Several studies have indicated that exposure to noise increases the risk of hypertension.1 2 The association with hypertension is relatively well established for occupational noise exposure as pointed out in recent reviews1^–3 and original papers,4 and some studies also support an effect of aircraft noise.1 2 5 Babisch1 identified approximately 10 studies examining the possible risk of hypertension from exposure to calculated road traffic noise, and concluded that the results were heterogeneous. However, studies performed in the last couple of years show significant positive associations between road traffic noise and hypertension.5^–7 Epidemiological studies on hypertension and road traffic noise generally have a cross-sectional design and do not include estimates of incidence of hypertension. Information on the temporal relationship between noise exposure and diagnosis is usually missing, and information on possible confounders is sometimes limited.1

Discussions on possible mechanisms behind an increased risk of hypertension from community noise have focused on physiological stress reactions with short-term changes in cardiovascular physiology and levels of stress hormones, as shown in short-term laboratory studies.2 3 8 Similar mechanisms may affect the risk of ischaemic heart disease.9

In the present study we report the prevalence of self-reported physician-diagnosed hypertension in a cross-sectional questionnaire study with detailed calculations of traffic-generated noise levels at the subjects’ residences, as well as information on major risk factors for hypertension. Using data on the year of moving into the present dwelling, and the year of diagnosis, we also estimated the incidence of hypertension. The technique of retrospective analysis of incidence based on information of year of onset in a cross-sectional survey has been used previously for other disease entities.10^–12

METHODS

Study area and noise levels

A detailed description of the study area and the methods for calculation of noise levels has been given elsewhere.13

The study area is part of the municipality of Lerum, which is east of Gothenburg, Sweden. Most of the population live along two major traffic routes and are exposed to noise from both road traffic from a highway (about 20 000 vehicles/24 h) and a major railway (about 200 trains/24 h). The study area is approximately 20 km by 6 km and consists largely of residential areas of detached, terraced and apartment houses.

As a first step, noise exposures in Lerum from different sources (road, rail and air traffic, as well as certain stationary noise sources) were mapped in 2003. The calculations were carried out by applying a grid with a distance of 15 m between each calculation point. Next, a geographical information system (GIS)-based method was used to determine the noise exposure adjacent to each residential building. Calculations of sound levels from road traffic and railway noise on the most exposed side were provided for each residential building using a validated model.14 All calculation points were determined at 2 m above the ground as free field values. For each type of noise source, several sound level metrics were calculated, but in the present study we used the A-weighed 24 h average, L_Aeq,24h.

Population, questionnaire and definition of hypertension

After determining the sound levels for all residential buildings, we linked this data file to information from the Land Registry which resulted in noise exposure data for a large number of dwellings that were linked to the local population registry. We selected individuals between 18 and 75 years of age who had resided at their present address for at least 6 months and who lived in dwellings with outdoor sound levels of L_Aeq,24h ⩾45 dB from rail and road traffic. The population sample was then divided into 5 dB categories based on outdoor sound levels from separate sources (L_Aeq,24h 45−50, 51−55, 56−60 and 61−70 dB). All individuals between 18 and 75 years of age were selected in each noise category except for the 45−50 and 50−55 dB categories where only half of the residents were randomly selected.

In total 2747 residents were thus selected for a postal questionnaire study. Questionnaires were posted in the spring of 2004, and 1953 persons (71%) responded. Table 1 shows the distribution of respondents stratified by road traffic noise exposure categories. For railway noise, the numbers of respondents were 925, 507, 331 and 190 in the respective noise categories (data not shown).

The questionnaire evaluated perceived symptoms and effects of traffic noise (eg, annoyance, sleep disturbance, general well-being). Furthermore, questions on various background factors were included: type of dwelling (detached house or apartment), year of moving into the dwelling, smoking (ever/never), weight, height, education (<12 or ⩾12 years of schooling) and occupational exposure to noise (yes/no). Results for annoyance, effects on activity, sleep disturbance and general well-being have been reported elsewhere.13

Questions on hypertension were: “Did you ever have hypertension diagnosed by a physician?”, “If so, in which year was hypertension diagnosed?”, “Do you take any antihypertensive drug?” and “Does/did any of your parents or siblings have hypertension?”.

Descriptive data on the population, as well as various background factors and possible confounders, stratified by road traffic noise, are shown in table 1. The differences in age between road traffic noise categories were small. However, in the strata with high noise levels there were more smokers and fewer respondents owned their own dwelling.

Data analysis

The prevalence of hypertension or use of antihypertensive drugs was calculated for different categories of road and railway noise. In addition, some analyses were performed using noise (L_Aeq,24h) as a continuous variable. Prevalence ratios for hypertension or use of antihypertensive drugs were calculated with 95% confidence intervals (CIs). Prevalence odds ratios were estimated using logistic regression models including noise categories and covariates. We initially included covariates associated with hypertension in the literature, namely, age, sex, body mass index (BMI), heredity for hypertension, ever smoking, education (<12 vs ⩾12 years’ schooling) and occupational noise exposure. However, based on a limited impact (p>0.1 and change in the risk estimates for road traffic noise of less than 10%) of smoking, education and occupational noise, the only covariates included in the final model were age, sex, BMI and heredity for hypertension (see Results section).

Incidence rates of hypertension were calculated for each noise stratum as the number of new cases of hypertension divided by the number of person-years under observation. The analysis was restricted to hypertension diagnosed after 25 years of age. Person-years were calculated from the year of moving into the current dwelling until diagnosis of hypertension or the end of follow-up (2004). Incidence rate ratios were calculated using the lowest noise category (L_Aeq,24h 45–50 dB) as reference category. Approximate 95% confidence limits for incidence rate ratios were calculated using the Wald method. In models including covariates as above, the relative risks for incident hypertension (after 25 years of age) were estimated using Poisson regression. Hazard ratios were calculated for the same models using Cox regression. Statistical analyses were performed using SAS 9.1 package (SAS Institute, Cary, North Carolina).

The study was approved by the ethics committee of the University of Gothenburg.

RESULTS

The overall prevalence of hypertension and use of hypertensive drugs is shown in figs 1 and 2.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Prevalence (SD) of self–reported hypertension and use of antihypertensive drugs, stratified by exposure to road traffic noise, in a questionnaire study of residents in Lerum, Sweden (n = 1953). Noise levels are expressed in L_Aeq,24h.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

Prevalence (SD) of self–reported hypertension and use of antihypertensive drugs, stratified by exposure to railway noise, in a questionnaire study of residents in Lerum, Sweden (n = 1953). Noise levels are expressed in L_Aeq,24h.

The prevalence ratio for use of antihypertensive drugs was 1.38 (95% CI 1.03 to 1.85) in the highest road noise category (56–70 dB) with the lowest (45–50 dB) as reference (table 2). The trend towards an exposure–response relationship was statistically significant (p = 0.03) for antihypertensive drugs but not for hypertension. An analysis stratified by sex showed that the positive association between road traffic noise and hypertension or use of antihypertensive drugs was mainly found in men (fig 1 and table 2). There was no similar trend for railway noise or combined noise. The prevalence ratios (men and women) for hypertension and use of antihypertensive drugs in the 51–55 dB category of railway noise were 0.81 (95% CI 0.61 to 1.07) and 0.82 (95% CI 0.60 to 1.12), respectively, with the lowest category (45–50 dB) as reference. In the highest railway noise category (56–70 dB), the prevalence ratios were 0.90 (95% CI 0.69 to 1.18) and 0.86 (95% CI 0.64 to 1.17).

As would be expected, several of the background factors (age, sex, BMI and heredity for hypertension) showed statistically significant associations with prevalence of hypertension. The impact of smoking, type of dwelling or education was small or non-existent.

Table 3 shows the prevalence ORs for hypertension when age, sex, heredity for hypertension and BMI were included in multiple logistic regression models together with road traffic noise. When the other covariates (smoking, education, occupational noise) were entered in the model, their regression coefficients were not statistically significant and in nearly all strata they changed estimates of road traffic noise by <10%. The ORs increased in those diagnosed with hypertension >10 years after moving into their present dwelling (table 3). There was a significant interaction between sex and exposure to high road traffic noise on the risk of hypertension, and this effect was mainly seen in men. For men diagnosed with hypertension >10 years after moving into their present dwelling, the trend of increasing ORs with road traffic category was statistically significant (linear trend, p = 0.04), while this was not the case for women or the combined group of both genders. Dividing the highest road noise category further resulted in small numbers, but for men diagnosed with hypertension >10 years after moving into their present dwelling, the adjusted ORs for hypertension were 2.8 (1.1 to 7.1) in the 56–60 dB stratum and 9.8 (2.8 to 34) at 61–70 dB. There were, however, only seven cases of hypertension (with data on covariates) in the highest road noise category.

When road traffic noise (L_Aeq,24h) was included as a continuous variable, results were consistent with those shown in table 3 in the sense that the overall impact of noise was limited and not statistically significant (point estimate 12% increase in the OR with a 10 dB increase in road traffic noise), while a significant increase in risk was shown in subjects with hypertension diagnosed after >10 years in the dwelling (point estimate 73% increase in the OR with a 10 dB increase in road traffic noise). For all subjects combined, there was a significant interaction between road traffic noise and sex, consistent with the stratified results shown in table 3.

Given the lack of an exposure–response relationship between the prevalence of hypertension and railway noise category (see above), no stratified analyses were performed for railway noise. In a model including railway noise categories and the same covariates as for road traffic noise, the OR (all men and women) was 0.8 (95% CI 0.5 to 1.1) for railway noise of 51–55 dB, and 1.0 (95% CI 0.7 to 1.5) for railway noise of 56–70 dB, using 45–50 dB as reference category.

The incidence rates of hypertension for men are shown in table 4. After a latency period of >1 year, there was a moderately increased incidence rate ratio in the highest noise category. When allowing for >10 years of latency, a more obvious increase in the incidence rate ratio was found in the two highest noise categories.

Results from Poisson regression models including only age and road traffic noise stratum showed, as would be expected, relative risks similar to the incidence rate ratios in table 4. When including the same covariates as in the logistic regression models (age, BMI and heredity for hypertension), the risk estimate in the highest road traffic noise category increased somewhat, showing a threefold increased risk of hypertension, with RR 2.9 (1.4 to 6.2). As indicated above for the logistic regression models, dividing the highest road noise category further resulted in small numbers. For subjects diagnosed with hypertension >10 years after moving into their present dwelling, the RRs for hypertension in the adjusted model were 2.4 (1.0 to 3.6) in the 56–60 dB stratum and 5.3 (1.9 to 15) at 61–70 dB.

Results from survival analysis using Cox regression models showed results very similar to those of the Poisson regression models (data not shown). Results were similar when analyses were repeated in the subgroup of the population aged >50 years. About 80% of the cases of hypertension had been diagnosed after 50 years of age.

DISCUSSION

The present study shows a positive association between calculated road traffic noise at their residence and risk of hypertension among men. The association was at least as strong when important potential confounders were included in the model. The prevalence ORs in the highest road traffic noise strata increased when the analysis was restricted to those who had lived in their present dwelling with high road traffic noise for a long time, especially when a latency period before the year of diagnosis of hypertension was allowed for. This was true also for the calculations of incidence rate ratios and relative risks in multivariate Poisson regression models, and it strengthens the case that there is a causal association between long-term exposure to road traffic noise and the risk of developing hypertension.

A merit of the present study compared with some previous cross-sectional studies is that the road traffic noise outside all individual dwellings was calculated using high-resolution GIS techniques and methods separating road traffic noise from other noise sources. Risk factors which could be possible confounders or effect modifiers with respect to hypertension were assessed. Also, the questionnaire design which enquired not only about physician-diagnosed hypertension but also year of diagnosis, as well as year of moving into the present dwelling, made it possible to conduct a more sophisticated analysis with assessment of a time window during which physiological reactions may have developed into hypertensive disease. This technique also made it possible to estimate the incidence of hypertension, making use of subjects free of hypertension at baseline when they moved into their present dwelling. This method has been used previously for other conditions,10^–12 and the validity of such retrospective information seems acceptable.12 15 For migraine, bias was found regarding age of onset, with subjects reporting onset of migraine at a later age than it actually occurred.10 For asthma, however, age of onset seems to be rather accurate.16

Although we knew the time relationship between exposure and disease and could perform a retrospective analysis, this does not eliminate all the disadvantages of a cross-sectional study. Selection bias cannot be excluded, since some people may move away from areas highly affected by road traffic noise. However, if this was the case, the impact of noise on hypertension would be underestimated. A study of incident hypertension in relation to aircraft noise was recently reported,17 but we found no such studies for road traffic noise.

In our opinion, information bias should be relatively small in the present study. Exposure classification was objective and performed without knowledge of the outcome. One limitation is that noise was mapped in 2003, while some of the cases of hypertension were diagnosed much earlier. However, more than two thirds of the cases of hypertension were diagnosed in or after 1995, and road traffic has not changed much since then. In addition, a reduction in traffic volumes of 50% would only change noise levels by 3 dB. The primary outcomes were physician-diagnosed hypertension and antihypertensive medication. These characteristics are relatively distinct and there was high agreement between them. It is, however, possible that some subject used other drugs to lower blood pressure than those primarily used as antihypertensives. The questionnaire also included a large number of other questions (on noise annoyance, sleep, etc). We find it highly unlikely that people with high levels of road traffic noise should falsely report physician-diagnosed hypertension. However, we performed no clinical examinations of the subjects; measuring blood pressure is not an optimal method of identifying subjects with physician-diagnosed hypertension, since the aim of pharmacological treatment is to keep blood pressure within the normal range. On the other hand, the criterion of physician-diagnosed hypertension will not catch subjects with undiagnosed hypertension. A validation of the diagnosis of self-reported hypertension in the US NHANES III study showed a sensitivity of 71%, and a specificity of 90%. With the prevalence of hypertension in that population, the positive predictive value was 72% and the negative predictive value was 89%.18 Given the good primary health care system in Sweden, we believe that the validity in our study area is somewhat better. Nevertheless, some misclassification is inevitable and would lead to a bias towards the null, and this would also be the effect of undiagnosed cases of hypertension.

Our findings of increased risk of hypertension from high road traffic noise are in agreement with several recent studies. In a large population-based cross-sectional study in the Netherlands, the use of antihypertensive medication was compared with residential noise exposure (L_den, with increased weighting for evening and night noise levels).7 Although there was no clear linear response across noise exposure strata, the OR for use of antihypertensive medication was significantly increased for subjects with L_den >55 dBA in models including a number of potential confounders. The risk was similar in men and women, and clustered in the 45−55-year-old age group. In the cross-sectional HYENA study, populations living near airports in six European countries were studied. A significant effect on hypertension was found for aircraft noise as well as for road traffic noise. The association with road traffic noise was stronger in men, and stratification by residential noise levels (L_Aeq,24h) showed ORs of 1.3 to 1.5 for noise categories of ∼50 dB and upwards. In a smaller, Swedish, cross-sectional study, there was a linear increase in the OR for hypertension with increasing residential noise levels (L_Aeq,24h).6 The effect was found only among those who had lived at their residence for >10 years. The response was clearer among women, although the sex difference was not statistically significant. Previous studies (before 2006) on the association between road traffic noise and hypertension have shown more heterogeneous results, as reviewed by Babisch.1

In the present study, an association between hypertension and road traffic noise was found only in men. The reason for this is unclear, but Babisch et al9 found an effect of road traffic noise on the risk of myocardial infarction only in men, and, as mentioned above, this was the case also for the risk of hypertension from high road traffic noise in the HYENA study.5

We found no association between railway noise and hypertension. In the daytime, the noise from trains is more intermittent than road traffic noise. Nevertheless, such intermittent noise from aircraft has been discussed as a risk factor for hypertension, especially when the noise events occur at night.5 19 Noise levels from train traffic were, however, relatively moderate13 compared with noise levels from aircraft reported around major European airports.5 We found only one previous study on railway noise and hypertension, a relatively small cross-sectional study performed in Sollentuna, Sweden,20 in which no association was found between the prevalence of self-reported hypertension and calculated noise levels (L_Aeq,24h).

The covariance between road traffic noise and air pollution is well known, and the potential effects on cardiovascular morbidity of these two factors may be difficult to disentangle. The fact that a Dutch cohort study21 showed a closer association between mortality and distance to major roads than between mortality and estimated air pollution levels was interpreted by Babisch1 as an indication that road traffic noise could also be a causative factor. Few studies have attempted to separate noise and air pollution levels. In their cross-sectional study, de Kluizenaar et al7 adjusting for estimated PM₁₀ concentrations, found that the associations between road noise exposure and risk of hypertension remained. In the present study we had no access to modelled air pollution data.

The results indicate an increased risk of hypertension at noise levels L_Aeq,24h >55 dB. A large number of people are exposed to such road noise levels. For example, in Sweden 15−20% of the population, and an even larger percentage in the rest of Europe, are exposed to road traffic noise at this level.22 In the recently presented report from the HYENA study, the risk of hypertension in men was increased in the road traffic noise stratum L_Aeq,24h 53−57 dB.5

In conclusion, the present study indicates that long-term residential exposure to road traffic noise L_Aeq,24h >55 dB increases the risk of hypertension in men. Such noise levels are common, and hypertension is a common disease and an independent risk factor for stroke and myocardial infarction. Therefore, if the association between road traffic noise and hypertension is causal, the impact on public health may be substantial.