Background Affective symptoms are common in subjects with mild cognitive impairment (MCI), but there is disagreement whether these symptoms are predictive for Alzheimer's disease (AD). We investigated the predictive accuracy of affective symptoms for AD during a follow-up study in subjects with MCI, and whether the predictive accuracy was modified by age, the presence of amnestic MCI or the length of follow-up.Method Newly referred subjects (n=263) with MCI older than 55 years were selected from a memory clinic and followed up after 2, 5 and 10 years. Predictors investigated were: symptoms of depression, anxiety, apathy and sleeping problems.Results Affective symptoms were present in 50-70% of the subjects. The average follow-up period was 5.4 years and 79 subjects (29%) developed AD. Sleeping problems were associated with a decreased risk for AD [odds ratio (OR) 0.35, p<0.001]. Symptoms of depression (OR 0.61, p=0.059) and anxiety (OR 0.58, p=0.051) showed a trend in the same direction. The OR of apathy for AD was 0.67 (p=0.14). Depression was associated with a decreased risk for AD only in subjects without amnestic MCI, but not in subjects with amnestic MCI. Moreover, anxiety was related to the risk for AD differently between subjects diagnosed with AD at the 5-year follow-up (OR 0.23) and subjects diagnosed with AD at the 10-year follow-up (OR 1.7).Conclusions Affective symptoms are associated with a decreased risk for AD. The risk may be dependent on MCI subtype or length of follow-up, but it does not depend on age. © Cambridge University Press 2009.