Ence of 0.5 points on the MMSE per follow-up visit (around five points over the entire follow-up period). In contrast, the piracetam group declined more rapidly. 
With respect to the IST and the BVRT, no significant difference was observed between the 1934-21-0 site EGb761H group compared to the `neither treatment’ group, whereas the piracetam group declined to a greater extent. A Pleuromutilin logistic regression model adjusted for the same confounding variables was performed to assess the association between EGb761H and psychotropic drug consumption (antidepressants, benzodiazepines or antipsychotics). The result showed that use of EGb761H was associated with significantly lower consumption of psychotropic drugs (OR 0.72, 95 Confidence Intervals: 0.57?0.91, p = 0.007). Due to the significant association between EGb761H use and reduced consumption of psychotropic drugs, the linear mixed effects model was reiterated adjusting for psychotropic drug consumption (Table 3). As can be seen, the beta coefficients remained unchanged, reflecting similar differences in cognitive decline between treatments groups after controlling for psychotropic drug use. The decline in MMSE score over time in the three treatment groups, as estimated by the model, is illustrated below in Figure 2. In a second step, the linear mixed effects model was reiterated to compare the EGb761H and piracetam treatment groups directly (Table 4). As can be seen, the effect size on the MMSE corresponded to a less rapid decline by around one point on the MMSE per follow-up visit in the EGb761H group. In addition, a significant difference in rate of change was observed not only for the MMSE but also the other two tests of memory and verbal fluency. The findings remained essentially unchanged after controlling for psychotropic drug use (Table 4).DiscussionThis analysis of prospectively collected data on cognitive function over a twenty-year period has shown that the decline of the MMSE score in a population of non-demented subjects was lower in the group of subjects who reported using EGb761H at some time than in those who did not. The difference in MMSE score at the end of the follow-up period was around five points, which can be considered an important and clinically relevant difference. The predicted MMSE score at the end of the follow-up period remained above the threshold of 24 (roughly normal cognitive function) in the group using EGb761H, which is also of clinical relevance. This effect appears to be a specific medication effect of EGb761H, since it was not observed for another nootropic medication, piracetam, prescribed for the same condition as EGb761H, whose users performed less well all along the follow-up period in the three tests studied. The latter finding suggests that the observed beneficial effect of EGb761H on cognitive decline is not an artefact of greater motivation to preserve cognitive function which encourages subjects to seek medication in general for the management of memory complaints. At first sight, these results may appear somewhat discordant with those trials such as the GEM [32] and the GuidAge [33] studies reporting no effect of the EGb761H on the risk of developing dementia, which led some authors to definitely conclude that ginkgo biloba is not effective for prevention of Alzheimer’s disease [45]. However, our results may not be so inconsistent if one considers the following issues. Firstly, it is important to emphasise that these studies relied on volunteers presenting mot.Ence of 0.5 points on the MMSE per follow-up visit (around five points over the entire follow-up period). In contrast, the piracetam group declined more rapidly. With respect to the IST and the BVRT, no significant difference was observed between the EGb761H group compared to the `neither treatment’ group, whereas the piracetam group declined to a greater extent. A logistic regression model adjusted for the same confounding variables was performed to assess the association between EGb761H and psychotropic drug consumption (antidepressants, benzodiazepines or antipsychotics). The result showed that use of EGb761H was associated with significantly lower consumption of psychotropic drugs (OR 0.72, 95 Confidence Intervals: 0.57?0.91, p = 0.007). Due to the significant association between EGb761H use and reduced consumption of psychotropic drugs, the linear mixed effects model was reiterated adjusting for psychotropic drug consumption (Table 3). As can be seen, the beta coefficients remained unchanged, reflecting similar differences in cognitive decline between treatments groups after controlling for psychotropic drug use. The decline in MMSE score over time in the three treatment groups, as estimated by the model, is illustrated below in Figure 2. In a second step, the linear mixed effects model was reiterated to compare the EGb761H and piracetam treatment groups directly (Table 4). As can be seen, the effect size on the MMSE corresponded to a less rapid decline by around one point on the MMSE per follow-up visit in the EGb761H group. In addition, a significant difference in rate of change was observed not only for the MMSE but also the other two tests of memory and verbal fluency. The findings remained essentially unchanged after controlling for psychotropic drug use (Table 4).DiscussionThis analysis of prospectively collected data on cognitive function over a twenty-year period has shown that the decline of the MMSE score in a population of non-demented subjects was lower in the group of subjects who reported using EGb761H at some time than in those who did not. The difference in MMSE score at the end of the follow-up period was around five points, which can be considered an important and clinically relevant difference. The predicted MMSE score at the end of the follow-up period remained above the threshold of 24 
(roughly normal cognitive function) in the group using EGb761H, which is also of clinical relevance. This effect appears to be a specific medication effect of EGb761H, since it was not observed for another nootropic medication, piracetam, prescribed for the same condition as EGb761H, whose users performed less well all along the follow-up period in the three tests studied. The latter finding suggests that the observed beneficial effect of EGb761H on cognitive decline is not an artefact of greater motivation to preserve cognitive function which encourages subjects to seek medication in general for the management of memory complaints. At first sight, these results may appear somewhat discordant with those trials such as the GEM [32] and the GuidAge [33] studies reporting no effect of the EGb761H on the risk of developing dementia, which led some authors to definitely conclude that ginkgo biloba is not effective for prevention of Alzheimer’s disease [45]. However, our results may not be so inconsistent if one considers the following issues. Firstly, it is important to emphasise that these studies relied on volunteers presenting mot.