RESEARCH ARTICLE

Assessing the cost-effectiveness of the rivastigmine transdermal patch for Alzheimers disease in the UK using MMSE- and ADL-based models
Balazs Nagy 1, Alan Brennan 1, Agnes Brandtmuller 1, Simu K. Thomas 2, Sean D. Sullivan 3 and Ron Akehurst 1
1 2

School of Health and Related Research, University of Shefeld, Shefeld, England Novartis Pharmaceuticals Corporation, East Hanover, NJ, USA 3 Pharmaceutical Outcomes Research and Policy Program, University of Washington, Seattle, WA, USA Correspondence to: B. Nagy, E-mail: b.nagy@sheffield.ac.uk

Objective: Assess long-term cost-effectiveness of rivastigmine patch in Alzheimers disease (AD) management in the UK, using cognitive and functional models based on clinical trial efcacy data. Methods: Incremental costs and Quality Adjusted Life Years (QALYs) associated with rivastigmine patch and capsule treatment versus best supportive care (BSC) were calculated using two economic models, one based solely on Mini-Mental State Examination (MMSE) scores, and one also incorporating activities of daily living (ADL) scores. The clinical pathway was populated with data from a clinical trial of rivastigmine patch (9.5 mg/24 h) and capsules (12 mg/day) versus placebo. Costs were based on the UK health and social care costs and basic UK National Health Service (NHS) prices. Disease progression was modelled beyond the trial period over 5 years using published equations to predict natural decline in AD patients. Base case costing variables included drugs, clinical monitoring, and institutionalization. Results: The MMSE model estimated incremental costs per QALY of 10 579 for rivastigmine patch and 15 154 for capsule versus BSC. The MMSE-ADL model estimated incremental costs per QALY of 9114 for rivastigmine patch and 13 758 for capsules. The main difference between the models was a greater number of institutionalized days avoided for rivastigmine versus BSC estimated by the MMSE-ADL model. Conclusions: Both the MMSE and MMSE-ADL models suggest that rivastigmine patch and capsules are cost-effective treatments versus BSC. Incorporating ADL evidence makes a marginal but important difference to estimates in this case. Future economic evaluations of AD treatment should include measures of both cognition and functioning. Copyright # 2010 John Wiley & Sons, Ltd. Key words: Alzheimers disease; rivastigmine; transdermal; cost-effectiveness; NHS History: Received 12 November 2009; Accepted 16 April 2010; Published online 15 September 2010 in Wiley Online Library (wileyonlinelibrary.com). DOI: 10.1002/gps.2551

Introduction An estimated 24.3 million people worldwide suffer from dementia, of which the most prevalent form is Alzheimers disease (AD) (Ferri et al., 2005). There is currently no cure for AD, although pharmacological therapies may provide symptomatic relief. Social support, home nursing, personal care, community care and nursing homes provide the mainstay of nonCopyright # 2010 John Wiley & Sons, Ltd.

pharmacological management (Cummings et al., 2002). The annual per-patient cost of AD in UK is estimated as 16 689 for a person with mild dementia living in the community, 37 473 for a person with severe dementia living in the community and 31 296 for an institutionalized patient (Knapp and Prince, 2007). These costs are expected to rise as the elderly population grows, and the incidence of AD increases (Ferri et al., 2005).
Int J Geriatr Psychiatry 2011; 26: 483494.

484

B. Nagy et al.

Cholinesterase inhibitors (ChEIs) are available in oral formulations for the symptomatic treatment of AD. In 2007, the rst transdermal therapy, rivastigmine patch, was approved in Europe for the treatment of mild-to-moderate AD. The once-daily rivastigmine patch provides comparable exposure to the highest dose of rivastigmine capsules (12 mg/day) (Mercier et al., 2007) and improved tolerability, with three-times fewer reports of nausea and vomiting compared with older oral treatment (Winblad et al., 2007a). Numerous studies have estimated the health economic impact of ChEI treatment (Stewart 1997; Stewart et al., 1998; Fenn and Gray, 1999; Jonsson et al., 1999; Neumann et al., 1999; OBrien et al., 1999; Hauber et al., 2000a, 2000b; Gareld et al., 2002; Ikeda et al., 2002; Migliaccio-Walle et al., 2003; Wimo et al., 2003; Loveman et al., 2006). Most previous economic evaluations of rivastigmine indicate potential cost savings compared with standard care, due to delays in institutionalization that offset the costs of drug treatment (Hauber et al., 2000a, 2000b). The UKs National Institute for Clinical Excellence (NICE) was established in 1999 to provide guidance on the clinical- and cost-effectiveness of new and existing interventions to the National Health Service (NHS) of England and Wales. In August 2007, NICE recommended ChEI therapy for the treatment of moderate-to-moderately severe AD (Mini-Mental State Examination (MMSE) scores of 1020). The current standard outcome measure for modelling clinical outcomes is the MMSE (Folstein et al., 1975). Recently, however, experts have suggested that measures of cognitive decline alone may not accurately reect progression of the aspects of AD that determine probability of institutionalization (Bavazzano et al., 1998; Bowie et al., 1999; Wolstenholme et al., 2002; Green, 2007). Behavioural and functional deterioration (e.g. loss of ability to perform routine tasks) are considered important contributing factors to determining whether patients require nursing home care (Desai et al., 2004). Indeed, recent European Medicines Agency guidelines highlight the importance of assessments of functional outcomes, recommending their inclusion as a co-primary endpoint (with assessment of cognition) in all clinical trials of AD treatments (European Medicines Agency, 2008). In conjunction with cognitive measures such as MMSE, measures of executive and global function, such as activities of daily living (ADL) assessments, may provide a more accurate prediction of time to institutionalization than MMSE alone (Hatoum et al., 2009).
Copyright # 2010 John Wiley & Sons, Ltd.

The objective of this study was to examine the longterm cost-effectiveness of rivastigmine patch and capsules based on UK health and social care costs, using a conventional MMSE-based cost-effectiveness model, and an additional economic model that also incorporates ADL data. Methods
Clinical pathway models

This study utilizes two cost-effectiveness models: one based on MMSE data alone (MMSE model), and a revised model also incorporating ADL data (MMSEADL model). Depending on the model, either MMSE or both MMSE and ADL scores were used to estimate the number of patients requiring institutionalization. MMSE scores were also used to quantify impact on health-related quality of life (HRQoL).
Data sources

Disease progression was modelled based on MMSE and Alzheimers Disease Cooperative Study-ADL (ADCS-ADL) scores from the pivotal IDEAL (Investigation of transDermal Exelon in ALzheimers disease) study (Winblad et al., 2007a; Grossberg et al., 2009). This 24-week, double-blind, doubledummy, placebo- and active-controlled trial with a 28-week open-label extension compared the efcacy and safety of rivastigmine patch (9.5 mg/24 h) with rivastigmine capsules (12 mg/day) and placebo in patients with mild-to-moderate AD (Winblad et al., 2007a; Grossberg et al., 2009). A 17.4 mg/24 h patch was also evaluated, but was not considered here as it is not an approved dose. The clinical pathways for this analysis were populated with data from the IDEAL 24-week double-blind study combined with data from a further 28-week open-label extension study (rivastigmine patch n 250, rivastigmine capsule n 256). BSC was modelled using the placebo arm (n 281) of the 24-week double-blind study.
Modelling of disease progression

To model the clinical pathway, patients were categorized over time as being on active rivastigmine treatment or discontinuing treatment and reverting to BSC. Patients also had a probability of death in each
Int J Geriatr Psychiatry 2011; 26: 483494.

Rivastigmine patch cost-utility in the UK

485

6-month time period. The IDEAL study provided 6and 12-month data for placebo and rivastigmine, respectively. As in previous cost-effectiveness analyses of rivastigmine in AD (Fenn and Gray, 1999; Hauber et al., 2000a, 2000b), MMSE scores were used to model disease progression beyond the clinical trial observation period. Longer-term MMSE outcomes were quantied using a published mathematical model (Equation 1) based on a longitudinal study of 719 untreated AD patients followed for up to 7 years (Mendiondo et al., 2000). This model links an individuals expected rate of worsening on the MMSE scale to their previous time period score, and was employed to predict MMSE decline after a patients last measured outcome. An iterative Mendiondo solution (6-month interval) was applied to the Intent-to-Treat Last Observation Carried Forward (ITT-LOCF) population in the double-blind phase of the IDEAL study, and the ITT population in the open-label extension phase. Corrections were made for potential responder bias for early discontinuations.
Time to 1-point MMSE decline years 0:00334 MMSE2 0:073MMSE 0:6013 (1)

NHS nursing homes). Linear regression of these data with MMSE scores derived Equation (2).
Probability of institutionalization 0:512 0:016 MMSE R2 0:711 (2)

Linking ADCS-ADL data to probability of institutionalization

Discontinuation and mortality

The MMSE-ADL model incorporates ADL data into the prediction of probability of institutionalization using a published relationship (McNamee et al., 2001) between probability of institutionalization and the TownsendADL (Townsend, 1979) and MMSE scales. Although the IDEAL study utilized the ADCS-ADL scale, no evidence is available on the relationship between this scale and probability of institutionalization. Therefore, ADCSADL data were translated into Townsend-ADL data by mapping these two scales together. The mapping procedure was qualitative: corresponding items on the two scales were identied, and severity categories were paired. Patient-level item responses on the ADCS-ADL were translated to the Townsend-ADL scale, giving each patient a total estimated Townsend-ADL score (Rozsa, 2007). Equation (3) was then applied to predict the probability of institutionalization using ADL (McNamee et al., 2001).
Probability of institutionalization

During the IDEAL study open-label extension a discontinuation rate of 19.1% every 6 months was reported (Grossberg et al., 2009). Discontinuing patients were presumed to immediately revert to the MMSE score expected if they had been on the BSC arm from the start of the study. Patients with MMSE scores <10 were assumed to withdraw from treatment, following current NICE guidelines on ChEI therapy (NICE, 2006a). As mortality rates in the IDEAL study were unusually low compared with previous epidemiological studies, an AD-specic annual mortality rate of 11.21% (Martin et al., 1987) was applied.
Probability of institutionalization

1 1 exp1:48 0:11 MMSE 0:07 Townsend-ADL

1

(3)

Estimation of long-term ADL scores

Logistic and linear regression analyses were performed to establish the relationship between total MMSE and Townsend-ADL scores, based on IDEAL study data. The relationship was best characterized with a simple linear regression using data from all patients in the open-label ITT population, including baseline visit data (Equation 4).
Townsend-ADL 13:214 0:438 MMSE R2 0:216; coefficients of the regression were significant; p < 0:001 4

In the MMSE-based model, probability of institutionalization was evaluated solely using MMSE scores. A detailed UK-based study (Stewart, 1997) provided residential status data (numbers of people living alone/with others in private households, or in
Copyright # 2010 John Wiley & Sons, Ltd.

In the absence of long-term ADL data, this function was applied to the estimated long-term MMSE
Int J Geriatr Psychiatry 2011; 26: 483494.

486

B. Nagy et al.

progression data for rivastigmine to estimate Townsend-ADL scores beyond the study period.
Validation of ADCS-ADL disease progression models

MMSE and ADCS-ADL data from the rivastigmine arm (n 495) of a 2-year, randomized, clinical trial comparing rivastigmine and donepezil in AD (Bullock et al., 2005) were used to validate the extrapolations from the IDEAL study. Linear regression using these data to describe the relationship between MMSE and ADCS-ADL resulted in a slope of 2.0670 0.0415, very similar to that arising from our modelling of the IDEAL dataset (2.0645 0.0298).
Utility values

included in the sensitivity analyses, based on Bell et al.s (2001) estimation of caregiver time stratied by type of setting and disease severity (Bell et al., 2001) and multiplied by the UK minimum wage (5.52 per hour). Drug acquisition costs of 1020 per annum were based on basic NHS prices (excluding VAT) (2009). Monitoring costs of 264 per annum (UK NHS reference costs, 2003) were included, assuming two outpatient visits per patient per year (NICE, 2006a). Discontinuing patients were assumed to incur drug/ monitoring costs for a further 3 months, but receive no further utility benet. All costs were updated to 2008 to compensate for ination (Curtis, 2008) and future costs and health benets were discounted by 3.5% following UK Treasury Department recommendations (NICE, 2004). All events (e.g. discontinuation) were assumed to occur halfway through each 6-month cycle.
Incremental cost-effectiveness ratios and net benefit

Utility values for both the MMSE and MMSE-ADL models were calculated using existing evidence correlating HRQoL to MMSE score (Brazier, 2001). Clinical trial data (Corey-Bloom et al., 1998; Rosler et al., 1999) were mapped to health status, as measured by the Health Utilities Index Version 3 (Furlong and Feeney, 1998). A regression function was derived, indicating that a 1-point deterioration in MMSE is equivalent to a reduction of approximately 0.03 utility (Equation 5) (Brazier, 2001). Quality-Adjusted Life Years (QALYs) were calculated by applying this function to the mean MMSE score for each cycle of observation, and multiplying by 0.5 to compensate for the 6-month cycle length.
Utility 0:0982 0:0298 MMSE R2 0:52 (5)

Following NICE guidance (NICE, 2008), this analysis estimates QALYs and costs for each treatment and compares the incremental cost per QALY gained between two treatments. A treatment might be considered cost-effective if the incremental cost per QALY gained is lower than a threshold willingness-topay (lambda) of 20 000 per QALY. Incremental net benet can be calculated as lambda incremental QALY incremental cost. A treatment is considered cost-effective relative to its comparator if incremental net benet is >0.
Sensitivity analyses

Calculation of cost

The cost of institutionalization in the base-case analysis (for both models) was represented by the health and social care cost of dementia patients living in supported accommodation in the UK. The sensitivity analyses considered other types of institutions. Institutionalization costs (29 948 per year; Personal Social Services Research Unit, University of Kent, UK, 2007) and standard community care costs (Community Care package for older people, low cost: 14 300 per year; Curtis and Netten, 2006) were applied consistently, irrespective of disease severity and type of treatment (Netten et al., 2001; Wolstenholme et al., 2002). Informal care costs, in line with the recommendations of the 2006 NICE analysis (Loveman et al., 2006), were
Copyright # 2010 John Wiley & Sons, Ltd.

One-way sensitivity analyses examined the effect of different sources of evidence on the MMSE-based probabilities for institutionalization, the impact of incorporating informal care, and the application of linear and logistic regressions to elucidate the relationship between ADL and MMSE using various populations of the IDEAL dataset (e.g. open-label/double-blind). Probabilistic sensitivity analyses investigated the probability that rivastigmine patch produces greater net benet compared with BSC and rivastigmine capsules over a range of willingness-to-pay thresholds. Uncertainty surrounding all model parameters was described by probability distributions (Table 1), and propagated through the model using Monte Carlo sampling techniques to produce distributions of expected costs and QALYs for each treatment group. Uncertainty surrounding the proportion of people at
Int J Geriatr Psychiatry 2011; 26: 483494.

Table 1 Uncertainty in model parameters used for probabilistic sensitivity analysesa Distribution Beta Number of observations Beta 10.1 33 048 937 15 848 523 0.0982 0.0298 0.5116 0.0160 Constant Slope 132 140 0.4380 Constant Slope Constant 0.011471415 0.00017999 Constant 4.317357935 0.005360524 Constant Slope 0.000157854 0.00000749 Constant Slope 0.00000038 Slope 0.000009 Slope 0.000324 Source 26 17.7 0.3 17.8 0.3 16.6 0.4 15.0 0.4 17.1 0.4 15.2 0.4 13.3 0.4 17.3 0.4 15.8 0.4 52 78 104 14.2 0.5 13.5 0.4 11.6 0.4 130 12.5 0.5 11.7 0.5 9.9 0.4 156 10.8 0.5 10.0 0.5 8.4 0.4 182 9.2 0.5 8.4 0.5 6.9 0.4 208 7.7 0.5 6.9 0.5 5.5 0.4 234 6.4 0.5 5.5 0.5 4.4 0.4 260 5.2 0.5 4.3 0.4 3.5 0.4 Authors Stewart, 1997, Authors PSSRU, 2007 PSSRU, 2006 ScHARR Normal Normal Multivariate normal Mean (slope) Multivariate normal Mean (slope) Multivariate normal Mean (slope) Mean (constant) Mean (constant) Mean (constant) Mean Standard error Mean Standard error 202 Alpha 141.4 Parameter Value Covariance Matrix Source Martin et al., 1987

Model parameter

Death rate

Rivastigmine patch cost-utility in the UK

Institutional cost

Copyright # 2010 John Wiley & Sons, Ltd.

Parameter Week Mean (patch) Standard error Mean (capsule) Standard error Mean (BSC) Standard error 16.7 0.2 16.7 0.2 16.7 0.2 0
b

Community cost

Utilities

Probability of institutionalization

Townsend-ADL MMSE relationship

Model parameter

Distribution

Value

MMSE score

Normal

Winblad et al.,2007a, Novartis (data on le), and patient-level simulation with Mendiondo equation

Int J Geriatr Psychiatry 2011; 26: 483494.

Number of iterations for the probabilistic sensitivity analyses 1000. Patient-level baseline parameters were adjusted to rivastigmine patch to allow all model populations to start from same average MMSE score. To model patch and capsule a combination of the IDEAL 24-week double-blind and the 28-week open-label patient-level data was used until Week 52. To model BSC the 24-week double-blind data were extrapolated from Week 24 using the Mendiondo equation. PSSRU, Personal Social Services Research Unit, University of Kent, UK; ScHaRR, School of Health and Related Research, University of Shefeld, UK.

487

488

B. Nagy et al.

each MMSE score was characterized using Dirichlet distributions based on the number and percentage of patients attributed to each MMSE score in each period from the disease progression model. Normal distributions were assigned to the average MMSE scores of patients in each period based on the patientlevel database of the IDEAL study. Two key sources of uncertaintythe variations of the patient-level MMSE scores in the IDEAL study and uncertainty surrounding the parameters of the Mendiondo MMSE disease progression equationwere taken into account by sampling parameters for the Mendiondo equation and propagating these through the patient-level database using Monte Carlo simulation. Results
Cost-effectiveness of rivastigmine patch

Results of the cost-effectiveness analysis using the MMSE and MMSE-ADL models to predict disease progression are shown in Table 2. The MMSE model predicted incremental costs over 5 years of 1174 and 1336 for patients treated with
Table 2 Cost-effectiveness analysis base casea Model used Benets QALYs gained MMSE scores gained Mean survival in years Costs () Drug costs Monitoring costs Institutionalization costs Community care costs Informal care costs Total costs Rivastigmine patch MMSE MMSE-ADL 1.6318 86.2 3.77 1678 533 37 216 37 698 77 126 1.6091 84.7 3.77 1678 533 47 082 33 046 82 339

rivastigmine patch and capsules, respectively, versus BSC. Corresponding incremental benets were estimated at 0.1109 and 0.0882 QALYs, respectively. Incremental costs per QALY were 10 579 with rivastigmine patch and 15 154 with capsules, suggesting that both treatments are cost-effective compared with BSC. Comparing patch and capsules directly, the patch appears marginally more benecial at an incremental 0.0228 QALYs, with similar, though slightly lower costs (a difference of 162 over 5 years). The MMSE-ADL model estimates incremental costs over 5 years of 1011 with rivastigmine patch and 1213 with capsules, versus BSC. These costs are lower than the MMSE-only model (by 14 and 9%, respectively). The QALY differences versus BSC are identical to those in the MMSE model; hence, the incremental cost-effectiveness ratios of 9114 for rivastigmine patch and 13 758 for capsules suggest that both treatments would be considered costeffective. Comparing patch and capsules, the patch is marginally more benecial, with both the MMSE and MMSE-ADL models estimating moderate 5-year cost savings (162 and 202, respectively). The MMSEADL model estimates larger cost savings because this model predicts more patients institutionalized at each

Rivastigmine capsule MMSE MMSE-ADL 1.5209 78.7 3.77 1639 521 37 619 37 508 77 288 MMSE model 1639 521 47 559 32 821 82 541

Best supportive care MMSE MMSE-ADL

39 179 36 773 75 952

49 353 31 975 81 328

MMSE-ADL model 9114 136 26.5 13 758 205 20.8 Patch dominates (8856) 132 5.7

Rivastigmine patch versus BSC Incremental cost per QALY gained () Incremental cost per MMSE scores gained () Number of institutional days avoided over 5 years Rivastigmine capsules versus BSC Incremental cost per QALY gained () Incremental cost per MMSE scores gained () Number of institutional days avoided over 5 years Rivastigmine patch versus capsules Incremental cost per QALY gained () Incremental cost per MMSE scores gained () Number of institutional days avoided over 5 years
a

10 579 158 22.8 15 154 226 18.0 Patch dominates (7124) 106 4.8

Where results do not differ between the two models, only one value is reported, for clarity. QALY, Quality Adjusted Life Year; MMSE, Mini-Mental State Examination; ADL, Activities of Daily Living; BSC, Best Supportive Care.

Copyright # 2010 John Wiley & Sons, Ltd.

Int J Geriatr Psychiatry 2011; 26: 483494.

Rivastigmine patch cost-utility in the UK

489

Figure 1 Implied probability of institutionalization related to MMSE score, using MMSE- (Stewart, 1997) and MMSE-ADL- (McNamee et al., 2001) based predictions for the cost-effectiveness model. MMSE, Mini-Mental State Examination; ADL, Activities of Daily Living.

level of disease severity, and also has a slightly steeper slope over the range of moderate-to-severe MMSE scores, resulting in greater cost-related differences between treatments (Figure 1).
Sensitivity analyses

Table 3 summarizes the results of one-way sensitivity analyses. Results were particularly sensitive to different sources of probability of institutionalization, with all

other sources (especially non-UK) showing higher probabilities of institutionalization and steeper sloping relationships with MMSE than the base case, implying greater cost-effectiveness. In each analysis, the MMSEADL model resulted in a superior estimate of incremental cost-effectiveness ratio than the MMSE model. The MMSE-ADL model proved insensitive to the method used to establish the relationship between MMSE and ADL scores (using open-label or doubleblind data, including/excluding baseline data and linear/logistic regression). The results of the probabilistic sensitivity analyses are presented as cost-effectiveness acceptability curves. With the MMSE model, the results suggest a high degree of certainty that both patch and capsule can be considered cost-effective versus BSC (Figure 2). The probability of cost-effectiveness at a 20 000 threshold was 100.0% for patch (Figure 2a) and 89.7% for capsule (Figure 2b). It is important to account for uncertainty when comparing patch and capsule. In the probabilistic sensitivity analyses (Figures 2c and 3c), incremental costs and QALYs between patch and capsule span every quadrant of the cost-effectiveness plane, showing an overlap of patch and capsule efcacy (and cost-effectiveness). In such a circumstance, costeffectiveness would be unlikely to drive a decision between patch and capsule, and patient, carer and physician views and preferences would likely be more

Table 3 One-way sensitivity analysis for rivastigmine patch versus best supportive care Sensitivity analysis MMSE model base case (Stewart, 1997) MMSE-ADL model base case (McNamee et al., 2001) Costs (rivastigmine, ) 77 126 82 339 Costs (BSC, ) 75 952 81 328 Incremental costs () 1174 1011 QALYs (rivastigmine) 1.6318 1.6318 QALYs (BSC) 1.5209 1.5209 Incremental QALYs 0.1109 0.1109 Incremental cost/QALY () 10 579 9114

Varying source of probability of institutionalization, MMSE model UK (Fenn and Gray, 1999) 72 235 71 439 USA (Hauber et al., 2000b) 89 326 90 338 Canada (OBrien et al., 1999) 86 567 87 401 Varying the perspective of the analysis Informal care costs included, 216 494 MMSE model Informal care costs included, 210 659 MMSE-ADL model 217 064 210 357

796 1012 833 570 301

1.6318 1.6318 1.6318 1.6318 1.6318

1.5209 1.5209 1.5209 1.5209 1.5209

0.1109 0.1109 0.1109 0.1109 0.1109

7171 9124 7510 5135 2716

Varying type of regression analysis and patient population to predict LinearOLno baseline 82 178 81 179 LinearDBall patients 82 412 81 374 LogisticOLall patients 82 523 81 505 LogisticOLno baseline 82 355 81 346 LogitDBall patients 82 412 81 374

Townsend-ADL scores, MMSE-ADL model 999 1.6318 1.5209 0.1109 1038 1.6318 1.5209 0.1109 1019 1.6318 1.5209 0.1109 1009 1.6318 1.5209 0.1109 1038 1.6318 1.5209 0.1109

9006 9358 9183 9097 9358

MMSE, Mini-Mental State Examination; ADL, Activities of Daily Living; BSC, Best Supportive Care; QALY, Quality Adjusted Life Year; Linear, Linear regression; Logistic, Logistic regression; OL, Open-Label dataset of the IDEAL study; DB, Double-Blind dataset of the IDEAL study; all patients, observations made on all patients; no baseline, baseline observations disregarded.

Copyright # 2010 John Wiley & Sons, Ltd.

Int J Geriatr Psychiatry 2011; 26: 483494.

490

B. Nagy et al.

Figure 2 Results of probabilistic sensitivity analyses for the MMSE model: (A) rivastigmine patch versus BSC; (B) rivastigmine capsule versus BSC; (C) rivastigmine patch versus capsule.

important. Figure 3 presents the equivalent uncertainty analyses for the MMSE-ADL model. The probability that both patch and capsules are cost-effective versus BSC at the 20 000 threshold is even higher than with the MMSE model. Discussion This study explores the effect of incorporating evidence on functioning, specically ADL, into cost-effectiveness analysis of AD treatments, and is the rst study to examine the cost-effectiveness of rivastigmine patch versus BSC. It provides a novel approach comparing traditional MMSE-based costCopyright # 2010 John Wiley & Sons, Ltd.

effectiveness modelling with an analysis incorporating ADL into the health outcomes and cost drivers of probability of institutionalization and costs of care. In the MMSE-based analysis, the cost-utility of rivastigmine versus BSC was estimated at 10 579 per QALY gained for patch, and 15 154 per QALY gained for capsules. The MMSE-ADL model predicts higher institutionalization cost differences and the resulting estimates show rivastigmine as being more cost-effective over a 5-year time horizon versus BSC. The MMSE-ADL model estimates are slightly more sensitive to the evidence used to quantify probability of institutionalization than the MMSE-only model (Table 2). Both models estimated costs per QALY lower than the frequently referenced NICE threshold
Int J Geriatr Psychiatry 2011; 26: 483494.

Rivastigmine patch cost-utility in the UK

491

Figure 3 Results of probabilistic sensitivity analyses for the MMSE-ADL model: (A) rivastigmine patch versus BSC; (B) rivastigmine capsule versus BSC; (C) rivastigmine patch versus capsule.

range (20 00030 000 per QALY), suggesting that both the rivastigmine patch and rivastigmine capsules might be considered cost-effective treatments for AD compared with BSC. There are several possible limitations to this analysis. Firstly, only one explanatory variable (MMSE score) was included in the regression analyses relating ADL scores to MMSE. Other patient characteristics (e.g. age, sex) were not included as they could not be incorporated in the cost-effectiveness model, which primarily characterizes patients according to MMSE. However, as the randomization process of the clinical trial produced groups of patients with very similar baseline characteristics between the study arms, this may not have biased our ndings. Second, these
Copyright # 2010 John Wiley & Sons, Ltd.

regression analyses were based on relatively short-term cross-sectional data. It was, therefore, not possible to test the validity of the MMSEADL relationship over a longer time scale. Long-term (5-year) ADL and MMSE evidence to validate the ADL-based predictions would be useful. A third limitation is that our mapping of the ADCS-ADL scale to the Townsend-ADL scale was based on a qualitative approach which we are not able to validate empirically. A survey of a patient population using both scales would be helpful for this purpose. Fourth, the analyses incorporating ADL data focussed mostly on estimation of institutionalization, the most inuential predictor of AD costs. Further evidence on the relationships between ADL and other inputs, especially utilities, community care costs, and
Int J Geriatr Psychiatry 2011; 26: 483494.

492

B. Nagy et al.

informal care costs, would enable renement of the cost-effectiveness estimates. In modelling long-term cost-effectiveness in AD, several assumptions need to be made regarding long-term disease progression and probability of institutionalization. The key assumption in our analysis is that the Mendiondo natural history model of disease progression applies to both the BSC arm and the rivastigmine arms in the model (Mendiondo et al., 2000). An alternative used in previous modelling exercises is Markov modelling of MMSE subgroups to predict transitions over the long-term, based on the placebo arm of the trial. However, using this approach with the placebo arm of the IDEAL dataset produced a long-term prediction of almost no disease progression over time (because mean MMSE change for these patients over 24 weeks was very small). This was so contradictory to clinical and real-life experience as to be an implausible modelling approach. Published estimates of probability and time to institutionalization vary greatly and there is no theoretical reason why the relationship should be linear. The base case assumption used here is the most recent available for the UK, but is nearly 10 years old. The use of alternative (non-UK and nonlinear) data sources (OBrien et al., 1999; Hauber et al., 2000b) alters the costutility ratio estimates (Table 3), although no analysis resulted in a costutility higher than the NICE threshold. In addition, a recent analysis suggested that the public sector may meet 70% of institutional care costs (NICE, 2006b). However, in reality many patients/carers cover varying degrees of these costs. More up-to-date evidence on the UK probability and timing of institutionalization related to MMSE and ADL, and on the proportion of costs paid by public sector and by individuals would be useful for decision makers. Similarly, while the relationship between utilities and MMSE is not necessarily linear, no better source of evidence to correlate utility to MMSE score was available. However, the Brazier equations used in this analysis seemed reasonably well established (R2 0.52; p < 0.001). Several factors are excluded from the calculations. Costs and benets incurred by carers were not included in our central analysis. When informal care costs were included in the model, the costutility ratio changed in favour of rivastigmine. This is largely because in community settings, informal caregiver time and costs are related to disease severity, whilst in institutional settings informal caregiver time and costs are independent of disease severity. The model estimates marginally higher disease severity for patients in the BSC arm compared with rivastigmine, so BSC arm
Copyright # 2010 John Wiley & Sons, Ltd.

patients are estimated to receive more informal care during the period that they are in a community setting. This in turn reduces the estimated difference in costs between the treatment arms. We excluded any potential reduction of caregiver workload related directly to the patch, because although it is conceived of as a once-daily, easy-to-administer treatment for AD, and is preferred by most caregivers to capsules (Winblad et al., 2007b), there was no denitive quantied evidence on caregiver time saved to factor this preference into either the quality of life or the costs side of a carer-based cost-effectiveness analysis. Also excluded were mild ChEI-associated side effects such as nausea and vomiting, because no direct evidence on the HRQoL or costs of such adverse events was available. However, withdrawals due to side effects are included in the discontinuation rates for both arms. The model framework presented here could be extended to assess other questions on cost-effectiveness in AD. Analyses using inputs specic to non-UK countries could determine whether rivastigmine patch could provide similar savings in the USA and elsewhere. This analysis compares rivastigmine patch against rivastigmine capsules and BSC. Further work analysing rivastigmine patch versus other ChEIs would be of interest. However, without access to individuallevel data for other ChEIs, a detailed analysis would be difcult. It would also be an indirect mixed treatment comparison, as head-to-head study data are not available. In conclusion, future health economic analyses should move beyond the exclusive use of cognitive outcomes such as MMSE to model cost-effectiveness, and incorporate measures of function and behavioural deterioration. Collection of additional data on MMSE, ADL, QoL of patients and/or carers, carer time and behavioural and psychological symptoms of dementia, and subsequent further research on the relationships between MMSE, ADL, disease progression, probability of institutionalization, utilities and costs would be welcomed by analysts and decision makers. In this analysis, both the MMSE and MMSE-ADL models suggest that rivastigmine patch and capsules can be considered cost-effective treatments for AD versus BSC in the UK context. The MMSE-ADL model indicated somewhat greater estimated cost-effectiveness, primarily due to a greater number of institutionalized days avoided with rivastigmine compared with BSC. Incorporation of ADL-based evidence into costeffectiveness analysis should be a priority. Decision makers in a variety of organizations and countries now use cost-effectiveness analyses such as this one to help determine reimbursement decisions and priorities.
Int J Geriatr Psychiatry 2011; 26: 483494.

Rivastigmine patch cost-utility in the UK

493
Bullock R, Touchon J, Bergman H, et al. 2005. Rivastigmine and donepezil treatment in moderate to moderately-severe Alzheimers disease over a 2-year period. Curr Med Res Opin 21: 13171327. Corey-Bloom J, Anand R, Veach J. 1998. A randomized trial evaluating the efcacy and safety of ENA 713 (rivastigmine tartrate), a new acetylcholinesterase inhibitor, in patients with mild to moderately severe Alzheimers disease. Int J Geriatr Psychopharmacol 1: 5565. Cummings JL, Frank JC, Cherry D, et al. 2002. Guidelines for managing Alzheimers disease: Part I. Assessment. Am Fam Physician 65: 22632272. Curtis L. 2008. Unit Costs of Health and Social Care 2008. Personal Social Services Research Unit, University of Kent: Canterbury. Curtis L, Netten A. 2006. Unit Costs of Health and Social Care 2006. Personal Social Services Research Unit, University of Kent: Canterbury. Desai AK, Grossberg GT, Sheth DN. 2004. Activities of daily living in patients with dementia: clinical relevance, methods of assessment and effects of treatment. CNS Drugs 18: 853875. European Medicines Agency. 2008. Guideline on medicinal products for the treatment of Alzheimers disease and other dementias. Rep. Doc. Ref. CPMP/EWP/553/95 Rev. 1. Fenn P, Gray A. 1999. Estimating long-term cost savings from treatment of Alzheimers disease. A modelling approach. Pharmacoeconomics 16: 165174. Ferri CP, Prince M, Brayne C, et al. 2005. Global prevalence of dementia: a Delphi consensus study. Lancet 366: 21122117. Folstein MF, Folstein SE, McHugh PR. 1975. Mini-mental state. A practical method for grading the cognitive state of patients for the clinician. J Psychiatr Res 12: 189 198. Furlong W, Feeney D. 1998. Multiplicative multi-attribute utility function for the Health Utilities Index Mark 3 (HUI 3) system: A technical report. McMaster University Centre for Health Economics and Policy Analysis. Working Paper no. 9811. Gareld FB, Getsios D, Caro JJ, Wimo A, Winblad B. 2002. Assessment of Health Economics in Alzheimers Disease (AHEAD): treatment with galantamine in Sweden. Pharmacoeconomics 20: 629637. Green C. 2007. Modelling disease progression in Alzheimers disease: a review of modelling methods used for cost-effectiveness analysis. Pharmacoeconomics 25: 735750. Grossberg G, Sadowsky C, Forstl H, et al. 2009. Safety and tolerability of the rivastigmine patch: results of a 28-week open-label extension. Alzheimer Dis Assoc Disord 23: 158164. Hatoum HT, Thomas SK, Lin SJ, Lane R, Bullock R. 2009. Predicting time to nursing home placement based on activities of daily living scores a modelling analysis using data on Alzheimers disease patients receiving rivastigmine or donepezil. J Med Econ 12: 98103. Hauber AB, Gnanasakthy A, Mauskopf JA. 2000a. Savings in the cost of caring for patients with Alzheimers disease in Canada: an analysis of treatment with rivastigmine. Clin Ther 22: 439451. Hauber AB, Gnanasakthy A, Snyder EH, et al. 2000b. Potential savings in the cost of caring for Alzheimers disease treatment with rivastigmine. Pharmacoeconomics 17: 351360. Ikeda S, Yamada Y, Ikegami N. 2002. Economic evaluation of donepezil treatment for Alzheimers disease in Japan. Dement Geriatr Cogn Disord 13: 3339. Jonsson L, Lindgren P, Wimo A, Jonsson B, Winblad B. 1999. The cost-effectiveness of donepezil therapy in Swedish patients with Alzheimers disease: a Markov model. Clin Ther 21: 12301240. Knapp M, Prince M. 2007. Dementia UK: A Report to the Alzheimers Society on the Prevalence and Economic Cost of Dementia in the UK Produced by Kings College London and London School of Economics. Alzheimers Society: London. Loveman E, Green C, Kirby J, et al. 2006. The clinical and cost-effectiveness of donepezil, rivastigmine, galantamine and memantine for Alzheimers disease. Health Technol Assess 10 (iii-iv, ix-xi): 1160. Martin DC, Miller JK, Kapoor W, Aerna VC, Boller F. 1987. A controlled study of survival with dementia. Arch Neurol 44: 11221126. McNamee P, Bond J, Buck D. 2001. Costs of dementia in England and Wales in the 21st century. Br J Psychiatr 179: 261266. Mendiondo MS, Ashford JW, Kryscio RJ, Schmitt FA. 2000. Modelling Mini Mental State Examination changes in Alzheimers disease. Stat Med 19: 16071616. ` Mercier F, Lefevre G, Huang HL, et al. 2007. Rivastigmine exposure provided by a transdermal patch versus capsules. Curr Med Res Opin 23: 31993204. Migliaccio-Walle K, Getsios D, Caro JJ, et al. 2003. Economic evaluation of galantamine in the treatment of mild to moderate Alzheimers disease in the United States. Clin Ther 25: 18061825. Netten A, Darton R, Bebbington A, et al. 2001. Residential and nursing home care of elderly people with cognitive impairment: prevalence, mortality and costs. Aging Ment Health 5: 1422. Neumann PJ, Hermann RC, Kuntz KM, et al. 1999. Cost-effectiveness of donepezil in the treatment of mild or moderate Alzheimers disease. Neurology 52: 11381145. NICE. 2004. Guide to the Methods of Technology Appraisal. National Institute for Clinical Excellence: London.

Key Points
 Measures of cognitive decline alone may not

allow accurate prediction of probability of institutionalization in patients with Alzheimers disease.  Rivastigmine patch and capsules are cost-effective treatments versus BSC, according to a traditional economic model based on MMSE and a revised model, which also incorporates activities of daily living.  Future economic evaluations of AD treatment should include measures of both cognition and functioning. Recent NICE guidance concluded that the three acetylcholinesterase inhibitors donepezil, galantamine and rivastigmine are recommended as options in the management of patients with Alzheimers disease of moderate severity (NICE, 2009). The health economic analysis presented here is broadly in line with these NICE recommendations, suggesting that policymakers and clinicians should consider rivastigmine patch and capsule as cost-effective interventions for the treatment of moderate AD in England and Wales.

Conflict of interest None declared.

Acknowledgements This study was supported by Novartis Pharmaceuticals Corporation, East Hanover, NJ, USA. Alpha-Plus Medical Communications Ltd (UK) provided editorial assistance with the production of the manuscript. The authors gratefully acknowledge John Brazier for his previous research on utility valuation, and Patricia Sacco for her assistance in the submission of this paper.

References
Bavazzano A, Magnol S, Calvani D, et al. 1998. Functional evaluation of Alzheimer patients during clinical trials: a review. Arch Gerontol Geriatr 26(Suppl 1): 2732. Bell C, Araki S, Neumann P. 2001. The association between caregiver burden and caregiver health-related quality of life in Alzheimer disease. Alzheimer Dis Assoc Disord 15: 129136. Bowie P, Branton T, Holmes J. 1999. Should the Mini Mental State Examination be used to monitor dementia treatments? Lancet 354: 15271528. Brazier J. 2001. Quality of Life evidence for patients with Alzheimers disease, University of Shefeld, School of Health and Related Research.

Copyright # 2010 John Wiley & Sons, Ltd.

Int J Geriatr Psychiatry 2011; 26: 483494.

494
NICE. 2008. Guide to the Methods of Technology Appraisal. National Institute for Clinical Excellence: London. NICE. 2006a. Dementia Supporting People with Dementia and their Carers in Health and Social Care NICE Clinical Guideline 42. National Institute for Clinical Excellence: London. NICE. 2006b. Donepezil, Galantamine, Rivastigmine (Review) and Memantine for the Treatment of Alzheimers Disease Final Appraisal Determination. National Institute for Clinical Excellence: London. NICE. 2009. Donepezil, Galantamine, Rivastigmine (Review) and Memantine for the Treatment of Alzheimers Disease (Amended). National Institute for Clinical Excellence: London. OBrien BJ, Goeree R, Hux M, et al. 1999. Economic evaluation of donepezil for the treatment of Alzheimers disease in Canada. J Am Geriatr Soc 47: 570578. Rosler M, Anand R, Cicin-Sain A, et al. 1999. Efcacy and safety of rivastigmine in patients with Alzheimers disease: international randomised controlled trial. Br Med J 318: 633638. Rozsa S. 2007. The Psychometric Properties of ADCS Activities of Daily Living Inventory and Comparison of Different ADL Scores. Department of Personality and Health Psychology, Lorand Eotvos University: Budapest, Hungary.

B. Nagy et al.
Stewart A. 1997. Costs of care for people with dementia aged 75 and over. 1303/2, Personal Social Services Research Unit (PSSRU). Stewart A, Phillips R, Dempsey G. 1998. Pharmacotherapy for people with Alzheimers disease: a Markov-cycle evaluation of ve years therapy using donepezil. Int J Geriatr Psychiatr 13: 445453. Townsend P. 1979. Poverty in the United Kingdom, a Survey of Household Resources and Standards of Living. Penguin Books and Allen Lane. London. Wimo A, Winblad B, Stofer A, Wirth Y, Mobius HJ. 2003. Resource utilisation and cost analysis of memantine in patients with moderate to severe Alzheimers disease. Pharmacoeconomics 21: 327340. Winblad B, Cummings J, Andreasen N, et al. 2007a. A six-month double-blind, randomized, placebo-controlled study of a transdermal patch in Alzheimers disease rivastigmine patch versus capsule. Int J Geriatr Psychiatr 22: 456 467. Winblad B, Kawata AK, Beusterien KM, et al. 2007b. Caregiver preference for rivastigmine patch relative to capsules for treatment of probable Alzheimers disease. Int J Geriatr Psychiatr 22: 485491. Wolstenholme J, Fenn P, Gray A, et al. 2002. Estimating the relationship between disease progression and cost of care in dementia. Br J Psychiatr 181: 3642.

Copyright # 2010 John Wiley & Sons, Ltd.

Int J Geriatr Psychiatry 2011; 26: 483494.

Copyright of International Journal of Geriatric Psychiatry is the property of John Wiley & Sons, Inc. and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.