The above letter was sent to the authors of the original article and their reply follows. In reply: We appreciate the interest of G. Gambassi et al. in our paper. The limitations of our data are acknowledged in our paper. The data analysis was cross-sectional in nature; therefore, attribution to causation has to be made with caution. No information was included on dosage, length of treatment, or compliance. Importantly, by not having data on blood pressure, our study does not indicate whether or not the drug effects were determined by control of hypertension. Their scenario, that our findings could be explained by a different prescribing (treatment?) pattern by the physician for the cognitively impaired subject, is implausible, however. Our research faculty, not the physician, made the diagnoses of cognitive impairment. It is very unlikely that the physician recognized this degree of impairment. Even in the case of dementia, it was only in a relatively small proportion of subjects that this diagnosis had been made previously by the treating physician. The strengths of our study were that it was conducted on a large, community-based population of older African Americans at high risk for vascular disease. The associations between the medications and cognitive function were consistent across four sets of diagnostic criteria for cognitive impairment. We regret that our study, with its modest conclusions, aroused in Gambassi et al. sentiments of “confusion” and that our diagnostic data were “not interpretable” to them. Part of their lack of comprehension may arise from their attempt to compare secondary data derived from their Systematic Assessment of Geriatric Drug Use via Epidemiology (SAGE) study to our data and to other primary data derived from population-based and clinical-based studies. The SAGE study consists of “a longitudinal data base that comprises data collected with the minimum data set on residents of nursing homes in five states.”1 Thus, their conclusions, based upon their data, appear to be at variance with those of other studies, which should not be surprising. Nursing home subjects are different from community-dwelling subjects and the medical care provided to them is not necessarily similar. Take a simple example. The age-adjusted prevalence rate for dementia in our community-dwelling sample of African Americans was 4.8%. In our nursing home sample of African Americans, age-specific prevalence rates rose from 45% in the group age 65 to 74 to 76% in the group age 85 and older.2 Diagnosis of dementia and its subtypes taken from existing nursing records, as in the SAGE study, are neither reliable nor comparable with diagnoses derived from standard comprehensive clinical and neurological assessments by neurologists, neuropsychologists, and psychiatrists (as in the SAGE study). In three major population-based studies of older community-dwelling African Americans in North Carolina,3 New York,4 and in our own study in Indianapolis,2 Alz-heimer's disease (AD) was the most common subtype of dementia. AD is associated with increased mortality, but the incidence and the prevalence of AD increases with age.5 Thus in all published studies; the rates of AD are highest in the oldest age categories. Women may or may not be at increased risk for AD, but they certainly live longer than men. Our finding that AD was most common in the oldest age category of women is consistent with all other studies.6 Comorbid conditions are common in dementia for both women and men. Studies of the association between diabetes mellitus and AD have produced mixed results, but at least one report has suggested that diabetic subjects are at increased risk for developing AD.7 A gender difference with regard to comorbidity and treatment for either AD patients or non-AD patients is certainly not a uniform finding and was not our experience in our study. To the best of our knowledge, none of the current expert panel treatment guidelines have gender issue recommendations. Our study is prospective in design. We hope that by following this population and conducting longitudinal analysis, some of the current uncertainties will be resolved.
Read full abstract