Assessing differential attrition in clinical trials: self-monitoring of oral anticoagulation and type II diabetes.
Heneghan C., Perera R., Ward A A., Fitzmaurice D., Meats E., Glasziou P.
BACKGROUND: Analyzing drop out rates and when they occur may give important information about the patient characteristics and trial characteristics that affect the overall uptake of an intervention. METHODS: We searched Medline and the Cochrane library from the beginning of the databases to May 2006 for published systematic reviews that compared the effects of self-monitoring (self-testing) or self-management (self-testing and self-dosage) of oral anticoagulation or self-monitored blood glucose in type 2 diabetics who were not using insulin. We assessed all study withdrawals pre-randomization and post randomization and sought information on the reasons for discontinuation of all participants.To measure the differential between groups in attrition we used the relative attrition (RA), which is equivalent to relative risk but uses attrition as the outcome (i.e. attrition in intervention group/attrition in control group). We determined the percentage drop outs for control and intervention groups and used DerSimonian and Laird random effects models to estimate a pooled relative attrition. L'abbe type plots created in R (version 2.0.2) were used to represent the difference in the relative attrition among the trials with 95% confidence areas and weights derived from the random effects model. RESULTS: With self-monitoring of blood glucose in type 2 diabetes, attrition ranged from 2.3% to 50.0% in the intervention groups and 0% to 40.4% in the control groups. There was no significant difference between the intervention and control, with an overall RA of 1.18 [95% CI, 0.70-2.01]. With self-monitoring of oral anticoagulation attrition ranged from 0% to 43.2% in the intervention groups and 0% to 21.4% in the control group. The RA was significantly greater in the intervention group, combined RA, 6.05 [95% CI, 2.53-14.49]. CONCLUSION: This paper demonstrates the use of relative attrition as a new tool in systematic review methodology which has the potential to identify patient, intervention and trial characteristics which influences attrition in trials.