytic samples were created comprising patients with EOD as primary, secondary or any other diagnosis, and patients having EOD as their primary diagnosis only. EOD as primary diagnosis only was used for cost analysis (i.e. expenses for services associated with EOD care) and descriptive characteristics, while the group identified with EOD codes as primary or non-primary diagnoses was also analyzed to describe characteristics of the overall EOD population. (Figure 1). Study outcome (dependent variable) was total annual per-patient cost for all health care service use associated with care for EOD, with independent variables represented by: age, gender, EOD subtypes, geography, and comorbidities. Differences in mean total costs between the variable categories were assessed using an ANOVA test. While evaluating the relationship between total cost and covariates of interest, and because of the skewed nature of cost, we utilized a Generalized Linear Model (GLM). [8][9][10] Predictors of cost were determined using the GLM using a forward predictive approach, and implementing Modified Park test to determine dependent variable distribution. 8-10 GLM was best method for the model using a "link" and "family" function due to skewed nature of the cost variable, and since log transformation could not achieve normality using Ordinary Least-Squares (OLS). [8][9][10][11] We used the link function to specify the relationship between the mean of our outcome variable, total cost, and predictors. [8][9][10][11] While family function which corresponds to the distribution of data which in this case was a Gamma distribution informed by the Modified Park Test. [8][9][10][11] For all analyses, statistical significance was considered a 2-sided P-value <0.05. All statistical analyses were performed using SAS version 9.4 (SAS Institute, Cary, NC). This study was reviewed and approved as exempt by the University of Rhode Island's Institutional Review Board.

Results
The cohort sample was created after applying inclusion and exclusion criteria ( Figure 1): patients with primary diagnosis of any subtype of dementia (n=2,150) were used for the descriptive statistics presented in Tables 1 and 2, health service utilization (Table 3), and used for the development of the GLM used for our cost analysis.
We initially identified 42,226 patients, from total number of 19 million patients with approximately 2 in 1,000 prevalence, having at least one primary or non-primary diagnosis for dementia ( Figure 1); from which 13,185 patients were excluded for not being continuously enrolled for a minimum of 18 months. A further 21,120 patients were excluded for being older than 64 years of age or younger than 21 years of age.
An additional 96 patients were excluded for having a cancer diagnosis.
In the sample frequency demographic (n=2150) of newly diagnosed EOD patients in our In Table 3  Dementia patients with 1 or more comorbidities can expect an increased overall cost of care, with cardiac comorbidities being the greatest contributors to cost.

Discussion
In this study, we estimated the total direct mean cost for disease specific patients with Early-Onset Dementia (EOD), and determined patients' clinical and demographic predictors of the overall care cost. The purpose of our study was to provide a source of cost information for an underrepresented population in studies conducted in both cost, and epidemiology fields. 1 Nursing Home care costs were included in addition to the outpatient, inpatient, and medication costs. 18,19 Our study focused on the total cost increases of the medical care settings that occurred in patients with specific clinical and demographic characteristics. In this 12 month period higher costs were associated with ages 43-53 years old, the Midwest region, and specific comorbidities. Dementia subtype patients were diagnosed with were not a predictor of cost, and neither was gender.
The majority, approximately 75%, of new cases with EOD as a primary diagnosis in our patient population, were diagnosed as either Mild Cognitive Impairment (MCI) or DNOS and 10% having an Alzheimer's disease (AD) diagnosis.
In this study the variable of Gender showed no statistically significant difference in terms of frequency of medical care setting utilization or their cost, or in frequencies among the various dementia subtypes except in AD were more females (60%) than males (40%) where diagnosed (P-value=0.05).
In this study we showed a direct relationship between increased cost of care in EOD patients per number comorbidities. We included each of the captured comorbidities into our model to estimate the ones associated with higher costs. Results We found cost differences across Geographical regions, attributed to differences seen in terms of care setting utilizations and cost among dementia patients across the US. [20][21][22][23] Patients in the Midwest showed the highest cost when compared to the other three regions of the country. These variations in cost across regions may be due to regional pricing differences, availability of services, socioeconomic differences, ethnic and racial differences which would need further examination 16,[20][21][22][23] .
We found the most frequent EOD was either MCI or DNOS, but with none of the types being a predictor of cost in our overall model. Expectations was that AD would account for majority of the diagnosis given that the few estimations found in the literature show that AD would account for about 30% of all EOD. 3,5,6,23 In addition most of the cost analyses done specifically target dementia and AD in ages 65 and older making it difficult to have comparable values in EOD patients, and relying on the assumption and the expectation that costs seen in EOD, and specifically AD, mirror what is produced in the older patient groups.

Limitations
There are several limitations in our study. The database did not provide information on race, education level, income or marital status, and these factors may have been associated with cost differences. The cost information collected was total direct cost for dementia care, and not incremental costs associated with development of the disease. In order to produce the incremental cost we would have to develop a non-EOD patient group with similar characteristics of those of our EOD to show the increasing or decreasing cost between developing EOD and not having EOD.
However, this was not the main concern of the study as we were trying to provide a descriptive informative review on direct costs associated with services for EOD and its subtypes. Being a commercial based health care plan our results might not be generalized to government or other non-private health care plans.

Conclusion
Results of our study provided needed information for the direct cost in Early-Onset Dementia patients. The study highlighted the significant variation of cost estimates for the different care settings of interest, and producing a total mean direct cost of care $10,932 (SD=$27,612) per patient. The study in addition provided an understanding of predictors associated with higher cost in the Early-Onset Dementia population; where patients with specific comorbidities were associated with increased cost, patients 43-53 years old recorded higher mean cost, and patients living in the Midwest region of the U.S. seen with higher costs. In providing these type of cost-ofillness studies decision makers are more informed as to the implications, and distribution of these direct cost on the Early-Onset Dementia population.