W. L. Marshall & L. E. Marshall
Rockwood Psychological Services, Ontario, Canada
[Sexual Offender Treatment, Volume 5 (2010), Issue 2]
In this paper we address issues raised by the talks Hanson and Rice gave at IATSO's 2010 conference in Oslo. Specifically they both indicated that in their views treatment for sexual offenders had not been satisfactorily demonstrated to be effective. Their basis for this claim was that treatment benefits can only be confidently inferred from Random Control Trials (RCT). Rice further suggested that until RCTs are conducted with sexual offender treatment, it may be that such treatment has negative effects (i.e., increases recidivism). We disagree with these suggestions and in this paper we outline the bases for our disagreement.
Key words: sexual offender treatment, effectiveness, randomized controlled trials
At the recent 2010 bi-annual conference of the International Association for the Treatment of Sexual Offenders (IATSO) held in Oslo, two plenary speakers, Karl Hanson and Marnie Rice, presented challenging papers. Both offered well-articulated points-of-view that represented their views on the appropriate way to evaluate treatment and on what the present literature indicates about the effectiveness of sexual offender treatment. In addition to features of their talks with which we agree, there were two points with which we take issue. We will deal with Dr Hanson's paper first and then the paper by Dr Rice.
Dr Hanson began by discussing the need for treatment providers to attend to the principles of effective offender treatment that Andrews and his colleagues (Andrews, 2001; Andrews & Bonta, 2006; Andrews, Bonta, & Hoge, 1990; Andrews, Bonta, & Wormith, 2006) have extracted from their meta-analyses of a large number of outcome studies of offender treatment programs. We are fully in agreement with Dr Hanson's position on this issue, but to be clear, we will briefly describe these principles. Andrews and his colleagues have shown that three principles are related to effective outcomes: (1) risk, (2) needs, and (3) responsivity. The first principle indicates that treatment resources should be allocated to the highest risk offenders. The second principle directs treatment programs to address those deficits (i.e., needs) that have been shown to predict reoffending (the so-called "criminogenic features"). The third principle has two components: general responsivity and specific responsivity. General responsivity requires treatment providers to employ an empirically sound program. Andrews and Bonta (2006) found that CBT programs were consistently effective across a range of offenders so they indicated that providing CBT would meet the general responsivity principle. The specific responsivity principle requires therapists to adjust their approach to treatment to meet the unique features of each client including each client's learning style.
Turning now to the other aspect of Hanson's (2010) talk, we will focus on the claim that to date the effectiveness of sexual offender treatment has not been satisfactorily demonstrated. This claim, which is shared by others (e.g., Quinsey, Harris, Rice, & Lalumière, 1993; Rice & Harris, 2003; Seto et al., 2008) and was restated by Rice (2010) in her talk, rests on framing the question as "Is sexual offender treatment effective?" The question is stated in this way because the appropriate empirical test is seen to be an examination of the null hypothesis. Seto et al. (2008) described the relevant null hypothesis as being "that treated and control groups do not differ in recidivism" (p. 254). The usual interpretation of the question phrased in this way is that a positive answer essentially requires that the bulk of acceptable studies show positive effects. Indeed the assumption guiding meta-analyses of sexual offender treatment outcome studies is that there is at least sufficient commonality among studies entering these analyses to justify viewing them as approximately equivalent in their programmatic features. In fairness, Hanson's two published meta-analyses (Hanson et al., 2002; Hanson, Bourgon, Helmus, & Hodgson, 2009) did distinguish cognitive-behavioral approaches (including those with relapse prevention components) from other types of treatment. Hanson's studies showed that it was only cognitive behavioral approaches (CBT) that reduced recidivism. Lösel and Schmucker's (2005) meta-analysis of sexual offender programs revealed much the same results. Nevertheless in both reports all programs described as CBT were treated as equivalent.
Describing programs as "Cognitive-behavioral", or "Cognitive-behavioral/Relapse prevention", implies a uniformity that does not appear to be present. In fact, the several surveys of sexual offender treatment programs completed by the Safer Society (Burton & Smith-Darden, 2001; McGrath, Cumming, & Buchard, 2003; McGrath, Cumming, Burchard, Zeoli, & Ellerby, 2010) reveal considerable differences across programs despite the fact that the majority of respondents described their programs as CBT or CBT/RP. In fact what is surprising about the results of these surveys is that very few programs target all known criminogenic features (i.e., those features shown by research to predict reoffending) and most target a variable number of features that as we (Marshall & Marshall, in press; Marshall, Marshall, & Ware, 2009) have shown do not predict reoffending (i.e., denial and various other so-called "distorted cognitions").
This is a potentially serious problem since Andrews and Bonta (2006) have shown that targeting noncriminogenic features of offenders reduces the beneficial effects that would otherwise result from treatment. The correlations between addressing various noncriminogenic needs and the overall effect size for treatment, ranged from r = -.18 to r = -.20. These are statistically significant negative influences. Just as problematic is the fact that the important aspects of the responsivity principle are given little systematic attention in the surveyed programs. According to Andrews and Bonta (2006) it is particularly that aspect of responsivity concerning therapist style and training that critically influences effectiveness. In their discussion of the general responsivity principle, Andrews and Bonta (2006) show that it is not so much the adoption of a CBT approach that produces effectiveness, but rather whether or not therapists are carefully selected for, and trained in, the appropriate skills. These skills include: empathy, warmth, respect, interest, and nonblaming communication. The latter has been described in our studies (Marshall, Serran, Moulden, et al., 2002; Marshall, Serran, Fernandez, et al., 2003) as "confrontation" which we showed to be a style of challenging sexual offenders that markedly reduced any positive effects of treatment. Dowden and Andrews (2003) demonstrated that when programs met these therapist style and training criteria, the treatment effect size was significant (ES = .39) whereas when these criteria were not met there were essentially no benefits from treatment (ES = .04).
Given this variability across treatment programs it makes little sense to expect them to have uniform effects. When different physicians administer the same medication to patients with the same disorder, we might reasonably expect the same results; if they were administering different medications we would definitely not expect the same result. So the question concerning the effectiveness of psychological treatment for sexual offenders, as it is phrased above, is really an inappropriate question. Our preferred way to state the question is "Can treatment for sexual offenders be effective?" Phrased in this way we only need one example of an effective program to confidently assert a positive answer.
If there is, indeed, one demonstrably effective program then the sensible response would be to ensure that all other programs match the content and delivery style of this program. Since both Hanson (2010) and Rice (2010) pointed to the effectiveness demonstrated by an RCT evaluation of Multi-systemic Therapy (MST) with juvenile sexual offenders (Borduin, 2010) then this program should be the basis for designing other programs with these young offenders. Our view is that the literature indicates that more than one program for adult sexual offenders have demonstrated effectiveness (see the review we completed in our latest book, Marshall, Marshall, Serran, & O'Brien, 2011). Indeed in the meta-analysis of sexual offender programs reported by Hanson et al. (2002) an overall effect for treatment was reported and Lösel and Schmucker (2005) also found an overall positive effect. Furthermore, as Hanson and Bussière (1998) noted regarding sexual offender programs "Even if we cannot be sure that treatment will be effective, there is reliable evidence that those offenders who attend and cooperate with treatment programs are less likely to offend than those who reject interventions" (p. 358).
In their talks, both Hanson and Rice declared the only basis upon which positive effects of treatment can be confidently inferred is an evaluation employing the Random Controlled Trial (RCT). The RCT design, to be clear, requires the random allocation of treatment volunteers to either treatment or no treatment. Both Hanson and Rice indicated that they hold California's Sex Offender Treatment and Evaluation Project (SOTEP) to be an exemplary application of the RCT design. Other advocates of the RCT design (e.g., Seto et al., 2008) likewise praise the SOTEP project for its experimental design. This project was initiated in the early 1980s (Marques, 1984) and its final results were reported in 2005 (Marques, Wiederanders, Day, Nelson, & van Ommeren, 2005). These results revealed no differences in recidivism rates between the treated group and the untreated volunteers, thus failing to reject the null hypothesis. This failure to reject the null hypothesis was taken by Rice and Harris (2003) to mean that "the effectiveness of psychological treatment for sex offenders remains to be demonstrated" (p. 428). However, in Marques et al.'s (2005) more detailed analyses of their results, it was reported that those treated offenders who achieved the goals of the program (i.e., changed in the appropriate directions on measures of the targets of treatment) had markedly reduced reoffense rates. Thus, our question "Can treatment be effective?" is clearly answered in the positive by this supposedly elegant RCT study. The questions that should be raised about the SOTEP study, then, should focus on why did these offenders succeed and why did the others fail? Examining why some succeeded might offer insights into the development of effective programs, rather than simply focussing on the overall failure of the treated group.
Marshall and Marshall (2007) indicated that one of the various problems they identified with the RCT approach is that it typically requires strict adherence to a treatment manual so that the internal validity of the study is assured. Internal validity refers to the degree to which treatment as delivered adheres to the detailed specifications of the treatment manual. This is done to ensure that it is the specifics of the program that produces the results and so that the program, and its results, can be replicated by others. The problem, as many commentators (Everitt & Wessely, 2004; Farrington, Gottfredson, Sherman, & Welsh, 2002; Gondolf, 2001; Hollin, 2006; Seligman & Levant, 1998) have pointed out, is that the RCT tends to sacrifice external validity (i.e., the generalizability of the results to clinical practice) in the interests of maintaining internal validity. Unfortunately since it is easier for manuals to specify treatment targets (and procedures to modify them) than it is to describe therapeutic process skills (i.e., therapist style and therapeutic alliance), the effect of adhering to detailed manuals is to diminish the role of the therapist in the delivery of treatment (see Marshall, 2009, for a detailed analysis of the problems involved in designing useful treatment manuals). It has been shown that for the treatment of sexual offenders (Serran, Fernandez, Marshall, & Mann, 2003), offenders in general (Marshall & Burton, 2010), and for all Axis 1 disorders (Norcross, 2002) the way in which treatment is delivered accounts for far more of the observed treatment benefits than does the application of appropriate procedures. Given these results, the RCT's design requirement of strict adherence to a treatment manual seems likely to diminish any potential benefits that may be derived from treatment. The fact that some treated offenders in the SOTEP project displayed reduced rates of recidivism as a result of successfully changing on the targets of treatment, suggest that the unique features of these successful clients were a match for the standardized treatment program offered by SOTEP.
It is the restricted capacity of RCT studies to fully implement the features of the responsivity principle outlined by Andrews and Bonta (2006) that limits the relevance of such studies for clinical practice. This limitation to RCT studies has been pointed out by numerous clinicians and researchers in the general clinical literature (Goldfried & Wolfe, 1996; Howard, Moras, Brill, Martinovich, & Lutz, 1996; Persons, 1991; Persons & Silberschatz, 1998; Seligman, 1995; 1996), so we are not alone in our concerns about the relevance of RCT studies. As Seligman and Levant (1998) note, outcome research has little practical relevance unless the study examines "therapy as it is actually delivered in the field" (p. 211). Persons and Silberschatz (1998) similarly note that the results of "RCTs have minimal impact on the practice of psychotherapy because the methods and findings do not address the issues and concerns of the practicing clinicians" (p. 128).
An additional interesting possibility was raised by Rice in her Oslo talk, a point she had made in an earlier paper (Rice & Harris, 2003) and that was also noted by Seto et al. (2008). Both Rice (2010) and Seto et al. (2008) suggested that it might be that the treatment of sexual offenders has negative effects. In fact, Seto et al. (2008) were quite clear on this point noting that unproven treatments (as they view sexual offender treatments to be) might "unintentionally increase recidivism" (p. 250). They even suggest how this might happen. "In sex offender treatment, recounting offense details in acceptance of responsibility and relapse prevention exercises might expose other offenders to new sexual content and new methods for accessing victims" and "The use of victim empathy exercises may fuel sadistic fantasies among the subgroup of sex offenders for whom victim suffering and distress are arousing rather than upsetting" (p. 250). Seto et al. (2008) describe these potential negative effects of involvement in treatment as "plausible". In the absence of data these are not unreasonable suggestions. Rice then reviewed medical interventions showing that several well established practices, when subjected to an RCT examination, revealed iatrogenic effects; patients got worse as a result of treatment. Is it really likely, or evident, that sexual offender treatment will produce negative effects and, if so, on which offenders?
We could find no evidence in the literature of negative effects. In fact, of all the studies entering various reviews, not one of the reports that failed to find benefits for sexual offender treatment indicated increased recidivism as a result of treatment. Since at least some offenders in the SOTEP study displayed significantly lower reoffense rates than did the untreated group, then, far from producing iatrogenic effects, it appears that even this highly structured approach to treatment can be effective for at least some clients. Perhaps even more salient to this issue, is the study reported by Rice, Quinsey and Harris (1991) which Rice mentioned in her presentation. Although this was not an RCT designed evaluation, this has not prevented those who advocate the value of the RCT design, and who also question the value of sexual offender treatment, from citing this and other non-RCT studies as illustrations of the failure of treatment to reduce recidivism. While the Rice et al. (1991) study failed to show any overall benefits, it certainly did not reveal any negative effects; the treated group and the untreated group showed essentially the same rates of recidivism at follow-up. Similarly a report by Quinsey, Khanna and Malcolm (1998) described their (non-RCT) evaluation of a Canadian prison-based program for very high risk sexual offenders. Note that Quinsey (Quinsey et al., 1993) also takes the position that sexual offender treatment has not been shown to be effective and that the RCT design is the only basis upon which conclusions about effectiveness can be made. In their report, Quinsey, Khanna and Malcolm compared recidivism rates for the treated subjects, all of whom were deemed on the basis of a comprehensive evaluation to need treatment, with a group of untreated offenders all whom were found, by the same evaluation procedures, to not be in need of treatment. Aside from the fact that it seems unlikely that with such high risk offenders the effects of treatment would reduce their reoffense rates to lower than that of those deemed not to need treatment, the results of this study did not indicate any negative effects from treatment. Both groups had essentially the same rates of recidivism.
In conclusion, we believe the empirical literature justifies a positive answer to the question "Can sexual offender treatment be effective?" We agree that there are at present limited studies reporting positive outcomes but there are more than one. We also agree that there are even fewer studies that reveal positive outcomes as the result of an RCT designed study. But there is at least one (Bourduin, Schaeffer, & Heiblum, 2009). When the question of effectiveness is phrased in our terms, it requires no more than one effective program for the answer to be in the affirmative. Furthermore, we confidently interpret the literature as indicating no evidence of negative effects arising from sexual offender treatment.
Finally we have been engaged in debates about the effectiveness of treatment for sexual offenders for two decades (Marshall, Jones, Ward, Johnston, & Barbaree, 1991; Marshall & Pithers, 1994). We respect those who disagree with us and we believe they offer sound arguments which we feel compelled to counter. Such debates reflect the scientific health of our field and we hope to see more debates on more issues as our field matures.
W. L. Marshall
Rockwood Psychological Services
303 Bagot Street, Suite 403
K7K 5W7, Canada