Study identification, data extraction and analysis
Experimental studies of disease modifying drugs in EAE were identified from PubMed and Web of Science by searching all articles from 1970 onwards. Additional publications were identified from the references of identified publications, non-systematic reviews and book chapters. The search strategy employed the following keywords: experimental autoimmune/allergic encephalomyelitis/encephalitis, animal models and multiple sclerosis, demyelination, interferon, interferon-alpha, interferon-beta, type 1 interferon, glatiramer, copaxone, copolymer, copolymer-1, integrin, α4β1, VLA-4, natalizumab, tysabri, mitoxantrone, anthracene dione, novantrone, FTY720, fingolimod, S1P, laquinimod, ABR-215062, roquinimex, linomide, cladribine, leflunomide, teriflunomide, panaclar, fumarate, fumaric acid esters, disease modifying treatments, disease modifying drugs. Conference abstracts were not included.
The profusion of EAE methodologies makes analysis difficult. Studies of the same treatment may use different animals, different methods of EAE induction, different routes and dosages of administration, different outcome measures or scoring systems and different time points. Moreover, raw data are frequently unpublished.
We have included in our analysis studies using clinical (EAE) score as an outcome measure. Where studies provided clinical data at a number of time points we chose the time point most relevant to the intervention and outcome. In studies attempting to prevent onset of EAE, we typically used the peak disease score attained. Distinct experiments within a paper were considered separately.
Heterogeneous outcome measures were standardised by converting all results to a percentage of maximum score. In the majority of studies, the maximum score (usually 4 or 5) is defined as terminal paralysis or death. Some older studies define a lower maximum such as complete hindlimb paralysis; these were treated in the same way for the purposes of standardisation.
Some studies use a binary outcome measure ‘incidence’ defined as presence or absence of paralysis; several such incidence studies has been performed with Glatiramer acetate and were therefore analysed separately. For other agents, these studies were excluded. Studies were also excluded if insufficient data were provided for the purposes of our analysis. These studies have nonetheless been included in tables for reference and been marked with a star. We employed the Cochrane Review Manager software version 5 (http://www.cc-ims.net/revman) to collate and analyse studies.
Data from trials of different sizes were incorporated in our meta-analysis. In order to offset the skewing of data resulting from these disparities, each trial was weighted. Different methods are available for this; our meta-analysis employed a random effects weighting which incorporates heterogeneity in true treatment effect (DerSimonian and Laird 1986).
The resulting data are summarised in forest plots where appropriate. The adjacent tables display the mean standardised clinical score and standard deviation, number of animals, study weighting and a standardised measure of effect, calculated as above. For incidence studies, data are expressed as odds ratio, that is, ratio of the odds of paralysis occurring in the treatment group to the odds of paralysis in the control group.
Funnel plots to assess for publication bias were drawn for the agents most intensively studied in EAE – glatiramer acetate (GA) and type-1 interferons.