The best magazine
Methods
This meta-analysis was planned and written according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines, using a methodology extensively described in a previous publication of our group.
Search Strategy
We screened PubMed, ISI Web of Science, Cochrane Library and Scopus databases for articles published until May 2014, without any language restrictions, but excluding abstracts and conference proceedings. Three of the authors (GeP, GR and MT) identified and assessed all articles under the major medical subject heading 'Parkinson Disease' and 'Cholinesterase Inhibitors' combined with text and key words (for PubMed as example): (("Parkinson Disease"[MeSH Terms] OR "Parkinson Disease" OR "Parkinson's Disease" OR "Idiopathic Parkinson's Disease" OR "Lewy Body Parkinson Disease" OR "Primary Parkinsonism" OR "Paralysis Agitans") AND ("Cholinesterase Inhibitors"[MeSH Terms] OR "Cholinesterase Inhibitors" OR "Anticholinesterase Drugs" OR "Anticholinesterases" OR "Anticholinesterase Agents" OR "Anti-Cholinesterases" OR "Anti Cholinesterases" OR "Acetylcholinesterase Inhibitors" OR "rivastigmine" OR "Exelon*" OR "SDZ ENA 713" OR "donepezil" OR "Aricept*" OR "E2020" OR "galantamine" OR "galanthamine" OR "reminyl*" OR "Nivalin*" OR "Razadyne*")). Two of the authors (GeP and GR) also reviewed the reference lists of the studies included in our analysis, narrative reviews and systematic reviews in order to identify additional eligible studies.
Study Selection
Two of the authors (GeP and GR) independently evaluated all identified titles and abstracts and excluded references not appropriate for our selection criteria. Eventual disagreement was sorted out by discussion with a third author (GiP). Studies were selected if they included: (1) patients with diagnosis of PD based on the UK Brain Bank criteria; (2) patients taking any dosage of rivastigmine, donepezil or galantamine; (3) more than 20 patients enrolled; (4) a duration of treatment at least 4 weeks; and (5) a double-blind, randomised placebo-controlled trials (RCTs) design. Reviews, case reports and animal studies were excluded. In synthesis, we focused only on RCTs comparing ChI versus placebo in patients with PD.
Risk of Bias
Two reviewers (GeP and GiP) independently classified RCTs according to the Cochrane assessment of risk of bias (RoB) and focusing, in particular, on sequence generation, allocation concealments, incomplete outcomes data, selective outcomes reporting, blinding of participants and blinding of outcome assessments. These features were classified as 'adequate' (low risk of bias), 'inadequate' (high risk of bias) or 'unclear'. If studies were adequate in all domains, they were considered to have a low risk of bias, but if they were inadequate or unclear in one or more key domain(s), they were considered to have a high or an unclear risk of bias, respectively. Disagreement was sorted out by discussion with a third author (MT). An Egger test for asymmetry (a funnel plot per each assessed outcome) was used to assess publication bias.
Data Extraction
Using an identical question sheet, two of the authors (GeP and GiP) independently recorded data including: study year, author first name, type of ChI, duration of the study, data regarding efficacy and safety outcomes, setting, clinical diagnosis, sample size and characteristics of patients. Disagreement was sorted out by discussion with a third reviewer (NF), when needed.
Outcomes
Primary efficacy outcomes were cognition, as measured by the Mini-Mental State Examination (MMSE) and rate of falls. Primary safety outcomes were rate of tremor and ADR. Secondary outcomes included: (1) language and memory functions, as measured by the Alzheimer Disease Assessment Scale-cognitive subscale (ADAS-cog); (2) global assessment, as measured by the Alzheimer Disease Cooperative Study (ADCS), the Clinician's Global Impression of Change (ADCS-CGIC or CGIC) or Global impression of improvement, the Clinician's Interview Based Impression of Change with caregiver/study partner input (CIBIC+); (3) behavioural disturbance, as measured by the 12-item or 10-item Neuropsychiatry Inventory (NPI); (4) disability, as measured by the Activities of Daily Living (ADL) or Schwab and England scale; (5) motor function, as measured by the Unified Parkinson Disease Rating Scale (UPDRS) III score; (6) mobility score, as measured by the gait and balance UPDRS III subscores; and (7) rate of death.
Statistical Analysis
For continuous end points, mean difference (MD) or standardised MD (SMD) was calculated for each outcome. The results were calculated using the method of the inverse variance. Using the guidelines of the Cochrane Handbook, missing SDs were calculated when not available.
For binary end points, ORs and 95% CIs were calculated for each outcome by using the intention-to-treat principle. We chose to use OR, instead of relative risk, because the published studies were heterogeneous in design, population, treatments, primary outcome measure and quality. An I statistic was used to assess the heterogeneity. Considering the latent clinical heterogeneity, a random effects model was used to summarise the data rather than the fixed effects model, independently of statistical evidence for heterogeneity. An I value greater than 25% has been suggestive of substantial heterogeneity.
All reported test results were two-tailed and statistical significance was set to a p value <0.05. Data analyses will be carried out with STATA V.12.
Sensitivity Analyses
We separately performed single meta-analysis evaluating efficacy and safety end points of rivastigmine, donepezil and galantamine as compared with placebo in order to confirm the consistency of the results. In addition, we performed one study-removed analysis to explore the influence of each study on the global risk estimate using the metaninf command (STATA V.12.0). To investigate the influence of potential effect modifiers on end points, we carried out a metaregression analysis with the metareg command (STATA V.12.0) to test age, sex (male %), disease duration in years, UPDRS III motor score baseline and MMSE baseline.
A random effects model was applied for metaregression analyses to take into account the mean of a distribution of effects across studies. Finally, in order to evaluate the additional component of variance, τ, we used the restricted maximum likelihood method to incorporate the occurrence of residual heterogeneity, not enucleated by the effect modifiers.
Source: ...