In this one-year trial, there were no significant changes in cognitive performance between AD participants receiving stimulation therapy or standard care. To our surprise, both groups retained their cognitive function and the results were consistent for three different cognitive tests as assessed at quarterly sessions during the study period. Among the participants receiving standard care, we expected an annual decline of 2–3 MMSE points, or an increase of 5–12 ADAS-Cog points, which is the natural course of AD [23, 24].
Our results are in agreement with others [25, 26] who reported small changes in mean MMSE score in controls receiving standard care. This is in contrast with Requena et al , who reported a significant decline in mean MMSE score from 19.4 (SD 4.9) to 13.1 (SD 5.9) after one year without intervention, but an increase in mean MMSE score from 19.4 (SD 8.2) to 21.9 (SD 7.9) after one year for AD subjects receiving stimulation therapy. However, these studies are limited by short follow up , retrospective design, poorly defined controls , small sample size and an open design .
In our study, there were no differences in cognitive performance between the groups given donepezil or placebo, irrespective of stimulation therapy or standard care. Chapman et al  and Matsuda et al  reported non-significant changes in mean MMSE score in the combined treatment group given donepezil and stimulation therapy after one year (−1.3 and +0.3 respectively), but a significant reduction in cognitive performance was observed in the donepezil only group (−2.9 and −2.0 respectively). Requena et al reported improved mean MMSE score (+1.5 points) after one year and a minimal reduction after two years in the combined treatment group (−1.3 points), whereas AD participants receiving donepezil without specific stimulation and AD participants in the control group had significantly lower MMSE score (−3.4 and −6.3, respectively). Our results partially support these studies [10–12] as the effect of stimulation on cognitive performance, with or without donepezil, did not deteriorate after one year. However, unlike Requena et al, our participants receiving donepezil or placebo added to standard care retained cognitive performance after one year. The differences could be due to smaller sample size and greater variation in cognitive function among participants in the Requena et al study compared to the present study. Differences in baseline cognitive function between mild and moderate AD may also give different cognitive deterioration slopes.
In a recent systematic review, Olazarán et al  has evaluated best effect of nonpharmacological therapies. Evidence of potential grade A treatment recommendation was reported for the effect of multicomponent intervention in delaying institutionalisation and grade B treatment recommendation for improvement in cognition and of activity of daily living. The intervention program of the present study is in accordance with these recommendations.
Several events and mechanisms may explain the similar cognitive performances between participants receiving stimulation therapy and standard care. In our study all groups were engaged in some treatment regimen also the standard care group that received placebo. We suggest that even this group was exposed to more than standard care throughout the study period because of overly enthusiastic co-workers including test technicians who became involved in activities that exceeded their predefined roles. Frequent monitoring and follow-ups do increase the attention given to the patients by family members, caregivers and study staff. This may act as stimulation in itself and generate an expectancy of a beneficial outcome . We know from the experiences of industrial companies that production may increase no matter what changes are introduced to the workers or working conditions, “the Hawthorne effect” . The Hawthorn effect has to our knowledge only been addressed in one single AD study . This effect is not sufficiently described in clinical AD trials [31, 32], and it may contribute to controls performing better than expected.
Beside the above-mentioned mechanisms, our study results may have been affected by a national AD campaign launched at the same time. That campaign focus on the beneficial effects of cognitive stimulation. This could have contributed to a diffusion of the therapeutic procedures across municipality borders and diluted the differences between stimulation therapy and standard care.
Strengths and weaknesses
Our study is population-based with few exclusion criteria. This is different from several other studies which recruited AD individuals from hospitals, memory clinics or nursing homes using restrictive inclusion and wide exclusion criteria that could influence study samples and results. Our study population is ethnically and socially homogenous, and the baseline characteristics did not differ between defined strata. The two-by-two factorial design in our study enables a head to head comparison between stimulation therapy and drug treatment
Participants remained in their own environment during the entire study period. None of the participants used memantin or other ChEI. Although 23% of the participants used anticholinergic drugs for co-morbidities , inappropriate drugs were equally distributed between groups and could hardly explain the results.Our participants received structured stimulation therapy on a daily basis and had few dropouts (22%), and the dropouts were equally distributed among subgroups. A possible weakness is that lower MMSE score among dropouts could have influenced mean cognitive deterioration during follow-up. However, repeated measure analyses confirmed the main result of the study.
Another possible weakness is that despite randomisation, participants in the donepezil group had a lower educational level compared to the placebo group. Less education is associated with an increased risk of AD  but it is questionable whether this risk factor has an influence on the cognitive trajectory in AD in any way. The non-randomized allocation of the participating municipalities to stimulation therapy or standard care could be considered a weakness. However, a sensitivity analysis done to control for possible clustering of data within the municipalities did not change the results. The ability of the applied tests to detect a change in cognitive performance is questioned in early stage AD . However, a stratified analysis of a subgroup presenting a MMSE score less than 21 points at entry (n = 43) showed no differences, and the results were consistent for all three cognitive tests (data not shown). It is therefore unlikely that a MMSE learning effect has occurred.
Furthermore, the stratified samples in the two-by-two factorial analysis that we did could be prone to type II errors due to relatively small sample sizes - especially the subgroup analyses of participants with MMSE < 21. However, the differences between all groups were consistent for all three tests. Another consequence of the two-by-two factorial design was that no subgroup was left without any intervention. Even the standard care group received either placebo or donepezil. This could have increased the expectancy of a favourable outcome in the control groups and diluted the results .