Epidemiologic surveillance of the tobacco epidemic.
The scientific evidence that cigarette smoking and other forms of tobacco use are a major cause of illness and mortality is now overwhelming. In populations in which cigarette smoking has been prevalent for some time, virtually all cases of lung cancer and a very substantial proportion of cases of coronary heart disease, cerebrovascular disease, peripheral vascular disease, chronic obstructive lung disease, and cancers of other sites are due to smoking . Smoking is thus a (if not the) major underlying cause of premature death from chronic disease in industrialized countries. This provides a strong argument that routine surveillance for tobacco use and its effects on health is an essential component of the public health system.
During the past two decades, there have been some major advances in epidemiologic techniques and the availability of data that permit one to quantify more precisely the mortality attributable each year to tobacco use. Certainly, for public health surveillance(*) and health policy formulation, information on trends in mortality attributable to smoking is likely to be at least as useful as information on smoking prevalence and consumption or trends in mortality or morbidity from lung cancer, ischemic heart disease, and other smoking-associated conditions.
These trends in disease are often difficult to interpret because of the influence of multiple risk factors that may or may not interact with tobacco use. Probably the most widely known calculations of smoking-attributable mortality are those for the United States prepared and disseminated by CDC's Office of Smoking and Health, although estimates based on a scaling of the American experience have recently become available for all developed countries.
There are, of course, some major similarities underlying the principles of public health surveillance for both communicable and noncommunicable health problems. The latter are almost always preceded by a period of cumulative exposure to one or more risk factors or other types of adverse health behavior. This is also true for such communicable diseases as human immunodeficiency virus (HIV) infection (which causes acquired immunodeficiency syndrome [AIDS]) and certain diarrheal diseases resulting from persistent exposure to unsanitary environmental conditions. At the risk of oversimplification, however, we can identify two principal differences between the etiology and natural history of noncommunicable and communicable health problems that have major implications for public health surveillance. First, noncommunicable health problems must be viewed along with the array of personal adverse health behaviors that typically precede them. Indeed, they extend further back along the causal chain, in the context of the sociocultural and occupational environment of the individual. Communicable diseases, on the other hand, are less frequently the result of detrimental health behavior by individuals. The second principal difference is the long time lag between first persistent exposure and subsequent mortality that usually distinguishes communicable from noncommunicable chronic diseases.
The sections below contain a brief overview of some of the principal issues associated with surveillance for smoking and its health consequences, focusing on lung cancer, the best known of the smoking-related conditions. Indeed, the interrelationship between smoking and lung cancer provides a useful illustration of the broader surveillance issues identified above.
EPIDEMIOLOGIC SURVEILLANCE OF THE TOBACCO EPIDEMIC
From these introductory remarks it is clear that, as a minimum, surveillance of the tobacco epidemic should include routine monitoring of both exposure to tobacco and subsequent health consequences. Nonetheless, the results of this surveillance might not be sufficient to influence health policy. When feasible, surveillance systems need to be complemented by periodic information from careful epidemiologic investigation into the health effects of smoking in the local population. Other countries--or even specific populations within the United States--might fail to be convinced of the extremely high and still-rising relative risks of dying from lung cancer for smokers compared with nonsmokers, as indicated by data from the American Cancer Society and other studies.
These components of an epidemiologic surveillance system are illustrated in Figure 1. Information needs to be routinely collected on input variables, particularly smoking prevalence and cigarette consumption by age, gender, and -- if possible -- socioeconomic status. The importance of these variables for targeting tobacco-control programs and evaluating their effectiveness is paramount. The process by which exposure ultimately leads to illness and death needs also to be monitored, preferably through prospective studies of smokers and nonsmokers. The continuous analysis of these studies will yield trends in relative risks for specific diseases, such as lung cancer, which provide compelling evidence of the extreme health hazard represented by smoking. Finally, it is imperative to monitor the evolution of the major chronic diseases that result from persistent smoking: vascular disease, various forms of cancer, and chronic lung disease.
It is important to keep in mind, however, that monitoring diseases in populations is at best an indirect indicator of the prevalence of specific adverse health behaviors. Multiple risk factors underline many of the major chronic diseases, and it is difficult to determine from routine data the contribution of a specific exposure (e.g., smoking) to a given disease.
INDICATORS AND THE AVAILABILITY OF DATA
If an adequate surveillance system of the smoking epidemic is to cover these various components, more specific indicators of exposure, relative risk, and disease outcome are required. Suggestions for variables to be included appear below, together with a brief discussion of the availability of data for monitoring the epidemic.
The extent of smoking exposure is generally measured through prevalence surveys and, in some cases, through a population census. The prevalence of smoking is perhaps the most relevant to policy formulation of all the surveillance tools for the tobacco epidemic. Prevalence is not generally difficult to assess and is a very sensitive and timely measure of the effectiveness of tobacco-control programs. In order to be useful for surveillance, however, data on smoking prevalence must be monitored at least according to age, gender, and some measure of socioeconomic status.
The importance of disaggregated data on smoking behavior by age and gender is perhaps best demonstrated by an example from Spain shown in Figure 2. Overall, 55% of men and 23% of women ages [is greater than or equal to]16 years were current tobacco smokers in 1987 . The marked gender differential in smoking prevalence seen in Spain is found in all countries except the few (e.g., United States and United Kingdom) in which the smoking epidemic has been entrenched for several decades. What is perhaps of greater importance is the marked age differential in smoking prevalence, particularly among females, which is concealed if only aggregate data are collected. An overall prevalence for females of 23% in Spain is thus quite misleading since it aggregates the experience of older Spanish women, among whom smoking is rare, with that of younger women, about half of whom smoke. The overall prevalence of lung cancer among Spanish women is still relatively low, although if young women become regular smokers, as is likely, we can expect to see a massive epidemic of lung cancer among Spanish women within the next decade or so.
Lack of disaggregation of data, together with a poor appreciation of the long delay between exposure and effect, has frequently led to major confusion regarding the effects of smoking on health . This point is illustrated in Figure 3, which shows the relationship between adult per capita consumption of tobacco during the late 1950s and early 1960s and the death rate (age-standardized) from lung cancer 25 years later. Although a general association is evident from the graph, the exceptions are undoubtedly much more important. Thus, within the solid lines defining the vertical box labeled A, one finds countries such as Japan and the Netherlands with similar per capita consumption in 1960 but with a more than twofold difference in lung cancer mortality in 1985. This example is frequently offered by the tobacco industry to support their claims of a lack of association between smoking and lung cancer. Clearly, however, such aggregate data, even with some attempt to control for the delay between cause and effect, are totally inappropriate for surveillance because they do not take into account effective duration of smoking (which was considerably higher in the Netherlands in 1960 than in postwar Japan) and because they do not allow for international differences in smoking characteristics, such as the tar and nicotine content of cigarettes consumed, degree and frequency of inhalation, length of cigarette remaining when discarded, and use of nonmanufactured cigarettes. Similar arguments could be advanced to account for the considerable difference in cigarette consumption patterns between the United States and Denmark. Death rates from lung cancer were similar in both countries in 1985, but cigarette consumption was more than twice as high in the United States in 1960 as in Denmark (Figure 3, horizontal box B). In the United States, as in the United Kingdom, Canada, and Ireland, smoking among men reached epidemic proportions well before they did in Denmark. By 1985, lung cancer death rates in the United States were declining among younger age groups for both men and women, but this is not true in Denmark, which again demonstrates the need for disaggregated trend data by age and gender.
Another essential axis of disaggregation is socioeconomic status as assessed by education, occupation, or some other proxy variable. In countries in which public knowledge about the health hazards of smoking is limited (e.g., in many developing countries), it is quite probable that smoking will be at least as common among those in higher socioeconomic classes as among the less wealthy. With the advent of health-promotion campaigns, however, the most responsive sector of the population is typically the most educated, with the result that smoking prevalence declines rather rapidly in this group but less rapidly in others . This is illustrated by the data for the United States shown in Figure 4. In the mid-1970s, only college graduates showed markedly different smoking behavior, with a prevalence about 10% lower than that for all other educational groups. Ten years later, this gap had widened to almost 20%; moreover, smoking prevalence in 1985 was significantly lower for those with some college education than for those who had completed no more than high school. Similar trends are apparent for other countries as well (Table 1). Clearly, the surveillance of trends in smoking prevalence among major socioeconomic groups is indispensable if health policies and health programs to reduce smoking are to be targeted correctly.
Other Smoking Parameters
Several other aspects of tobacco use in addition to prevalence could be useful in public health surveillance efforts. Daily or some other periodic measure of consumption is obviously highly desirable because the risk of smoking-related illness is directly proportional to the amount smoked. More importantly, the duration of smoking has a much greater impact on risk, with the incidence of lung cancer, for instance, rising in proportion to the fourth or fifth power of the effective duration of smoking . Duration can be assessed through simple questions about the age at onset of regular smoking. Other parameters likely to affect the risk of having smoking-induced illness include the degree of inhalation and, in particular, the average tar and nicotine content of cigarettes. There have been substantial declines in tar and nicotine levels over the past two or three decades as a result of research on carcinogens. Lower tar and nicotine levels appear to reduce the risk of lung cancer, but they do not appear to affect the risk of vascular disease, which is the principal cause of smoking-related deaths.
Overall, data on smoking prevalence are available separately for males and females for about 80 countries or areas. However, the utility of these data for public health surveillance is often limited by the lack of age- and gender-specific estimates and by severe biases arising from the way in which samples were drawn, questionnaires were administered, or data were analyzed. For example, very few countries have a behavioral risk factor surveillance system comparable with the telephone-interview approach used in the United States. Data on degree of inhalation are also scarce, and, despite their importance, questions on effective duration of smoking are not always included in surveys. Time series of average consumption levels since the early 1960s can be estimated for most countries by using trade and production data, but the utility of these series for epidemiologic surveillance is hampered by the fact that the estimates are not age or gender specific. [TABULAR DATA OMITTED]
Mortality and Morbidity
An important distinction needs to be made at the outset among the various types of data needed for surveillance of disease outcomes resulting from tobacco use. The first category of data allows surveillance for diseases for which smoking is a substantial contributory factor. These include diseases such as lung cancer, laryngeal cancer, esophageal cancer, cancers of selected other sites (e.g., pancreas), ischemic heart disease, cerebrovascular disease, chronic bronchitis, and emphysema. However, all of these diseases have multiple risk factors, and, hence, trends in mortality and morbidity from these conditions are only indirectly indicative of the health effects of smoking.
Data on deaths causally or otherwise associated with smoking are available for about 80 countries or areas representing about 35% of the world's population . These data are primarily from industrialized countries, as well as parts of Latin America and East Asia, and reflect the cost of establishing and maintaining vital-registration systems. These data need to be viewed cautiously before they are used for disease surveillance because of the variety of factors that can influence the completeness, diagnostic accuracy, and comparability of cause-of-death statistics. Data on incidence of cancer are even more sporadic. According to the International Agency for Research on Cancer , annual incidence data from cancer registries are available for only about 15% of the world's population -- mostly representing developed countries. Nevertheless, where cancer incidence registries have well-established procedures to ensure the completeness and quality of registration data, they are likely to be highly relevant for the surveillance of the major forms of cancer resulting from tobacco use.
The second category of mortality (rarely morbidity) data that might be used for surveillance are those obtained in epidemiologic follow-up studies designed to compare the mortality of smokers with that of nonsmokers. Studies of this type (e.g., the American Cancer Society's prospective studies of over 1 million adults with follow-up covering the two periods 1959--1965 and 1982--1988) yield the most relevant and, if the studies are properly conducted, precise data on the health impact of smoking . Repeated prospective studies with sequential periods of follow-up that are more or less representative of the same population (e.g., adults in the United States) are thus the optimal source of surveillance data on the health effects of smoking. Not surprisingly, however, this type of information is not widely available because of the cost and relative complexity of large-scale prospective studies.
The question arises as to whether these two categories of surveillance data might in some way be merged to define a third category of data on trends in smoking-attributable mortality and morbidity. That is, can estimates of the comparative incidence or mortality of smokers compared with nonsmokers derived from epidemiologic follow-up studies be applied to routine incidence or death rates by age and gender to estimate trends in smoking-attributable illness for each major cause? Time series that explicitly provide such estimates (with the same periodicity as routinely collected data) are likely to be valuable for setting health policies and monitoring programs to control smoking. Estimates of smoking-attributable mortality have been prepared for the United States on this basis since the mid-1980s and, more recently, Peto et al. have estimated time series of cause-specific smoking-attributable mortality for each of the industrialized countries since 1965 . A similar exercise to estimate the smoking-attributable fraction of new cases of lung cancer on the basis of incidence data from cancer registries has also recently been conducted by Parkin and Sasco .
A summary of the estimated trend in smoking-attributable deaths from all forms of cancer is shown in Table 2 for selected countries. The rapid emergence of the epidemic is quite clear in all countries -- initially for males, but now for females as well. Given the specificity of the estimates by age, gender, and cause, it is not difficult to monitor trends in associated indicators that might have considerable policy relevance in themselves, such as the proportion of smoking-attributable deaths that can justifiably be considered "premature." This proportion, for cancer, is also shown in Table 2 (for 1985 only) and indicates that 40%-60% of all cancer deaths from smoking occur, in fact, in middle-age.
When the individual national estimates are aggregated, smoking is estimated to account for about 42% of all deaths among men who die of cancer in industrialized countries each year. For women, the average proportion is currently about 8% (15%-20% in the United States and Great Britain) but is rising rapidly . [TABULAR DATA OMITTED]
Surveillance of the tobacco epidemic will undoubtedly become increasingly urgent as the magnitude of the burden of illness and mortality caused by smoking becomes more and more evident. Smoking is currently estimated to cause about 3 million deaths per year throughout the world, but by the 2020s, if current trends continue, this toll is expected to rise to 10 million deaths per year -- most of which will occur in developing countries . Health policies and programs to control tobacco use are thus urgently needed in all countries, as are adequate surveillance systems to provide the necessary information support.
It is essential that any surveillance system for monitoring the tobacco epidemic provide disaggregated data on both tobacco use and tobacco-related morbidity and mortality. Wherever possible, surveillance systems should be complemented by careful prospective epidemiologic studies that compare morbidity and mortality among smokers and nonsmokers. These two sources of information, when used in conjunction, can provide specific, policy-relevant time trends for smoking-attributable illness. Careful consideration also needs to be given to the long delay between the onset of persistent smoking and illness or death from the habit. Disaggregated data according to age, gender, and socioeconomic status will help to avoid gross analytical errors and are essential for the proper evaluation of tobacco-control policies.
(*) Throughout this paper, the concepts of "epidemiologic surveillance" and "public health surveillance" of the tobacco epidemic will be used interchangeably, although certainly epidemiologic surveillance is a component of a broader public health surveillance system that might well include such aspects as costs of smoking-induced illnesses, public attitudes and beliefs, health education, and legislation.
[1.] Centers for Disease Control. Reducing the health consequences of smoking: 25 years of progress: a report of the Surgeon General. DHHS Publication No. (CDC) 89-8411. Washington, D.C.: U.S. Government Printing Office, 1989.
[2.] Peto R, Lopez AD, Boreham J, Thun M, Heath C. Mortality from tobacco in developed countries: indirect estimation from national vital statistics. Lancet 1992;339:1268-78.
[3.] de Onis M, Villar J. La consommation de tabac chez la femme espagnole. World Health Stat Q 1991;44:80-8.
[4.] Lopez AD. Changes in tobacco consumption and lung cancer risk: evidence from national statistics. In: Hakama M, Beral V, Cullen JW, Parkin DM, eds. Evaluating effectiveness of primary prevention of cancer. IARC Scientific Publications No. 103. Lyon, France: International Agency for Research on Cancer, 1990.
[5.] Pierce JP, Fiore MC, Novotny TE, Hatziandreu EJ, Davis RM. Trends in cigarette smoking in the United States: educational differences are increasing. JAMA 1989;261:56-60.
[6.] Doll R, Peto R. Cigarette smoking and bronchial carcinoma: dose and time relationships among regular smokers and lifelong non-smokers. J Epidemiol Community Health 1978; 32:303-13.
[7.] Ruzicka LT, Lopez AD. The use of cause of death statistics for health situation assessment: national and international experiences. World Health Stat Q 1990;43:249-58.
[8.] International Agency for Research on Cancer. Cancer: causes, occurrence and control. IARC Scientific Publications No. 100. Lyon, France: International Agency for Research on Cancer, 1990.
[9.] Garfinkel L. Selection, follow-up and analysis in the American Society prospective studies. In: Garfinkel L, Ochs O, Munshinski M, eds. Selection, follow-up and analysis in prospective studies: a workshop. NCI Monograph 67. Bethesda, MD: National Cancer Institute, NIH Publication No. 85-2713, 1985:49-52.
[10.] Parkin MD, Sasco AJ. Lung cancer: worldwide variation in occurrence and proportion attributable to tobacco use. Lung Cancer (in press).
[11.] World Health Organization. Tobacco-attributable mortality: global estimates and projections. Tobacco Alert 1991;1:4-7.
|Printer friendly Cite/link Email Feedback|
|Title Annotation:||Proceedings of the 1992 International Symposium on Public Health Surveillance; speech of Alan D. Lopez of World Health Organization|
|Publication:||Morbidity and Mortality Weekly Report|
|Date:||Dec 1, 1992|
|Previous Article:||Communicating with decision makers: experiences from Zimbabwe.|
|Next Article:||Surveillance for environmental diseases in Hungary.|