Smoking is a major contributor to all-cause mortality in Europe and accounts for one-fifth of the cancer-related deaths. Monitoring the tobacco epidemic via an analysis of lung cancer trends is essential in helping countries arrest the effects of tobacco epidemic in the region. The study aims to provide a comprehensive and up-to-date overview of the temporal patterns of lung cancer mortality in Europe, emphasizing country- and sex-specific differences. National lung cancer mortality data were extracted from the WHO mortality databank by age, sex, year of death (1970-2007) for 36 countries in Europe. Trends in lung cancer mortality in men have tended to decrease in many European countries during the last two decades, particularly in North and Western Europe. Among women, mortality rates are still increasing in many countries, although in a few populations, rates are beginning to stabilize, notably in the high-risk countries within Eastern Europe (Hungary, Poland and the Czech Republic), and in Northern Europe (Denmark, Iceland and the United Kingdom). Men and women are clearly in very different phases of the smoking epidemic, and, as reflected in the mortality rates by birth cohort, the stage varies widely by country within each European region. That lung cancer mortality trends in men are on a downwards path in most European countries while female rates continue to rise, points to an urgent need for national and European prevention strategies that target tobacco cessation and prevention among European women.
Smoking is a major contributor to all-cause mortality in Europe, accounting for one-fifth of all cancer-related deaths, with lung cancer a major component of this burden. The disease has now surpassed breast cancer as the leading cause of cancer death in a number of developed countries, including the United States, Denmark, Sweden, The Netherlands, Poland and the United Kingdom.The level of rates and temporal patterns of lung cancer incidence or mortality largely reflect the different phases of the smoking epidemic in men and women in different European countries. These tend to differ markedly from country to country, but in general, the decreases or stabilizations of rates in men observed in many European countries in recent years, are in contrast to the uniformly increasing rates seen among women, often within the same country.
Careful monitoring of the trends of the smoking epidemic has been recently put forward by the WHO as one of six key policies in tobacco control. The close relation between the temporal patterns of lung cancer mortality and smoking prevalence indicates the dissemination of trends in the former provides key information for the targeted primary prevention of cancer and other noncommunicable diseases.
The aim of this study is to provide such a comprehensive and up-to-date overview of the temporal patterns of lung cancer mortality in the European region with an emphasis on the country- and sex-specific differences observed. The long-established generational influence of tobacco consumption and subsequent lung cancer provides the rationale for an inclusion of time trends according to birth cohort. Such trends provide an early indication of the stage of the lung cancer epidemic in men and women in each country and region of Europe, enabling a timely assessment of the relative success of tobacco control, drawing attention to specific populations for which further concerted action is needed.
Material and Methods
National mortality were extracted from the WHO mortality databank, which contains data on all deaths officially reported by WHO Member States by topography, coded and tabulated in successive revisions of the International Classification of Diseases (from ICD-7 through to ICD-10), 5-year age group, sex and year of death. Corresponding population data were extracted from the same source.
To provide an overview of the current geographical patterns and recent trends, countries were included if they were defined as within the European region (according to the United Nations classification), and data spanned at least 10 years, with availability including at least the year 1995. The U.N. classification was used to define the geographical region of each country (Eastern, Northern, Southern and Western). Data for 36 countries met these criteria, and the availability in terms of most recent year of death and the span of the series in each country is provided in Tables 1 and 2, respectively.
To enable meaningful birth cohort analyses (see methods later), countries were excluded from the aforementioned list if the mortality data spanned less than 15 consecutive years. Mortality data were missing for certain years in a number of countries with spans of data greater than 15 years. The years were reselected in these countries to ensure a study period of the least 15 years with no gaps, or they were removed from the analysis where this could not be achieved. 31 countries were included in this part of the analysis.
Aggregated age-specific and age-standardized (Europe) mortality rates per 100,000 were estimated, with 5-year moving averages used to smooth out some of the random variation inherent in the year-specific adjusted rates in the presentation. Moving averages of the age-truncated rates for the ages 20-44 are also presented. Restricting the age range to 35-79, synthetic 10-year cohorts were obtained on subtracting the midpoints of 5-year age groups from the corresponding midyears of 5-year calendar time. These trends are presented as rates versus birth cohort by age, with quasi-parallelism of the age-specific curves an indication of generational influences. Such cohort effects may be established by environmental determinants acting during birth or early in life, but for lung cancer, they approximate patterns of tobacco consumption that are shared in generations as they age together. This coupled with the protracted but time-varying duration of lung cancer carcinogenesis means that changes in mortality rates with time will commonly show up as a birth cohort phenomenon.
Assuming the mortality rates were constant within the 5-year age classes a = 1,2,,A and 5-year periods of diagnosis p = 1,2,,P, an age-period-cohort (APC) model was fitted, with the number of cancer deaths assumed to be distributed as a Poisson random variable with the log of the person-years at risk specified as an offset:
Birth cohorts were thus derived from period and age such that c = p – a for c = 1,2,,C with C = A + P – 1, and a, p and c the parameters of the fixed effects of age group a, period p and birth cohort c. Given the limitation of APC analyses – the inherent inability to identify the individual slopes of age, period or cohort due to a linear dependency between the time components, the estimated effects are presented in this paper using the full APC model, with an a priori allocation of drift to obtain a unique solution. Such a solution was obtained on constraining the linear slope of the period effects to zero, assuming the linear changes in lung cancer mortality rates over time arise purely from cohort-related factors. As the solution is entirely dependent on this allocation of drift, the results should be treated with caution. Stata 10 was used for data management and plotting the observed trends, and R used for modelling.
Geographical variations circa 2003
There is little correlation in the age-adjusted mortality rates in men and women according to geographical location (Fig. 1). A 4-fold variation in lung cancer mortality rates in men is observed circa 2003 (the year for which mortality data is available in almost all countries), with rates highest in Hungary and Belgium, and lowest in the Nordic countries of Sweden and Iceland. The rates in women tend to vary to the same extent as men, although unlike men, lung cancer mortality rates in many of Northern European countries are among the highest in Europe, including those of Iceland and Denmark. As with men, female rates are highest in Hungary.
Trends by calendar period 1970-2007
The trends by region in Figures 2a-2d indicate the general extent to which men and women in different countries are in different phases of the temporal development of lung cancer. The variations in the order of magnitude and the temporal patterns of the sex-specific rates in countries within each region are quantified in Tables 1 and 2, and described later.
Eastern Europe (Fig. 2a)
Rates in Eastern European males peaked in the early-to-mid-1990s, with the high rates in the Czech Republic declining earlier, and in the Ukraine, a reduction in rates yet to materialise. In Romania, mortality is historically low in the region, but has been increasing in both sexes. There is evidence that, in countries with the highest female rates (Hungary, Bulgaria and the Czech Republic), there are also the first signs of a plateau in the trends. Rates in women in lower-risk countries – including the Russian Federation and the Ukraine – have been decreasing at an average 2-3% per year since around 1990 (Table 2).
Table 2. Countries studied by European area (as defined by the U.N. population division), years and span available for trends analysis, age-standardized rates using European standard (Asr1), estimated annual percentage change (EAPC1) in most recent 10-year period
Northern Europe (Fig. 2b)
Rates and trends within the region are quite diverse. The Baltic countries have the highest lung cancer burden, but rates plateau around 1990. The historically high rates in the U.K. and Finland have been on the decline for several decades, with Ireland and Denmark following suit circa 1985. Rates in Sweden have been rather low among men, and on the decline as early as 1980. In contrast, rates in Icelandic men have rapidly increased from a low order of magnitude in the 1970s, but as with Norway, rates are now rather stable. Among Northern European women, trends have tended to stabilize within the last decade among the highest-risk countries, including Denmark, Iceland and the U.K. The intermediate female rates in Norway and Sweden (and to a lesser extent, Finland) have been steadily increasing year on year, with little or no evidence of a peak as yet emerging.
Southern Europe (Fig. 2c)
Rates in men have flattened out or declined in many Southern European countries (Spain, Italy, Greece, Slovenia) around 1990, with slightly later peaks in Albania and high-risk Croatia. Exceptions include Portugal where rate increases of 0.5% have been observed in the last decade (Table 2), with a plateau suggested, but not yet clearly observed. In women, trends indicate an early phase in the temporal development of lung cancer, with rates relatively low across countries but increasing, most notably in higher-risk Slovenia and lower-risk Spain.
Western Europe (Fig. 2d)
Some of the most homogeneous patterns of lung cancer are seen within the region, although they are quite different between the sexes. In men, rates have tended to peak in the high-risk Benelux countries circa 1985, with this plateau observed earlier and later in the lower-risk countries of Austria and Switzerland, respectively. In contrast, the historically low rates observed among females have been rapidly increasing in allsix countries, most notably in France and the Netherlands, where mortality rates have increased at 4-5% per year on average (Table 2). A relative stabilization in the level of mortality has yet to emerge in any of the female populations in the region.
Trends in young ages (20-44)
Mortality trends among the youngest members of the population provide an early indication of the status of the epidemic; in Figure 3, they are shown for five countries, selected given they represent varying mortality levels in Europe, with increasing rates among females. It is evident that rates in women at younger ages in these countries follow those of men, albeit with a delay ranging from 5 to 25 years. It is interesting to contrast, for example, the similar rates and trends of lung cancer in Greece and Spain in men, yet quite heterogeneous patterns exhibited for women.
Trends by birth cohort 1891-1966
Figures 4a-4d illustrate the trends versus birth cohort by age at death for (i) males and (ii) females. The unambiguous parallelism exhibited on the semi-log scale illustrates the generational influence at the population level, with a general observation of initial increases in lung cancer mortality across successive birth cohorts of men and women. This phenomenon tends to be followed by a short-term stabilization and then a continuous decline in rates among more recently-born cohorts. The specific cohort for which such a stabilization or downturn occurs clearly varies by region and country and is generally seen later in women than in men.
Cohorts of men in the United Kingdom, Sweden and Finland already experienced consecutively lower rates of lung cancer mortality during the first decade of the 20th century or earlier (Fig. 4b(i)), while in France and many Southern European countries, generational declines are not seen until after the Second World War. With the notable exception of Portugal, there are declines apparent in male lung cancer rates across Europe among recently born cohorts.
Among females, the pattern among recent cohorts varies considerably by country within region. In the United Kingdom and Ireland, rates of lung cancer have been diminishing in generations born after 1925, followed by postwar declines seen in several Nordic countries. In much of Western and Southern Europe, however, rates have been increasing steadily in cohorts up to the 1950s, and in the likes of the Netherlands and France appear to have accelerated among those born after World War II. There are however early signs of a cohort-led decline in a number of countries including Austria, Switzerland, Hungary and the Czech Republic. In contrast to the high rates in the aforementioned Eastern European countries, rates in the Russian Federation and the Ukraine have been consistently low, trends stable or declining in recent generations.
Age-period-cohort model analyses
Figures 5a-5d summarizes the observed sex-specific trends in a single measure by country according to European area. The cohort parameters are obtained on assuming the sex-specific linear trends in each country are due to generational influences. On the basis of this assumption, it is possible to contrast the relative stages of the epidemic in men and women by country, and pinpoint the specific cohorts for which the male and female cohorts reach their maximum risk, where applicable. In Western Europe, in particular, there are clearly very different phases of the temporal development of lung cancer in relation to gender, with generations of women (1955-1970) tending to reach their maximum risk 50 years after men (circa 1905-1920) in six of the seven countries. This pattern is repeated in the Nordic countries, the United Kingdom, Italy and the Czech Republic. There are many instances, however, in France, Portugal, Spain and Hungary, where maximum lung cancer risk is observed in the very recent male and female generations born around 1960. Another pattern emerges in certain Eastern European countries including Russia and the Ukraine, where maximum risk appears set in certain cohorts born between the Wars, again, irrespective of sex.
This study represents a comprehensive and up-to-date analysis of the time trends of lung cancer mortality in Europe, taking into consideration the major differentials across 36 countries in 4 regions by sex and the components of time (age, calendar period and birth cohort). The article serves to highlight the varying phases of the lung cancer epidemic in different European countries and the disparities in rates and time trends between the sexes. These observations reflect the smoking habits of generations of men and women born from the late 19th century onwards, but also point to the relative successes or failures of smoking prevention and cessation efforts over the continent in the last few decades.
Men and women are clearly in very different phases of the smoking epidemic, and these may differ considerably between neighboring countries. The trends in lung cancer mortality in men are largely encouraging, in that there are general declines in most European countries, particularly during the last one or 2 decades, and among successive birth cohorts born as early as 1900 (the United Kingdom) or as late as 1955 (France and Italy). While the overall male rates are still increasing in parts of Southern Europe (Portugal and the Republic of Macedonia), and in Eastern Europe (Bulgaria, Romania, and the Republic of Moldova), the declines observed in recent birth cohorts indicate an overall decline may emerge in the next years.
Among women, lung cancer mortality rates have reached a plateaux or are beginning to decline in a number of Eastern European countries (notably in the high-risk countries of Hungary, Poland and the Czech Republic), and in Northern Europe (in particular Denmark, Iceland and the United Kingdom). This is reflected in the generational trends and successive declines in risk among females born predominantly after 1950 in these regions. While there are increasing trends in the all-ages rates among women in the Nordic countries of Sweden, Norway and Finland, declines may be anticipated in the future assuming that the successive decreases in risk among female cohorts born after the Second World War adequately reflect recent changes in smoking habits. More concerning are the continuing generational increases in risk among Bulgarian and Romanian women.
In contrast to much of Eastern and Northern Europe, the increasing trends in Western Europe (uniformly rising in France and The Netherlands), and in Southern Europe (rapidly among young women in Spain), are far from encouraging. Women born 1900 through to 1960 are at increasing higher risk of lung cancer death in many of the countries in these regions. Exceptions include Austria, Switzerland, Germany and Greece, where declines are seen in female rates among recent birth cohorts.
These observations indicate the evolution of the lung cancer epidemic as a reproducible birth cohort phenomenon. The increases in rates among successive birth cohorts tend to give way to short-term stabilizations followed by continuous declines among more recently-born cohorts. This pattern is observed across Europe with no evidence of any exceptions. There are however large sex- and country-specific variations in the phase and impact of the epidemic (in terms of absolute levels of rates), with women at an earlier phase than men in every country. Our estimates suggest the generation of men at maximum risk of lung cancer can be observed as much as 50 years ahead of the equivalent female cohort in some countries (e.g., the Netherlands), but as little as 10 years in others (e.g., Hungary). The differential largely relates to a varying interval between the adoption of the smoking habit in men and women and the later temporal development of lung cancer mortality within each population.
It has been noted that each phase of the study of cancer trends – data collection, analysis and presentation – bring their own set of problems. Although mortality data is prone to erroneous death certification and changes in coding practice, and there are likely age-specific variations in the accuracy and completeness between countries, the data should be a reasonably robust indicator of the accumulated hazards of smoking. Although the historically high and stable case-fatality of lung cancer implies that mortality rates and trends are good proxies of their equivalents in terms of incidence, case-fatality may have diminished in recent years with the advent of newer diagnostic techniques, better staging and treatment options for selected patients. These interventions are unlikely to have materially affected the direction or magnitude of the trends reported here, however.
The APC model is beset by identifiability problems, and we have focused on an analysis and presentation of the identifiable linear trend (drift), and a single solution assuming a linear period slope of zero. While the latter represents an arbitrary parameterization to circumvent non-identifiability that lung cancer is predominantly a birth cohort phenomenon is long established (as is evident in Fig. 4), and one that is driven by generational changes in smoking habits. The approach used here still enables non-linear period effects to be identified, potentially informative, for example, where there have been improvements in survival across all studied age groups.
Caveats aside, the disparities in mortality rates reflect sex-specific historical differences in smoking; these are constantly shifting as the phase of the smoking epidemics evolve. Globally, the prevalence of smoking in women is one quarter that of men, but in some countries in Europe, adolescent boys and girls have about the same smoking prevalence, and about one quarter of all smokers start smoking by the age of 10. The tobacco industry has been specifically targeting women as an important consumer base.
In 1995, an estimated one-third of all cancer deaths in developed countries (47% of male cancer deaths and 14% of female cancer deaths) were attributable to smoking. Besides lung cancer, other cancers of the oral cavity and pharynx, oesophagus, larynx, bladder, pancreas, kidney, cervix and breast are elevated among smokers. The WHO have estimated that during the 20th century, smoking killed 100 million people worldwide, and that in the 21st century it will kill one billion people unless effective tobacco control policies to counter the tobacco epidemic are implemented worldwide.
The key tobacco prevention measures are established but not effectively and strategically implemented across Europe. In 2008, the WHO launched its MPOWER programme comprising of a series of tobacco control measures that include the monitoring of the epidemic, as well as strategies to protect people from smoking, offer help in cessation, warn about the dangers of tobacco, enforce bans on tobacco advertising and lobby for higher tobacco taxes. With the provision of such evidence-based measures, it is hoped that much of the future lung cancer burden in the forthcoming decades can be avoided. It is evident from this surveillance exercise that progress in tobacco control has been made in Europe, but there are still clear disparities across countries and regions, with several countries continuing to have very high mortality rates, with increasing trends in women seen in many. Smoking prevention and cessation policies are essential, particularly those targeting younger women and lower socioeconomic backgrounds, in order to avert many premature and wholly preventable lung cancer and smoking-related deaths in the next decades.
The study was undertaken as part of the Women in Europe against Lung Cancer and Smoking (WELAS) project funded by the Public Health Programme of the European Commission SANCO A/800124.