Context

More than half of persons with Alzheimer's disease are at moderate or severe stages, and evidence defining best practice for them is limited. Previous clinical trials in these patients, as summarised by Howard et al,1–5 have shown that treatment with donepezil or memantine only, or the addition of memantine to donepezil, confers benefit in cognition and daily function. Building upon these, Howard and colleagues aimed to address the relative benefit of continuing donepezil, adding or substituting memantine, or stopping donepezil in these patients.

Findings

A blinded interim analysis showed reduced SD, so the planned sample size was reduced to 430 participants. Enrolment was slow and was eventually stopped by the funding agency (ie, Medical Research Council) as it was thought that reporting the results sooner outweighed the benefits of accruing a complete sample. The result was based on a sample size of 295 participants. Follow-up was achieved in two-thirds of the patients. Analysis was both by intention-to-treat and by per protocol analysis. Participants assigned to donepezil withdrawal only showed marked decline in both cognition and daily function in comparison with the other three treatment groups, which showed significantly less decline, without significant differences between them. Memantine treatment was associated with fewer behavioural symptoms. There were no unexpected adverse effects.

Commentary

Patients with moderate-to-severe Alzheimer's disease who were previously treated with donepezil showed better outcomes as assessed by simple measures of cognition and daily function if they stayed on donepezil or started on memantine therapy. In contrast, those who were switched to placebo only declined markedly in terms of cognition and daily function. The limited instruments used in the trial are not well suited to informing clinical decision-making, but the results were significant and clinically relevant.

Although the funder intended for the study to address the impact of adding memantine therapy, and contrary to assertions in the study that ‘combined (sic) treatment with donepezil and memantine was not significantly superior to treatment with donepezil alone,’ there was no evidence of lack of effect. After a 2-year enrolment period, only 295 of the originally planned 800 participants were enrolled, and sample sizes at the end of the study were small; ranging from 20 to 38 participants per group. There were markedly different dropout rates and important baseline differences among the groups. As a result of these factors, this aim simply could not be addressed and the failure to demonstrate a ‘statistically significant’ difference among groups does not mean or even imply that there was no effect. Furthermore, upon inspection of the data, there were differential effects by treatment group for a large part of the treatment period: this is not a clinically irrelevant finding.

The decline in measures of cognition and daily functioning among individuals treated with placebo only raises the question of how the data monitoring committee viewed the risks versus benefits of stopping all active treatment. The study report also states that, ‘The findings of the study showing that combination (sic) therapy with memantine that a cholinesterase inhibitor was more effective than treatment with a cholinesterase inhibitor alone have not been replicated’. This is not correct. First, all of the relevant studies addressed adjunctive (one form of treatment is added to another that was started previously), not combination (two agents are started simultaneously) therapy: this conceptual error could lead to confusion, since the two treatment approaches are quite different. Second, two of the three prior adjunctive therapy studies showed benefit,5–7 and the citation used to indicate lack of replication refers to a study in patients with milder, rather than moderate to severe Alzheimer's disease and is not pertinent.8

This study achieved some very important results—that stopping donepezil is associated with a decline in cognition and function. However, it did not achieve all its intended goals—it did not definitively answer all of the questions posed, leaving room for additional research on combinations of medications, substitution of one for another and the comparative effectiveness of treatment options.