LAUSR

Question In adults with normal cognition or mild cognitive impairment (MCI), what are the efficacy and safety of drugs for preventing or delaying cognitive decline or dementia? Review scope Included studies compared prescription drugs approved by the US Food and Drug Administration with placebo or control treatment in adults with normal cognition or MCI, and had follow-up 6 months. Primary outcomes were MCI, or Alzheimer or unspecified dementia. Secondary outcomes included cognitive performance on validated tests. Review methods MEDLINE, EMBASE/Excerpta Medica, PsycINFO, and Cochrane Library (2009 to Jul 2017); an Agency for Healthcare Research and Quality (AHRQ) review published in 2010; and reference lists were searched for English-language randomized and nonrandomized controlled trials. Of 102 eligible reports, 51 unique trials had low or medium risk for bias (AHRQ criteria) and were included in the analyses. Overall strength of evidence for all comparisons and outcomes was rated as moderate, low, or insufficient (AHRQ criteria); results with insufficient strength of evidence are not reported here. Data were not pooled because of heterogeneity. Main results The main findings are in the Table. Conclusion In adults with normal cognition or mild cognitive impairment, drugs do not prevent or delay cognitive decline or dementia. Drugs vs control in patients with normal cognition or MCI* Comparisons Main findings Patients with normal cognition Anti-HTN drugs vs placebo No difference for dementia at 2.2 to 4.3 y (4 trials, n =21831), incident cognitive impairment at 4.7 y (1 trial, n =5926, OR 0.90, 95% CI 0.80 to 1.01), or BCTs at 2 to 4.7 y (4 trials, n =15811). Intensive vs standard anti-HTN treatment No difference for BCTs, EF/A/PS, or memory at 40 mo (1 trial, n =1439). Anti-HTN drugs vs anti-HTN drugs No difference for incident cognitive impairment at 4.7 y (1 trial, n =25620) or EF/A/PS at 24 wk to 9 mo (3 trials, n =2614). Intensive vs standard diabetes treatment No difference for incident cognitive impairment at 6.2 y (1 trial, n =11685), BCTs or EF/A/PS at 3.3 to 6.2 y (2 trials, n =14479), or memory at 3.3 y (1 trial, n =2794). Statin + fenofibrate vs statin No difference for BCTs, EF/A/PS, or memory at 3.3 y (1 trial, n =1538). NSAID vs placebo No difference for dementia at 8 y (celecoxib HR 1.03, CI 0.72 to 1.50; naproxen HR 0.92, CI 0.62 to 1.35) or multidomain tests, EF/A/PS, or memory at 4 y (1 trial, n =2117). Aspirin vs placebo No difference for BCTs, multidomain tests, or memory at 9.6 y (1 trial, n =6377). Estrogen vs placebo Estrogen increased a composite of dementia or MCI (HR 1.38, CI 1.01 to 1.89) but not dementia (HR 1.49, CI 0.83 to 2.66) or MCI (HR 1.34, CI 0.95 to 1.89) separately at 5.2 y (1 trial, n =2947). Estrogen improved EF/A/PS for 2 of 19 tests and memory for 2 of 35 tests at 6 mo to 5.2 y (6 trials, n =4117**), and decreased performance for 1 of 2 BCTs at 2 to 5.4 y (2 trials, n =3364). Estrogen + progestin vs placebo Estrogen + progestin increased dementia (HR 2.05, CI 1.21 to 3.48) but not MCI (HR 1.07, CI 0.74 to 1.55) or a composite of dementia or MCI (HR 1.37, CI 0.99 to 1.89) at 4.1 y (1 trial, n =4532). Estrogen + progestin decreased performance for 1 of 4 BCTs at 3.2 to 5.4 y (3 trials, n =4353), 1 of 9 EF/A/PS tests at 6 mo to 5.4 y (4 trials, n =4376), and 4 of 16 memory tests at 6 mo to 5.4 y (5 trials, n =4518). Selective estrogen receptor modulator vs placebo Raloxifene, 120 mg/d, reduced MCI (RR 0.67, CI 0.46 to 0.98) but not dementia (RR 0.91, CI 0.47 to 1.76) or a composite of dementia or MCI (RR 0.73, CI 0.53 to 1.01), and raloxifene, 60 mg/d, did not differ from placebo for MCI (RR 1.18, CI 0.85 to 1.64), dementia (RR 0.90, CI 0.47 to 1.74), or a composite of dementia or MCI (RR 1.12, CI 0.84 to 1.49) at 3 y in 1 trial (n =5386). No difference between raloxifene, 60 or 120 mg/d, and placebo for EF/A/PS, memory, or multidomain tests at 1 to 3 y (2 trials, n =7621**). Patients with MCI Cholinesterase inhibitors vs placebo No difference for dementia (RR 0.80, CI 0.57 to 1.13) or multidomain tests at 3 y (1 trial, n =512) or BCTs at 1 to 3 y (2 trials, n =533). Testosterone vs placebo No difference for EF/A/PS or memory at 6 to 12 mo (2 trials, n =515). *BCT = brief cognitive test; EF/A/PS = executive function/attention/processing speed; HR = hazard ratio; HTN = hypertension; MCI = mild cognitive impairment; NSAID = nonsteroidal antiinflammatory drug; OR = odds ratio; RR = risk ratio; other abbreviations defined in Glossary. Results are based on low strength of evidence unless reported otherwise. Results with insufficient strength of evidence are not reported here. Borderline benefit with treatment in 1 trial (RR 0.50, CI 0.24 to 1.00); no group difference in 3 trials. Composite of dementia, cognitive impairment, or Mini-Mental State Examination score <24. Moderate strength of evidence confirmed by author. Estimated n provided by author. Composite of incident dementia or Mini-Mental State Examination score <24. **Low strength of evidence confirmed by author. Commentary Fink and colleagues concluded that, among the many treatments reviewed, cholinesterase inhibitors (ChEI) also did not improve cognition or prevent progression to dementia in patients with MCI. Because MCI and dementia are part of a continuum, it is paradoxical that clinical trials have shown that ChEI can improve cognitive performance in early dementia but not in MCI. Criteria for MCI include subjective elements, such as complaints about cognitive problems and reported functional dependence. In the studies reviewed by Fink and colleagues, it is possible that cognitive improvements in MCI were below the threshold of detection for the tests used or that included patients who met the definition of MCI did not actually have it. Up to 38% of patients with MCI revert to normal cognition (1), suggesting a high rate of false-positive diagnoses. After removing the 30% of patients with a probable false-positive diagnosis in a trial of donepezil in MCI, reanalysis found that patients randomized to donepezil had lower progression rates and better cognitive performance than control patients (2). Observational studies have shown positive associations between such midlife vascular risk factors as hypertension and development of Alzheimer disease (3). In Fink and colleagues' analyses, randomized trials of standard and intensive control of diabetes and hypertension did not show differences in cognitive outcomes. These trials should not be considered definitive as the null results could have resulted from insufficient trial duration, inadequacy of addressing only 1 risk factor, or intervening too late along the pathophysiologic pathway toward dementia.