Wong, Chit-Ming; Ou, Chun-Quan; Lee, Nga-Wing; Chan, King-Pan; Thach, Thuan-Quoc; Chau, Yuen-Kwan; Ho, Sai-Yin; Hedley, Anthony Johnson; Lam, Tai-Hing
Click on the links below to access all the ArticlePlus for this article.
Please note that ArticlePlus files may launch a viewer application outside of your web browser.
In the Asia Pacific region, smoking and air pollution are among the most important causes of avoidable mortality. The health hazards of both have been widely studied,1–4 but there has been limited assessment of their joint effects. A recent analysis5 of the American Cancer Society cohort showed that mortalities from cardiovascular disease and diabetes were related to both fine particulates (PM2.5) and smoking. The relative risks of death from cardiovascular diseases per 10 μg/m3 change in average exposure were slightly higher in current smokers than never-smokers for all categories of cardiorespiratory diseases under study, but the differences were not significant except for hypertension. A case-crossover analysis6 of a cohort in Bordeaux, France, also showed that mortality risks caused by all natural causes and cardiorespiratory diseases per 10 μg/m3 change in black smoke were slightly higher in current smokers than never-smokers. Smokers in more polluted areas, compared with never-smokers in less polluted areas, had greater declines in lung function, higher excess risks for nonadenocarcinoma lung cancer and for deaths from all natural causes.7–10 However the roles of individual pollutants in the interactions between air pollution and smoking were not assessed in any of these studies. It has been recently recognized that adverse effects of air pollution are stronger in some subgroups of the population. Persons with specific diseases or disadvantaged socioeconomic conditions experience greater excess risks.11–13 However, it is not clear whether unhealthy lifestyle factors such as smoking are responsible for an increased susceptibility to the adverse effects of air pollution. In this study we tested the hypothesis that smokers are more adversely affected by ambient particulate air pollution than never-smokers.
Study Population and Exposure
This study included 10,833 Chinese men in Hong Kong who died during the period 1 January to 31 December 1998. The informants registering an adult death in 1 of 4 death registries in Hong Kong were asked by a trained interviewer to provide information through a standardized questionnaire about the decedent, while waiting in the registry for the death certificate. The questionnaire included individual information for each decedent about social and demographic characteristics, and smoking status, and other lifestyle habits 10 years before death. Assistance in completion of the questionnaire by trained interviewers was provided for illiterate informants and for clarification of questions. A detailed description of the design and conduct of the investigation had been published elsewhere.2
The present analysis included male never-smokers and daily smokers who smoked at least one cigarette per day 10 years before death. Only men were included, as regular daily smoking in women was less prevalent (13%) than in men (51%).
The unique death registration number was linked to the Hong Kong Department of Health databases to obtain the certified underlying cause of death. The date of death was linked to the Environmental Protection Department databases14 from 8 monitoring stations for daily 24-hour concentrations of particulate matter with aerodynamic diameter <10 μm (PM10) and to the Hong Kong Observatory15 for daily temperature and relative humidity data. The daily PM10 concentrations among the 8 monitoring stations are highly correlated (Spearman correlation coefficient ranging from 0.86 to 0.96), and the population density (6.8 millions in 1092 km2) of Hong Kong is very high; it is therefore most likely valid to compute the overall arithmetic mean daily PM10 concentrations of the 8 stations after eliminating discrepancies between stations, and use them as a proxy measure for environmental exposure. The Ethics Committee of the institute where this study was carried out has approved the study proposal.
The underlying cause of death was coded according to the International Classification of Disease 9th revision (ICD-9). The health outcomes were deaths from all natural causes (ICD-9 1-799) and for cardiorespiratory diseases (ICD-9 390-519). We did analysis for all ages (ie, age 30+) and for the 65 or older age (age 65+) group to show the results for the complete data and for the older population.
We defined 2 measures of mortality risks; excess risk (ER) of death per 10 μg/m3 increase in pollutant for smokers and never-smokers, and difference in ER for air pollution effects in smokers relative to never-smokers. The 95% confidence intervals (CIs) for the ERs were also computed.
Time-series count data were stratified according to smoking status of the deceased 10 years before death. Generalized additive Poisson regression model for daily counts Yt at calendar day t (t = 1, 2…365) of each health outcome was constructed. The variables in the model were day of the week, holiday, influenza epidemic indicators, temperature and relative humidity for observable confounding variables Zt, and a smoothing function of t (S(t, df), t = 1, 2, 3…365) defined with certain degrees of freedom (df) for confounding variables that are not directly observable but may be related to long-term trends and seasonal variations of the data.16 Thus,
where A and g are coefficients for the constant term and Zt the confounding variables of the regression model.
We then combined the stratified data with a dummy variable defined for smoking status (C, 1 for smokers and 0 for never-smokers). Poisson regression model for the combined data was specified as
where h is the coefficient for the variable C of the regression model.
Poisson regression Model 1 for stratified data and Model 2 for the combined data were fitted with a certain degrees of freedom for temperature, humidity and time until coefficients of partial autocorrelation function were less than around 0.1 and free from discernible patterns. Then, PM10 concentration variable (X) at concurrent and previous 1–3 days was separately entered into Equation 1, and the effect of PM10 in smokers and never-smokers was assessed by the coefficient associated with X. A product of C and X (CX) and the individual variable X were then entered into Equation 2, from which the additional effect (ie, difference in effects of air pollution for smokers relative to never-smokers) was assessed by the coefficient of the CX term. Additionally, we fitted an unconstrained distributed lag model, simultaneously with inclusion of PM10 concentrations at concurrent and previous 1–3 days and their interaction terms with smoking, to estimate the overall interaction between smoking and air pollution.
Two types of sensitivity analyses were performed. Using case-crossover analysis, we compared PM10 levels at the day of individual death (ie, case) with the levels 7 days before and 7 days after death (ie, 2 controls). We also evaluated excess risk of mortality associated with PM10 and additional ER in smokers compared with never-smokers using conditional logistic regression.17 Case-only logistic regression was also applied, with smoking status for individual deaths (no controls) as the dependent variable and PM10 concentration as the independent variable, to detect the interaction between PM10 and smoking status. The case-only approach does not depend on the restrictive assumption for other time-varying factors, or on the development of the core model required in Poisson regression.18,19 Finally we assessed copollutant effects on estimates for the interaction effects between smoking and PM10 by entering the terms for main effect and the product between the smoking variable and the concentration for each of the 3 copollutants. All analysis was performed using R 2.0.1 program and Stata 8.2 statistical package (StataCorp, College Station, TX).
A total of 4182 male never-smokers and 6901 male smokers were included in the study. Compared with smokers, never-smokers were more likely to be older, locally born, better educated, and living in self-owned housing (Table 1).
During the study period, the mean daily number of all deaths attributable to natural causes was 37.4, of which 16.8 were from cardiorespiratory diseases. The mean PM10 concentration was 48.1 μg/m3, temperature 24.0°C, and relative humidity 79.2% (Table 2).
Table 3 presents separate estimates of PM10 effects for smokers and never-smokers. In smokers, the most significant effects of PM10 were associated with exposures at zero or 2 days before death. Among smokers age 30 or more, the excess risks for exposures 2 days before death were 1.8% per 10 μg/m3 increase in PM10 (95% CI = 0.5% to 3.1%) for all natural causes, and 2.3% (0.2% to 4.4%) for cardio-respiratory diseases. The same excess risks for smokers of age 65+ were 2.4% (0.7% to 4.1%) and 2.6% (0.3% to 5.0%). In never-smokers, no excess risks were observed.
Table 4 shows the results for interaction between smoking and PM10, namely, the additional ER of death associated with PM10 in smokers compared with never-smokers. The additional ER of death for exposures 2 days before death from all natural causes in smokers compared with never-smokers, were 1.9% (0.3% to 3.6%) for ages 30+ and 2.2% (0.2% to 4.2%) for ages 65+. The corresponding additional ER caused by cardiorespiratory diseases were 2.2% (−0.4% to 4.8%) and 2.4% (−0.2% to 5.2%; Table 4). The estimates of additional ER from the unconstrained distributed lag model were generally larger than those at individual lag days (Table 4).
The estimates by case-crossover analyses were roughly similar but less precise than estimates from single-lag Poisson regression model (Table 4). Additional ERs in smokers compared with never-smokers using case-only logistic regression were almost identical to those obtained by single-lag Poisson regression model. The estimated effect modification of smoking on PM10 effects changed slightly after adjustment for the main effects of each of the 3 copollutants (data now shown) and increased to a greater extent after also adjustment for their interactions with smoking (Supplementary Table 1, available with the online version of the article).
Health effects of air pollution estimated from daily time-series modeling have been increasingly used in public health decision-making.20,21 Health effect estimates based on data from routinely collected whole population mortality, health service utilization, or territory-wide environmental surveillance data,21,22 are representative of the population. Observable confounding effects from environmental covariates, which vary with time, can be controlled by their inclusion in the core model; unobservable confounding responsible for seasonal and long-time trends can be controlled by inclusion of a smoothing function of the time variable into the Poisson regression model.16,23 In all the core models we rigorously checked that the residuals are independent and random, by well-established model diagnostics using partial autocorrelation function plots, to minimize any residual confounding effects.
Although personal risk factors are unlikely to confound time-series studies of air pollution effects because they usually do not change over a short period, these factors may modify the short-term effects of air pollution. The health effect assessment based on the whole population in daily time-series studies is a good measure of effects for all subpopulations provided that health effects are homogenous within the population. If air pollution effects vary among subpopulations, (such as the differing effects we have shown between smokers and never-smokers), the effects for more susceptible subgroups (such as smokers) would be underestimated using overall effect estimates that do not take into account the interaction between this risk factor and air pollution exposure.
We observed substantial effects of PM10 in smokers, whereas the effects among never-smokers were not conclusive. We cannot determine whether the true effects for never-smokers are close to or equal to the null. One year of data, may not provide enough power to address this question; a previous simulation study showed that with one-year data the power to detect an excess risk of 2% is almost 100% and to detect an excess risk of 0.5% is 30%.24 However, even with this limitation it is apparent that smokers are at higher risk than nonsmokers from air pollution effects.
Although time-series analysis is the most commonly used method for assessing the short-term effects of air pollution, to our knowledge, this is the first time-series study to examine effect modification by an individual lifestyle factor through examining the potential interaction between the fixed individual factor and time-varying concentrations of air pollution. We found positive additional ER of mortality associated with PM10 in smokers compared with never-smokers. Sensitivity analyses illustrated that the estimates of additional ER from Poisson regression were robust against 2 alternative statistical models. The estimates by case-crossover analysis were roughly similar but less precise, with larger standard error of effect estimates than those using Poisson regression, consistent with a previous report.25 In this analysis, the case-only approach provided almost identical estimates to those using Poisson regression, confirming the relationship between these 2 methods demonstrated previously.18,19 Time-invariant factors, such as socioeconomic status, are associated with smoking but do not vary between the levels of ambient air pollution, so such factors should not confound the estimation in case-only approach. Further, the interaction effects did not diminish after adjustment for each copollutant nitrogen dioxide, ozone or sulfur dioxide, for the main effects and for the interaction effects between copollutants and smoking status of the individuals.
A number of mechanisms could explain the positive interaction between smoking and ambient particulate pollution on mortality. A possible mechanism induced by smoking may operate through decreased clearance and increased deposition and retention of particles. In a chamber study after exposure to iron oxide particles (2.9 μm aerodynamic diameter), the alveolar long-term clearance kinetics revealed a mean half-time of 124 days in healthy nonsmokers and 208 days in smokers.26 Mortensen et al27 reported faster mucociliary clearance in lifelong nonsmokers than in ex-smokers. Both mainstream and sidestream smoke inhibit ciliary beat frequencies and in some cases completely stop ciliary action.28,29 Chronic smoking has been shown to induce ciliary damage, nonreversible even after a long period of smoking cessation.30 On the other hand, on-site measurement has shown that smokers had a significantly higher total respiratory system deposition of PM2.5 than nonsmokers.31
It has been established that ultrafine particles are able to penetrate the human lung and enter the systemic circulation after inhalation.32 Many laboratory and epidemiologic studies have indicated that cigarette smoke induces structural disruption of the airway epithelial barrier33 and causes vascular endothelial dysfunction,34,35 thus increasing particulate entry (and even uptake) into the arterial wall and exacerbating consequent harmful effects. The clearance of particles penetrated and deposited is mainly subject to the macrophage-mediated phagocytosis and digestion. However, smoking has a long-term chronic effect on many important aspects of immune responses, such as neutrophil kinetics (eg, suppression of chemotaxis and phagocytosis),36 function of lymphocytes and cytokines levels (eg, interleukin 1β and interleukin-6).37 A recent case-control study found that PM2.5 concentrations were associated with absolute neutrophil counts and white blood cell counts in nonsmokers but not in smokers.38 This impairment of proper immune response in smokers may inhibit the recognition and removal of particulate matter in the body. Furthermore, smoking is associated with oxidation and decreased concentrations of the major endogenous antioxidant, glutathione39 which would exacerbate oxidative stress induced by particulate air pollution.40 Lastly, heritability may play a role in etiologic mechanism, although little about this has been studied to date. The prevalence of some specific genotypes (eg, DRD2 Taq1A and GSTP1-105) is higher in smokers than that in nonsmokers.41 Such genotypes influence smoking behavior, including the initiation and dependence,42 and effects of smoking on health,43 and people with such genotypes may be more susceptible to the effects of exposure to air pollution.44 Taken together, these factors may explain why smoking may exacerbate the adverse effects of inhaled particulate air pollution, and why smokers may suffer more than never-smokers from air pollution.
Our findings could have public health impact. A coherent public health policy aimed at the reduction of avoidable mortality from air pollution should target both environmental air quality and tobacco control. The elimination of either one of these 2 exposures can lead to 2 benefits: one from avoiding the adverse effects of the exposure that has been eliminated, and the other from avoiding the interaction of the 2 exposures. Public health policy in Hong Kong, as elsewhere, needs to be evidence-driven; otherwise the actions of policy makers and lawmakers are likely to be inadequate or challenged by vested interests. In addition, the results of this study can be incorporated into health promotion programs to motivate more smokers to quit, especially those who are concerned about air pollution.
We thank Deacon Lee, Aberdeen University Medical School, for his contribution to the literature review and assistance in clerical work.
1. Doll R, Peto R, Boreham J, et al. Mortality in relation to smoking: 50 years' observations on male British doctors. BMJ
2. Lam TH, Ho SY, Hedley AJ, et al. Mortality and smoking in Hong Kong: case-control study of all adult deaths in 1998. BMJ
3. Samoli E, Touloumi G, Zanobetti A, et al. Investigating the dose-response relation between air pollution and total mortality in the APHEA-2 multicity project. Occup Environ Med
4. Braga AL, Zanobetti A, Schwartz J. The lag structure between particulate air pollution and respiratory and cardiovascular deaths in 10 US cities. J Occup Environ Med
5. Pope CA III, Burnett RT, Thurston GD, et al. Cardiovascular mortality and long-term exposure to particulate air pollution: epidemiological evidence of general pathophysiological pathways of disease. Circulation
6. Filleul L, Rondeau V, Cantagrel A, et al. Do subject characteristics modify the effects of particulate air pollution on daily mortality among the elderly? J Occup Environ Med
7. Tashkin DP, Detels R, Simmons M, et al. The UCLA population studies of chronic obstructive respiratory disease: XI. Impact of air pollution and smoking on annual change in forced expiratory volume in one second. Am J Respir Crit Care Med
8. Xu X, Wang L. Synergistic effects of air pollution and personal smoking on adult pulmonary function. Arch Environ Health
9. Krzyzanowski M, Wojtyniak B. Ten-year mortality in a sample of an adult population in relation to air pollution. J Epidemiol Community Health
10. Katsouyanni K, Trichopoulos D, Kalandidi A, et al. A case-control study of air pollution and tobacco smoking in lung cancer among women in Athens. Prev Med
11. Bateson TF, Schwartz J. Who is sensitive to the effects of particulate air pollution on mortality? A case-crossover analysis of effect modifiers. Epidemiology
12. Sunyer J, Schwartz J, Tobias A, et al. Patients with chronic obstructive pulmonary disease are at increased risk of death associated with urban particle air pollution: a case-crossover analysis. Am J Epidemiol
13. Gouveia N, Fletcher T. Time series analysis of air pollution and mortality: effects by cause, age and socioeconomic status. J Epidemiol Community Health
16. Dominici F, McDermott A, Zeger SL, et al. On the use of generalized additive models in time-series studies of air pollution and health. Am J Epidemiol
17. Lee JT, Kim H, Schwartz J. Bidirectional case-crossover studies of air pollution: bias from skewed and incomplete waves. Environ Health Perspect
18. Armstrong BG. Fixed factors that modify the effects of time-varying factors: applying the case-only approach. Epidemiology
19. Schwartz J. Who is sensitive to extremes of temperature? A case-only analysis. Epidemiology
20. Rabl A, Spadaro JV, van der ZB. Uncertainty of air pollution cost estimates: to what extent does it matter? Environ Sci Technol
21. Wong CM, Atkinson RW, Anderson HR, et al. A tale of two cities: effects of air pollution on hospital admissions in Hong Kong and London compared. Environ Health Perspect
22. Wong CM, Ma S, Hedley AJ, Lam TH. Effect of air pollution on daily mortality in Hong Kong. Environ Health Perspect
23. Samet JM, Dominici F, Curriero FC, et al. Fine particulate air pollution and mortality in 20 U.S. cities, 1987–1994. N Engl J Med
24. Wong CM, Ma S. A pilot study for the validation and elucidation of public health related environmental effect estimates from statistical modeling of daily counts of health outcomes. 2004. Available at: http://www.hwfb.gov.hk/grants/
. Accessed June 27, 2006.
25. Fung KY, Krewski D, Chen Y, et al. Comparison of time series and case-crossover analyses of air pollution and hospital admission data. Int J Epidemiol
26. Moller W, Barth W, Kohlhaufl M, et al. Human alveolar long-term clearance of ferromagnetic iron oxide microparticles in healthy and diseased subjects. Exp Lung Res
27. Mortensen J, Lange P, Nyboe J, et al. Lung mucociliary clearance. Eur J Nucl Med
28. Riveles K, Roza R, Arey J, et al. Pyrazine derivatives in cigarette smoke inhibit hamster oviductal functioning. Reprod Biol Endocrinol
29. Knoll M, Shaoulian R, Magers T, et al. Ciliary beat frequency of hamster oviducts is decreased in vitro by exposure to solutions of mainstream and sidestream cigarette smoke. Biol Reprod
30. Verra F, Escudier E, Lebargy F, et al. Ciliary abnormalities in bronchial epithelium of smokers, ex-smokers, and nonsmokers. Am J Respir Crit Care Med
31. Frosig A, Bendixen H, Sherson D. Pulmonary deposition of particles in welders: on-site measurements. Arch Environ Health
32. Nemmar A, Hoet PH, Vanquickenborne B, et al. Passage of inhaled particles into the blood circulation in humans. Circulation
33. Sharman JE, Cockcroft JR, Coombes JS. Cardiovascular implications of exposure to traffic air pollution during exercise. QJM
34. Puranik R, Celermajer DS. Smoking and endothelial function. Prog Cardiovasc Dis
35. Maresh JG, Xu H, Jiang N, et al. Tobacco smoke dysregulates endothelial vasoregulatory transcripts in vivo. Physiol Genomics
36. Palmer RM, Wilson RF, Hasan AS, et al. Mechanisms of action of environmental factors-tobacco smoking. J Clin Periodontol
. 2005;32(Suppl 6):180–195.
37. Soukup JM, Becker S. Human alveolar macrophage responses to air pollution particulates are associated with insoluble components of coarse material, including particulate endotoxin. Toxicol Appl Pharmacol
38. Kim JY, Chen JC, Boyce PD, et al. Exposure to welding fumes is associated with acute systemic inflammatory responses. Occup Environ Med
39. Moriarty SE, Shah JH, Lynn M, et al. Oxidation of glutathione and cysteine in human plasma associated with smoking. Free Radic Biol Med
40. Donaldson K, Stone V, Borm PJ, et al. Oxidative stress and calcium signaling in the adverse effects of environmental particles (PM10). Free Radic Biol Med
41. Li MD, Ma JZ, Beuten J. Progress in searching for susceptibility loci and genes for smoking-related behaviour. Clin Genet
42. Munafo M, Clark T, Johnstone E, et al. The genetic basis for smoking behavior: a systematic review and meta-analysis. Nicotine Tobacco Res
43. Cao W, Cai L, Rao JY, et al. Tobacco smoking, GSTP1 polymorphism, and bladder carcinoma. Cancer
44. Lee YL, Lin YC, Lee YC, et al. Glutathione S-transferase P1 gene polymorphism and air pollution as interactive risk factors for childhood asthma. Clin Exp Allergy
© 2007 Lippincott Williams & Wilkins, Inc.