OBJECTIVE To examine changes in the prevalence of cigarette smoking in 35 study populations of the World Health Organisation’s MONICA Project.
DESIGN Data from two independent, community-based surveys conducted, on average, five years apart.
SETTING Geographically defined populations in 21 countries mainly in eastern and western Europe.
SUBJECTS Randomly selected men and women aged 25–64 years. Numbers of participants in each study population ranged from 586 to 2817 in each survey.
MAIN OUTCOME MEASURES Changes in proportions of current smokers, ex-smokers, and never-smokers by age and sex using data collected by standardised methods.
RESULTS Among men, smoking prevalence decreased in most populations, by three to four percentage points over five years. In Beijing, however, it increased in all age groups—overall by 11 percentage points. Among women there were increases in smoking in about half the populations. The increases were mainly in the age group 35–54 years and often in those populations where smoking prevalence among women has been relatively low.
CONCLUSIONS Smoking initiation by middle-aged women in parts of southern and eastern Europe and among men of all ages in Beijing is a matter of concern. The various public health measures that have helped to reduce smoking among men in developed countries should be vigorously extended to these other groups now at growing risk of smoking-related disease.
- cigarette smoking
- World Health Organisation MONICA Project
Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.
The role of cigarette smoking as a risk factor for several cancers and cardiovascular disease is well established. During the 1970s preventive measures were formally recommended by the World Health Organisation (WHO) and were initiated in many countries. Since then the overall trends in cigarette smoking among men in western countries have been downwards. The prevalence of smoking has been increasing, however, among women in many western countries, and among men (and sometimes women) in eastern Europe and in most developing countries.1
During the 1980s, tobacco control activities intensified in many countries and there was a growing public awareness of the adverse effects of smoking on health.2 Thus it is of interest to know whether the previously reported trends in cigarette smoking continued during the mid-1980s and which populations changed their smoking behaviour most. The WHO MONICA Project provides an opportunity for investigating some of these issues using data collected in a standardised way.
The WHO MONICA Project is a multinational study to monitor trends and determinants of cardiovascular disease. It involves collaborating centres in 21 countries over a 10-year period starting in the early 1980s. The project consists of population-based surveys of cardiovascular risk factors, surveillance of all suspected coronary events (and, optionally, strokes), and monitoring of acute coronary care in geographically defined communities.3 4
This paper examines changes in cigarette smoking in the study populations between the first risk factor survey and another conducted, on average, five years later. The following main questions are addressed.
What changes have occurred in prevalence of smoking in the populations between the two surveys?
Have the changes been similar for men and women and in different age groups or birth cohorts?
Are the magnitudes of change related to the levels of prevalence of smoking among the differing populations?
Within populations, are the magnitudes of changes related to educational level?
The study populations for the WHO MONICA Project lie predominantly in Europe, where they are widely distributed geographically and cover a range of social, economic, and political conditions. There are also a few centres in North America, Asia, and Australasia. Rates of mortality and incidence of coronary disease vary markedly among the MONICA populations. Brief descriptions of the populations were given in a previous report documenting rates of coronary events.5 The first surveys of risk factors were conducted in the early 1980s with the earliest starting in 1979 and the latest concluding in 1988. The middle surveys were carried out in the late 1980s with the earliest beginning in 1985 and the last completed in 1992. (Final surveys were conducted a decade after the first ones.) The same methodology was used for both surveys for each population.
Independent samples were selected for each survey, as the WHO MONICA Project is designed as a longitudinal study of populations, not individuals. The duration of the surveys varied among the study populations. In most populations single-stage sampling was used with stratification by age and sex. In 11 populations the sampling had two or more stages—for example, in the first stage, villages were selected, and in the second stage individuals from the chosen villages were sampled. In one population the household was the unit of sampling. The samples were all approximately self-weighting so that within strata defined by sex and 10-year age group, each person in the study population had approximately the same probability of being selected. Therefore the data were analysed as though they were collected by simple random sampling within each stratum.
Response rates for the surveys were calculated in two ways, using different denominators. For populations where the sampling frame was inaccurate and migration was common, there were subjects selected for the sample who could not be contacted and for whom it was not even possible to determine if they lived in the study area. Response rates were calculated firstly regarding these people as non-respondents and hence in the denominator (definition A) and secondly regarding them as ineligible and so not in the denominator (definition B). If no information was available on the size of this group, the second response rate was not calculated. The true response rate lies between the two rates reported. Basic information on the populations, sampling schemes, survey periods, and response rates is shown in table1.
Information on smoking was obtained using standardised methods during risk factor surveys which included physical measurements and taking blood samples. Data were obtained either by self-completed questionnaires which were usually completed at the time of the physical examination and then checked by the research staff, or by face-to-face interview. A current cigarette smoker was defined as a person who reported smoking cigarettes regularly at the time of the survey. An ex-smoker was anyone who reported having smoked cigarettes regularly in the past but was not a current smoker at the time of the survey. A never-smoker was a person who reported not being a current smoker and who had never smoked cigarettes regularly in the past. Occasional smokers (those who usually smoked less than one cigarette per day) were excluded. The proportions of occasional smokers were below 5% in most populations. Smoking of pipes or cigars was not taken into consideration.
Trends in smoking prevalence within age groups reflect both the effects of quitting and of never starting to smoke—both important targets for prevention. In populations where adoption of smoking occurs mainly at ages younger than those in the WHO MONICA Project, birth cohort-specific trends should reflect patterns of quitting smoking. In other populations, where smoking may start later in life, age-specific trends rather than birth cohort-specific trends may be more readily interpreted. For these reasons patterns of smoking were examined by both age group and birth cohort.
The age of a subject was calculated as the age at the last birthday on or before the date of participation in the survey. This was then categorised as 25–34, 35–44, 45–54, or 55–64 years. Some centres did not include the youngest age group (25–34 years) in their surveys. Any subject whose age was outside the range 25–64 years was excluded. Five-year birth cohorts were defined by the years of birth corresponding most closely to the five-year age groups 25–29, 30–34, and so on, at the middle of the first survey in each population.
Age-specific average changes over five years in smoking prevalence were estimated by fitting simple linear regression models separately for each sex and 10-year age group in each population. The models were of the form where y is the prevalence in each calendar year (obtained from either survey), t is the calendar year, and ̅t̅ ̅ is the mean value oft. Then the coefficient a is the prevalence at the mid-point of the two surveys and the coefficient bis the average annual change. This parameterisation is chosen so that the estimators for a and b are uncorrelated.6 The average change over five years was estimated by b×5 and its standard error was calculated as five times the standard error of b. For some populations the surveys were less than five years apart so the five-year averages are extrapolations from the data. To enable comparisons to be made of smoking prevalence among populations for the same period, equation 1 was also used to estimate the prevalence in 1986.
Age-standardised changes for the age group 35–64 years were calculated from the estimates of average changes for the 10-year age groups by direct standardisation using the world standard population weights of 12/31, 11/31, and 8/31 for the age groups 35–44, 45–54, and 55–64 years, respectively.7 Birth cohort-specific changes in smoking prevalence for each sex were calculated similarly.
To examine the relationship between magnitude of change and average prevalence of smoking, Pearson correlation coefficients were calculated between the estimates of a and the estimates ofb (obtained from equation 1) for all populations.
Level of education was divided into three, approximately equal categories defined by years of schooling within each five-year birth cohort and sex group in each population. This approach was adopted to take into account the rapid changes in educational level that have occurred in many of the populations. Cut-points between whole years of schooling were chosen so that the proportions in the two extreme categories, those with least and most years of schooling, were as close as possible to a third. Because of clumping of the distributions of years of schooling, however, these cut-points were changed if necessary to ensure that each of the two extreme categories had at least 15% of the participants.
To compare the changes in smoking prevalence between subgroups defined by level of education, a two-stage procedure was used. First the changes (b in equation 1) were estimated for each five-year birth cohort/sex/population/education category. Then a model was fitted with the change in smoking prevalence as the response variable, and education category, population, birth cohort, and smoking prevalence at the mid-point between the surveys (a from equation 1) as the explanatory variables. These models were estimated for each sex separately.
Some of the MONICA collaborating centres are conducting the project in more than one population. There were 37 populations in which the first and middle surveys were carried out and data of acceptable quality were available in the MONICA Data Centre in Helsinki, Finland. Of these populations, two were excluded because it was not possible to distinguish unequivocally, from the questions used, between people who smoked regularly and those who smoked only occasionally. In addition, one centre included only men in the middle survey. Thus for most analyses there were 35 populations of men and 34 of women. For eight populations the age group 25–34 years was not included in one or both surveys. In one population it was not possible to define birth cohorts. For analyses involving level of education based on years of schooling, six populations were excluded because either the information was not collected or it failed to meet the MONICA quality control criteria.
The median response rate for the first survey was 76% (range 51–91%) and for the middle survey was 74% (range 57–88%) by the more conservative definition A (table 1).
Among men the estimated age-standardised prevalence of current cigarette smoking in 1986 varied from 20.5% in Stanford (California, United States) to 59.7% in Novosibirsk (Russia) (table 2a). In most populations, smoking prevalence was highest in the youngest age group and decreased with age. Among women, smoking prevalence varied from less than 3% in Novosibirsk to almost 50% in Glasgow (United Kingdom) (table 2b). In almost all populations, prevalence among women was highest in the youngest age group and decreased with age, most dramatically from 32% in the age group 25–34 years to 1% in the oldest age group of women in Catalonia (Spain). Prevalences of smoking for birth cohorts (not shown here) exhibited patterns similar to those seen in tables 2a and 2b for age groups.
Between the two surveys, the age-standardised prevalence of current smoking decreased among men in 29 of the 35 populations although many of the changes were not statistically significantly different from zero (table 3a). The average decline in all populations over five years was three to four percentage points and this decline was similar in all age groups. The greatest declines were in Stanford and Ticino (Switzerland), whereas smoking prevalence increased significantly in all age groups of men in Beijing.
Among women, between the two surveys there were fewer overall changes in patterns of current smoking (table 3b). There were, however, large increases in smoking prevalence in urban Augsburg (Germany), Tarnobrzeg Voivodship (Poland), and Catalonia, and consistent decreases in Stanford, Belfast (United Kingdom), and Newcastle (Australia).
For men the correlations between changes in smoking prevalence and prevalence levels were mainly small and positive: 25–34 years,r = 0.09, 95% confidence interval (CI) = −0.30 to 0.46; 35–44 years, r = 0.20, 95% CI = −0.15 to 0.50; 45–54 years, r = −0.24, 95% CI = −0.53 to 0.10; and 55–64 years, r = 0.25, 95% CI = −0.09 to 0.54). The populations with lower prevalence of smoking among men showed larger decreases. For women, in contrast, the correlations were all negative as there were increases in smoking in populations with low prevalence and decreases in population with higher prevalence: 25–34 years, r = −0.20, 95% CI = −0.54 to 0.20; 35–44 years, r = −0.19, 95% CI = −0.49 to 0.16; 45–54 years, r = −0.12, 95% CI = −0.44 to 0.22; and 55–64 years, r = −0.35, 95% CI = −0.61 to −0.01.
Tables similar to tables 2 and 3 were also constructed for the proportions of ex-smokers and never-smokers by age group (not shown). The results are summarised in table 4. Among men, the usual pattern was that the prevalences of current smokers and never-smokers decreased with age and the prevalence of ex-smokers increased. One exception was Beijing, where the prevalences of current smokers decreased and never-smokers increased with age, but there were few ex-smokers.
The average trends showed a decline in current smoking among men during the study period. Changes in the proportions of ex-smokers were generally small and inconsistent. In several populations, however, there were large (more than seven percentage points) and consistent increases in the proportions of ex-smokers in all age groups. This pattern of fewer current smokers and more ex-smokers shows clearly that smoking cessation occurred in Ticino, Newcastle, and Stanford. In contrast, in Beijing never-smoking declined by 17 percentage points whereas current smoking and, to a lesser extent ex-smoking, increased.
Among women, the prevalences of current and ex-smokers decreased and the percentages of never-smokers increased with age (table 4). Over the study period, however, there was evidence of an increase in smoking cessation among younger women. In contrast to men, the prevalence of ex-smokers in women decreased with age. The percentages of ex-smokers increased and current smokers decreased (giving credence to an actual decline in smoking) in Stanford and Newcastle and among younger women in Glostrup (Denmark). There was also evidence of increases in the adoption of smoking (demonstrated by declines in never-smoking and increases in current and ex-smoking) in most age groups in several populations: Catalonia, Augsburg (urban), North Karelia (Finland), Tarnobrzeg Voivodship, and the Swiss centres.
Trends in smoking behaviour by birth cohort were examined to elucidate further the patterns of change (results not shown here). In populations where smokers are quitting, decreases in prevalence of current smoking and corresponding increases in prevalence of ex-smoking are to be expected. This pattern was indeed apparent in many populations. Downward trends in prevalence of never-smoking indicate people taking up smoking for the first time. This was evident among men in Beijing and among women in the urban population of Augsburg and in Ticino. Increases in the prevalence of never-smoking within birth cohorts are logically impossible. Where such changes are seen, they suggest either changes in the representativeness of the study samples between the two surveys due to differences in response rate or demographic changes—for example, because of migration—or changes in reporting behaviour. Large increases in the reported prevalence of never-smoking within birth cohorts were apparent among men and women in some populations in East Germany, and in the Czech Republic and Gothenburg (Sweden); among men only in Warsaw, Kaunas (Lithuania), and the populations in Russia; and among women only in Novi Sad (Yugoslavia) and Friuli (Italy).
Among men but not women, there were statistically significant differences in the prevalence of smoking between groups defined by level of education, adjusted for birth cohort and population (table 5). The prevalence of smoking was highest in the groups with least education and decreased with increasing education. Between the two surveys the decreases in smoking prevalence showed only small and not statistically significant differences between educational groups in both men and women.
Among men, the prevalence of smoking was generally high in eastern European and lowest in some northern European populations. In most populations smoking among men declined during the study period, continuing trends that had begun earlier. There was consistent evidence of reductions in the proportions of current smokers and increases in the proportions of ex-smokers. In Beijing, however, the prevalence of smoking in men increased by 11 percentage points overall. Among women, smoking increased in most populations where the prevalence was low. In populations where prevalence was high, smoking among women declined and the patterns were similar to those for men. Our results are thus consistent with the model proposed by Lopez et al.8
The strength of this study is that data were collected by methods standardised over time and among populations in countries with widely varying prevalences of smoking for men and women. Only survey results satisfying specified criteria for acceptable quality are included in the analyses presented here.
In this study the changes in smoking refer to populations, not individuals, as the data were obtained on independent samples taken at different times. This multiple cross-sectional study design is appropriate for estimating changes in populations. Estimates of smoking prevalence, and changes in prevalence, all have associated standard errors. These were not taken into consideration in the calculation of correlations and as a result the strength of association may have been underestimated somewhat.9 Only trends in cigarette smoking are discussed and not trends in other uses of tobacco, such as smoking cigars, cigarillos, or cheroots, or taking oral snuff.
In most of the study populations, smoking prevalence among men was lower among those with higher education. The declines in smoking prevalence in groups defined by level of education were very variable and there were no consistent patterns. For example, in Beijing where the prevalence of smoking is rising, increases were greater among less educated men than more educated men in each age group. Among women, there were no consistent patterns of prevalence or change in smoking in relation to education. It is possible that five years is too short a period to demonstrate different rates of change between educational groups.
It is possible that under-reporting of current smoking and over-reporting of never-smoking may have increased between the two surveys, possibly because of growing attitudes against smoking in many populations. This could account for some of the large increases in prevalence of never-smoking, as they occurred in populations with relatively high and stable response rates for the two surveys. Major demographic changes could also be the cause in some of the populations. As the strength of public opinion against smoking increases, the possibility of mis-reporting may increase so that validation of self-reported behaviour will become an important issue.
The problem of smoking among women requires special attention. In populations where few women smoke, our data show clear evidence of middle-aged women taking up smoking. This may reflect responses to intensive, targeted advertising campaigns, as well as profound societal changes in some populations. It presents a new challenge for public health. The increases in smoking among middle-aged women are additional to the increases in smoking among girls and young women reported from many countries.10 The youngest group of women included in the WHO MONICA Project, those aged 25–34 years, showed the greatest declines in smoking prevalence. However, this group is older than the adolescents about whom concern is usually expressed and it is the age group where most pregnancies occur in many developed countries. Pregnancy could be a reason for cessation of smoking, and our results provide some evidence that women, who may have begun to smoke as teenagers or in early adulthood, give up smoking in their twenties and thirties.
In many of the populations in this study, there have been intensive programmes to control tobacco and there is evidence that these have been effective. For example, some of the largest declines were in California where there has been a long-established, multifaceted, anti-smoking campaign. In many populations there have been extensive health promotion campaigns. The implications for public health policy of the findings reported here are that activities aimed at changing community and individual attitudes need to be maintained or intensified to reduce cigarette smoking, especially among women and in those countries where smoking prevalence is high.11 The International Union Against Cancer tripartite strategy provides a comprehensive approach to tobacco control, covering legislation (including, for example, taxation, restrictions on sales and promotion, and cigarette yield), education (to limit adoption and to encourage cessation of cigarette smoking), and cessation activities (using a variety of methods).12 As such strategies are more widely adopted, significant declines in smoking may be anticipated. The WHO MONICA Project, which is monitoring smoking over a 10-year period, provides an opportunity for evaluating the effectiveness of campaigns in a large number and range of communities.
MONICA centres are funded predominantly by regional and national governments, research councils, and research charities. The study is co-ordinated by the World Health Organisation. The MONICA Data Centre in Helsinki is supported by the National Public Health Institute of Finland and a contribution to WHO from the National Heart, Lung, and Blood Institute, National Institutes of Health, Bethesda, Maryland, United States. Grants from ASTRA Hässle AB, Sweden, Hoechst AG, Germany, Hoffman-La Roche AG, Switzerland, and the Institut de Recherches Internationales Servier, France, help to support data analysis and preparation of publications.
Sites and key personnel of contributing MONICA centres
University of Western Australia, Nedlands
Principal investigator: MST Hobbs; key personnel: K Jamrozik, PL Thompson, BK Armstrong
University of Newcastle, Newcastle
Principal investigator: A Dobson; key personnel: H Alexander, R Heller
Beijing Heart, Lung and Blood Vessel Research Institute
Beijing principal investigator: Wu Zhaosu; former principal investigator: Wu Ying-Kai; for key personnel risk factor survey: Yao Chonghua, Zhang Ruisong
Institute for Clinical and Experimental Medicine, Prague
Principal investigator: Z Skodová; key personnel: Z Pisa, L Berka, Z Cicha, J Cerovská, R Emrová, M Hoke, M Hronkova, J Pikhartová, R Poledne, P Vojtisek, J Vorlicek, E Wiesner
Copenhagen University Hospital, Glostrup
Principal investigator: M Schroll; key personnel: M Kirchhoff, A Sjøl, S Quitsau-Lund
National Public Health Institute, Helsinki
Principal investigator: J Tuomilehto; former principal investigator: P Puska; key personnel for risk factor surveys: C-G Gref, H Korhonen, M Jauhiainen
Country coordinator: J Richard
National Institute of Health and Medical Research (U258), Paris
Key personnel: A Bingham
National Institute of Health and Medical Research (INSERM 326), Toulouse
Principal investigators: P Douste-Blazy, JP Cambou; key personnel: MP Branchu, V Delmas, P Rodier
GSF-Institute of Epidemiology, Neuherberg/Munich
Principal investigator: U Keil; key personnel: J Stieber, A Döring, B Filipiak, U Härtel, HW Hense
Centre for Epidemiology and Health Research, Berlin (from October 1990—previously German Democratic Republic)
Principal investigator: W Barth, L Heinemann; key personnel: A Assmann, S Böthig, G Voigt, S Brasche, D Quietsch, E Classen
Medical University of Pécs, Institute of Public Health, Pécs
Key personnel: I Ember, J Tényi, I Szilard
Heart Preventive Clinic, Reykjavik
Principal investigator: N Sigfússon; key personnel: II Gudmundsdóttir, I Stefánsdóttir, T Thorsteinsson, H Sigvaldason
Institute of Cardiology, Regional Hospital, Udine
Principal investigator: GA Feruglio; key personnel: D Vanuzzo, M Palmieri, M Spanghero, M Scarpa, L Pilotto, G Cignacco, R Marini, GZilio
University of Milan, Institute of Occupational Health, Milan
Principal investigators: GC Cesana, M Ferrario; key personnel: R Sega, P Mocarelli, G DeVito, F Valagussa
Kaunas Medical Academy Institute of Cardiology, Kaunas
Principal investigator: J Bluzhas; key personnel for risk factor surveys: S Domarkiene, A Tamosiunas, R Reklaitiene
University of Auckland, Auckland
Principal investigator: R Beaglehole; key personnel: R Jackson, R Bonita, A Stewart, D Mahon, W Bingley
Unit of Clinical Epidemiology and Population Studies, School of Public Health, Jagiellonian University, Krakow
Principal investigator: A Pajak; former principal investigator: J Sznajd; key personnel: E Kawalec T Pazucha, M Malczewska, I Mórawska
National Institute of Cardiology, Warsaw, Department of Cardiovascular Epidemiology and Prevention
Principal investigator: SL Rywik; key personnel: G Broda (coordinator), M Polakowska, P Kurjata
State Research Centre for Preventive Medicine, Moscow
Principal investigator: TA Varlamova; key personnel: A Britov, V Konstantinov, T Timofeeva, A Alexandri, O Konstantinova
Institute of Internal Medicine, Novosibirsk
Principal investigator: YP Nikitin; key personnel: S Malyutina, V Gafarov, V Feigin
Institute of Health Studies, Department of Health and Social Security, Barcelona
Principal investigators: S Sans, I Balaguer-Vintró; key personnel: L Balanà, G Paluzie, T Puig
Department of Medicine, Ostra Hospital, Göteborg
Principal investigator: L Wilhelmsen; key personnel: P Harmsen, A Rosengren, G Lappas
Umea University Hospital, Department of Medicine
Principal investigator: K Asplund, F Huhtasaari; key personnel: B Stegmayr, V Lundberg
University Institute of Social and Preventive Medicine, Lausanne
Principal investigator: F Gutzwiller (Zürich); key personnel: M Rickenbach, V Wietlisbach, F Barazzoni, F Mainieri, B Tullen
The Queen’s University of Belfast, Belfast, Northern Ireland
Principal investigator: A Evans; key personnel: E McCrum, T Falconer, S Cashman, C Patterson, M Kerr, D O’Reilly, A Scott, M McConville, I McMillan
University of Dundee, Dundee, Scotland
Principal investigator: H Tunstall-Pedoe; former co-principal investigator (population surveys): WCS Smith; key personnel: R Tavendale, K Barrett, C Brown, M Shewry; former key personnel: I Crombie, M Kenicer
Stanford Center for Research in Disease Prevention, Stanford, California
Principal investigator: SP Fortmann; key personnel: A Varady, M Hull
Health Centre “Novi Sad”, Novi Sad
Principal investigator: M Planojevic; former principal investigator: D Jakovljevic; key personnel: A Svircevic, M Mirilov, T Strasser
MONICA Management Centre World Health Organisation, Geneva
Responsible officer: I Martin; former responsible officers: I Gyarfas; Z Pisa, SRA Dodu, S Böthig; key personnel: MJ Watson, M Hill
MONICA Data Centre—National Public Health Institute, Helsinki, Finland
Responsible officer: K Kuulasmaa; former responsible officer: J Tuomilehto; key personnel: V Moltchanov, A Molarius, E Ruokokoski
MONICA Steering Committee
A Evans (chair), M Hobbs (chair publications subcommittee), M Ferrario, H Tunstall-Pedoe (rapporteur), I Martin, K Kuulasmaa, A Shatchkute (WHO, Copenhagen), consultants: A Dobson, Z Pisa, OD Williams
Previous Steering Committee members: S Sans, F Gutzwiller, SP Fortmann, A Menotti, P Puska, SL Rywik, U Keil, R Beaglehole, and former chiefs of CVD/HQ, Geneva (listed above)—V Zaitsev, J Tuomilehto, I Gyarfas
Former consultants: MJ Karvonen, RJ Prineas, M Feinleib, FH Epstein (Zurich, Switzerland)