Daily smoking among adolescents is a significant public health problem. Smoking-related health problems are a function of the duration (years of smoking) and the intensity of use (number of cigarettes smoked) [1]. Most adult smokers began to smoke or were already addicted to nicotine before the age of 18 [2, 3]. Besides, a lot of adolescents want to quit smoking, but only a small number of them really succeed [1, 2]. Tobacco control policies varied widely in European countries in the last 20 years [4]. The smoking prevalence among adolescents is important for policy makers to monitor their current policy and to make decisions for future policies. Information on recent smoking trends within a country and comparison of trends between countries is therefore urgently needed. This information is important to have a benchmark, in order for countries to see how large their smoking problem is compared with other countries. These data can also help to explain the observed differences and trends in smoking prevalence, by relating it to potentially relevant circumstances in the different countries. Relating smoking trends to country-specific policies regarding smoking, can help policy-makers to determine which actions to take in order to reduce smoking. Studies gathering this information according to a standardised research protocol are rare.

According to the WHO European report on Tobacco Control Policy [5], gender differences in smoking prevalence among young people in Europe are smaller than those for adults. Similarities and differences in smoking trends among boys and girls need consideration for future developments. Here, we present the results of a large international study concentrating on the evolution of daily smoking prevalence among boys and girls between 1990 and 2002. The study targeted 14 and 15 year olds in 10 European countries and Canada.

The present paper is based on observations made in the Health Behaviour in School-aged Children study (HBSC). This is a four-yearly cross-national research study conducted in collaboration with the WHO Regional Office for Europe [6]. The data of the 4 last surveys are used (1989–1990, 1993–1994, 1997–1998, and 2001–2002). The HBSC-study is carried out in a growing number of countries (from 16 countries in 1989 to 36 countries in 2001). Only countries participating in the 4 survey periods were included in the analyses: Austria, Belgium, Canada, Finland, Hungary, Latvia, Norway, Poland, United Kingdom (Scotland and Wales), Sweden, and Switzerland. The HBSC study aims to gain insight into young people's health and well-being, health behaviours and their social context. The target population of the study is young people 11, 13 and 15 years old attending school. Cluster sampling (school or classes) is used as sampling method in the study. The survey is carried out on a nationally representative sample in each participating country. The sample consists of more than 1200 students in each year, country and age-category. In this paper, 14-year-old and 15-year-old students were selected (n = 75 745), as daily smoking is still rather rare in younger age groups. More details can be found in the international HBSC protocol [6]. The survey is approved by the Ethics Committee of the University Hospital of Ghent, project 2001/304.

Detailed information on non-response in all countries and all survey years is not available. Non-response at school-level varies between countries and survey years and a decreasing trend can be observed. However, non-response at pupil-level (for this study most important) is more constant between countries and survey years and remains high.

The self-administered questionnaire is completed in the classroom and consists of a standard questionnaire developed by the HBSC international research network. Besides questions on smoking and other health-risk behaviours, there were also questions on health outcomes, individual and social resources... The question used in this paper that remained unchanged over the 4 survey periods, is:

'How often do you smoke tobacco at present?' 'Every day'; 'at least once a week, but not every day'; 'less than once a week'; 'I do not smoke'.

In countries of the European Union with membership before 2004, a converging trend among adult smokers has been observed [8]. However, this trend was not observed in daily smoking among adolescents. Taking into account also some new member states, in 2002 the smoking prevalence among boys varied from 5.5% to 20.0%. Among girls, it varied from 8.9% to 24.7%. It is far from easy to explain this important variation between countries. Policy differences as well as differences in youth cultures can play a role.

Interestingly enough, smoking prevalence within countries is not linked with the observed smoking trends between 1990 and 2002. Among boys as well as girls, three different trends were observed showing the same geographical pattern. Among boys, the Nordic countries show a declining or stabilising smoking trend; in the Western countries an initial increase is followed by a decrease in daily smoking; and in the Eastern European countries an increase is followed by a stabilisation in smoking prevalence between 1998 and 2002. Among girls, similar daily smoking trends can be found, with only a few exceptions. First, no country in this study shows a continuous decline in daily smoking prevalence among girls. Second, Austria and Hungary show an increasing smoking trend in girls, while in boys a stabilisation is observed. Third, Hungary is the only country in this study where smoking prevalence among girls has increased since the last two surveys.

Pirkins et al. [9] state that cross-national data of adolescent substance use should be interpreted cautiously. When comparing data from cross-national surveys, the list of problems includes differences in population focus, differences in sampling method, a different survey context and question wording. The HBSC study attempts to control these problems by adapting standardised methods [6]. Literature on smoking trends using the same methods over different periods and in different countries is very scarce [8].

A weakness in large scale school-based studies is the self-report of substance use. In general, self-reported smoking prevalence has been considered as a good indicator of the actual smoking status, compared with biochemical validated smoking prevalence [10, 11], especially in epidemiology. But it may still give an underestimation of the problem in adolescents [11]. Although the questionnaire had to be completed anonymously, cultural differences in answering questions (especially questions with a social stigma) can be a problem (like tobacco use in some countries and certain periods for girls and/or boys). Validation studies in this respect are mostly done in Western countries. It would be interesting to repeat such validation studies in countries with a different cultural background. Another limitation of this school-based study is the fact that school drop outs, which may be a high-risk group for smoking, are not included in the survey (at least in some countries). And finally, information referring to smokeless tobacco is lacking. For instance in Sweden, smokeless tobacco is much used among youngsters, especially boys (14.5% used snuff weekly in 2002) [12]. Hence, in some of the participating countries, the daily smoking prevalence can be an underestimation of the tobacco-related problem in reality.

This paper concentrates only on daily smoking among adolescents, which may give a misleading picture of the whole smoking epidemic. When daily smoking is declining, this behaviour can be overtaken by occasional smoking. According to McNeil [13], smoking among adolescents may well show important fluctuations in regularity, from weekly to daily smoking. However, since daily smoking is defined as an important part of nicotine dependence [14], we decided to use this indicator in order to get a clear picture of the current and future burden of smoking on the public health. Daily smoking adolescents are more likely to smoke in the future and to develop smoking-related health problems leading to premature deaths.

This is a descriptive epidemiological study. To help policy makers, analytical epidemiological studies explaining differences in smoking prevalence and trends are needed. Further analyses are needed on different levels of information (individual, population and country characteristics).

Among adults and, as observed in this study, also among adolescents, gender and country differences in smoking trends follow the four stage model of the smoking epidemic [15], and 'Diffusion of Innovations' theory proposed by Rogers [16]. In the first stage of the smoking epidemic model, smoking begins as a male habit; after men have adopted smoking, females begin to smoke in the second stage; in the third stage, male prevalence begins to decline, while female smoking prevalence remains stable; the fourth phase is characterised by a decline in both genders. It may well be that different countries are facing different stages. However, if this theory holds, most of the countries studied here are found in stage three. This should be further examined. Following Rogers' theory, innovations, such as smoking, are taken up first by communities marked out by their relative advantage in terms of educational level, socioeconomic status and upward social mobility [16]. The observed geographical pattern in smoking trends reflects this theory.

However, these theories do not explain the large differences in smoking prevalence between the countries. As documented in the 2004 ENSP report (European Network for Smoking Prevention), effective tobacco control efforts targeting adolescents are not taken in all countries [4]. Among adolescents, most effects are obtained by increasing taxes and prices, restricting advertising, sponsoring media campaigns and subsidising cessation treatment [4]. Although the whole smoking prevalence pattern cannot be explained by the implementation of these measurements, it is noteworthy that countries scoring high on these components (like the UK, Sweden and Norway) have also a relatively low smoking prevalence, especially among boys. Countries scoring low on these components (like Latvia and Austria) have relatively high smoking prevalence, again especially among boys.

From this paper, we can conclude that among European adolescents, three groups of countries in a different stage of the smoking epidemic curve can be identified, with girls being in an earlier stage than boys.

As smoking-attributable mortality is most closely related to smoking patterns from thirty or more years earlier and not to the current smoking prevalence [15], the results in this paper predict a huge burden on the health care systems of Eastern European countries over the next 20–30 years. Policy makers in these countries must be encouraged to initiate cost-effective strategies for tobacco control as proposed by the World Bank [17]. But equally important, countries with a declining or stabilising daily smoking trend among adolescents must remain alert. Policy makers there should face the challenge to keep the smoking prevalence declining or at least stable. This can be done by developing initiatives that are innovative and suitable for both boys and girls.