Association Between Noise Pollution and Prevalent Ischemic Heart Disease
Association Between Noise Pollution and Prevalent Ischemic Heart Disease
Association Between Noise Pollution and Prevalent Ischemic Heart Disease
1515/folmed-2016-0041
Correspondence: Background: Noise pollution is considered a risk factor for ischemic heart disease
Angel М. Dzhambov, Department (IHD). Both are highly prevalent in Bulgaria, but their association has not been
of Hygiene and Ecological Medi- studied sufficiently.
cine, Faculty of Public Health, Medi-
cal University of Plovdiv, 15A Vassil Aim: The aim of the present study was to examine the risk of IHD associated with
Aprilov Blvd., 4002 Plovdiv, Bulgaria road traffic (Lden) and lifetime occupational noise exposure (LONE) in a Bulgarian
E-mail: [email protected] sample.
Tel: +35932580562 Materials and methods: A cross-sectional survey was carried out among 513
Received: 06 Sept 2015 residents of Plovdiv, Bulgaria. A questionnaire asked about doctor-diagnosed
Accepted: 08 July 2016 IHD, LONE and confounding factors. Lden was derived from official noise map after
Published Online: 18 Oct 2016 geocoding participants’ addresses. In log-link Poisson regressions we investigated
Published: 23 Dec 2016 the relative risks of prevalent IHD. Sensitivity analyses examined subgroup differ-
ences.
Key words: road traffic noise; oc-
cupational noise, ischemic heart Results: Lden ≥ 65 dB was associated with higher risk (RR=1.84, 95% CI: 0.61, 5.57)
disease, coronary heart disease; of IHD in long-term residents (≥ 20 years). LONE was associated with RR=1.76
cardiovascular disease (0.82, 3.78) for ever-exposed; and RR=2.35 (1.00, 5.52) for 15 - 47 years exposure.
Citation: Dzhambov AM, Dimitro- Conclusions: Exposure to Lden≥65 dB was associated with non-significantly
va DD. Association between noise higher risk of IHD. Longer LONE was consistently associated with higher risk. In
pollution and prevalent ischemic some subgroups the effect of noise was more pronounced.
heart disease.
Folia Medica 2016;58(4):273-281
doi: 10.1515/folmed-2016-0041
myocardial infarction were estimated at about € 11.6 loud enough to disturb normal conversations?”; after
million.12 According to the Sixth European Working geocoding participants’ addresses, we estimated the
Conditions Survey, in 2015, 28.8% of the Bulgarian Euclidean distance to the nearest major road (> 10
workers were exposed to noise so loud that they 000 vehicles/24h) based on satellite imagery (Google
would have to raise their voice to talk to people EarthTM) as a proxy for traffic-related air pollution;
during at least one fourth of the time.13 Neverthe- day-evening-night equivalent sound level (Lden) was
less, few studies have explored the effects of noise extracted from official noise map of Plovdiv.16
on IHD. Tzenova et al.14 surveyed 1062 residents • factors relevant to IHD (body-mass index,
of Sofia and found significantly higher prevalence diagnosis with type 2 diabetes mellitus and arterial
of IHD (17.38% vs. 10.56%) in those exposed to hypertension, pack-years of smoking)
residential Leq 6-22 h>60 dB for 21 years, from which • other potential confounders: noise sensitiv-
we could calculate RR=1.65 (95% CI:1.21,2.24). ity17, sleep disturbance (“0, very good” to “10,
Dimitrova and Karaslavova reported no ef- cannot sleep at all”), bedroom location (having a
fect of self-reported exposure to “excess produc- noisy façade or not) and duration of residence at
tion noise” on myocardial infarction among men the current address
(OR unadjusted =0.680, 95% CI:0.379,1.220), but
SAMPLE SIZE
significantly higher odds (ORunadjusted=4.01, 95%
CI:0.84,19.11) among women.15 Given this insuf- An a priori sample size was estimated using G*Power
ficient evidence, the aim of the present study was v.3.1.9.2. For Lden≥65 dB it was based on the risk
to examine the risk of IHD associated with road estimate and the base rate of IHD in the non-exposed
traffic and occupational noise exposure in a Bulgar- group reported by Tzenova et al.14 – 817 for 80%
ian population. power and 482 for 60% power. For long-term LONE
the calculations were based on the risk of CHD in
MATERIALS AND METHODS long-term exposed workers reported by Gan et al.6
and the same base rate from Tzenova et al.14–391
DESIGN AND VARIABLES
for 80% power. A final sample of 513 residents
This was a cross-sectional study carried out between was included in the analyses, after excluding those
July and November, 2014 in Plovdiv, Bulgaria. meeting exclusion criteria.
Participants were sampled via two procedures – a
non-probability snowball method and field interviews STATISTICAL ANALYSIS
(see ref. 16). Briefly, in the snowball sampling, the Variables were initially screened for patterns of miss-
initial seeds were asked to recruit other members ing data, univariate normality (D’Agostino-Pearson
of their social network to ensure continuity of the K2 test) and outliers (modified outlier labeling rule).
procedure. In the field survey, the authors selected The variable with highest percentage of missing
different neighborhood blocks representative of values (40.9%) was Lden due to underreporting of
the geomorphological and land-use urban fabric residential addresses and it was not missing com-
of Plovdiv; the first author visited those sites and pletely at random. Outliers were kept if they were
invited local residents to participate. Current resi- considered true and honest answers. Ethnic minor-
dence in Plovdiv and age 18 years or older were ity members were combined into one group due to
the inclusion criteria; those reporting hearing loss the lower number of Armenian, Roma, and Jewish
or uncorrected hearing impairment were excluded. participants, whereas the Turks were prevalent.
Participants completed a questionnaire compris- Descriptive statistics were computed. In case
ing questions about: of non-normally distributed/categorical variables
• demographics: age, gender, ethnicity, highest we used Spearman correlations and Pearson Chi-
educational attainment, marital status, occupation, Square/Fisher’s Exact Test/Fisher-Freeman-Halton
perceived socio-economic status Test. Welch’s t-test and ANOVA were used even
• self-reported doctor-diagnosis with IHD with non-normally distributed interval data due
• environmental exposures – lifetime occupa- to their robustness to violations of normality and
tional and residential noise exposure: “For how long homogeneity of variance.18,19
during your lifetime have you worked at a place The main analyses investigated the risk of preva-
where noise was loud enough to disturb normal lent IHD associated with lifetime occupational noise
conversations?” and “For how long during your exposure (LONE) and Lden exposure at the current
lifetime have you lived in a place where noise was address. We conducted log-link Poisson regressions
to approximate the relative risk (RR).20,21 LONE 95% CI: 0.18, 0.46), pack years of smoking (rs =
was included as categorical variable (ever vs. never 0.20, 95% CI: 0.05, 0.35), and sleep disturbance (rs
exposed and different quartiles of exposure vs. = 0.19, 95% CI: 0.04, 0.34). It was longer in men
never exposed); Lden was dichotomized (≥65 dB and pensioners (data not shown). Lden levels in the
vs.<65 dB), according to the threshold relevant to dataset ranged from 50 to 80 dB. One hundred and
cardiovascular diseases (WHO, cited by Ref. 12). ninety-seven (38.40%) participants were exposed to
We specified several models with increasing adjust- Lden≥65 dB and it was associated with higher body
ments: unadjusted model, basic model (age and mass index (23.55, SD=4.60 vs. 22.29, SD=3.67),
gender-adjusted), main model, and fully adjusted marital status, and occupation (data not shown).
model. The tool DAGitty v. 2.2 was used to con- Table 2 shows associations between IHD and
struct directed acyclic graphs (DAGs) to determine Lden for people with different duration of residence.
a priori the sufficient adjustment sets for estimat- No significantly elevated risk was observed across
ing the total effect of Lden and LONE in the main the models regardless of duration of residence and
model without overadjusting (we forced pack-years adjustment sets. Regarding LONE, ever-exposed
of smoking and lifetime residential noise exposure people had RR=1.76 increased risk; those exposed
in the main model for LONE, and pack-years of for 15–47 years had RR=2.35 increased risk, with
smoking, LONE, distance to major road, and dura- evidence of linear trend across the quartiles of
tion of residence in the main model for Lden). The exposure (Table 3).
models were run on 50 imputed datasets.22 They After stratification by participants’ characteris-
were tested for multicollinearity. Finally, sensitiv- tics, there was marginally significantly increased
ity analyses were conducted on the main model for risk associated with Lden among people with lower
LONE (per one interquartile range increase) and Lden socioeconomic status, whereas in those with middle
(≥ 10 years at the current address) to study possible and upper socioeconomic status there was no
effect modification by participants’ characteristics. risk. In ethnic minorities the risk reached RR=4.2
Results were considered statistically significant (95% CI: 1.14, 15.48). The other interaction terms
at p < 0.05 (two-tailed). Data were processed with were non-significant at p < 0.25 (Fig. 1). The risk
Statistical Package for the Social Sciences v. 17.0. associated with LONE was significantly higher
among ethnic minority members in comparison to
RESULTS Bulgarians (Fig. 2). The risk estimates themselves
A total of 249 questionnaires were distributed via were significant among women, ethnic minorities,
snowball sampling (85.5% response rate); 1906 smokers, and participants without diabetes.
people were approached during the field sampling
DISCUSSION
(19.3% response rate). Data from 513 question-
naires were analyzed after additionally excluding Overall, this study suggested non-significantly in-
68 cases due to unacceptably high percentage of creased risk of prevalent IHD in association with
missing data on all key variables or due to meeting Lden. Some foreign studies reviewed by Babisch also
exclusion criteria. reported null results.4 The exposure threshold of 60
Out of 513 participants, 35 (6.82%) reported dB Leq 6-22 h previously used by Tzenova et al.14 can
doctor-diagnosed IHD. According to Table 1, they be converted to approximately 62 dB Lden23 which is
were older, more often male, with lower educational somewhat comparable to the 65-dB threshold used
level, widowed/divorced, retired, and with lower in our study. We corroborated the high risk among
socio-economic status. They were more often dia- those exposed to these levels for over 20 years,
betic and hypertensive, they had higher body mass although our estimate failed statistical significance.
index, had smoked more, were more sensitive to LONE, conversely, was associated with signifi-
noise, reported more sleep disturbance, and had cantly increased risk among long-term employees
been exposed to occupational noise longer. and other subgroups.
Overall, 163 participants (31.77%) had been With respect to occupational noise, among ethnic
exposed to loud occupational noise at some point minorities there was evidence of more pronounced
in their life. Median duration of exposure among risk of IHD for every 13 years of employment, but
those participants was 7 years (IQR=14 years) and the results were inconclusive due to the wide confi-
it was positively associated with age (rs = 0.79, dence interval of the risk estimate. In similar vein
95% CI: 0.72, 0.84), body mass index (rs = 0.33, as Dimitrova and Karaslavova15, we found higher
Table 1. Sample characteristics according to participants’ ischemic heart disease (IHD) status
Note: aPercentages are reported within the IHD status; bp-value for Welch’s t-test or Pearson Chi-Square/Fisher’s Exact
Test/Fisher-Freeman-Halton Test. Lden – day-evening-night equivalent sound level.
risk among women exposed to noise at work. Gan explanation – in addition to environmental health
et al.6 conducted a cross-sectional survey using self- disparities, the access to health services, lifestyle,
rated noise exposure and found more pronounced religious, nutritional, and genetic factors might
odds of CHD among men and long-term exposed play a modifying role. Some minorities in Plovdiv
participants. like the Roma live in neighborhoods where leisure
The ethnic differences that we found might be time noise from different cultural and religious
quite important and need to be studied in future practices (e.g., loud music at weddings on the
research. The fact that noise exposure and IHD outside) is a serious environmental nuisance, which
did not differ considerably between Bulgarians and could amplify the effects of occupational noise. It
ethnic minority members suggests a more complex is also possible that ethnic minority members get
Table 2. Associations between ischemic heart disease (IHD) and exposure to Lden ≥ 65 dB among people with
different duration of residence at the current address (log-link Poisson regression models)
Note: Models are based on 50 imputed datasets. aModel is adjusted for age and gender; bAdditionally adjusted for ethnic-
ity, socio-economic status, educational attainment, pack-years of smoking, lifetime occupational noise exposure, distance
to major road and duration of residency; cAdditionally adjusted for body mass index, type 2 diabetes mellitus, arterial
hypertension, noise sensitivity, sleep disturbance and bedroom location; Lden – day-evening-night equivalent sound level.
Table 3. Associations between ischemic heart disease (IHD) and quartiles of lifetime occupational noise exposure
(log-link Poisson regression models)
RR (95% CI)
among ever RR (95% CI) according to the quartiles of exposurea Trende
exposeda
0.02 – 1 1–7 7 – 15 15 – 47
years years years years
No. of cases with IHD 23 0 2 6 16
Model (sample size)
Unadjusted (510) 3.55 (1.78, 7.05)* < 0.01 0.91 (0.21, 4.02) 3.48 (1.31, 9.25)* 9.79 (4.89, 19.62)* 0.001
Basicb (510) 2.08 (1.05, 4.11)* < 0.01 1.29 (0.31, 5.29) 1.52 (0.53, 4.36) 3.04 (1.45, 6.37)* 0.025
Mainc (500) 1.76 (0.82, 3.78) < 0.01 1.33 (0.37, 4.82) 1.39 (0.40, 4.76) 2.35 (1.00, 5.52) 0.069
Fully adjustedd (500) 1.29 (0.67, 2.49) < 0.01 1.18 (0.44, 3.18) 1.60 (0.52, 4.94) 1.24 (0.58, 2.66) 0.444
Note: Models are based on 50 imputed datasets. aReference category: never exposed (14 cases with IHD). bModel is
adjusted for age and gender; cAdditionally adjusted for ethnicity, socio-economic status, educational attainment, pack-
years of smoking and lifetime residential noise exposure; dAdditionally adjusted for distance to major road, body mass
index, type 2 diabetes mellitus, arterial hypertension, noise sensitivity, sleep disturbance, bedroom location and duration
of residence; ep-value for linear trend across the means of the exposure categories; *statistically significant at p<0.05.
employed as blue-collar workers more often due temperatures, chemicals, physical strain, etc. Thus,
to prejudice or bigotry; we did not have job title un-modeled residual confounding cannot be ruled
description, which would have allowed us to deter- out. This merits further research given the ethnic
mine the noise source in participants’ occupational structure of our society and the unfavorable milieu
environment, but the exposure to machine noise, and physical environment of some minorities.
prevalent among blue-collar workers, might differ
STRENGTHS AND LIMITATIONS
in its effects on IHD from “social noise” due to
differences in frequency spectrum or impulse-noise This study was one of the few to specifically focus
characteristics, despite the fact that both may elicit on the effects of both occupational and road traffic
the same vocal effort to talk to others; blue-collar noise on IHD in Bulgaria. It used objective measure
workers are also co-exposed to vibrations, extreme of road traffic noise and adjusted for most of the
Note: Models are based of 50 imputed datasets. Models are Note: Models are based of 50 imputed datasets. Models are
adjusted for age, gender, ethnicity, socio-economic status, adjusted for age, gender, ethnicity, socio-economic status,
education, pack-years of smoking, lifetime occupational noise education, pack-years of smoking and lifetime residential
exposure, distance to major road and duration of residence noise exposure (unless the results are stratified by the respec-
(unless the results are stratified by the respective covariate); tive covariate); SES – socio-economic status, T2DM – type 2
SES – socio-economic status, AH – arterial hypertension, diabetes mellitus, AH – arterial hypertension.
Lden – day-evening-night equivalent sound level.
Figure 1. Association between ischemic heart disease Figure 2. Association between ischemic heart disease
and exposure to road traffic Lden ≥ 65 dB among long- and one interquartile range increase in lifetime occupa-
term residents (≥ 10 preceding years) stratified by tional noise exposure stratified by participants’ charac-
participants’ characteristics (log-link Poisson regres- teristics (log-link Poisson regression models).
sion models).
relevant covariates.4,5 noise sensitivity) were relevant only for the past year
Like some previous studies6,14 ours has limita- and therefore, temporally unstable. The low response
tions. Its cross-sectional design means that we could rate and mixed sampling inhibits generalization of
not confirm that the exposure preceded the outcome. the findings, but this was not our objective and,
However, it met important criteria for causation given the sufficient breadth of the data, it should
(consistency, biological plausibility, coherence) not be concerning.
which makes it acceptable.24 The smallish sample Self-reported IHD might underestimate the true
size might explain some non-significant estimates; prevalence of the disease. However, in a population-
Babisch, on the other hand, advocated for interpreta- based male survey self-reported doctor-diagnosis with
tion of confidence intervals rather than p-values25, angina pectoris was found to be a valid measure.27
which are, according to some, misleading in epi- Another study of 34616 employees found 78% sen-
demiological research26. However, because of the sitivity and 99% specificity of self-reported coronary
low number of different ethnic minority members, heart disease.28
they were all analyzed under one group. Self-rated LONE is another possible source of
Some of the covariates (e.g., sleep disturbance, bias, as we had no data on whether participants’ oc-
Болгария
2 Менеджмент здоровья, Экономика здравоохранения и общая медицина, Факультет общественного здоровья, Медицин-