Recently, a team of US scientists resurrected a virus that has since been labelled ‘perhaps the most effective bioweapons agent now known’ (von Bubnoff, 2005). In 1918, a highly virulent strain of influenza virus killed up to 50 million people worldwide. The virus – later dubbed the Spanish Flu – killed more people than any other disease of similar duration in the history of humankind. Until last year, this virus was extinct, preserved only as small DNA fragments in victims buried in Alaskan permafrost, or in tissue specimen of the United States Armed Forces Pathology Institute. Now the full sequence of the Spanish Flu virus has been published (Taubenberger et al., 2005) and the virus itself reconstructed. It proved to be as fatal as the original. When tested on mice, it killed the animals more quickly than any other flu virus ever tested (Tumpey et al., 2005).

What sounds like a brilliant piece of science is at the same time a recipe for disaster. Given the availability of the virus' full-genome sequence and the detailed method for its reconstruction on the Internet, its production by rogue scientists is now a real possibility. The Spanish Flu case underlines the need for biosecurity regulations to prevent the proliferation of particularly dangerous knowledge. It is not only a matter of publishing sensitive dual-use information; there are good arguments to stop experiments that are likely to produce sensitive knowledge from the very beginning.

Yes, it is proposed here to restrict science. The outcry is predictable, and yet misplaced. There is no such thing as freedom of science. We as scientists have for decades applied and accepted clear restrictions on the scientific process, and in many – albeit not all – cases for very good reasons. Human experimentation is but one example. The real question is not whether anybody would dare to prevent the course of science, but whether there is a good reason to do so and whether a restriction would have any effect at all – the standard argument being ‘if we don't do it, somebody else would do it’.

Theory of science might suggest that any knowledge that is achievable by humankind will be retrieved by humankind, but the practice of science is a different issue. The story of the resurrection of the Spanish Flu is a unique scientific crime story, resembling more a Michael Crichton plot than a NIH-project. The first attempt, in the early 1950s, to unbury 1918 flu victims from permafrost and revive the virus from lung tissue samples, failed. Then, in the early 1990s, US scientist Jefferey Taubenberger mastered the technology to retrieve and sequence nucleic acid fragments from preserved tissues and decided to try the new method on the Spanish Flu. He had a resource for his research that is probably unique in the world and very likely not accessible to any rogue scientist: the tissue archives of the US Armed Forces Pathology Institute. After years of trial and error, his team finally succeeded. In one tissue sample of a soldier who died in 1918 of the Spanish Flu, they identified flu gene fragments and were able to partly sequence them.

But the RNA was too fragmented to get the full sequence. Then Taubenberger was lucky enough to get access to a second source. A colleague unburied one flu victim from an Alaskan permafrost cemetery and got a tissue sample that did indeed contain additional RNA fragments of the Spanish Flu virus. Again, this resource is not easily accessible, as demonstrated by another expedition to Norwegian flu victims buried in permafrost, which failed in 1998 (Kolata, 1999). Knowing all this, it appears to be rather unlikely that another group of scientists – however brilliant and skilful – without these unique resources would have been in a position to sequence and resurrect the Spanish Flu virus. If Taubenberger et al. had not done it, nobody else could have done it; at least not for quite a while.

The key question, whether there is a good reason to prevent an experiment like the Spanish Flu resurrection, can only be answered by an in-depth risk-benefit analysis. The risk side of the equation is rather straightforward: a highly virulent and contagious influenza virus is an ideal weapon; state-run biological warfare programmes with Internet-access and some expertise in virology and molecular biology would be able to repeat the work; and there are currently no arms-control measures in place to manage the risk of misuse.

Much more difficult is the assessment of potential benefits. It has been claimed, for example, that this work will increase our understanding of virulence and pathogenic factors in influenza viruses and that it might contribute to identifying the next pandemic strain or to developing appropriate drugs (von Bubnoff, 2005). These rather general statements are probably true for the vast majority of biomedical research projects; they contribute in some way to our understanding of a particular disease. But such generalities are of little help for a systematic risk-benefit analysis, in which concrete risks are weighed against concrete benefits. A more scientific approach to benefit assessment would include, for example, the following questions: does this research address an important health or humanitarian problem? Are there alternative ways to achieve the same scientific or humanitarian goals? How big the added value is, that is, to what extent does a particular experimental approach increase the knowledge and the likelihood of reaching a specific goal?

If applied to the Spanish flu work, a rather mixed picture emerges. Influenza pandemics are, without doubt, a very important public health problem. But it is less obvious that a reconstructed virus from 1918 is crucial for increasing our understanding of the genesis, prevention and management of influenza pandemics. Hundreds of influenza strains from the past five decades, including some pandemic strains, are available to researchers and are indeed used by initiatives such as the Influenza Genome Sequencing Project to investigate genetic virulence factors. The added value of one additional strain, even one with an exceptional high mortality rate, is limited, given that strains with varying degree of contagiousness and pathogenicity are already available and provide a wealth of research resources for comparative studies.

Thus, the tangible societal benefits of sequencing and reconstructing the 1918 pandemic influenza virus remain poorly defined. Considering the high risk of abuse, the availability of alternative research avenues and its limited added value to public health, this particular research project appears to be one of the few cases in which the risks outweigh the benefits and which should not have proceeded over the course of 10 years.