Methodological issues concerning evaluation of treatment for sexual offenders: randomization, treatment dropouts, untreated controls, and within-treatment studies.

Abstract

The value of randomized controlled trials in evaluation of sexual offender treatment has been questioned. Concern was expressed that randomization fails to produce equivalent samples, without apparent appreciation this is inevitable when variables are distributed by chance; lack of equivalence is controlled by use of tests of significance. A further uncriticized and inappropriate procedure in treatment evaluation is separation of the results of subjects who did not complete treatment from those who did, when the outcome of the former group was known. Despite an APA Task Force recommendation, no attention has been given to the consistent finding that no treatment is less effective than placebo psychological therapies. The significance of Type II errors is discussed and the recommendation criticized that within-treatment research be encouraged as an alternative to outcome research. Demonstrating a within-treatment response when that response is associated with a better outcome does not necessarily mean that the treatment was effective. Subjects with a good prognosis could be more able to demonstrate a within-treatment response to the treatment. Nonrandomized matched samples do not adequately control all sample differences. The post hoc statistical reversal of a reported trend for sexual offenders treated with relapse prevention to show a worse outcome than untreated offenders, in order to correct lack of equivalence of the two groups, is considered inappropriate. That relapse prevention was found less effective than no treatment raises the possibility that it has a negative effect. To continue the use of relapse prevention other than in randomly controlled evaluative studies would appear to be unethical.