This post comments generally on several points raised in the recent discussion of various treatments and comparative outcome research with trauma victims:
From my perspective, one difficulty that seems specific to trauma research is that traumatic events frequently "stir-up" unresolved, affectively similar (but often logically unrelated) incidents from an individual's past. These apparently associated memories, some of them "forgotten" before being stirred-up by the recent trauma, can add to the trauma response (e.g., arousal, flashbacks) or increase motivation to avoid the whole, much larger, issue -- obviously complicating effective treatment. This is more my clinical observation than an established fact, BTW.
Clearly, if some subjects among the comparative treatment groups bring one or several such additional but unrecognized issues into a study focused on a specific ("presenting") trauma, this can greatly increase the within subject error variance -- more, it seems to me, in trauma work than with other typical clinical populations. Moreover, this appears to be a more serious problem in studies with smaller Ns, and perhaps also in more rigorous studies where tightly standardized treatments are focused only on the "presenting" trauma (and consequently more apt to ignore any prior but related traumatic incidents of some subjects).
To return to the previously raised issue of long-term fluctuations in adaptive functioning well after the trauma has been treated, I suspect that one component of this fluctuation would be other (especially untreated) traumatic experiences in the individual's history (before the study began), and another component would be similar experiences as may naturally occur over the longitudinal outcome periods (e.g., crime, etc.). It seems very difficult for human studies to either prevent or control for these potential sources of variance in later adaptive functioning.
One related point: There is evidence that chronic emotional trauma has the potential to cause permanent physical damage in at least the hippocampus [see Sapolsky, in Why Zebras Don't Get Ulcers (1994), who argues that chronic stress is a significant cause of aging in several species]. If there are sub-types within PTSD, I suspect they may relate to variations in the capacity to recover from stressful events, and a history of chronic stress may diminish this capacity cumulatively (this seems consistent, for example, with the proposed DESNOS diagnosis). So aside from the clinical difficulties of possible emotional differences in trauma victims, there may be coincident differences in subjects' physical ability to perceive or recover from stress. I'm unaware of reasonable measures for such variation that could be used in treatment outcome studies -- as far as I know, this is still unmeasurable variance.
Finally, LeDoux and others (LeDoux in June 1994 Scientific American; see also Jacobs & Nadel, 1985, Psychological Review, 92, 512-531) would argue that exposure to a feared stressor can produce permanent changes in how the brain responds to ambiguous but potentially dangerous similar stimuli. If this is true, then perhaps trauma interventions really ought to concentrate on the subsequent (more cognitive and less limbic) realization that an ambiguous stimuli is not so fearful after all (e.g., "its just an aftershock") -- thus, the aim would be to cut short a fight/flight response, rather than to prevent it (managing rather than curing trauma). This may remain an open empirical question for some time.
But, if this view is right, one implication for outcome research is that immediate psychophysiological measures might not be sensitive to effective trauma treatments if they are just measuring the permanently altered initial limbic response. In that case, we might be better off looking for psychophysical measures of the _duration_ of the startle response or its resistance to habituation, rather than initial startle magnitude.
I'll try to end with a less pessimistic point: I remember reading a paper by John Reid (1990; Journal of Applied Social Psychology, 20, 1695-1703) suggesting that trauma response really might best be studied using normal subjects already in longitudinal studies before the "presenting" trauma, since we would have detailed information about life events and functioning for these subjects prior to the traumatic event. I think that is a great idea.
Comments are most welcome.