Recent studies indicate that there is a marked reduction in trkA-containing nucleus basalis neurons in end-stage Alzheimer's disease (AD). We used unbiased stereological counting procedures to determine whether these changes extend to individuals with mild cognitive impairment (MCI) without dementia from a cohort of people enrolled in the Religious Orders Study. Thirty people (average age 84.7 years) came to autopsy. All individuals were cognitively tested within 12 months of death (average MMSE 24.2). Clinically, 9 had no cognitive impairment (NCI), 12 were categorized with MCI, and 9 had probable AD The average number of trkA-immunoreactive neurons in persons with NCI was 196, 632 +/- 12,093 (n = 9), for those with MCI it was 106,110 +/- 14,565, and for those with AD it was 86,978 +/- 12,141. Multiple comparisons showed that both those with MCI and those with AD had significant loss in the number of trkA-containing neurons compared to those with NCI (46% decrease for MCI, 56% for AD). An analysis of variance revealed that the total number of neurons containing trkA immunoreactivity was related to diagnostic classification (P < 0.001), with a significant reduction in AD and MCI compared to NCI but without a significant difference between MCI and AD. Cell density was similarly related to diagnostic classification (P < 0.001). There was a significant correlation with the Boston Naming Test and with a global score measure of cognitive function. The number of trkA-immunoreactive neurons was not correlated with MMSE, age at death, education, apolipoprotein E allele status, gender, or Braak score. These data indicate that alterations in the number of nucleus basalis neurons containing trkA immunoreactivity occurs early and are not accelerated from the transition from MCI to mild AD.