IS POTROOM ASTHMA DUE MORE TO SULPHUR DIOXIDE THAN FLUORIDE? – an inception cohort study in the Australian aluminiumindustry
Michael John Abramson, Geza Paul Benke, Jisheng Cui, Nicholas Hubert de Klerk, Anthony Del Monaco, Martine Dennekamp, Lin Fritschi, Arthur William Musk, Malcolm Ross Sim
Online Data Supplement
The population for this analysis included all new employees who commenced after 1 July 1995 at two aluminium smelters in Victoria, Australia. Smelter A included a small nearby coal mine, power station and rolling mill. Participants were recruited between 1995 and 2003 (Table 1). Employees were not eligible to take part if theyhad been previously employed by Alcoa at any of the eligible sites, were contractors or vacation students. The numbers and proportions of eligible employees tested at each site are given in Table E1. Whilst the median interval between successive interviews was between 300 and 400 days, a small proportion were conducted more than 500 days after the previous interview. Questionnaires were administered, spirometry and methacholine challenge tests were performed in that order at each assessment. Skin prick testing was only performed at the entry interview.
Exposure Assessment
All sampling of fluorides, SO2,inhalable and inhalable dust was collected by personal samplers. The main components of the inhalable dust were calcined coke (electrode plant), alumina (potrooms), cryolite and aluminium fluoride, with minor levels of other metal oxides and refractory materials. The inhalable dust included fluoride compounds, but theses were measured when monitoring for fluorides. Most of the exposure data were log-normally distributed for the different full shift tasks, however in the analysis we used the arithmetic mean which was appropriate for a linear dose response model.[1]
Exposures to contaminants of interest were firstly sampled in full shift tasks across the smelters. The proportion of time reported by the subjects in each of the tasks was then used to calculate each subject’s exposure in all tasks he/she worked at the smelter during the course of the study. A quantitative cumulative exposure was then calculated for each subject between interviews. We constructed a database called the Task Exposure Database (TED)[2] , together with a database called the Personal Exposure Matrix (PEM), which we used for calculation of each subject’s exposure to all contaminants of interest for the exposure metric.
Other Assessments
The questionnaire included items relating to cough, shortness of breath, wheeze and chest tightness from the British Medical Research Council questionnaire[3] and a detailed smoking history. A rolling seal spirometer (Graseby Anderson Model S401, Smyrna, GA, USA) was used in accordance with ATS guidelines.[4] Calibration and quality assurance procedures followed those of an accredited hospital lung function laboratory ( Measured lung function was compared with values predicted from the equations of Gore et al.[5] Methacholine was delivered by a dosimeter (Mefar MB3, Bovezzi, Italy) following the protocol used in the European Community Respiratory Health Survey.[6] The best FEV1 measurement meeting ATS reproducibility criteria[7] was recorded after each dose. If these were not satisfied, the average of technically satisfactory FEV1 measurements was recorded. Bronchial reactivity was also expressed on a continuous scale as Log Dose Response Slope.[8] Skin prick testing was performed in a standardized fashion[9] on the flexor surface of the forearm with extracts of four common aeroallergens (D. pteronyssinus, grass pollen mix, cat dander and Aspergillus mix).
Sample size and statistical analysis
Sample size for the inception cohort study was based upon preliminary cross-sectional data. To achieve a power of 80% to detect differences significant (p < 0.05) between each “exposed” group and the unexposed office group, it was estimated that 60 to 80 employees in each exposure group would enable the following differences to be detected: 25 ml/year decline in FEV1, 20% increase in BHR over 2 years and between 5% and 12% occurrence of symptoms over 2 years. Given that around a quarter of eligible employees were unexposed office workers, this was estimated to require the recruitment of 1,000 new employees, or about 500 at the Victorian operations.
Smoothed prevalence curves and their 95% confidence intervals were produced (eg. Figure 1), in which smoothing was achieved by using the moving average method with weights 0.4, 0.3 and 0.4, respectively. Three types of correlation structures were examined for the repeated measurements of each symptom. The first method assumed that the correlation coefficient of a symptom between any two interviews was constant (also called the exchangeable correlation). The second method assumed that the correlation coefficients were an exponential function of the lag between interview numbers (also called the autoregressive correlation). The third method assumed that the correlation coefficients were arbitrary numbers and they were estimated from the data (also called the unstructured correlation). Robust variance estimates were used in the analysis.[10] This produced valid standard errors even if the true correlation structure between repeated measurements was not as hypothesized by the specified working correlation structure.
Trends in odds ratios between the tertiles of exposure and the unexposed (reference) category were assessed with the Mantel-extension test.[11] Because of skewed distributions of contaminants, Spearman rank correlations were estimated.
Additional Figure Legends
Figure E1. Smoothed incidence curves (and 95% confidence intervals) for wheeze categorized by gender
Figure E2. Smoothed incidence curves(and 95% confidence intervals) for chest tightness categorized by gender
Table E1.: Number and proportions of interviews conducted by site
Variable / Smelter A (n = 985) / Smelter B (n = 675)Number / % / Number / %
Year of interview
1995 / 30 / 3.0 / 5 / 0.7
1996 / 55 / 5.6 / 35 / 5.2
1997 / 51 / 5.2 / 60 / 8.9
1998 / 70 / 7.1 / 63 / 9.3
1999 / 97 / 9.8 / 74 / 11.0
2000 / 126 / 12.8 / 105 / 15.6
2001 / 155 / 15.7 / 107 / 15.9
2002 / 195 / 19.8 / 117 / 17.3
2003 / 206 / 20.9 / 109 / 16.1
Interval between consecutive interviews (days)†
< 300 / 90 / 12.9 / 62 / 12.0
300 – 400 / 406 / 58.2 / 301 / 58.3
400 – 500 / 165 / 23.6 / 111 / 21.5
≥ 500 / 37 / 5.3 / 42 / 8.1
Sequence of interviews
Entry interview / 287 / 29.1 / 159 / 23.6
1st follow-up interview / 223 / 23.6 / 137 / 20.3
2nd follow-up interview / 155 / 15.7 / 112 / 16.6
3rd follow-up interview / 120 / 12.2 / 94 / 13.9
4th follow-up interview / 84 / 8.5 / 68 / 10.1
5th follow-up interview / 50 / 5.1 / 46 / 6.8
6th follow-up interview / 31 / 3.1 / 34 / 5.0
7th follow-up interview / 22 / 2.2 / 20 / 3.0
8th follow-up interview / 13 / 1.3 / 5 / 0.7
Number of interviews per participant / (n = 287) / (n = 159)
1 / 62 / 21.6 / 17 / 10.7
2 / 67 / 23.3 / 27 / 17.0
3 / 37 / 12.9 / 21 / 13.2
4 / 35 / 12.2 / 29 / 18.2
5 / 38 / 13.2 / 18 / 11.3
6 / 20 / 7.0 / 17 / 10.7
7 / 7 / 2.4 / 12 / 7.5
8 / 10 / 3.5 / 13 / 8.2
9 / 11 / 3.8 / 5 / 3.1
† There were 446 participants at the entry interview.
Table E2: Number of interviews and range (mg/m3-year) in each tertile of cumulative exposure between interviews to each contaminant
Contaminant / Range / N†BSF
Unexposed / 0 / 848
Tertile 1 / (0, 0.0033] / 263
Tertile 2 / (0.0033, 0.0106] / 263
Tertile 3 / > 0.0106 / 262
Fluoride
Unexposed / 0 / 515
Tertile 1 / (0, 0.0271] / 374
Tertile 2 / (0.0271, 0.4412] / 374
Tertile 3 / > 0.4412 / 373
Inhalable dust
Unexposed / 0 / 460
Tertile 1 / (0, 0.6247] / 392
Tertile 2 / (0.6247, 3.1524] / 392
Tertile 3 / > 3.1524 / 392
SO2
Unexposed / 0 / 712
Tertile 1 / (0, 0.1138] / 308
Tertile 2 / (0.1138, 0.2923] / 308
Tertile 3 / > 0.2923 / 308
Oil mist
Unexposed / 0 / 807
Tertile 1 / (0, 0.0125] / 52
Tertile 2 / (0.0125, 0.0861] / 51
Tertile 3 / > 0.0861 / 51
† There were 24 participants whose exposure levels were missing.
Abbreviations: BSF = Benzene Soluble Fraction, SO2 = Sulphur Dioxide
Table E3. Odds Ratios (and 95% Confidence Intervals) for the co-variates and potential confounders of associations between trends in SO2 and wheeze
Males (n=1112) / Females (n=440)OR (95% Confidence Intervals)
Age / 1.002 (0.978, 1.026) / 0.997 (0.952, 1.043)
Non Smoker / 1.0 / 1
Current Smoker / 3.53 (2.35, 5.30)** / 1.68 (0.89, 3.17)
Ex-Smoker / 1.63 (1.02, 2.62)* / 0.53 (0.24, 1.18)
Site (Smelter B) / 1.72 (1.09, 2.74)* / 1.30 (0.63, 2.68)
Atopy / 0.99 (0.66, 1.49) / 3.68 (1.88, 7.24)**
* p < 0.05** p < 0.01
Table E4. Comparison of single and two exposure models for wheeze, stratified by gender † Odds ratios present odds of wheeze for a one tertile increase in exposure to each contaminant between interviews.
Contaminant / Male (n=290) / Female (n=102)OR / 95% CI / p-value / OR / 95% CI / p-value
Single exposure models
Fluoride / 1.20 / (1.01, 1.42) / 0.035 / 1.16 / (0.86, 1.58) / 0.33
SO2 / 1.33 / (1.15, 1.54) / <0.001 / 1.26 / (0.94, 1.69) / 0.12
Inhalable dust / 1.35 / (1.16, 1.58) / <0.001 / 1.14 / (0.83, 1.57) / 0.42
Two exposure models
Fluoride / 0.78 / (0.56, 1.09) / 0.14 / 0.95 / (0.64, 1.40) / 0.78
SO2 / 1.61 / (1.20, 2.16) / 0.001 / 1.31 / (0.90, 1.90) / 0.16
Inhalable dust / 1.19 / (0.93, 1.52) / 0.17 / 0.96 / (0.66, 1.4) / 0.84
SO2 / 1.17 / (0.93, 1.48) / 0.19 / 1.29 / (0.93, 1.79) / 0.12
Inhalable dust / 1.65 / (1.26, 2.15) / <0.001 / 1.04 / (0.66, 1.63) / 0.88
Fluoride / 0.78 / (0.58, 1.04) / 0.09 / 1.13 / (0.75, 1.69) / 0.57
† Adjusted for age, smoking status, smelter site, and atopy
Table E5: Associations between lung function†, bronchial reactivity* and cumulative exposures between interviews
FEV1(ml) ‡ / FVC(ml) ‡ / LogDRS**Contaminant / Regression coefficient (95% Confidence Intervals)
BSF
Tertile 1 / 82.5 (11.9, 153.1) / 143.8 (69.7, 217.8) / -178.7 (-373.4, 16.0)
Tertile 2 / 35.1 (-33.8, 104.0) / 121.9 (50.0, 193.9) / -44.6 (-243.7, 154.5)
Tertile 3 / 67.9 (-10.9, 146.6) / 135.1 (55.4, 214.9) / -19.6 (-227.7, 188.6)
Trend / 1.4 (-16.6, 19.4)
P = 0.88 / 15.3 (-2.7, 33.3)
P = 0.10 / 28.2 (-29.1, 85.5)
P = 0.34
Fluoride
Tertile 1 / 48.5 (-2.0, 99.0) / 65.7 (11.6, 119.7) / -130.0 (-359.6, 99.7)
Tertile 2 / 53.3 (-0.07, 106.6) / 63.4 (3.6, 123.2) / -54.2 (-273.6, 165.1)
Tertile 3 / 12.2 (-45.8, 70.2) / 52.2 (-9.8, 114.2) / 18.5 (-200.8, 237.8)
Trend / -6.7 (-24.8, 11.4)
P = 0.47 / 6.8 (-12.4, 26.1)
P = 0.49 / 23.9 (-39.6, 87.4)
P = 0.46
Inspirable dust
Tertile 1 / 58.9 (4.9, 113.0) / 102.2 (47.6, 156.8) / -116.5 (-318.1, 85.2)
Tertile 2 / 56.1 (4.0, 108.2) / 99.2 (46.2, 152.2) / -78.3 (-280.8, 124.2)
Tertile 3 / 51.1 (-5.2, 107.3) / 100.6 (46.4, 154.8) / -4.7 (-210.4, 201.0)
Trend / 3.1 (-11.8, 18)
P = 0.68 / 11.2 (-3.9, 26.2)
P = 0.15 / 24.2 (-25.9, 74.3)
P = 0.34
SO2
Tertile 1 / 71.4 (6.6, 136.1) / 132.1 (65.7, 198.5) / -40.9 (-237.0, 155.1)
Tertile 2 / 65.9 (-0.4, 132.3) / 131.2 (64.5, 198.0) / 23.5 (-182.4, 229.4)
Tertile 3 / 33.6 (-31.3, 98.6) / 131.4 (64.9, 197.9) / 82.7 (-120.8, 286.1)
Trend / -6.1 (-20.9, 8.6)
P = 0.42 / 17.1 (1.3, 32.9)
P = 0.03 / 40.1 (-15.4, 95.6)
P = 0.16
Oil mist
Tertile 1 / 1.7 (-49.4, 52.7) / -28.8 (-88.2, 30.6) / -222.7 (-545.9, 100.6)
Tertile 2 / -8.4 (-60.8, 44.1) / -30.6 (-88.6, 27.5) / -199.0 (-423.3, 25.2)
Tertile 3 / -16.4 (-81.4, 48.6) / -61.7 (-139.7, 16.3) / -459.2 (-775.1, -143.3)
Trend / -5.2 (-26, 15.6)
P = 0.63 / -19 (-43.3, 5.3)
P = 0.13 / -136.8 (-229.0, -44.5)
P = 0.004
† Number of interviews used in analysis = 920 for oil mist and = 1547 for other exposures
* Number of interviews used in analysis = 780 for oil mist and = 1308 for other exposures
‡ Adjusted for age, gender, age and gender interaction, smoking status, smelter site, height at entry interview, and interaction between gender and smoking status.
** Adjusted for age, gender, age and gender interaction, smoking status, smelter site, atopy, and interaction between gender and smoking status.
Table E6. Comparison of single and two exposure models for FEV1/FVC, stratified by gender † Coefficients present the change in FEV1/FVC(‰) for a one tertile increase in exposure to each contaminant between interviews.
Contaminant / Male (n=305) / Female (n=111)β (‰) / 95% CI / p-value / β (‰) / 95% CI / p-value
Single exposure models
Fluoride / -4.04 / (-6.90, -1.18) / 0.006 / -3.88 / (-9.66, 1.89) / 0.19
SO2 / -5.54 / (-7.91, -3.18) / <0.001 / -3.67 / (-7.92, 0.59) / 0.091
Inhalable dust / -2.73 / (-5.03, -0.43) / 0.02 / -0.32 / (-4.40, 3.76) / 0.88
Two exposure models
Fluoride / 2.09 / (-1.72, 5.90) / 0.28 / -1.99 / (-9.12, 5.14) / 0.58
SO2 / -6.83 / (-9.97, -3.69) / <0.001 / -2.52 / (-7.60, 2.56) / 0.33
Inhalable dust / 2.62 / (-0.27, 5.50) / 0.075 / 2.96 / (-1.79, 7.71) / 0.22
SO2 / -7.52 / (-10.58, -4.45) / <0.001 / -5.57 / (-10.5, -0.60) / 0.028
Inhalable dust / -0.51 / (-3.73, 2.71) / 0.76 / 4.67 / (-1.97, 11.3) / 0.17
Fluoride / -3.54 / (-7.62, 0.54) / 0.089 / -8.02 / (-17.23, 1.18) / 0.088
β Regression coefficient
† Adjusted for age, smoking status, smelter site, and height at the entry interview
Table E7: Association between cumulative exposure to gaseous fluoride (HF) andwheeze stratified by gender andadjusted for age, smoking, smelter site, and atopy
Cumulative / Male (n = 1114 ) / Female (n =440 )exposure to HF / OR / 95% CI / OR / 95% CI
Tertile1 / 0.75 / (0.45, 1.26) / 0.84 / (0.42, 1.69)
Tertile2 / 0.95 / (0.58, 1.56) / 1.00 / (0.47, 2.13)
Tertile3 / 1.52 / (0.94, 2.46) / 1.51 / (0.58, 3.93)
Trend / 1.19 / (1.01, 1.41)
P=0.035 / 1.14 / (0.84, 1.54)
P=0.40
TableE8: Association between cumulative exposure to gaseous fluoride (HF) and FEV1/FVC stratified by genderand adjusted for age, smoking, smelter site and height
Cumulative / Male (n = 1123 ) / Female (n =424 )exposure to HF / β / 95% CI / β / 95% CI
Tertile1 / -2.58 / (-12.61, 7.45) / -8.47 / (-18.42, 1.47)
Tertile2 / -2.05 / (-11.08, 6.98) / -13.26 / (-25.64, 0.88)
Tertile3 / -11.61 / (-21.54, -1.69) / -11.74 / (-25.75, 2.26)
Trend / -4.48 / (-7.24, -1.73)
P=0.001 / -3.28 / (-7.86, 1.29)
P=0.16
TableE9: Association between cumulative exposure to gaseous fluoride (HF) and consecutive change in FEV1stratified by gender and adjusted for age, smoking, smelter site and height
Cumulative exposure / Male (n = 818 ) / Female (n =313)to HF / β / 95% CI / β / 95% CI
Tertile1 / -17.67 / (-48.09, 12.74) / -10.59 / (-35.95, 14.77)
Tertile2 / -14.98 / (-47.24, 17.29) / -38.57 / (-63.81, -13.32)
Tertile3 / -26.85 / (-50.11, -3.58) / -40.41 / (-76.92, -3.91)
Trend / -8.23 / (-15.56, -0.90)
P=0.034 / -7.97 / (-15.20, -0.74)
P=0.031
Table E10: Association between cumulative exposure to gaseous fluoride (HF) and BHR stratified by gender andadjusted for age, smoking, smelter site and atopy
Cumulative / Male (n = 984 ) / Female (n =336 )exposure to HF / OR / 95% CI / OR / 95% CI
Tertile1 / 1.16 / (0.65, 2.08) / 0.89 / (0.44, 1.80)
Tertile2 / 1.43 / (0.84, 2.46) / 1.20 / (0.52, 2.73)
Tertile3 / 1.98 / (1.13, 3.47) / 1.20 / (0.46, 3.14)
Trend / 1.27 / (1.07, 1.51)
P=0.006 / 1.08 / (0.79, 1.48)
P=0.62
Table E11. Odds Ratios (and 95% Confidence Intervals) for the co-variates and potential confounders of associations between trend in SO2 and BHR
Males (n=984) / Females (n=336)OR (95% Confidence Intervals)
Age / 0.977 (0.952, 1.003) / 0.98 (0.94, 1.02)
Non Smoker / 1.0 / 1
Current Smoker / 1.43 (0.98, 2.09) / 2.88 (1.33, 6.25)**
Ex-Smoker / 1.07 (0.70, 1.62) / 1.12 (0.55, 2.25)
Site (Smelter B) / 1.21 (0.74, 2.00) / 0.83 (0.37, 1.84)
Atopy / 2.84 (1.76, 4.58)** / 5.87 (2.80, 12.27)**
* p < 0.05** p < 0.01
Table E12. Correlationsbetween cumulative exposuresto contaminants
Fluoride / Inhalable Dust / SO2 / Oil MistBSF / 0.77** / 0.76** / 0.84** / -0.29**
Fluoride / 0.92** / 0.88** / -0.34**
Inhalable Dust / 0.81** / -0.32**
SO2 / -0.33**
** p < 0.01
References
1Seixas NS, Robins TG, Moulton LH. The use of geometric and arithmetic mean exposures in occupational epidemiology. Am J Ind Med 1988;14:465-77.
2Benke GP, Sim MR, Fritschi L, et al. A task exposure database for use in the alumina and primary aluminium industry. Appl Occup Environ Hyg 2001;16:149-53.
3Medical Research Council’s Committee on Environmental and Occupational Health. Questionnaire on respiratory symptoms. London, UK: MRC 1986.
4American Thoracic Society. Standardization of spirometry. 1994 update. American Journal of Respiratory & Critical Care Medicine 1995;152:1107-36.
5Gore CJ, Crockett AJ, Pederson DG, et al. Spirometric standards for healthy adult lifetime nonsmokers in Australia. Eur Respir J 1995;8:773-82.
6Chinn S, Burney P, Jarvis D, et al. Variation in bronchial responsiveness in the European Community Respiratory Health Survey (ECRHS). Eur Respir J 1997;10:2495-501.
7American Thoracic Society. ATS statement--Snowbird workshop on standardization of spirometry. Am Rev Respir Dis 1979;119:831-8.
8O'Connor G, Sparrow D, Taylor D, et al. Analysis of dose-response curves to methacholine. An approach suitable for population studies. Am Rev Respir Dis 1987;136:1412-7.
9Nelson HS. Diagnostic procedures in allergy: allergy skin testing. Ann Allergy 1983;51:411-7.
10Huber PJ. The behavior of maximum likelihood estimates under nonstandard conditions. Proceedings of the Fifth Berkeley Symposium on Mathematical Statistics and Probability. Berkeley, CA: University of California Press 1967:221-33.
11Rothman KJ, Greenland S, Lash T. Modern Epidemiology. New York: Lippincott–Raven 2008.
1