IJE Advance Access originally published online on August 18, 2005
International Journal of Epidemiology 2005 34(5):1054-1062; doi:10.1093/ije/dyi172
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Article |
The contribution of smoking to inequalities in mortality by education varies over time and by sex: two national cohort studies, 1981–84 and 1996–99
Department of Public Health, Wellington School of Medicine and Health Sciences, University of Otago, PO Box 7343, Wellington, New Zealand
* Corresponding author. E-mail: tblakely{at}wnmeds.ac.nz
 |
Abstract |
|---|
 Top Abstract IntroductionMethodResultsDiscussionSummary statistics New Zealand...Declaration of InterestReferences |
|---|
Background The contributions of tobacco smoking to overall mortality and socioeconomic inequalities in mortality vary between populations and over time. We determined how these contributions varied by sex and over time in two national New Zealand cohort studies.
Methods Poisson regression and modelling were conducted on linked census–mortality cohorts for people aged 45–74 years in 1981–84 and 1996–99 (2.0 and 2.7 million person-years, respectively).
Results Contribution to socioeconomic inequalities in mortality. Adjusting for current and former smoking reduced the all-cause mortality rate ratios for men with nil educational qualifications compared with men with post-school qualifications from 1.34 to 1.29 in 1981–84 and from 1.31 to 1.25 in 1996–99, or 16 and 21% reductions in relative inequalities. Equivalent results for women were 1.42–1.41 in 1981–84 and 1.42–1.37 in 1996–99, or 3 and 11% reductions in relative inequalities.
Contribution to overall mortality. Using 1996–99 data, we estimated that if all current smokers quit and became ex-smokers, mortality rates would reduce by 11% for men and 5% for women. If everyone was a never smoker (i.e. a historically smoke-free society), mortality rates would have been 26% lower for men and 25% lower for women.
Conclusions The contribution of smoking to educational inequalities in mortality was greater for males, and increased over time for both males and females, reflecting the historically differential phasing of the tobacco epidemic by sex and socioeconomic position. Complete cessation of smoking in contemporary New Zealand would reduce both overall mortality and educational inequalities in mortality.
Keywords Tobacco, smoking, educational inequalities, mortality
Accepted 28 July 2005
|
Introduction |
|---|
|
TopAbstractIntroductionMethodResultsDiscussionSummary statistics New Zealand...Declaration of InterestReferences |
|---|
Reducing socioeconomic inequalities in health is increasingly recognized as an important goal—as echoed in recent definitions and goals of public health.1 One potential pathway to reduce such inequalities is by improving tobacco control since gradients exist for smoking by socioeconomic position in most developed countries.2–4
The estimated extent to which socioeconomic gaps in total mortality in developed countries are attributable to smoking varies widely. For specific causes of mortality such as cardiovascular disease (CVD), the contribution of smoking to the socioeconomic gradient appears to vary from a minimum,5 to around a quarter,6 and to a third or more.7,8 Similarly, lung cancer mortality differences by educational level in the 1990s have been found to vary widely between 10 European countries, especially for females, clearly indicative of varying socioeconomic differences in smoking.9 This last study also used the lung cancer patterns to extrapolate to the percentage contribution of smoking to all-cause mortality using the method of Peto et al.10 They estimated that 15–30% of male inequalities and a wide-ranging –15 to 35% of female inequalities were attributable to smoking.9 In contrast, a recent World Bank report estimates that smoking accounts for the majority of socioeconomic differences in male mortality in four countries: Canada, Poland, the United Kingdom, and the United States.11 Some within country studies, using direct analyses on individual-level data with smoking status as a covariate, have found a relatively modest contribution (less than a third).6,12–14
In addition to varying study designs, assumptions, and other sources of error, such variations in the contribution of smoking to socioeconomic differences in mortality will also depend on the cause of death and how smoking was patterned by socioeconomic position in the relevant years prior to the analyses (i.e. allowing for time lags). Indeed, the Lopez et al.15 model of stages of the smoking epidemic predicts that smoking will be more concentrated in higher socioeconomic groups in Stage 1, but more concentrated among lower socioeconomic groups by Stage 4. Many developed countries (including New Zealand16) are now well advanced into Stage 4 of the tobacco epidemic.3,9
To determine the contribution of smoking to both overall mortality, and socioeconomic inequalities in mortality, we analysed data from cohort studies of all New Zealanders during 1981–84 and 1996–99. These large studies permit a precise comparison of the contributions of smoking to mortality and inequalities in mortality both by sex and over time. We chose the highest level of educational qualifications as our measure of socioeconomic position for two reasons. First, education is obtained earlier in the life-course than when most adult mortality occurs making it less susceptible to health selection. Second, as the most distal socioeconomic factor, it is not as important to adjust for other socioeconomic factors that may be on the causal pathway from educational attainment to mortality (e.g. labour force status and income) before determining the contribution of smoking to educational inequalities.
|
Method |
|---|
|
TopAbstractIntroductionMethodResultsDiscussionSummary statistics New Zealand...Declaration of InterestReferences |
|---|
Data
The 1981 and 1996 census records were each anonymously and probabilistically linked to 3 years of subsequent mortality data,17,18 creating two separate cohort studies of the New Zealand population followed-up for 3 years. Of those mortality records linked, at least 98% are estimated to be true links.19 Also, 73% of eligible mortality records (45–74 years of age) were linked for the 1981–84 cohort, and 81% of eligible mortality records were linked for the 1996–99 cohort.20 There was variation in this percentage of mortality records linked to a census record by sex, age, ethnicity, and neighbourhood deprivation.18 To adjust for this linkage bias, we calculated weights for use in analyses. For example, if 20 out of 30 M
ori male decedents who were aged 45–64 and living (at the time of death) in moderately deprived small areas of New Zealand were linked to a census record, each of the 20 linked records received a weight of 1.5 (=30/20). Similar weights were calculated by 200–300 strata for each census cohort.19 Highest educational qualification was classified as nil, school, and post-school level qualification. The smoking variable was specified as never, ex-smoker, and current smoker. Mortality was treated as all-causes, CVD, cancer (malignant), and unintentional injury. Although, unintentional injury deaths may have some causal association with tobacco smoking,21,22 the strength of the association is probably considerably less than that for CVD or cancer. We, therefore, included unintentional injury deaths in our analyses to ensure that smoking was contributing more to inequalities in cancer and CVD; a specificity ‘test’ that would provide some reassurance that our main findings are not just owing to smoking being a proxy for other variables that confound or mediate the education-smoking association.23
Ethnicity is a major determinant of health, and health inequalities, in New Zealand.24,25 Socioeconomic position and smoking also vary profoundly with ethnicity. Therefore, we control for ethnicity (Mori, Pacific, non-Mori, and non-Pacific) in addition to age (5 year age groups) in all analyses.
Analyses
Analyses were conducted for person-years in the 45–74 years age range. We excluded younger people as tobacco is not a major contributor to death. The New Zealand Census Mortality Study (NZCMS) does not attempt to link people aged 
75 years on census night. Observations with one or more missing values for sex, ethnicity, education, and smoking were excluded, leaving 86.7 and 91.7% of the total eligible person-time available for analyses (2.0 and 2.7 million years, respectively). We used Poisson regression, adjusted for age and ethnicity, to determine the rate ratios of mortality by education before and after adjusting for smoking. We used the percentage reduction in the excess rate ratio (i.e. rate ratio minus 1) after adjusting for smoking to estimate the contribution of smoking to educational inequalities in mortality. For example, if the rate ratio before adjusting for smoking was 2.0 and after adjusting for smoking was 1.90, then the percentage reduction in the excess rate ratio was 100% x (2.0 – 1.9)/(2.0 – 1.0) = 10%.
To determine the nature of any effect measure modification between smoking and education on mortality, we also examined rate ratios of mortality for a cross-classified education and smoking variable (i.e. reference value of ‘post-school education and never smoker’, and eight other strata of smoking by education). Analyses were conducted separately by sex and cohort.
Counterfactual modelling
We modelled the percentage reduction in mortality, both for all 45–74 year olds combined and by educational strata, that would result from various counterfactual scenarios (see Table 4 for scenarios). The percentage reductions in overall mortality were simply the population attributable risk (PAR) percentages that would result from the stated shift in the smoking distribution, using the following PAR formula26
 |
= counterfactual proportion of population in ith strata of smoking and jth strata of education. This modelling was conducted for all-cause mortality only among males and females using 1996–99 cohort data only for RRij and Pij. To avoid introducing instability into the above calculations due to statistical imprecision in the observed cross-classified rate ratios, we used modelled rate ratios that assumed rate ratio homogeneity. That is, the rate ratio for current smokers with nil qualifications compared with never smokers with post-school qualifications was assumed to be equal to the product of the rate ratios for: current vs never smokers, adjusted for age, ethnicity, and education; and nil qualifications vs post-school qualifications, adjusted for age, ethnicity, and smoking. (This assumption of rate ratio homogeneity was a better fit to the observed rate ratios than an assumption of rate difference homogeneity—workings available from the authors on request.)
|
Results |
|---|
|
TopAbstractIntroductionMethodResultsDiscussionSummary statistics New Zealand...Declaration of InterestReferences |
|---|
The distributions of the person-time and deaths in the cohorts available for analyses in 1981–84 and 1996–99 are shown in Table 1.
|
The age and ethnicity adjusted rate ratios of mortality for nil compared with post-school educated people were relatively constant from 1981–84 to 1996–99 for both men [rate ratios of 1.34 (95% confidence interval (95% CI) 1.28–1.41) and 1.31 (95% CI 1.26–1.37), respectively] and women [1.42 (1.32–1.54) and 1.42 (1.34–1.50), respectively; Table 2]. Further adjusting for smoking reduced the mortality rate ratio for sex by cohort by cause of death analyses. Examining the percentage reduction in the excess rate ratio (using unrounded rate ratios), several patterns are evident. First, the contribution of smoking to educational inequalities was greater for men in both time periods. Second, the contribution of smoking to educational inequalities in mortality (all-cause, cancer, and CVD) increased over time for both men and women. Sixteen per cent of the excess rate ratio for all-cause mortality was explained by current and former smoking in 1981–84 for men, and this increased to 21% in 1996–99. The equivalent percentages for women were 3% increasing to 11%. Third, although, the mortality rate ratios were less for cancer compared with CVD, the percentage contribution of smoking to educational inequalities was higher for cancer within each sex by cohort.
|
Fourth, considering our specificity test by including unintentional injury, only the 1981–84 result for men suggested that smoking might contribute to educational inequalities in unintentional injury (i.e. 18%). However, this result was based on rate ratios with wide CIs. Further, across the four sex by cohort analyses the average percentage of the injury association supposedly explained by smoking was close to nil (i.e. average of 18, 7, –4, and –14% = 1.75%). Finally, unlike CVD and cancer, there was no pattern of an increasing contribution of smoking to educational inequalities over time for unintentional injury. These findings considered together provide support for the contention that the above results for cancer, CVD, and all-cause mortality were not just a result of smoking being a proxy for other (confounding) variables.
Table 3 presents the mortality rate ratios cross classified by smoking and education. It is evident that educational inequalities in mortality existed within all strata of smoking, and smoking was strongly associated with mortality among all strata of education. Modelled rate ratios for all-cause mortality in 1996–99, assuming rate ratio homogeneity for education and smoking, are shown in the last panel of Table 3.
|
The percentage reductions in all-cause mortality, overall and within strata of education, are shown in Table 4 for various counterfactual scenarios. Scenario (i) estimates the impact of complete smoking cessation with all smokers becoming ex-smokers within their educational group; mortality in 45–74 year olds might decrease by
11 and 5% for males and females. Further, percentage reductions for Scenario (i) are pro-equity in that they are greatest among people with nil qualifications. Scenario (ii) is a historically smoke-free population where nobody had ever smoked. The reductions in overall mortality are now profound (estimated at 26% for men and 25% for women) and greatest among people with nil qualifications.
|
Scenario (iii) is an imaginary country with no educational inequalities in mortality (everyone had the mortality rates of post-school qualified people) but still one with the same smoking distribution and mortality rate ratios associated with smoking. Scenarios (iv) and (v) are combinations of smoking interventions and eradication of educational inequalities in mortality. An alternative New Zealand c.1996–99 that was both smoke-free and with no educational inequalities in mortality might have had
34% (men) to 38% (women) lower mortality rates. An important point from these combined scenarios is that the mortality reductions are not much less than just adding the mortality reductions that would result from either changing smoking rates or eradicating educational inequalities. That is, it is very worthwhile to tackle both smoking and educational inequalities. |
Discussion |
|---|
|
TopAbstractIntroductionMethodResultsDiscussionSummary statistics New Zealand...Declaration of InterestReferences |
|---|
Main findings
This analysis indicates that inequalities in smoking contribute to educational inequalities in mortality by moderate amounts in the New Zealand population (i.e. 11% for women and 21% for men for 1996–99). Compared with estimates for European countries, this contribution in New Zealand is about mid-range.9 But as stated in the Introduction section, understanding the context and the phasing of the tobacco epidemic is critical to interpretation. Our New Zealand findings are potentially illuminating in this regard. First, the percentage contribution of tobacco increased since 1981–84 for both sexes, reflecting the increasing socioeconomic inequalities in smoking over time.16 One likely reason for this is that smoking uptake has declined more steeply amongst those New Zealanders in the highest socioeconomic level over recent decades.16 Furthermore, there is evidence that quit rates among higher-income New Zealanders have also increased more than other income groups over this time period.16 The socioeconomic differential in quit rates may have been attributable to previous economic barriers to quitting technologies (e.g. the price of nicotine replacement therapy), the impact of education on motivation and knowledge of how to quit, and being in a supportive environment.
The variation by sex in the contribution of smoking to mortality inequalities is also likely to reflect the different historical patterns of smoking prevalence by sex. In particular, the low contribution of smoking to educational inequalities in mortality for women in 1981–84 (3%) matches the small (by education) or absent (by income) gradient in female lung cancer mortality during this period.27
When considering the most recent data (1996–99) it appears that the complete cessation of tobacco use (i.e. if current smokers became ex-smokers) would be expected to reduce overall mortality rates by 11% for men and 5% for women. If the New Zealand society had historically been smoke-free, then these figures would be higher at 26 and 25% for men and women, respectively. As with any PAR% analyses, these results should be interpreted with caution owing to problems with assuming causality, the effects of group sizes (particularly the reference group), and moderate variations in the strength of association (e.g. the ex-smoker to never-smoker rate ratio was stronger among females than males in our 1996–99 cohort, possibly just due to chance). Nevertheless, our findings reinforce the view that tobacco is one of the most important single risk factors for mortality in this population.
Limitations of this analysis
Non-differential misclassification of smoking status in the census (given limitations of self-reported smoking status28) would probably have caused us to underestimate the contribution of smoking to the inequality gradient. Further underestimation of the contribution of smoking may have arisen owing to differential misclassification bias if the number of pack-years or smoking intensity varied across education (e.g. one national survey found that New Zealand smokers with less education smoked significantly more cigarettes per day29). There is also likely to have been variation in the type of cigarettes smoked (e.g. hand-rolled vs filter cigarettes) and the number of years since quitting might have varied among ex-smokers.
Second-hand smoke (SHS) exposure may also increase the overall contribution of tobacco to socioeconomic inequalities in mortality. This is because there is New Zealand work indicating that SHS increases mortality risk by 15%30 and New Zealanders with lower levels of education have higher levels of SHS exposure.31 Such exposure is also likely to have varied over time and may partly account for the trends over time found in this analysis. For example, a national law change in 1990 to control smoking in selected workplaces may have reduced SHS among office workers more than factory workers.32
Taken together, the likely misclassification biases we can identify all suggest we might have underestimated the contribution of tobacco smoking to the educational inequalities in mortality.
On the other hand, smoking is correlated with other variables that may also be mediators or confounders of the education–smoking association. For example, smoking is associated with other socioeconomic factors and lifestyle factors such as diet. Therefore, when we adjust for smoking we may also be adjusting (in part) for non-smoking contributions to educational inequalities in mortality. This particular problem exists for all epidemiological analyses that seek to partition direct and indirect effects as has been theoretically demonstrated elsewhere;33,34 although, the severity and pervasiveness of this potential bias is not clear.35 Our analyses of unintentional injury mortality did not find that smoking explained much, on average, of the education–mortality association. That is, the contribution of smoking to educational differences in mortality was specific to CVD and cancer, providing some reassurance that these main findings are not just attributable to smoking being a proxy for other variables.
In summary, our estimates of the contribution of tobacco smoking to educational differences in mortality may not be exactly accurate. Indeed, in our view, our results probably moderately underestimate the contribution of tobacco to educational inequalities in mortality. But any such bias should not vary too much between cohorts or sexes (although, the association of SHS with education probably did increase over time). Therefore, we believe our findings of sex variation, and variations over time, in the contribution of smoking to educational inequalities in mortality are robust.
Implications for research and tobacco control policies
The heterogeneity over time and by sex of the contribution of smoking to the educational gradient in mortality in this national population indicates the need for country-specific analyses that are repeated regularly over time—especially in countries with dynamic tobacco epidemics. These may especially be desirable in countries with dynamic tobacco epidemics. We controlled for ethnicity in these analyses and yet ethnic group specific analyses of the educational gradient in mortality are also desirable research outcomes in themselves. All these analyses are possible if smoking data are collected as part of the national census or other large population surveys.
These findings indicate that there is scope to reduce the educational gradient in mortality through tobacco control measures. Interventions of particular relevance include those relating to controls of marketing, raising prices, workplace smoking restrictions, and support for smoking cessation36—all of which are now being utilized in the New Zealand setting. A review of 16 studies that aimed to reduce smoking in low socioeconomic groups in various countries found that half of these have demonstrated effectiveness.37 Out of another nine studies that were not targeted at low socioeconomic groups, in five the intervention was at least as effective in low as in high socioeconomic groups. In particular, smoking cessation services in the United Kingdom appear to be reducing local inequalities in smoking prevalence rates38 and tobacco control measures may account for the declining national level educational inequalities in smoking rates for British and Italian men.2 Some of these interventions may start to yield mortality-reduction benefits within a matter of years (e.g. in terms of CVD mortality), whereas interventions that reduce the uptake of smoking would generally take decades to yield significant mortality reduction benefits. Similarly, any interventions to enhance educational levels for those with poorer education would generally take many decades to impact on mortality levels.
The counterfactual model of a historically smoke-free society is of potential relevance for governments that might wish to seek financial compensation from the tobacco industry for previous harm done to population health (e.g. via litigation). Our findings also reinforce the urgent need for governments to enhance evidence-based tobacco control interventions that reduce socioeconomic inequalities in tobacco consumption, or at least prevent the full manifestation of socioeconomic inequalities in smoking as predicted by Lopez et al.15 in Stage 4 of the tobacco epidemic. As with many research reports before, these findings also emphasize the need to protect both overall population health and socioeconomic inequalities in health by urgently ending the free market in tobacco products.
|
Summary statistics New Zealand security statement |
|---|
|
TopAbstractIntroductionMethodResultsDiscussionSummary statistics New Zealand...Declaration of InterestReferences |
|---|
The New Zealand Census Mortality Study (NZCMS) is a study of the relationship between socioeconomic factors and mortality in New Zealand, based on the integration of anonymized population census data from Statistics New Zealand and mortality data from the New Zealand Health Information Service.
The project was approved by Statistics New Zealand as a Data Laboratory project under the Microdata Access Protocols in 1997. The datasets created by the integration process are covered by the Statistics Act and can be used for statistical purposes only. Only approved researchers who have signed Statistics New Zealand's declaration of secrecy can access the integrated data in the Data Laboratory. For further information about confidentiality matters in regard to this study please contact Statistics New Zealand.
|
Declaration of Interest |
|---|
|
TopAbstractIntroductionMethodResultsDiscussionSummary statistics New Zealand...Declaration of InterestReferences |
|---|
The authors have declared no conflicts of interest.
KEY MESSAGES
|
|
Acknowledgments |
|---|
The NZCMS was funded by the Health Research Council of New Zealand and now receives ongoing funding from the Ministry of Health. The NZCMS is conducted in collaboration with Statistics New Zealand and within the confines of the Statistics Act 1975. The authors thank James Hogan and June Atkinson who conducted the analyses reported in this paper. Helpful comments on drafts were received from Martin Tobias, Darren Hunt, and two anonymous reviewers.
|
References |
|---|
|
TopAbstractIntroductionMethodResultsDiscussionSummary statistics New Zealand...Declaration of InterestReferences |
|---|
1 Beaglehole R, Bonita R, Horton R et al. Public health in the new era: improving health through collective action. Lancet 2004;363:2084–6.[CrossRef][ISI][Medline]
2 Giskes K, Kunst A, Benach J et al. Trends in smoking behaviour by education between 1985 and 2000 in nine European countries. J Epidemiol Community Health 2005;59:395–401.
3 Cavelaars AE, Kunst AE, Geurts JJ et al. Educational differences in smoking: international comparison. BMJ 2000;320:1102–7.
4 Stellman SD, Resnicow K. Tobacco smoking, cancer and social class. IARC Sci Publ 1997:229–50.
5 Suadicani P, Hein HO, Gyntelberg F. Socioeconomic status and ischaemic heart disease mortality in middle-aged men: importance of the duration of follow-up. The Copenhagen Male Study. Int J Epidemiol 2001;30:248–55.
6 Emberson JR, Whincup PH, Morris RW, Walker M. Social class differences in coronary heart disease in middle-aged British men: implications for prevention. Int J Epidemiol 2004;33:289–96.
7 Strand BH, Tverdal A. Can cardiovascular risk factors and lifestyle explain the educational inequalities in mortality from ischaemic heart disease and from other heart diseases? 26 year follow up of 50,000 Norwegian men and women. J Epidemiol Community Health 2004;58:705–9.
8 Pekkanen J, Tuomilehto J, Uutela A et al. Social class, health behaviour, and mortality among men and women in eastern Finland. BMJ 1995;311:589–93.
9 Mackenbach JP, Huisman M, Andersen O et al. Inequalities in lung cancer mortality by the educational level in 10 European populations. Eur J Cancer 2004;40:126–35.[CrossRef][ISI][Medline]
10 Peto R, Lopez AD, Boreham J et al. Mortality from tobacco in developed countries: indirect estimation from national vital statistics. Lancet 1992;339:1268–78.[CrossRef][ISI][Medline]
11 Jha P, Chaloupka FJ. Curbing the Epidemic: Government and the Economics of Tobacco Control. Washington DC: The World Bank, 1999.
12 Schrijvers CT, Stronks K, van de Mheen HD, Mackenbach JP. Explaining educational differences in mortality: the role of behavioral and material factors. Am J Public Health 1999;89:535–40.
13 van Rossum CT, Shipley MJ, van de Mheen H et al. Employment grade differences in cause specific mortality. A 25 year follow up of civil servants from the first Whitehall study. J Epidemiol Community Health 2000;54:178–84.
14 van Oort FV, van Lenthe FJ, Mackenbach JP. Cooccurrence of lifestyle risk factors and the explanation of education inequalities in mortality: results from the GLOBE study. Prev Med 2004;39:1126–34.[CrossRef][ISI][Medline]
15 Lopez AD, Collishaw NE, Piha T. A descriptive model of the cigarette epidemic in developed countries. Tob Control 1994;3:242–7.
16 Hill S, Blakely T, Thomson G, Howden-Chapman P. Smoking Inequalities: Policies and Patterns of Tobacco Use in New Zealand, 1981–1996 (Also at http://www.wnmeds.ac.nz/nzcms-info.html). Wellington, New Zealand: Department of Public Health, Wellington School of Medicine and Health Sciences, University of Otago, 2003.
17 Hill S, Atkinson J, Blakely T. Anonymous record linkage of census and mortality records: 1981, 1986, 1991, 1996 census cohorts. NZCMS Technical Report No. 3. ISBN 0-473-09110 (Also at http://www.wnmeds.ac.nz/nzcms-info.html). Wellington: Department of Public Health, Wellington School of Medicine and Health Sciences, University of Otago, 2002.
18 Blakely T, Salmond C, Woodward A. Anonymous linkage of New Zealand mortality and Census data. Aust N Z J Public Health 2000;24:92–5.[ISI][Medline]
19 Blakely T, Salmond C. Probabilistic record linkage and a method to calculate the positive predictive value. Int J Epidemiol 2002;31:1246–52.
20 Fawcett J, Blakely T, Atkinson J. Weighting the 81, 86, 91 & 96 census-mortality cohorts to adjust for linkage bias. NZCMS Technical Report No. 5. ISBN 0-473-09112-7 (Also at http://www.wnmeds.ac.nz/nzcms-info.html). Wellington: Department of Public Health, Wellington School of Medicine and Health Sciences, University of Otago, 2002, p. 92.
21 Leistikow BN, Martin DC, Samuels SJ. Injury death excesses in smokers: a 1990–95 United States national cohort study. Inj Prev 2000;6:277–80.
22 Leistikow BN, Shipley MJ. Might stopping smoking reduce injury death risks? A meta-analysis of randomized, controlled trials. Prev Med 1999;28:255–9.[CrossRef][ISI][Medline]
23 Davey Smith G, Ebrahim S. Data dredging, bias, or confounding. BMJ 2002;325:1437–8.
24 Blakely T, Ajwani S, Robson B et al. Decades of Disparity: Widening ethnic mortality gaps from 1980 to 1999. N Z Med J 2004;117:U995.[Medline]
25 Blakely T, Tobias M, Robson B, Ajwani S, Bonne M, Woodward A. Widening ethnic mortality disparities in New Zealand, 1981–99. Soc Sci Med (in press).
26 Murray C, Lopez A. On the comparable quantification of health risks: lessons from the Global Burden of Disease Study. Epidemiology 1999;10:594–605.[CrossRef][ISI][Medline]
27 Blakely T, Fawcett J, Atkinson J et al. Decades of Disparity II: Socioeconomic Mortality Trends in New Zealand 1981–1999. Wellington: Ministry of Health, 2005.
28 Caraballo RS, Giovino GA, Pechacek TF, Mowery PD. Factors associated with discrepancies between self-reports on cigarette smoking and measured serum cotinine levels among persons aged 17 years or older: Third National Health and Nutrition Examination Survey, 1988–1994. Am J Epidemiol 2001;153:807–14.
29 Ministry of Health. Taking the pulse: The 1996–97 New Zealand Health Survey. Wellington: Ministry of Health, 1999.
30 Hill S, Blakely T, Kawachi I, Woodward A. Mortality among "never smokers" living with smokers: two cohort studies, 1981–4 and 1996–9. BMJ 2004;328:988–9.
31 Whitlock G, MacMahon S, Vander Hoorn S et al. Association of environmental tobacco smoke exposure with socioeconomic status in a population of 7725 New Zealanders. Tob Control 1998;7:276–80.
32 Brander P. Evaluation of the Smoke-Free Environments Legislation Affecting Workplaces. Wellington: Health Research Services, Department of Health, 1992.
33 Robins JM, Greenland S. Identifiability and exchangeability for direct and indirect effects. Epidemiology 1992;3:143–55.[ISI][Medline]
34 Cole SR, Hernán MA. Fallibility in estimating direct effects. Int J Epidemiol 2002;31:163–5.
35 Blakely T. Commentary: estimating direct and indirect effects-fallible in theory, but in the real world? Int J Epidemiol 2002;31:166–7.
36 Kunst A, Giskes K, Mackenbach J. Socioeconomic Inequalities in Smoking in the European Union: Applying an Equity Lens to Tobacco-Control Policies. Brussels: European Network of Smoking Prevention, 2004, pp. 1–40.
37 Platt S, Amos A, Gnich W, Parry O. Smoking policies. In: Mackenbach J P, Bakker M, (eds). Reducing Inequalities in Health: A European Perspective. Routledge: London, 2002.
38 Lowey H, Tocque K, Bellis M, Fullard B. Smoking cessation services are reducing inequalities. J Epidemiol Community Health 2003;57:579–80.
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  S. E Gilman, L. T Martin, D. B Abrams, I. Kawachi, L. Kubzansky, E. B Loucks, R. Rende, R. Rudd, and S. L Buka Educational attainment and cigarette smoking: a causal association? Int. J. Epidemiol., June 1, 2008; 37(3): 615 - 624. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 T. Blakely, M. Tobias, and J. Atkinson Inequalities in mortality during and after restructuring of the New Zealand economy: repeated cohort studies BMJ, February 16, 2008; 336(7640): 371 - 375. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 M. Leinsalu, M. Tekkel, and A. E. Kunst Social determinants of ever initiating smoking differ from those of quitting: a cross-sectional study in Estonia Eur J Public Health, December 1, 2007; 17(6): 572 - 578. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 J Fawcett and T Blakely Cancer is overtaking cardiovascular disease as the main driver of socioeconomic inequalities in mortality: New Zealand (1981-99) J. Epidemiol. Community Health, January 1, 2007; 61(1): 59 - 66. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||