Contribution of smoking and air pollution exposure in urban areas to social differences in respiratory health

We used data of a large cohort with detailed information regarding respiratory health, SES, and other risk factors. Details of that cohort study have been described previously [9, 20]. Briefly, the baseline investigation of the SALIA (Study on the influence of Air pollution on Lung function, Inflammation and Aging) study group has been carried out between 1985 and 1994. These cross-sectional studies were part of the Environmental Health Surveys of the government of North-Rhine Westphalia, Germany, to obtain information regarding the effect of air pollution on respiratory health and allergies in women from the highly industrialized Ruhr district and two rural reference towns. These areas were chosen to represent a variety of environmental and socio-economic conditions. All women from predefined areas were asked to participate in the study. The study population comprised 4874 women aged 54–55 years at the time of baseline investigation and of German nationality. All participants gave informed consent for the interview, the health examination and the blood analysis. The response rate was stable throughout the consecutive cross-sections and was between 66% and 74%. Due to capacity reasons, only a subgroup, was invited to have their lung function measured (n = 2593). To avoid heterogeneity caused by specific social differences between rural and urban areas, only women from areas in the Ruhr district were included in the analysis (n = 3072). Additionally, we used only data from women with successful spirometry and with complete questionnaire data about education and covariates in the current analysis (n = 1251). The reason for this exclusion criterion was to be able to compare effects on lung function with effects on respiratory symptoms in the same study group.

The SALIA cohort study complies with the Helsinki Declaration and has received approval of the Ethics Committee of the Ruhr University in Bochum/Germany.

Lung function was assessed by spirometry using VICATEST (VICATEST 4, Mijnhardt, Odijk, The Netherlands) by trained technicians. Identical protocols were used for all lung function measurements. The measuring instruments were calibrated prior to each session. At least two acceptable spirometric measurements were obtained from a minimum of four forced expirations. This procedure is in accordance with the ATS criteria [21]. In our analysis, we used the forced expiratory volume in one-second (FEV₁) and the forced vital capacity (FVC). Linear regression models were used to predict the lung function parameter FEV₁ and FVC based on age, height, race and sex. We used the equations which are recommended by the American Thoracic Society [22]. The prediction equations for creating reference values for these women were:

FEV₁ ^predicted = 0.433-0.0036*age-0.00019*age²+0.000115*height²

FVC^predicted = -0.356 + 0.0187*age*0.00038*age² *0.000148*height²

We considered lung function to be impaired if FEV₁ < 80% or FVC < 80% of the predicted value of the respective parameter. These cut-offs were also applied in the re-analysis of the Harvard Six City Study [13]. Furthermore, we applied the GOLD criterion [23] to identify chronic obstructive pulmonary disease (COPD). According to the GOLD criterion, a post-bronchodilator measurement of FEV₁/FVC <0.7 defines COPD, at least stage one.

However, a post-bronchodilator measurement was not used in our study. Therefore, we excluded 44 women with bronchial asthma from further analysis of the association between COPD and SES and respiratory health risk factors, to avoid confounding of COPD with bronchial asthma. Bronchial asthma was considered present when diagnosed by a physician or if asthma medication was used.

Symptoms and diagnoses were assessed by a self-administered standardised questionnaire. For the respiratory health status, the questionnaire included questions regarding chronic bronchitis and bronchial asthma diagnosed by a physician, frequent cough, and frequent cough with phlegm production as well as medication. Furthermore, relevant potential respiratory health risk factors such as occupational exposure to dust, gases, vapours, wet conditions or extreme temperatures, smoking, environmental tobacco smoke (ETS) and level of education were included in the questionnaire. All returned questionnaires were checked by the investigating physician. We grouped the women into three categories according to their reported current smoking habits: non-smoker (including ex-smoker), passive smoker (ETS at home and/or work place), and current smoker. We determined the SES by categorizing the women into three levels of education using the highest school level completed by either the women or her husband as low (< 10 years), medium (= 10 years), high (> 10 years).

We used ambient air concentration of particulate matter of less then 10 μm dynamic diameters (PM₁₀) and nitrogen dioxide (NO₂) to assess environmental pollution for our analysis. Total suspended particles (TSP) and NO₂ are routinely measured by the State Environmental Agency of North Rhine Westphalia. The concentrations of NO₂ are measured half-hourly by means of chemiluminescence. TSP were gathered with a low-volume sampler (air flow: 1 m³/h) and continuously measured using beta-ray absorption. To be able to compare our results with other air pollution studies where PM₁₀, measurements were used, we multiplied the TSP measurements with a conversion factor of 0.71. This conversion factor was calculated from seven monitoring sites in the Ruhr area, where parallel measurements of TSP and PM₁₀ were performed between 1998 and 2004 [20]. Across the Ruhr district, concentrations of ambient air pollutants were measured at stationary monitoring sites representing urban background levels. The monitoring stations are approximately located on an 8 km grid in the women's residential areas. Individual ambient air pollution exposure of the women was assessed by the measurements of the monitoring station nearest to the women's home addresses and by using 5-year-mean concentrations before the baseline investigations. We used 8 mean values from 6 monitoring stations. Furthermore, we used geographic information system (GIS) (Arc GIS 9.0, ESRI Redlands, Cato, U.S.A.) to calculate the distance of a woman's residential address to the nearest major road with more than 10.000 cars/day. A distance of less than 100 metres to the nearest major road was considered to indicate an increased exposure to local traffic-related air pollution. The daily traffic counts were obtained from the State Environmental Agency of North Rhine Westphalia.

The statistical analysis was performed by using SAS^® statistical software package, version 9.1.3 (SAS Institute, Cary, NC). The statistical description of respiratory health indicators and risk factors were provided per level of SES. We used the Cochran-Mantel-Haenszel test and linear regression for binary and continuous variables, respectively, to test a linear association of health outcomes and risk factors with the levels of SES. In order to test for a linear trend, SES was coded as follows:

One if educational level was less than 10 years school education, 0.5 if equal to 10 years and 0 if more than 10 years school education. An undirected association was assessed by chi-square test. The bivariate unadjusted associations between respiratory health indicators and risk factors were assessed by logistic regression. A stepwise procedure of multiple logistic regressions with pre-specified steps was used to analyse alterations of the associations between respiratory health indicators and SES (model I) by consecutively additional inclusion of occupational exposure (model II), tobacco smoke (model III) and environmental air pollution exposure (model IV) into the regression models. The results of the regression analyses are presented as odds ratios with 95% confidence limits and their p-values. The alteration of the crude SES effect by additional allowance of other risk factors in the model was assessed by the percentage reduction in the 'excess odds', i.e. OR-1, due to adjustment for the other risk factors. This percentage change can be taken as a measure of explained SES effect by other risk factors. Odds ratios of the continuous risk factors PM₁₀ and NO₂ were calculated for an interquartile range increase of 7 μg/m³ and 16 μg/m³, respectively. The odds ratios presented for the SES factor indicate the risk ratios of the lowest versus the highest SES level. Interaction terms were included into the regression models to evaluate whether the risk factors modified the effect of the SES. Outcome variables which showed no association (p > 0.05) with SES were not analyzed. Likewise, risk factors, which showed no univariate association with SES, were not included in the analysis of alterations of the associations between respiratory health indicators and SES. If the association between SES and respiratory health was not homogeneous for different strata of the risk factors, the analysis was restricted to the stratum, where the association between SES and respiratory health indicator was highest.

The cohort has been previously described in detail [9, 20]. In brief, lung function measurements and complete information were obtained from 1251 women in the urban areas of the Ruhr district. In particular, information was available on school education of either the women or their husbands. The average age of the participants was 55 years (mean: 54.5, SD 0.6). Table 1 shows the distribution of SES expressed as educational levels. The prevalence of women with impaired respiratory health and a reduction in lung function according to their educational level are also presented in Table 1. The results show the differences in prevalence of respiratory impairment in women with different school education. Respiratory symptoms were more frequently reported in women with less than 10 years of school education. However, chronic bronchitis by a physician's diagnosis was more common in women with 10 years of school education. The percentage of women with impaired lung function (FVC and FEV₁ <80% of predicted) was higher in the group with less than 10 years of school education. A similar result can be seen for COPD. The prevalence of impaired respiratory health decreased with increased education. A linear trend was seen for all respiratory health indicators with the exception of chronic bronchitis by a physician's diagnosis. For chronic bronchitis, no significant association of any kind with the educational level could be assessed using chi-square test. We therefore excluded chronic bronchitis from the further analysis. The association between respiratory health indicators and SES was stronger for the indicators based on lung function measurements than for those based on questionnaire variables. The percentage of women with reduced lung-function who, however, did not report symptoms was lower in better educated women. This might be due to selective underreporting of these women. For instance, 81% of women with more than 10 years of education and reduced lung function (FVC < 80%) reported frequent cough in comparison to only 67% of women with less than 10 years of education.

The distributions of respiratory health risk factors in the study group are presented in Table 2. Unfavourable occupational exposure to dust, gases, vapours, wet conditions or extreme temperatures concerned 14.2% of the study group. Current smoking on entry to the study was in 17.5%, 36.6% were never smokers but did report environmental tobacco smoke (ETS) exposure, 45.9% were never smokers without environmental tobacco smoke, and 9.6% were ex-smokers. The mean distance to the nearest road with more than 10.000 cars per day was 519 m. 22.1% of all women lived in a distance of less than 100 m from a road with more than 10.000 cars a day (major road). The mean values for exposure to PM₁₀ and NO₂ were 49.4 μg/m³ and 49.2 μg/m³, respectively. Women with less than 10 years of school education had a higher prevalence of occupational exposure. Women with lower SES had a higher exposure to ETS and were more likely to be current smokers. Further, women with less than 10 years of school education lived closer to major roads than women with higher education, but traffic exposure was not linearly associated with SES. Long-term exposure to increased levels of PM₁₀ was significantly associated with lower school education. Women with less than 10 years of school education were exposed to higher levels of PM₁₀ compared to women with higher education. The prevalence of occupational exposure and current smoking status and the long-term average concentration of PM₁₀ were significantly increased with decreased level of education. NO₂ exposure did not exhibit any association with SES. We therefore excluded the risk factor NO₂ exposure from further analysis.

We tested all associations of risk factors with health outcomes in separate bivariate models (Table 3). Current smoking exhibited a significant adverse effect with all respiratory health indicators, whereas ETS did not. Likewise, occupational exposure was a significant respiratory health risk factor, but not significant for bronchial asthma. A reduction in FVC and FEV₁ (<80% predicted) was significantly associated with exposure of high levels of PM₁₀. An elevated prevalence of COPD and reduced FEV₁ (<80% predicted) with women exposed to high traffic was indicated. The parameter estimates were similar but slightly smaller compared to those in our previous publication [9], where we included women from urban and rural areas. As exposure of ETS was not a risk factor for respiratory impairment in this study group we excluded ETS from further statistical analysis.

Before assessing the contribution of respiratory health risk factors to the association between SES and respiratory health indicators, we checked whether the effect between educational level and respiratory health risk factors was homogeneous. For chronic cough with and without phlegm production a significant (p < 0.1) modification of the SES effect by the distance to a major road was observed. The association was stronger in better educated women and stronger for women living in a distance of more than 100 m to the nearest major road. Therefore for the symptoms of chronic cough with and without phlegm production, we only considered the stratum of low traffic exposure, i.e. women living in a distance of more than 100 m to the nearest major road, in the following step-wise regression analysis. Table 4 shows the results of the step-wise regression analysis of the association between respiratory health indicators and SES adjusted by additionally including other risk factors (occupational exposure, current smoking and air pollution exposure). These results indicate that a low level of education was associated with a reduction in lung function, i.e. an increased prevalence of FEV₁ or FVC less than 80% of the predicted value, as well as an increased prevalence of COPD, chronic cough with and without phlegm production, and bronchial asthma. The associations remained stable when adjusting for occupational exposure. However, the association was attenuated after additionally adjusting for smoking. After further adjustment for the environmental risk factors exposure of PM₁₀ and distance to major road an additional attenuation of the SES effect was observed for COPD, FEV₁ < 80% predicted and FVC <80% predicted. The explanatory measure of SES effects by other risk factors, the percentage reduction of excess odds ratio, showed a range between 5% to 10% attributable risk for the self-reported respiratory health indicators and 21% to 33% change for the measured lung function indicators. The smallest change was observed for bronchial asthma with current smoking showing the best explanation of the SES effect.