NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.

Hansen RA, Gartlehner G, Kaufer DJ, et al. Drug Class Review: Alzheimer's Drugs: Final Report [Internet]. Portland (OR): Oregon Health & Science University; 2006 Jun.

Cover of Drug Class Review: Alzheimer's Drugs

Drug Class Review: Alzheimer's Drugs: Final Report [Internet].

Show details

Results

We identified 1,167 citations from searches and reviews of reference lists; we identified no unpublished trials from dossiers submitted by pharmaceutical companies. In total we included 46 studies: 31 RCTs, 8 meta-analyses or systematic reviews, and 7 studies of other design. Furthermore, we retrieved 59 articles for background information. We could not retrieve six articles after multiple attempts.21–26 For some studies, the investigators published more than a single article; therefore, numbers of referenced articles may not always sum to the number of studies (Figure 1, QUORUM Tree).

Figure 1

Figure 1

Results of literature search

Reasons for exclusions were based on eligibility criteria or methodological criteria. We excluded nine studies that met the eligibility criteria but were later rated as poor quality for internal validity from the analysis (Appendix D). The main reasons for a poor quality rating were high study attrition rates among RCTs and lack of systematic literature search for meta-analyses. Lack of a systematic literature search leads to a selected spectrum of trials and biased results.

Of 46 included studies, 74 percent were supported financially by pharmaceutical companies; 17 percent were funded by governmental agencies or independent funds. We could not determine the funding source for 9 percent of included studies.

Studies reviewed for this report employed several different instruments for assessing symptoms, health status, or quality of life. Table 3 summarizes symptom assessment scales and health status or quality-of-life instruments encountered in this literature and used in this report.

Table 3. Abbreviations and full names of assessment scales and other instruments.

Table 3

Abbreviations and full names of assessment scales and other instruments.

Key Question 1

How do donepezil, galantamine, rivastigmine, tacrine, and memantine or combinations of these drugs (i.e., acetylcholinesterase inhibitor plus memantine) compare in their efficacy or effectiveness for stabilizing symptoms and treating behavioral disturbances in patients with Alzheimer's Disease?

We included 26 RCTs and 8 systematic reviews/meta-analyses. Of the RCTs, 3 were head-to-head trials; 23 were placebo-controlled trials. Only one trial was deemed to be an effectiveness trial.

A. Description of studies

We did not identify any head-to-head, randomized, double-blind, parallel-group study. Of the three head-to-head trials we identified,27–29 all were open-label trials blinding only the rater to treatment allocation; two trials27, 28 compared donepezil to galantamine and one trial29 compared donepezil to rivastigmine. We included two systematic reviews30, 31 that pooled placebo-controlled trials of donepezil, galantamine, and rivastigmine to represent ChEIs as a class. Several other systematic reviews pooled placebo-controlled trials for specific medications.32–37

Of the included placebo-controlled trials, 11 compared donepezil to placebo,32, 33, 38–48 6 compared galantamine to placebo,49–54 3 compared rivastigmine to placebo,55–57 1 compared tacrine to placebo,58 and 2 compared memantine to placebo.59, 60 Most trials were 3 months to 1 year in duration; one trial followed patients for more than 3.5 years.38 Only one trial was deemed to be an effectiveness trial.38 Doses generally were given within the range of the approved package labeling (see Table 1), although several galantamine trials used doses above the recommended 24 mg/day and rivastigmine trials commonly included a low dose arm of 1–4 mg/day.

B. Study populations

We included studies with a sample size greater than 100; the largest trial included in our review randomized 978 patients with probable AD.52 On average, the mean patient age was between 70 and 75 years; one trial was conducted in a nursing home population with a mean age of 86 years.47 Most studies were conducted in patients with mild to moderate AD; one donepezil40 and two memantine trials59, 60 were conducted in patients with moderate to severe AD. Most trials specifically excluded patients with vascular dementia and clinically significant neurologic disease other than AD. Some trials did not specify such exclusion criteria or report the proportion of patients with such comorbid disease. Most studies allowed patients to use other medications except for drugs with cholinomimetic effects or anticholinergic medications.

C. Outcome measures

Studies commonly included measures to assess symptom stabilization (e.g., cognition, global assessment of change) and behavioral disturbances. Most studies included a measure of cognition (e.g., ADAS-cog) as the primary outcome; other commonly included measures of cognition were the MMSE and SIB. Global change often was measured using scales such as the CGI-C, CIBIC-plus, or GDS; functional status was commonly assessed using measures such as the ADCS-ADL, DAD, Bristol ADL, and PDS. Changes in mood, behavior, and personality were assessed with measures such as the NPI or BEHAVE-AD. Some studies included other instruments that assessed quality of life or caregiver burden.

D. Head-to-head comparisons

We did not identify any randomized, double-blind, comparative trials. We did identify three open-label head-to-head trials.27–29 One trial compared donepezil to galantamine over 52 weeks,27 one compared donepezil to galantamine over 12 weeks,28 and one compared donepezil to rivastigmine over 12 weeks.29 These trials blinded only the rater to treatment allocation. Although open-label trials are subject to "fatal flaws" for internal validity, we review their results because they provide the only comparative evidence. We do not provide quality ratings for these trials.

We included two meta-analysis30, 31 that evaluated evidence comparing donepezil, galantamine, and rivastigmine with placebo. Although these reviews do not make indirect comparisons among included ChEIs, the quantitative summary of placebo-controlled trials is useful for summarizing evidence for ChEIs in general.

Donepezil vs. Galantamine

One 52-week open-label trial compared donepezil 10 mg/day to galantamine 24 mg/day in 182 patients with probable AD and MMSE scores between 9 and 18 at screening.27 Although raters were blinded to treatment allocation, patients, caregivers, and physicians were not blinded because of differences between the two study drugs in dosing frequency, escalation schedules, and physical appearance. On average, study participants were 73 years of age with a mean baseline MMSE score of 15. The primary study endpoint was based on function assessed by the Bristol ADL; cognitive outcome measures included the MMSE and ADAS-cog, behavioral disturbances were assessed with the NPI, and caregiver burden was measured using the SCGB scale. At endpoint no statistically significant differences were observed in functional abilities, cognitive symptoms, behavioral disturbances, or caregiver burden between the donepezil and galantamine treatment groups. This trial was funded by the makers of galantamine.

One 12-week open-label trial compared flexible doses of donepezil 5–10mg/day (once daily) and galantamine 8–24mg/day (twice daily) in 120 patients with probable or possible AD;28 as in the 52-week trial, only raters were blinded to treatment allocation. The mean age of study participants was 74 years with a mean baseline MMSE score of 21. On average, baseline MMSE scores for patients in this trial indicated less severe disease than in the 52-week trial. At baseline, patient demographics and disease characteristics were similar in both groups. The primary outcome measure was unblinded physician and caregiver satisfaction as measured on a scale specifically developed by the makers of donepezil for use in another head-to-head trial (presumably this instrument had not been previously validated).29 Secondary outcome measures included cognition (ADAS-cog, MMSE) and disability (DAD). At 12 weeks, both physician and caregiver satisfaction ratings were significantly better for donepezil (P < 0.001 and P < 0.01, respectively). Furthermore, donepezil-treated patients demonstrated significantly more improvement on the ADAS-cog, MMSE, and DAD (P < 0.05). In contrast to the 52-week study that demonstrated no difference between donepezil and galantamine, this trial was funded by the makers of donepezil. Additionally, this trial demonstrated the worst reported galantamine response among all other clinical studies. Both trials utilized similar dosing protocols.

Donepezil vs. Rivastigmine

One 12-week open-label trial compared flexible doses (5–10 mg/day) of donepezil to flexible doses (6–12 mg/day) of rivastigmine in 111 patients with mild to moderate AD.29 The mean age of study participants was 74 years with a mean baseline MMSE score of 20; 54% of donepezil-treated patients and 64% of rivastigmine-treated patients were female. With regard to baseline disease severity, patients in this trial most closely resembled the 12-week trial comparing donepezil to galantamine. Cognitive symptoms and disease severity were assessed with the ADAS-cog and MMSE, respectively. ADAS-cog raters were blinded to treatment allocation, but unblinded clinicians administered the MMSE. At 12 weeks no statistically significant differences in ADAS-cog or MMSE were reported for the two treatment groups. These investigators also administered an unidentified measure of clinician and caregiver satisfaction.

Although physicians and caregivers reported significantly higher scores on the satisfaction measure for donepezil than for rivastigmine, this measure was designed and initially used in this trial and had not been previously validated. This trial was funded by the makers of donepezil.

E. Placebo-controlled trials

We identified 8 systematic reviews or meta-analyses of placebo-controlled trials and 23 RCTs that met the inclusion criteria for our review of placebo-controlled evidence. When good-rated systematic reviews provided comprehensive evidence for a specific drug-placebo comparison, we did not include individual trials already covered in the systematic review. However, in cases where individual trials were too heterogeneous or not adequately described by existing systematic reviews (i.e., donepezil and memantine), we include these trials in our review in addition to the pooled analysis.

Donepezil, Galantamine, and Rivastigmine vs. Placebo (Meta-Analysis)

Two methodologically sound meta-analysis30, 31 evaluated placebo-controlled evidence for donepezil, galantamine, and rivastigmine. These reviews cannot be used to compare one drug to another directly, but quantitative analyses from these studies are relevant to the question of the general effectiveness of ChEIs as a class. The most recently published review31 included 22 trials. The authors attributed a 1.5 to 3.9 point reduction in ADAS-cog scores and a 0.26 to 0.54 point improvement in CIBIC-plus scores to the included drugs, citing serious methodologic flaws in this evidence base.31 The older review30 included 16 trials. The authors defined "global responders" as subjects rated as minimally to very much improved on the CGIC or CIBIC-plus; "cognitive responders" were defined as patients with a 4-point or greater improvement (decrease) from baseline on the ADAS-cog. Compared to placebo the pooled number needed to treat (NNT) to yield one additional ChEI global responder was 12 (95% CI 9–16); the NNT to yield one additional cognitive responder was 10 (95% CI 8–15). These pooled NNT calculations should be interpreted cautiously, as some heterogeneity exists among trials included in this analysis.

Compared to patients receiving placebo,30 significantly more patients receiving ChEIs had adverse events (8%; 95% CI 5%–11%), dropped out (8%; 95% CI 5%–11%), or dropped out because of adverse events (7%; 95% CI 3%–10%). Pooled rates of dropouts and adverse events were not reported for each drug. However, adverse event rates in excess of those for placebo were lowest for donepezil (6%; 95% CI 2%–9%), followed by rivastigmine (8%; 95% CI 1%–10%), and galantamine (12%; 95% CI 7%–18%). Similarly, drop out rates in excess of the rate for placebo were lowest for donepezil (3%; 95% CI 1%–6%), followed by rivastigmine (9%; 95% CI 5%–12%), and galantamine (14%; 95% CI 8%–21%). Drop out rates due to adverse events demonstrated a similar trend.

Donepezil vs. placebo

We included two meta-analyses32, 33 and 11 trials38–48, 61 comparing donepezil to placebo. A good meta-analysis pooled data from 13 trials lasting 12 or more weeks and involving 4,365 participants.32 Pooled results demonstrated statistically significantly better ratings for 5mg/day and 10mg/day donepezil on all outcomes measures at 24 weeks. For 10mg/day doses, the global assessment with CIBIC-plus, dichotomized into those showing no change or decline and those showing improvement yielded an odds ratio (OR) of 2.18 ( 95% CI 1.53 – 3.11; P < 0.001) and assessment of cognition with MMSE a weighted mean difference (WMD) of 1.50, (95% CI 0.97 – 2.04; P < 0.0001) and with ADAS-Cog a WMD of −2.92 (95% CI −3.74 –−2.10; P < 0.001). The size of the effect was dose-related and did not differ by severity of the disease. Furthermore, pooled data from two trials assessing activities of daily living (DAD, IADL, PSMS, CMCS) presented a statistically significant benefit for 5mg/day and 10mg/day donepezil treatment at week 12 and week 24. No difference was reported on a patient-rated Quality of Life Scale between donepezil and placebo. These findings were consistent with those of a fair-rated meta-analysis using individual patient data of placebo-controlled trials.33

Of 11 placebo-controlled trials that we examined, all but one46 had been included in the meta-analysis by Birks et al.32 Because some42, 43, 45 of these included trials provide specific results on quality of life and activities of daily living we summarize results in Table 4.38, 40, 44

Table 4. Summary of trials assessing symptoms and behavioral disturbances.

Table 4

Summary of trials assessing symptoms and behavioral disturbances.

The only effectiveness study we identified was the only trial on donepezil that was not funded by the pharmaceutical industry.38 This UK study enrolled 565 patients and assessed the effectiveness of long-term (3 years and 36 weeks) donepezil treatment in community-residents with mild to moderate AD with or without concomitant vascular dementia. Primary outcome measures were rate of institutionalism and functional capacity (Bristol ADL). No significant differences could be observed in the rates of institutionalism between donepezil and placebo at 1 year (9% vs. 14%; P = 0.15) and at 3 years (42% vs. 44%; P = 0.4). After 12 weeks until the end of the trial, the Bristol ADL scores of donepezil-treated patients were statistically significantly better, though the difference was modest (average +1.0 point, 95% CI 0.5 – 1.6; P = 0.0004). Similarly, MMSE scores were modestly but statistically significantly higher in donepezil- than in placebo-treated patients (average 0.8 points, 95% CI 0.5 – 1.2; P = 0.001); the clinical significance of these findings is questionable. No significant differences were detected in progression of disability (Bristol ADL) or behavioral and psychological symptoms (NPI).

A fair US-based study (n = 431) examined the functional decline of donepezil compared to placebo-treated patients over 1 year.42 The primary endpoint was time to clinically evident decline in function (defined in study protocol). A higher proportion of placebo than of donepezil- treated patients reached the primary endpoint (56% vs. 41%; P < 0.005). The median time to clinically evident functional decline was significantly shorter for placebo than for donepezil-treated patients on donepezil (208 vs. 380 days; P = 0.0051).

The placebo-controlled study not included in any meta-analysis assessed the efficacy of donepezil on cognitive outcomes46. Findings are consistent with results from meta-analyses. Results reported significantly better outcomes for the donepezil than for placebo groups after 24 weeks of treatment.

Galantamine vs. placebo

One good-rated systematic review34 (updated in 200562), two good-rated RCTs,49, 53 and four fair-rated RCTs50–52, 54 compared galantamine to placebo. We focus the majority of our discussion on the updated systematic review62 because it provides a comprehensive summary of four of the five RCTs identified in our search. However, for measures of behavior and functional capacity we focus our discussion on individual trials because data in these domains were not pooled in the systematic review.

Trials ranged from 12 to 52 weeks in duration. The most frequent galantamine dose tested was 24mg/day; in most trials patients began at 8 mg/day and increased over time to the daily maximum. Patients reached their maximum daily dose 2 to 8 weeks into the respective trials. All trials used the ADAS-cog to assess cognitive change; other measures of symptomatic change included the European adaptation of the ADAS scale, the expanded ADAS-cog, and the Digit Symbol Substitution Test. Most trials used global rating scales such as the CIBIC-plus or the ADCS-CGIC. Changes in behavior were assessed by the NPI and functional status was assessed using the PDS, DAD, and ADCS-ADL.

Overall, galantamine was significantly better than placebo for improving intermediate outcome measures of cognitive symptoms and global rating scales.34 Pooled analyses of ADAS-cog scores from trials lasting 5 to 6 months revealed statistically significant differences for all doses of galantamine compared to placebo (8mg: WMD −1.3; 95% CI −2.6–0.3; 16mg: WMD −3.1; 95% CI −4.1–−2.1; 24mg/day: WMD −3.3; 95% CI −3.9–−2.7; 32mg/day: WMD −3.3; 95% CI −4.1–−2.4). Results from trials of 3 months' duration were similar. Pooled ITT analyses for global rating scales also favored galantamine over placebo. Trials lasting 5 to 6 months demonstrated similar differences (16mg/day: OR 2.04; 95% CI 1.4–2.9; 24mg/day: OR 1.82; 95% CI 1.4–2.3; 32mg/day: OR 1.79; 95% CI 1.3–2.4), except for the 8mg/day dose, which was not significantly different from placebo. Trials lasting 3 months demonstrated statistically significant differences between galantamine and placebo on global rating scales for doses of 18mg/day (OR 2.44; 95% CI 1.2–5.0), 24mg/day (OR 2.11; 95% CI 1.0–4.6), and 36 mg/day (OR 2.7; 95% CI 1.2–6.2). The good-rated trial not included in the systematic review49 provided consistent results. The LOCF mean change in ADAS-cog from baseline to 26 weeks was −1.6 (± 0.36) for galantamine, −1.3 (± 0.31) for galantamine prolonged release capsule (PRC), and +1.2 (± 0.33) for placebo. Both galantamine and galantamine PRC were numerically superior to placebo in CIBIC-plus scores, but differences failed to reach statistical significance at 26 weeks.

Although most trials assessed behavior or functional status, the authors of the systematic review did not pool these data, presumably because of differences in study design and reporting. Evidence from individual trials is mixed. Two good-rated trials assessed activities of daily living with the ADCS-ADL scale; ITT results statistically favored galantamine over placebo at 26 weeks in both trials.49, 53 Another trial that assessed activities of daily living using the PDS found no significant differences between galantamine and placebo.54 Three trials measured disability using the DAD scale; one reported statistically significant differences between galantamine and placebo for doses of 24mg/day and 32mg/day,51 one reported statistically significant differences for doses of 32mg/day but not for 24mg/day,53 and one reported no differences for doses of 24mg/day or 32mg/day.50 One trial assessed sleep quality using the NPI sleep score and the PSQ1; no differences were found between galantamine and placebo on either measure.63 Three trials assessed behavioral symptoms using the NPI; two reported no statistically significant differences in NPI scores at 26 weeks49, 51 and the other reported statistically significant differences at 22 weeks for doses of 16mg/day and 24mg/day.52

Only one trial reported caregiver burden.52, 64, 65 This study reported the caregiver distress component of the NPI in a 22 week trial comparing galantamine 16mg/day and 24mg/day to placebo. At endpoint, only the 24mg/day dose was significantly better than placebo (P = 0.05).

No galantamine trial specifically reported the effect of drug treatment on rates of institutionalization or death.

Rivastigmine vs. placebo

One good-rated systematic review,32 one fair-rated systematic review,35 and three placebo-controlled trials were included in our review of rivastigmine.55–57 The good-rated systematic review included 3 published trials55, 57, 66 and 5 unpublished phase II and phase III clinical trials involving 3,450 patients. All trials but one66 were sponsored by rivastigmine's manufacturer. The fair-rated systematic review included data from two published trials55, 57 and one unpublished phase III clinical trial.

Although both systematic reviews included data from two of the same trials, we include them both because each study drew unique conclusions. However, because the Cochrane review received a better quality rating and was more comprehensive, we believe the good-rated Cochrane review gives the best overall summary.

The good-rated systematic review32 included data from eight trials; studies ranged in duration from 9 to 26 weeks. In most trials, the mean baseline MMSE score was between 18 and 20. Analyses were stratified by dose, characterizing rivastigmine 1–4mg/day as low dose and rivastigmine 6–12mg/day as high dose. Common outcome measures included the ADAS-cog, CIBIC-plus, GDS, MMSE, and PDS. Caregiver activities also were assessed using the CAS. Pooled results suggest significantly greater improvement on the CIBIC-plus for all doses of rivastigmine compared to placebo. Significantly greater improvement also was found for high-dose rivastigmine (6–12mg/day) compared to placebo on the ADAS-cog, MMSE, GDS, and the PDS; pooled results were not significant for low-dose rivastigmine (1–4mg/day) for these outcome measures. The high-dose regimen currently is the recommended dosing range for rivastigmine.

The fair-rated systematic review35 included data from two published trials55, 57 and one unpublished trial (B351). In contrast to the good-rated Cochrane review,32 this review reported statistically significant differences favoring all doses of rivastigmine compared to placebo. Statistically significant differences were reported for the ADAS-cog, CIBIC-plus, GDS, MMSE, and PDS. Although this pooled population includes data from the three largest placebo-controlled trials conducted by Novartis, it does not include a similarly designed phase III trial (i.e., B304). Furthermore, this review presents observed cases analyses for the ADAS-cog and CIBIC-plus but uses LOCF analyses for the PDS. The less conservative LOCF data may allow the natural course of the disease to overestimate treatment effect. This review was funded by Novartis, the makers of rivastigmine.

To contrast differences in the pooled evidence from these reviews, we review data from three published placebo-controlled trials that met the criteria for our review.55–57 In contrast to the Cochrane review,32 one trial found statistically significant differences in the ADAS-cog and GDS for all doses of rivastigmine compared to placebo;55 a second reported statistically significant differences in these measures only for rivastigmine 6–12mg/day (but not 1–4mg/day).57 A third trial56 reported statistically significant differences in cognitive and behavioral measures between rivastigmine 6mg/day and placebo; similar differences were not observed for patients treated with rivastigmine 4mg/day.

No rivastigmine trial specifically reported the effect of drug treatment on caregiver burden, institutionalization, or death.

Tacrine vs. placebo

A fair-rated meta-analysis pooled individual patient data on 1,984 patients with probable AD from 12 published and unpublished placebo-controlled trials.36 Dosages in the component trials varied from 20 mg/day to 160 mg/day. Trials lasted 3 to 36 weeks. Pooled results at 12 weeks presented a small beneficial effect of tacrine over placebo for cognitive function (MMSE: + 0.62 points, 95% CI: 0.23 – 1.00; P = 0.002), clinical global impression (CGI: OR 1.58, 95% CI: 1.18–2.11; P = 0.002), and behavioral disturbance (ADAS: 0.58 points, 95% CI: 0.17–1.00; P = 0.006). No significant difference could be detected in functional autonomy at 6 weeks (PDS: 0.75 points, 95% CI: −0.34 – 1.93; P = 0.21). The authors did not report if the component studies were critically appraised for methodological quality before inclusion. In studies without a dose titration phase (i.e., no active drug run-in phase before randomization) before the efficacy study, significantly more patients on tacrine than on placebo withdrew from the study (OR: 3.63, 95% CI: 2.80 – 4.71; no absolute numbers reported).

Four placebo-controlled trials met our eligibility criteria.58, 67–69 We excluded three of these studies for poor methodological quality because of high overall67, 68 or high differential69 loss to follow-up. In all three trials, the high attrition rate reflected frequent adverse events, in particular elevated liver function tests in tacrine-treated patients. The fourth study compared three fixed dosing regimens (20mg/day, 40mg/day, 80mg/day) to placebo in 468 patients with mild to moderate Alzheimer's disease.58 We were unable to determine the differential loss to follow-up from the provided data. Thus, differential loss to follow-up may exceed our cut-off level of 15 percentage points. The differential loss to follow-up because of adverse events in this study was 18 percentage points (placebo: 7%; tacrine: 25%). Efficacy results reported statistically significant improvements only for tacrine at 80 mg/day on the CGIC (P = 0.015), ADAS-total (P = 0.029), and caregiver-rated CGIC (P = 0.028) compared to placebo. No significant differences could be detected for ADAS-cog, MMSE, PDS, or for dosages less than 80 mg/day on CGIC.

No tacrine trial specifically reported the effect of drug treatment on caregiver burden, institutionalization, or death.

Memantine vs. placebo

Two fair-rated RCTs59, 60 comparing memantine to placebo met the inclusion criteria for our review. Although we identified one good-rated systematic review,70 it included only one of the two RCTs that met our inclusion criteria so we do not discuss it further.

Both placebo-controlled trials randomized moderate to severe AD patients to memantine 20mg/day or placebo.59, 60 One trial required patients to be receiving stable treatment with donepezil prior to randomization,60 and thus cannot be directly compared to the trial that did not allow concomitant use of donepezil. Population demographics were similar across trials. Outcome measures consistently used in both trials included the CIBIC-plus, SIB, ADCS-ADL, and NPI. In both trials, memantine-treated patients did significantly better on the SIB and ADCS-ADL than placebo-treated patients (the primary outcome measures in both trials). However, only patients randomized to both memantine and donepezil faired significantly better on the CIBIC-plus and NPI than patients randomized to placebo plus donepezil.60 In the memantine monotherapy study, no differences in MMSE, CIBIC-plus, GDS, or NPI were reported between memantine- and placebo-treated patients.

Both included trials assessed caregiver burden.59, 60 One trial incorporated a resource utilization scale,59 and the other trial used a behavioral rating scale (BGP) that assesses caregiver dependence.60 Both trials showed significantly greater improvement in caregiver burden (P < 0.01) for memantine compared to placebo.

F. Summary of the evidence

Comparative evidence for drugs used to treat AD is limited to three open-label head-to-head efficacy trials; two trials compared donepezil to galantamine27, 28 and one compared donepezil to rivastigmine.29 Evidence for the comparison of donepezil with galantamine is mixed. In one 52-week trial,27 donepezil and galantamine did not differ in stabilizing symptoms or improving behavior and functional status. In a shorter trial (12 weeks),28 donepezil was superior to galantamine in its effects on cognition, functional status, and caregiver and clinician satisfaction. The comparison of donepezil to rivastigmine is limited to a single 12-week trial;29 it produced similar improvement in cognitive scores for both drugs, although clinician and caregiver satisfaction ratings were significantly better for donepezil. Because of limitations in the quantity of evidence, design of available trials (i.e., open-label), use of outcome measures not previously validated in AD populations (e.g., caregiver satisfaction), suspicious directionality of findings favoring the funding drug company, and the minimal differences observed between compared drugs (i.e., clinical significance of differences is inconclusive), we conclude that the evidence is inadequate to draw conclusions about the effectiveness of one AD drug compared to another.

Evidence from placebo-controlled trials and systematic reviews of placebo-controlled trials provide general evidence of the efficacy and effectiveness of these drugs. Overall, the ChEIs as a class are modestly effective in reducing the rate of decline in cognition.30, 31 The NNT to yield one additional ChEI (excluding tacrine) global responder is 12; the NNT to yield one additional cognitive responder is 10.30 Evidence from placebo-controlled trials and a systematic review of placebo-controlled trials provide general evidence of the efficacy of memantine.

Evidence from one placebo-controlled effectiveness trial38 and 22 efficacy trials39–60, 64, 65 supports modest effects on symptom stabilization, behavior, and functional status as measured by various scales. Although some trials did not support statistically significant differences between active treatment and placebo on all outcome measures,38, 39, 42, 45, 48–51, 54, 58, 59 most trials yielded data supporting a slower rate of decline or modest improvement in measures of cognition and global assessment. Fewer trials supported differences in measures of behavior or functioning. Caregiver burden was infrequently assessed or reported, although 4 trials found significantly greater improvement for active treatment compared to placebo.52, 59, 60, 64, 65 Only one study assessed nursing home placement as a function of medication treatment.38 This trial did not detect significant differences in institutionisations between donepezil and placebo after 1 and 3 years.

The clinical significance of some statistical differences is controversial. Although some trials defined clinical and global responders a priori, inconsistencies in trial design and reporting make it difficult to assess the clinical relevance of differences across trials.

Overall, the quality of evidence of general efficacy of ChEIs and memantine is fair; the quality of evidence of effectiveness of ChEIs and memantine is limited to one study on donepezil38 and therefore poor. On the basis of current evidence, we cannot demonstrate substantial differences in efficacy between one AD drug and another.

Key Question 2

How do donepezil, galantamine, rivastigmine, tacrine, and memantine compare in their time to effect and in the time required to assess the clinical response?

We did not identify any study that directly compared the time to effect or time required to assess the clinical response of one AD drug compared to another. One open-label head-to-head trial provides evidence on the time to effect between donepezil and galantamine.28 The study reports a trend favoring donepezil in cognition at weeks 4 and 8 that reached statistical significance at week 12 (P < 0.05).28 DAD scores were significantly greater in donepezil-treated patients at weeks 4 and 12. Other head-to-head trials reported only long-term outcomes.

Placebo-controlled trials are too heterogeneous with respect to study design, outcomes assessment, and populations to allow any inferences about the comparative time to effect. Given that the overall placebo-controlled evidence indicates that long-term treatment with ChEIs and memantine will produce only modest beneficial effects on cognition and global assessment, the clinical significance of time to effect is likely to be of minimal importance to physicians and patients.

In general determining time to effect and time required to assess clinical response is difficult, given the design of most trials and the nature of measurement scales. First, trials commonly were not designed to measure the time required to produce a statistically different response. In most trials, the first follow-up visit was not conducted until 4 to 12 weeks after randomization. Given this relatively large and inconsistent gap in follow-up between randomization and first clinical assessment, comparison across placebo-controlled trials cannot provide accurate information. Second, different studies used different outcome scales that are not necessarily comparable to assess effect sizes. Third, the ability of a trial to detect statistically significant difference depends on the sample size of each respective trial; trials with large sample sizes have greater power to present statistically significant findings at earlier time points.

Interpretation of clinical response (and time to assess it) is also of concern. Three published studies have sought to shed light on the clinical significance of treatment effects in AD trials.51, 78, 79 In one78 the authors calculated standardized effect sizes from ChEI trials to assess clinically detectable responses. Effect sizes greater than 0.20 were considered to be clinically detectable, but one cannot determine from the article if this assumption was derived from validated evidence. In another study using a survey of specialists, the investigators established a change in MMSE score of 3.72 points as a clinically significant difference.

Most of the included studies in this report have used arbitrary cut-off points on cognitive measures such as the ADAS-cog (≥ 4 points improvement from baseline) to define a clinical response. Others have considered any improvement on global assessment scales such as the CGI-C or the CIBIC-plus to define a clinical response. These definitions are arbitrary and have not been validated; consequently, comparisons across trials are impossible.

One generic indicator that influences time to effect is the time to titration of therapeutic dose. Statistically significant differences between donepezil and placebo were reported in most trials for 5mg and 10mg daily doses; because the recommended starting dose of donepezil is 5mg/day (titrating to 10mg/day at 4 to 6 weeks), this finding suggests that donepezil-treated patients are given a therapeutic dose from day 1 of treatment (although steady state of therapeutic concentrations is not achieved for approximately 2 weeks). Titration of rivastigmine-treated patients to a therapeutic dose (i.e., 6mg to 12mg/day) is recommended at week 2, again inferring a relatively short time to therapeutic dose. Conversely, patients treated with galantamine, tacrine, or memantine typically are not titrated to therapeutic doses until 3 weeks or later. Although titration schedules are designed to minimize potential adverse events, some patients may be titrated sooner than recommended. Furthermore, titration schedules do not reflect the time it takes to maintain steady state concentrations. Given the typically long natural course of disease and the modest treatment effects, the clinical significance of these differences is questionable, however.

Key Question 3

What are the comparative incidence and severity of complications of donepezil, galantamine, rivastigmine, tacrine, and memantine?

In general, adverse events depend on dose and mechanism of action for individual AD drugs. In most trials assessing a range of doses specific adverse events were reported more frequently among patients randomized to higher doses of study drugs.43, 44, 50, 52–55, 57, 58 In some trials the speed of dose titration also was believed to be related to greater reporting of adverse events.55, 57

Based on three open-label head-to-head trials,27–29 evidence suggests some differences between compared drugs. In one 12-week trial comparing donepezil with rivastigmine,29 gastrointestinal-related adverse events were significantly more common among rivastigmine-treated patients; nausea and vomiting were reported by 41.8% and 23.6% of rivastigmine-treated patients compared to 10.7% and 7.1% of donepezil-treated patients, respectively (P = NR). Two trials compared donepezil to galantamine; the evidence is mixed. The incidence of gastrointestinal-related adverse events was not different in a 52-week trial comparing donepezil and galantamine.27 In one 12-week trial gastrointestinal-related events were reported by 46.4% of patients in the galantamine group compared to 25% of patients in the donepezil group.28

Indirect comparisons based on evidence from placebo-controlled trials are difficult to make given differences in trial design, study populations, and assessment and reporting of specific events. Overall, adverse events were reported by 40% to 96% of randomized patients. In general ChEI- and memantine-treated patients appear to report a similar number of adverse events, although evidence is insufficient to compare the incidence of specific adverse events across drugs. Overall discontinuation rates are similar among memantine and ChEIs except for tacrine.

Table 5 presents the mean incidence of specific adverse events based on data provided by placebo-controlled trials of ChEIs and memantine. Statistics are descriptive only. Comparisons across different drugs are limited and should be made with caution. Large confidence intervals for some estimates indicate lack of precision due to a small number of component studies for some medications.

Table 5. Mean incidence of specific adverse events in placebo-controlled trials.

Table 5

Mean incidence of specific adverse events in placebo-controlled trials.

Some of the ChEIs appear to have a higher incidence of vomiting; this is likely attributable to their cholinergic mechanism of action. The highest incidence of nausea and vomiting was reported in rivastigmine trials, although these trials utilized a faster titration schedule than recommended by the product labeling and the rate of adverse events was also higher than normal in the placebo group. However, these estimates are consistent with available comparative evidence, which suggest that the risk of gastrointestinal-related adverse events is greater with rivastigmine and galantamine than with donepezil.

The incidence of specific adverse events reported by memantine-treated patients was generally low. From the single trial in our review that assessed only memantine59 (rather than memantine plus donepezil),60 the only adverse events reported by more than 10% of memantine-treated patients were agitation, diarrhea, somnolence, and urinary incontinence; no adverse event was reported in significantly more memantine-treated patients than in placebo-treated patients. The rate of agitation was significantly different in memantine-treated patients than in those on placebo, although significantly more placebo- than memantine-treated patients reported agitation (32% vs. 18%, respectively; P = NR). Urinary tract infections also were more common in placebo-treated patients than in memantine-treated patients (13% vs. 6%, respectively; P = NR).

Discontinuation rates varied across trials. Based on one trial that compared donepezil to rivastigmine,29 more patients randomized to rivastigmine than donepezil discontinued treatment (30.9% vs. 10.7%, respectively; P = NR). Two open-label trials compared donepezil to galantamine;27, 28 overall discontinuation rates did not differ significantly. Trials assessing tacrine consistently reported significantly higher discontinuation rates for tacrine than for placebo patients.58, 67–69 The high withdrawal rates were mainly attributable to elevated serum alanine aminotransferase (ALT; a feature of liver toxicity).

Withdrawals because of adverse events in donepezil, galantamine, rivastigmine, and memantine trials varied. Evidence from one open-label head-to-head comparison of donepezil and rivastigmine29 suggests a higher number of withdrawals due to adverse events among rivastigmine-treated compared to donepezil-treated patients (21.8% vs. 10.7%, respectively; P = NR). Based on two open-label trials comparing donepezil and galantamine, withdrawals due to adverse events were higher among galantamine-treated patients than among donepezil-treated patients in one 12-week trial (21.4% vs. 9.4%, respectively; P = NR),28 but not in a 52-week trial (13.4% vs. 13.2%, respectively; P not significant).27 From placebo-controlled evidence, no obvious trend favored one drug over another. Patients treated with higher doses were more likely to discontinue because of an adverse event. A meta-analysis of discontinuation rates did not find a statistically significant difference between donepezil and placebo,32 even though the incidence of anorexia, diarrhea, dizziness, fatigue, insomnia, muscle cramps, nausea, vomiting, tremor, vertigo, and weight loss were statistically significantly more common in the donepezil than in the placebo group.

We did not identify any study that assessed temporary or permanent adverse events due to discontinuation of donepezil, galantamine, rivastigmine, tacrine or memantine.

A. Specific adverse events

Hepatotoxicity

A major safety concern of tacrine treatment is hepatotoxic effects. A retrospective review of tacrine-trials involving 2,446 AD patients reported that 49% of tacrine-treated patients had elevated ALT levels.80 Among all patients, 25% presented an ALT elevation three times the upper normal limit; 2% had ALT levels 20 times higher than normal. Patients with elevated ALT levels were generally asymptomatic, although sometimes they experienced eosinophilia, rash, and fever. Few patients developed signs of severe hepatocellular injury (e.g., jaundice); no death attributable to liver toxicity was reported.

Results of this retrospective analysis are consistent with individual trials included in this review. All four placebo-controlled RCTs of tacrine reported high elevations of ALT.58, 67–69 We excluded three of these studies from the efficacy analysis on grounds of quality because of high overall67, 68 or differential69 loss to follow-up. In all three trials the high drop-out rate was attributable to a high rate of elevated liver function tests in tacrine-treated patients. The differential loss to follow-up because of adverse events in the fourth study58 was 18 percentage points (placebo: 7%; tacrine: 25%). Hepatotoxicity has not been reported for donepezil, galantamine, rivastigmine, or memantine.

Gastrointestinal adverse events and loss of body weight

ChEI trials commonly reported nausea and vomiting by more than 10% of patients (and as many as 50% of patients) randomized to active treatment. In the only memantine trial the incidence of nausea and vomiting did not differ between the active drug and placebo. Nausea, vomiting, and diarrhea are thought to reflect excessive activation of intestinal muscarinic cholinergic receptors and tend to be dose related. Anorexia and loss of body weight are associated gastrointestinal adverse events.

We did not find any trials directly comparing the incidence of gastrointestinal adverse events among ChEIs and memantine.

In a systematic review of donepezil, galantamine, and rivastigmine trials,14 nausea and vomiting were 3 to 5 times more common in patients randomized to active treatment compared to placebo (P < 0.0001). The odds of having nausea or vomiting with rivastigmine compared to placebo (OR 5.28; 95% CI 4.19–6.65) were consistently higher than with donepezil or galantamine compared to placebo (donepezil OR 2.73; 95% CI 1.86–4.00; galantamine OR 3.01; 95% CI 2.15–4.21), although this finding could likely be attributed, at least in part, to the faster than recommended dose titration used in rivastigmine trials.55, 57

Diarrhea was also common in the pooled analysis,14 although the pooled odds ratio was significant for donepezil and rivastigmine (donepezil OR 2.83; 95% CI 2.01–4.00; rivastigmine OR 1.77; 95% CI 1.38–2.28) but not for galantamine (OR 1.37; 95% CI 0.91–2.05). The higher incidence of gastrointestinal events may be related to the significant loss of body weight commonly reported for donepezil-, galantamine-, and rivastigmine-treated patients. Pooled analysis suggests a 2- to 4-fold increase in the risk of anorexia for active treatment compared to placebo. Although tacrine was not included in this analysis, relative trends in gastrointestinal adverse events and loss of body weight reported in tacrine trials are consistent with those seen in donepezil, galantamine, and rivastigmine trials.58, 67–69

A retrospective data review of the mean incidence rates of gastrointestinal adverse events of some RCTs shows that the following percentages of patients suffered nausea: donepezil, 11%; rivastigmine, 35%; and tacrine, 28%.81 Similarly, the relative proportions of patients who experienced vomiting were 5%, 21% and 28%, respectively; diarrhea occurred in 10%, 16% and 16%, respectively. Another review reported a loss of body weight of 0.5 to 2.5 kilogram for galantamine at doses of 16mg/day to 32mg/day and a loss of body weight of 1.39 to 1.78 kilogram for rivastigmine at doses of 6 mg/day to 12 mg/day.82

Data from the Réseau sur la Maladie d'Alzheimer Francais (REAL.FR) cohort was used to assess the risk of weight loss with AChEI.83 This long-term observational study found the risk of clinically significant weight loss to be similar for Alzheimer's patients taking AChEIs and patients not taking these drugs (21.1% vs 19.5%, respectively; P = 0.81). However, we excluded this study for reasons of quality because we were unable to assess the similarity or differences between the two populations, and little information was provided with regard to the type, intensity, or duration of drug treatment.

Cardiovascular adverse events

Bradycardia and subsequent dizziness or syncope originates from central and peripheral muscarinic cholinergic stimulation. Cardiovascular adverse events can lead to falls and other types of injury-causing accidents. We did not find any trials directly comparing the incidence of cardiovascular adverse events among ChEIs and memantine.

Cardiovascular adverse events may be of particular concern in patients with cardiac conduction disorders or a sick sinus syndrome. One head-to-head study reports no statistically significant differences in changes of heart rates between donepezil and galantamine.28 Two open-label comparative trials reported no difference in cardiovascular events between donepezil and galantamine28 and rivastigmine.29 Most placebo-controlled trials revealed no other significant differences in cardiovascular events, vital signs, or electrocardiogram (ECG) findings. One trial described a statistically significantly larger reduction of heart rate in patients treated with donepezil than in those given placebo.43 However, the incidence of bradycardia (heart rate < 50 beats per minute) was not significantly different among treatment groups. An analysis of prescription-event monitoring (n = 1,762) in general practice in the UK did not find evidence for cardiac arrhythmias with donepezil treatment.84

One pooled data-analysis of RCTs including 2,791 patients evaluated ECG results from four clinical trials of rivastigmine;85 rivastigmine had no apparent effect on heart rate. However, patients with underlying ECG abnormalities did not meet eligibility criteria of the RCTs.

B. Summary of the evidence

The overall grade of the evidence on comparative tolerability is poor to fair. Evidence of the comparative incidence of adverse events and tolerability comes from three open-label trials comparing donepezil with galantamine and rivastigmine. One 52-week trial27 and one 12-week trial28 compared donepezil to galantamine. Although the number of adverse events and loss to follow-up differed between trials, withdrawals and withdrawals because of adverse events were not significantly different in the 52-week trial and only minor differences favoring donepezil were observed in the 12-week trial. In one trial that compared donepezil to rivastigmine,29 total withdrawals and withdrawals because of adverse events were significantly greater among rivastigmine-treated patients. Gastrointestinal-related events were most commonly reported among rivastigmine-treated patients. Indirect comparison of the pooled mean incidence of adverse events from placebo-controlled trials also suggests a higher rate of gastrointestinal-related events among rivastigmine-treated patients. However, this comparison is limited by the tremendous variability observed among placebo-controlled evidence.

Evidence of hepatotoxicity and cardiovascular events comes from comparative trials, meta-analyses, and indirect comparison of placebo controlled evidence. Evidence from one meta-analysis and four placebo-controlled trials indicate substantially higher rates of hepatotoxicity for tacrine.67, 80 Donepezil, galantamine, rivastigmine, and memantine did not present hepatotoxic effects in placebo controlled trials. Two open-label comparative trials reported no difference in cardiovascular events between donepezil and galantamine28 and rivastigmine.29 Placebo-controlled trials revealed no other significant differences in cardiovascular events.

Key Question 4

Does efficacy, effectiveness, or adverse events of donepezil, galantamine, rivastigmine, tacrine, or memantine differ in subgroups of patients with (1) different demographic profiles (age, race, or sex), (2) Parkinsonian features or vascular dementia, or (3) use of other commonly prescribed drugs?

A. Age

We did not identify any study specifically designed to compare the effect of donepezil, galantamine, rivastigmine, tacrine, or memantine in a younger versus an older population.

We did find age-related information in two sources: one subgroup analysis of rivastigmine-treated patients35 and a placebo-controlled donepezil trial conducted in a population of nursing home residents who were, on average, older than the typical population for donepezil studies.47 The subgroup analysis pooled data from four rivastigmine trials and reported an age-related treatment effect. Patients 75 years and older revealed a greater benefit of rivastigmine than did patients younger than 75 years; 15% of older patients and 6% of younger patients were considered responders on the ADAS-cog.35

A single trial47 conducted in nursing home residents with a mean age of 85 years (range 64 to 102 years) provides indirect evidence about age effects when compared to findings from other similarly designed trials in which the mean age was less than 75 years.40, 41, 43–45, 48 Overall, no difference in efficacy or adverse events was apparent in the data on the older population compared to data from the trials in younger populations.

B. Race

We did not identify any study specifically designed to compare the effect of donepezil, galantamine, rivastigmine, tacrine, or memantine in one racial group compared to another. In general, trials were conducted predominantly in white populations.

One study used pooled data from 2,126 patients in three placebo-controlled rivastigmine trials to analyze differences in efficacy among racial subgroups.35 The pooled population was 93.6% white, 4.4% black, and 2% other races. Treatment response did not differ across racial subgroups.

One donepezil trial41 was conducted in a Japanese population and one tacrine trial67 was conducted in a Chinese population. Overall, effect sizes observed in these trials are similar to effect sizes reported in trials conducted predominantly in non-Asian populations. However, the trial conducted in the Japanese population presented treatment effects on low-dose donepezil, which suggests ethnic differences in major enzymes that metabolize ChEIs.30 This finding was supported by a meta-analysis of ChEIs.30

C. Sex

We did not identify any study specifically designed to compare the effect of donepezil, galantamine, rivastigmine, tacrine, or memantine in females compared to males. On average, study populations comprised more females than males; this fact reflects population and disease demographics and does not provide insight into treatment differences.

One review35 of pooled data from rivastigmine trials conducted a subgroup analysis by sex but reported no differences. No other indirect evidence suggests that effectiveness or adverse events differ between females and males.

D. Parkinsonian features

Dementia with Parkinsonian features, or dementia with Lewy bodies (DLB), is characterized by abnormal protein inclusions (Lewy bodies) in selected areas of the brain. Because these structures, and many of the symptoms of dementia with Lewy bodies, are associated with Parkinson's and Alzheimer's diseases, it remains unclear whether DLB is a distinct clinical entity or perhaps a variant of Alzheimer's or Parkinson's disease.

We did not identify any trial conducted in patients with AD that compared effectiveness or adverse events in a population with Parkinsonian features to a population without Parkinsonian features. Although some trials specifically excluded patients with suspected PD, trials that did not specifically exclude patients with Parkinsonian features did not report differences among these patients.

Evidence from a recently published large-scale placebo-controlled study supports the general efficacy of rivastigmine in treating patients with PD dementia.89 This 24-week multicenter European study enrolled 541 subjects with PD dementia (defined as the onset of cognitive symptoms 2 or more years after the onset of PD) who were randomized to either placebo or rivastigmine (1:2 ratio) beginning at 1.5mg twice a day and increased at 4-week intervals as tolerated up to 12 mg/day. Primary efficacy analyses showed better ADAS-cog scores and global ratings in the rivastigmine-treated group compared to placebo group.

E. Comorbid vascular dementia

Vascular dementia is the second most common form of dementia. In many patients with AD, vascular factors contribute to the development or expression of dementia. Mixed vascular dementia includes those patients that have clinical features of AD and clinically significant cerebrovascular disease. Most studies included in our review specifically excluded patients with mixed vascular dementia; studies that did not explicitly exclude patients with comorbid cerebrovascular disease often did not report the prevalence or stratify the results for this subgroup.

Although evidence is difficult to interpret given the inconsistencies in trial design and lack of differentiation between AD and vascular dementia, we discuss four studies that provide general evidence of the efficacy of donepezil,38 galantamine,90 rivastigmine,91 and memantine70 in populations with comorbid vascular dementia.

The only effectiveness study included in our review38 randomized patients with or without a coexisting diagnosis of vascular dementia to long-term treatment with donepezil or placebo. Although results are not stratified by coexisting vascular dementia, results support the general efficacy of donepezil in this mixed population.

One placebo-controlled RCT90 examined the effect of galantamine in patients with probable vascular dementia and AD with cerebrovascular disease. This 26-week trial randomized 592 patients to galantamine or placebo in a 2:1 ratio. Diagnosis of vascular dementia was based on National Institute of Neurological Disorders and Stroke and the Association Internationale pour la Recherche et l'Enseignement en Neurosciences (NINDS-AIREN) International Workshop criteria; computed tomography or magnetic resonance imaging was used to confirm evidence of cerebrovascular disease. Overall, galantamine was significantly better than placebo (P < 0.05) on cognitive, functional, behavioral, and global assessment measures. Treatment differences were of similar size to those seen in galantamine studies in patients with AD.50, 52, 53 Galantamine was significantly better than placebo (P < 0.05) only in the subgroup of patients with AD and cerebrovascular disease. Although the study was not powered to detect treatment differences in the subgroups, differences between galantamine and placebo were not significant in patients with vascular dementia.

We identified one subgroup analysis of AD patients with concurrent vascular risk factors from a placebo-controlled RCT of rivastigmine.91 Patients from this trial55 were categorized by their Modified Hachinski Ischemic Score (MHIS); MHIS scores greater than zero were used to identify the presence of vascular risk factors. At 26 weeks, rivastigmine was significantly better than placebo on cognitive, functional, and global assessment measures for patients with and without vascular risk factors. Larger treatment differences between rivastigmine and placebo were found for patients with vascular risk factors.

One systematic review70 of placebo-controlled memantine trials included trials conducted in populations with AD, vascular dementia, and mixed or unspecified AD with vascular dementia. Although individual trials were different with regard to design, duration, dose, and outcome measures, comparison of evidence across populations suggests that results of trials conducted in populations with mixed or unspecified vascular dementia are similar to trials conducted in populations with AD only.

F. Other drugs

We did not identify any published study that specifically compared outcomes among subgroups of patients taking a ChEI or memantine concurrently with another drug to patients not concurrently taking the same medications. To characterize potential and known drug-drug interaction risks as much as possible in this situation, we summarize indirect evidence and pharmacokinetic properties.

In general, ChEIs (i.e., donepezil, galantamine, rivastigmine, tacrine) may interfere with the activity of anticholinergic medications. Likewise, a synergistic effect may be expected when ChEIs are given with cholinomimetics or other ChEIs. Concurrent use of such drugs should be approached with caution.

The NMDA antagonist memantine is believed to be safe when administered in combination with a ChEI. In a 24-week trial, memantine was safely administered in combination with donepezil60 without evidence of altering the pharmacokinetic properties of either drug; evidence of additional benefit of this combination is not clear

The potential for other drug-drug interactions with donepezil, galantamine, rivastigmine, tacrine, and memantine should be evaluated on an individual basis. Pharmacokinetic parameters and information submitted to the FDA for approval provide useful information.

Donepezil

Donepezil is metabolized by CYP450 isoenzymes 2D6 and 3A4. Because other drugs may compete for or inhibit these metabolic enzymes, a potential for interaction exists with drugs metabolized by the same isoenzymes. Although to our knowledge no in vivo studies have been conducted, in vitro evidence suggests that donepezil has little effect on the metabolism of other drugs (e.g., theophylline, cimetidine, warfarin, digoxin, etc.). Drugs that inhibit 2D6 and 3A4 (e.g., ketoconazole, miconazole, quinidine, ritonavir, selective serotonin reuptake inhibitors [SSRIs], etc.) have been shown to inhibit donepezil metabolism but clinically significant interactions are rare. Patients taking donepezil in combination with other drugs metabolized by CYP450 isoenzymes 2D6 and 3A4 should be monitored closely.

Although donepezil is highly protein bound (96%) drug displacement studies performed in vitro have shown little effect of other highly bound drugs on the binding of donepezil to human albumin. Similarly, donepezil did not affect binding of other drugs to human albumin.

Galantamine

Like donepezil, galantamine is metabolized by CYP450 isoenzymes 2D6 and 3A4. In vivo studies have shown increased bioavailability of galantamine when it is administered together with inhibitors of these isoenzymes (e.g., cimetidine, ranitidine, ketoconazole, erythromycin, paroxetine). By contrast, galantamine is believed to have little effect on other drugs metabolized by the CYP system.

Rivastigmine

Because rivastigmine is metabolized primarily through hydrolysis by esterases, minimal interaction with drugs metabolized by CYP450 enzymes is anticipated. No other drug-drug interactions have been demonstrated.

In a subgroup analysis of nicotine users randomized to rivastigmine, a statistically significant relationship in the dose-response relationship was reported;35 this analysis suggests that nicotine attenuates the benefits of rivastigmine. Another post-hoc analysis of 2,459 patients from 4 placebo-controlled rivastigmine trials evaluated drug interactions with 22 classes of medications.92 This analysis did not reveal any significant pattern of increase in adverse events that would indicate a drug-drug interaction.

Tacrine

Tacrine is metabolized primarily by the CYP450 isoenzyme 1A2. Drug-drug interactions may occur with other medications metabolized by this enzyme (e.g., theophylline). Administration of tacrine and theophylline has been shown to increase average plasma theophylline concentrations 2-fold. Likewise, administration of cimetidine with tacrine has been shown to increase plasma concentrations of tacrine.

Memantine

Because memantine is eliminated predominantly by the kidney, drugs that are inhibitors and/or substrates of the CYP450 system are not expected to interact with it. However, because memantine is eliminated via renal mechanisms, concurrent administration of drugs that use the same renal mechanisms (e.g., hydrochlorothiazide, triamterene, cimetidine, ranitidine, quinidine, nicotine) could alter the plasma levels of both agents. Additionally, drugs that make the urine alkaline (e.g., sodium bicarbonate, carbonic anhydrase inhibitors) may reduce the clearance of memantine. Patients using these drugs and memantine concurrently should be monitored closely.

G. Summary of the evidence

The overall grade of the evidence on efficacy and tolerability in subgroups is poor. We did not identify any study specifically designed to compare the effect of donepezil, galantamine, rivastigmine, tacrine, or memantine in one subgroup of patients compared to another. Subgroup analyses and indirect evidence from placebo controlled trials provide evidence for some AD drugs.

One subgroup analysis reported greater benefit for rivastigmine in patients older than 75 years. Indirect comparison of evidence from one donepezil trial conducted in nursing home residents to trials conducted in younger populations suggests no apparent difference in efficacy or adverse events.

Subgroup analyses of pooled data from four rivastigmine trials suggest no differences in efficacy or adverse events by sex or race.

No evidence addressed patients with comorbid PD.

Four studies provide general evidence of the efficacy of donepezil, galantamine, rivastigmine, and memantine in populations with comorbid vascular dementia. Only one study stratified patients by vascular risk factors; larger treatment differences between rivastigmine and placebo were found for patients with vascular risk factors compared to patients without vascular risk factors.

No study compared outcomes among subgroups of patients taking a ChEI or memantine concurrently with another drug to patients not concurrently taking the same medication.

Table 6Summary of trials assessing adverse events

Author, YearNStudy designResultsQuality Rating
Hepatotoxicity
Watkins et al., 1994802446secondary data review49% of tacrine-treated patients presented ALT elevationsN/A
Farlow et al., 199258468RCT25% of tacrine-treated patients had elevated ALT levelsFair
*Knapp et al., 199468, 86–88663RCT54% of tacrine-treated patients had elevated ALT levelsPoor
*Wong et al ., 199967100RCT51% of tacrine-treated patients had elevated ALT levelsPoor
*Wood et al.,199469154RCT44% of tacrine-treated patients had elevated ALT levelsPoor
Gastrointestinal adverse events
Grimley Evans et al., 200414NRpooled analysisNausea, vomiting, and diarrhea 3 to 5 times more likely for donepezil, galantamine, and rivastigmine than for placeboN/A
Cutler et al., 1994813350pooled data analysisTacrine had a higher rate of adverse events than donepezil and rivastigmineN/A
Gauthier et al. 200182NRRetrospective data reviewdose dependent rates of gastrointestinal adverse events for ChEIsN/A
*Gillette-Guyonnet et al. 200583486Cohort studySimilar incidence of weight loss for patients taking ChEIs compared to patients not taking ChEIsPoor
Cardiovascular adverse events
Dunn et al., 2000841762Prescription- event monitoringNo cardiac arrhythmias were reported for donepezilN/A
Morganroth et al. 2002852791Pooled analysis of RCTsNo effect on heart rate for rivastigmineFair

* Poor quality rating for efficacy but included for adverse events

Table 7Summary of trials assessing subgroups

Author, YearNStudy designResultsQuality Rating
Age
Schneider et al., 1998352,126Pooled AnalysisBetter cognitive scores (ADAS-cog) for RIV in patients older than 75 yearsFair
Race
Schneider et al., 1998352,126Pooled AnalysisNo differences in response to RIV between black and white patientsFair
Sex
Schneider et al., 1998352,126Pooled AnalysisNo differences in response to RIV between male and female patientsFair
Comorbid Vascular Dementia
AD200038565RCTResults not stratified; general efficacy of DON supported in this mixed populationGood
Erkinjuntti et al., 200290592RCTNo comparison of patients with comorbid vascular disease to population with only AD; general evidence of GAL efficacy in population with comorbid vascular diseaseFair
Kumar et al., 200091699RCT subgroupRIV better than placebo for patients with and without vascular risk factors; larger differences for patients with vascular risk factorsFair

Table 8Key questions and summary of the evidence

Key QuestionQuality of EvidenceConclusion
Key Question 1: Efficacy / EffectivenessPoor to fairNo double-blind head-to-head trial compared one AD drug to another. Three open-label head-to-head trials compared the efficacy of one AD medication to another; two trials compared donepezil to galantamine and one trial compared donepezil to rivastigmine. Evidence for the comparison of donepezil with galantamine is mixed. In one52-week trial, donepezil and galantamine did not differ in stabilizing symptoms or improving behavior and functional status. In a shorter trial (12 weeks), donepezil was superior to galantamine in its effects on cognition, functional status, and caregiver and clinician satisfaction. The comparison of donepezil to rivastigmine is limited to a single 12-week trial; similar improvements in cognitive scores were reported for both drugs, although clinician and caregiver satisfaction ratings were significantly better for donepezil. Both trials that reported significant differences were funded by the manufacturer of donepezil while the trial reporting no differences was funded by the manufacturer of galantamine.

Evidence of general efficacy for donepezil, galantamine, rivastigmine, tacrine, and memantine is fair; 1 placebo- controlled effectiveness trial, 22 efficacy trials, and 8 systematic reviews support modest effects on symptom stabilization, behavior, and functional status as measured by various scales. Although some trials did not support statistically significant differences between active treatment and placebo on all outcome measures, most trials yielded data supporting modest improvement or a slower rate of decline in measures of cognition and global assessment. Fewer data supported differences in measures of behavior, functioning, rate of institutionalization, or caregiver burden.

Although evidence of general efficacy is fair, evidence of effectiveness is poor. We identified only one trial considered to demonstrate effectiveness.
Key Question 2: Time to EffectPoorWe did not identify any study that directly compared the time to effect or time required to assess the clinical response of one AD drug compared to another. Placebo- controlled trials are too heterogeneous with respect to study design, outcomes assessment, and populations to allow any inferences about the comparative time to effect or time required to assess clinical response.
Key Question 3: Adverse eventsPoor to FairHead-to-head trials did not present differences in adverse events between donepezil and galantamine, and donepezil and rivastigmine. Indirect evidence from placebo-controlled trials indicates a substantially higher risk of hepatotoxicity for tacrine than for donepezil, galantamine, rivastigmine, and memantine.
Key Question 4: SubgroupsPoorWe did not identify any study specifically designed to compare the effect of donepezil, galantamine, rivastigmine, tacrine, or memantine in one subgroup of patients compared to another. Subgroup analyses and indirect evidence from placebo controlled trials provide evidence for some AD drugs.

One subgroup analysis reported greater benefit for rivastigmine in patients older than 75 years. Indirect comparison of evidence from one donepezil trial conducted in nursing home residents to trials conducted in younger populations suggests no apparent difference in efficacy or adverse events.

Subgroup analyses of pooled data from four donepezil trials suggest no differences in efficacy or adverse events by sex or race.

No evidence addressed patients with comorbid Parkinson's disease.

Four studies provide general evidence of the efficacy of donepezil, galantamine, rivastigmine, and memantine in populations with comorbid vascular dementia. Only one study stratified patients by vascular risk factors; larger treatment differences between rivastigmine and placebo were found for patients with vascular risk factors compared to patients without vascular risk factors.

No study compared outcomes among subgroups of patients taking a ChEI or memantine concurrently with another drug to patients not concurrently taking the same medication.
Copyright © 2006, Oregon Health & Science University, Portland, Oregon.
Bookshelf ID: NBK10317
PubReader format: click here to try

Views

Other titles in this collection

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...