Reviews of epidemiological studies of cardiovascular effects of occupational noise have been published regularly in the literature. The first part of this paper summarizes the conclusions which have been drawn. The second part focuses in more detail on the results of studies in which noise exposure was not assessed by objective measurements, but on the basis of subjective ratings as given by the study subjects in a questionnaire. Finally, the paper addresses some issues of occupational noise studies which deserve special attention from the methodological and conceptual point of view. These are concerned with study design, statistical inference, selection and confounding bias, overcontrolling, effect modification and exposure misclassification.
Keywords: Occupational noise, Cardiovascular effects, Epidemiology, Review, Subjective exposure assessment, Methodology
|How to cite this article:|
Babisch W. Epidemiological studies of the cardiovascular effects of occupational noise - a critical appraisal. Noise Health 1998;1:24-39
| Introduction|| |
Most of the occupational noise studies found in the literature to date relating to cardiovascular disease are concerned with arterial blood pressure. Response variables other than blood pressure have rarely been examined. In early reviews of noise and cardiovascular disorders, it was concluded that prolonged exposure to high intensity industrial noise increases the risk of hypertension (Westman and Walters, 1981). Probably the most comprehensive review of epidemiological studies on non-auditory health effects ever written was published in 1981 by the United States Environmental Protection Agency (Thompson, 1981). Although some of the studies identified were concerned with traffic noise, the vast majority of the studies dealt with occupational noise and were of cross-sectional design. Based on consistent findings from 44 out of 55 studies on blood pressure, Thompson concluded that the strongest evidence of an association - if one existed - was between exposure to high noise levels and elevated blood pressure (Thompson, 1983). The systolic blood pressure differences in these studies were between 5 and 10 mmHg and the prevalence ratio of hypertension between 1.6 and 2.8. A ranking system was applied to evaluate the studies with regard to their validity as far as analytic power, assessment of exposure and health outcome, possible selection bias, and control for potential confounding. Most of the studies were rather poor in this respect. This was partly due to the lack of statistical evaluation. All in all, it was concluded that occupational noise must be considered as a cardiovascular risk factor for high blood pressure. There was sufficient evidence to support the argument for further investigation of the effect of long-term exposure to high noise on blood pressure, especially among industrial workers (Thompson, 1981; Westman and Walters, 1981).
In another review published about ten years later, Thompson discussed 26 epidemiological studies on blood pressure as the cardiovascular end point, which had been published in the English literature since 1980 (Thompson, 1993). Although the newer studies showed some improvement in methodology, the review concluded that the picture was no longer clear, neither confirming nor invalidating the hypothesis that prolonged exposure to high noise levels increases the risk of high blood pressure. More rigorous studies showed the lower associations. Prevalence ratios for hypertension in these studies ranged from 0 to 3.1 and mean blood pressure differences were between 0 and 10 mmHg (Thompson, 1993). (Note: Prevalence (-rate) ratio: [rate (proportion) of prevalent cases of disease in the exposed group] / [rate (proportion) of prevalent cases of disease in the unexposed group] at a specific point in time). Again, a few traffic noise studies were included in this review. The need for better control of confounding and effect modification was stressed in the conclusions - particularly, since the effects are likely to be relatively small and therefore more susceptible to confounding effects by intervening variables. More powerful study designs were demanded, including analytical studies such as case-control and cohort studies to allow for causal reasoning.
This interpretation of findings was in line with that of other reviewers who also concluded that no definite answer could be given to the question of whether occupational noise causes high blood pressure, since there are conflicting results of studies of relatively good quality (DeJoy, 1984; Dijk, 1990). Van Dijk, for example, concluded that the epidemiology of non-auditory effects of occupational noise was still in its years of childhood (Dijk, 1990). A more "pro-noise" conclusion came from the review by Kristensen, who evaluated 47 studies on the relationship between noise and cardiovascular diseases (Kristensen, 1989). These studies were grouped with respect to their methodological strength. A clear correlation was found between study quality and results, so that the share of positive studies increased with increasing quality of the study. However, it was also noted that the epidemiology in this field was still of relatively low quality. The overall conclusion was that the research which was primarily conducted in the 1980's yielded reasonable support for the hypothesis of a causal relationship between noise and cardiovascular diseases. This does not mean that the causal relationship itself was accepted to be true. A recent review prepared by PasschierVermeer for the Dutch Health Council (Passchier-Vermeer, 1993) took 20 occupational noise studies into account which met certain methological standards. There, a more rigorous conclusion was drawn. Namely, that there is sufficient evidence of a causal relationship between occupational noise at levels above 85 dB(A) and the prevalence of hypertension. An average relative risk of 1.7 was estimated for workers exposed to high occupational noise as compared to workers in less exposed areas. Increases in mean systolic and diastolic blood pressure of 3.9 and 1.6 mmHg, respectively, were given. Over the past years, there has been a long standing controversy as to which of the two blood pressure measurements indicates better subsequent cardiovascular diseases. However, since systolic and diastolic blood pressure are highly correlated, they appear to be of similar predictive value for the incidence of ischaemic heart disease (The Pooling Project Research Group, 1978; Shaper et al., 1985; Selmer, 1992). In the elderly, systolic blood pressure seems to be a better predictor for cardiovascular mortality (Taylor et al., 1991).
In contradiction to this, the resume of a literature review and an expert meeting held recently by the German Federal Institute for Occupational Medicine was that the precise quantitative significance of noise effects could not be estimated with confidence (Schust, 1995; Bundesanstalt fur Arbeitsmedizin, 1996). The review given by Schwarze at this meeting particularly concluded that the results of studies on high blood pressure conducted in the past 12 years were highly contradictory (Schwarze, 1996). But the experts agreed that occupational noise must be suspected to be a risk factor contributing to the development of cardiovascular disorders (Bundesanstalt fur Arbeitsmedizin, 1996). This statement is actually not much different from those given in the 1980's reviews.
Thompson recognised in her latest review a few studies from the 1990's that used a longitudinal approach in the study design (Hirai et al., 1991; Hessel, 1994), indicating some methodological progress in noise epidemiology (Thompson, 1996). Both studies gave negative results. The findings in earlier longitudinal studies were contradictory (Lees, 1980; Delin, 1984; Aro, 1984; Vermel et al., 1988). A study carried out in China offers strong evidence of an association between noise level and prevalence of hypertension. In this study, a dose-response relationship was found showing a relative risk (odds ratio) of 1.36 per 10 dB(A) increase in sound pressure level within the range of 75 to 105 dB(A) in women who had worked their entire lives in high noise with unprotected ears (Zhao et al., 1991). Note: Relative risk is used as a general term including prevalence ratio (crosssectional study), risk ratio, incidence rate ratio (cohort study), odds ratio (case-control study). Schwarze and Thompson pointed out that in particular those studies support the hypothesis that occupational noise may lead to sustained blood pressure elevations. These refer to very high sound pressure levels (100 dB(A)), long exposure periods and unprotected ears (Schwarze and Thompson, 1993; Schwarze, 1996). It was emphasised that future noise research should focus on mediating and effect modifying factors and their quantitative role in the causal chain between noise and health (Thompson, 1996). This is in line with Jansen who suggested that preventive medicine should spend greater effort in ascertaining (physiologically) sensitive subjects (Jansen et al., 1996). Attempts have been made to identify subjects on the basis of a stress test (Tomei et al., 1991; Jansen et al., 1996). From the methodological point of view, the identification of vulnerable or sensitive groups of people would help to eliminate random (statistical) noise from the data and thus improve statistical inference.
| Discussion|| |
The state of the art of the research of cardiovascular health effects of occupational noise exposure still shows a heterogeneous picture. Methodological problems do not seem to have been overcome. As stated recently in the document prepared by Berglund and Lindvall for the World Health Organisation, more epidemiological studies are required to clarify the nature of nonauditory health risk associated with occupational noise exposure (Berglund and Lindvall, 1995). Due to the equivocal findings, no guideline values may be given at the moment. A difficulty with the detection and interpretation of cardiovascular noise effects in epidemiological studies is the fact that the noise reactions are non-specific (Ahrlin and Ohrstrom, 1978). Any other stimulus that acts via the same pathway may cause the same health outcome. This means that possible confounding is a major problem in epidemiological noise research, particularly when small effects are investigated. Individual factors can be strong effect modifiers either reinforcing or diminishing the potential noise effects. Furthermore, manifest cardiovascular health effects which are measured with epidemiological methods have long induction periods, making any study design extremely susceptible to factors other than the noise. Analytic epidemiological studies, such as case-control and cohort studies, which allow causal inference are extremely rare to date.
[Figure - 1] gives the principal reaction model to which the epidemiological test hypothesis refers (Babisch, 1997). It outlines the causal chain of the presumed noise induced cardiovascular effects and is based on the general stress concept (Henry and Stephens, 1977; Cohen, 1995). The noise stressor arouses the sympathetic and endocrine system (sympathetic-adrenalmedullary axis and pituitary-adrenal-cortical axis) which in the long run causes physiological and biochemical dysfunction in cardiovascular risk factors - thus increasing the risk for high blood pressure and myocardial infarction. What is to be stressed here are the two different mechanisms in which the sound may affect the organism - the "direct" and the "indirect" pathway in the model. On the one hand the relatively low level environmental noise which acts primarily via the cortical (cognitive) and subcortical (emotional) perception of the sound - thus making the sound become noise. On the other hand, sound of higher intensity which can also affect the autonomous nervous system directly via synaptic nervous interactions in the reticular activating system and parts of the between brain including the hypothalamus (Westman and Walters, 1981). When dealing with occupational noise one is usually concentrating on the left branch of this model - the direct pathway. However, it was emphasised in the literature that not only in traffic noise but also in occupational noise, research should pay more attention to the perception of the noise on the disease outcome - in order to see more stringent results in line with the hypothesis (Delin, 1984; Dijk, 1990; Schwarze and Thompson, 1993). It was suggested that indirect effects could occur at noise levels less than necessary to affect hearing (Carter, 1986). Noise effects on blood pressure changes were found to be more pronounced in white collar workers than in blue collar workers, presumably, because white collar workers were more disturbed by the noise, even at lower noise levels (Aro, 1984).
Subjective assessment of noise exposure
If noise exposed subjects who are not annoyed by the sound pressure level are presumably not at risk, one would expect stronger associations of any health sequela with subjective indicators of exposure (noise annoyance) than with objective indicators (sound pressure level) (Gierke and Harries, 1990). Grouping subjects by noise levels would then diminish the noise (annoyance) effect, because the relation between annoyance due to work noise and the sound pressure level was found to be weak (Dijk et al., 1987a; Dijk et al., 1987b; Melamed et al., 1992). [Table - 1],[Table - 2] list the results of studies where noise was assessed by subjective measures of noise such as annoyance, noisiness, loudness or perceived noise.
The results of epidemiological studies by van Dijk and co-workers do not support this idea. Weak but significant negative associations were found between noise annoyance and blood pressure in shipyard and mixed industrial workers in two cross-sectional studies (Dijk et al., 1987a; Dijk et al., 1987b). In comparison, no significant relationship was found with the noise level. Selection bias due to the healthy worker effect and the inability to test possible interactions with mental efforts were discussed in the explanation of the negative findings. In a study of cross-sectional and longitudinal design, Aro compared the discomfort caused by noise (5 categories) with the noise level as predictors on blood pressure and blood pressure change, respectively (Aro, 1984). The correlation of systolic/diastolic blood pressure with the subjective noise measurement (r = 0.09 (n. s.)/0.00 (n. s.)) was smaller than with the objective noise measurement (r = 0.14 (sig.)/0.03 (n. s.)). The longitudinal findings were similar. The standardised regression coefficients found for noise discomfort as a predictor of blood pressure change ((3 = 0.07 (n. s.)/0.06 (n. s.) mmHg per standard deviation (b = 0.71/0.38 mmHg per category)) were smaller than those found for the noise level as a predictor (β = -0.08 (n. s.)/0.13 (sig.) mmHg per standard deviation). However, subjective and objective noise exposure were treated simultaneously in the statistical model. It was mentioned in the publication that when only one noise variable was included in the model at a time, each variable was a significant predictor of diastolic blood pressure change. Again, the noise level showed the stronger positive association. All in all, this study, too, does not support the idea of a closer relationship between subjective measures of exposure and blood pressure than between objective measures of exposure and blood pressure.
In a community-based cross-sectional study carried out by Lercher et al. in the Tyrol, Austria, the effect of occupational noise annoyance (2 categories) on blood pressure was studied. Mean differences of 2.1 (n. s.) and 3.5 (sig.) mmHg for systolic and diastolic blood pressure, respectively, were found for the two differently annoyed groups, the latter being statistically significant (Lercher et al., 1993). The results regarding interactions with other work stressors supported the idea that noise effects are more pronounced in combination with other stressors. In the matched case-control study carried out in car-industries by Mann and co-workers, the association of the incidence of myocardial infarction with a subjective descriptor of perceived work noise (2 categories) was not stronger than with the objective noise measurement (> 90 dB(A) vs. < 90 dB(A)). A relative risk of 1.6 (n. s.) was found regarding the subjective exposure as compared to a relative risk of 2.0 (n. s.) regarding the objective exposure (Mann et al., 1994). The Berlin population-based case-control study focussed primarily on the incidence of myocardial infarction in relation to traffic noise. It also provides information about occupational noise exposure (Ising et al., 1997). The subjects estimated their work noise exposure to a scale of typical noise sources. An increasing trend towards higher relative risks across four subjective noise categories was found, ranging from 1.5 (sig.) to 3.7 (sig.) as compared to the lowest category. All were significant. No objective noise measurements were available. The relative risks found were very high, suggesting that subjective work noise is a major risk factor for myocardial infarction.
Two cohort studies regarding subjective occupational noise exposure were carried out by Bellach and co-workers. It was stated that noise was not the primary matter of interest when the studies were initiated. In a random sample of the German population, relative risks of 2.8 (sig.) and 0.6 (n. s.) for the cumulative incidence of myocardial infarction and hypertension, respectively, were found for subjects who reported that they perceived occupational noise, as compared to those who denied this (Bellach et al., 1995). The follow-up period considered was 11 years. Morbidity was assessed subjectively using a standardised questionnaire. The authors claimed that the validity of this instrument of disease assessment was high when severe diseases were considered (Bormann et al., 1990). In the other community-based study, the study population was again interviewed after a followup interval of 16 years (Bellach et al., 1995). No relationships were found between the noise exposure reported at the beginning of the study and the incidence of cardiovascular diseases (figures not given in the publication). A crosssectional analysis carried out at the end of the follow-up revealed that males, who reported that they were exposed to noise at work had nonsignificant relative risks of 1.1 (n. s.) and 1.2 (n. s.) for myocardial infarction and hypertension, respectively, as compared to those not exposed (Bellach et al., 1995). Again, subjective morbidity was assessed. Also clinical blood pressure measurements were carried out in the follow-up phase. The observed relative risk of hypertension of 1.7 (sig.) for the subjects who reported that they were exposed to occupational noise as compared to those who were not exposed, was significantly higher regarding subjective morbidity.
| Discussion|| |
The number of studies on the relationship between cardiovascular effects and subjective indicators of noise exposure is extremely small. Therefore, no distinction was made in this review between descriptors of noise annoyance, loudness, discomfort due to noise, and noise perception. Presumably, answers to the questions of whether a work place is noisy or not and how loud a work place is, may to some extent reflect the unpleasantness of noise and thus the noise annoyance. However, this needs further clarification. The magnitudes of the effects observed in the studies where subjective noise exposure and blood pressure or hypertension were investigated, do not give much support to the idea that subjective indicators of the noise are stronger predictors than the sound pressure level. On the other hand, the relative risks found with regard to the association between myocardial infarction and subjective noise exposure, which range between 1.0 and 3.7, tend to be higher than those found in studies where noise level was considered as exposure (Lees et al., 1980; Theriault et al., 1988; Kent et al., 1986). The entire field may be very complex. In the literature on stress research, it is reported that studies relying on objective measures have yielded fairly consistent positive associations between hypertension and objective measures of stress. In comparison, studies using self-reported measures of stress have often yielded null or negative associations with the haemodynamic health status (Winkleby et al., 1988). Some persons may cope with adverse circumstances by suppressing their emotional reactions to preserve psychological well-being. But also physiological effects on central nervous system functioning, due to alterations in blood pressure, may change the perception of external events (Winkleby et al., 1988).
The population-based studies (Ising et al., 1997; Lercher et al., 1993; Bellach et al., 1995) may be susceptible to a confounding bias of an unknown direction being introduced by the mixture of blue and white collar workers in the study population. A criticism which applies to all cross-sectional and case-control studies in general, is concerned with observation bias due to over reporting. Subjects who are aware of their disease by the time they become interviewed about the noise may give distorted answers in the presence of a disease. For example, cases may try to blame the noise for their health problem, although there is another (unknown) reason. Or, "tough" controls may try to play down the noise in the interview. Both effects tend to overestimate the true noise effect. The issue of possible recall bias may become even more relevant, when both exposure and morbidity are measured on a subjective basis at the same point in time, as was done in the cross-sectional part of the studies by Bellach et al. (Bellach et al., 1995). On the other hand, the problem of recall bias may be less of a problem there because the noise objective was not immediately obvious.
In conclusion: The research regarding perceived noise and its possible health effects is in its beginnings. No definite answers can be given or trends extracted. The study by Ising et al. suggest that it may be worth following the concept in future noise research. Methodological arguments call for prospective studies in this field. Subjective exposure should be assessed at a time when the participants are free of the disease under study.
In the following section, some methodological issues are addressed which seem to be problems in many of the studies on the relationship between cardiovascular effects and occupational noise exposure.
In the literature, information about the use of ear protection is largely missing. This does not only apply to the early studies but also to the newer ones, which is surprising since hearing conservation programmes have been widely introduced in the work environment. Hearing protection is most likely to act as an effect modifier on the study results (Talbott et al., 1996). The ear protectors are designed to provide an overall reduction in exposure of about 15 to 30 dB(A) (Berger, 1993). In an experimental study, Ising and co-workers demonstrated that the blood pressure readings and the urinary excretion of catecholamines was lower on days when the workers wore ear protectors as compared to days with unprotected ears, for average noise levels of 95 dB(A). The effective reduction of the sound pressure level in this experiment was measured to be between 10 to 16 dB(A) which accounts for the practical aspects of the individual's handling of the devices (Ising et al., 1980). When hearing protection was not considered in a study, it may be that fitting of the devices was simply not provided. From the researcher's point of view this may be an "advantage" of the early noise studies that hearing protection programmes were not so broadly implemented during the first post war decades - and even until now in developing countries. Misclassification of exposure would then be less of a problem in the older studies but a problem definitely in the newer ones. Herewith it is also important to recognise that even if workers are obliged to use hearing protection by the time the study is conducted, the duration of employment may include many years of exposure to noise when hearing conservation programmes were not in place (Talbott et al., 1985). The length of exposure to be considered for the development of manifest cardiovascular diseases may be rather long (Kavoussi, 1973; Lang et al., 1992).
In many occupations the appropriate use of hearing protection is likely to reduce the effective exposure levels to below 85 dB(A). If the analysis does not take into account hearing protection, it may very well be that noise "exposed" subjects who use ear protection will be compared with "unexposed" subjects who do not use ear protection. The difference of the effective noise level between the compared groups will then be very small. It would not be surprising if such studies give negative results.
When epidemiological noise research deals with populations that protect their ears - and this probably holds for all new studies due to occupational safety regulations - one could speculate that the indirect noise effects (due to annoyance) may dominate the possible health outcome. The direct noise effects (due to the sound pressure level) then play a lesser role. In other words, in most industrialised countries we may be dealing with a scenario where the sound pressure level is no longer the primary noise stressor but the annoyance caused by the work noise - similar to the situation with regard to environmental noise. Annoyance, however, is modified by several other factors and only partly determined by the noise level (Dijk et al., 1987b; Job, 1996). This means that even if hearing protection is adequately assessed and treated in the analysis, associations may be small when the health effects become related to the sound exposure level.
The issue becomes further complex when the discomfort of the use of hearing protection is considered. Temporal sequence is interesting here: Do annoyed subjects use hearing protection more often because they are annoyed by the noise, which should reduce the perceived noise stress? Or, are subjects more annoyed by the noise because they have to wear ear protection, which would enhance the stress reactions (Kristal-Boneh et al., 1995)? Studies show that the use of hearing protection is positively associated with noise annoyance at work (Dijk et al., 1987b). This raises the question of how hearing protection and annoyance can be adequately considered in the statistical analysis. The simple treatment as a covariate in a multiple regression model - the conception of a confounder (Kristal-Boneh et al., 1995) - would be inappropriate.
Cross sectional studies
A general argument against most work noise studies by reviewers is concerned with the cross sectional character of the vast majority of studies. Cross-sectional studies are usually viewed as being purely descriptive. Epidemiological reasoning, however, should be based on analytical studies such as case-control studies, cohort studies or experimental trials (Hennekens and Buring, 1987). In a crosssectional or prevalence study, respectively, only one specific point in time is looked at. It is unclear what the temporal sequence is. Does the cause precede the effect or vice versa? However, this general criticism may be a minor problem in noise studies. It is certainly likely that the exposure of an individual is determined by his illness, meaning that subjects with health problems may tend to move away from highly exposed areas within the company if possible. Such kind of self-selection would be a source of differential exposure misclassification, because diseased subjects who have formerly been exposed will be allocated to the unexposed group based on their present exposure. As a consequence, the true noise effect would be underestimated. On the other hand, misclassification in the opposite direction does not seem to be very plausible. It is unreasonable that subjects with health problems would be moved into exposed work places because of these health problems.
However, temporality in cross-sectional noise studies can be a bigger problem with regard to control of confounding bias even if potential confounders and effect modifiers are largely assessed within a study. For example, the current smoking behaviour, body weight, physical activity and other risk factors which one wants to account for in the analysis are likely to be determined by the presence of the chronic disease in prevalent cases. As a consequence, a considerable amount of uncontrolled or residual confounding may remain in the data. If no retrospective components of data assessments are considered in the study design, confounding bias of an unknown direction may yield to an over- or underestimation of the true noise effect.
Healthy worker effect
A well recognised source of confounding bias is associated with the "survivor" or "healthy worker" effect. If the noise hypothesis is true, at the time when the study is started an excess rate of exposed workers may have died from cardiovascular diseases due to noise or may have left the place of work because of the noise. Noise sensitive persons and those with health problems may not even be interested in taking a noisy employment. The over-representation of survivors in the exposed group tends to diminish the magnitudes of the effect estimates towards zero or relative risk one. This, in principle, applies to all observational studies including cross-sectional, case-control and cohort studies.
White/blue collar workers
A large variety of factors may affect the selection of individuals into the different occupational noise categories. Noisy working conditions are often associated with other factors of the work place which may be related to the health outcome, e. g. heat, cold, solvents and other chemical substances, light, danger, dust and socio-professional conditions (Rosenman, 1984; Rosenstock and Cullen, 1986). Besides factors related to the work place, white and blue collar workers may differ with regard to lifestyle, social status and psycho-social factors. These are are potential confounding factors (Fouriaud et al., 1984; Tyroler et al., 1987). Therefore, white collar workers are not a good control group for noise exposed blue collar workers (Dijk, 1990). It may be difficult to identify unexposed blue collar workers in industrial settings, but the comparison of white and blue collar workers (Malchaire and Mullier, 1979; Belli et al., 1983) must be eliminated due to the potential of confounding bias. Relative homogeneous occupational groups are required for comparison.
Some studies on cardiovascular effects of occupational noise utilised hearing loss as a surrogate for noise exposure. The idea is that hearing loss is an indicator for cumulative noise exposure. However, these studies have been strongly criticised (Kristensen, 1989; Dijk, 1990). There is considerable inter-individual variability in the susceptibility for noise-induced sensorineural hearing loss. Hearing loss may be due to causes other than the work noise (e. g. leisure noise, military service). Because of the low specificity of hearing loss as a marker for noise exposure, there will be noise exposed subjects who develop cardiovascular diseases due to the noise (if the noise hypothesis is true) who are grouped as being unexposed because no substantial hearing impairment is present - and vice versa. Such kind of differential exposure misclassification dilutes the true noise effect. Hearing loss and cardiovascular changes may be biologically interrelated (one causing the other) or have the same (unknown) genesis other than noise (Rosen et al., 1964; Rubinstein et al., 1977; Pillsburg, 1986). Not only hypertension was found to be associated with hearing loss but also hypotension (Meinhard and Renker, 1970). Presbycusis (hearing impairment related to aging) shows a large variability among individuals and may affect study results derived from older populations. The argument sometimes made that loss of hearing is likely to protect against non-auditory noise effects, does not apply to typical work noise of high sound pressure levels. Moreover, hearing impaired subjects suffer from recruitment (that is the abnormal growth of loudness).
In conclusion: The results of studies on the relationship between hearing loss and cardiovascular diseases are contradictory (Thompson, 1981; Fogari et al., 1994). The number of negative studies that use hearing loss as a marker for noise exposure was found to be relatively larger than the number of negative studies that refer to the noise level (Kristensen, 1989). Thompson stated in the EPA report (Thompson, 1981) that blood pressure in worker groups with different degrees of hearing impairment must be interpreted with caution when inferring causal relationships to noise exposure. In general, other comments are more rigorous in discouraging hearing loss as a biological marker for noise exposure in studies on stress-related non-auditory health effects (Wu et al., 1987; Dijk, 1990).
In any epidemiological study, control for possible confounding is required. Since we are investigating cardiovascular outcomes, control variables that the analyses should account for, comprise risk factors such as age, sex, social class, body mass index, physical activity at work and during leisure time, smoking, alcohol consumption, salt intake and family history of cardiovascular diseases. Generally, a distinction must be made between exogeneous and endogeneous factors. The above mentioned factors can be characterised as exogeneous factors. They act from outside the organism on the health outcome in the broadest way and include behavioural factors. In contradiction, endogeneous risk factors are factors that act from the inside of the body. These comprise factors that belong to the metabolism and physiological functioning of the organism, such as stress hormones, blood pressure, blood lipids, serum glucose or blood clotting activity, to mention but a few.
A problem arises when endogeneous and exogeneous factors are treated simultaneously as covariates in the statistical analyses because it is not clear whether endogeneous factors are the cause, effect or both. The idea of controlling for endogeneous risk factors comes from the fact that noise stress related effects are unspecific. Changes in endogeneous risk factors may also occur due to exogeneous factors and stressors other than the noise. These may be statistically associated with the noise and thus formally fulfil the criterium of a confounder. Consequently, studies account for these variables in the analyses (Belli et al., 1984; Theriault et al., 1988; Mann et al., 1994). This is a questionable strategy from the methodological point of view, because it incorporates the problem of overcontrolling (Kramer, 1988). A confounder is a variable that is statistically (either biologically or simply for distributional reasons) related to the exposure and the outcome. Multiple regression analysis is often applied to adjust the effect under study for the influence of the confounder. However, having the scheme of the noise stress model in mind, it is obvious that endogeneous risk factors play the role of a mediator in the chain of a causal pathway of the effect. For example, according to the model, noise increases serum cholesterol. Hypercholesterolaemia, on the other hand, is a risk factor for arteriosclerosis leading to myocardial infarction. This means that controlling for the cholesterol in a regression model would diminish the true noise effect because the degree of variance in the dependent variable, which is explained by elevated cholesterol due to noise, would at least partly be attributed to the cholesterol factor. The remaining variance explained by the noise factor is mediated by other endogeneous factors (other than the cholesterol). The more endogeneous factors are included in the model, the less variance will be attributed to the noise factor. For example, our own data from the Caerphilly study show a crude relative risk for the incidence of IHD in men of the highest traffic noise category - adjusted for room orientation (N = 2094) - of 1.23. Entering cholesterol into the model reduces the relative risk to 1.13. The risk reduces further to 1.08 when plasma viscosity is entered.
It is not that one should not control for endogeneous factors, but one must be aware of the fact that overcontrolling acts conservatively in the study results. Since the epidemiology of noise research is dealing with small effects, one may end up with a lot of negative studies due to overcontrolling. Rather than controlling endogeneous risk factors one should assess exogeneous factors that compete with the noise as fully as possible. The judgement to include or exclude a control variable from the statistical analyses should not be made on grounds of significance of its univariate associations with disease or exposure but on biological grounds and knowledge from other studies. Several nonsignificant small effects may accumulate to a significant impact on the study result. (Monson, 1990).
Overcontrolling also occurs when groups are matched with regard to endogeneous factors of the study subjects. This incorporates the well recognised problem of overmatching in case control studies (Rothman, 1986). Furthermore, overcontrolling may occur when effect modifiers or mediators of effect are treated as confounders in a statistical model, e. g. hearing protection and noise annoyance. In a study where subjective and objective noise exposure were treated simultaneously in the statistical model, no noise effect was detected (Aro, 1984). However, when discomfort of noise was excluded, the noise level was a significant predictor of diastolic blood pressure change. Overcontrolling occurs when subjects are excluded from the study population on the basis of criteria which are associated with the disease outcome. For example, the exclusion of subjects who are on medication - e. g. for high blood pressure or coronary heart disease (Cottington et al., 1983; Pulles et al., 1990; Kristal-Boneh et al., 1995) - is likely to diminish the true noise effect. Cases will be lost that are possibly due to the noise exposure. Antihypertensive treatment showed a strong positive correlation with mean blood pressure (Fouriaud et al., 1994).
| Conclusions|| |
The results of epidemiological studies of cardiovascular health effects of occupational noise are contradictory. This applies also to newer studies which show methodological improvement in design and analysis to control for possible bias. However, methodological problems seems to have not been overcome. The use of hearing protection may be a source of differential exposure misclassification, particularly in the newer studies because of the implementation of hearing conservation programmes (Tarter and Robins, 1990). Subjective measures of exposure can be stronger predictors of the health outcome than the sound level, but there is little epidemiological information available at psesent.
In future research of non-auditory health effects of occupational noise exposure, special emphasis should be given to analytical studies (e. g. cohort studies). The impact of hearing protection on the health effects should be investigated quantitatively. Subjective indicators of noise exposure (e. g. annoyance) should be assessed as well as objective noise indicators (noise level). It has been recognised in the literature that intervening factors must be assessed, but there may be problems in the conceptual understanding of confounders, effect modifiers and mediators of effect and exposure, and the strategies of how to treat them in the statistical analyses. This includes, for example, the adequate treatment of duration of exposure which is strongly correlated to age (Dijk, 1987a). The combination of noise with other stressors may act synergistically on the health outcome (Dijk, 1987c).
This paper was presented at the 3rd European Conference on Protection Against Noise (PAN), 12-15 March 1998, Stockholm Sweden organised and supported by the European Commission BIOMED 2 concerted action PAN (Contract BMH 4-CT96-0110).
| References|| |
|1.||Ahrlin U. & Ohrstrom E. (1987) Medical effects of environmental noise on humans. Journal of Sound and Vibration 59: 79-87 |
|2.||Aro S. (1984) Occupational stress, health related behaviour, and blood pressure: a 5-year follow-up. Prev. Med. 13: 333-348 |
|3.||Babisch W. (1998) Epidemiological studies on cardiovascular effects of traffic noise. In Advances in Noise Research. Biological Effects of Noise. Prasher D. & Luxon L., eds. Whurr Publishers Ltd., London, in print |
|4.||Belli S., Sani L., Scarficcia G. & Sorrentino R. (1984) Arterial Hypertension and noise: a cross-sectional study. Am. J. Ind. Med. 6: 59-65 |
|5.||Bellach B., Dortschy R., Muller D. & Ziese T. (1995) Gesundheitliche Auswirkungen von Larmbelastung - Methodische Betrachtungen zu den Ergebnissen dreier epidemiologischer Studien. Bundesgesundhbl. 38: 84-89 |
|6.||Berger E.H. (1993) The naked truth about NRR's. Cited in Community noise. Document prepared for the World Health Organisation. Berglund B. & Lindvall T. eds. (1995) Archives of the Center for Sensory Research 2, Stockholm, appendix 1, pp5 |
|7.||Berglund B. & Lindvall T. (1995) Community noise. Document prepared for the World Health Organisation. Archives of the Center for Sensory Research 2, Stockholm, pp1-195 |
|8.||Belli S., Sani L., Scarficcia G. & Sorrentino R. (1984) Arterial hypertension and noise: a cross-sectional study. American Journal of Industrial Medicine 6: 59-65 |
|9.||Bormann C., Hoeltz J., Hoffmeister H., Klaes L., Kreuter H., Lopez H., Stolzenberg H. & Weilandt C. (1990) Subjektive Morbiditat, BGA Schriften 4/90. Munchen: MMV Medizin Verlag. |
|10.||Bundesanstalt far Arbeitsmedizin, ed. (1996) Noise at the work-place and cardiovascular diseases (in German). Schriftenreihe der Bundesanstalt fMr Arbeitsmedizin, Tagungsbericht 12. Wirtschaftsverlag NW, Bremerhaven, pp4 |
|11.||Carter N.L. (1986) Non-auditory health effects of occupational noise. The Journal of Occupational Health and Safety 2: 210-215 |
|12.||Cohen S., Kessler R.C. & Underwood Gordon, L. (1995) Strategies for measuring stress in studies of psychiatric and physical disorders. In Measuring Stress, A Guide for Health and Social Scientists. Cohen S., Kessler R. C. & Underwood Gordon L. eds. Oxford University Press, New York |
|13.||Cottington E., Matthews K., Talbott E. & Kuller L. (1983) Occupational stress and diastolic blood pressure in a bluecolour population: the Pittsburgh noise-hypertension project. Am. J. Epidemiol. 118: 442 |
|14.||DeJoy D.M. (1984) A report on the status of research on the cardiovascular effects of noise. Noise Control Eng. J. 23: 32-39 |
|15.||Delin C.-O. (1984) Noisy work and hypertension. The Lancet 10/1984: 931 |
|16.||Dijk F.J.H.v. (1987a) Non-auditory effects of noise in industry, V. a field study in a shipyard. Int. Arch. Occup. Environ. Health 59: 55-62 |
|17.||Dijk F.J.H.v. (1987b) Non-auditory effects of noise in industry, VI. a final field study in industry. Int. Arch. Occup. Environ. Health 59: 133-145 |
|18.||Dijk F.J.H.v. (1987c) Non-auditory effects of noise in industry, VII. evaluation, conclusions and recommendations. Int. Arch. Occup. Environ. Health 59, 147-152. |
|19.||Dijk F.J.H.v. (1990) Epidemiological research on nonauditory effects of occupational noise exposure since 1983. In New advances in Noise Research, part 1. Proceedings of the fifth International Congress on Noise as a Public Health Problem, Stockholm 1988. Berglund B. & Lindvall T., eds. Swedish Council for Building Research, Stockholm, pp285-292 (vol 4) |
|20.||Fogari R., Zoppi A., Vanasia A., Marasi G. & Villa G. (1994) Occupational noise exposure and blood pressure. Journal of Hypertension 12: 475-479 |
|21.||Fouriaud C., Jacquinet-Salord M.C., Degoulet P., Aime F., Lang T., Laprugne J., Main J., Oeconomos J., Phalente J. & Prades A. (1984) Influence of socioprofessonal conditions on blood pressure levels and hypertension control. American Journal of Epidemiology 120: 72-86 |
|22.||Gierke H.E.v. & Harries C.S. (1990) On the potential associations between noise exposure and cardiovascular disease. In New Advances in Noise Research, part 1. Proceedings of the fifth International Congress on Noise as a Public Health Problem, Stockholm 1988. Berglund B. & Lindvall T. eds. Swedish Council for Building Research, Stockholm, pp349-361 (vol 4). |
|23.||Hennekens C.H. & Buring J.E. (1987) Epidemiology in Medicine. Little, Brown and Company, Boston |
|24.||Henry J.P., Stephens P.M. (1977) Stress, health, and the social environment. Springer-Verlag, New York |
|25.||Hessel P.A. (1994) Occupational noise exposure and blood pressure: longitudinal and cross-sectional observations in a group of underground miners. Archives of Environmental Health 49: 128-134 |
|26.||Hirai A., Takata M., Mikawa M. & Yasumoto K. (1991) Prolonged exposure to industrial noise causes hearing loss but not high blood pressure: a study of 2124 factory laborers in Japan. Journal of Hypertension 9: 1069-1073 |
|27.||Ising H., Gunther T., Havestadt C., Krause C., Markert B., Melchert H.U., Schoknecht G., Thefeld W. & Tietze K.W. (1980) Studies to quantify the risk of heart and circulatory diseases in workers exposed to noise (in German). Bundesanstalt fur Arbeitschutz und Unfallforschung (ed), Report No 225. Wirtschaftsverlag NW, Verlag fur neue Werbung GmbH, Bremerhaven |
|28.||Ising H., Babisch W., Kruppa B., Lindthammer A. & Wiens D. (1996) Subjective work noise: a major risk factor in myocardial infarction. Soz Praeventivmed. 42: 216-222 |
|29.||Jansen G., Schwarze S. & Notbohm G. (1996) Noiseinduced hazards to health with special respect to the physiological noise sensitivity (in German). Zeitschrift fiir Larmbekampfung 43: 31-40 |
|30.||Job R.F.S. (1996) The influence of subjective reactions to noise on health effects of the noise. Environment International 22: 93-104 |
|31.||Kavoussi N. (1973) The relationship between the length of exposure to noise and the incidence of hypertension at a silo in Terran. Med. Lavoro. 64: 292-295 |
|32.||Kramer M.S. (1988) Clinical Epidemiology and Biostatistics. Springer Verlag, Berlin |
|33.||Kent S.J., Gierke H.E.v. & Tolan G.D. (1986) Analysis of the potential association between noise-induced hearing loss and cardiovascular disease in USAF aircrew members. Aviation, Space, and Environmental Medicine 57: 348-361 |
|34.||Kristal-Boneh E., Melamed S., Harari G. & Green M.S. (1995) Acute and chronic effects of noise exposure on blood pressure and heart rate among industrial employees: the Cordis study. Archives of Environmental Health 50: 298-304 |
|35.||Kristensen T.S. (1989) Cardiovascular diseases in the work environment. A critical review of the epidemiologic literature on nonchemical factors. Scand. J. Work Environ. Health 15: 165-179 |
|36.||Lercher P., Hortnagl J. & Kofler W.W. (1993) Work noise annoyance and blood pressure: combined effects with stressful working conditions. Int. Arch. Occup. Environ. Health 65: 23-28 |
|37.||Lang T., Fouriaud C. & Jacquinet-Salord M.-C. (1992) Length of occupational noise exposure and blood pressure. Int. Arch. Occup. Environ. Health 63, 369-372 |
|38.||Lees R.E.M., Romeril C.S. & Wetherall L.D. (1980) A study of stress indicators in workers exposed to industrial noise. Canadian Journal of Public Health 71: 261-265 |
|39.||Malchaire J.B. & Mullier M. (1979) Occupational exposure to noise and hypertension: a retrospective study. Ann. Occup. Hyg. 22: 63-66 |
|40.||Mann H.B., Labrot B., Stork J., Cremer P. & Nagel D. (1994) Case-control study regarding myocardial infarction, shiftwork and occupational exposure to noise. In Proceedings of the 9th International Symposium in Epidemiology in Occupational Health, Cincinnati, Ohio 1992. International Commission on Occupational Health, ed. U. S. Department of Health and Human Services, National Institute for Occupational Safety and Health, Cincinnati, pp582-587 |
|41.||Meinhart P. & Renker U. (1970) Untersuchungen zur Morbiditat an Herz- und Kreislauferkrankungen durch Dauerlarmexposition. Z. Ges. Hyg. 16: 853-857 |
|42.||Melamed S., Luz J. & Green M.S. (1992) Noise exposure, noise annoyance and their relation to psychological distress, accident and sickness absence among blue-collar workers - the Cordis study. Isr. J. Med. Sci. 28: 629-635 |
|43.||Monson R.R. (1990) Occupational Epidemiology. CRC Press Inc, Boca Raton |
|44.||Passchier-Vermeer W. (1993) Noise and Health. Health Council of the Netherlands, The Hague |
|45.||Pillsbury H.C. (1986) Hypertension, hyperlipoproteinemia, chronic noise exposure: is there synergism in cochlear pathology? Laryngoscope 96: 1112-1138 |
|46.||Pulles M.P.J., Biesiot W. & Steward R. (1990) Adverse effects of environmental noise on health: an interdiciplinary approach. Environment International 16: 437-445 |
|47.||Rosen S., Plester D., El-Mofty A. & Rosen H.V. (1964) Relation of hearing loss to cardiovascular disease. Trans. Am. Acad. Ophtalmol. Otolaryngol. 68: 433-444 |
|48.||Rosenstock L. & Cullen M.R. (1986) Clinical occupational medicine. Saunders, Philadelphia, pp71-80 |
|49.||Rosenman K.D. (1984) Cardiovascular disease and work place exposures. Arch. Environ. Health. 39: 218-224 |
|50.||Rothman K.J. (1986) Modern Epidemiology. Little, Brown And Company, Boston |
|51.||Rubinstein M., Hildesheimer M., Zohar S. & Chilarovitz T. (1977) Chronic cardiovascular pathology and hearing loss in the aged. Gerontology 23: 4-9 |
|52.||Schust M. (19995) Wirkung von Larm am Arbeitsplatz auf das Herz-Kreislauf-System In Kommentierte Literatursammlung, Schriftenreihe der Bundesanstalt fur Arbeitsmedizin, Literaturdokumentation 2. Bundesanstalt fur Arbeitsmedizin, ed. Wirtschaftsverlag NW, Verlag far neue Wissenschaft GmbH, Bremerhaven |
|53.||Schwarze S. & Thompson S.J. (1993) Research on nonauditory physiological effects of noise since 1988: review and perspectives. In Noise and Man '93, Proceedings of the sixth International Congress on Noise as a Public |
|54.||Health Problem, Nice 1993. Vallet M., ed. Institut National De Recherche Sur Les Transports Et Leur Securite, Arcueil Cedex, pp252-259 (vol 3) |
|55.||Schwarze S. (1996) Forschungen zur Wirkung von Larmbelastungen am Arbeitsplatz auf das Herz-KreislaufSystem - Ubersicht und Ausblick In Ldrm am Arbeitsplatz und Herz-Kreislauf-Erkrankungen, Schriftenreihe der Bundesanstalt fur Arbeitsmedizin, Tagungsbericht 12. Bundesanstalt far Arbeitsmedizin, ed. Wirtschaftsverlag NW, Bremerhaven, pp6-20 |
|56.||Selmer R. (1992) Blood pressure and twenty-year mortality in the city of Bergen, Norway. American Journal of Epidemiology 136: 428-440. |
|57.||Shaper A.G., Pocock S.J., Walker M., Phillips A.N., Whitehead T.P. & MacFarlane P.W. (1985) Risk factors for ischaemic heart disease: the prospective phase of the British Regional Heart Study. Journal of Epidemiology and Community Health 39: 197-209 |
|58.||Talbott E., Helmkamp J., Matthews K., Kuller L., Cottington E. & Redmond G. (1985) Occupational noise exposure, noise-induced hearing loss, and the epidemiology of high blood pressure. American Journal of Epidemiology 121: 501-514 |
|59.||Talbott E.O., Brink L.L., Burks C., Palmer C., Engberg R., Cioletti M. & Inman C. (1996) Occupational noise exposure, use of hearing protectors over time and the risk of high blood pressure: the results of a case/control study. In InterNoise 96, Proceedings of the 25th International Congress on Noise Control Energineering, Liverpool 1996. Hill F.A. & Lawrence R., eds. Institute of Acoustics, St. Albans, pp2131-2137 |
|60.||Tarter S.K. & Robins T.G. (1990) Chronic noise exposure, high-frequency hearing loss, and hypertension among automotive assembly workers. Journal of Occupational Medicine 32: 685-689 |
|61.||Taylor J.O., Cornoni-Huntley J., Curb J.D., Manton K.G., Ostfeld A.M., Scherr P. & Wallace R.B. (1991) Blood pressure and mortality risk in the elderly. American Journal of Epidemiology 134: 489-501 |
|62.||The Pooling Project Research Group (1978) relationship of blood pressure, serum cholesterol, smoking habit, relative weight and ECG abnormalities to incidence of major coronary events: final report of The Pooling Project. J. Chron. Dis. 31: 201-306 |
|63.||Theriault G.P., Tremblay C.G. & Armstrong B G (1988) Risk of ischemic heart disease among primary aluminium production workers. American Journal of Industrial Medicine 13: 659-666 |
|64.||Thompson S. (1981) Epidemiology feasibility study: effects of noise on the cardiovascular system. In Report No. EPA 550/9-81-103A. United States Environmental Protection Agency, ed. Office of Noise Abatement and Control, Washington D.C. |
|65.||Thompson S. (1981) Epidemiology feasibility study: effects of noise on the cardiovascular system. In Report No. EPA 550/9-81-103B. United States Environmental Protection Agency, ed. Office of Noise Abatement and Control, Washington D.C. |
|66.||Thompson S.J. (1983) Effects of noise on the cardiovasular system: appraisal of epidemiologic evidence. In Proceedings of the Fourth International Congress on Noise as a Public Health Problem, Turin 1983. Rossi G., ed. Centro Richerche E Studi Amplifon, Milano, pp711-714 (vol 1) |
|67.||Thompson S.J. (1993) Review: extraaural health effects of chronic noise exposure in humans. In Noise and Disease, Proceedings of the International Symposium, Berlin 1991. Ising H. & Kruppa B., eds. Gustav Fischer Verlag, Stuttgart, pp100-117 |
|68.||Thompson S. (1996) Non-auditory health effects of noise: updated review. In Proceedings of the 25th Aniversary Congress Internoise 96, Liverpool 1996. Hill F.A. &, Lawrence R, eds. Institute of Acoustics, St Albans, pp2177-2182 (vol 4) |
|69.||Tomei F., Tomao E., Papaleo B., Baccolo T.P., Alfi P. (1991) Study of some cardiovascular parameters after chronic exposure to noise. Int. J. Cardiol. 33: 393-399 |
|70.||Tyroler A.H., Haynes S.G., Cobb L.A., Irvin C.W., James S.A., Kuller L.H., Miller R.E., Shumaker S.A., Syme S.L. & Wolf S. (1987) Task force 1: environmental risk factors in coronary artery disease. Circulation 76: suppl I, 139-144 |
|71.||Vermel A.E., Zinenko G.M., Kochanova E.M., Saures L.T., Bogotov K.M. & Chernova E.Y. (1988) Intense industrial noise and AH morbidity (according to the results of a prospective epidemiological survey of organized female populations in Moscow). Ter. Arch. 60: 88-91 |
|72.||Westman J.C. & Walters J.R. (1981) Noise and stress: a comprehensive approach. Environmental Health Perspectives 41: 291-309. |
|73.||Winkleby M.A., Ragland D.R. & Syme S.L. (1988) Selfreported stressors and hypertension: evidence of an inverse association. American Journal of Epidemiology 127: 124-133 |
|74.||Wu T.N., Chou F.S. & Chang P.Y. (1987) A study of noiseinduced hearing loss and blood pressure in steel mill workers. Int. Arch. Occup. Environ. Health 59: 529-536 |
|75.||Zhao Y., Zhang S., Selvin S., Spear R.C. (1991) A dose response relation for noise induced hypertension. British Journal of Industrial Medicine 48: 179-184 |
Institute for Water-, Soil and Air Hygiene, Federal Health Office, P.O. Box 330022, D- 14191 Berlin
Source of Support: None, Conflict of Interest: None
[Figure - 1]
[Table - 1], [Table - 2]