Policy-making is a difficult practice. It lies halfway between art and science, and science can mislead as much as conjecture. Often we are forced to make policy based on an uncertain mixture and evaluation may later determine that we were wrong.

A cautionary tale is provided by the paper in this issue of Addiction by Schnohr et al. [1], who obtained rather inconclusive results from an analysis that suffered from poorly defined inputs. They found an association between school smoking bans and adolescent smoking prevalence; however, the environment of the study must surely be clouded, as the degree to which the bans were implemented in the 27 countries studied is unknown and is certainly very variable. On the other hand, they found no association with price—a well-known disincentive to smoke as shown by the references the authors quote—and a reverse association with laws governing age at which tobacco sales are allowed. This latter finding is somewhat unsurprising, as such laws go back far into history (often before tobacco was proven to be harmful) and implementation is very variable. Such implementation needs a work-force and a budget which is rarely available. Thus good statistical outputs require precise inputs, which are usually something of a luxury.

We are then left with a dilemma as to what evidence we need to make effective legislative policy. During the 1970s the Union Internationale Contre le Cancer (UICC) espoused the idea of Comprehensive Tobacco Policy [2] which focused upon implementing everything that was intuitively expected to work, as there was virtually no evidence base on which to proceed. Major areas which attracted attention included: eradication of tobacco promotion of all types; health warnings on packets; taxation manipulation; restrictions on smoking opportunities; encouragement of non-smokers' rights; education programmes (including cessation programmes); and for smokers to reduce tar exposure (set out in text as ‘harmful components of tobacco smoke’).

We had to wait several decades for proof that some of these policies worked and some were indeed unsuccessful. The efficacy of smoke-free public places is now known, as a result of graduated research over time [3], and eventually national experiments as seen in Ireland are shown to have dramatic effects [4]. Similarly, the concept of reducing tar exposure was eventually shown in 2001 to be wrong [5]—not because of the initial logic being incorrect but because of tobacco industry manipulation of cigarette design and advertising aimed at misleading the public into the view that low-tar cigarettes were indeed lower-exposure products. In this case public health advocates were simply outsmarted by the tobacco industry.

There still remain difficulties in obtaining suitable legislative policy even in advanced countries, and two examples illustrate the distance we have to travel:

  • • Graphic health warnings on packets have been adopted widely when research showed that they were efficacious after their introduction in Canada [6], but the next logical step is the introduction of plain packaging devoid of trade marks. It would be easier to fight for these if research existed that proved they would work; however, what we currently know about trade marks suggests strongly that plain packaging would be more effective [7].
  • • Despite severe restrictions on marketing practice, no country has legislated to restrict the composition of cigarette smoke by setting upper limits on certain major toxicants, as has been proposed by the World Health Organization (WHO) [8]. This applies to motor exhausts but not to cigarette smoke. Thus, continuing smokers are ingesting unnecessarily high amounts of carcinogenic and toxic material. Is this because we lack proof that such action would reduce risk? The precautionary principle suggests that toxicants should be reduced wherever they are reducible; for example, no one proposes that water in developed countries should have unnecessary levels of arsenic simply because it would be difficult to actually prove its removal was beneficial. Proof that reducing toxicants in cigarette smoke could come only after decades of research and we need to proceed on basic principles.

Much of our difficulty in making effective legislative policy lies in the complex way we legislate, with each small policy step requiring specific legislation in a parliament or congress. There is a lesson in history—the first significant anti-tobacco legislation ever passed was the Norwegian Tobacco Act (1975) which, among other specific prohibitions, granted ‘Permission to the Ministry of Health to issue provisions concerning the content, weight, filters, packaging, etc., of tobacco products’. [2]. This was a short, simple piece of legislation that gave power to health officials to make decisions about tobacco.

Perhaps the next legislative target is a simplified form of regulation that emulates the Norwegian principle so that we could proceed to control tobacco on the basis of common sense.