David Dobbs, our regular Mind Matters columnist, is taking advantage of the balmy weather to get some much-needed R&R. So we dug up one of the very first Mind Matters columns on account of it being newly topical...
Welcome to the second installment of
This week's paper is
by Bruce Dohrenwend, J. Blake Turner,
Mind Matters is Sciam.com's "seminar blog" on the sciences of mind and brain. Each week, top researchers describe their disciplines' most significant new findings -- and what they, as fellow researchers, find most exciting, maddening, significant, odd, or otherwise noteworthy in the research driving their fields. Blog visitors can participate. We hope you'll join us.
Nicholas A. Turse, Ben G. Adams,
Karestan C. Koenen, and Randall Marshall
From Science, 18 August 2006.
by David Dobbs, Editor, Mind Matters
"The Psychological Risks of Vietnam for U.S. Veterans: A Revisit with New Data and Methods" addresses what seems a simple question: What proportion of U.S. veterans of the Vietnam War developed post-traumatic stress syndrome, or PTSD, in reaction to their service there? That answering this question proves difficult shouldn't surprise, for the definition of PTSD has a history almost as controversial
as that of the Vietnam War. Indeed, PTSD was recognized as a legitimate disorder only in the 1980s, and only because Vietnam veterans forced the issue
. Research since then has shown that PTSD rises from distinct changes
in neuroanatomy and endocrinology.
Yet if PTSD's clinical basis is solid, its prevalence among veterans remains controversial. This is partly because the stakes are high: As both essays note below, we can't properly treat PTSD in veterans, whether of past, present, or future wars, if we don't know its prevalence. In addition, assessing our troops' trauma inevitably feeds the ongoing debate over war's cost-benefit ratio. This debate is difficult at any time -- and torturous indeed when a war's benefits prove elusive. This happened in Vietnam and appears to be happening now in Iraq. Small wonder that this simple question -- How much trauma have we inflicted on our veterans? -- can prove excruciatingly painful to answer.
Below, Harvard psychologist Richard J. McNally, psychiatrist Charles Marmar, of the University of California, San Francisco, and psychologist William Schlenger, of Abt Associates, reflect on Dohrenwend's attempt. Almost 20 years ago, all three of these researchers helped collect the data for the study that Dohrenwend reevaluates in his Science
paper -- Marmar and Schlenger as a principal co-investigators, McNally as a field researcher interviewing veterans. As you'll see, they bring no end of perspective and interest to this tangled but vital topic.
Recalculating War's Psychic Cost
by Richard J. McNally
Though the Vietnam War ended more than three decades ago, the battle over its psychiatric cost continues. For researchers and clinicians, this debate centers on the prevalence of post-traumatic stress disorder (PTSD) among Americans who served in that controversial conflict. It's not an academic question. We must accurately gauge how many Vietnam veterans suffer from PTSD so we can provide adequate treatment for them -- and for psychiatric casualties from our wars in Iraq and Afghanistan.
Last August Columbia University epidemiologist Bruce Dohrenwend and colleagues stirred this debate anew when they published in the journal Science a reanalysis of data collected in the late 1980s by the National Vietnam Veterans Readjustment Study (NVVRS)
. Outraging some and pleasing others, they concluded that actual rates of PTSD in Vietnam veterans were roughly 40 percent lower than what the NVVRS had concluded.
A Revision Revisited
In one sense Dohrenwend's report was a revision of a revision. The NVVRS, published in 1990, had superseded earlier estimates, done in 1988 by the U.S. Centers for Disease Control, that many critics had thought far too low. The 1988 Centers for Disease Control study had concluded that about 15 percent of Vietnam veterans had developed PTSD at some point in their lives (what epidemiologists call the lifetime prevalence) and that just 2.2 percent still suffered the disorder when interviewed in the late 1980s (the current prevalence). A chorus of veterans' advocates and clinicians said these numbers were far too low.
The 1990 NVVRS, more rigorously conducted than the earlier study, produced numbers far more troubling. It found lifetime prevalences of 31 percent for full-blown PTSD and 22.5 percent for partial PTSD, bringing the total rate for lifetime sufferers to more than 53 percent. The NVVRS also found a current (that is, late 1980s) prevalence of 15.2 percent. The NVVRS researchers argued convincingly that the earlier study had severe methodological flaws that created an undercount of the PTSD rate. Consequently, the NVVRS figures became the accepted count.
Many historians of military psychiatry found the NVVRS figures puzzling. How, they asked, did more than 53 percent of veterans develop either partial or full-blown PTSD when only about 15 percent had been assigned to combat duty? To be sure, one didn't need to be an infantry rifleman in Vietnam to suffer trauma. Medical and other support personnel often witnessed or experienced horrific events. Yet the data were still perplexing. After all, clerks and truck drivers get exposed to far less danger and violence than do infantrymen, and most people exposed to trauma never develop PTSD anyway. Skeptics wondered whether the NVVRS might have misconstrued normal emotional responses to war-zone duty as symptoms of psychiatric illness. This flaw seemed especially likely because the prevailing diagnostic criteria for PTSD in the 1980s, when the NVVRS was done, did not -- unlike standards developed during the 1990s -- require that symptoms cause functional impairment in those reporting them.
To address these concerns, the Dohrenwend study accepted a case as PTSD-positive only if three criteria were met. First, symptoms had to have arisen from exposure to a trauma during the war, not before or after it. Second, personnel files and other archival data had to corroborate veterans' reports of trauma. Third, veterans had to meet the Dohrenwend study's criteria for impairment on the Global Assessment of Functioning (GAF) scale -- a scale of general functional impairment used by clinical interviewers during the NVVRS survey.
It was these revised criteria that led Dohrenwend to conclude that PTSD actually affected about 40 percent fewer veterans than the NVVRS had estimated.
Fewer, but Still Many
This new estimate clearly represents a huge revision. Yet Dohrenwend didn't even use the most stringent criteria possible. Had they done so, they would have plausibly generated and defended an even greater correction.
Here's why. Dohrenwend called a case PTSD-positive if the veteran scored 7 or below on the 1-to-9-point GAF scale
. (A score of 9 denotes high, "normal" function, whereas 1 denotes extremely low function.) The criteria for a score of 7 were: "Some difficulty in social, occupational, or school functioning, but generally functioning pretty well; has some meaningful interpersonal relationships OR some mild symptoms (e.g., depressed mood and mild insomnia, occasional truancy, or theft within the household)." This description falls well short of clinically significant impairment -- that is, impairment likely to require treatment. Managed-care companies, for example, seldom reimburse for treatment of patients scoring higher than 6 on the GAF, and someone who is "unable to keep a job" gets a score of 5. Dohrenwend and colleagues could have plausibly used the slightly more stringent score of 6 as a diagnostic cutoff. If they had, they would have found just 5.4 percent of Vietnam veterans suffering PTSD in 1988 -- a 65 percent drop from the NVVRS's 15 percent. In other words, the Dohrenwend study criteria produced a current prevalence 40 percent lower than the NVVRS study found, and applying slightly more stringent criteria for clinical impairment would have found a prevalence rate 65 percent lower.
In that mixed light, Dohrenwend's finding of 9 percent current prevalence can seem either high or low. Yet this 9 percent quadrupled the 2.2 percent rate found by the flawed Centers for Disease Control study, and even the stricter, hypothetical 5.4 percent prevalence rate that you'd get using a GAF cutoff of 6 would have more than doubled the Centers' numbers. Perhaps most important is that even at Dohrenwend's 9 percent current-prevalence and 15 percent lifetime-prevalence rates, we're still talking about hundreds of thousands of veterans requiring treatment.
Rage or Relief?
The discovery that a disease affects fewer people than previously thought should produce joy and relief. Yet this hasn't been the case. Among some -- not just veterans' advocates, but many researchers and clinicians devoted to studying and treating trauma victims -- pointing out that the NVVRS overestimated the rate of PTSD is more likely to provoke rage than relief.
Why is that? Cynics might say that researchers and doctors whose funding and prestige hinge on an alarmingly high prevalence of the disorder will resist news that it's not epidemic after all. Yet that would ignore the genuine concern for trauma victims among those who study and treat PTSD. The real fear -- and a real danger -- is that reduced prevalence estimates will encourage a financially strapped government to cut clinical resources for both Vietnam veterans and those returning from Iraq and Afghanistan. Many vets returning from Iraq already complain of poor access to psychiatric care.
Perhaps the most important question Dohrenwend's reanalysis should provoke is whether veterans who have war-related psychiatric illness, whether from Vietnam or Iraq today, are getting prompt access to good treatment. If not, we need to increase our resources -- no matter what the epidemiological studies show.
Richard J. McNally, a professor of psychology at Harvard University, studies information-processing dynamics in anxiety disorders and PTSD. He has served on the PTSD committee of the Ameircan Psychiatric Association's DSM-IV Task Force and the National Institute of Mental Helth's consensus panel for the assessment of PTSD.
by William E. Schlenger and Charles R. Marmar
The Dohrenwend study has generated a lot of reaction that misses the point. Much of the press, following Richard McNally's views expressed in the Perspective
he wrote accompanying the Dohrenwend study in Science
, has focused on the apparent discrepancy between the prevalence rates found by the Dohrenwend study and those found by the NVVRS study in the late 1980s. McNally wrote in Science
that this discrepancy was "the most newsworthy finding" in the study and "confirmed the critics" who considered PTSD prevalence rates overestimated. Many news reports, following his lead, emphasized a sharp correction of apparently inflated numbers.
Yet the Dohrenwend study's lower numbers should have surprised no one who looked clearly at the two studies' methods. Indeed, the Dohrenwend study's treatment of the original NVVRS data differed from the NVVRS's in ways that made higher numbers impossible and lower numbers a certainty. The studies used such different diagnostic and statistical criteria that comparing their results is like comparing apples and oranges or inches and centimeters.
This is unfortunate, for it has obscured the Dohrenwend study's real significance. Dohrenwend is newsworthy not because it corrects prevalence rates but because it confirms that the vast majority of PTSD cases in Vietnam veterans are legitimate. Dohrenwend also confirmed a predictable "dose-response" relationship
for PTSD in these veterans with records-based exposure measures, showing that the risk and potential severity of PTSD corresponded to the amount and intensity of a soldier's exposure to trauma. This emphatically affirms the medical reality of a syndrome that some continue to doubt. The Dohrenwend study should put to rest the most troubling charges levied against both the NVVRS study and Vietnam veterans with PTSD, which is that many vets are exaggerating or faking their symptoms.
Dohrenwend and colleagues, for example, closely examined the case records of veterans who reported high trauma exposure but whose combat and other records classified them as low probable exposure -- the group likely to hold those inflating or faking symptoms. The researchers found "no indication of dissembling and little evidence of exaggeration" (pp. 980-981). They also found virtually no evidence of attempts to inflate disability claims. So much for widespread malingering and fraud.
Apples and Oranges
But what about the numbers? What about the proclaimed 40 percent difference between the NVVRS current (i.e., 1987) prevalence estimate and Dohrenwend's? Both studies sought to estimate PTSD prevalence. But the tighter definitions of PTSD used today, along with several statistical filters used by Dohrenwend but not the NVVRS, meant that the two studies categorized PTSD so differently that they were effectively measuring different phenomena.
- Dohrenwend excluded veterans whose diagnoses could not be confirmed by combat records of trauma exposure. Combat records, however, are notoriously incomplete. And most exposures reported by clinically diagnosed cases that could be checked against records proved positive. Thus excluding all that could not be confirmed surely eliminated some legitimate cases.
- The NVVRS included cases of combat-related PTSD in veterans who had experienced trauma before going to Vietnam, but Dohrenwend excluded them. While exclusion avoids confounding of causes, it ignores an important practical reality: many people experience multiple exposures over their lifetimes. As a result, PTSD's genesis is often cumulative. And PTSD grown from a seed planted in childhood but watered to fruition in war is still PTSD, and it's still war-related.
- Although the Dohrenwend study carefully excluded potential false positives, it made no attempt to unearth false negatives (i.e., undiagnosed true cases). Yet because many PTSD sufferers do not seek help or deny or underreport symptoms, this almost guarantees an undercount.
All of these factors pushed Dohrenwend's prevalance numbers down. No offsetting methodological factors were used to identify errors that would push them up. The Dohrenwend numbers, then, really provide only a conservative lower bound: We can now say that current prevalence of PTSD among Vietnam veterans in 1987 was at least
9.1 percent. Meanwhile, documented false negatives and other findings from other studies suggest that actual prevalence was probably near the original NVVRS estimate of 15 percent.
The Dohrenwend study provides not a refutation or even a "huge revision" of the NVVRS findings, but rather a rigorous and vital confirmation of the picture those findings painted of war's psychological consequences. It refutes the critics' idea that we've overestimated PTSD or coddled veterans who claim to have it, and it documents that, for most part, the suffering is real. It also shows that the charge of the critics--that veterans are exaggerating or dissembling--is false.
Psychologist William Schlenger is a principal scientist at Abt Associates, a private research and consulting firm, where he researches psychiatric and substance abuse epidemiology and evaluation. He has almost 30 years of experience conducting extensive research on post-traumatic stress disorder (PTSD), including a study of the psychological impact of the terrorist attacks of September 11, 2001, on residents of New York City and Washington, D.C.
Charles Marmar, M.D., is a professor of psychiatry at the University of California San Francisco Medical School and chief of mental health services at the San Francisco Veterans Administration Medical Center, where he directs the PTSD treatment program.