Background: Optimal treatment to prevent or delay cognitive decline, mild cognitive impairment (MCI), or dementia is uncertain. Purpose: To summarize current evidence on the efficacy and harms of pharmacologic interventions to prevent or delay cognitive decline, MCI, or dementia in adults with normal cognition or MCI. Data Sources: Several electronic databases from January 2009 to July 2017, bibliographies, and expert recommendations. Study Selection: English-language trials of at least 6 months' duration enrolling adults without dementia and comparing pharmacologic interventions with placebo, usual care, or active control on cognitive outcomes. Data Extraction: Two reviewers independently rated risk of bias and strength of evidence; 1 extracted data, and a second checked accuracy. Data Synthesis: Fifty-one unique trials were rated as having low to moderate risk of bias (including 3 that studied dementia medications, 16 antihypertensives, 4 diabetes medications, 2 nonsteroidal anti-inflammatory drugs [NSAIDs] or aspirin, 17 hormones, and 7 lipid-lowering agents). In persons with normal cognition, estrogen and estrogen–progestin increased risk for dementia or a combined outcome of MCI or dementia (1 trial, low strength of evidence); high-dose raloxifene decreased risk for MCI but not for dementia (1 trial, low strength of evidence); and antihypertensives (4 trials), NSAIDs (1 trial), and statins (1 trial) did not alter dementia risk (low to insufficient strength of evidence). In persons with MCI, cholinesterase inhibitors did not reduce dementia risk (1 trial, low strength of evidence). In persons with normal cognition and those with MCI, these pharmacologic treatments neither improved nor slowed decline in cognitive test performance (low to insufficient strength of evidence). Adverse events were inconsistently reported but were increased for estrogen (stroke), estrogen–progestin (stroke, coronary heart disease, invasive breast cancer, and pulmonary embolism), and raloxifene (venous thromboembolism). Limitation: High attrition, short follow-up, inconsistent cognitive outcomes, and possible selective reporting and publication. Conclusion: Evidence does not support use of the studied pharmacologic treatments for cognitive protection in persons with normal cognition or MCI.