Even by government standards, a near 30-year delay in getting a study approved is extreme. But that is essentially what has happened with a proposed large-scale epidemiological study of the possible effects of the defoliant Agent Orange and other combat factors on US veterans of the Vietnam War. This US administration, or the next, would do well to heed the advice issued last month by the National Academy of Sciences that such a study should now proceed.

The academy argues that new data, and advances in geographical information systems (J. M. Stellman et al. Nature 422, 681–687; 2003), could plug a significant gap in earlier epidemiological studies (see page 786). Today, a study should be able to provide key information about when and where troops were exposed to the defoliant, which contained highly toxic compounds called dioxins. And advances in computing and databases mean that the study would cost a fraction of earlier estimates.

The question is whether the US Department of Veterans Affairs (VA) will cooperate. In the 1980s, pilot studies by the VA and several other government agencies found little evidence for a link between Agent Orange and veterans' health problems. They concluded that poor data made doing the large-scale study requested by Congress in 1979 impossible. The VA has been sceptical of the health claims ever since. But critics questioned the independence of those studies. So in 1991, Congress asked the National Academies to assess both the evidence for the health links and the feasibility of full-blown epidemiological studies. It disagreed with the previous analysis, saying that a large-scale study was indeed feasible. But there matters have stood ever since. An independent study cannot happen until the VA provides the funding — which has not been forthcoming.

It is time to stop stalling. It is true that veterans would not be the first choice for a study on the toxicity of herbicides and dioxins — better-documented accidental exposures would be preferable. But that misses the point. The proposed study is both a legitimate research opportunity and a moral imperative. There are privacy issues with medical and other records, but these are not insurmountable. And even if the data and science are not perfect, the new models and techniques can be refined. A lot of data on some 3 million veterans are already sitting in the government's databases — or in some cases are gathering dust, undigitized, on its shelves. Cleaning up and tying those data together would itself be valuable archival housekeeping, and would aid research.

The process would also benefit the VA. More eyes on the data would stimulate creativity, and a greater involvement of outside researchers would bolster credibility and public acceptance of the study's results. This is particularly crucial following the debate over 'Gulf War syndrome', in which government studies lasting more than a decade failed to properly identify or exclude debilitating war-related illnesses in veterans from the first Gulf War. With an increasing number of soldiers from the current war reporting symptoms of post-traumatic stress disorder, it is essential that the VA includes external scientific input as early as possible in the process of helping to treat them.

And there is a bigger picture. US veterans are not the only ones suffering as a result of the Vietnam War. So scientists and the US and Vietnamese governments need to revive stalled bilateral collaborations (see Nature 434, 687; 2005) on the health effects of Agent Orange on the Vietnamese population itself. The responsibilities of warfare do not end on the battlefield, nor can they be limited by the battle lines that once divided friend from foe.