Comparative safety and efficacy of cognitive enhancers for Alzheimer's dementia: a systematic review with individual patient data network meta-analysis

Abstract

Objective: To examine the comparative efficacy and safety of cognitive enhancers by patient characteristics for managing Alzheimer's dementia (AD).

Design: Systematic review and individual patient data (IPD) network meta-analysis (NMA) based on our previously published systematic review and aggregate data NMA.

Data sources: MEDLINE, Embase, Cochrane Methodology Register, CINAHL, AgeLine and Cochrane Central Register of Controlled Trials up to March 2016.

Participants: 80 randomised controlled trials (RCTs) including 21 138 adults with AD, and 12 RCTs with IPD including 6906 patients.

Interventions: Cognitive enhancers (donepezil, rivastigmine, galantamine and memantine) alone or in any combination against other cognitive enhancers or placebo.

Data extraction and synthesis: We requested IPD from authors, sponsors and data sharing platforms. When IPD were not available, we used aggregate data. We appraised study quality with the Cochrane risk-of-bias. We conducted a two-stage random-effects IPD-NMA, and assessed their findings using CINeMA (Confidence in Network Meta-Analysis).

Primary and secondary outcomes: We included trials assessing cognition with the Mini-Mental State Examination (MMSE), and adverse events.

Results: Our IPD-NMA compared nine treatments (including placebo). Donepezil (mean difference (MD)=1.41, 95% CI: 0.51 to 2.32) and donepezil +memantine (MD=2.57, 95% CI: 0.07 to 5.07) improved MMSE score (56 RCTs, 11 619 participants; CINeMA score: moderate) compared with placebo. According to P-score, oral rivastigmine (OR=1.26, 95% CI: 0.82 to 1.94, P-score=16%) and donepezil (OR=1.08, 95% CI: 0.87 to 1.35, P-score=30%) had the least favourable safety profile, but none of the estimated treatment effects were sufficiently precise when compared with placebo (45 RCTs, 15 649 patients; CINeMA score: moderate to high). For moderate-to-severe impairment, donepezil, memantine and their combination performed best, but for mild-to-moderate impairment donepezil and transdermal rivastigmine ranked best. Adjusting for MMSE baseline differences, oral rivastigmine and galantamine improved MMSE score, whereas when adjusting for comorbidities only oral rivastigmine was effective.

Conclusions: The choice among the different cognitive enhancers may depend on patient's characteristics. The MDs of all cognitive enhancer regimens except for single-agent oral rivastigmine, galantamine and memantine, against placebo were clinically important for cognition (MD larger than 1.40 MMSE points), but results were quite imprecise. However, two-thirds of the published RCTs were associated with high risk of bias for incomplete outcome data, and IPD were only available for 15% of the included RCTs.

Prospero registration number: CRD42015023507.

Keywords: dementia; epidemiology; statistics & research methods.

Figure 1

Flow diagram for study inclusion in the review (A) and studies retrieved with individual patient data (B). AD, Alzheimer’s dementia; IPD, individual patient data.

Figure 2

Network diagrams for (A) MMSE and (B) AE outcomes. The size of each node and line indicates the number of studies included in each treatment comparison. The number of studies per treatment comparison is presented on each edge, and the number of studies with individual patient data (IPD) is depicted in a parenthesis. Orange coloured edges are informed by both IPD and aggregate data, whereas black coloured edges are informed by aggregate data only. AE, adverse event; DONE, donepezil; GALA, galantamine; MEMA, memantine; MMSE, Mini-Mental State Examination; PLAC, placebo; RIVA_O, oral rivastigmine; RIVA_T, transdermal rivastigmine.

Figure 3

Forest plot of network meta-analysis (NMA) results for all cognitive enhancers versus placebo in (A) MMSE outcome, and (B) AE outcome. NMA results are presented for (i) aggregate data (AD) and fully adjusted results from studies with available individual patient data (IPD), (ii) AD and crude results from studies with available IPD, (iii) AD only (studies with available IPD are not included in the analysis) and (iv) crude results from individual studies with IPD. AD, Alzheimer’s dementia; AE, adverse events; DONE, donepezil; GALA, galantamine; MEMA, memantine; MMSE, Mini-Mental State Examination; PLAC, placebo; RIVA_O, oral rivastigmine; RIVA_T, transdermal rivastigmine.

Figure 4

Rank-heat plot of P-scores for nine treatments, including placebo, studied in randomised clinical trials with patients with Alzheimer’s dementia assessing Mini-Mental State Examination. Circles from inside out present results for different network meta-analyses including: (i) aggregate data (AD) only (studies with available IPD are not included in the analysis), (ii) crude results from individual studies with individual patient data (IPD), (iii) AD and crude results from studies with available IPD and (iv) AD and fully adjusted results from studies with available IPD. Numbers within each sector correspond to the P-score values as calculated in each model. AD, Alzheimer’s dementia; adjMD, adjusted mean difference; DONE, donepezil; GALA, galantamine; MEMA, memantine; PLAC, placebo; RIVA_O, oral rivastigmine; RIVA_T, transdermal rivastigmine; unadjMD, unadjusted MD.