fluorodeoxyglucose positron emission tomography scanning to predict cognitive decline in a group of subjects selected from the Sacramento Area Latino Study on Aging (SALSA) cohort. 1 We agree that there are few other studies with which to compare our own results. The published articles cited by Dr Panza include estimates of incidence ranging from 8.5 to 26 per 1,000 person-years for mild cognitive impairment (MCI) syndromes. We also agree that the most appropriate figure for comparison with our own data would be incidence of MCI or cognitive impairment/no dementia combined with incidence of dementia, as indeed we reported the "conversion" from normal to both of these states. The authors cite their own data from the Italian Longitudinal Study on Aging in this regard as 34 per 1,000 person-years. Although we did quote published incidence rates for dementia from two population-based studies, it is also worth noting that published dementia incidence rates vary widely, from 10.7 to more than 30 per 1,000 person-years. 2 Given the large variability in methodology, populations, and the difficulties in establishing and differentiating MCI, cognitive impairment/no dementia, Alzheimer's disease, and other dementias, the threefold differences in estimates of MCI or dementia incidence are not surprising. In this light, we are more impressed with the similarity than the difference of our crude incidence rate of 26 per 1,000 to the figure of 34 per 1,000 person-years that Dr Panza and colleagues reported.
However, our study was not designed to determine the incidence of cognitive impairment/no dementia or dementia. The subjects were a subsample of a larger cohort, may not reflect the incidence in the entire cohort, and are a relatively small group from which to draw firm conclusions about incidence rates. Incident disease or conversion was not an outcome measure in our study, which used cognitive decline as a continuous variable. We presented incidence data to make a simple point: Because the rate of conversion was not unusually high, the subjects were not already in an early stage of dementia. Such an incidence rate would thus bias against finding a predictive value to positron emission tomography imaging, making the tests of our hypotheses more conservative. Although generalizability is always a concern in small samples, we believe that this is the appropriate sort of cohort in which to test the predictive value of imaging.