Noise Health Home 

[Download PDF]
Year : 2008  |  Volume : 10  |  Issue : 38  |  Page : 27--33

Road traffic noise and cardiovascular risk

Wolfgang Babisch 
 Department of Environmental Hygiene, Federal Environment Agency, Germany

Correspondence Address:
Wolfgang Babisch
Corrensplatz 1, 14195 Berlin


Studies on the association between community noise and cardiovascular risk were subjected to a meta-analysis for deriving a common dose-effect curve. Peer-reviewed articles, objective assessment of exposure and outcome as well as control for confounding and multiple exposure categories were all necessary inclusion criteria. A distinction was made between descriptive (cross-sectional) and analytical (case-control, cohort) studies. Meta-analyses were carried out for two descriptive and five analytical studies for calculating a pooled dose-effect curve for the association between road traffic noise levels and the risk of myocardial infarction. No increase in risk was found below 60 dB(A) for the average A-weighted sound pressure levels during the day. An increase in risk was found with increasing noise levels above 60 dB(A) thus showing a dose-response relationship. A risk curve was estimated for the association using a polynomial fit of the data that can be used for risk assessment and the environmental burden of disease calculations.

How to cite this article:
Babisch W. Road traffic noise and cardiovascular risk.Noise Health 2008;10:27-33

How to cite this URL:
Babisch W. Road traffic noise and cardiovascular risk. Noise Health [serial online] 2008 [cited 2022 Sep 26 ];10:27-33
Available from:

Full Text


It is widely accepted that noise may be detrimental to health if the daytime noise emission level exceeds 65 dB(A). [1] A common dose-response curve is required for a quantitative risk assessment and the derivation of guidelines for public health noise policies. The risk estimates obtained from different noise studies can be summarized using the statistical approach of a meta-analysis. Classical, systematic and quantitative reviews have been published in the past, summarizing the results of studies that were carried out until the end of the last century. [2],[3] In a meta-analysis by Van Kempen et al. , it was concluded that the risk of ischemic heart disease (IHD) increased by 1.09 per 5 dB(A) increase of road traffic noise level (95% confidence interval (CI): 1.05-1.13, L day = 51-70 dB(A)) when two cross-sectional studies were considered. [3] However, the pooled estimate of two prospective studies was calculated to be 0.97 per 5 dB(A) (95% CI: 0.90-1.04, L day, 6-22 h = 51-70 dB(A)). Limiting the diagnosis of ischemic heart disease to myocardial infarction resulted in the inclusion of three studies in the meta-analysis. The linear effect estimate was 1.03 per 5 dB(A) increase in the road traffic noise level (95% CI: 0.99-1.09, L day, 6-22 h = 51-80 dB(A)). New studies have however appeared in the meantime.

Recently, the author published an update of his previous review from 2000 in the Journal 'Noise and Health', [4] which was based on a larger scientific report. [5] The present article refers to this review, which is used as a data source for a meta-analysis in order to derive a dose-response relationship between transportation noise and cardiovascular risk.

Source of information

By 2005, 61 epidemiological studies were recognized as having either objectively or subjectively assessed the relationship between transportation noise and cardiovascular endpoints. Nearly all of the studies referred to road traffic noise or commercial aircraft noise and a few to military aircraft noise. Most of the studies were of the cross-sectional type (descriptive studies) but there were also observational studies such as case-control and cohort studies (analytical studies). Confounding factors were not always adequately considered in some older studies. Not many studies provided information on dose-response relationships because only dichotomous exposure categories were considered. Study subjects were children and adults. Focusing on long-term effects of chronic noise exposure, effects on mean blood pressure readings in children were excluded from further evaluation because these might reflect short-term arousals of the autonomous system rather than arteriosclerotic changes of the vascular system. Thirty seven of the reviewed studies assessed the prevalence or incidence of manifest diseases including hypertension. The prevalence and incidence of diseases was either assessed in a self-administered questionnaire, a clinical interview or by clinical measurements.

Noise exposure

Noise exposure was divided into five dB(A)-categories for the daytime outdoor average A-weighted sound pressure level (L day,16 h : 6-22 h). This was considered in most of the studies, which were conducted before the European Noise Directive was in force. [6] Some aircraft noise studies used national calculation methods, e.g ., Dutch 'Kosten Units', Swedish 'FBN', Japanese 'WECPNL'. Information on night-time exposure (L night : 22-6 h or 23-7 h) was seldom available. Newer studies used nonweighted or weighted average noise levels (L Aeq , L dn , L den ) of the 24 hour day-evening-night sound exposure. Sound levels were converted on the basis of best-guess approximations to L day,16 h for comparison and pooling. In urban settings, night-time average noise levels (22-6 h) for road traffic tend to be approximately 7-10 dB(A) lower than daytime average noise levels and are relatively independent of the traffic volume of the street (no motorways), [7],[8] which means that L dn is a good approximation for L day,16 h . Measurements showed that L den was approximately 1-3 dB(A) higher than L day,16 h for differences where L day,16 h - L night,8 h ranged from 5 to 10 dB(A). [9] The following empirical formula was given for the conversion:

L den ≈ L day,16 h - 2*ln[(L day,16 h - L night,8 h )/22.4)]

In the above formula, if the difference between day and night emission sound levels is of the order of 7-8 dB(A), then this accounts for L den values which are higher than the L day,16 h levels by approx. 2 dB(A) units. This is commonly found for road traffic noise in urban streets. A conversion factor of 2 dB(A) was also suggested by others. [10] The differences between L den and L dn are usually small. The difference range, L den - L dn is between 0 and 1.5 dB depending on whether the noise level L Aeq drops in the evening. [11] Therefore, in epidemiological studies in which the relative effects of road traffic noise are studied, the sound emission during the daytime can be taken as an approximate relative measure of the overall sound emission including the night. This seems to be further justified because existing noise regulations usually consider a 10 dB(A) difference between the day and the night. However, this approximation can only be made with respect to urban road traffic noise. No such approximation can be made for motorways, train and aircraft noise. Approximate formulas for the conversion of different noise indicators are also given in the "Good Practice Guide for Strategic Noise Mapping". [12]

Cardiovascular endpoints

Thirty-seven of the studies comprising this review assessed the prevalence or incidence of manifest diseases including hypertension. The strongest evidence of an association between community noise and cardiovascular endpoints was found for ischemic heart diseases including myocardial infarction. Ischemic heart diseases (coded as 410-414 in ICD 9) included: 410: acute myocardial infarction, 411: other acute and subacute forms of ischemic heart disease, 412: old myocardial infarction, 413: angina pectoris, 414: coronary atherosclerosis. Most of these studies referred to road traffic noise and were carried out on males because the incidence rates of ischemic heart disease are higher in middle-aged male subjects. In general, relative risks were found to be higher when mediating exposure factors like residence time, room orientation and window-opening habits were considered in the analyses. Both the "objective" exposure (sound level) and the "subjective" exposure (annoyance) were associated with a higher risk of ischemic heart disease. The results were not as consistent for hypertension as for ischemic heart disease.

[Figure 1] shows relative risk of ischemic heart diseases (calculated as odds ratios, ORs) found in 17 studies (12 road traffic noise, five aircraft noise) of transportation noise as given in the publications (stratified by sex if the information was available). The data is taken from the major review. [4] The mean value for the 5 dB(A) noise level category given in the tables of reference was used to determine the x-axis value. If only single IHD components (see ICD-9 codes in the introduction) were given in a study, the most definite and clinically manifest ones (IHD ischemia or acute myocardial infarction) were considered for the graph. The study location, the year, the sexes (m = males, f = females, mf = males + females) and the sound source are given in the legend. The graph illustrates the heterogeneity of the results of different studies pointing to the fact that very strict criteria are needed to derive a generalized dose-effect curve. On the other hand, it shows a mostly consistent increase in risk with increasing noise level (qualitative argument). Most of the studies refer to road traffic noise (12 studies).

Evaluation of studies

All epidemiological noise studies were evaluated with respect to their feasibility in a meta-analysis. The following judgment criteria for the inclusion or exclusion in the analysis process were applied: (1) peer-reviewed in the international literature, (2) reasonable control of possible confounding (stratification, model adjustment (regression), matching), (3) objective assessment of exposure (sound level) and (4) objective assessment of outcome (clinical assessment). Additional criteria for the ranking were: (5) type of study (analytical or descriptive) and (6) multi-level dose-response assessment (not only dichotomous exposure categories).

The odds ratios calculated for the different 5 dB(A) noise categories (L day,16 h ) within a single study were then pooled between studies for each 5 dB(A) noise category. Since higher exposure categories usually consist of smaller numbers of subjects than the lower categories, regression coefficients across the whole range of noise levels within a study tend to be largely influenced by the lower categories. This may lead to an underestimation of the risk in higher noise categories. The approach used here pools the effect estimates of single studies within each noise category, thus giving true weight to the higher noise categories and accounting for possible nonlinear associations.

Based on the judgment criteria, five analytical (prospective case-control and cohort) and two descriptive (cross-sectional) studies were suitable for the derivation of a common dose-response curve for the association between road traffic noise and the risk of myocardial infarction. Two separate meta-analyses were made by considering the analytical studies that were carried out in Caerphilly and Speedwell (pooled six years' follow-up data) [13],[14] and Berlin ("Berlin I", "Berlin II", "Berlin III") [15],[16] on the one hand and the descriptive studies that were carried out in Caerphilly and Speedwell (cross-sectional studies) [17] on the other hand. All studies referred to the road traffic noise level during the day (L day, 6-22 h ) and the incidence (analytical studies) or prevalence (descriptive studies) of myocardial infarction as the outcome. All study subjects were men. In all analytical studies, the orientation of rooms was considered for the exposure assessment (at least one bedroom or living room facing the street). In all descriptive studies, the traffic noise level referred to the nearest facades that were facing the street and did not consider the orientation of rooms/windows. The individual effect estimates of each study were adjusted for the covariates given in these studies. This means that different sets of covariates were considered in each study. This pragmatic approach accounts best for possible confounding in each study and provides the most reliable effect estimates derived from each study.

The common set of covariates considered in the descriptive studies were age, sex (only males), social class, body mass index, smoking, family history of IHD, physical activity during leisure and prevalence of preexisting diseases. The common set of covariates considered in the analytical studies were age, sex (only males), social class, school education, employment status, shift work, smoking and body mass index. Some of the analytical studies also considered physical activity during leisure, family history of IHD, prevalence of preexisting diseases, work noise and marital status. In one study, the effect estimates were further adjusted for hypertension, diabetes mellitus and myocardial infarction. This may be considered a conservative approach due to overcontrolling because according to the reaction model, biological (risk) factors are mediators along the pathway from exposure (noise stress) to disease.

Dose-effect relationship (pooled data)

[Table 1] shows individual and pooled effect estimates with confidence intervals (rounded brackets), statistical weights (square brackets) for the individual studies and the Q-test of heterogeneity between studies. [5] According to the Q-test, the nil-hypothesis of nonheterogeneity was never discarded.

[Figure 2]a and b show the results of the meta-analyses (pooled estimates and 95% confidence intervals) for the descriptive and analytical studies. Whereas the cross-sectional studies [Figure 2]a cover the sound level range of L day,16 h from >50-70 dB(A), the cohort and case-control studies [Figure 2]b cover only the range from ≤60-80 dB(A). However, both curves together can serve as a basis for interpretation. From [Figure 2]a, it can be seen that no noticeable increase in myocardial infarction risk can be detected ≤60 dB(A) for L day,16 h . The myocardial infarction risk increases for noise levels >60 dB(A) and is >1.2 for noise levels of 70 dB(A) as seen in [Figure 2]b. Fixed-effect meta-analyses of a linear trend (STATA 9, proc METAREG) revealed an odds ratio per 10 dB(A) of OR = 1.17 (95% CI = 0.87-1.57, P = 0.301).

The confidence intervals of the effect estimates shown in [Figure 2]a and b for the association between traffic noise and myocardial infarction include relative risks of 1.0. It should be noted in this respect that the purpose of the meta-analysis was not significance testing. Rather, it was anticipated to derive a 'best-guess' pooled dose-effect relationship that can be used for a quantitative risk assessment. Individual studies however, showed significant ( P P [15],[16] When the meta-analysis is carried out for subsamples of subjects that had lived for at least 10 or 15 years in their dwellings (dependent on the data given in the reference), larger effect estimates are also obtained in the meta-analysis. The result is shown in [Figure 2]c. The lower confidence intervals are very close to an odds ratio of 1.0. When the upper two noise categories are combined, the pooled effect estimate is OR = 1.25 ( P = 0.068) in the total sample and OR = 1.44 ( P = 0.020) in the subsample, the latter being statistically significant. Meta-analyses of a linear trend carried out in the sub-sample revealed an odds ratio per 10 dB(A) of OR = 1.44 (95% CI = 0.97-2.12, P = 0.067). For the calculation of population-attributable risk percentages however, the weaker effect estimates referring to larger numbers of the total sample must be used because information about residence time is usually not available on a population level.

Dose-effect curve (polynomial fit)

A polynomial function was fitted through the data points shown in [Figure 2]b to generate a continuous exposure-response curve [Figure 3]a that can not only be applied to categorized noise data but also to continuous noise data. The data points are weighted by the number of subjects ("N-weighting"). 96% of the variance (R 2 ) is explained by the polynomial function. Mean category values of the decibel-x axis are considered for the calculation. For the reference category " ≤60 dB(A)," a value of 55 dB(A) was used because this category also includes a large number of noise levels below 55 dB(A). Changing this value to other levels, e.g ., 52.5 or 57.5 had only a very marginal impact on the coefficients and the "fit statistics". In urban settings, background levels during the day do not often fall below 50 dB(A). The exposure-response function refers to road traffic noise and for historical reasons and the way studies were designed, to the daytime noise indicator L day,6-22 h . It is defined for noise levels ranging from 55 to approx. 80 dB(A).

Exposure-response function

OR = 1.629657 - 0.000613(L day,16 h ) 2

+ 0.000007357(L day,16 h ) 3 , R 2 = 0.96

The analytical studies were chosen for the risk curve because of their generally accepted higher credibility with respect to causal inference. However, both curves support each other very well. This is illustrated in [Figure 3]b where both descriptive and analytical studies were considered together for one polynomial fit. The development of exposure-response curves, in general, is a dynamic process.

According to the European Directive on the assessment and management of environmental noise, [6] the noise indicators L den and L night shall be assessed for noise mapping in European member states. L den consists of the L day,12 h , L evening,4 h (+5 dB weighting) and L night,8 h (+10 dB weighting). Approximate conversion to other noise indicators can be made under certain assumptions and conditions. For example, if only the aggregated L den data is available for road traffic noise, L den - 2 dB(A) should be considered in the exposure-response equation for the calculation of risk. Note: Although L day,16 h is the nonweighted energetic average of L day,12 h + L evening,4 h , both are necessary determinants of L den . In the future, conversions from L day to L den can be made when widespread noise assessments data are available for Europe according to the EU directive on the assessment and management of environmental noise. [6] However, one has to bear in mind that the weights which are used for the calculation of L den may not adequately reflect the physiological response to noise. In general, in health effects research, nonweighted exposure indices appear to be more appropriate.


The biological plausibility of the hypothesis of the effects of noise is well-documented. Acute noise effects have been studied extensively during the '60s to '80s of the last century and a general noise reaction model was well established before noise effects research moved from the laboratory into the field to test epidemiological noise hypotheses with respect to long-term effects of noise.

The current approach to assess common effect estimates or dose-effect curves obtained from different studies is different from the regression-oriented approach. Van Kempen et al. [3] calculated uniform regression coefficients across all noise categories within individual studies, which were then pooled for all studies ('regression approach'). On the other hand, this author [5] calculated pooled relative risks for individual noise categories from different noise studies, which were then considered for a dose-response relationship ('categorial approach'). Both approaches have advantages and disadvantages. The regression approach has the advantage that regression coefficients can easily be pooled regardless of actual noise levels; only the slope (regression coefficient) of the dose effect curve is taken into account. For example, one aircraft noise study showed high risks at relatively low noise levels [18] while another showed an increased risk only at higher noise levels. [19] Presuming that the relative relationships within the studies are unbiased, it may be possible to pool the studies on the basis of their regression coefficients (slopes) although the noise assessment may not be comparable in absolute terms. The disadvantage of the regression approach is that the linear regression does not account for nonlinear associations and possible thresholds of effects, particularly when the regression lines of different studies include different ranges of noise levels.

The categorial approach is noise level-oriented. Only relative risks from different studies referring to the same noise category are pooled to derive a dose-effect curve. This has the advantage that possible thresholds of effects can be determined. Furthermore, this approach accounts for nonlinear associations. It is less likely to obscure possibly higher risks in higher noise categories where the numbers of subjects are relatively small, which is usually the case given the empirical noise distribution in populations. For example, in case of j-shaped or quadratic associations, an overall regression coefficient underestimates the risks in higher noise categories simply because the slope of the regression line is primarily determined by the larger numbers of subjects in the lower exposure categories where effects may be smaller. The disadvantage of the categorial approach is that it relies on relatively homogeneous and comparable noise indicators in order to pool effect estimates from different studies within the noise categories. Considering the two studies mentioned above, it would be difficult to pool both studies on the basis of noise levels. The noise assessment in studies may differ due to various methodological reasons, e.g. , measurement vs. modeling, different calculation methods, different time period considered, inclusion of weighing factors, different distances of reference point, different sides of the house, etc.

In any case, it is essential that critical decisions are made as to which studies should be included in a meta-analysis. Studies that are not suitable with respect to issues of exposure misclassification, selection bias and observation bias or confounding should be excluded from the meta-analysis, which is the basis for a quantitative risk assessment. Only a very few epidemiological studies were considered in the pooled dose-response curve. With the exception of one new study, [16] these studies were also identified in the previous Van Kempen meta-analysis. However, the cross-sectional data of one study was excluded, because the outcome was assessed only by a postal questionnaire. [15] In the present meta-analysis, a distinction was made between prevalence (obtained from cross-sectional studies) and incidence (obtained from case-control and cohort studies) of disease. The validity of prospective studies is usually considered higher than that of cross-sectional studies. In all studies, L Aeq,6-22 h was assessed in a comparable and reliable way.

The exposure-response curve derived from male study subjects may be generalized for both sexes. The subjects in the noise studies were mostly men due to considerations of statistical power in the study design. Cardiovascular diseases are more frequent in middle-aged male subjects. [20] Females did not show a noise effect in one study (1/4 of the study sample). [16] However, the hormonal/menopausal status was not assessed, which could act as a confounder and dilute the effect. [21] For reasons of homogeneity, the relatively small number of females was excluded from the calculation of pooled effect estimates. In laboratory studies, the focus was primarily on "before-after" effects of noise exposure in the same test persons rather than on gender differences. In the large body of occupational noise studies, gender was often considered as a confounding factor but not as a potentially effect-modifying factor in the statistical analyses. Male blue collar workers were predominantly found in high-noise workplaces. Studies on the association between environmental noise and high blood pressure showed no consistent pattern with respect to higher relative risks in either men or women. [5] Although there are differences in the absolute risk between males and females, it seems to be reasonable to assume that in relative terms, females may be just as affected by noise stress as males as long as all confounding or modifying factors are considered. However, the possible effect-modifying impact of gender should be assessed in future research. The dose-response curve shown here will be regularly updated with respect to information coming from new studies.


1EEA. Traffic noise: exposure and annoyance. Available from: [Last accessed on 2004 Jun 9]. Copenhagen: European Environmental Agency: 2001.
2Babisch W. Traffic noise and cardiovascular disease: Epidemiological review and synthesis. Noise Health 2000;2:9-32.
3Kempen EE, Kruize H, Boshuizen HC, Ameling CB, Staatsen BA, Hollander de AE. The association between noise exposure and blood pressure and ischaemic heart disease: A meta-analysis. Environ Health Perspect 2002;110:307-17.
4Babisch W. Transportation noise and cardiovascular risk: Updated review and synthesis of epidemiological studies indicate that the evidence has increased. Noise Health 2006;8:1-29.
5Babisch W. Transportation noise and cardiovascular risk: Review and synthesis of epidemiological studies, Dose-effect curve and risk estimation. WaBoLu-Hefte 01/06. [Last accessed on 2004 Apr 10]. Available from: = KennummerandSuchwort = 2997. Dessau: Umweltbundesamt; 2006.
6Directive 2002/49/EC. Directive of the European Parliament and of the Council of 25 June 2002 relating to the assessment and management of environmental noise. Official J Eur Commun 2002;189:12-25.
7Evans GW, Lercher P, Meis M, Ising H, Kofler WW. Community noise exposure and stress in children. J Acoust Soc Am 2001;109:1023-7.
8Ullrich S. Lürmbelastung durch den Straβenverkehr. Z Lδrmbekδmpfung 1998;45:22-6.
9Bite M, Bite PZ. Zusammenhang zwischen den Straβenverkehrslδrmindizes LAeq(06-22) und LAeq(22-06) sowie Lden. Zeitschrift für Lδrmbekδmpfung 2004;51:27-8.
10Gjestland T. Lden conversion. Personal Communication 2006.
11Miedema HM, Oudshoorn CG. Annoyance from transportation noise: Relationships with exposure metrics DNL and DENL and their confidence intervals. Environ Health Perspect 2001;109:409-16.
12European Commission Working Group Assessment of Exposure to Noise (WG-AEN). Good practice guide for strategic noise mapping and the production of associated data on noise exposure, Version 2, 13 January 2006. Position paper, final draft. WG-AEN 016.2003.doc, Available from: [Last accessed on 2006 Nov 8].
13Babisch W, Ising H, Gallacher JE, Sweetnam PM, Elwood PC. Traffic noise and cardiovascular risk: The Caerphilly and Speedwell studies, third phase - 10 years follow-up. Arch Environ Health 1999;54:210-6.
14Babisch W, Ising H, Gallacher JE. Health status as a potential effect modifier of the relation between noise annoyance and incidence of ischaemic heart disease. Occup Environ Med 2003;60:739-45.
15Babisch W, Ising H, Kruppa B, Wiens D. The incidence of myocardial infarction and its relation to road traffic noise - the Berlin case-control studies. Environ Int 1994;20:469-74.
16Babisch W, Beule B, Schust M, Kersten N, Ising H. Traffic noise and risk of myocardial infarction. Epidemiology 2005;16:33-40.
17Babisch W, Ising H, Elwood PC, Sharp DS, Bainton D. Traffic noise and cardiovascular risk: The Caerphilly and Speedwell studies, second phase. Risk estimation, prevalence and incidence of ischemic heart disease. Arch Environ Health 1993;48:406-13.
18Rosenlund M, Berglind N, Pershagen G, Jδrup L, Bluhm G. Increased prevalence of hypertension in a population exposed to aircraft noise. Occup Environ Med 2001;58:769-73.
19Matsui T, Uehara T, Miyakita T, Hitamatsu K, Osada Y, Yamamoto T. The Okinawa study: Effects of chronic aircraft noise on blood pressure and some other physiological indices. J Sound Vibration 2004;277:469-70.
20Yusuf S, Reddy S, Ounpuu S, Anand S. Global burden of cardiovascular diseases, Part I: General considerations, the epidemiologic transition, risk factors and impact of urbanization. Circulation 2001;104:2746-53.
21Farley TM, Meirik O, Chang CL, Poulter NR. Combined oral contraceptives, smoking and cardiovascular risk. J Epidemiol Commun Health 1998;52:775-85.