OBJECTIVE: Demographic changes are increasing the pressure to improve therapeutic strategies against cognitive decline in Alzheimer disease (AD) and mild cognitive impairment (MCI). Besides drug treatment, physical activity seems to be a promising intervention target as epidemiological and clinical studies suggest beneficial effects of exercise training on cognition. Using comparable inclusion and exclusion criteria, we analyzed the efficacy of drug therapy (cholinesterase inhibitors, memantine, and Ginkgo biloba) and exercise interventions for improving cognition in AD and MCI populations. METHODS: We searched The Cochrane Library, EBSCO, OVID, Web of Science, and U.S Food and Drug Administration data from inception through October 30, 2013. Randomized controlled trials in which at least one treatment arm consisted of an exercise or a pharmacological intervention for AD or MCI patients, and which had either a non-exposed control condition or a control condition that received another intervention. Treatment discontinuation rates and Standardized Mean Change score using Raw score standardization (SMCR) of cognitive performance were calculated. RESULTS: Discontinuation rates varied substantially and ranged between 0% and 49% with a median of 18%. Significantly increased discontinuation rates were found for galantamine and rivastigmine as compared to placebo in AD studies. Drug treatments resulted in a small pooled effect on cognition (SMCR: 0.23, 95% CI: 0.20 to 0.25) in AD studies (N = 45, 18,434 patients) and no effect in any of the MCI studies (N = 5, 3,693 patients; SMCR: 0.03, 95% CI: 0.00 to 0.005). Exercise interventions had a moderate to strong pooled effect size (SMCR: 0.83, 95% CI: 0.59 to 1.07) in AD studies (N = 4, 119 patients), and a small effect size (SMCR: 0.20, 95% CI: 0.11 to 0.28) in MCI (N = 6, 443 patients). CONCLUSIONS: Drug treatments have a small but significant impact on cognitive functioning in AD and exercise has the potential to improve cognition in AD and MCI. Head-to-head trials with sufficient statistical power are necessary to directly compare efficacy, safety, and acceptability. Combining these two approaches might further increase the efficacy of each individual intervention. IDENTIFIER: PROSPERO (2013:CRD42013003910).