Seeking Safety Therapy: Clarification of Results
To the Editor: Our study of Seeking Safety therapy was published in the September issue (
1 ). We would like to correct two omissions, which may help clarify outcomes for Seeking Safety and the comparison condition (enhanced residential services). In the article we addressed outcomes over a one-year time frame. However, we did not present outcomes at six months, which was the end of the Seeking Safety intervention. Second, we failed to identify the full extent and asymmetry of the data loss that characterized the study after the end of treatment.
At six months, two significant differences favored Seeking Safety over the comparison condition after Bonferroni correction: avoidant behavior and social support. Two additional variables that favored Seeking Safety were not significant after Bonferroni correction: the PTSD Checklist and days worked. On eight variables, both Seeking Safety and the comparison condition showed significant improvements over baseline, with no difference between conditions: days of drug use, days of alcohol use, the 30-item Symptom Checklist Revised, self-esteem, the psychiatric composite of the Addiction Severity Index (ASI), the medical measure on the 12-Item Short-Form Survey (SF-12), hypervigilant behavior, and days homeless. Two additional variables (ASI drug and alcohol subscales) had a similar pattern; although the difference from baseline was a non-significant trend for the Seeking Safety condition, this likely reflects the greater statistical power in the much larger comparison condition. Finally, only two variables, intrusive thoughts and the medical measure of the SF-12, did not show significant improvements from baseline for either condition.
Sample sizes decreased substantially over time. At three months, about 80% of participants in both conditions completed the assessment, and at six months the rate for both was about 63%. However, at nine months only 40% of Seeking Safety participants and 56% of comparison participants were available, and at 12 months the proportions dropped to 27% and 53%, respectively. It is thus difficult to draw conclusions about the later time points, and we suggest caution in interpreting the one-year outcomes reported in the paper.
In that spirit, we also note that an interaction analysis of condition-by-time showed that participants in both conditions improved on number of days of drug use during the first six months, but that during the second six months participants in the Seeking Safety intervention experienced an increase. The latter may reflect, however, selective dropout from the study. Thus the statement in the abstract that "the Seeking Safety cohort was significantly more likely to have used drugs within the past 30 days" reflected only a small sample during the follow-up period after Seeking Safety had ended, and this finding should be interpreted with caution.
We believe that these additions clarify and elaborate on the results of this project. In sum, at the end of treatment, participants in both Seeking Safety and the comparison condition evidenced consistent and positive outcomes on substance use and related areas. On two of 12 outcomes, differences between conditions favored the Seeking Safety intervention. Later time points (nine and 12 months) were a follow-up period for Seeking Safety, and sample attrition was substantial (the majority of the Seeking Safety sample was not assessed).