R.J.K. Khan, M. Parker, D. Fick, and A. Keogh reply:
We thank Mr. Dobson and Mr. Nguyen for their letter. The issue that they
raise is one of semantics. Randomized trials may be blinded (e.g., with use of
random numbers or sealed identical envelopes) or quasi-randomized (e.g., with
use of alternation or odd/even numbers). Therefore, we believe that our title
is not misleading. Moreover, the inclusion of such studies is accepted
protocol for other evidence-based organizations such as Cochrane.
We were explicit about the inclusion criteria in the methodology section of
our paper, stating that we would consider quasi-randomized controlled trials
as well as trials with inadequate concealment of treatment allocation. We also
were transparent and classified the randomization type of each study in the
results section (see Table I for the characteristics of the included
studies)
It would be ideal to include only pure randomized trials in a
meta-analysis. However, in the field of surgery, and in orthopaedics in
particular, there are, regrettably, few such studies. Therefore, to exclude
quasi-randomized controlled trials would reduce the numbers substantially.
Like other researchers, we hope that the number of quasi-randomized trials
will continue to decline as researchers use a blinded pure method of
randomization.