USAT quote
Washington Post quote
nav bar

Attributed Risk, Veterans, and Lou Gehrig's Disease
February 01, 2002
Manipulating data to "prove" the existence of a syndrome

It must be frustrating for a Gulf War veteran to read the morning newspaper. Though definitive studies have failed to establish a link between military service and adverse health outcomes, headlines persist in suggesting otherwise.

No one doubts that Gulf War veterans suffer a variety of very real maladies. But the heart of the policy debate lies elsewhere. Do they suffer afflictions at higher rates than the general population? Do they suffer afflictions at higher rates than comparably matched veterans who didn't deploy to the Gulf? Were the afflictions caused by deployment, rather than being the standard ills of mortality? To date, studies have failed to affirm these questions.

For instance, in 1996 two studies in the New England Journal of Medicine found neither increased deaths nor hospitalizations among Gulf vets when they were compared to other vets who did not deploy. A National Academy of Sciences review in 2000 likewise was unable to find an association between complaints and deployment.

All that apparently changed on December 10 with the announcement of a new study undertaken by the Departments of Defense and Veterans Affairs. According to ABC News, "The government says that Gulf War veterans are twice as likely to develop Lou Gehrig's disease as other servicemen. There are only 40 cases of the disease among 700,000 veterans of that war, but it's a very high rate."

Lou Gehrig's disease is formally known as amyotrophic lateral sclerosis (ALS), a peripheral nerve disorder of unknown causation that kills approximately 50 percent of its victims within three years of the first symptoms. About 30,000 Americans currently suffer the disorder, with approximately 5,000 new cases being diagnosed each year.

This study of 2.5 million veterans (700,000 who deployed and 1.8 million who didn't) found 40 cases among the former group, against 67 cases among the latter. Because of the study (which was termed "preliminary" and has yet to undergo peer reviewed publication), the US will offer disability compensation for veterans who suffer ALS.

According to The New York Times, "The study found that among the nearly 700,000 soldiers sent to the gulf (sic) from August 1990 to July 1991, the risk of contracting ALS was 6.7 per million. Among the 1.8 million who were not deployed, the risk was 3.5 per million." The rate figures represent the number of new cases per year over a ten year period, hence the respective 40 and 67 case totals.

But are vets suffering ALS at rates that are higher than expected (after statistical adjustments have been made for their age and characteristics)? According to the study, the 700,000 deployed vets should have experienced only 33 cases, and not 40. Hence, the number of cases is elevated over expectations by 21 percent. Therefore, the study tried to determine the effect of deployment itself on the elevated risk.

This is known in epidemiology as "attributed risk," or the proportion of excess cases of those exposed to some factor when compared to those unexposed, the difference being attributed to the factor itself. It turns out that the attributed risk for deployment is 18 percent.

But how compelling is this attributed risk figure? According to the way attributed risk is calculated, the outcome is also a function of the number of cases in the unexposed group. This follows because attributed risk is commonly determined by subtracting the cases in the unexposed group from the cases in the exposed group. Thus, an attributed risk differential could also appear if the number of cases in the unexposed group is unusually low compared to their expected number.

As it turns out, the number of cases in the unexposed group (vets who didn't deploy) is unusually low. The non-deployed vets experienced only 67 cases, when their population of 1.8 million would have been expected to experience 85 (again, the difference is 21 percent, only in this case, fewer cases). In epidemiologic terms, these vets could be described as having received an "attributed benefit" from some (unknown) factor.

If we combine their respective "expected" cases, they equal 118. But the two actual populations combined experienced only 107 cases of ALS. That is, the veterans taken together are still on the "benefit" side of the calculation. It is only by virtue of the study's division of that population into the 700,000 who deployed vs. 1.8 million who didn't that the former stand out as at greater risk. This could be good epidemiology, in that it investigated whether some factor affected some deployed vets, and succeeded in finding such evidence.

Alternatively, the figure could represent an artifact of the researchers' decision to divide and compare these populations in just this way. The danger is that chance variation (statistical "noise") continues to plague the outcomes. This is especially true when the overall number of those afflicted is not large.

Moreover, of the 57,000 British veterans who deployed, no differential ALS has been found. Even among US veterans, deployed Naval and Marine personnel had the same rates as those who didn't deploy. The difference only affected the Air Force and Army personnel. Hence, the concern over the "preliminary" status of the results.

America today takes new-found pride in its men and women in the military. That Gulf War vets who suffer will now be compensated is perhaps the least we can do for those who served. Still, questions remain. What are we to say to the 67 veterans who didn't deploy but are nevertheless sick? What is the presumed "cause" of their disease? These findings appear far from definitive. The study revealed two anomalies, finding both higher and lower rates than expected. Not surprisingly, only one anomaly has led to causal attribution and compensation. No one has proposed an investigation into the "cause" of 21 percent fewer cases among the non-deployed vets.

Whatever the policy disposition, fairness must acknowledge that there is another risk for Gulf War veterans -- being routinely alarmed and confused by incomplete reports of their presumed fate.

This article originally appeared in the Providence Journal on December 28, 2001.

about staff publications email contact search
email this page to a friend