Dropout rates in placebo-controlled and active-control clinical trials of antipsychotic drugs: a meta-analysis
Kemmler G, Hummer M, Widschwendter C and Fleischhacker WW;
Commented by , 17 Jan 2006
Aim
Some years ago an analysis of the controlled clinical trials register of the Cochrane Schizophrenia Group found increasing dropout rates since 1950. One reason for these high dropout rates may be the use of placebo groups.
The argument made is the following: If the investigators know that a proportion of the patients receives placebo, they might be afraid of giving ineffective care to some participants. Therefore, in the case of non-response they might be more likely to take the patient out of the study and thus increase dropout rates.
The authors investigated this assumption in a meta-analysis of randomized double-blind trials.
Method
The authors searched MEDLINE to identify double-blind, randomized, controlled clinical trials of the new generation antipsychotics risperidone, olanzapine, quetiapine, amisulpride, ziprasidone, and aripiprazole. The following criteria had to be met in addition: the study participants had to be patients with schizophrenia or schizoaffective disorder, the primary outcome had to be change of psychopathology, and they only included short term studies with a duration below 12 weeks. Dropout rates in placebo-controlled trials (PCTs) were compared with those of active-control trials.
Results
Thirty-one trials with 10.058 participants were included. The active treatment arms of placebo-controlled trials had significantly higher weighted mean dropout rates than active-control trials.
The percentages were 48.1% (PCTs) vs 28.3% (active-control trials) for second-generation antipsychotics (odds ratio, 2.3; 95% confidence interval, 1.6-3.5) and 55.4% (PCTs) vs 37.2% (active-control trials) for conventional antipsychotics (odds ratio, 2.1; 95% confidence interval, 1.3-3.4).
As expected, the attrition rates of the placebo groups within PCTs were significantly higher than in the second-generation antipsychotics groups (60.2% vs 48.1%; odds ratio, 1.6; 95% confidence interval, 1.4-1.9). In the subset of studies which had both a second-generation and conventional antipsychotic arm the dropout rates were significantly higher in the conventional antipsychotic groups.
Dr Leucht's comments
The high dropout rates in antipsychotic drug trials are a major problem for research in this area. In recent years even in short term (e.g. 4-12 weeks) studies the dropout rates have often been higher than 40%. It is unclear how such studies can be analysed in a reasonable way.
One approach is last-observation-carried forward. Here, if a patient drops out after for example two weeks in a six weeks study, his two weeks result is used in the endpoint evaluation ("carried forward to endpoint"). But this model depends on the correctness of the assumption that the symptoms of those who dropped out would not have improved (or worsened) if the patients had stayed in the study.
Another model is to only include those patients in the analysis who completed the study. But this analysis usually overestimates treatment effects, because those patients who drop out usually have worse outcomes.
Newer approaches are imputation strategies with which it is tried to extrapolate the outcome of dropouts by using the results of study completers with similar characteristics (for example so called mixed models analyses).
Nevertheless, these models are limited, as well, if the dropout rates are very high – e.g. 72% in the famous CATIE study. Strategies to reduce dropouts in both active comparator and placebo-controlled trials are urgently needed.