To the Editor

To the Editor : Breslau et al. (32, 573–576) were critical of our editorial (Neria&Bromet, 2000) on the strengths and limitations of their analysis of the relationship between trauma exposure, post-traumatic stress disorder (PTSD) and depression (Breslau et al. 2000). Their research, as they point out, assumes ‘the validity of the DSM definition of PTSD’. They argue that our comments were misguided because we do not appreciate the difference between PTSD and trauma-related depression. Our concern is not about the definition of PTSD but rather about whether their secondary analysis was able to address directly the occurrence of ‘major depression (or another disorder) associated with traumatic events ’. We were especially concerned because the data came from the DSM-III-R version of the Diagnostic Interview Schedule (DIS) and the Composite International Diagnostic Interview (CIDI) used in the National Comorbidity Survey (NCS); in neither case were respondents asked about ‘major depression associated with traumatic events ’. Moreover, in the NCS analysis, 25% of the exposed sample was excluded because age of onset was unavailable, including those with childhood physical abuse and childhood neglect. Kessler himself has shown that such exposures are associated with subsequent anxiety, mood and addictive disorders (Kessler et al. 1997) as well as PTSD (Kessler et al. 1995). In general, secondary analyses are often limited because the data are rarely optimal for addressing the new questions being asked. Epidemiological research has several functions (Morris, 1964). The most basic descriptive function is to estimate the rate of clinical conditions in the general population and the risk factors associated with their occurrence. In this respect, studies like the National Comorbidity Survey (NCS) and the ongoing international World Mental Health (WMH) project in which we are participating produce valuable basic data, and longitudinal spin-offs, like the Baltimore ECA follow-up, provide a platform for examining relationships longitudinally. However, these studies were not designed to address a different and equally important function of epidemiology, and that is ‘completing the clinical picture ’ (Morris, 1964). As we read the Breslau et al. (2000) analysis, this was indeed the underlying intent, and their paper is in fact subtitled ‘the post-traumatic stress disorder-major depression connection’. Unfortunately, as noted above, respondents were not asked to report on post-trauma symptoms other than those contained in the diagnosis of PTSD. The situation was further biased by the order in which the modules were administered (Stone et al. 2000). In the NCS, the CIDI depression module was administered in ‘Part 1’ of the interview, while PTSD was covered much later in ‘Part 2’. One could hypothesize that if trauma exposure had been explored before the depression screen was administered, the actual rate of post-trauma depression might have been different in exposed individuals. Again, our point is simply that statistical analysis cannot substitute for lack of direct data. The relationship between trauma, PTSD and the development of depression and other disorders is a terribly important issue with huge treatment implications, and it deserves to be studied directly in both general population and trauma-exposed samples. When Breslau et al. suggested that we were ‘speculating beyond the DSM classification’, they were partly correct. While sociologists and epidemiologists have successfully translated existing criteria into structured interview schedules, the field of trauma research is badly in need of investigators who can ‘think outside the box’ and develop tools that will enable us to complete the clinical picture that arises and evolves over time after a traumatic event. Indeed, in addition to depression (Shalev et al. 1998), a number of other symptoms and disorders have been described in trauma survivors