Quality of meta-analyses in terms of actual statistical analysis (an empirical investigation)

A paper by Jadad et al suggests that the quality of Cochrane Collaboration (CC) reviews is

generally better than those done by others. However, as has been pointed out, the quality tool that

has been used to assess such reviews places little premium on correct analysis. An informal

investigation suggests that CC analyses are often deficient in some respects. In particular, there is

a tendency to count control arms twice, to handle cross-over and cluster randomised trials

inappropriately, to ignore covariate information and to impute missing values when it would

probably be better not to include the trials.

This project will consist of establishing a suitable quality instrument, devising an appropriate

sampling plan, selecting some studies from the CC database http://www3.interscience.wiley.com/cgi-bin ,

summarising the results and drawing suitable inferences.

References

Jadad AR, Moher M, Browman GP, Booker L, Sigouin C, Fuentes M, et al. Systematic reviews

and meta-analyses on treatment of asthma: critical evaluation. British Medical Journal

2000;320(7234):537-40.

Senn SJ. Review is biased. British Medical Journal 2000;321:p297.

Senn SJ. An unreasonable prejudice against modelling? Pharmaceutical Statistics 2005;4:87-89.

Back to Stephen Senn's Research Page