Several studies have identified an association between body mass index (BMI) and the incidence and severity of Alzheimer's disease (AD) but this relationship is not fully understood. The primary objective of this study was to assess the possible association between BMI and cerebrospinal fluid (CSF) biomarkers of AD pathology in subjects with normal cognition and cognitive impairment. The secondary objective was to test whether BMI may contribute to improve the accuracy of a clinical model to predict AD pathology in memory clinic patients with cognitive impairment. One hundred and seven elderly subjects with cognitive impairment (91 memory clinic patients with mild cognitive impairment [MCI] and 16 with dementia of AD type) and 55 cognitively healthy volunteers were included in this study. All subjects received a comprehensive clinical and neuropsychological evaluation and a lumbar puncture for CSF biomarker analysis. Multiple linear regressions and receiver operating characteristic (ROC) analyses were carried out to assess the association between BMI and the CSF biomarkers of AD pathology. BMI was positively correlated with the CSF levels of Aβ 42 and negatively with tau and P-tau181 in participants with cognitive impairment. The associations were independent of age, sex, educational level, type and severity of cognitive impairment, cerebrovascular risk factors and the presence of the APOEε4 allele. Furthermore, BMI significantly improved the sensitivity and specificity of a multi-factorial model to predict the presence of an AD CSF biomarker profile. Lower BMI is associated with cerebral AD pathology rather than with cognitive impairment in elderly subjects with MCI and mild dementia. Along with other clinical factors, decreasing BMI may help the clinician to identify patients with cognitive impairment due to AD.