[Page728]

Although we agree with Rind et al. ( 1998) that CSA does not inevitably lead to intense and pervasive harm in all individuals, our conclusions, which are summarized in Table 9, differ from those of Rind et al. in almost every other area.

It is also important to note that although the results of Rind et al.'s meta-analysis support those of previous reviews that show that extreme long-term effects from CSA are not inevitable (Beitchman et al., 1991, 1992: Browne & Finkelhor, 1986; Kendall- Tackett et al., 1993), their findings also demonstrate a significant association between acknowledging a history of such abuse and an increased vulnerability to a wide range of mental health and social problems in adult life.

The fact that many of these associations were small should not be considered surprising given that the use of correlations coupled with attenuation problems served to minimize the appearance of meaningful effects. In addition, it should be remembered that Rind et al. studied a healthy sample and that the meta-analysis tapped a very broad range of sexual experiences, many of which involved no physical contact. It is well recognized that heterogeneity in abuse severity can distort estimates of the consequences of CSA, as the lack of measurable consequences for the majority who experienced milder forms of abuse are likely to obscure the significant consequences experienced by the smaller number of people who experienced more severe forms of CSA (Haugaard, 2000).

An interesting aspect of the data presented by Rind et al. (1998) is the fact that abused students were less well adjusted than controls in 17 of 18 symptoms examined, with a high degree of consistency of the individual effect size magnitudes (most uncorrected rs = .11 + .02) across different dependent variables (e.g., anxiety, depression, psychotic symptoms; see Rind et al., 1998, Table 3).

Although these results are similar to those of other studies indicating a global and diffuse impact of trauma on children (McCloskey & Walker, 2000), Rind et al. averaged the 18 different symptoms together and treated them as if they measured equal portions of a single construct (i.e., maladjustment). This is necessarily an underestimate unless there is perfect overlap of symptoms.

Although the literature does show substantial co-morbidity of mental disorders (Gotlib, 1984), on the order of

most studies conclude that the overlap is only partial at best (Sandanger et al.. 1998).

Moreover, syndromes often occur at different times and at different periods in the life span, and co-morbidity is lower among those with less severe symptoms (Schatzberg et al., 1990).

Thus, in the young and relatively high functioning samples reviewed in the Rind et al. (1998) meta-analysis, there should be substantial independence of symptom clusters, on the order of at least 50%. Thus, a practical assessment of the clinical impact of CSA on mental health would be more than just the average of the effects of CSA on the different outcomes.

To the extent that symptoms such as alcohol problems, sexual adjustment, dissociation, and eating disorders are distinct, effects would be additive in their effect on overall adjustment. For example, clinical studies have shown that co-morbid depression plus anxiety results in more severe symptoms than either anxiety or depression alone (Angst, 1997).

Epidemiologic surveys have also

[Page729]

shown that even when symptoms are not sufficiently severe to fulfill criteria for a specific diagnosis of either depressive or anxiety disorder, patients presenting with both anxiety and depressive symptoms often display significant levels of functional impairment, have a high use of non psychiatric medical care, have long-lasting symptoms, and are at risk for more severe psychiatric disorders (Boulenger, Foumier, Rosales, & Lavallee, 1997).

Overall, the significant co-morbidity suggested by Rind et al.'s data suggests that some abused students may experience significant levels of functional impairment, especially if the majority of the morbidity originates in the smaller number of students who experienced the most severe forms of abuse.

Actually, when placed in the proper context, the results of the current meta-analysis are quite sobering. Despite

the meta-analysis yielded effect sizes that could be interpreted as quite meaningful, especially when one considers that small r values can reflect serious effects when a condition is as prevalent as CSA.

As a comparison, Ondersma, Chaffin and Berliner (1999) meta-analyzed data from 14 studies on smoking and the development of lung cancer and found an effect size of r = .17. [*15]

[*15] Although Ondersma et al. (1999) reported that the effect size was .12, an erratum was later published saying the actual effect size was .17. (Erratum, 1999).

Thus, the effect sizes Rind et al. (1998) reported for the correlation between CSA and anxiety. depression, paranoia, and psychotic symptoms (.13 for each) are only slightly smaller than the effect of cigarette smoking on lung cancer in the general population. Based on a correlation of .17, many might be tempted to argue that smoking only plays a negligible role in the development of lung cancer. Those who did so would, however, greatly underestimate the health risks associated with higher levels of smoking. The risk of lung cancer varies with amount smoked, and for individuals who smoke two or more packs a day, the risk exceeds 20 times that of nonsmokers (Blot & Fraumeni. 1996).

When interpreting the results of both our own and Rind et al.'s (1998) various analyses, several important caveats are in order.

First, it is likely that those most severely affected by CSA will be underrepresented in college samples. Therefore, it is not appropriate to generalize from studies of college students to sexual abuse victims in other populations, especially those encountered in the legal or clinical settings.

Second, a prominent epidemiologist, Sander Greenland (1994), has suggested that the worst abuse of meta-analysis stems from the temptation to produce a single estimate of effect from disparate study results and then to treat this estimate as a definitive literature synthesis. We reiterate this concern.

Finally, a single estimate of adjustment obtained by averaging data on a wide range of exposure levels to CSA cannot be used to draw conclusions about the risk of problems and distress in an individual victim.

Accordingly, attempts to use Rind et al.'s study to argue that an individual has not been harmed by sexual abuse constitute a serious misapplication of its findings.