Some years ago nine areas were identified where either there was no consensus about the best approach or where inadequate information was available. The following comments briefly outline those areas and indicate progress made by SMG members.

1. Meta-analysis of continuous data
More work is needed on methods for combining results from trials with continuous outcome measures, especially with regard to the choice of effect measure, distribution of the data, baseline assessments, and handling missing data. Another issue is combining results from trials where continuous outcome data have been categorised using varying cut-points and numbers of groups, or where some trials have presented results grouped and others as continuous.

2. Meta-analysis of cross-over trials
Extensive work has been carried out to evaluate strategies for combining information from parallel group and cross-over trials, resulting in four publications (Elbourne et al 2002; Curtin et al 20002a,b,c). Further work is needed to consider the information that needs to be supplied in reports of cross-over trials to enable them to be used in this way.

3.Issues related to heterogeneity, subgroups and meta-regression
Recent work has seen advances in the quantification of heterogeneity (Higgins and Thompson 2002; Higgins et al 2003) and methods of meta-regression (Thompson and Higgins 2002, Higgins and Thompson 2004). It would be valuable to gain more insight into the merits of various strategies for combining trials when statistical heterogeneity is found. More empirical research into the effects of variation in methodology on heterogeneity would be valuable.

4. Summary statistics and study weights
More work is needed in examining the relative merits of odds ratios, relative risks, NNTs etc for meta-analysis and the presentation of results, and in examining the relative merits of different methods for combining these statistics, with the different weighting schemes they induce.

5. Meta-analysis of survival data
Methods exist to combine results from published survival studies where individual data are not available (Williamson et al 2002), but further work in this area may be warranted. It would be useful to study the loss of information (and power?) arising from classifying patients by presence or absence of the event of interest rather than by time to event. The effect of the varying length of follow up should be examined. Guidelines for minimum standards for reporting such trials should be developed.

6. Trial quality and reporting
Several members of the Cochrane Methods Statistics group were heavily involved in the 2001 update of the CONSORT statement (Moher et al 2001a; Altman et al 2001) and two papers evaluating the impact of the original (1996) CONSORT statement (Egger et al 2001b; Moher et al 2001b). Cochrane Methods Statistics members are involved in extending CONSORT to other trial designs - crossover, cluster (Campbell et al 2004), factorial, multi-arm, within-individual, and equivalence trials - and to the reporting of harms (Ioannidis et al 2004).

7. Meta-analysis of sparse data
Guidance has been needed on which methods to use when events are rare, or observed numbers of events are low. Two papers (Sweeting et al 2004; Bradburn et al 2007) will enable the SMG to develop better guidance in this area.

8. Meta-analysis of cluster randomised trials
Cluster randomized trials pose particular problems, not least because of limitations in the analysis and reporting of existing cluster randomized trials. Guidance is needed on how such trials should be incorporated into meta-analyses.

9. Bayesian methods of meta-analysis
Compared with existing methods for meta-analysis, Bayesian methods (e.g. Warn et al 2002) offer both advantages and disadvantages. Software limitations are a key obstacle to the use of Bayesian methods in Cochrane Reviews, and options need to be explored. Research issues are: Use of Bayesian methods to allow incorporation of 'external' data i.e. relevant data in awkward forms, such as qualitative data; investigation of influences of prior choice on results of Bayesian Meta-analysis; how to weight the various types of evidence in Bayesian meta-analysis; what does 'a priori' mean when retrospectively reviewing research?; Can we obtain a 'black box' implementation of simple Bayesian meta-analysis?

The preceding list is not an inclusive list of all current or intended topics of investigation. Note too that several members of the Cochrane Methods Statistics group have coauthored publications relevant to other methods groups: Individual Patient Data Meta-analysis, Screening and Diagnostic Tests, Reporting bias, and Non-Randomised Studies.

Further topics on the Cochrane Methods Statistics's research agenda include:

a. Different endpoints
It would be useful to consider what, if anything, can be done to combine trials that use different endpoints (this is a generalisation of the issue of combining data from trials where the same endpoint is assessed in different ways).

b. Graphical presentation
Various aspects of graphical presentation vary among published meta-analyses. The relative merits of these should be systematically reviewed. If possible, recommendations should be developed for standard graphical presentation. Aspects to consider include ordering of trials, symbols used, whether or not log scale used for treatment effect, and whether any additional graphs might usefully supplement the standard type.