DONEPEZIL FOR NURSING HOME PATIENTS WITH DEMENTIA: A REINTERPRETATION OF THE EVIDENCE ‐ (Letter referred to the authors of the original paper)

To the Editor: We read with interest Tariot et al.'s recent study examining the effect of donepezil in nursing home residents with Alzheimer's disease1 and believe that the authors' discussion of the results was overly enthusiastic about findings favoring the use of donepezil and excessively dismissive of findings that should give evidence-based practitioners reason to pause. First, the authors report no significant differences between donepezil and placebo for the primary endpoint of behavioral problems, as measured by the neuropsychiatric inventory (NPI). Although the authors argue that the negative finding may be an artifact of “uncontrolled methods of behavior management” in the nursing home (apparently the usual pharmacological and nonpharmacological care for patients with behavioral disturbances), this explanation is unconvincing. Either way, one would be hard pressed to support the use of donepezil if reasonable alternative approaches (especially nonpharmacological care for behavior disturbances) rendered its use irrelevant. In a series of additional analyses, the authors examined the effect of donepezil on each of the 12 behaviors measured by the NPI. Although there was a trend toward worse outcomes for 11 of the 12 behaviors, the discussion focused only on the one behavior favoring donepezil. Given the testing of multiple hypotheses, it is not surprising that one of these analyses would favor donepezil by chance. The discussion of this isolated finding should have been more cautious in light of the number of statistical tests applied and the contradictory findings for the other 11 behavioral outcomes. The discussion puts considerable emphasis on the moderate improvement in clinical dementia rating scores and early differences in Mini-Mental State Examination (MMSE) scores in donepezil-treated patients. However, it appears that by the end of the study the differences in MMSE scores had disappeared, although no P-value is presented to clarify this. Given the last observation carried forward methodology, the discussion should have focused primarily on differences at the final study visit, with differences at other time points viewed as tertiary outcomes. In addition, the authors are too quick to dismiss the lack of benefit in activities of daily living—an important metric for patients and caregivers. The explanations for the lack of effect seem post hoc. Finally, the discussion of adverse events should have been more cautious. The substantial difference in the incidence of weight loss between donepezil and placebo (19% vs 10%) is particularly troubling. Although the P-value for this difference is not reported in the paper, we calculate it to be statistically significant (P < .05). The mean weight loss of 3 kg in patients suffering this adverse event is clinically important and can initiate a cascade of other problems leading to functional decline and worsening quality of life. Clinicians must seriously balance this adverse effect against any potential effect on cognitive function. In sum, this study shows no evidence of benefit in the treatment of behavioral problems in nursing home patients with dementia. Secondary measures of cognitive and daily function are mixed and do not provide convincing evidence of substantial benefit in patients with the most severe cognitive impairment. Furthermore, the higher rates of potentially serious adverse events such as weight loss are troubling. Clinicians should be cautious in using donepezil for nursing home patients with dementia.