A systematic review of psychological interventions for sexual offenders I: Randomised control trials

Abstract

Despite the growth in literature on the merits of treatment, opinion remains divided about the lack of relevant evaluative research on the effectiveness of psychological treatments for sexual offenders. This study reports a systematic review of randomised control trials reporting the effectiveness of sexual offender treatment programmes. Electronic and hand searches were carried out for randomised control trials published between 1998 and 2003. Searches revealed nine randomised control trials (RCTs) and these contained data on the treatment of over 500 men, 231 of whom have been followed up for 10 years. Analysis of the nine trials showed that cognitive behavioural therapy (CBT) in groups reduced re-offence at one year compared with standard care (n = 155) but increased re-arrest at 10 years. The implications of these findings and the subsequent use of randomised control trials in clinical policy and research are discussed.

Keywords:

• Sexual
• offender
• treatment
• randomised
• control trial
• review

Acknowledgements

A section of this study was published electronically as a Cochrane Review (Issue 3, 2004) with the support of the NHS Research and Development Programme. We are grateful to members of the Cambridge Socio-Legal Group, Conor Duggan, Clive Adams, Mark Fenton, and Tracey Kenworthy, for comments and contributions on a much earlier version of this article. Thanks also to Dr Julian Higgins at the MRC Biostatistics Unit, Cambridge, UK, from whom statistical advice was sought.

Notes

1 This outcome and all other outcomes with ∗ are primary outcomes. These were chosen before analysing the data in order to focus the review on outcomes that would make differences to the practices of, primarily, clinicians, recipients of care, and policy-makers. Other outcomes were included in the recognition that, should they be reported, they too may well be of interest. The so-called objective measures of outcome, such as pupilography or eye gaze direction, were not a focus of this review. We are aware of the difficulties of getting accurate outcome measures in this area and that objective measures are used. These proxy measures are difficult to interpret and we recorded them only in Table I in outcomes of included studies.

2 This is calculated using the mean number of participants per cluster (M) and the intraclass correlation co-efficient (ICC). Design effect = 1 + (M– 1) ∗ ICC (Donner & Klar, 2002). A figure of .1 was assumed, should the ICC not be reported (Ukoumunne, Gulliford, Chinn, Sterne, & Burney, 1999).

3 As studies have shown that poor reporting of randomisation is associated with increased estimates of effect (Chalmers, Celano, Sacks, & Smith, 1983; Schulz, Chalmers, Grimes, & Altman, 1994) it has to be assumed that all studies in this systematic review may report results prone to favour the experimental interventions.