Authors take issue with commentary on their paper

Editor—We commend the development of the BMJ's guidelines on educational interventions, which were used to assess our paper.¹ At several points in Ker's commentary on our paper, however, she raises the need for further information. We recognised this need when we prepared the manuscript. We were also aware, though, of the constraints on space; articles in the BMJ should not exceed 2000 words, and providing the requisite information would have made our paper substantially longer. An alternative would have been to give the information on the journal's website; this has the disadvantage of increasing the demands on authors but should be considered if the guidelines are to be fully applied.

We had some difficulty in following the comments on the randomisation approach. Firstly, it is unclear why Ker thinks the term “controlled” should be dropped; this is the conventional term for a trial such as this one, entailing randomisation to two or more groups, one of which is a control or comparison group.

Secondly, Ker suggests that contamination threatens the status of the control group. Presumably she is referring to the control group experiencing the effects of the intervention. It is certainly true that contamination is a major consideration in a study of this kind. We recognised this before we started the study and chose the sampling design specifically to minimise its likelihood. We believe that contamination was minimal, but even if this was not so it should have reduced the chances of finding the differences that emerged. This does not represent an absence of control.

Thirdly, Ker suggests that imbalances in the demographic variables in the intervention and control groups threaten the study's internal validity. Such an imbalance can arise in randomised controlled trials, and steps were taken in the data analysis to adjust for baseline differences between intervention and control groups.

Fourthly, Ker comments on the randomisation process. We believe that randomisation took place at the highest level possible for this kind of community based intervention. The suggested purposive sampling method, with presumably some form of pairwise or stratified randomisation, may be the ideal but raises a further barrier to the recruitment of adequate numbers of general practitioners. By making recruitment more difficult it would probably reduce further the representativeness of those practitioners who eventually participate.

Lastly, Ker's final sentence is ambiguous. A simple adjustment to the wording, replacing “the study” with either “a study” or “this or any other study,” would more adequately describe her meaning.