Author's reply

The multivariate analysis of variance demonstrated no significant difference between the group with DSM-IV PTSD and the group without DSM-IV PTSD in terms of the severity of disability. This finding is unrelated to the other mental health problems measured in the study, as shown by the analysis of covariance. The power of the study was 0.85 (assuming a detectable difference of 3 out of 30 on the Sheehan Disability Scale and α=0.01). This is acceptable for limiting the chances of type II error. Multicollinearity is only of importance when trying to draw inferences about the relative contribution of more than one predictor variable to the success of the model. In this study the Beck Depression Inventory (BDI) (or its variant the MBDI) was the only variable retained in the regression models and so multicollinearity is not an issue. Disability is not an absolute requirement in DSM-IV. The utility of objective measures of disability v. subjective measures was discussed in the paper. However, the subjective experience of the patient is probably of most value in clinical terms. Other studies have found contrasting results, as discussed in the paper's introduction. However, most have methodological limitations. The treatment of PTSD, as opposed to depression, may be relevant to the DSM-IV diagnostic criteria but may not be relevant to the patient.