The authors describe some situations where meta-analyses would have be
en of great benefit if they have been done or sufficiently published.
They emphasize several problems ensued by this new method : if possibl
e, meta-analyses have to be done each time a new trial is analysed, ac
cording to a precise method. Cumulative meta-analysis should rely on a
regular and prospective registration of clinical trials. In all cases
, meta-analyses should be available to a large number of physicians, i
n the main interest of patients.