Battle and non-battle injury and posttraumatic stress disorder in military personnel.

I read with interest the recent article by Macgregor et al. [1]. The authors conducted risk assessment of posttraumatic stress disorder (PTSD) in patients of military personnel with battle injury (BI). They also conducted sub-analysis on predictive factors of PTSD in patients with BI. As a result, odds ratio (OR) and 95% confidence interval (CI) of patients with BI against patients with non BI for PTSD were 2.10 and 1.60–2.75, respectively. Furthermore, ORs (95% CIs) of BI patients with moderate injury, with serious-severe injury, with previous mental health diagnosis within 1 year of deployment, and with prior BI for PTSD were 1.49 (1.12–2.00), 1.64 (1.01–2.68), 2.69 (1.50–4.81), and 1.96 (1.22–3.16), respectively. I have two concerns on their study. First, the authors used many independent variables for the adjustment of OR. The number of independent variables to determine predictors of PTSD in patients with BI was 15. Among them, 5 independent variables were composed of 3 items of reply. This means that 20 independent variables were used for logistic regression analysis. The number of events per independent variable in multivariate regression analysis should be 10–20 to keep statistical power [2], and the authors should prepare 200–400 events for stable estimates. As the number of PTSD in patients with BI was 447, their study satisfied the criteria. More events would make stable estimate for predicting PTSD. Second, caution should be paid for their procedure of PTSD judgement. The authors used questionnaire for the screening of PTSD with 4-item instrument, named PC-PTSD [3]. A validation study was also quoted as a reference [4]. As they used 3/4 as a cutoff point of positive screen for PTSD, and I recommend quoting other references [5,6], because Veteran Administration changed the cut-off point of PC-PTSD from 2/3 to 3/4 in 2005. In addition, PC-PTSD is not an instrument for clinical diagnosis of PTSD, which was conducted by Diagnostic and Statistical Manual of Mental Disorders (DSM-IV) PTSD symptoms [7]. Compared with PC-PTSD, PTSD checklist (PCL) was constructed as self-report measures of PTSD with keeping validity [8], and there is a risk of overestimation of PTSD prevalence by questionnaire survey.