Document GLG2XdqG2LN6K224Nqby0Eqr

Occupational and Environmental Medicine 1997;54:27-31 27 Relation between decline in FEV1 and exposure to dust and tobacco smoke in aluminium potroom workers Vidar S0yseth, Jacob Boe, Johny Kongerud Hydro Aluminium Ardal, N-5870 0vre Ardal, Norway V Soyseth Department of Thoracic Medicine, Rikshospitalet, University of Oslo, Norway V Soyseth J Boe J Kongerud Correspondence to: Dr Vidar Soyseth, Lillehammer Hospital, Medical Department, N-2600 Lillehammer, Norway. Accepted 29 July 1996 Abstract Objectives-To investigate the relation between pulmonary function and occupational exposure in aluminium pot operators. Methods-2795 observations were obtained in 630 workers over six years of follow up. An autoregressive method of analysis was used. Results-After adjustment for FEV, in the three previous years, the effect of smoking v no smoking on FEV, was -43-1 ml, 95% confidence interval (95% CI) - 72X3 to - 13-9. Similarly, an increase in the exposure to particulates by 1 mg/m3 corresponded to a decrease in FEV, of - 11 9 ml, 95% CI - 19 9 to -39. Age was a sig- nificant predictor of both FEV, and FVC. Conclusion-Exposure to particulates in aluminium potrooms seems to increase the decline in FEV,, thereby increasing the risk of development of chronic obstructive lung disease in pot operators. (Occup Environ Med 1997;54:27-31) Keywords: aluminium; autoregressive time series; pulmonary function; occupational exposure; tobacco smoke It has been found that forced expiratory volume in one second (FEVy) is a good predictor of mortality due to chronic obstructive pulmonary disease (COPD).'2 Hence, in the prevention of COPD it is important to identify determinants of the annual decline of FEV,. It is generally well, accepted that smoking is associated with increased annual decline of FEV,.' The relation between the development of COPD and occupational exposure to airborne pollutant is less well documented although the evidence for such an association is increasing.45 Aluminium is produced by electrolysis of alumina (A1203). Alumina is a powder with a median mass diameter of 100 um.6 The range of the respiratory fraction of the powder is < 2%.6 Nevertheless, about 50% of the particles in the potroom atmosphere are < 6-15 pm.6 As well as alumina, the work atmosphere is polluted with fluoride dust and carbon particles from the anode. Gases that act as airway irritants, such as hydrogen fluoride and sulphur dioxide, are also emitted from the pots. We have previously shown that these exposures increase bronchial responsiveness7, and the incidence of respiratory symptoms,8 and that the prevalence of airways obstruction increases with the duration of exposure in the potrooms.9 An increased mortality due to COPD in aluminium potroom workers has recently been reported by R0nneberg."0 The relation between occupational exposures and the development of airway obstruction is, however, lacking." We have performed repeated spirometries annually in potroom workers for six years at an aluminium smelter in western Norway. The objective of the study was to investigate the chronic effect of exposure to potroom pollutants on the development of lung function in pot operators. Materials and methods POPUIATION AND THE PLANT The study was conducted at Hydro Aluminium Plant in Ardal in 1986-92. During this period the production of aluminium has increased from 170 000 to 190 000 tonnes, whereas the emission of fluoride and sulphur dioxide to the environment has decreased from 40 kg/h to 15 kg/h and from 270 kg/h to 50 kg/h, respectively. These improvements have been achieved by introduction of new technology and better pot operation routines. The plant has three potroom departments, two prebake and one S0derberg. All employees working in these potrooms in September 1986 or later were invited to participate in the study. Those starting work in the potrooms during the follow up were also recruited to the cohort. The workers were examined annually between 1 September and 1 November. The attendance rate was 95%98% at each of these surveys. New employees attending the potrooms were examined before the first day at work. Those who left the potrooms were examined during the last 14 days before they finished work. Workers who temporarily left the potrooms were examined before leaving work and after returning to the potrooms. PULMONARY FUNCTION AND QUESTIONNAIRE Pulmonary function was measured with a dry bellow spirometer Jones Pulmonaire, Illinois). It was calibrated monthly with a 11 syringe. The subject performed at least three expira- tory manoeuvres; the two best should not differ by more than 5% or 100 ml, whichever was the largest.'2 The results were converted to body temperature, pressure, and saturation (BTPS). Three technicians (nurses) were trained to perform the test by one of us (JK) 28 Soyseth, Boe, Kongerud before the start of the study and annually thereafter. Reference values were taken from an asymptomatic Norwegian urban population.13 Information on smoking habit was obtained from a validated questionnaire.'4 OCCUPATIONAL EXPOSURE Since 1986 measurements of exposure to particulates and fluoride have been performed annually with personal samplers.9 The work in the potrooms is divided into several job categories. Each year during this follow up, workers have been randomly selected to wear such samplers for eight hour shifts; 874 measurements were taken. The annual exposure in each job category was expressed as the geometric mean of these measurements. Information on the job category in each operator who participated in the study was obtained from a questionnaire'4 at each examination. It was assumed that the workers were exposed to the estimated exposure in their job category in the time interval between the surveys. The availability of information about deviation from this assumption was limited. Consequently, we were not able to adjust exposure for absence from work, such as sickness absence and military service. Such deviations were likely to cause irregular time spacing between the examinations. STATISTICAL ANALYSES In this longitudinal study we chose to use an autoregressive method that has been described by Rosner and coworkers.'5 Other alternative methods of longitudinal analyses could also be considered, such as the random effects model'6 and generalised estimation equations.'7 These methods enable use of all the available data. The advantage of the method chosen is its simplicity. The data can be analysed with software that offers ordinary multiple regression methods, and no complex algorithms for non-linear calculations are required. The limitation of this approach is the assumption that the observations must be equally spaced and that the number of excluded subjects increases as the order of the autoregression model increases. The problem of excluding subjects is of minor importance as the objective of the study was to investigate the chronic effects of exposure on lung function-that is, in the subjects who stayed at work. Briefly, the outcome (FEVy or FVC) and the covariates were entered into an ordinary multiple regression model, adjusting for one or more of the previous values of the outcome. The appendix explains details about the model and how the number of previous values of the outcome were determined. The method is based on the assumption that observations are equally spaced over time.'5 Thus, only those observations that were separated by between 10 and 14 months were included in the analyses-that is, only workers who had worked in the potrooms for at least 10 months were included. The analyses were conducted in two steps. Firstly, the number of previous values of the outcome was found (appendix). Next, height, sex, age, total fluoride, total particulates, and smoking habits were included as covariates. Two indices of smoking were used; nonsmoker and current smoker, and the amount of tobacco smoked (g/week) was used as a continuous variable. Finally, the model was reduced by backward elimination by removal of covariates that did not contribute significantly to the model, provided that this removal caused < 10% change in the remaining coeffi- cients. 18 Results In all, 2795 spirometries were carried out during the follow up in 630 workers. Table 1 shows the characteristics of the workers at inclusion to the cohort. It shows that 58% of the workers were excluded from the final analyses, covering 32% of the total follow up time. Hence, the final analyses covered 68% of the observation time. Except for annual decline in FEV1 and age the differences between the subjects were neglectable. Table 2 shows the number of examinations during the follow up. In the total workforce, a mean (SD) time lag was 12 (2) months between two examinations over 1500 observa- Table 1 Personal characteristics at baseline of the workers who were included in the cohort and those who were available in the final analyses Not included in the final analyses < 3 yfollow up n = 251 > 3 y follow up n = 114 Final analyses n = 265 Sex (F (%)) Current smoking habits (%) Age (y) FEV: ml Predicted (%) FVC: ml Predicted (%) Annual decline in FEV ml/y Duration of employment (months) Ended employment (%) Particulates (mg/M3) Follow up time (y) 32 (12-8) 131 (52 2) 25-1 (19-2 to 57 5) 4100 (2940 to 5090) 88-9 (73.5 to 104-0) 5140 (3860 to 6200) 92-7 (79 3 to 108-4) -35-6 (- 162-2 to 152-9)* 0 (0 to 315) 213 (84-9) 2-17 (0 to 6-64) 209-5 21 (18-4) 66 (57 9) 26-0 (18-8 to 48-1) 4100 (2960 to 5360) 88-8 (71-9 to 107-7) 5110 (3800 to 6410) 92-4 (79-3 to 107-4) - 14-4 (-75-7 to 73 6) 3 (0 to 277) 35 (30 7) 2-35 (1-24 to 6 64) 508-2 29 (10-9) 155 (58 5) 33-5 (20-1 to 51-8) 4050 (3080 to 5060) 90 5 (77 9 to 105-2) 5130 (399 to 6330) 94-8 (82-2 to 110-7) -41-8 (-90 3 to 22 7) 70 (0 to 329) 31 (11-7) 3-19 (1-70 to 6-64) 1493 Continuous variables are medians (10th to 90th percentiles). *Estimated in 72 workers who had three or more recordings. Relation between decline in FEVJ and exposure to dust and tobacco smoke in aluminium potroom workers 29 Table 2 Number of spirometries that were carried out during the follow up Number of examinations Year offollow up 1986 1987 1988 1989 1990 1991 1992 Sum 1 371 83 81 41 18 16 20 630 2 1 298 100 65 38 21 8 531 3 6 243 96 47 37 19 448 4 2 5 221 77 57 31 393 5 2 14 180 90 51 337 6 4 8 164 93 269 7 4 6 161 171 8 5 9 14 9 11 2 Sum 372 389 431 441 372 397 393 2975 tions of 478 workers. The mean (SD) time between two successive examinations in these workers was 12-0 (0-6) months. The exposure to fluorides and particulates decreased during the follow up (table 3). The decline in exposure seemed to be greatest between 1987 and 1989, when manual refill of alumina was replaced by automatic refilling of the prebake pots. The tobacco consumption was, however, nearly unchanged during follow up (table 3). In the first step of the analyses, the order of the autoregressive model was settled among the 478 workers in whom the time between two consecutive examinations was within 12 (2) months (table 4). It was found that the three last measurements of the outcome were significantly related to the dependent variable. Because of this the number of observations was reduced to 658 from 265 workers available for the final analyses (tables 1 and 4). Despite the large reduction of data available for analysis, the difference between the original population and the final population did not differ in the characteristics at inclusion (table 1). The final analyses started with the full model (AR3) including height, sex, total fluoride, age, total particulates, and tobacco smoke as covariates. During the backward model reduction, height, sex, and total fluoride did not contribute significantly to the model, and removal caused only minor changes to the coefficients of the remaining variables when FEVI was used as the depen- dent variable. When smoking (dichotomised) was replaced by the amount of tobacco smoked (g/week), the coefficient was estimated to be -0 39 ml/g/week (SEM 0-18, P = 0 03). Table 5 shows the results of the final analysis of FEVI. Age, current smoking, and exposure to particulates were the significant covariates. After adjustment for three previous measurements of FEV1, a significant effect of age on FEVy was found. If FEVy is related lin- early with age, the association between them should have the same magnitude independent of age when previous FEVI is taken into account. It therefore seems that the decline in FEV, accelerates with age. Also, the age effect expresses the association between the outcome and age conditionally on previous levels of the outcome (that are themselves dependent on age). Thus, the absolute age effect on FEVy is not estimated with this method. The association between exposure to partic- ulates and FEVI was investigated in non- smokers separately. An increase in exposure to particulates by 1 mg/M3 corresponded to a decline in FEV, of - 13A4 ml (SEM 5.7, P = 002) in non-smokers. Modification of the effect of exposure to particulates by smoking was investigated by adding a product term between these two covariates. This product term was not significant (P > 0-05), indicating that effect of particulates was not greater in smokers than in non-smokers. The analyses of FVC as the dependent variable showed no significant association with smoking or the occupational exposure indices (table 5). It was, however, significantly related to age. Finally, we have used the results from table 5 to predict FEV, in a worker for different levels of exposure to particulates and smoking habits (figure). The models predict FEVI in four workers who all have the same baseline FEVy (4 1) at 30 years of age. After having reached 62 years, the difference in FEV, between a smoker who was exposed to 5 mg/mi (the hygienic threshold level in the Norwegian aluminium industry) and a non-smoker who has had a negligible exposure to particulates, is 1518 ml. It is also indicated that the effect of exposure to particulates on FEVI is of the same size as the effect of smoking in this cohort (75 g/week). The decline in FEV1 seems to accelerate with increasing age. Table 3 Exposure to totalfluorides (mg/m3), particlulates (mgmr3), and smoking habits during the follow up Year offollow up Type ofexposure 1986 1987 1988 1989 1990 Total fluoride: Median 1Oth to 90th percentile Total particulate: Median 10th to 90th percentile Smoking: Amount (mean (SD), g/week) Prev. (%) 0-72 0-37 to 1-16 3-19 1-70 to 6-64 74 (33) 55-8 0-52 0-34 to 1-04 3-05 1-64 to 4 07 75 (35) 58-0 0-37 0-28 to 0 70 3-43 1-63 to 4-36 78 (38) 59-2 0-38 0-26 to 0-61 1-31 1 01 to 2-17 77 (38) 57-4 0r35 0-14 to 0-48 1-56 0-68 to 1-87 73 (36) 59-4 1991 0.24 0-12 to 0-38 0-99 0-60 to 1-67 74 (38) 58-1 1992 0-36 0-13 to 0 40 1-47 1-13 to 2-10 75 (39) 62-7 30 S0yseth, Boe, Kongenrd Table 4 Model (equation 2, appendix) avith no covariates classification of exposure in the long term Ln r Model Parameter parameters estimate SEM workers who were periodically absent from P value work is likely to be incorrect causing misclassi- 1 478 1500 ca 7y 2 328 1002 I 71 3 265 72 658 ae 71 72 4 210 73 370 at 71 72 73 Y4 76-1 0-971 18653 0-338 -91 3 0-527 0-292 0184 -97-1 0545 0-220 0-181 0-055 28-9 0 009 fication of exposure, thereby distorting the 0 007 < 0 001 results. The finding that they actually had a 3029 <0 001 lower decline in FEV, supports the association 0-029 < 0 001 between exposure and outcome. As the associ- 40 8 0-042 0-025 < 0001 ation between FEVI and particulates was 0 044 < 0 001 found in non-smokers, the results seem not to 0-036 < 0001 be confounded by smoking. 53-8 0-07 0055 <0001 There are, however, some problems with 0-064 0-056 < < 0001 the interpretation of the coefficients. They 0-048 0-25 express the effect of a covariate conditionally L = the order of the autoregressive model, n - number of individuals, r = number of rows in the on the three previous measures of the out- design matrix to fit specific models. come. Thus, they do not express the annual change of the outcome directly. Rosner and Table 5 Resultfrom the autoregressive model ofFEV, and FVC as the dependent coworkers offer a formula for calculation of variables annual change in cases of first order regres- Independent variable Intercept Particulates (mg/M3) Current smoking FEVI (ml) Coefficient 176-0 - 11-9 SEM 72-2 4-1 P value 0 015 0-004 FVC (ml) sion-that is, after adjustment for the previous Coefficient SEM P value value of the outcome.1I In the case of third -271L0 308 order regression this task is much more com0-38 plex. Therefore we chose to use the estimated - 81 5-7 0-15 coefficients to predict lung function for differ- Yesvno Age (y) -43-1 - 2-75 149 0-79 0001 -3-80 .060 <0001 ent alternatives of smoking and occupational exposure. As the characteristics of the total work force and the subjects included in the final model Prediction lines ofFEV, in Non-smokers were similar it seems that the results should be four workers with different exposures to articulates and smoking habits. All had the same baseline FEV, at 30 years of age. --N-SSmmioSokmkmoeeerrorss,rrprrrtrcruplrattreitci5ucmlugal/taietnse3asrhrmr3rrrrrvvpmoaalttlriidoodofnfmonrdsreppfcoortrteaooaspteeelsreaaatstothoes1r0pswrwmeohocnoihtshihavosvn.ewoLofoorstkkshedeoficioneffofhri-- 4000 % cients. As significant results were obtained, we 3500 - 3000 E 2500 2000- U- 1500 - \ -.. regard this as a minor problem. As the final analysis included only those who had three previous spirometries or more, the workers must have worked in the potrooms for three years or more. Thus, the results are not applicable to the short term 1000 _ effect of these exposures on FEV,. Never- 500- 030_ 35 40 45 50 55 Age (y) 60 65 I 70 theless, the development of COPD needs sev- eral years. Therefore, we think that the current analyses are relevant in this context. Furthermore, it seems to us that previous levels of FEVy are confounded by previous exposure, Discuss ion and this effect has to be taken into account in In this 1Longitudinal study we have found that the estimation of the current effect. However, the declline in FEV1 is increased in smokers we have data on operators with short term compare-d with non-smokers, and that it exposure, and we plan to make a separate increases with the amount of tobacco smoked. analysis of these data. Furthermore, it increases with occupational In previous studies we have found that respi- exposur(e to particulates. ratory symptoms and bronchial responsiveness The choice of analytical methodology might were associated with exposure to fluoride but be questioned, as only one quarter of the origi- not the exposure to particulates.78 Therefore it nal obs4 ervations were available in the final seems that the present results deviate from our analyses . This number might have been previous results. This inconsistency has at slightly:increased if a method that allows for least two explanations. Firstly, both bronchial unequally spaced observations have been responsiveness and respiratory symptoms are used.19 ]However, inclusion of more workers probably more related to reversible obstruc- with unequally spaced examinations during tion than the development of COPD. follow uip might have decreased the validity: Although there is some overlap in the mor- accordingig to the protocol, the operators phology of these entities, they represent differ- should imeet to do the examinations during ent diseases, and they may be associated with Septemtber and October-that is, 12 (2) different exposures. Next, exposure to particu- months apart. Those who were observed out- lates and fluoride (and exposure to other air- side this range were likely to have been unex- way irritants) are closely correlated. Thus, the posed to potroom fumes in the period between capability of the statistical models to differen- the obse!rvations. Although about 60% of the tiate between them is limited. In a follow up of workers were excluded from the final analyses, a cross sectional study of aluminium pot oper- nearly 75% of the follow up time of the long ators, Chan-Yeung and coworkers were not term enaployees was included. Furthermore, able to show any increased decline in FEV1 Relation between decline in FEVI and exposure to dust and tobacco smoke in aluminium potroom workers 31 compared with an unexposed control group between the original and follow up surveys.20 It should be noted that about 50% of the work force on the original survey left the industry before the follow up. It seems likely that workers who develop respiratory impairment are more prone to leave the industry than those who remain healthy. Hence, the negative result may be explained by a selective loss of follow ups. From a preventive point of view, our results have several implications. Firstly, a significant effect of dust exposure on the development on FEV1 was found. Therefore efforts should be made to reduce the exposure to particulates. Several alternatives should be considered, such as increase of particle diameter, decrease of contamination of the work atmosphere, and improvement of airway protection. Routine surveillance of the decline in FEV1 in each worker, and removal of workers with increased decline from exposure as well as selective intervention against smoking, should also be considered. Finally, the Norwegian compensation legislation has considerable consequences for the industry. It states that in a subject with a disease that might be caused by any occupational exposure, the contribution of lifestyle exposure on the disease should be ignored. Thus, the aluminium industry must be prepared to compensate for the smoking habits in workers with COPD, as smokers are more likely to develop the disease than non-smokers. It is therefore profitable to prohibit smoking in pot operators, at least at work. This is a serious problem, as more than 50% of the workforce were smokers, and the fact that neither the prevalence of smoking or the amount of tobacco smoked has decreased during the past years. In conclusion, FEV1 is negatively related to exposure to tobacco smoke and occupational exposure in aluminium pot operators. Great efforts should be made to decrease the occupational exposures as well as tobacco consumption. We thank Professor 0 Aalen for his statistical comments on the manuscript, the technicians E Jevnaker, K Moen, and T Nes who performed the spirometries and the interviews, and finally, all the workers who participated in the study. The study was supported by grants from the Norwegian Aluminium Sectretariat for Health, Environment, and Safety. Appendix The autoregressive method used in this paper allowed for inclusion of independent variables that are time dependent and fixed over time and partial use of data for people with missing data. For the ith person the outcome yi,(FEV, or FVG) at time t is expressed as L Jf K yui= a + ,,1ys,-+IJ.xiit + Efk ,.,1 i- i k-lI ziA + e,(i where I = 1 to n; t = L to T; yi, = value of the outcome variable for the ith person at the tth examination, ei, is statistically independent for all it with common N(O,&) distribution. x,,, is the jth time dependent exposure variable for ith subject ascertained at time t. The zs represent exposure variables that do not change over time. The es represent the effect of the previous ys on the current level ofy, whereas the fis and the I*s repre- sent the effect of the independent variables on the level of the outcome variable at time t after adjusting for the levels of the outcome variable at the previous L time points. To find the number of previous measurements of outcome to be included, the following equation was used: L yit = a + ylYiz - I + eit , =, (2) L was the highest order of previous outcome that contributed significantly to the model. Table 4 shows the results of these analyses. 1 Burrows B, Earle RH. Course and prognosis of chronic obstructive lung disease. N Engl Med 1969;280: 397-404. 2 Anthonisen NR, Wright EC, Hodgkin JE. Prognosis in chronic obstructive pulmonary disease. Am Rev Respir Dis 1986;133:14-20. 3 Cotes JE. Lung function. Assessment and application in medicine. 4th ed. Oxford: Blackwell, 1979:358-61. 4 Becklake M. Occupational exposures: evidence for a causal association with chronic obstructive pulmonary disease. Am Rev Respir Dis 1989;140:s85-91. 5 Oxman AD, Muir DCF, Shannon HS, Stock SR, Hnizdo E, Lange HJ. Occupational dust exposure and chronic obstructive pulmonary disease. Am Rev Respir Dis 1993; 148:38-48. 6 Bjorseth 0. Some hygienic aspects related to use of reacted (dry scrubbed) alumina in primary aluminium production. Trondheim: The Norwegian Institute of Technology, Universtity of Trondheim, 1983. (Dissertation.) 7 Soyseth V, Kongerud J, Ekstrand J, Boe J. Relation between exposure to fluoride and bronchial responsive- ness in aluminium potroom workers with work-related asthma-like symptoms. Thorax 1994;49:984-9. 8 Kongerud J, Samuelsen SO. A longitudianal study of respi- ratory symptoms in aluminum potroom workers. Am Rev RespirDis 1991;144:10-5. 9 Kongerud J, Gronnesby JK, Magnus P. Respiratory symp- toms and lung function of aluminium potroom workers. ScandJ Work Environ Health 1990;16:270-7. 10 Ronneberg A. Mortality and cancer morbidity in workers from an aluminium smelter with prebaked carbon anodes-Part m: mortality from circulatory and respiratory diseases. Occup Environ Med 1995;52:255-61. 11 Abramson MJ, Wlodarczyk JH, Saunders NA, Hensley MJ. Does aluminium smelting cause lung disease? Am Rev RespirDis 1989;139:1042-57. 12 Gardner RM, Baker CD, Broennlen AM, et al. ATS- Statement-Snowbird workshop on standardization of spirometry. Am Rev Respir Dis 1979;119:831-8. 13 Gulsvik A. Obstructive lung disease in an urban population. Rikshospitalet, University Hospital, Oslo, 1979. (Dissertation.) 14 Kongerud J, Aalen 00, Vale JR. Questionnaire reliability and validity for aluminium potroom workers. Scand Jf Work Environ Health 1989;15:364-70. 15 Rosner B, Mufioz A, Tager I, Speizer F, Weiss S. The use of an autoregressive model for the analysis of longitudinal data in epidemiologic studies. Stat Med 1985;4:457-67. 16 Laird NM, Ware JH. Random-effects model for longitudinal data. Biometrics 1982;38:963-74. 17 Zeger S, Liang K. Longitudinal data analysis for discrete and continuous outcome. Biometrics 1986;42:121-30. 18 Kleinbaum DG, Kupper IL., Muller KE. Applied regression analysis and other multivarible methods, 2nd ed. Boston: PWS-KENT, 1987:325-6. 19 Rosner B, Muftoz A. Autoregressive modelling for the analysis of longitudinal data with unequally spaced examinations. Stat Med 1988;7:59-71. 20 Chan-Yeung M, Enarson D, MacLean L, Irving D. Longitudinal study of workers in an aluminum smelter. Arch Environ Health 1989;44:134-9.