Resource Utilisation and Cost Analysis of Memantine in Patients with Moderate to Severe Alzheimer s Disease

ORIGINAL RESEARCH ARTICLE Pharmacoeconomics 2003; 21 (5): 1 1170-7690/03/0099-0001/$30.00/0 Adis International Limited. All rights reserved. Resource Utilisation and Cost Analysis of Memantine in Patients with Moderate to Severe Alzheimer s Disease Anders Wimo, 1 Bengt Winblad, 2 Albrecht Stöffler, 3 Yvonne Wirth 3 and Hans-Jörg Möbius 3 1 Division of Geriatric Epidemiology (Sector of Health Economy), Neurotec, Karolinska Institute, Huddinge, Sweden 2 Division of Geriatric Epidemiology, Neurotec, Karolinska Institute, Huddinge, Sweden 3 Merz Pharmaceuticals GmbH, Frankfurt/Main, Germany Abstract Background: Alzheimer s disease (AD) is a devastating illness that causes enormous emotional stress to affected families and is associated with substantial medical and nonmedical costs. Objective: To determine the effects of 28 weeks of memantine treatment for patients with AD on resource utilisation and costs. Study design and methods: Multicentre, prospective, double-blind, randomised, placebo-controlled clinical trial performed in the US. The Wilcoxon-Mann- Whitney test was used to examine the resource utilisation variables and logistic regression models were used for multivariate resource utilisation analyses. Analysis of covariance (ANCOVA) models (log and non-log) were computed to examine costs from a societal perspective. All costs were calculated in 1999 US dollars. Study population: Outpatients with moderate to severe AD. Overall, 252 patients received randomised treatment, and 166 patients (placebo n = 76, memantine n = 90) formed the treated-per-protocol (TPP) subset for the health economic analyses, on which the main cost analysis was based. Main outcome measure: Resource Utilisation in Dementia (RUD) scale, measuring patient and caregiver resource utilisation, and various sources for cost calculations. Results: Controlling for baseline differences between the groups, significantly less caregiver time was needed for patients receiving memantine than for those receiving placebo (difference 51.5 hours per month; 95% CI 95.27, 7.17; p = 0.02). Analysis of residential status also favoured memantine: time to institutionalisation (p = 0.052) and institutionalisation at week 28 (p = 0.04 with the chi-square test). Total costs from a societal perspective were lower in the memantine group (difference $US1089.74/month [non-overlapping 95% CI for treatment difference 1954.90, 224.58]; p = 0.01). The main differences between the groups were total caregiver costs ($US 823.77/month; p = 0.03) and direct nonmedical costs ($US 430.84/month; p = 0.07) favouring memantine

2 Wimo et al. treatment. Patient direct medical costs were higher in the memantine group (p < 0.01), mainly due to the cost of memantine. Conclusion: Resource utilisation and total health costs were lower in the memantine group than the placebo group. The results suggest that memantine treatment of patients with moderate to severe AD is cost saving from a societal perspective. Alzheimer s disease (AD) is a chronic illness associated with a progressive loss of cognitive and intellectual abilities, such as memory, judgement, and abstract thinking. AD affects about 25 million people worldwide. [1] Management of the functional disabilities of patients with AD is highly cost-intensive. The yearly healthcare costs of AD (in 1996 $US) are estimated to range from $US1.8 billion in England to $US155 billion in the US ($US3.3 billion in Canada, $US3.4 billion in Sweden). [2] In addition, families and caregivers of patients with AD undergo increasing stress, both financial and emotional, as the disease progresses. There is strong support in the literature that caregivers of patients with dementia often suffer from stress-burden depression. [3] Furthermore, caregivers act as unpaid providers of informal care, a contribution that is substantial from a societal perspective. [4] Studies have indicated that the costs of family contribution to the care of patients with AD may be as high as 70% or more of the total home-care costs. [5,6] Pharmacological treatment of AD focuses mainly on improving memory and cognition. To date the therapeutic options have included only drugs augmenting cholinergic neurotransmission. More recently, pharmacological intervention addressing glutamatergic neurotransmission has become of interest, resulting from a greater recognition that several neurotransmitter systems are involved in the aetiology of AD. [7] While the majority of previous clinical studies in AD focused on patients with mild to moderate disease, the present study included patients with moderate to severe AD. Severe AD is a devastating condition, causing great suffering among patients and caregivers. [8] At this stage of the disease, patients still have an average life expectancy of about 3 years. [9] Furthermore, about 75% of the total healthcare costs associated with AD occur during the stage of severe dementia; they are mainly attributable to admissions to nursing homes. [10] Consequently, any drug that prolongs the onset of severe dementia or improves the functional capacity of a patient with advanced AD would significantly lower the overall healthcare costs. Memantine (1-amino-3,5-dimethyladamantane hydrochloride, CAS 4100-52-1) is a noncompetitive N-methyl-D-aspartate (NMDA) antagonist in development for the treatment of AD. By inhibiting excessive stimulation of NMDA receptors, memantine has the potential to provide both symptomatic improvement and neuroprotective effects. [11-14] In Germany, memantine has been used clinically for various central nervous system indications since the 1980s and has been approved recently by the European Medicine Evaluations Agency for the treatment of moderate to severe AD. A first controlled clinical trial of memantine in severe dementia (mean Mini-Mental State Examination [MMSE] score of 6.3 at baseline) was published by Winblad and Poritis in 1999. [15] So far, only limited data on the health economics of antidementia drugs are available. The present double-blind, placebo-controlled study was prospectively included as part of a phase III pharmacoeconomic trial. [16] The primary objective of this pharmacoeconomic study was to compare resource utilisation and costs in patients receiving memantine with those in patients receiving placebo for 28 weeks. Methods Study Design This was a multicentre, prospective, double-

Economics of Memantine in Alzheimer s Disease 3 blind, randomised, placebo-controlled 28-week clinical trial performed in the US. The costing process in the pharmacoeconomic part of the trial consisted of two phases; in the first phase the use of resources was quantified and in the second phase, costs were calculated by the multiplication of the resource use by a per item/per diem cost. Patient and caregiver resource utilisation data were captured by the Resource Utilisation in Dementia (RUD) questionnaire [17] at baseline, week 12, and week 28. The study complied with the principles of the Declaration of Helsinki, was approved by the local Ethics Committees, and was conducted in accordance with the US FDA regulations pertaining to Good Clinical Practice in clinical studies of investigational products and the International Conference on Harmonisation Good Clinical Practice Consolidated Guideline (May 1997). Patients Patients eligible for this study were outpatients aged at least 50 years with probable AD according to the Diagnostic and Statistical Manual of Mental Disorders (DSM-IV) [18] and the National Institute of Neurologic and Communicative Disorders and Stroke and AD and Related Disorders Association (NINCDS-ADRDA) [19] criteria. Eligibility criteria included baseline MMSE [20] scores of 3 14, stage 5 or 6 on the Global Deterioration Scale (GDS), [21] and stage 6a or greater on the Functional Assessment Staging (FAST). [22] Procedures Clinical Assessments The effects of study medication on function and cognition were assessed by the use of the modified Alzheimer s Disease Cooperative Study (ADCS) Activities of Daily Living (ADLsev) Inventory, FAST, and the severe impairment battery (SIB). [22-24] The clinical relevance of these assessments was further confirmed by a clinical global rating (Clinicians Interview-Based Impression of Change plus caregiver input [CIBIC-Plus]). [25] Trial Medication Tablets containing memantine (5mg or 10mg memantine HCl; Merz Pharmaceuticals) or placebo tablets of identical appearance were used. The total daily dose was 20mg (10mg given twice daily). Resource Utilisation Resource utilisation was measured in days spent in nursing homes, hospitals or emergency rooms, as well as in hours of home support or informal care. Costs of resource utilisation were calculated by using a set of per-term costs. Data were generated by using the RUD instrument [17] that assesses resource utilisation in AD. The RUD instrument is a structured interview of the patient s caregiver consisting of a baseline questionnaire (assessing basic patient demographic data and specific events that occurred during the past 1 month) and a follow-up questionnaire (assessing specific events that occurred during the past 3 months). The baseline questionnaire of the RUD instrument captured 1 month of caregiver time, caregiver direct medical resource utilisation, patient direct medical resource utilisation and patient direct nonmedical resource utilisation. Certain relevant caregiver variables, such as direct medical resources and productivity losses are also assessed. The baseline questionnaire was completed at week 0 (baseline), with the follow-up questionnaire completed at weeks 12 and 28, or at the time of dropout if the study was terminated prematurely. The specific events included time spent caring for the patient, changes in the caregiver s work status, healthcare resource utilisation by the caregiver, healthcare resource utilisation by the patient, and the patient s residential status. Caregiver time included time spent for tasks such as using the toilet, eating, dressing, grooming, walking, bathing, shopping, cooking, housekeeping, laundry, transportation, taking medication, managing finances, and providing supervision. Caregiver work loss included permanent and temporary reductions of paid labour. Residential status was categorised as community living (patient s own home or friend s/relative s home) or institutional

4 Wimo et al. living (intermediate-care facility, home oriented toward dementia care, nursing home or other longterm facility). Most resource data were obtained during the full 3 months preceding the visits at weeks 12 and 28. An exception was caregiver time, which was assessed for 1 month preceding the visits at weeks 12 and 28. Thus, most variables were calculated for a total of 6 months, while caregiver time was assessed for 2 months in total. To maintain consistency, utilisation data were interpolated for the whole study period and presented as average monthly costs. Statistical Analyses Analysis of Populations To avoid potential biases in the analysis of the pharmacoeconomic data, the treated-per-protocol (TPP) population was predefined to serve as the basis for the main RUD analysis. The TPP population excluded patients with missing values due to cancelled or missed visits (such as patients not completing the study) and those with major protocol violations. The variables total societal cost and caregiver time defined in the TPP analysis were also analysed by a last-observation-carried-forward (LOCF) approach on the intention-to-treat (ITT) population. The ITT population for this analysis was defined as all randomised patients with baseline data (n = 252). Analysis of Resource Utilisation Variables Resource utilisation data [17] are expressed as average monthly utilisation. The Wilcoxon-Mann- Whitney test was used to examine the differences between study groups because most resource utilisation variables were not normally distributed, with the exception of caregiver time. Because the distribution of caregiver time approached normality, an analysis of covariance (ANCOVA) model was used. However, ANCOVA models could not be used for other analyses, including baseline characteristics, because of the low levels of utilisation reported for those resource items. Thus, logistic regression models were computed instead. A chi-square test was used to assess whether the proportion of patients in community and institutional residential settings differed with respect to treatment group at week 28. For patients living in a community setting at baseline, time to institutionalisation was analysed using Kaplan-Meier curves. An event occurred if a patient living in a community setting moved to an institutional setting. Differences in time to institutionalisation between the treatment groups were evaluated using a log-rank test. Analysis of Costs Cost data are expressed as average monthly costs from a societal perspective in 1999 $US. An average unit of cost was assigned to each relevant resource utilisation variable. Cost units and their sources are shown in table I. Costs were computed by multiplying resources by their cost units and summing them over the double-blind treatment period. An opportunity cost approach was used for caregiver time. Total societal costs were calculated by adding up all cost components, including total direct medical costs (such as hospital care), total direct nonmedical costs (such as nursing home care) and the cost of caregiver time (opportunity cost approach). Results Patient Disposition Patient demographics and baseline characteristics were similar in the two arms (figure 1 and table II). The ITT population of the efficacy analysis consisted of 252 (memantine n =126, placebo n = 126) patients. Of these, 181 completed the 28- week treatment period (memantine 97, placebo 84) and formed the observed cases population (OC). The discontinuation rate was higher in the placebo group (42 patients; 33%) than in the memantine group (29 patients; 23%). Adverse events led to premature study termination in 22 (17%) placebo versus 13 (10%) memantine recipients. Of these 181 patients, 10 patients were excluded because of

Economics of Memantine in Alzheimer s Disease 5 Table I. Unit cost and source of resource utilisation items Resource utilisation item Unit of measurement Unit cost ($US; Source 1999 values) Hospitalisations Hospitalisation 467.10 Literature [26] Emergency room visits Visit 143.42 Literature [26] Physician visits a Visit 64.00 1999 Medicare physician fee and coding schedule Other medical visits b Visit 54.40 1999 Medicare nurse practitioner fee schedule Visiting nurse Visit c 27.10 Mean value from a survey of 5 providers Home health aid Visit c 13.03 Mean value from the literature [27,28] and a survey of 1 provider Adult day care Day 38.09 Mean value from the literature [29] and a survey of 2 providers Caregiver time d male <25y Hour 10.30 Bureau of labour statistics male 25 34y Hour 17.05 Bureau of labour statistics male 35 44y Hour 21.55 Bureau of labour statistics male 45 54y Hour 23.65 Bureau of labour statistics male 55 64y Hour 23.25 Bureau of labour statistics male 65y Hour 17.75 Bureau of labour statistics female <25y Hour 9.18 Bureau of labour statistics female 25 34y Hour 13.65 Bureau of labour statistics female 35 44y Hour 15.33 Bureau of labour statistics female 45 54y Hour 16.55 Bureau of labour statistics female 55 64y Hour 14.98 Bureau of labour statistics female 65y Hour 11.15 Bureau of labour statistics Institutionalisation e Day 99.07 Literature [22] Transportation service Trip 1.43 Mean value from a survey of 4 providers Food delivery service Meal 3.43 Mean value from a survey of 3 providers Memantine f Month 143.96 Merz Pharmaceuticals Non-study drugs Tablet Varied 1999 Red book g a Assumes the same costs for all types of physician visits. b Assumes the same costs for all types of other medical visits. c Assumes 1 visit = 1 hour. d An opportunity cost approach was used to estimate caregiver time. e To maintain consistency with the resource utilisation analysis, residential status was dichotomised (community or institutional). Because the majority of institutionalised patients were in an assisted living setting, the cost of assisted living was used for the unit cost of institutionalisation. f Memantine is not yet approved in the US. After EU approval it was launched on 1 August 2002 in Germany with a daily treatment cost of 4.72 euros at a 20mg daily dose. Assuming parity of the US dollar and euro, we calculated 4.72 30.5 = $US143.96 monthly treatment cost. g See Appendix A of the 1999 Red Book [30] for a complete list of medication unit costs for patients and caregivers. major protocol deviations. The resulting TPP population included 171 patients (memantine n = 93, placebo n = 78). For the TPP analysis of the RUD data, five additional patients (memantine n = 3, placebo n = 2) had to be excluded per protocol because of at least one missed RUD visit (as predefined by the complete absence of data for the RUD at any given visit). Thus, the TPP analysis of the RUD data consisted of 166 patients (memantine n = 90, placebo n = 76; table II). Efficacy Findings The clinical efficacy results of this trial are reported in greater detail elsewhere. [16] The decline in various domains was consistently smaller for the memantine treatment group at endpoint. At week

6 Wimo et al. Screen failures n = 93 Completers n = 84 (67%) Placebo n = 126 (100%) TPP population for RUD analysis n = 76 (60%) Dropouts n = 42 (33%) Protocol violators (or missing RUD visits) n = 8 (6%) Patients randomised n = 345 Patients randomised n = 252 Completers n = 97 (77%) Memantine n = 126 (100%) TPP population for RUD analysis n = 90 (71%) Dropouts n = 29 (23%) Protocol violators (or missing RUD visits) n = 7 (5%) Fig. 1. Disposition of study population. RUD = Resource Utilisation in Dementia; TPP = treated-per-protocol. 28, there was a statistically significant superiority of memantine over placebo for all key outcomes (global, functional and cognitive domains). For the CIBIC-Plus, a mean difference of 0.36 points (p = 0.025; n = 181) was found. For the functional AD- Lsev total score, the mean difference between treatment groups was 3.37 points (p = 0.003; n = 181) and for the FAST the difference was 0.4 points (p = 0.007; n = 181). This was consistent with memantine s effect on cognitive performance: for SIB total scores, a 5.70-point mean difference (p = 0.002; n = 179) was shown. The behavioural parameter Neuropsychiatric Inventory (NPI) and the scaling instrument MMSE did not show significant differences at week 28.There were no clinically important differences between the treatment groups in the overall number of patients who experienced adverse events (AEs). Furthermore, all medical aspects of AEs were covered by the RUD questionnaire and therefore the cost by AEs were not analysed separately. Resource Utilisation No significant differences were detected in patient or caregiver resource utilisation data at baseline. However, there was a trend toward higher utilisation of monthly patient psychologist and geriatrician visits at baseline in the memantine group. As the distributions of all variables were skewed, with the exception of caregiver time, which approximated a normal distribution, the Wilcoxon- Mann-Whitney test was computed to provide a more robust test in the presence of non-normality. During the following double-blind study period, caregivers of patients in the placebo group spent 455 hours per month (about 15 hours per day) on caring tasks and supervision, which was 42 hours per month in excess of the time spent by caregivers of patients receiving memantine (table III). The difference between treatment groups for work loss of caregivers was 4 hours and also favoured memantine. There were few relevant differences in resource utilisation between the memantine and placebo groups. The ANCOVA (controlled for baseline caregiver time, patient gender, caregiver gender, caregiver relationship to patient and baseline residential status) showed that caregiver time for patients receiving memantine was significantly less on average than caregiver time for patients receiving placebo (difference 51.52 hours per month; 95% CI 95.27, 7.17; p = 0.02). No significant differences were found in the logistic regression models, where caregiver primary care visits, caregiver total outpatient visits, patient neurologist visits, and patient total outpatient visits were analysed. One patient in the memantine group and five patients in the placebo group living in a community setting moved to an institutional setting in the course of the study. At week 28, the chi-square test showed a statistically significant advantage for the memantine-treated patients residential status (p =

Economics of Memantine in Alzheimer s Disease 7 Table II. Demographics at baseline (treated-per-protocol population) Variable Placebo (n = 76) [no. (%)] Memantine (n = 90) [no. (%)] p value (χ 2 test) Age 0.78 <75y 32 (42.11) 36 (40.00) >75y 44 (57.89) 54 (60.00) Patient gender 0.12 Male 28 (36.84) 23 (25.56) Female 48 (63.16) 67 (74.44) Caregiver gender 0.15 Male 32 (42.11) 48 (53.33) Female 44 (57.89) 42 (46.67) Patient marital status 0.81 Married 58 (76.32) 65 (72.22) Widowed 17 (22.37) 24 (26.67) Divorced/separated 1 (1.32) 1 (1.11) Severity of Alzheimer s disease 0.61 MMSE <10 51 (67.11) 57 (63.33) MMSE 10 25 (32.89) 33 (36.67) Caregiver relationship to patient 0.13 Friend 2 (2.63) 1 (1.1) Spouse 51 (67.11) 57 (63.33) Other family member 18 (23.68) 31 (34.44) Paid caretaker 5 (6.58) 1 (1.1) Residential status 0.16 Community 66 (86.84) 84 (93.33) Institutional 10 (13.16) 6 (6.67) Patient race/ethnicity 0.68 Caucasian 68 (89.47) 80 (88.89) African-American 4 (5.26) 3 (3.33) Other 4 (5.26) 7 (7.78) Patient smoking status 0.36 Nonsmoker 43 (56.58) 59 (65.56) Smoker 4 (5.26) 6 (6.67) Ex-smoker 29 (38.16) 25 (27.78) MMSE = Mini-Mental State Examination. 0.04). The log-rank test of the differences in time to institutionalisation between the treatment groups supported this finding, as it showed a trend towards a statistically significant advantage for memantine patients (p = 0.052). In the LOCF analysis on the ITT population, the required caregiver time yielded results similar to those obtained for the TPP population, showing a statistically significant superiority of memantine over placebo (difference 45.82 hours, with nonoverlapping 95% CI 81.27, 10.37; p = 0.01). Costs In most cases, the mean costs were greater than the median costs (table IV), indicating that the data were positively skewed, a phenomenon often observed in cost data analyses. Moreover, the data

8 Wimo et al. Table III. Monthly resource utilisation (caregivers and patients) during the double-blind study period (treated-per-protocol population) Resource utilisation item Unit of measurement Placebo (n = 76) [mean (±SD)] Memantine (n = 90) [mean (±SD)] Caregiver resource utilisation Productivity loss Caregiver time a Hours 455.64 (232.70) 413.46 (234.27) Work loss b Hours 14.05 (38.72) 9.81 (29.60) Direct medical resources Hospitalisations Events 0.01 (0.05) 0.02 (0.06) Hospital days Days 0.04 (0.23) 0.05 (0.19) Emergency room visits Visits 0.02 (0.06) 0.01 (0.06) Primary-care physician visits Visits 0.23 (0.29) 0.18 (0.22) Geriatrician visits Visits 0.00 (0.00) < 0.01 (0.02) Neurologist visits Visits 0.00 (0.00) 0.01 (0.05) Psychiatrist visits Visits <0.01 (0.02) 0.01 (0.06) Social-worker visits Visits 0.00 (0.00) < 0.01 (0.04) Psychologist visits Visits 0.00 (0.00) 0.04 (0.42) Physical therapist visits Visits 0.03 (0.25) 0.04 (0.29) Occupational therapist visits Visits 0.00 (0.00) 0.00 (0.00) Nurse visits Visits 0.02 (0.08) < 0.01 (0.03) Total outpatient visits Visits 0.45 (0.65) 0.49 (1.09) Patient resource utilisation Direct medical resources Hospitalisations Events 0.01 (0.06) 0.01 (0.05) Hospital days Days 0.07 (0.42) 0.05 (0.23) Emergency room visits Visits 0.03 (0.07) 0.03 (0.10) Primary-care physician visits Visits 0.23 (0.24) 0.21 (0.21) Geriatrician visits Visits 0.01 (0.06) 0.01 (0.06) Neurologist visits Visits 0.03 (0.10) 0.06 (0.14) Psychiatrist visits Visits 0.01 (0.03) 0.01 (0.05) Social-worker visits Visits 0.01 (0.03) 0.01 (0.04) Psychologist visits Visits 0.03 (0.08) 0.08 (0.43) Physical therapist visits Visits 0.08 (0.59) 0.00 (0.00) Occupational therapist visits Visits 0.00 (0.00) 0.00 (0.00) Nurse visits Visits 0.28 (1.18) 0.10 (0.24) Total outpatient visits Visits 0.82 (2.05) 0.67 (0.78) Visiting nurse Visits 0.05 (0.21) 0.06 (0.25) Direct nonmedical resources Home-health aid Visits 0.84 (3.51) 1.41 (3.23) Adult day care Visits 1.15 (3.51) 0.85 (2.04) Transportation service Uses 0.19 (0.97) 0.24 (1.04) Food delivery service Uses 0.23 (1.37) 0.24 (1.72) a Caregiver time includes time spent in caregiving tasks such as using the toilet, eating, dressing, grooming, walking, bathing, shopping, cooking, housekeeping, laundry, transportation, taking medication, managing finances, and providing supervision. Caregiver time was capped at 720 hours, the maximum number of hours in a month. b Work loss includes permanent and temporary reductions in paid labour.

Economics of Memantine in Alzheimer s Disease 9 were highly variable as indicated by the large standard deviations. The largest subcategory of costs was total caregiver costs, mainly due to the substantial caregiver time expended. Without correcting for possibly confounding effects, average monthly caregiver costs during the study period were $US296.19 less in the memantine group than the placebo group. Total societal costs also favoured memantine treatment. Average monthly societal costs during the study period were $US587.29 less in the memantine group than the placebo group. Monthly patient direct medical costs were higher in the memantine group, due to the cost of memantine; the difference was $US159.68. AN- COVA models were computed to examine the difference in costs between memantine and placebo, controlling for total societal costs at baseline, patient gender, caregiver gender and caregiver relationship to the patient. Total caregiver costs and total societal cost factors were statistically significantly lower in the memantine group as compared with the placebo group (table IV). Correcting for confounding effects, average monthly caregiver costs were $US823.77 lower and average monthly societal costs were $US1089.74 lower in the memantine group than the placebo group (nonoverlapping 95% CI for treatment difference 1954.90, 224.58). In addition, there was a trend toward lower direct nonmedical costs among patients receiving memantine, with a between-group difference of $US430.84/month (95% CI 897.93, 36.24; p = 0.07). Since costs were non-normally distributed and positively skewed, a logarithmic transformation of costs (log model) was computed, and the AN- COVA models were reanalysed. Covariates included in the models were total societal log costs at baseline, patient gender, caregiver gender and caregiver relationship to the patient (table IV). The results of the log cost models were similar to those obtained in the linear models, in that the statistically significant factors were patient direct medical costs, total caregiver costs, and total societal costs. As in the linear models, both of the log models of total caregiver costs and total societal costs favoured memantine. In contrast, there were higher patient direct medical costs in the memantine group, due to the cost associated with memantine. A predefined sensitivity analysis was conducted for the costs of institutionalisation. Institutional costs were reanalysed using a lower figure ($US70.68/day) for a group home for patients with dementia, and a higher value ($US126.32/day) for nursing-home care. The group-home cost was the mean value obtained from a survey of two providers, and the nursing-home cost was obtained from the literature. [31] The differences in total societal costs and total societal log costs did not change significantly when the high and low institutionalisation costs were used. Likewise, the differences in direct nonmedical costs and direct nonmedical log costs did not change significantly. Intention-to-Treat Analysis of Costs In addition to the described TPP analysis of total societal costs, the LOCF analysis on the ITT population (table V) confirmed the results obtained in the TPP analysis. The adjusted mean difference between groups was $US840.40 in favour of the memantine group (non-overlapping 95% CI 1501.00, 179.79) and was statistically significant (p = 0.01). Discussion Our findings suggest that the use of memantine instead of placebo in patients with moderate to severe AD saves costs. The societal cost savings were due to lower institutionalisation rates and reduced caregiver time required in the memantine group. Our findings are in good agreement with the reduced total healthcare and societal costs in mild to moderate patients with AD previously reported in a prospective, randomised, controlled health economics study with the choline esterase inhibitor donepezil. [32] However, the benefit of donepezil over placebo was less marked than that of memantine shown in our study, possibly because the patients participating in the earlier study had mild to mod-

10 Wimo et al. Table IV. Monthly medical and nonmedical costs (in 1999 $US) Statistic Placebo Memantine Difference (p value) Patient direct medical costs Unadjusted mean (±SD) 74.32 (129.68) 234.00 (96.01) 159.68 Adjusted mean a (SE) 72.09 (12.70) 236.96 (11.64) 164.87 (<0.01) Adjusted mean of log costs b (SE) 3.49 (0.11) 5.42 (0.10) 1.94 (<0.01) Median 40.12 201.09 160.98 Minimum, maximum 0.00, 1012.25 143.96, 594.63 Caregiver direct medical costs Unadjusted mean (±SD) 67.80 (90.11) 60.45 (76.49) 7.35 Adjusted mean a (SE) 66.70 (9.56) 61.30 (8.76) 5.40 (0.68) Adjusted mean of log costs b (SE) 2.97 (0.23) 2.78 (0.21) 0.19 (0.56) Median 33.78 25.24 8.54 Minimum, maximum 0.00, 439.30 0.00, 322.61 Total direct medical costs Unadjusted mean (±SD) 142.12 (156.83) 294.45 (126.36) 152.33 Adjusted mean a (SE) 138.79 (16.21) 298.26 (14.85) 159.47 (<0.01) Adjusted mean of log costs b (SE) 4.43 (0.10) 5.63 (0.09) 1.20 (<0.01) Median 99.96 267.40 167.44 Minimum, maximum 0.71, 1058.43 143.96, 789.77 Total caregiver costs Unadjusted mean (±SD) 7231.18 (4025.14) 6934.99 (4348.25) 296.19 Adjusted mean a (SE) 7602.52 (279.43) 6778.75 (256.11) 823.77 (0.03) Adjusted mean of log costs b (SE) 8.72 (0.06) 8.55 (0.05) 0.17 (0.03) Median 7338.77 6707.15 631.62 Minimum, maximum 53.25, 16878.31 168.35, 18841.31 Direct nonmedical costs Unadjusted mean (±SD) 529.01 (2148.29) 78.23 (260.36) 450.78 Adjusted mean a (SE) 519.57 (172.73) 88.73 (158.31) 430.84 (0.07) Adjusted mean of log costs b (SE) 1.84 (0.30) 1.94 (0.27) 0.10 (0.81) Median 0.00 0.00 0.00 Minimum, maximum 0.00, 13968.87 0.00, 2377.68 Ambulatory medical costs Unadjusted mean (±SD) 58.15 (114.87) 48.72 (61.34) 9.43 Adjusted mean a (SE) 56.32 (10.45) 50.96 (9.58) 5.36 (0.71) Adjusted mean of log costs b (SE) 3.16 (0.19) 3.10 (0.18) 0.06 (0.84) Median 32.25 32.50 0.25 Minimum, maximum 0.00, 937.31 0.00, 486.90 Hospital costs Unadjusted mean (±SD) 18.52 (49.88) 21.38 (42.14) 2.86 Adjusted mean a (SE) 17.61 (5.17) 21.23 (4.74) 3.61 (0.61) Adjusted mean of log costs b (SE) 0.85 (0.21) 1.08 (0.19) 0.23 (0.42)

Economics of Memantine in Alzheimer s Disease 11 Median 0.00 0.00 0.00 Minimum, maximum 0.00, 319.60 0.00, 136.45 Total societal costs Unadjusted mean (±SD) 7834.51 (4279.54) 7247.22 (4346.65) 587.29 Adjusted mean a (SE) 8194.17 (319.95) 7104.43 (293.24) 1089.74 (0.01) Adjusted mean of log costs b (SE) 8.82 (0.06) 8.62 (0.05) 0.20 (<0.01) Median 7915.45 7109.37 806.08 Minimum, maximum 206.70, 20831.45 333.44, 19360.26 a Controlling for baseline total societal costs, patient gender, caregiver gender and caregiver relationship to the patient (ANCOVA). b Controlling for baseline total societal log costs, patient gender, caregiver gender and caregiver relationship to the patient (ANCOVA). ANCOVA = analysis of covariance; SD = standard deviation; SE = standard error. erate rather than moderate to severe AD. Thus, the difference in resource utilisation would be expected to be smaller if the total resource utilisation is lower for less severely affected patients. Head to head trials of memantine versus choline esterase inhibitors would, of course, provide a better basis for a direct comparison of the compounds than placebo-controlled trials. There has been concern with regard to a possible paradoxical effect on caregivers when antidementia drugs are introduced. [33,34] In the event that these drugs are effective in postponing institutional care, there may be an increased burden on informal caregivers, in terms of both caregiver time and stress. Our study, however, indicates that caregiver time was significantly lower if patients received memantine rather than placebo. It appears unlikely that this result may be attributable to a major extent to behavioural changes induced by memantine, since the behavioural assessment NPI did not show significant differences at endpoint. It is more logical to consider that it is the ADL-effects that are important. The dropout rate of 28% on average might be a reason for concern regarding the interpretation of the data. However, in addition to the predefined analyses, several additional approaches to replace missing values resulted invariably in statistically significant clinical efficacy data favouring memantine treatment (Reisberg et al. [16] ). There is a paucity of comparable data in well-controlled AD trials in such an outpatient population with a mean MMSE score of 7.9 at baseline. The percentage of patients with any adverse event and the number of events related to the study drug were similar between the treatment groups. Slightly more placebo recipients than memantine recipients prematurely discontinued the study due to an AE. The most frequent reason for discontinuation was agitation in both treatment groups, which was slightly more frequent in the placebo group (7% with placebo and 5% with memantine). The discontinuation rate can most likely be attributed to the severe disease stage of the patients. This assumption is supported by a recent clinical trial with a less severely affected patient population (mean MMSE at baseline of about 12) and a shorter observation period (24 weeks) in which a 50% dropout rate was reported. [35] The LOCF method is the most commonly used procedure to replace missing data in efficacy analyses, but its usefulness in AD trials is debated. [36] In this study, the TPP population was predefined for the main RUD analysis because there is no final consensus on how best to avoid bias in the handling of missing data in pharmacoeconomic studies. Resource utilisation represents a flow of resources between two measurement points and is not a point estimate. Since dementia disorders are progressive even after drug treatment, the LOCF approach may lead to overestimation of missing data. Alternative methods to replace missing data, such as mean or median value imputation, regres-

12 Wimo et al. Table V. Total societal costs per month (in 1999 $US, last-observation-carried-forward analysis on the intention-to-treat population) Statistic Placebo (n = 126) Memantine (n = 126) Difference CI for treatment p value difference Unadjusted mean (±SD) 7760.01 (4385.52) 7231.61 (4299.39) 528.40 Adjusted mean a (SE) 8032.85 (235.62) 7192.46 (235.62) 840.40 1501.00, 179.79 0.01 Adjusted mean of log costs b 8.78 (0.04) 8.63 (0.04) 0.15 0.27, 0.03 0.01 (SE) Median 8028.00 7495.39 532.61 Minimum, maximum 46.10, 20831.45 143.96, 19360.26 a Controlling for baseline total societal costs, patient gender, caregiver gender and caregiver relationship to the patient (ANCOVA). b Controlling for baseline total societal log costs, patient gender, caregiver gender and caregiver relationship to the patient (ANCOVA). ANCOVA = analysis of covariance; SD = standard deviation; SE = standard error. sion imputation (to predict the missing value), modelling procedures, hot deck imputation (the value for a patient with similar characteristics is imputed), etc., have been discussed, but the use of these methods in health economics studies should be analysed critically. Standard-of-practice health economic analyses evaluate the differences in cost relative to differences in effectiveness, i.e. they determine incremental cost effectiveness. Effectiveness is demonstrated in the real world, as opposed to efficacy, which is demonstrated in ideal circumstances, such as those encountered in a controlled clinical trial. [37] Effectiveness may differ from efficacy if, for example, compliance with drug therapy is poor. However, because memantine has a favourable adverse event profile and convenient twice-daily administration, compliance is unlikely to be a major concern. In fact, only two patients in the memantine group and no patient in the placebo group were excluded from the TPP population because of poor drug compliance. Overall, the findings in the present study [16] appear to translate into the effectiveness of memantine, but of course further studies are needed to confirm this assumption. Pragmatic designs may be an alternative for such studies. Conclusions In the present study, memantine was more efficacious than placebo in slowing cognitive and functional decline in patients with AD and was associated with reduced total healthcare and societal costs. When a drug is both more effective and less costly than an alternative, it is said to exhibit dominance. [38] Thus, from a societal perspective, memantine dominated placebo in the present study. The healthcare system often focuses predominantly on direct medical costs rather than on total societal costs. Of course, because of the cost of pharmacotherapeutic intervention, which is not uniformly included in such analyses, direct medical costs were higher in the memantine group than the placebo group. However, memantine can be regarded as cost effective if the incremental costs are compared with the incremental effectiveness between the groups. Such a statement relies on the assumption that the efficacy data in terms of the primary outcomes from the clinical part of the study are generalisable to clinical effectiveness in daily care. This trial was conducted in the US. Evidently, there are considerable cultural differences within the developed world and even within Europe regarding the care for patients with AD as their disease progresses relentlessly. With regard to the international extrapolation of such data, it has been suggested previously that this may have some fundamental validity when the essence of care requirements is considered. [39] Thus, it appears conceivable that the necessary future pharmacoeconomic trials in other cultural settings may confirm the memantine benefits observed while using other definitions for the patient s transition from familybased living to intensive nursing care.

Economics of Memantine in Alzheimer s Disease 13 Acknowledgements Anders Wimo and Bengt Winblad have been acting as consultants to most pharmaceutical companies who have developed antidementia drugs (e.g. Parke-Davis, Pfizer, Novartis, Janssen-Cilag, Aventis, Merz). They have no employment, stocks or research grants from these companies. H.J. Möbius, A. Stöffler and Y. Wirth are employees of Merz Pharmaceuticals. References 1. Winblad B. Socio-economic perspectives of dementia and costeffectiveness of treatments. 7th International Geneva/Springfield Symposium on Advances in Alzheimer Therapy; 2002 Apr 3-6; Geneva 2. Jonsson B, Jonsson L, Wimo A. Cost of dementia. In: May M, Sartorius N, editors. Dementia: WPA series evidence and experience in psychiatry. London: John Wiley & Sons, 2000: 335-63 3. Grafstrom M, Winblad B. Family burden in the care of the demented and nondemented elderly: a longitudinal study. Alzheimer Dis Assoc Disord 1995; 9 (2): 78-86 4. Max W, Webber P, Fox P. Alzheimer s disease: the unpaid burden of caring. J Aging Health 1995; 7 (2): 179-99 5. Rice DP, Fox PJ, Max W, et al. The economic burden of Alzheimer s disease care. Health Aff (Millwood) 1993; 12 (2): 164-76 6. Stommel M, Collins CE, Given B. The costs of family contribution to the care of persons with dementia. Gerontologist 1994; 34: 199-205 7. Palmer AM, Gershon S. Is the neuronal basis of Alzheimer s disease cholinergic or glutamergic? FASEB J 1990; 4: 2745-52 8. Winblad B, Wimo A, Möbius HJ, et al. Severe dementia: a common condition entailing high costs at individual and societal levels. Int J Geriatr Psychiatry 1999; 14 (11): 911-4 9. Aguero-Torres H, Fratiglioni L, Winblad B. Natural history of Alzheimer s disease and other dementias: review of the literature in the light of the findings from the Kungsholmen Project. Int J Geriatr Psychiatry 1998; 13 (11): 755-66 10. Wimo A, Winblad B. Health economic aspects of Alzheimer s disease and its treatment. Psychogeriatrics 2001; 3: 189-93 11. Barnes CA, Danysz W, Parsons CG. Effects of the uncompetitive NMDA receptor antagonist memantine on hippocampal long-term potentiation, short-term exploratory modulation and spatial memory in awake, freely moving rats. Eur J Neurosci 1996; 8: 565-71 12. Müller WE, Mutschler E, Riederer P. Noncompetitive NMDA receptor antagonists with fast open-channel blocking kinetics and strong voltage-dependency as potential therapeutic agents for Alzheimer s dementia. Pharmacopsychiatry 1995; 28 (4): 113-24 13. Danysz W, Parsons CG, Möbius HJ, et al. Neuroprotection and symptomatological action of memantine relevant for Alzheimer s disease: an unified hypothesis on the mechanism of action. Neurotox Res 2000; 2 (2-3): 85-98 14. Parsons CG, Danysz W, Quack G. Memantine is a clinically well tolerated NMDA receptor antagonist: a review of preclinical data. Neuropharmacology 1999; 38 (6): 735-67 15. Winblad B, Poritis N. Memantine in severe dementia, results of the M-BEST study. Int J Geriatr Psychiatry 1999; 14 (2): 135-46 16. Reisberg B, Windscheif U, Ferris SH, et al. Memantine in moderately severe to severe Alzheimer s disease (AD): results of a placebo-controlled 6-month trial. Neurobiol Aging 2000; 21 (1S): S275 17. Wimo A, Wetterholm AL, Mastey V, et al. Evaluation of the resource utilization and caregiver time in anti-dementia drug trials: a quantitative battery. In: Wimo A, Jönsson B, Karlsson G, editors. The health economics of dementia. London: John Wiley & Sons, 1988: 465-99 18. American Psychiatric Association. Diagnostic and statistical manual of mental disorders. 4th ed. Washington, D.C: American Psychiatric Association, 1994 19. McKhann G, Drachman D, Folstein M, et al. Clinical diagnosis of Alzheimer s disease: report of the NINCDS-ADRDA Work Group under the auspices of Department of Health and Human Services Task Force on Alzheimer s Disease. Neurology 1984; 34: 939-44 20. Folstein MF, Folstein SE, McHugh PR. Mini-mental state: a practical method for grading the cognitive state of patients for the clinician. J Psychiatr Res 1975; 12: 189-98 21. Reisberg B, Ferris SH, de Leon MJ, et al. The Global Deterioration Scale for assessment of primary degenerative dementia. Am J Psychiatry 1982; 139: 1136-9 22. Sclan SG, Reisberg B. Functional assessment staging (FAST) in Alzheimer s disease: reliability, validity, and ordinality. Int Psychogeriatr 1992; 4 Suppl. 1: 55-69 23. Galasko D, Bennett D, Sano M, et al. An inventory to assess activities of daily living for clinical trials in Alzheimer s disease. Alzheimers Dis Assoc Disord 1997; 11 (2): 33-9 24. Schmitt FA, Ashford W, Ernesto C, et al. The severe impairment battery: concurrent validity and the assessment of longitudinal change in Alzheimer s disease. The Alzheimer s Disease Cooperative Study. Alzheimers Dis Assoc Disord 1997; 11 (2): 51-6 25. Reisberg B, Schneider L, Doody R, et al. Clinical global measures of dementia: position paper from the International Working Group on Harmonization of Dementia Drug Guidelines. Alzheimers Dis Assoc Disord 1997; 11 Suppl. 3: 8-18 26. Gutterman EM, Markowitz JS, Lewis B, et al. Cost of Alzheimer s disease and related dementia in managed-medicare. J Am Geriatr Soc 1999; 47: 1065-71 27. Ernst RL, Hay JW. The US economic and social costs of Alzheimer s disease revisited. Am J Public Health 1994; 8: 1261-4 28. Neumann PJ, Hermann RC, Kuntz KM, et al. Cost-effectiveness of donepezil in the treatment of mild or moderate Alzheimer s disease. Neurology 1999; 52: 1138-45 29. Weinberger M, Gold DT, Divine GW, et al. Expenditures in caring for patients with dementia who live at home. Am J Public Health 1993; 83: 338-41 30. Red book, 1999 edition. Montvale (NJ): Medical Economics Company/Thompson Medical Economics, 1999 31. Leon J, Moyer D. Potential cost savings in residential care for Alzheimer s disease patients. Gerontologist 1999; 39 (4): 440-9 32. Wimo A, Winblad B, Mastey V, et al. Donepezil reduces total healthcare and societal costs in patients with mild to moderate

14 Wimo et al. Alzheimer s disease: results of a one-year, double-blind randomized trial [abstract]. Eur J Neurol 2000; 7 (3): 25 33. Max W. The cost of Alzheimer s disease: will drug treatment ease the burden? Pharmacoeconomics 1996; 9 (1): 5-10 34. Wimo A, Winblad B, Grafstrom M. The social consequences for families with Alzheimer s disease patients: potential impact of new drug treatment. Int J Geriatr Psychiatry 1999; 14 (5): 338-47 35. Feldman H, Gauthier S, Hecker J, et al. A 24-week, randomized, double-blind study of donepezil in moderate to severe Alzheimer s disease. Neurology 2001; 57 (4): 613-20 36. Committee for Proprietary Medicinal Products (CPMP). Points to consider on missing data. (CPMP/EWP/1776/99 draft). London, 2001 Jan. London: The European Agency for the Evaluation of Medical Products (EMEA), 2001 37. Gold MR, Siegel JE, Russel LB, et al. Cost-effectiveness in health and medicine. Oxford: Oxford University Press, 1996 38. Drummond MF, O Brien B, Stoddart GL, Torrance GW. Methods for the economic evaluation of health care programmes. Oxford: Oxford University Press, 1997 39. Jönsson L. Guidelines for cost-effectiveness analysis of antidementia drugs: the role of models. Oral presentation, 3rd International Pharmacoeconomical Conference on Alzheimer s Disease; 2002 Jul 19-20; Stockholm Correspondence and offprints: Dr Anders Wimo, HC Bergsjö, Box 16, S-82070 Bergsjö, Sweden. E-mail: Anders.Wimo@telia.com