[Page 30]

The major findings of the present study may be summarized as follows:

Considerable confidence can be placed in these findings, because they are based on a large number of participants (i.e., 25,367), relatively narrow confidence intervals, and a large file drawer requirement. Because CSA is related to serious mental health problems for a significant number of victims (in the millions in the United States and Canada; Violato & Genuis, 1994), the need to study the phenomenon of CSA and develop prevention and treatment programs to counteract its effects is pressing.

Based on ANOVAs of the mediating variables, there did not appear to be differences in the risk of developing negative outcomes among CSA victims depending on their gender or socioeconomic level. Furthermore, statistical differences were not observed depending on

These results do not corroborate the findings of other studies that have found an increased risk for the development of negative outcomes depending on the type of sexual abuse experienced, the child's age, repetition of abuse, and familiarity with the perpetrator.

For instance, others have found that the trauma associated with CSA is related to variables such as the severity of the abuse, chronicity of abuse, recency of abuse experience, relationship of perpetrator, and use of

[Page 31]

violence or neglect in abuse (Collings, 1995; Conte & Schuennan, 1987; Higgins & McCabe, 1994; Trickett et al.. 1994). It may be that the experience of CSA itself stands out as a negatively significant event, which does not discriminate across individuals but rather affects human development in a consistent manner.

Alternatively, it may be that current data are too crude and imprecise to allow for detection of differences. It seems theoretically evident, for instance, that chronicity of abuse or violence should be related to at least some outcomes. In any case it appears that CSA researchers need to resolve the methodological flaws pervasive in the existing research base. To advance knowledge and focus further research, we recommend a more rigorous approach to investigations, whereby contrast groups are included and specific mediator influences are examined.

The statistically non-significant results that emerged from the uni-variate analyses conducted on the effect sizes of the dependent measures and various mediating variables are also very informative. Similar to Jumper's (1995) meta- analyses, the results of the current meta-analysis did not confirm a differential effect by gender on effect size. In other words, there was no statistical difference in negative outcomes between male and female victims of CSA. These results suggest that men and women who were sexually abused as children do not differ significantly in terms of PTSD outcome, depression, suicide, sexual promiscuity, sexual perpetration, and academic achievement. Moreover, the results indicate that SES does not mediate the relationship either. Therefore, regardless of gender or SES, individuals who were sexually victimized as children may not accommodate abuse, but rather may display negative symptoms over time.

Given that type of abuse (e.g., fondling, penetration) was not found to statistically mediate the relationship between CSA and negative outcomes, one could argue that sexual abuse consists of both physical molestation and emotional violation. Again. the failure to find significant differences may be a reflection of poor quality of the data rather than a real phenomenon. In either case, treatment interventions focusing on the physical dimensions of the abusive experience also need to focus on the resolution of the abuse at a more profound emotional and psychological level. We need to continue to develop and refine strategies that may help CSA victims resolve the emotional, as well as the physical, impact of early abuse experiences.

The results for effect sizes on each of the dependent measures studied demonstrate that individuals who were sexually victimized in their childhood appear to be at an increased risk for developing a variety of symptoms. According to the BESD results, CSA substantially increases the risk for fPTSD, depression, suicidality, sexual promiscuity, sexual perpetration, and poor academic performance. These are large effects that should be of great concern to professionals and to the general public. Such an effect size and increased risk should be considered extremely serious.

Among the medical community, for example, results from a study examining the influence of aspirin on the mortality rate of American physicians were

[Page 32]

discontinued immediately when attention was directed to the practical significance of the study's findings (Rosenthal, 1991). Although a statistically non-significant r of .34 was found, the impact of the correlation suggested that 34 of every 1,000 lives were being saved when aspirin was taken. Based on the study's findings, the study was immediately terminated, with the recommendation that all people at risk for coronary and heart disease should take aspirin. The results from the aspirin study were much more modest than our current findings but, nevertheless, resulted in immediate action.

While the overall effect sizes can be considered robust, there are limitations of the present meta-analysis. One limitation is that the "fugitive literature" was not retrieved (Rosenthal, 1995). Our search was limited to major journals; few studies came from books, conference proceedings, or technical reports. Thus, inherent in the meta-analysis may be the retrieval bias of including studies selectively published because of statistically significant results. Conversely, of course, only studies that had passed the rigor of peer review were included.

A second limitation is that PTSD and depression symptomatologies overlap, thereby making it difficult at times to identify a particular study in one versus the other outcome category. We expected that the effect sizes would change (perhaps demonstrate a stronger effect between CSA and PTSD) if the PTSD and depression studies were collapsed, given that depression is a main symptom of the PTSD diagnosis. Nonetheless, given the existence of numerous studies focusing  on CSA and depression as an outcome, we considered it prudent to conduct a separate and independent meta-analysis examining that specific relationship.

A third limitation of the meta-analysis is also a limitation of all the existing research in CSA: The present meta-analysis was affected by the large gaps and amount of missing data on many of the mediating variables. In many of the studies coded, specific demographic information and proportions, for example, were not provided on variables such as race or SES. Moreover, the age of abuse onset was most frequently reported as "under 17 years of age", thereby rendering it in an unspecific form not amenable to statistical analyses. It is crucial that effects of CSA be examined in relation to identified chronological and developmental age periods. Thus, more original research is needed to determine the effect of potentially important mediating variables such as age when abused, duration of abuse, race, and SES, as well as more detailed and comprehensive reporting of the results.

Another potential limitation of the present meta-analysis involved the assumption that the presence of disorder symptoms suggested the possible presence of a disorder (e.g., PTSD). Of course, the presence of symptoms is not the same as the presence of a disorder, and prevalence rates are not the same between or within outcomes across the life cycle. Furthermore, because many studies of CSA use clinically referred samples, it is possible that the pathologies were associated with additional or different types of abuse. As argued by Kendall- Tackett et al. (1993), most clinical comparison groups probably do contain children who

[Page 33]

have experienced abuse and have shown numerous other forms of symptomatic behavior. Consequently, it is recommended that investigators use appropriate contrast groups (e.g., the general population) in future studies of CSA.

The results of the present study do not support a specific sexual abuse syndrome involving clear PTSD outcomes with sexualized behavior and affect. Rather, the results support the multifaceted model of traumatization in accordance with Kendall-Tackett et al. (1993) and Finkelhor and Browne (1985). In this view, it appears that CSA produces multifaceted effects and that distinct mechanisms and processes may operate to account for the variety of outcomes. Moreover, as Conte and Schuerman (1987) indicated, CSA may be embedded in dysfunctional families and a generally maltreating environment. We concur with Green (1993), therefore, in interpreting CSA as a profoundly traumatizing event or events rather than as a specific syndrome or disorder. These traumatizing events probably affect outcomes in a variety of ways (e.g., arrested developmental processes) that require further clarification.

Irrespective of the specific mechanisms, the present meta-analysis provides compelling evidence of the negative impact of CSA on human development. The results are clear; CSA is associated with the development of PTSD and depression, as well as with suicide, sexual promiscuity, the victim-perpetrator cycle, and poor academic performance, regardless of victim age, gender, or socioeconomic status. This important social, political, and mental health problem requires urgent action.