by MEC van Eiik et al (The Netherlands)
This paper describes the results of a randomised controlled trial with three arms (individual visits, group visits and control) to assess the effect of academic detailing visits on the prescribing of anticholinergic antidepressants in the elderly.
There are a number of points relating to the methodology and the statistical analysis of the data that need to be addressed by the authors.
Design and Methods:
| 1. | The participation rate appears rather low, 61% (43/70) of GPs in the individual visit arm and 5 of the 7 groups allocated to the group visit arm. What reasons were given for non-participation? |
| 2. | What was the power of the study to show significant differences in prescribing between the three groups? The 95% confidence intervals of the IRRs (Table 4) seem rather wide. |
Statistical Analysis and Presentation of Results:
| 3. | The description of the statistical analysis (Page 12) is inadequate to determine whether the correct adjustment for cluster randomisation was carried out. More details than a simple reference to the software program Spida are required. Furthermore, the presented results are taken from the unadjusted analyses. The statement that the adjusted analyses, "..did not materially influence outcome" (Page 12) is unsatisfactory. What difference did it make? |
| 4. | Baseline incidence rates were very different between the three arms of the trial (Table 1). Although the statistical analyses adjusted for these differences, it is difficult to judge whether the observed changes over the study period might have been due to a 'regression to the mean' effect. |
| 5. | There is no direct comparison of the 'group' and 'Individual' arms. |
| 6. | Abstract. "Both the individual and the group visits significantly decreased the incidence of anticholinergic ADs and increased the incidence of non-anticholinergic AD use in the elderly". This statement is rather misleading. The decrease in anticholinergic ADs was not statistically significant for the individual arm and the increase in non-anticholinergic ADs was not significant in the group arm. |
Julie Morris
CHECKLIST FOR STATISTICAL REVIEW OF PAPERS FOR THE BMJ
"Unclear" and "No" answers to be addressed by authors
when revising their papers – in particular any marked with *
| BMJ Ref No: | Date of Review: |
| Design Features | ||
| 1. | Was the objective of the study sufficiently described? | Yes |
| 2 | Was an appropriate study design used to achieve the objective? | Yes |
| 3. | Was there a satisfactory statement given of source of subjects? | Yes |
| 4. | Was a pre-study calculation of required sample size reported? | No* |
| Conduct of study | ||
| 5. | Was a satisfactory response rate achieved? | See report |
| Analysis and Presentation | ||
| 6. | Was there a statement adequately describing or referencing all statistical procedures used? | No* |
| 7. | Were the statistical analyses used appropriate? | Unclear* |
| 8. | Was the presentation of statistical material satisfactory? | Yes |
| 9. | Were confidence intervals given for the main results? | Yes |
| 10. | Was the conclusion drawn from the statistical analyses justified? | Unclear – see report |
| Abstract | ||
| 11. | Does the Abstract give a fair summary of the paper? | No |
| Recommendation on Paper | ||
| 12. | Is the paper of acceptable statistical standard for publication? | No |
| 13. | If 'No' to Question 12 could it become acceptable with suitable revision? | Yes |
| Reviewer: | Julie Morris | See again? | Yes |