A systematic literature review was performed to assess the suitability of cerebrospinal fluid (CSF) levels of protein amyloid-beta42 (Abeta42) and tau as markers to detect the disease-modifying effects of drugs in clinical trials of AD treatments. All databases were searched for observational studies, single-arm clinical trials, and randomized controlled trials involving patients with AD in which CSF Abeta42 and tau were measured. A meta-analytic random-effects model was used to evaluate the mean absolute change in protein concentration over time. Spearman correlation was used to assess the association between change in CSF protein concentration and change in cognitive function. The mean changes per month in observational studies were -0.4 pg/ml/month (95% CI: -1.9 to 1.1) for 8 CSF Abeta42 and 1.5 pg/ml/month (95% CI: 0.1 to 3.0) for 12 CSF tau studies. The correlation coefficients for the relationship between CSF protein concentration and cognition were 0.43 (p = 0.068) for all 18 Abeta42 studies and -0.05 (p = 0.857) for all 18 tau studies. A trend in which CSF Abeta42 decreases and tau protein increases over time was identified in AD patients. CSF Abeta42 and tau concentrations should be used with caution as surrogate endpoints in early-phase clinical trials for AD.