Several human studies imply that the trace element selenium and its species may influence the onset of neurological disease, including Alzheimer's dementia (AD). Nevertheless, the literature is conflicting, with reported associations between exposure and risk in opposite direction, possibly due to biases in exposure assessment. After conducting a cohort study that detected an excess AD risk associated with higher levels of inorganic-hexavalent selenium in subjects with mild cognitive impairment (MCI), we investigated the relation between selenium and AD using a case-control study design. We determined cerebrospinal fluid levels of selenium species in 56 MCI participants already included in the cohort study, considered as referents, and in 33 patients with established AD. AD risk was inversely correlated with inorganic selenium species and with the organic form bound to selenoprotein P. Selenium bound to other organo-selenium species was positively correlated with AD risk, suggesting compensatory selenoprotein upregulation following increased oxidative stress. The finding of an increased AD risk associated with inorganic-hexavalent selenium from the cohort study was not replicated. This case-control study yielded entirely different results than those generated by a cohort study with a partially overlapping participant population, suggesting that case-control design does not allow to reliably assess the role of selenium exposure in AD etiology. This inability appears to be due to exposure misclassification, falsely indicating an etiologic role of selenium deficiency likely due to reverse causation, and involving most selenium species. The case-control design may instead lend insights into the pathologic process underlying disease progression.