Jake M. Najman^{1, 2, 4} , Michael P. Dunne³ and Frances M. Boyle¹

Archives of Sexual Behavior, December 2006

The commentary offered by Rind and Tromovitch (2007) provides a welcome opportunity to revisit the findings presented in our article (Najman, Dunne, Purdie, Boyle, & Coxeter, 2005) and to reconsider their interpretation. We identify four main issues addressed in their commentary.

The first is concerned with the distinction between statistical significance and magnitude of effect. Rind and Tromovitch (2007) argued that our use of the words “significant impairment” implied that there was a stronger association than was, in fact, the case. They re-analyzed some of our data and reported correlation coefficients of around .10 – statistically significant associations but of modest magnitude.

We suggest that the measure of effect size presented by Rind and Tromovitch (2007) understates the magnitude of association. Their preference for the correlation coefficient involves a preference for an average measure of association over all values of both variables, yet the effects in which we have an interest involve a minority of cases at one extreme of the distribution. The advantage of our original form of presentation was that it enabled researchers to access much of the data in close to its original form, and to derive their own judgements about our claims, an opportunity which has been taken up in the commentary.

The second issue raised by the commentary concerns the suggestion that CSA and sexual dysfunction have a cause-effect association. Rind and Tromovitch (2007) argued that our data involve retrospective recall using a cross-sectional survey. They emphasised that this renders doubtful claims involving cause and effect.

We agree with this suggestion and made this point in the text of the article. However, this is not the totality of available evidence, and we reiterate our view that there is sufficient evidence to suggest that the claim of a cause-effect association is plausible and probable.

Thus, CSA was defined as occurring prior to 16 years of age while the sexual dysfunction was reported long afterwards. There is the evidence that CSA is a salient event. There is the evidence that the association is of moderate strength. There is also some face validity to the case for an association between CSA and sexual dysfunction. Granted the above are suggestive only; nevertheless, we argue that there is a reasonable case for asserting the plausibility of a causal association.

We suggest that Rind and Tromovitch (2007) have overestimated the capacity that research offers to disentangle cause and effect associations. Both cross-sectional and longitudinal data enable inferences about cause and effect to be drawn, but it is unusual for these inferences to be able to be drawn with great confidence.

In the case of longitudinal data, it is occasionally possible to identify a discrete event which precipitates a particular outcome. More commonly, a pre-existing health problem (e.g., anxiety or depression) or social circumstances may lead to differential exposures which precipitate particular outcomes (e.g., CSA).

From this perspective, not all children are equally liable to experience CSA. Rather, some are more vulnerable or more likely to be exposed. It makes more sense to think of cause-effect as a continuing sequence of related exposures and outcomes.

For research purposes, some variables are selected for study but the real world situation is only partially reflected in this research. It requires frequent, preferably prospective, assessments of exposure and outcome to enable more confident assertions of a cause-effect association. We suggest that Bradford Hill’s (1965) early article on causal reasoning remains the most sensible way to think about the plausibility of cause-effect associations.

Bradford Hill suggested that we consider eight criteria in judging whether an association may be causal. These are

Based upon a consideration of these criteria, we maintain that, on the balance of the available evidence, CSA is likely to “cause” sexual dysfunction.

Thirdly, Rind and Tromovitch (2007) also raised the possibility that the apparent association between CSA and sexual dysfunction is attributable to confounding. According to this view, family conflict/marital problems, for example, might lead to both CSA and to sexual dysfunction rather than CSA being a direct cause of sexual dysfunction.

There are three responses to this plausible claim.

Lastly, Rind, and Tromovitch (2007) raised the issue of the measurement of CSA. They argued that our definition of CSA was only one of a number that might be used. Of course, they are correct. We used a definition of CSA which involved the experience of “unwanted” sexual activity prior to 16 years of age.

This definition has advantages and disadvantages:

It is likely that there would be quite different health consequences depending upon which of the above types of CSA has occurred. If the former type of CSA constitutes a significant proportion of all CSA, then this might explain why the effect sizes were not as great as one might anticipate. There is certainly a need for more discriminating distinctions of the type, timing, and duration of CSA. Until such data become available, it is difficult to be more specific than assert, on the basis of our (and others’) research, that CSA appears to lead to sexual dysfunction.

Discussions about the scientific basis of CSA research are important. Along with other topics that attract prejudiced responses and negative stereotypes (e.g., illicit drug use, prostitution), many who engage in these debates use available research to advance a pre-existing set of moral preferences. These “moral” interpretations may be more pronounced when the topic of the research does not yield to the level of research control that an idealized view of science might recommend. However, good science is not done from a cookbook and is not a matter of sticking to a recipe. Difficult and sensitive topics demand study but frequently they are not able to be studied in the way other topics are addressed.

Rind and Tromovitch’s (2007) rejection of the external validity of clinical research illustrates this point. They argue that much of the research into the effects of CSA is of limited value because it has relied on retrospective interviews with people recruited from clinical contexts. They dismiss clinical research as invalid because the samples are unrepresentative, and they suggest (as did Pope & Hudson, 1995) that sick people attribute their condition to salient putative causes such as CSA.

An alternative viewpoint is that clinical research is valuable because we have learned much about the possible health consequences of CSA. Although some findings from individual clinical studies may be misleading, the composite picture from accumulated clinical research may be valid. It is only partly sensible to think in terms of distinct, clinical and non-clinical populations from which samples are drawn.

Over a one year period, the vast majority of the population reports attending clinical (medical) services. For example, in Australia, for the year 2002–2003, there were an average of 11.2 medical services per head of population per year (Australian Institute of Health and Welfare, 2004). Between 18% and 31% of the population have had a mental illness, meeting clinical criteria, in the previous year (Bijl, Ravelli, & van Zessen, 1998; Henderson, Andrews, & Hall, 2000; Jacobi et al., 2004; Kessler et al., 1994). To exclude these groups from consideration is as potentially misleading as to base one’s conclusions solely on samples derived from clinical settings.

When studying CSA at the population level, both perpetrators and subjects may be hidden and they may have reasons for avoiding researchers. Some may prefer not to be reminded of traumatic events experienced in the past. Some subjects may fear legal retribution.

The difficulties associated with doing good scientific research on hidden populations does not reduce the need for such research to be undertaken nor do the difficulties associated with doing such research deny the researcher the right to make judgements about what the research appears to indicate.

Interestingly, Rind and Tromovitch (2007) did not argue that CSA does not lead to sexual dysfunction. Rather, they argued that we have overstated the magnitude of association and that there is a need to have better control of the data before they are convinced.

We share some of their views but not others. Researchers need to be open in their practices, transparent in their reasoning, and present the evidence on which they have made their judgements. It is from these open discussions, we suggest, that knowledge is advanced. The contribution of Rind and Tromovitch to this debate is thus encouraging and we acknowledge their thoughtful commentary.

References