Socioeconomic Status and Environmental Noise Exposure In

Dale et al. BMC Public Health (2015) 15:205 DOI 10.1186/s12889-015-1571-2

RESEARCH ARTICLE Open Access Socioeconomic status and environmental noise exposure in Montreal, Canada Laura M Dale1, Sophie Goudreau2, Stephane Perron2, Martina S Ragettli2,4, Marianne Hatzopoulou3 and Audrey Smargiassi4,5*

Abstract Background: This study’s objective was to determine whether socioeconomically deprived populations are exposed to greater levels of environmental noise. Methods: Indicators of socioeconomic status were correlated with LAeq24h noise levels estimated with a land-use regression model at a small geographic scale. Results: We found that noise exposure was associated with all socioeconomic indicators, with the strongest correlations found for median household income, proportion of people who spend over 30% of their income on housing, proportion of people below the low income boundary and with a social deprivation index combining several socio-economic variables. Conclusion: Our results were inconsistent with a number of studies performed elsewhere, indicating that locally conducted studies are imperative to assessing whether this double burden of noise exposure and low socioeconomic status exists in other contexts. The primary implication of our study is that noise exposure represents an environmental injustice in Montreal, which is an issue that merits both investigation and concern. Keywords: Noise, Socioeconomic status, Deprivation, Environmental equity, Burden

Background adequate social relationships with members of one’s family, Chronic exposure to noise has been linked to various ad- community, or workplace. While each form of “relative” verse effects such as annoyance, sleep disturbance, im- deprivation may have its own public health implication, paired cognitive performance, as well as to the onset of socioeconomic status is often used as an indicator of cardiovascular diseases [1]. Environmental noise is one “relative” deprivation. of the most widespread sources of stress and discomfort Living in disadvantaged communities can be deleteri- in urban areas and few studies have assessed its associ- ous for health as a result of any and all of at least five ation with social deprivation [2-5]. health-influencing characteristics described by Stokols Deprivation is defined as “a state of observable and [7]. As such, one’s environment may act as 1) a medium demonstrable disadvantage relative to the local commu- for disease transmission, 2) a stressor, 3) a source of safety nity or the wider society or nation to which an indivi- or danger, 4) an enabler or hinderer of healthy behavior, dual, family or group belongs” [6]. Also, Townsend [6] and/or 5) a provider (or not) of health resources. Further- describes two main forms of “relative” deprivation. The more, poorer individuals are less empowered and may face first, “relative” material deprivation refers to a deficiency fewer choices of where to live, often forcing them to reside of fundamental goods and conveniences such as a safe in dwellings with inadequate conditions, and near a larger place to live, an adequate diet, and basic amenities. The number of environmental stressors such as toxic waste second, “relative” social deprivation refers to a lack of dumps, industrial sites, and roads with high traffic density [8-10]. However, higher exposures to environmental * Correspondence: [email protected] stressors have also been noted in wealthier populations. 4Département de santé environnementale et de santé au travail, Université de Montréal, Montréal H3C 3J7, Canada For example, Cesaroni et al. [11] noted that individuals 5Institut National de Santé Publique du Québec, Montréal, Canada Full list of author information is available at the end of the article

© 2015 Dale et al.; licensee BioMed Central. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly credited. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated. Dale et al. BMC Public Health (2015) 15:205 Page 2 of 8

living in areas of high road traffic were of higher social and social geography of Montreal can be found in position in Rome. Crouse et al.’s study on air pollution and socioeconomic Few studies have assessed associations between socio- deprivation [17]. economic status and exposure to noise. The study conducted by Hoffmann et al. [12], in Germany, noted a Socioeconomic indicators negative correlation with noise pollution (from traffic The socioeconomic characteristics of Montreal’spopula- noise) along the entire gradient of socioeconomic status tion were described using 2006 census data from Statistics based on four social indicators. Similarly in Hong Kong Canada aggregated at the dissemination area level. Al- [2] and in the Twin Cities, Minnesota [5], socially disad- though more recent data is available (i.e. 2011), the vali- vantaged groups were more exposed to traffic noise. In dity of the last Canadian census data has been questioned Paris (France), Havard et al. [3] reported that people liv- given that the questions related to the socio-economic sta- ing in advantaged neighborhoods were more exposed to tus are, since 2011, provided on a voluntary basis and thus road traffic noise than their less affluent counterparts. not mandatory as in the past [18]. In Marseille (France), a non-linear relationship, with the Dissemination areas are the smallest standard geo- highest exposure to road traffic noise at a middle level graphic areas for which all census data are distributed socio-economic status was noted [4]. Thus, studies on and they respect several delineation criteria. They 1) re- such trends are scarce and methodologically inconsistent, spect the boundaries of census subdivisions and census and have yet to be done in Montreal. tracts, allowing them to remain fairly stable over time, 2) The objective of the present study was to determine follow linear features such as roads, 3) are uniform in whether there is a correlation between the socioeconomic terms of population size, containing 400–700 persons status of populations in Montreal and exposure to envi- (larger or smaller sizes may result in order to respect cri- ronmental noise. terion 1), 4) are delineated based on block population counts from the previous census, and 5) are compact in Methods shape as much as possible [19]. Study area The dataset included 3147 dissemination areas with a Our study took place on the island of Montreal, where population greater than zero. 28 dissemination areas we examined neighborhood scale social and physical en- were without a population, due to their location in in- vironmental characteristics. The island has an area of dustrial areas or in large parks such as the Mount Royal 500 km2 that contains 19 boroughs. Road traffic among Park. The area of the dissemination areas ranged from the several expressways throughout the city is heavy, par- 0.07 km2 to 17.8 km2 and the mean was 0.16 km2.The ticularly along the number 13, 15, 20, 25 and 40 highways, populations of the 3147 dissemination areas considered which span the whole island. A number of railway tracks ranged from 113 to 4877 and the mean was 585. also reach an extensive portion of the island. There exists Eight indicators of socioeconomic status were studied a large international airport in the Dorval region (to the for each dissemination area :1) proportion of households West of the Island), and in certain areas, particularly in with only one person, 2) unemployment rate, 3) pro- the eastern portion of the city, there are clusters of indus- portion of people over the age of 25 without a diploma, trial activity. Whereas the most densely populated areas 4) proportion of people below the low-income boundary, are located around the city center and between highways 5) median household income, and 6) proportion of people 13 and 25, the territory west of highway 13 is more subur- who spend over 30% of their income on housing. Further- ban in character with low residential density; the area east more, two indicators combining several socio-economic of highway 25 is mainly of low and medium residential variables and developed by Pampalon and Raymond [20] density (Figure 1). were used, 7) the material deprivation index and 8) the so- Montreal differs from many other large North American cial deprivation index. Each of these indicators has been cities in a number of ways. For example, visible minority used regularly in studies linking deprivation with health or immigration statuses tend not to correlate to a great outcomes [17,20,21]. extent with low socioeconomic status [13,14]. Regard- less, Montreal represents one of the most economically Noise levels segregated cities in Canada, with the least equal income A-weighted outdoor summer noise levels (LAeq24h) distribution for the year 2000, as determined by the me- were computed for cells of 20 m × 20 m on the Island of dian share of income received by the less well-off half of Montreal, using a Land Use Regression (LUR) model the population [15]. This inequality translates into real [22]. This LUR model was developed based on LAeq24h health disparities, for example, a variation in life ex- from a two-week sampling period at 87 sites during the pectancy among men in different parts of the city by six summer 2010, along with determinants of the built en- years [16]. A more detailed description of the physical vironment in Montreal (e.g. vegetation, land use, road Dale et al. BMC Public Health (2015) 15:205 Page 3 of 8

Figure 1 Map of Montreal Island. network, etc.). These determinants focused on transporta- for the noise and the value from one to five for each so- tion and industrial noise sources, but did not include all cioeconomic indicator were added for each dissemin- neighborhood outdoor noise sources such as bars, ation area. Thus dissemination areas with a value of ten parking, etc. We assigned, to each dissemination area, represented both the worst noise and lowest socioeco- the average of all summer noise level estimates within nomic status. that area. Results Statistical analyses LAeq24h noise levels for the sampling period were in The strength and the direction of the relation between the range of 50.5-68.8 dBA with an arithmetic mean of mean predicted dissemination area LAeq24h noise levels 58.3 ± 3.2 dBA and median level of 58.3 dBA. The Island and each of eight indicators was assessed by computing of Montreal is noisier mainly where highways and indus- Pearson correlation coefficients. trial areas are present in the north-east of the island, and in the west where highways and the international Map Montreal Airport are also found (Figure 2). A map illustrating the double burden of noise exposure The Pearson correlation coefficients for LAeq24h levels and deprivation was produced as follows. First, quintiles with the indicators of the socioeconomic status were in of noise levels for the dissemination areas and quintile the range of |0.232| to |0.426|, all in their expected direc- of each the indicators of the socioeconomic status were tion so as to indicate a relationship between socioeco- calculated; the fifth quintile represented the worst noise nomic deprivation and elevated noise exposure (Table 1). or deprivation level. Then, the values from one to five, Correlations were usually linear, as represented in the Dale et al. BMC Public Health (2015) 15:205 Page 4 of 8

Figure 2 LAeq24h noise levels by dissemination areas in Montreal.

Table 1 Descriptive statistics of dissemination area variables (n = 3147) and their correlations with noise levels Variables (by dissemination area) Mean Median Standard 1st 99th Pearson correlation deviation percentile percentile with mean LAeq24h Coefficient p value Proportion of households with 1 person 34.7 35.3 14.8 4.3 68.2 0.259 <0.0001 Unemployment rate 8.8 7.7 6.3 0 28.6 0.232 <0.0001 Proportion of people over age 25 without 20.8 19.0 12.4 0 50.0 0.311 <0.0001 a diploma Proportion of people below the low 21.8 20.0 14.7 0 61.8 0.360 <0.0001 income boundary Median household income* 47 735 40 420 26 263 17 006 147 114 −0.426 <0.0001 Proportion of people who spend over 21.2 20.5 14.3 0 57.4 0.378 <0.0001 30% of their income on housing Material deprivation index −0.003 −0.004 0.048 −0.116 0.112 0.380 <0.0001 Social deprivation index 0.014 0.020 0.043 −0.097 0.100 0.337 <0.0001 *Distributions of income, after tax and earnings distributions have been suppressed where the total number of units (persons, families or households) in the reference year estimated number is less than 250. Dale et al. BMC Public Health (2015) 15:205 Page 5 of 8

scatter plots (Additional file 1: Figure S1), except for The information from the map is also summarized in median income, which approximated a negative expo- Table 2. There were few dissemination areas affected by nential relation with noise levels. a single factor: 43 dissemination areas presented only Figure 3 presents a map that combines the sound the worst average income and 37 dissemination areas, levels and the median household income. The areas the worst noise levels. presenting a double burden of noise exposure and lowest household income are presented in black in the Discussion figure. 15.8% (n = 501) of dissemination areas presented In this study we found that environmental noise expo- a double burden (class 9 and 10). These dissemination sure appears to be higher in areas of greater socio- areas were mainly found in the center of the Island of economic disadvantage. Associations between noise and Montreal, although not really clustered in one location. socioeconomic status indicators were moderate (<0.5) Noise levels were low and median incomes were high and usually linear. in the dissemination areas in the west of the Island Our results are consistent with those of a few studies (Additional file 2: Figures S2). Similar results were that showed that increased noise levels are associated obtained with other indicators of the socioeconomic with decreased socioeconomic status [2-5,12-23]. How- status (maps not shown). For example, using the un- ever, the heterogeneity of results in other locations employment rate which is the socioeconomic indicator makes it clear that the present study is inconsistent with the least correlated with noise levels, we noted that a number of others. For example, in Paris, France, 15.3% of dissemination areas presented a double bur- Havard et al. [3] associated higher noise exposure from den (data not shown). road traffic noise with higher socioeconomic status. A

Figure 3 Sum of quintile values of noise and median household income by dissemination areas in Montreal. Dale et al. BMC Public Health (2015) 15:205 Page 6 of 8

Table 2 Number of dissemination areas by quintiles of median household income and average noise levels (LAeq24h), Montréal, 2006 Median household income by Average LAeq24h by dissemination area (quintile) Total dissemination area (quintile) 1 (low) 2 3 4 5 (high) 1 (high) 283 126 121 68 37 635 2 100 129 159 140 106 634 3 56 135 143 162 138 634 4 38 97 179 169 150 633 5 (low) 43 74 143 163 188 611 Total 520 561 745 702 619 3147 case study in the Rijnmond region of the Netherlands by status, it is possible that such a discrepancy may also exist Kruize and Bouwman [23] similarly associated higher in- in Montreal. Crouse et al. [17] demonstrated a positive come levels with higher noise exposure levels, except in correlation between air pollution and socioeconomic sta- the case of aircraft noise for which opposite trend was tus but with notable deviations from the overall trend in observed. In Marseilles, France, Bocquier et al. [4] found certain neighborhoods and for certain indicators. For a non-linear relationship between social inequalities and example, the low education indicator was inconsistent exposure to road traffic noise, with the highest levels of with the idea of a double burden of socioeconomic dis- exposure at a middle level of deprivation. Brainard et al. advantage and air pollution exposure. In addition, two in Birmingham, UK [24] reported inconclusive trends other indicators (proportion of lone-parent families and between noise exposure from combined sources (road, proportion of adults separated, divorced, or widowed) rail and airport) and deprivation. were non-linearly associated with air pollution exposure. There are a number of considerations that should be Finally, the proportion of adults separated, divorced, or taken into account when comparing our study with widowed had a positive association with air pollution until others. First, our study did not differentiate between dif- a specific point, at which the relationship became nega- ferent noise sources, or between nighttime and daytime tive. Conversely, for each indicators used in our study, a noise. Second, we used different indicators for socioeco- double burden of low socioeconomic status and higher nomic status than each of the studies described above noise exposure was observed. Apparently, noise and air (although overlap did occur). pollution are not uniformly linked in Montreal. Third, it should be noted that each of the studies men- Our work is imperfect for a number of reasons. First, tioned above was conducted in Europe, where cities are since it is impossible to measure noise everywhere, any- frequently configured differently than in North America. time and for the entire population, we used a land use Indeed the inner city in Western Europe is often weal- regression model that predicts average sound levels over thier than the suburbs whereas in Montreal, the propor- 24 hours (LAeq24h), with an error associated with the tion of people with a low socioeconomic statusis higher prediction of LAeq24h of 3.3 dBA. in the center of the Island, where higher noise levels can Second, LAeq24h may not be the best indicator of ex- be found. Each of these considerations limits the com- posure as sporadic noise events, such as those coming parability of our results. from aircrafts, buses or trucks, may be more disruptive Disregarding these methodological differences how- than more constant streams of noise [26,27]. In addition, ever, it would appear that the relationship between so- the 24 hr indicator does not allow us to differentiate bet- cioeconomic status and noise exposure is highly variable ween nighttime and daytime noise. Given the numerous and dependent on local contexts. There is a need for fur- potential effects of sleep disturbance on health, the pos- ther research in this field to reconcile and provide expla- sibility of certain areas having higher nighttime noise nations for these regional discrepancies. By identifying represents an important consideration. Similarly, our explanatory factors, we may then begin to extrapolate analysis considered only noise levels at people’s places of local results to similar areas elsewhere in the world. residence. Since not all people are home all day, they A study done in New York City on noise, air pollutants could have different overall exposure profiles than those and traffic suggested that an imperfect correlation exists suggested by our results. between noise and air pollution in the context of road Our results are also limited to outdoor noise levels; it traffic [25]. Comparing the results from the present study is conceivable that a comparison of indoor noise would on noise and socioeconomic status with those from further emphasize this double burden. For example, low- Crouse et al. [17] on air pollution and socioeconomic income dwellings may be more likely to have features Dale et al. BMC Public Health (2015) 15:205 Page 7 of 8

such as thinner windows and ineffective insulation, areas. There is an apparent lack of environmental equity making them more susceptible to permeation by noise. on the Island of Montreal. Our results indicate that Furthermore, the noise model by Goudreau et al. [22] deprived groups endure a double burden of low eco- used in this study is based on two-week measurements nomic status and higher exposure to environmental in the summer only, and does not consider potential sea- noise. These findings thus highlight a societal externality sonal differences due to weather conditions and noise imposed on an already disadvantaged group. Interven- sources. Nevertheless, in our previous work in Montreal, tions to reduce noise levels in Montreal should be tar- we noted a relatively good correlation between winter geted to lower income neighborhoods. and summer noise levels (Pearson r = 0.74; Goudreau 2014, personal communication). Additional files Furthermore, our assessment of the spatial distribution of the socioeconomic position is imperfect as we used Additional file 1: Figure S1. Scatterplots correlating LAeq24h with aggregated socioeconomic indicators at the dissemi- socioeconomic indicators. nation area level. Small-scale spatial variations of the so- Additional file 2: Figure S2. Median household income by dissemination cioeconomic status within a dissemination area may not areas in Montreal. be detected. It is indeed worth noting that concentrated pockets of individuals with low socio-economic status Competing interests are generally small in Montreal, rarely covering an entire The authors declare that they have no competing interests. dissemination area [28]. Authors’ contributions Another consideration in our study design is the LD wrote the manuscript and interpreted the results. SG was involved in the modifiable areal unit problem (MAUP), which arises due data collection, performed the analyses and contributed to the interpretation to the arbitrary nature by which the boundaries of areal of the results. SP and MH contributed to the design of the study and to the writhing of the manuscript. AS conceived the study, directed the analyses, units are chosen. If different boundaries were used in- contributed to the interpretation of results and to the writing of the stead (i.e. if the areal units were modified), then the re- manuscript. All authors reviewed and approved the final manuscript. sults based on these delineations could be quite different [29]. Bowen [30] suggests that the appropriate level of Acknowledgement spatial aggregation varies based on the objectives of the This work was financially supported by the Direction de Santé Publique de l’Agence de la Santé et des Services Sociaux de Montréal. study, and that studies on exposure to environmental hazards call for smaller geographic units. The logic for Author details 1 this recommendation follows from spatial concerns re- McGill School of Environment, McGill University, Montréal, QC, Canada. 2Direction de santé publique de Montréal, Montréal, Canada. 3Department of garding areal unit definitions. For example, it is con- Civil Engineering, McGill University, Montréal, QC, Canada. 4Département de ceivable that two populations on the boundaries of their santé environnementale et de santé au travail, Université de Montréal, 5 respective areal units may be more similar to each other Montréal H3C 3J7, Canada. Institut National de Santé Publique du Québec, Montréal, Canada. than to the populations within their own units. Also, given that dissemination areas are larger outside the city Received: 1 October 2014 Accepted: 17 February 2015 core, the error that can result from attributing the same noise levels to all individuals in a dissemination area References may be greater outside the metropolitan center. The lo- 1. Basner M, Babisch W, Davis A, Brink M, Clark C, Janssen S, et al. Auditory and gical solution here would be to use a smaller spatial unit non-auditory effects of noise on health. Lancet. 2014;383(9925):1325–32. that better represents exposure [30]. We thus chose the 2. Lam KC, Chan PK. Socio-economic status and inequalities in exposure to transportation noise in Hong Kong. Transportation. 2006;2(1):107–13. smallest census unit available, dissemination areas. 3. Havard S, Reich BJ, Bean K, Chaix B. Social inequalities in residential A final consideration in our analysis is that the census exposure to road traffic noise: an environmental justice analysis based on data used in our study was from the year 2006 while our the RECORD cohort study. Occup Environ Med. 2011;68:322–74. http://dx.doi.org/10.1136/oem.2010.060640. noise data was obtained in 2010. It is possible that the so- 4. Bocquier A, Cortaredona S, Boutin C, David A, Bigot A, Chaix B, et al. cioeconomic and geographic context has changed since Small-area analysis of social inequalities in residential exposure to road then, although we do not have reason to believe so. traffic noise in Marseilles, France. Eur J Pub Health. 2012;23(4):540–6. http://dx.doi.org/10.1093/eurpub/cks059. 5. Nega TH, Chihara L, Smith K, Jayaraman M. Traffic noise and inequality in Conclusion the twin cities, Minnesota. Human Ecological Risk Assessment: An Int J. The main implications of the present study are two-fold. 2013;19(3):601–19. 6. Townsend P. Deprivation. J Social Policy. 1987;16(2):125–46. First, it is apparent that studies linking environmental http://dx.doi.org/10.1017/S0047279400020341. exposures to socioeconomic status require careful exa- 7. Stokols D. Establishing and maintaining healthy environments: toward a mination in local contexts. This finding arose in light of social ecology of health promotion. Am Psychol. 1992;47(1):6–22. http://dx.doi.org/10.1037/0003-066X.47.1.6. the many results on noise exposure and social condi- 8. Evans G. The environment of childhood poverty. Am Psychol. tions that have proven difficult to generalize to other 2004;59(2):77–92. http://dx.doi.org/10.1037/0003-066X.59.2.77. Dale et al. BMC Public Health (2015) 15:205 Page 8 of 8

9. Braubach M, Fairburn J. Social inequities in environmental risks associated with housing and residential location—a review of evidence. European J PublHealth. 2010;20(1):36–42. 10. Mitchell G, Dorling D. An environmental justice analysis of British air quality. Environ Planning A. 2003;35(5):909–29. 11. Cesaroni G, Badaloni C, Romano V, Donato E, Perucci CA, Forastiere F. Socioeconomic position and health status of people who live near busy roads: the Rome Longitudinal Study (RoLS). Environ Heal. 2010;9(1):41. 12. Hoffmann B, Robra BP, Swart E. Social inequality and noise pollution by traffic in the living environment—an analysis by the German Federal Health Survey. Gesudheitswesen. 2003;65(6):393–401. 13. Seguin AM, Germain A. The social sustainability of Montreal: a local or state matter? In: Polèse M, Stren R, editors. Social Sustainability of cities: Diversity and the Management of change. Toronto: University of Toronto Press; 2000. p. 39–67. 14. Bauder H, Sharpe B. Residential segregation of visible minorities in Canada’s gateway cities. Can Geogr. 2002;46(3):204–22. http://dx.doi.org/10.1111/ j.1541-0064.2002.tb00741.x. 15. Ross NA, Wolfson MC, Dunn JR, Berthelot JM, Kaplan GA, Lynch JW. Relation between income inequality and mortality in Canada and in the United States: cross sectional assessment using census data and vital statistics. BMJ. 2000;320:898–902. http://dx.doi.org/10.1136/bmj.320.7239.898. 16. Leblanc MF, Raynault MF, Lessard R. Direction de santé publique. Social inequalities in health in Montreal – Progress to date. Agence de la santé et des services sociaux de Montréal. ISBN 978-2-89673-133-6. 2011;144. 17. Crouse DL, Goldberg MS, Ross NA. A prediction-based approach to modelling temporal and spatial variability of traffic-related air pollution in Montreal. Canada Atmosphéric Environ. 2009;43(32):5075–84. 18. Direction de la méthodologie et de la qualité et Direction des statistiques sociodémographiques. L’Enquête nationale auprès des ménages de Statistique Canada : État des connaissances à l’intention des utilisateurs du Québec. Institut de la statistique du Québec. 2013. p13. Available at: http://www.stat.gouv.qc.ca/statistiques/enm-note-information.pdf 19. Statistics Canada. 2006 Census Dictionary. More information on dissemination area (DA). 2009. Available at: http://www12.statcan.ca/census-recensement/ 2006/ref/dict/geo021a-eng.cfm. (October 18, 2013 date last accessed). 20. Pampalon R, Raymond G. A deprivation index for health welfare planning in Quebec. Chronic Diseases Canada. 2000;21(3):104–13. 21. Twigge-Molecey, A. Is gentrification taking place in the neighbourhoods surrounding the MUHC? A census-based analysis of relevant indicators. Analysis Report RR09-02E. Montréal: CURA Making Megaprojects Work for Communities - Mégaprojets au service des communautés; 2009. 80. 22. Goudreau S, Plante C, Fournier M, Brand A, Roch Y, Smargiassi A. Estimation of spatial variations in urban noise levels with a land use regression model. Environment and Pollution; (4):48 23. Kruize H, Buowman AA. Environmental (in) equity in the Netherlands: A Case Study on the Distribution of Environmental Quality in the Rijnmond Region. RIVM 2004. http://hdl.handle.net/10029/8982. 24. Brainard JS, Jones AP, Bateman IJ, Lovett AA. Exposure to environmental urban noise pollution in Birmingham, UK. Urban Stud. 2004;41(13):2581–600. http://dx.doi.org/10.1080/0042098042000294574. 25. Ross Z, Kheirbek I, Clougherty JE, Kazuhiko I, Thomas M, Markowitz S, et al. Noise, air pollutants and traffic: continuous measurement and correlation at a high-traffic location New York City. Environ Res. 2011;111:1054–63. 26. Clark C, Martin R, van Kempen E, Alfred T, Head J, Davies HW, et al. Exposure-effect relations between aircraft and road traffic noise exposure at school and reading comprehension. Am J Epidemiol. 2005;163(1):27–37. http://dx.doi.org/10.1093/aje/kwj001. http://dx.doi.org/10.1016/j. Submit your next manuscript to BioMed Central atmosenv.2009.06.040. and take full advantage of: 27. Seto EYW, Holt A, Rivard T, Bhatia R. Spatial distribution of traffic induced noise exposure in a US city: an analytic tool for assessing the health • Convenient online submission impacts of urban planning decisions. Int J Health Geogr. 2007;6:24–39. http://dx.doi.org/10.1186/1476-072X-6-24. • Thorough peer review 28. Montpetit C. Groupe de travail sur la pauvreté à Montréal/Centraide. Un • No space constraints or color ﬁgure charges portrait de la pauvreté sur le territoire de Centraide du Grand Montréal. Montréal: Centraide/Centre Léa-Roback; 2007. • Immediate publication on acceptance 29. Baden BM, Noonan DS, Turaga RMR. Scales of justice: is there a geographic • Inclusion in PubMed, CAS, Scopus and Google Scholar bias in environmental equity analysis? J Environ Plan Manag. 2007;50 • Research which is freely available for redistribution (2):163–85. http://dx.doi.org/10.1080/09640560601156433. 30. Bowen WM. Environmental justice through research-based decision-making. New-York: Routledge; 2001. Submit your manuscript at www.biomedcentral.com/submit International Journal of Environmental Research and Public Health

Article Sleep Disturbance from Road Trafﬁc, Railways, Airplanes and from Total Environmental Noise Levels in Montreal

Stéphane Perron 1,2, Céline Plante 1, Martina S. Ragettli 3,4, David J. Kaiser 1,5, Sophie Goudreau 1 and Audrey Smargiassi 6,7,8,* 1 Public Health Department of Montreal, Montreal, QC H2L 1M3, Canada; [email protected] (S.P.); [email protected] (C.P.); [email protected] (D.J.K.); [email protected] (S.G.) 2 Department of Social and Preventive Medicine, School of Public Health, University of Montreal, Montreal, QC H3C 3J7, Canada 3 Swiss Tropical and Public Health Institute, Basel 4002, Switzerland; [email protected] 4 University of Basel, Basel 4003, Switzerland 5 Department of Epidemiology, Biostatistics and Occupational Health, McGill University, Montreal, QC H3A 1A2, Canada 6 National Institute of Public Health Quebec, Montreal, QC H3C 2B9, Canada 7 Public Health Research Institute, University of Montreal, QC H3C 3J7, Canada 8 Department of Environmental Health and Occupational Health, School of Public Health, University of Montreal, Montreal, QC H3C 3J7, Canada * Correspondence: [email protected]; Tel.: +1-514-343-6111 (ext. 38528)

Academic Editors: Peter Lercher, Ronny Klaeboe and Mariola Sliwinska-Kowalska Received: 18 February 2016; Accepted: 4 August 2016; Published: 11 August 2016

Abstract: The objective of our study was to measure the impact of transportation-related noise and total environmental noise on sleep disturbance for the residents of Montreal, Canada. A telephone-based survey on noise-related sleep disturbance among 4336 persons aged 18 years and over was conducted. LNight for each study participant was estimated using a land use regression (LUR) model. Distance of the respondent’s residence to the nearest transportation noise source was also used as an indicator of noise exposure. The proportion of the population whose sleep was disturbed by outdoor environmental noise in the past 4 weeks was 12.4%. The proportion of those affected by road trafﬁc, airplane and railway noise was 4.2%, 1.5% and 1.1%, respectively. We observed an increased prevalence in sleep disturbance for those exposed to both rail and road noise when compared for those exposed to road only. We did not observe an increased prevalence in sleep disturbance for those that were both exposed to road and planes when compared to those exposed to road or planes only. We developed regression models to assess the marginal proportion of sleep disturbance as a function of estimated LNight and distance to transportation noise sources. In our models, sleep disturbance increased with proximity to transportation noise sources (railway, airplane and road trafﬁc) and with increasing LNight values. Our study provides a quantitative estimate of the association between total environmental noise levels estimated using an LUR model and sleep disturbance from transportation noise.

Keywords: transportation noise; sleep disturbance; land use regression

1. Introduction Systematic reviews have shown associations between environmental noise from transportation-related sources and multiple health outcomes, including ischemic heart disease, hypertension, sleep disturbance and annoyance [1–4]. Recent studies have also shown an association between road

Int. J. Environ. Res. Public Health 2016, 13, 809; doi:10.3390/ijerph13080809 www.mdpi.com/journal/ijerph Int. J. Environ. Res. Public Health 2016, 13, 809 2 of 21 traffic noise and diabetes [5], stroke [6,7], mental health [8] and all-cause mortality [7]. Other studies also showed an association between aircraft noise and stroke [9,10]. In several studies, night-time noise has been linked to sleep disturbance [11]. In a systematic review of epidemiological studies, exposure to aircraft noise was associated with multiple indicators of sleep disturbance, including awakenings, decreased slow wave sleep time and the use of sleep medication [12]. Finally, in one experimental study, combined noise exposure from multiple transportation sources (road, railway and air traffic noise) led to more awakenings and arousals than for single noise sources; a similar effect was seen when the number of noise events per unit time increased [13]. To our knowledge, the population health impact of combined train and road traffic noise sources on sleep disturbance has been evaluated in one survey-based study [14]. From a public health perspective, sleep disturbance in and of itself is relevant to health; in addition, it is suspected to be in the causal pathway linking exposure to noise with cardiovascular diseases. Even at very low night-time noise levels, physiological reactions such as an increase in heart rate, blood pressure, catecholamine production and endothelial dysfunction [15,16], body movements and arousals have been measured [3,17]. Night-time activation of the autonomic nervous system has been shown to incompletely habituate to noise [17]. Non-habituating autonomic reactions such as increased heart rate or blood pressure may be important precursors of long-term cardiovascular outcomes [18]. The relationship between environmental noise levels and sleep disturbance has mostly been studied in Europe [19,20]. Sensitivity and exposure to noise may however differ between countries, seasons and individuals. Noise annoyance, an indicator closely linked to sleep disturbance, has been shown to be higher during the summer in European and American cities [21]. Miedema et al. [21] reviewing 42 studies conducted at different times of the year estimated that a 15 ˝C difference in temperature has about the same effect on noise annoyance as does a 1- to 3-dB difference in noise exposure. Noise sensitivity has been shown to influence sleep disturbance. Indeed, in survey studies and laboratory studies of exposure to transportation noise, subjectively assessed sleep disturbance is higher amongst individuals who report being sensitive to noise [22–25]. In addition, the link between environmental noise and sleep disturbance has been studied in experimental or quasi-experimental settings, using either direct measurements or noise propagation models that predict noise for a specific source by means of physical rules of noise propagation and attenuation [3,12,24,25]. Recent noise monitoring campaigns in the United States [26] and in three Canadian cities (Halifax [27], Toronto [28] and Montreal [29]) have shown that measured noise levels tend to exceed the 55 dBA LDay and LNight threshold that has been associated with negative health impacts when using propagation noise models [3,4]. In order to accurately quantify the burden of disease associated with environmental noise exposure, it is important to use methods that are based on actual measured noise and control for factors such as sensitivity to noise in the population. In the absence of a noise monitoring system covering the entirety of a large urban area such as Montreal, LUR noise models provide an opportunity to estimate health impacts on the basis of a sample of actual noise measurements. To our knowledge, the effect of environmental noise on sleep disturbance has not been studied using LUR noise models. The objective of this study was to estimate the prevalence of sleep disturbance associated with exposure to road traffic, railway, airplane and total environmental noise using two exposure metrics—the distance to each transportation noise source; and total environmental noise levels estimated with a land use regression (LUR) noise model amongst residents of the Island of Montreal, Canada. The effect of combinations of noise sources on sleep disturbance was also investigated.

2. Methods

2.1. Study Population The study population consists of the population of the Island of Montreal—approximately 1.5 million people 18 years or older in 2011 [30]. A signiﬁcant proportion of the population lives along major roads such as arterials and national highways; some residences are also located in Int. J. Environ. Res. Public Health 2016, 13, 809 3 of 21 Int. J. Environ. Res. Public Health 2016, 13, 809 3 of 21 proximity toto railwayrailway tracks, tracks, shunting shunting yards yards and/or and/or in the in vicinitythe vicinity of the of Montreal the Montreal International International Airport Airport(Figure1 ).(Figure Proximity 1). Proximity to the airport to the was airport deﬁned was on thedefined basis ofon the the 2009 basis estimate of the of 2009 the Noiseestimate Exposure of the NoiseForecast Exposure zone 25 Forecast (hereafter zone referred 25 (hereafter to as NEF25), referred which to circumscribesas NEF25), which an area circumscribes where annoyance an area is wherelikely toannoyance occur because is likely of aircraftto occur noise because [29]. of aircraft noise [29].

FigureFigure 1. 1. MapMap of of the the study study area, area, the the island island of of Montreal, Montreal, with potential transportation noise sources.

2.2. Survey Survey and and Sleep Sleep Disturbance Disturbance by Noise Data The data on sleep disturbance used in this stud studyy are derived from a telephone-based survey of thethe health effects of environmental noise carried out in 2014 on the Island of Montreal. A detailed methodology of of the survey used fo forr this study has been presented elsewhere [31]. [31]. Briefly, Briefly, survey respondentsrespondents 18 18 years years of of age or older were randomly selected from four pre-defined pre-defined survey strata (road, rail,rail, aircraft,aircraft, non-exposed) non-exposed) based based on on proximity proximit toy theto the noise noise sources sources illustrated illustrated in Figure in Figure1, in order 1, in orderto ensure to ensure sufficient sufficient respondents respondents exposed exposed to various to transportationvarious transportation noise sources. noise First,sources. exposure First, exposurecategories categories were defined were on defined the basis on of the distance basis toof eachdistance transport-related to each transport-related source: living source: within living 100 m withinof an artery 100 m or highwayof an artery (road); or withinhighway 150 (road); m of a within railway 150 line m or of main a railway line of aline railroad or main shunting line yardof a railroad(rail); within shunting 1 km ofyard the (rail); NEF25 within zone of 1 thekm Montrealof the NEF25 International zone of Airportthe Montreal (aircraft); International and not within Airport one (aircraft);of the three and defined not distanceswithin one to transportationof the three noisedefined sources distances (non-exposed). to transportation The exposure noise categories sources (non-exposed).were then combined The exposure to generate categories four survey were strata: then (1)combined road (road to generate only); (2) four rail (railsurvey and strata: road and(1) road rail); (road(3) NEF25 only); (NEF25, (2) rail (rail road and and road NEF25, and rail rail); and (3) NEF25, NEF25 road(NEF25, and road rail andand NEF25);NEF25, rail and and (4) non-exposedNEF25, road and(an arearail and not NEF25); directly exposedand (4) non-exposed to transportation (an area noise not sources). directly exposed Noise exposure to transp categoriesortation noise were sources). created Noiseby means exposure of ArcGIS categories 9.2 (ESRI, were Redlands, created by CA, means USA) of using ArcGIS a digital 9.2 (ESRI, road Redlands, network (AddressesCA, USA) using Quebec a digital2013), aroad railway network network (Addresses (CanMap Quebec® Rail V2010.1), 2013), a railway and the network geo-referenced (CanMap NEF25® Rail map V2010.1), of the Montreal and the geo-referencedInternational Airport NEF25 [ 29map,32]. of We the aimed Montreal to obtain Intern 4500ational respondents Airport (i.e., [29,32]. 1125 We respondents aimed to for obtain each 4500survey respondents strata). (i.e., 1125 respondents for each survey strata).

Int. J. Environ. Res. Public Health 2016, 13, 809 4 of 21

The survey questionnaire was adapted from the European LARES survey (Large Analysis and Review of European housing and health Status) [33] and included questions on socio-demographic characteristics, self-estimated noise sensitivity and annoyance. The survey was carried out between 10 April 2014 and 20 June 2014. Results on annoyance for the same population are reported in Ragettli et al. 2015 [34]. Sleep disturbance was assessed with two questions. First, a general question on sleep disturbance was phrased as follows: “Was your sleep disturbed by noise in the past 4 weeks?” If the respondent answered yes, the following question was asked: “If yes, what was the source of this noise?” Respondents could choose from the following list of twelve noise sources and/or by an open response item: (1) noise in the neighbourhood (bars, discos, demonstrations); (2) playgrounds, schools; (3) road trafﬁc; (4) airplanes; (5) trains; (6) parking lots; (7) apartments nearby (e.g., voices, music, TV, work, animals); (8) animals or birds (outdoors); (9) shopping centers, industrial or construction zones; (10) Someone using the stairs; (11) ventilation, heating, garbage chute; and (12) noise coming from your own apartment. For the purposes of the present study, four sources of sleep disturbance were retained: individual transportation noise sources—road trafﬁc, railways, airplanes; and total outdoor environmental noise (all sources previously listed except apartments nearby, someone using the stairs, ventilation, heating, garbage chute and noise coming from your own apartment). Sleep disturbance from any of the three transportation sources was also retained as total transportation noise. Finally, the questionnaire included a question on noise sensitivity phrased as follows: “Please indicate if you are not at all, a little, somewhat, quite a bit or a lot sensitive to noise”. Answers were grouped into two categories: 1. not at all and a little; 2. somewhat, quite a bit or a lot.

2.3. Noise Levels and Distance to Noise Sources

A-weighted night-time outdoor total environmental noise levels (LNight) for the Island of Montreal were derived at the geographic coordinate of the postal code of each subject using a land use regression model (LUR) developed by Ragettli et al. [34]. On the island of Montreal, each postal code corresponds to a very small geographical area, usually a block-face or a large apartment complex. The median, 5th and 95th percentile of the area of postal codes included in the present study are 7000 m2, 1180 m2 2 and 16,018 m , respectively. The LUR model for LNight was developed based on 204 noise samples (175 different geographical sites) collected during a two-week sampling period in the summer of 2010 [29] and in the spring of 2014. The sampling sites were selected in order to be close to residential areas and to cover various transportation noise sources (road traffic, trains and aircraft). At each location, noise levels were measured continuously during at least five days recording two-minute averages. For each site, the equivalent LNight (average) was calculated by integrating all values sampled between 11:00 PM and 6:00 AM. The LUR model was built using a general additive model (GAM) by establishing a statistical relationship between noise measurements and surrounding determinants of the built and natural environments. The model is described in Ragettli et al. [34] and includes predictor variables related to all traffic noise sources: road (length of major roads within 50 m, annual average traffic counts at the nearest road, distance to highway), railway (presence of a railroad shunting yard within 100 m) and air traffic (1 km or less from the NEF25 contour). Characteristics of the built environment known to be associated with noise levels were also added to the model: normalized difference vegetation index (NDVI); geographical sector (West, Downtown, East, Center); residential density; proximity to a bus line; proximity to a river; and presence of commercial land use. Finally, the estimation error of the LNight was assessed for each sampling site using the Root Mean Square Error (RMSE) obtained by leave one out cross validation (LOOCV). The model explained 59% of the spatial variability in LNight, and the RMSE was 3.9 dBA. In addition to LNight for each subject, the distance of the residential six-digit postal code to the nearest transportation noise source (major road, railway, NEF25 contour) was computed using PostGresSQL 9.1 (PostgreSQL Global Development Group, Berkeley, CA, USA). As with Int. J. Environ. Res. Public Health 2016, 13, 809 5 of 21

Ragettli et al. [34], the distance in meters to the closest major road was divided into six categories (in meters; <50, 51–100, 101–150, 151–200, 201–500, >500), distance in meters to the closest railway (which also includes main railway lines of railroad shunting yards) was divided into ﬁve categories (in meters; ď100, 101–200, 201–500, 500–1000, >1000) [35], distance to the NEF25 contour was categorized into 1000 meter increments (within the NEF contour, >0–1000, 1001–2000, >2000) [36]. To explore the effect on sleep disturbance of being exposed to a single transport-related source compared to multiple sources, we created two additional exposure categories based on the same distance criteria as used to generate the four survey strata (see Section 2.2) but reﬂecting simultaneous exposure to two sources: road and airplane; and road and rail.

2.4. Survey Weights and Statistical Analysis Data were weighted to generate estimates that are representative of the total adult population of Montreal. First, we computed the sampling weights for each stratum by multiplying the reciprocals of the proportions of respondents by strata and the corresponding proportion of the total population aged 18 and more (obtained from the 2011 Census). These sampling weights were then “raked” so that weighted totals would match census totals for sex and nine classes of age and education (i.e., three classes of education ˆ three classes of age) [37]. Descriptive data and statistical models are all based on weighted data. In order to quantify the associations between sleep disturbance by source and exposure to noise (categories of distances to source and LUR-estimated LNight as a continuous variable), marginal proportions were computed using the age-sex-education adjusted log-binomial regressions and the “margins” function of STATA 13 (StataCorp, College Station, TX, USA), for the categories of distance (categories described above) and 5-dBA increments in LUR-estimated LNight. We also ran a model with all three distance variables. The weighted prevalence of sleep disturbance (i.e., marginal proportions) due to each source (roads, rail, and aircraft) was estimated for categories of distance to the corresponding source in three separate regression models, e.g., proportion of sleep disturbance caused by airplanes according to categories of distance to the NEF25 contour. In addition, the marginal prevalence of sleep disturbance due to road traffic, all transport sources and all environmental sources was computed for each category of LUR-estimated LNight (three models). LNight was used as a continuous variable since the association between LNight and sleep disturbance by environment, transport and roads was linear. We report marginal proportions by increment of 5 dBA. To explore the effect on sleep disturbance by transportation noise, of being exposed to a single transport-related source, compared to combined sources, we computed a variable classifying the respondents into one of the six following exposure groups: 1-road traffic (ď100 m of major road); 2-aircraft (ď1000 m or inside NEF25 contour); 3-train (ď150 m of a railway); 4-road & aircraft (ď100 m of major road & ď1000 m or inside NEF25 contour); 5-road & railway (ď100 m of major road & ď150 m of a railway); and 6-not exposed (>100 m of major road & >150 m of a railway & >1000 m of NEF25 contour). The categories are mutually exclusive and exhaustive. Note that these thresholds correspond to the two first categories of distances to the noise sources described in Section 2.3. Combined exposure to airplanes and railways and combined exposure to all three sources are excluded from the analysis due to an insufficient number of cases. We then computed the marginal proportion of sleep disturbance for each group, based on the regression model relating sleep disturbance by transportation to this variable. Since noise sensitivity may be an important co-factor in negative health outcomes, we also performed regression models adjusted for sensitivity. Furthermore, regression models were performed with the inclusion of an interaction term between sensitivity and distance or LNight. Using these models, we computed marginal proportions of sleep disturbance for categories of distance, and categories of LUR-estimated LNight levels, stratified by sensitivity to noise (not at all/a little and somewhat/quite a bit or a lot). Tests of the significance of interactions terms were performed with the function testparm on Stata. STATA version 13.0 (StataCorp, College Station, TX, USA) was used for all statistical analyses. Int. J. Environ. Res. Public Health 2016, 13, 809 6 of 21

3. Results In total, 15,697 randomly selected telephone numbers were dialed (of which 21.5% were out of service or not valid) to obtain 4500 respondents. Using the American Association for Public Opinion Research (AAPOR) standards [38], this corresponds to a response rate of 46.8%. After excluding ineligible observations (missing data for age and/or education for generating survey weights), 4336 observations were available for computing the survey weights and carrying out the analyses. Table1 presents the characteristics of the unweighted and weighted survey sample. A third of the weighted sample lives within 100 m of a major road. The median of nighttime noise levels LNight is 56.5 and about 60% of the population is exposed to night-time noise levels above 55 dBA according to the LUR model (not shown). Slightly more than half of the respondents are sensitive to noise. Environmental noise is a common cause of sleep disturbance. In this sample, 15.4% of individuals have bad or very bad sleep quality (not shown in table).

Table 1. Weighed characteristics of the study population.

Sample (4336) Unweighted Weighted Age (mean (Standard Deviation (SD))) 53.8 (16.1) 51.0 (15.0) Female (%) 51.8 53.0 Years lived in the current residence (mean (SD)) 13.8 (12.4) 13.0 (10.5) Educational level (%) No diploma or elementary school 8.0 11.0 High school 17.4 25.7 College 26.2 26.0 University degree 48.3 37.2 Exposed to airplane noise: ď1 km from NEF25 curve (%) 24.4 4.8 Exposed to road trafﬁc noise: ď100 m from a major road (%) 39.1 33.4 Exposed to railway noise: ď150 m from a railway (%) 19.1 7.3 Night-time noise level (dBA) Min 44.6 44.6 Median 56.1 56.5 Max 69.7 69.7 Sensitivity to noise (%) Somewhat/Quite a bit/a lot vs. Not at all/A little 53.8 52.2 Sleep disturbed in the past 4 weeks (%) by Environmental noise 15.2 12.4 Total transportation noise 8.9 6.1 Trafﬁc noise 4.5 4.2 Railway noise 2.2 1.1 Aircraft noise 3.4 1.5

Figures2–4 present the population-weighted prevalence of sleep disturbances according to distance to each transportation noise source. Using log-binomial regression models with the most exposed as the reference categories, negative relationships between sleep disturbance and distance to the noise source are observed for all three transportation sources. The summaries of the regression models are presented in Tables A1–A3 of the AppendixA. The strongest effects are observed within 1000 m of the NEF25 contour, 50 m of a major road and 100 m of a railway. We observed an increase in sleep disturbance due to transportation noise (as deﬁned in Section 2.2) for those exposed to more than one transportation noise source as compared to those not exposed (not close to) to any transportation noise source (Figure3). Those exposed to both road and rail noise are also more sleep disturbed than those exposed to road noise only or rail noise only but the difference is only statistically signiﬁcant for road noise. There is also a difference between individuals within 1000 m of the NEF25 contour and those within a 100 m of a major road (Figure3). Results using distance variables estimated with multivariate models were similar to those obtained with univariate models. Int. J. Environ. Res. Public Health 2016, 13, 809 7 of 21

The marginal predictions of sleep disturbance in the total population resulting from models adjusted for sensitivity are very similar to results from unadjusted models (data not shown), although the marginal proportions are higher in sensitive than in the non-sensitive subgroup. Sensitivity to noise does not modify the association between sleep disturbance and proximity to railway sources (Table A4). There appears to be a modification of the association between sleep disturbance and proximity to the airport according to noise sensitivity (Table A6), but the variability of the estimates is large and the interaction term is not statistically significant (p = 0.07, Stata testparm function). Individuals reporting sensitivity to noise have a higher prevalence of sleep disturbance with increasing proximity to major roads whereas there is no statistical relationship between distance to major roads and sleep disturbance amongst those not sensitive to noise (Table A8). The interaction term for proximity to roads and noise sensitivity is, however not statistically significant (Stata testparm function, p = 0.34). The marginal predictions of sleep disturbance in the total population resulting from models with interaction terms between noise sensitivity and distance variables are very similar to models unadjusted for noise sensitivity and without interaction terms (Tables A5, A7 and A9). Figure 4 presents prevalence of sleep disturbance in relation to LNight from the LUR noise model. Noise levels derived from the LUR model are correlated with an increase in sleep disturbance caused by environmental, transportation and road traffic noise. At nighttime levels of 60 dBA, 5.3% of individuals have their sleep disturbed by road traffic noise, 7.0% by transportation noise and 13.4% by environmental noise. The relationship between distance to the three single noise sources and LNight levels are presented in Figure A1; there is no relationship between LNight and distance to NEF25 contour or railways. The marginal predictions of sleep disturbance in the total population (by total environmental noise, transportation and road traffic) resulting from models with LNight adjusted for sensitivity are very similar to the unadjusted predictions (data not shown), although the marginal proportions are higher in the noise sensitive subgroups (Tables A10–A15). The effect of the interaction between noise sensitivity and LNight is statistically significant for all three transportation sources (Stata testparm function, p = 0.01). The marginal predictions of sleep disturbance in the total population resulting from models with interaction terms between noise sensitivity and Lnight are however, very similar to models Int.unadjusted J. Environ. Res. for Public noise Health sensitivity2016, 13, 809and without interaction terms (Tables A11, A13 and A15). 7 of 21

25 25 Sleep disturbed by air planes Sleep disturbed by road traffic 20 20

15.1 15 15

10 10 8.7 7.4 6.0

5 disturbance(%) Sleep 5 Sleep disturbance(%) 4.8 3.9 3.1 3.9 0.9 1.2 0 0 within NEF25 1-1000 1001-2000 >2000 <50 51-100 101-150 151-200 201-500 >500

Distance from the postal code to the NEF25 curve (m) Distance from the postal code to major road (m) (a) (b) 25 Sleep disturbed by railways 20

10.7 10

5 4.4

Sleep disturbance (%) disturbance Sleep 3.7 0.7 0.8 0.3 0 <=100 101-150 151-200 201-500 501-1000 >1000

Distance from the postal code to railway (m) (c)

Figure 2. Distance to aircraft NEF25 (a), major road (b), railways (c) and marginal proportions and 95% CI of sleep disturbance by railways in the weighted study sample.

The marginal predictions of sleep disturbance in the total population resulting from models adjusted for sensitivity are very similar to results from unadjusted models (data not shown), although the marginal proportions are higher in sensitive than in the non-sensitive subgroup. Sensitivity to noise does not modify the association between sleep disturbance and proximity to railway sources (Table A4). There appears to be a modification of the association between sleep disturbance and proximity to the airport according to noise sensitivity (Table A6), but the variability of the estimates is large and the interaction term is not statistically significant (p = 0.07, Stata testparm function). Individuals reporting sensitivity to noise have a higher prevalence of sleep disturbance with increasing proximity to major roads whereas there is no statistical relationship between distance to major roads and sleep disturbance amongst those not sensitive to noise (Table A8). The interaction term for proximity to roads and noise sensitivity is, however not statistically significant (Stata testparm function, p = 0.34). The marginal predictions of sleep disturbance in the total population resulting from models with interaction terms between noise sensitivity and distance variables are very similar to models unadjusted for noise sensitivity and without interaction terms (Tables A5, A7 and A9). Figure4 presents prevalence of sleep disturbance in relation to LNight from the LUR noise model. Noise levels derived from the LUR model are correlated with an increase in sleep disturbance caused by environmental, transportation and road traffic noise. At nighttime levels of 60 dBA, 5.3% of individuals have their sleep disturbed by road traffic noise, 7.0% by transportation noise and 13.4% by environmental noise. The relationship between distance to the three single noise sources and LNight levels are presented in Figure A1; there is no relationship between LNight and distance to NEF25 contour or railways. The marginal predictions of sleep disturbance in the total population (by total environmental noise, transportation and road traffic) resulting from models with LNight adjusted for sensitivity are very similar to the unadjusted predictions (data not shown), although the marginal proportions are higher in the noise sensitive subgroups (Tables A10–A15). The effect of the interaction between noise sensitivity and LNight is statistically significant for all three transportation sources (Stata testparm Int. J. Environ. Res. Public Health 2016, 13, 809 8 of 21

Int. J. Environ. Res. Public Health 2016, 13, 809 8 of 21 function,Int. J. Environ.p = 0.01).Res. Public The Health marginal 2016, 13, 809 predictions of sleep disturbance in the total population resulting8 of 21 from modelsFigure with2. Distance interaction to aircraft terms NEF25 between (a), major noise road sensitivity (b), railways and (c) Landnight marginalare however, proportions very and similar to Figure 2. Distance to aircraft NEF25 (a), major road (b), railways (c) and marginal proportions and models95% unadjusted CI of sleep for disturbance noise sensitivity by railways and in the without weighted interaction study sample. terms (Tables A11, A13 and A15). 95% CI of sleep disturbance by railways in the weighted study sample.

25 25

20 20

15 15

10 10

5 5

0 Sleep disturbance by transportation noise % by transportationdisturbance Sleep 0 Sleep disturbance by transportation noise % by transportationdisturbance Sleep Planes Roads Railways Road and plane Road and rail Not close to Planes Roads Railways Road and plane Road and railtransportation Not close to Close to one source only transportationnoise sources Close to one source only noise sources Category of exposition Category of exposition Figure 3. Marginal proportions of sleep disturbance by transportation noise according to proximity FigureFigure 3. Marginal3. Marginal proportions proportions of of sleep sleep disturbance disturbance by by transportation transportation noise noiseaccording according toto proximityproximity to to single and combined sources of transportation noise: Airplanes (≤1000 m from NEF25 or in singleto single and combined and combined sources sources of transportation of transportation noise: noise: Airplanes Airplanes (ď1000 (≤ m1000 from m NEF25from NEF25 or in NEF25),or in NEF25), roads (≤100 m from an artery or highway) and railways (≤150 m from a railway line or main roadsNEF25), (ď100 roads m from (≤100 an m artery from oran highway)artery or highway) and railways and railways (ď150 m ( from≤150 m a railwayfrom a railway line or line main or line main of a line of a railroad shunting yard). railroadline of shunting a railroad yard). shunting See yard). Table A16 for the regression model.

25 2 25 25 Sleep disturbed by environment noise 2 25 Sleep disturbed by transport noise Sleep disturbed by environment noise Sleep disturbed by transport noise 20 2 20 20 2 20 16.7 14.9 16.7 15 15 15 13.4 14.9 12.0 13.4 15 10.7 12.0 10 9.6 10.7 10.7 10 10.7 10 9.6 10 8.6 7.0 8.6 Sleep disturbance (%) disturbance Sleep 5 5.6 7.0 Sleep disturbance (%) disturbance Sleep 5 5 4.5 5.6 5 3.7 4.5 3.7 0 0 0 45 50 55 60 65 70 0 45 50 55 60 65 70 45 50 55 60 65 70 45 50 55 60 65 70 Lnight (dBA) Lnight (dBA) Lnight (dBA) (a)(bLni) ght (dBA) (a)(25 b) 25 Sleep disturbed by Road traffic noise Sleep disturbed by Road traffic noise 20 20

15 15 13.0 13.0 10 10 8.3 8.3

Sleep disturbance(%) 5 5.3 Sleep disturbance(%) 5 3.4 5.3 2.2 3.4 1.4 2.2 0 1.4 0 45 50 55 60 65 70 45 50 55 60 65 70 Lnight (dBA) Lnight (dBA) (c) (c) Figure 4. Marginal proportions and 95% CI (confidence interval) of sleep disturbance by Figure 4. Marginal proportions and 95% CI (confidence interval) of sleep disturbance by Figureenvironmental 4. Marginal noise proportions (a); transportation and 95% CI noise (conﬁdence (b); andinterval) road traffic of sleep in relation disturbance (c) to byLNight environmental estimated environmental noise (a); transportation noise (b); and road traffic in relation (c) to LNight estimated noiseby the (a); LUR transportation noise model noise in the (b weighted); and road study trafﬁc sample. in relation (c) to LNight estimated by the LUR noise by the LUR noise model in the weighted study sample. model in the weighted study sample.

Int. J. Environ. Res. Public Health 2016, 13, 809 9 of 21

4. Discussion In Montreal, 12.8% of the population reports sleep disturbance related to environmental noise exposure. We observed a strong relationship between sleep disturbance because of road, railway and aircraft noise sources and distance to the respective noise sources. On a population level, road traffic is the most frequent transport-related cause of sleep disturbance reported by respondents in our study. Our data also indicate that certain groups report very high prevalence of sleep disturbance; for example, nearly one in six people living within the NEF25 contour reports sleep disturbance caused by airplanes noise. There was an association between LNight as modeled with our LUR noise model and sleep disturbance caused by road traffic, transport and outdoor environmental noise. As LNight levels estimated with the LUR model increase, the proportion of individuals that report having their sleep disturbed by environmental noise sources, transport-related sources and by road traffic also increases. Variability in exposure, measurements methods, confounding factors and populations may explain why results in our study are not transposable to other studies and why there is no consensus on a single dose-response relationship between noise levels and sleep disturbance [39]. In a pooled data analysis of 24 studies by Miedema and Vos (2007), at an estimated exposure level of 60 dBA LNight, about 20% of individuals reported sleep disturbance due to road traffic or aircraft noise and 10% due to noise from railways [11]. At equivalent sound levels, sleep disturbance because of road traffic is four times higher in the pooled studies as compared to our study. Some of the differences may be attributable to previous studies’ reliance on numerical propagation models, as compared to the LUR models based on real measurements used in our study. It is possible that under some circumstances, noise levels estimated with propagation models underestimate the true noise levels encountered in the urban environment. LUR models provide estimates for total environmental noise whereas the propagation models estimate source-specific noise levels. A 60 dBA “pure” traffic noise at a level of 60 dBA may be more disturbing than environmental noise at an equivalent level even when close to a road. A study of three cities in Europe comparing LUR models with propagation models showed that for two of the three cities, noise propagation models may underestimate noise levels when compared to LUR models [40]. Although highly correlated, measured noise levels were also found to be higher by 6.2 dBA in Vancouver when compared to propagation models [41]. In our study, the question on sleep disturbance is dichotomous whereas in the pooled analysis by Miedema and Vos, assessment of sleep disturbance varied across studies, which may also influence the results [11]. Finally, almost all studies in the pooled analysis were performed in areas with more temperate climates than Montreal; part of the difference in the relationship between estimated noise exposure and sleep disturbance may stem from better noise insulation in the context of a colder temperature profile [21]. In addition, our survey was performed in the spring when the temperature, in particular night-time temperature, is cooler in Montreal requiring windows to be closed, especially at night. Closed windows strongly attenuate noise exposure from outside sources. In a study conducted in Brussels, Belgium, indoor noise levels in quiet areas were similar to levels in noisy areas because households in noisier areas kept their windows closed, as opposed to those in the quiet areas [42]. In a past Canadian survey, 14% of respondents indicated that road traffic noise often interfered with their ability to sleep in the last year and 26.8% of those living within 30 m of a major road reported that road traffic noise often interfered with sleep [43]. In contrast, our results indicate that 7.4% of individuals living within 50 metres of a major road report having their sleep disturbed in the last month. Discrepancies in the association between proximity to major roads and sleep disturbance between the two studies may be explained by the fact that in Michaud et al.’s study [43], a major road was defined as being either one that has four or more lanes or has a posted speed limit of 80 km an hour and over in both urban and rural contexts. In addition, In Michaud et al.’s (2007) study, distances to the road were self-reported and not defined by GIS [43]. Finally, in our study, major roads were defined as highways or arterial roads, the latter probably emitting less noise. Int. J. Environ. Res. Public Health 2016, 13, 809 10 of 21

We did not assess the health status or chronic disease status of the study participants, but have no reason to believe that Montreal’s population is healthier than the populations sampled in the other studies. Although 37.2% of our weighted study sample had a university degree, this is similar to other survey-based studies in European cities such as Malmo, Sweden (44%) [14], Belgrade, Serbia (approximately 34%) [25] and Oslo, Norway (46%) [8]. Although we have no a priori reason to believe that the individual characteristics of our study population account for the differences in the prevalence of sleep disturbance between our study and others, we cannot rule out that certain population characteristics may have contributed to some of the observed differences. In our study, the overall prevalence of sleep disturbance in the total population was not influenced by noise sensitivity although those sensitive to noise reported higher levels of sleep disturbance, irrespective of the noise source. This is consistent with a Korean and a Norwegian study that reported that sleep disturbance was more pronounced in individuals sensitive to noise [23,24]. We observed more sleep disturbance for those exposed to two transportation noise sources (i.e., rail and road noise sources) compared to those exposed to one source. This is concordant with an experimental study where those exposed to more than one noise source had more sleep fragmentation and worse sleep [13] but not with another study where those exposed were sleep quality decreased equally for those exposed only to road traffic or road traffic and ventilation noise [44]. In another experimental study, individuals reported more annoyance from aircraft noise when exposed to multiple noise sources [45]. We did not detect a trend with regards to aircraft noise in our study. It is however, possible that we did not have the statistical power to detect all trends or that one noise source predominates over the other. The dynamics of two noise sources in a non-experimental setting may also differ according to the noise source; for example, background road traffic noise may attenuate the effects of aircraft peak noise levels. Finally, there may be unmeasured confounding: for example, those exposed to both rail and traffic noise may also be more exposed to noise by heavy vehicles than those exposed only to rail noise. Future studies are needed to better quantify the possible link between the nighttime transportation noise exposure, sleep disturbance and cardiovascular disease. One of the potential limitations of our study is the cross-sectional design. This design should however, be adequate for measuring outcomes such as sleep disturbance that have a close temporal relationship with the exposure. Our study is limited by the fact that we have no objective measures of sleep disturbance (actigraphy, polysomnography etc.). The relatively low response rate (46.8%), even if the data is weighted to be representative of the population age, sex and education, may compromise the generalizability of our results to the entire population. The use of sleep questionnaires has however, been shown to be a simple, reproducible and inexpensive method for gathering information regarding sleep outcomes [46,47]. There are also limitations related to the use of a LUR model to estimate night noise exposure; our model only explained 59% of the spatial variation of LNight and its error obtained by crossvalidation was 3.9 dBA. In the absence of an independent set of measured noise levels for our study area, we could not assess the performance of the LUR model against actual levels of environmental noise; this would be an important additional step in determining the true error associated with the LUR model. The LUR model cannot isolate a specific noise source whereas propagation models can estimate source-specific noise levels. In addition, since the model integrates noise values between 11:00 PM and 6:00 AM, it does not capture night-time variations; noises of short duration and high intensity (impact noises) and intermittent noise sources are thus less well represented. Recent reviews have stressed that intermittent noise sources have more impact on sleep quality than continuous noise sources [47,48]. This may explain the absence of an observed relationship between LNight and distance to NEF25 contour as well as railways for sleep disturbance and for annoyance as shown in Ragettli et al. (2016) [31]. Our model may better represent road traffic noise, which would explain why a relation was noted between LNight and distance to major roads. The use of the metric termed intermittency ratio (IR) has recently been proposed to predict sleep disturbance [49]. The intermittency ratio expresses the proportion of the total dose of acoustic energy that is attributable to individual Int. J. Environ. Res. Public Health 2016, 13, 809 11 of 21 noise events above a certain threshold. It is possible that in Montreal, the IR is low in proximity to major roads and higher in proximity to railways and the airport; if this is the case, the IR would be more closely related to distance from these sources than the LNight and enable a better prediction model for sleep disturbance. Further research is needed to develop models that capture the influence of intermittent noise sources such as airplanes and trains. Another limit to be considered is that the use of the postal code to represent the location of respondents’ homes introduces a certain degree of noise exposure misclassification. However, there is no reason to believe that this error is directional and the effect of noise sources on sleep is thereby likely underestimated. Finally, the position of the sleeping quarters relative to the various noise sources, as well as characteristics of the respondents’ homes (insulation, placement of bedrooms, window opening behaviour, etc.) could not be considered. Indeed, a study in Stockholm, Sweden demonstrated that those more exposed to high noise levels kept their windows closed [50]. In addition, there was a better dose-response relationship to noise levels and sleep quality when noise was measured outside bedroom windows instead of the most exposed façade [50].

5. Conclusions Our results clearly demonstrate an association between distance to noise sources and prevalence of sleep disturbance in Montreal. Our study also provides a quantitative estimate of the association between total environmental noise levels estimated using an LUR model and sleep disturbance from transportation noise. Our study results indicate that sleep disturbance due to transportation noise constitutes a significant public health challenge for an urban area such as the Island of Montreal, in particular amongst individuals living in proximity to major roadways, railways and an international airport. The high proportion of individuals exposed to noise from multiple transportation-related sources and consequent impacts on health and quality of life call for city-wide measures, including zoning restrictions, traffic calming measures and improved soundproofing of private dwellings situated in proximity to noise sources. Future studies are needed to better quantify the possible link between nighttime transportation noise exposure and other health effects such as cardiovascular disease. The health benefits of noise mitigation policies and interventions should also be assessed.

Acknowledgments: This work was financially supported by the Direction de Santé Publique de l’Agence de la Santé et des Services Sociaux de Montréal and by the Institut National de Santé Publique du Québec. We are grateful to Michel Fournier for his advices on the statistical analyses. Author Contributions: Stéphane Perron conceived the study, contributed to the design of the study and writing of the manuscript, conducted the analysis, interpreted the results and wrote the first draft of the manuscript. Sophie Goudreau was involved in the data collection and analyses. Céline Plante contributed to the statistical analysis and interpretation of results. Martina S. Ragettli was involved in the data collection and analyses and interpretation of the results and David Kaiser contributed to the interpretation of the results and the writing of the manuscript. Audrey Smargiassi conceived the study, contributed to the design of the study and writing of the manuscript. All authors reviewed and approved the final manuscript. Conflicts of Interest: The authors declare no conflict of interest. Int. J. Environ. Res. Public Health 2016, 13, 809 12 of 21 Int. J. Environ. Res. Public Health 2016, 13, 809 12 of 21

AppendixAppendix 80 80 70 70 60 60 Lnight (dba) Lnight Lnight (dba) Lnight 50 50 40 40 within NEF25 1-1000 1001-2000 >2000 <=100 101-150 151-200 201-500 501-1000 >1000

(a) (b) 80 70 60 Lnight (dba) Lnight 50 40 <50 51-100 101-150 151-200 201-500 >500

(c)

Figure A1. Relation between predicted LNight and distance to: (a) NEF25; (b) railway; (c) major road. Figure A1. Relation between predicted LNight and distance to: (a) NEF25; (b) railway; (c) major road.

TableTable A1. Log-binomial A1. Log-binomial regression regression models models for for sleep sleep disturbancedisturbance due due to to airplanes airplanes in inrelation relation to the to the distance to the NEF25 contour of the Montreal International Airport. All variables are in a same model. distance to the NEF25 contour of the Montreal International Airport. All variables are in a same model. Variable PPR SE p-Value (95% CI) Distance Variableto NEF25 (m) PPR SE p-Value (95% CI) within NEF25 1 Distance to NEF25 (m) within1–1000 NEF25 0.525 1 0.136 0.013 (0.315, 0.874) 1001–20001–1000 0.346 0.525 0.229 0.136 0.108 0.013 (0.315,(0.095, 0.874) 1.263) 1001–2000>2000 0.051 0.346 0.015 0.229 0.000 0.108 (0.095,(0.028, 1.263) 0.092) Age>2000 0.051 0.015 0.000 (0.028, 0.092) 18–29Age 1 30–3918–29 8.030 1 8.633 0.053 (0.976, 66.087) 40–4930–39 10.507 8.030 9.871 8.633 0.012 0.053 (0.976,(1.665, 66.087) 66.285) 50–5940–49 8.828 10.507 7.997 9.871 0.016 0.012 (1.665,(1.495, 66.285) 52.141) 60–6950–59 11.741 8.828 10.773 7.997 0.007 0.016 (1.495,(1.943, 52.141) 70.951) 70–7960–69 5.596 11.741 5.188 10.773 0.063 0.007 (1.943,(0.909, 70.951) 34.457) ≥70–7980 4.723 5.596 5.179 5.188 0.157 0.063 (0.909,(0.550, 34.457) 40.547) Educationě80 4.723 5.179 0.157 (0.550, 40.547) NoEducation diploma 1 HighNo diplomaschool 3.754 1 2.552 0.052 (0.990, 14.231) HighCEGEP school 2.843 3.754 1.872 2.552 0.113 0.052 (0.990,(0.782, 14.231) 10.340) UniversityCEGEP degree 2.055 2.843 1.279 1.872 0.2470.113 (0.782, (0.607, 10.340) 6.963) University degree 2.055 1.279 0.247 (0.607, 6.963) Sex MenSex 1 WomenMen 1.268 1 0.403 0.454 (0.681, 2.363) Women 1.268 0.403 0.454 (0.681, 2.363)

Constant 0.008 0.008 0.000 (0.001, 0.064) CEGEP: equivalent to technical school education and is the equivalent of the last year of high school in American colleges and the ﬁrst year of university. PPR: prevalence proportion ratio. CI: conﬁdence interval. Int. J. Environ. Res. Public Health 2016, 13, 809 13 of 21

Table A2. Log-binomial regression models for sleep disturbance due to road trafﬁc in relation to the distance to major roads. All variables are in a same model.

Variable PPR SE p-Value (95% CI) Distance to a major road (m) >50 1 51–100 0.684 0.173 0.132 (0.417, 1.122) 101–150 0.418 0.140 0.009 (0.217, 0.807) 151–200 0.532 0.210 0.111 (0.245, 1.155) 201–500 0.521 0.128 0.008 (0.322, 0.842) >501 0.160 0.085 0.001 (0.056, 0.455) Age 18–29 1 30–39 1.668 0.880 0.333 (0.592, 4.693) 40–49 1.813 0.923 0.243 (0.668, 4.919) 50–59 2.685 1.358 0.051 (0.996, 7.238) 60–69 1.793 0.980 0.286 (0.614, 5.233) 70–79 1.002 0.644 0.998 (0.284, 3.530) ě80 1.620 1.204 0.516 (0.377, 6.954) Education No diploma 1 High school 0.431 0.192 0.059 (0.180, 1.032) CEGEP 0.851 0.338 0.684 (0.390, 1.854) University degree 0.723 0.271 0.387 (0.346, 1.508) Sex Men 1 Women 1.438 0.284 0.066 (0.977, 2.117) Constant 0.047 0.030 0.000 (0.013, 0.167)

Table A3. Log-binomial regression models for sleep disturbance due to railways in relation to the distance to railways. All variables are in a same model.

Variable PPR SE p-Value (95% CI) Distance to a railway (m) >100 1 101–150 0.409 0.124 0.003 (0.225, 0.743) 151–200 0.347 0.175 0.036 (0.129, 0.932) 201–500 0.066 0.031 0.000 (0.026, 0.167) 501–1000 0.076 0.035 0.000 (0.031, 0.188) >1000 0.027 0.016 0.000 (0.008, 0.087) Age 18–29 1 30–39 1.010 0.569 0.986 (0.334, 3.049) 40–49 0.483 0.251 0.161 (0.174, 1.337) 50–59 0.771 0.399 0.616 (0.280, 2.128) 60–69 1.116 0.627 0.845 (0.371, 3.355) 70–79 1.482 0.947 0.538 (0.424, 5.188) ě80 1.743 1.233 0.433 (0.435, 6.979) Education No diploma 1 High school 1.849 1.312 0.387 (0.460, 7.431) CEGEP 1.871 1.390 0.399 (0.436, 8.032) University degree 2.688 1.884 0.158 (0.680, 10.623) Sex Men 1 Women 0.879 0.237 0.634 (0.518, 1.492) Constant 0.060 0.051 0.001 (0.011, 0.318) Int. J. Environ. Res. Public Health 2016, 13, 809 14 of 21

Table A4. Distance to railways and proportions of sleep disturbance due to railways in the weighted study sample, adjusted for sensitivity to noise: regression model.

Variable PPR SE p-Value (95% CI) Distance to a Railway (m) ď100 1 101–150 0.420 0.212 0.09 (0.156, 1.131) 151–200 0.267 0.178 0.05 (0.072, 0.989) 201–500 0.027 0.028 <0.01 (0.003, 0.209) 501–1000 0.014 0.010 <0.01 (0.004, 0.054) >1000 0.024 0.026 <0.01 (0.003, 0.192) Sensitivity Not sensitive 1 Sensitive 1.387 0.603 0.45 (0.591, 3.255) Distance to railway ˆ Sensitivity interaction (base levels omitted) 101–150—Sensitive 0.924 0.602 0.90 (0.257, 3.318) 151–200—Sensitive 1.688 1.597 0.58 (0.264, 10.788) 201–500—Sensitive 3.628 4.379 0.29 (0.340, 38.668) 501–1000—Sensitive 8.142 6.781 0.01 (1.591, 41.669) Age 18–29 1 30–39 0.941 0.526 0.91 (0.314, 2.816) 40–49 0.428 0.225 0.11 (0.153, 1.199) 50–59 0.662 0.339 0.42 (0.242, 1.806) 60–69 0.972 0.550 0.96 (0.320, 2.948) 70–79 1.359 0.848 0.62 (0.400, 4.621) ě80 1.813 1.296 0.41 (0.446, 7.361) Education No diploma 1 High school 1.840 1.302 0.39 (0.459, 7.367) CEGEP 1.710 1.259 0.47 (0.404, 7.243) University degree 2.352 1.701 0.24 (0.569, 9.712) Sex Men 1 Women 0.836 0.230 0.52 (0.488, 1.433) Constant 0.0610225 0.0522719 <0.01 (0.011, 0.327)

Table A5. Distance to railways and proportions of sleep disturbance due to railways in the weighted study sample, adjusted for sensitivity to noise: marginal prediction by sensitivity.

Sensitive Proportion Not Sensitive Proportion All Proportion Variable (95% CI) (95% CI) (95% CI) ď100 12.32 (6.21, 18.44) 8.89 (3.36, 14.41) 10.67 (6.87, 14.46) 101–150 4.78 (2.06, 7.50) 3.73 (0.78, 6.68) 4.28 (2.29, 6.27) 151–200 5.56 (´1.40, 12.52) 2.37 (´0.40, 5.14) 4.02 (0.08, 7.97) 201–500 1.19 (0, 2.39) 0.24 (´0.23, 0.70) 0.73 (0.10, 1.37) 501–1000 1.44 (0.17, 2.70) 0.13 (´0.02, 0.28) 0.8 (0.16, 1.45) >1000 0.35 (´0.14, 0.84) 0.21 (´0.21, 0.64) 0.28 (´0.04, 0.61) Int. J. Environ. Res. Public Health 2016, 13, 809 15 of 21

Table A6. Distance to the NEF25 contour and marginal proportions of sleep disturbance by airplanes in the weighted study sample adjusted for sensitivity to noise: regression model.

Variable PPR SE p-Value (95% CI) Distance to NEF25 contour (m) within NEF25 1 1–1000 0.323 0.135 0.01 (0.142, 0.733) 1001–2000 0.126 0.090 0.00 (0.031, 0.512) >2000 0.023 0.014 0.00 (0.006, 0.079) Sensitivity Not sensitive 1 Sensitive 0.975 0.438 0.95 (0.404, 2.350) Distance to NEF25 ˆ Sensitivity interaction (base levels omitted) 1–1000—Sensitive 2.223 1.240 0.15 (0.744, 6.635) 1001–2000—Sensitive 3.815 4.127 0.22 (0.458, 31.806) >2000—Sensitive 3.233 2.548 0.14 (0.689, 15.161) Age 18–29 1 30–39 7.240 7.884 0.07 (0.856, 61.224) 40–49 8.793 8.229 0.02 (1.404, 55.079) 50–59 7.167 6.388 0.03 (1.249, 41.133) 60–69 10.149 9.249 0.01 (1.700, 60.581) 70–79 4.807 4.428 0.09 (0.790, 29.255) ě80 4.765 5.251 0.16 (0.549, 41.337) Education No diploma 1 High school 3.443 2.447 0.08 (0.854, 13.873) CEGEP 2.323 1.574 0.21 (0.616, 8.768) University degree 1.591 0.996 0.46 (0.467, 5.427) Sex Men 1 Women 1.204 0.385 0.56 (0.644, 2.252) Constant 0.011 0.013 0.00 (0.001, 0.100)

Table A7. Distance to the NEF25 contour and marginal proportions of sleep disturbance by airplanes in the weighted study sample adjusted for sensitivity to noise: marginal prediction by sensitivity.

Sensitive Proportion Not Sensitive All Proportion Variable (95% CI) Proportion (95% CI) (95% CI) Within NEF25 15.25 (7.15, 23.35) 15.57 (7.78, 23.37) 15.4 (9.61, 21.19) 1–1000 11.57 (8.38, 14.75) 5.80 (2.72, 8.88) 8.92 (6.69, 11.14) 1001–2000 8.15 (´2.30, 18.60) 2.37 (´0.53, 5.27) 5.49 (´0.23, 11.21) >2000 1.36 (0.69, 2.04) 0.44 (´0.06, 0.94) 0.94 (0.51, 1.36)

Table A8. Distance to major roads proportions of sleep disturbance due to road trafﬁc in the weighted study sample, adjusted for sensitivity to noise: regression model.

Variable PPR1 SE p-Value (95% CI) Distance to a major road (m) ď50 1 51–100 ´0.611 0.583 0.29 (´1.754, 0.531) 101–150 ´0.455 0.723 0.53 (´1.873, 0.962) 151–200 ´0.643 0.757 0.40 (´2.127, 0.841) 201–500 0.289 0.512 0.57 (´0.714, 1.293) >500 ´2.968 0.808 0.00 (´4.552, 1.383) Int. J. Environ. Res. Public Health 2016, 13, 809 16 of 21

Table A8. Cont.

Variable PPR1 SE p-Value (95% CI) Sensitivity Not sensitive 1 Sensitive 1.566 0.410 0.00 (0.763, 2.369) Distance road ˆ Sensitivity interaction (base levels omitted) 51–100—Sensitive 0.201 0.646 0.76 (´1.066, 1.468) 101–150—Sensitive ´0.356 0.826 0.67 (´1.976, 1.264) 151–200—Sensitive 0.021 0.877 0.98 (´1.698, 1.740) 201–500—Sensitive ´1.255 0.583 0.03 (´2.398, 0.112) >500—Sensitive 1.369 0.980 0.16 (´0.552, 3.289) Age 18–29 1 30–39 0.461 0.551 0.40 (´0.62, 1.541) 40–49 0.482 0.530 0.36 (´0.557, 1.521) 50–59 0.800 0.521 0.13 (´0.223, 1.822) 60–69 0.381 0.571 0.50 (´0.737, 1.500) 70–79 ´0.132 0.666 0.84 (´1.438, 1.174) ě80 0.470 0.762 0.54 (´1.025, 1.964) Education No diploma 1 High school ´1.023 0.453 0.02 (´1.912, 0.135) CEGEP ´0.419 0.413 0.31 (´1.229, 0.390) University degree ´0.665 0.392 0.09 (´1.434, 0.103) Sex Men 1 Women 0.308 0.199 0.12 (´0.082, 0.698) Constant ´3.785 0.711 0.00 (´5.18, 2.391)

Table A9. Distance to major roads proportions of sleep disturbance due to road trafﬁc in the weighted study sample, adjusted for sensitivity to noise: marginal prediction by sensitivity.

Sensitive Proportion Not Sensitive Proportion All Proportion Variable (95% CI) (95% CI) (95% CI) ď50 11.62 (8.14, 15.10) 2.43 (0.61, 4.24) 7.29 (5.18, 9.39) 51–100 7.71 (4.15, 11.27) 1.32 (0.15, 2.48) 4.70 (2.86, 6.54) 101–150 5.16 (1.59, 8.74) 1.54 (´0.34, 3.42) 3.45 (1.40, 5.51) 151–200 6.24 (1.10, 11.37) 1.28 (´0.37, 2.92) 3.90 (1.13, 6.67) 201–500 4.43 (2.31, 6.54) 3.24 (1.04, 5.45) 3.87 (2.35, 5.38) >500 2.35 (´0.13, 4.83) 0.12 (´0.05, 0.30) 1.30 (´0.01, 2.61)

Table A10. LNight and marginal proportions of sleep disturbance by total environmental noise in the weighted study sample, adjusted for sensitivity to noise: regression model.

Variable PPR SE p-Value (95% CI)

LNight 0.988 0.021 0.57 (0.948, 1.03) Sensitivity Not sensitive 1 Sensitive 0.106 0.158 0.13 (0.006, 1.990) Lnight ˆ Sensitive interaction 1.052 0.027 0.05 (1.000, 1.108) Age 18–29 1 30–39 1.321 0.331 0.27 (0.808, 2.159) 40–49 1.356 0.320 0.20 (0.854, 2.154) 50–59 1.523 0.346 0.06 (0.975, 2.378) 60–69 1.322 0.323 0.25 (0.819, 2.134) 70–79 1.107 0.315 0.72 (0.634, 1.932) ě80 1.236 0.439 0.55 (0.616, 2.481) Int. J. Environ. Res. Public Health 2016, 13, 809 17 of 21

Table A10. Cont.

Variable PPR SE p-Value (95% CI) Education No diploma 1 High school 0.732 0.191 0.23 (0.438, 1.222) CEGEP 1.147 0.274 0.57 (0.718, 1.831) University degree 0.973 0.223 0.90 (0.620, 1.526) Sex Men 1 Women 0.884 0.092 0.24 (0.721, 1.085) Constant 0.139 0.170 0.11 (0.013, 1.532)

Table A11. LNight and marginal proportions of sleep disturbance by total environmental noise in the weighted study sample, adjusted for sensitivity to noise: marginal prediction by sensitivity.

Sensitive Proportion Not Sensitive Proportion All Proportion Variable (95% CI) (95% CI) (95% CI) 45 dBA 11.59 (7.77, 15.41) 6.81 (4.32, 9.30) 9.36 (6.27, 12.45) 50 dBA 13.34 (10.43, 16.26) 7.39 (5.41, 9.36) 10.57 (8.30, 12.83) 55 dBA 15.36 (13.26, 17.46) 8.02 (6.40, 9.63) 11.94 (10.51, 13.37) 60 dBA 17.68 (15.08, 20.28) 8.70 (6.95, 10.45) 13.50 (11.87, 15.13) 65 dBA 20.36 (15.52, 25.20) 9.44 (6.91, 11.97) 15.27 (11.97, 18.57) 70 dBA 23.44 (15.20, 31.68) 10.24 (6.49, 14.00) 17.29 (11.54, 23.04)

Table A12. LNight and marginal proportions of sleep disturbance by transportation noise in the weighted study sample, adjusted for sensitivity to noise: regression model.

Variable PPR SE p-Value (95% CI)

LNight 0.969 0.031 0.32 (0.91, 1.031) Sensitivity Not sensitive 1 Sensitive 0.009 0.020 0.03 (0, 0.691) Lnight ˆ Sensitive interaction 1.105 0.043 0.01 (1.024, 1.192) Age 18–29 1 30–39 1.554 0.611 0.26 (0.719, 3.359) 40–49 1.484 0.556 0.29 (0.712, 3.094) 50–59 1.825 0.674 0.10 (0.885, 3.763) 60–69 1.687 0.660 0.18 (0.783, 3.634) 70–79 1.042 0.489 0.93 (0.415, 2.612) ě80 1.505 0.833 0.46 (0.508, 4.457) Education No diploma 1 High school 0.614 0.225 0.18 (0.299, 1.261) CEGEP 0.837 0.295 0.61 (0.419, 1.671) University degree 0.721 0.239 0.33 (0.376, 1.383) Sex Men 1 Women 1.175 0.180 0.29 (0.87, 1.586) Constant 0.162 0.311 0.34 (0.004, 6.979) Int. J. Environ. Res. Public Health 2016, 13, 809 18 of 21

Table A13. LNight and marginal proportions of sleep disturbance by transportation noise in the weighted study sample, adjusted for sensitivity to noise: marginal prediction by sensitivity.

Sensitive Proportion Not Sensitive Proportion All Proportion Variable (95% CI) (95% CI) (95% CI) 45 dBA 4.82 (2.38, 7.26) 0.96681 (0.97, 3.5) 3.60 (1.78, 5.42) 50 dBA 6.17 (4.09, 8.24) 1.52281 (1.52, 3.73) 4.50 (3.02, 5.98) 55 dBA 7.89 (6.26, 9.53) 2.10621 (2.11, 4.06) 5.63 (4.62, 6.64) 60 dBA 10.10 (7.98, 12.23) 2.50965 (2.51, 4.74) 7.05 (5.87, 8.24) 65 dBA 12.93 (8.5, 17.36) 2.56426 (2.56, 5.95) 8.85 (6.1, 11.6) 70 dBA 16.55 (8.07, 25.03) 2.30118 (2.3, 7.71) 11.12 (5.71, 16.52)

Table A14. LNight and marginal proportions of sleep disturbance by road noise in the weighted study sample, adjusted for sensitivity to noise: regression.

Variable PPR SE p-Value (95% CI)

LNight 1.002 0.039 0.96 (0.928, 1.082) Sensitivity Not sensitive 1 Sensitive 0.006 0.015 0.05 (0, 0.918) LNight ˆ Sensitive interaction 1.117 0.050 0.01 (1.022, 1.22) Age 18–29 1 30–39 1.645 0.839 0.33 (0.606, 4.471) 40–49 1.734 0.875 0.28 (0.644, 4.666) 50–59 2.449 1.222 0.07 (0.921, 6.512) 60–69 1.609 0.878 0.38 (0.552, 4.687) 70–79 0.993 0.637 0.99 (0.282, 3.492) ě80 1.867 1.343 0.39 (0.456, 7.65) Education No diploma 1 High school 0.343 0.153 0.02 (0.144, 0.82) CEGEP 0.688 0.278 0.36 (0.312, 1.518) University degree 0.563 0.213 0.13 (0.268, 1.183) Sex Men 1 Women 1.338 0.265 0.14 (0.907, 1.974) Constant 0.015 0.036 0.08 (0, 1.745)

Table A15. LNight and marginal proportions of sleep disturbance by road noise in the weighted study sample, adjusted for sensitivity to noise: marginal prediction by sensitivity.

Sensitive Proportion Not Sensitive Proportion All Proportion Variable (95% CI) (95% CI) (95% CI) 45 dBA 1.93 (0.82, 3.05) 0.25 (1.25, 2.15) 1.38 (0.59, 2.17) 50 dBA 3.12 (1.87, 4.37) 0.52 (1.66, 3) 2.17 (1.33, 3.01) 55 dBA 5.04 (3.73, 6.35) 0.89 (2.28, 4.21) 3.42 (2.63, 4.21) 60 dBA 8.13 (6.25, 10.01) 1.30 (3.3, 6.44) 5.41 (4.37, 6.44) 65 dBA 13.12 (8.61, 17.63) 1.63 (5.06, 11.23) 8.55 (5.86, 11.24) 70 dBA 21.18 (10.34, 32.02) 1.73 (8, 20.13) 13.55 (6.9, 20.2) Int. J. Environ. Res. Public Health 2016, 13, 809 19 of 21

Table A16. Log-binomial regression models for sleep disturbance by transportation noise according to proximity to single and combined sources of transportation noise *.

Variable PPR SE p-Value (95% CI) Not exposed 1 Airplane only 3.440 0.589 <0.01 (2.459, 4.814) Road only 1.852 0.348 <0.01 (1.281, 2.678) Railway only 2.615 0.576 <0.01 (1.698, 4.026) Road & Aircraft 2.406 0.656 <0.01 (1.409, 4.107) Road & Railway 3.764 0.860 <0.01 (2.405, 5.890) Age 18–29 1 30–39 7.240 7.884 0.07 (0.856, 61.224) 40–49 8.793 8.229 0.02 (1.404, 55.079) 50–59 7.167 6.388 0.03 (1.249, 41.133) 60–69 10.149 9.249 0.01 (1.700, 60.581) 70–79 4.807 4.428 0.09 (0.790, 29.255) ě80 4.765 5.251 0.16 (0.549, 41.337) Education No diploma 1 High school 3.443 2.447 0.08 (0.854, 13.873) CEGEP 2.323 1.574 0.21 (0.616, 8.768) University degree 1.591 0.996 0.46 (0.467, 5.427) Sex Men 1 Women 1.204 0.385 0.56 (0.644, 2.252) Constant 0.011 0.013 <0.01 (0.001, 0.100) * Airplane (ď1000 m from NEF25 or in NEF25), Road (ď100 m from an artery or highway), Railway (ď150 m from a railway line or main line of a railroad shunting yard).

References

1. Babisch, W. Updated exposure-response relationship between road traffic noise and coronary heart diseases: A meta-analysis. Noise Health 2014, 16, 1–9. [CrossRef][PubMed] 2. Babisch, W.; Kamp, I. Exposure-response relationship of the association between aircraft noise and the risk of hypertension. Noise Health 2009, 11, 161–168. [CrossRef][PubMed] 3. WHO. Night Noise Guidelines for Europe; WHO Regional Office for Europe: Copenhagen, Danmark, 2009; pp. 1–162. 4. WHO. Burden of Disease from Environmental Noise. Quantification of Healthy Life Years Lost in Europe; WHO Regional Office for Europe: Copenhagen, Danmark, 2011; pp. 1–106. 5. Sorensen, M.; Andersen, Z.J.; Nordsborg, R.B.; Becker, T.; Tjonneland, A.; Overvad, K.; Raaschou-Nielsen, O. Long-term exposure to road traffic noise and incident diabetes: A cohort study. Environ. Health Perspect. 2013, 121, 217–222. [PubMed] 6. Sorensen, M.; Hvidberg, M.; Andersen, Z.J.; Nordsborg, R.B.; Lillelund, K.G.; Jakobsen, J.; Tjønneland, A.; Overvad, K.; Raaschou-Nielsen, O. Road traffic noise and stroke: A prospective cohort study. Eur. Heart J. 2011, 32, 737–744. [CrossRef][PubMed] 7. Halonen, J.I.; Hansell, A.L.; Gulliver, J.; Morley, D.; Blangiardo, M.; Fecht, D.; Toledano, M.B.; Beevers, S.D.; Anderson, H.R.; Kelly, F.J.; et al. Road traffic noise is associated with increased cardiovascular morbidity and mortality and all-cause mortality in London. Eur. Heart. J. 2015, 36, 2653–2661. [CrossRef][PubMed] 8. Sygna, K.; Aasvang, G.M.; Aamodt, G.; Oftedal, B.; Krog, N.H. Road traffic noise, sleep and mental health. Environ. Res. 2014, 131, 17–24. [CrossRef][PubMed] 9. Floud, S.; Blangiardo, M.; Clark, C.; de Hoogh, K.; Babisch, W.; Houthuijs, D.; Swart, W.; Pershagen, G.; Katsouyanni, K.; Velonakis, M.; et al. Exposure to aircraft and road traffic noise and associations with heart disease and stroke in six European countries: A cross-sectional study. Environ. Health 2013, 12, 89. [CrossRef] [PubMed] Int. J. Environ. Res. Public Health 2016, 13, 809 20 of 21

10. Hansell, A.L.; Blangiardo, M.; Fortunato, L.; Floud, S.; de Hoogh, K.; Fecht, D.; Ghosh, R.E.; Laszlo, H.E.; Pearson, C.; Beale, L.; et al. Aircraft noise and cardiovascular disease near Heathrow airport in London: Small area study. BMJ 2013, 347, f5432. [CrossRef][PubMed] 11. Miedema, H.M.; Vos, H. Associations between self-reported sleep disturbance and environmental noise based on reanalyses of pooled data from 24 studies. Behav. Sleep Med. 2007, 5, 1–20. [CrossRef][PubMed] 12. Perron, S.; Tetreault, L.F.; King, N.; Plante, C.; Smargiassi, A. Review of the effect of aircraft noise on sleep disturbance in adults. Noise Health 2012, 14, 58–67. [CrossRef][PubMed] 13. Basner, M.; Muller, U.; Elmenhorst, E.M. Single and combined effects of air, road, and rail traffic noise on sleep and recuperation. Sleep 2011, 34, 11–23. [PubMed] 14. Bodin, T.; Bjork, J.; Ardo, J.; Albin, M. Annoyance, sleep and concentration problems due to combined traffic noise and the benefit of quiet side. Int. J. Environ. Res. Public Health 2015, 12, 1612–1628. [CrossRef][PubMed] 15. Schmidt, F.; Kolle, K.; Kreuder, K.; Schnorbus, B.; Wild, P.; Hechtner, M.; Binder, H.; Gori, T.; Münzel, T. Nighttime aircraft noise impairs endothelial function and increases blood pressure in patients with or at high risk for coronary artery disease. Clin. Res. Cardiol. 2015, 104, 23–30. [CrossRef][PubMed] 16. Schmidt, F.P.; Basner, M.; Kroger, G.; Weck, S.; Schnorbus, B.; Muttray, A.; Sariyar, M.; Binder, H.; Gori, T.; Warnholtz, A.; et al. Effect of nighttime aircraft noise exposure on endothelial function and stress hormone release in healthy adults. Eur. Heart J. 2013, 34, 3508–3514. [CrossRef][PubMed] 17. Munzel, T.; Gori, T.; Babisch, W.; Basner, M. Cardiovascular effects of environmental noise exposure. Eur. Heart J. 2014, 35, 829–836. [CrossRef][PubMed] 18. Basner, M.; Brink, M.; Bristow, A.; de Kluizenaar, Y.; Finegold, L.; Hong, J.; Janssen, S.A.; Klaeboe, R.; Leroux, T.; Liebl, A.; et al. ICBEN review of research on the biological effects of noise 2011–2014. Noise Health 2015, 17, 57–82. [CrossRef][PubMed] 19. Babisch, W.; Swart, W.; Houthuijs, D.; Selander, J.; Bluhm, G.; Pershagen, G.; Dimakopoulou, K.; Haralabidis, A.S.; Katsouyanni, K.; Davou, E.; et al. Exposure modifiers of the relationships of transportation noise with high blood pressure and noise annoyance. J. Acoust. Soc. Am. 2012, 132, 3788–3808. [CrossRef] [PubMed] 20. Öhrström, E.S.A.; Svensson, H.; Gidlöf-Gunnarsson, A. Effects of road traffic noise and the benefit of access to quietness. J. Sound Vib. 2006, 1, 40–59. [CrossRef] 21. Miedema, H.M.; Fields, J.M.; Vos, H. Effect of season and meteorological conditions on community noise annoyance. J. Acoust. Soc. Am. 2005, 117, 2853–2865. [CrossRef][PubMed] 22. Marks, A.; Griefahn, B. Associations between noise sensitivity and sleep, subjectively evaluated sleep quality, annoyance, and performance after exposure to nocturnal traffic noise. Noise Health 2007, 9, 1–7. [PubMed] 23. Aasvang, G.M.; Moum, T.; Engdahl, B. Self-reported sleep disturbances due to railway noise: Exposure-response relationships for nighttime equivalent and maximum noise levels. J. Acoust. Soc. Am. 2008, 124, 257–268. [CrossRef][PubMed] 24. Hong, J.; Kim, J.; Lim, C.; Kim, K.; Lee, S. The effects of long-term exposure to railway and road traffic noise on subjective sleep disturbance. J. Acoust. Soc. Am. 2010, 128, 2829–2835. [CrossRef][PubMed] 25. Jakovljevic, B.; Belojevic, G.; Paunovic, K.; Stojanov, V. Road traffic noise and sleep disturbances in an urban population: Cross-sectional study. Croat. Med. J. 2006, 47, 125–133. [PubMed] 26. Lee, E.Y.J.M.; Ross, Z.; Coogan, P.F.; Seto, E.Y.W. Assessment of traffic-related noise in three cities in the United States. Environ. Res. 2014, 132, 182–189. [CrossRef][PubMed] 27. King, G.; Roland-Mieszkowski, M.; Jason, T.; Rainham, D.G. Noise levels associated with urban land use. J. Urban Health 2012, 89, 1017–1030. [CrossRef][PubMed] 28. Zuo, F.; Li, Y.; Johnson, S.; Johnson, J.; Varughese, S.; Copes, R.; Liu, F.; Wu, H.J.; Hou, R.; Chen, H. Temporal and spatial variability of traffic-related noise in the City of Toronto, Canada. Sci. Total Environ. 2014, 472, 1100–1107. [CrossRef][PubMed] 29. Goudreau, S.; Plante, C.; Fournier, M.; Brand, A.; Roche, Y.; Smargiassi, A. Estimation of spatial variations in urban noise levels with a land use regression model. Environ. Polluti. 2014, 3, 48–58. [CrossRef] 30. Statistics Canada. Population and Dwelling Counts, for Canada and Economic Regions, 2011 and 2006 Censuses. Available online: https://www12.statcan.gc.ca/census-recensement/2011/dp-pd/hlt-fst/pd- pl/Table-Tableau.cfm?LANG=Eng&T=1401&SR=1&S=9&O=D&RPP=25 (accessed on 22 December 2015). Int. J. Environ. Res. Public Health 2016, 13, 809 21 of 21

31. Ragettli, M.S.; Goudreau, S.; Plante, C.; Perron, S.; Fournier, M.; Smargiassi, A. Annoyance from Road Traffic, Trains, Airplanes and from Total Environmental Noise Levels. Int. J. Environ. Res. Public. Health 2016, 13, 90. [CrossRef][PubMed] 32. BOEING. Courbe D’ambiance Sonore. Available online: http://www.boeing.com/commercial/noise/ montreal2011.pdf (accessed on 11 November 2011). 33. Niemann, H.; Bonnefoy, X.; Braubach, M.; Hecht, K.; Maschke, C.; Rodrigues, C.; Röbbel, N. Noise-induced annoyance and morbidity results from the pan-European LARES study. Noise Health 2006, 8, 63–79. [CrossRef][PubMed] 34. Ragettli, M.S.; Goudreau, S.; Plante, C.; Fournier, M.; Hatzopoulou, M.; Perron, S.; Smargiassi, A. Statistical modeling of the spatial variability of environmental noise levels in Montreal, Canada, using noise measurements and land use characteristics. J. Expo. Sci. Environ. Epidemiol. 2016, 1.[CrossRef][PubMed] 35. Ohrstrom, E.; Barregard, L.; Andersson, E.; Skanberg, A.; Svensson, H.; Angerheim, P. Annoyance due to single and combined sound exposure from railway and road traffic. J. Acoust. Soc. Am. 2007, 122, 2642–2652. [CrossRef][PubMed] 36. Dale, L.M.; Debia, M.; Mudaheranwa, O.C.; Plante, C.; Smargiassi, A. An exploration of transportation source contribution to noise levels near an airport. Environ. Pollut. 2013, 3, 73–81. [CrossRef] 37. Frankel, M. Sampling theory. In Handbook of Survey Research, 2nd ed.; Marsden, P.V., Wright, J.D., Eds.; Emerald Group Publishing Limited: Bingley, UK, 2010. 38. American Association for Public Opinion Research (AAPOR). Response Rates—An Overview. Available online: http://www.aapor.org/AAPORKentico/Education-Resources/For-Researchers/Poll-Survey-FAQ/ Response-Rates-An-Overview.aspx (accessed on 10 June 2015). 39. Hume, K.I.; Brink, M.; Basner, M. Effects of environmental noise on sleep. Noise Health 2012, 14, 297–302. [CrossRef][PubMed] 40. Aguilera, I.; Foraster, M.; Basagana, X.; Corradi, E.; Deltell, A.; Morelli, X.; Phuleria, H.C.; Ragettli, M.S.; Rivera, M.; Thomasson, A.; et al. Application of land use regression modelling to assess the spatial distribution of road traffic noise in three European cities. J. Expo. Sci. Environ. Epidemiol. 2015, 25, 97–105. [CrossRef][PubMed] 41. Gan, W.Q.; McLean, K.; Brauer, M.; Chiarello, S.A.; Davies, H.W. Modeling population exposure to community noise and air pollution in a large metropolitan area. Environ. Res. 2012, 116, 11–16. [CrossRef] [PubMed] 42. Pirrera, S.; De Valck, E.; Cluydts, R. Field study on the impact of nocturnal road traffic noise on sleep: The importance of in- and outdoor noise assessment, the bedroom location and nighttime noise disturbances. Sci. Total. Environ. 2014, 500, 84–90. [CrossRef][PubMed] 43. Michaud, D.S.; Fidell, S.; Pearsons, K.; Campbell, K.C.; Keith, S.E. Review of field studies of aircraft noise-induced sleep disturbance. J. Acoust. Soc. Am. 2007, 121, 32–41. [CrossRef][PubMed] 44. Öhrström, E.; Skånberg, A. Sleep disturbances from road traffic and ventilation noise—Laboratory and field experiments. J. Sound. Vib. 2004, 271, 279–296. [CrossRef] 45. Elmenhorst, E.M.; Quehl, J.; Muller, U.; Basner, M. Nocturnal air, road, and rail traffic noise and daytime cognitive performance and annoyance. J. Acoust. Soc. Am. 2014, 135, 213–222. [CrossRef][PubMed] 46. Muzet, A. Environmental noise, sleep and health. Sleep Med. Rev. 2007, 11, 135–142. [CrossRef][PubMed] 47. Pirrera, S.; De Valck, E.; Cluydts, R. Nocturnal road traffic noise: A review on its assessment and consequences on sleep and health. Environ. Int. 2010, 36, 492–498. [CrossRef][PubMed] 48. Kawada, T. Noise and health—Sleep disturbance in adults. J. Occup. Health 2011, 53, 413–416. [CrossRef] [PubMed] 49. Wunderli, J.M.; Pieren, R.; Habermacher, M.; Vienneau, D.; Cajochen, C.; Probst-Hensch, N.; Röösli, M.; Brink, M. Intermittency ratio: A metric reflecting short-term temporal variations of transportation noise exposure. J. Expo. Sci. Environ. Epidemiol. 2015, 1–11. [CrossRef][PubMed] 50. Öhrström, E.; Hadzibajramovic, E.; Holmes, M.; Svensson, H. Effects of road traffic noise on sleep: Studies on children and adults. J. Environ. Psychol. 2006, 26, 116–126. [CrossRef]

© 2016 by the authors; licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC-BY) license (http://creativecommons.org/licenses/by/4.0/). Int J Public Health (2013) 58:649–666 DOI 10.1007/s00038-013-0489-7

REVIEW

Cardiovascular health, trafﬁc-related air pollution and noise: are associations mutually confounded? A systematic review

Louis-Franc¸ois Te´treault • Ste´phane Perron • Audrey Smargiassi

Received: 31 March 2013 / Revised: 25 June 2013 / Accepted: 27 June 2013 / Published online: 26 July 2013 The Author(s) 2013. This article is published with open access at Springerlink.com

Abstract Conclusions Results from this review suggest that con- Objectives This review assessed the confounding effect founding of cardiovascular effects by noise or air pollutants of one traffic-related exposure (noise or air pollutants) on is low, though with further improvements in exposure the association between the other exposure and cardio- assessment, the situation may change. More studies using vascular outcomes. pollution indicators specific to road traffic are needed to Methods A systematic review was conducted with the properly assess if noise and air pollution are subjected to databases Medline and Embase. The confounding effects in confounding. studies were assessed by using change in the estimate with a 10 % cutoff point. The influence on the change in the estimate Keywords Noise Air pollution Confounding of the quality of the studies, the exposure assessment methods Cardiovascular and the correlation between road noise and air pollutions were also assessed. Results Nine publications were identified. For most studies, Introduction the specified confounders produced changes in estimates \10 %. The correlation between noise and pollutants, the Studies have shown that exposure to road traffic noise and quality of the study and of the exposure assessment do not air pollutant emissions can lead to adverse health effects seem to influence the confounding effects. such as annoyance (World Health Organisation (2011), sleep disturbance (Pirrera et al. 2010), respiratory problems This article is part of the special issue: ‘‘Environment and Health (Health Effects Institute Panel on the Health Effects of Reviews’’. Traffic-Related Air Pollution 2010) and cancer (Beelen et al. 2008). A few studies (Finkelstein et al. 2004; Gan et al. 2010; L.-F. Te´treault A. Smargiassi (&) Hoffmann et al. 2006, 2009; Maheswaran and Elliott 2003) De´partement de sante´ environnementale et sante´ au travaill, have also reported associations with cardiovascular (CV) University of Montreal, 2375 Chemin de la Coˆte-Sainte- Catherine, Montreál, QC H3T 1A8, Canada outcomes. CV health effects could be explained either by e-mail: [email protected] noise or air pollution associated with traffic. L.-F. Te´treault On the one hand, there is evidence linking traffic noise e-mail: [email protected] to ischemic heart diseases (World Health Organisation 2011; Babisch and Kamp 2009; Babisch 2006) and S. Perron hypertension (World Health Organisation Regional Office De´partement de me´decine sociale et pre´ventive, University of Montreal, 1301 Sherbrooke est, for Europe 2011; van Kempen and Babisch 2012). On the Montreal, QC H2L 1M3, Canada other hand, there is also evidence linking traffic-generated e-mail: [email protected] air pollution to CV diseases (Hoek et al. 2002; Brook et al. 2010). However, in a recent report of the Health Effects S. Perron Direction de sante´ publique de l’Agence de la sante´ et des Institute Panel on the Health Effects of Traffic-Related Air services sociaux de Montreál, Quebec, Canada Pollution 2010), the evidence of the effects of traffic- 123 650 L.-F. Te´treault et al. related air pollution on cardiovascular mortality was con- noise and air pollution. This is fundamental to better steer sidered suggestive, but not sufficient. Nonetheless, there is public health interventions and policies aimed at reducing increasing evidence connecting air pollution to overall CV effects of road traffic. For example, if road traffic induces cardiovascular mortality (Chen et al. 2013), myocardial CV effects through noise levels, then regulations to reduce car infarction (Nuvolone et al. 2011; Rosenlund et al. 2006), air pollutant emissions may not tackle the problem and zoning atherosclerosis (Adar et al. 2013) and atrial fibrillation by-laws could be more appropriate. In this article, we (Liao et al. 2010). Although noise or air pollution can reviewed epidemiological evidences that looked at the con- confound the effect of each other, the underlying physio- founding effect of one traffic-related exposure (noise or air logical mechanisms are likely different. pollution) on the association between the exposure to traffic- A few mechanisms are postulated to explain the effect related noise or air pollution and CV outcomes. of noise on CV diseases. Noise can act as a general stressor disturbing the body homeostasis through the ‘‘stress syn- drome’’ (Babisch and Kamp 2009; Westman and Walters Methods 1981; World Health Organisation Regional Office for Europe 2011; Amato et al. 2010; Gan et al. 2012). Noise Data sources can induce stress by two different pathways. The stress response generated by the direct pathway consists of a The bibliographic databases used were Medline and Else- neural activation by the noise. In the indirect pathway, the vier Embase on the Ovid SP portal. Only studies published activation of the stress response is created by a cognitive until November 2012 were considered. No other temporal interpretation of the noise (Westman and Walters 1981; limitation was applied. Gan et al. 2012). The activation of this pathway can be influenced by the perception of the noise, the perceived Extraction strategy control over the sound and noise sensitivity of the subjects (European Environment Agency (EEA) 2010). There are The strategy used to conduct this review consisted of a also pathways by which noise could increase the risk of combination of keywords representing three distinct cate- cardiovascular diseases. One of those is the disruption of gories: (1) exposure to traffic air pollution, (2) traffic noise sleeping patterns. Studies have associated the lack of exposure and (3) cardiovascular outcomes (see ‘‘Appen- reduction of blood pressure (BP) during the night (‘‘BP dix’’ for specific keywords). Only peer review articles dipping’’) caused by noise with an increased risk of car- written in English or French on road traffic were reviewed. diovascular outcomes (Vardeny et al. 2011; Haralabidis The studies were then selected manually according to the et al. 2011). It is also suggested that short sleep durations following exclusion criteria: may result in higher ghrelin and lower leptin concentra- • Commentaries, editorials, review articles tions (Taheri et al. 2004). The deregulation of those • Studies not related to road traffic hormones linked with appetite regulation could potentially • Studies not performed on humans lead to obesity and higher risk of cardiovascular diseases. • Studies with no simultaneous exposure to noise and air For air pollution, several pathways are hypothesized to pollutants explain its impact on CV diseases. Firstly, air pollutants could • Studies with no assessment of cardiovascular effects generate an increase in lung oxidative stress and inflamma- • Studies not reporting the impact of confounding tion. Some pollutants could also migrate through the variables pulmonary epithelium into blood. Those pathways could lead to a systemic and vascular inflammation, increasing the risk of Studies using the same cohort but assessing different CV hypertension and thrombogenesis. Another hypothesized outcomes were included. The references of each selected mechanism is the activation of the pulmonary reflex by par- article were consulted to ensure that all pertinent infor- ticulate matter leading to the activation of the sympathetic mation was gathered in our review. Experts in the field system. A chronic activation of this system could lead to were also consulted to see if further articles could be hypertension, plaque instability and cardiac arrhythmias added. Finally, gray literature (OAIster database, WHO (Brook et al. 2004;Burganetal.2010; Mills et al. 2007). and the New York Academy of Medicine) was also con- As motor vehicles are the predominant source of both air sulted. Studies presenting the associations before and after pollution and noise in many cities (Allen and Adar 2011), the authors controlled for the co-variables of interest (e.g., the reported associations between road traffic exposure and traffic noise or air pollution) were reviewed. CV diseases could be influenced by a confounding or The articles selected were then separated into two cat- interaction between those two pollutants. Yet, few studies egories: studies of CV outcomes where the noise exposure have aimed to untangle the possible effects of road traffic effect was adjusted for air pollutants and studies evaluating 123 Cardiovascular health, traffic-related air pollution and noise: are associations mutually confounded 651 the effect of air pollutants, adjusted for the noise effect. looked at blood pressure or hypertension (Clark et al. 2012; In this review, cardiovascular findings of the original Dratva et al. 2012; Sorensen et al. 2012b; de Kluizenaar studies are reported before and after the control of either et al. 2007), and one article was identified for both heart road traffic noise or air pollution effects. The effect of failure and cardiac dysrhythmia (Beelen et al. 2009). the supposed confounders was assessed using the per- Tables 1 and 2 present the associations between car- centage of variation in the estimate as follows: the diovascular outcomes and either road traffic air pollution or confounding effect (C) of noise or air pollution was noise. The tables also summarize the population charac- evaluated by a change in the point estimate (CIE) with a teristics of each study, the health outcomes monitored, as cutoff point of 10 % (Eq. 1) (Bliss et al. 2012; Vitting- well as the noise and air pollutants indicators used. When hoff et al. 2012). available, the background exposure levels to pollution and Adjusted point estimate Unadjusted point estimate 100 ¼ C noise of the subjects were reported. Unadjusted point estimate Six studies used day–evening–night noise levels (Lden), ð1Þ as indicators of exposure to noise levels from road traffic (Beelen et al. 2009; de Kluizenaar et al. 2007; Gan et al. To assess the quality of the study and identify possible 2012; Selander et al. 2009; Sorensen et al. 2011, 2012a, b). biases within the studies reviewed, two authors (LFT and Others studies used equivalent noise levels over 24 h

SP) reviewed independently each study. The methodology (Leq,24h) (Selander et al. 2009), night noise levels (Lnight) applied has been described previously (Perron et al. (Dratva et al. 2012) or daily noise levels (Lday) (Clark et al. 2012). Briefly, we considered the study’s design selection 2012; Dratva et al. 2012). All studies reported used dis- and classification bias relevant to our review. For each persion noise models, but only four reported validation type of bias, the distinction between major and minor information on the model used (Beelen et al. 2009; biases was assessed qualitatively by both authors. We Sorensen et al. 2011, 2012a, b). In those studies, the defined major biases as those that could invalidate the average difference between the estimated noise levels and results presented in the study. Minor biases are expected measured noise levels was 3.1 dB or less. Regarding to affect the association studied, but unlikely to invalidate exposure to air pollutants, seven studies used nitrogen the conclusion. oxide (NOx or NO2) as an indicator (Clark et al. 2012; Dratva et al. 2012; Gan et al. 2012; Selander et al. 2009; Sorensen et al. 2011, 2012a, b). Two articles reported

Results results based on particulate matter (PM10 or PM2.5) (de Kluizenaar et al. 2007; Gan et al. 2012). Two studies used Figure 1 presents a flowchart describing the stages of measurements of black smoke or black carbon to account selection for the studies used in this review. Our keyword for road traffic pollutants (Beelen et al. 2009; Gan et al. search yielded 221 starting articles (‘‘Appendix’’), which 2012). Of those studies, two applied land use regression were reduced to 16 articles by applying the various inclu- models (Beelen et al. 2009; Gan et al. 2012) to assess the sion and exclusions criteria. A large number of irrelevant air pollution exposure, while the other studies employed articles had to be sifted by hand because of the double dispersion models (Clark et al. 2012; de Kluizenaar et al. meaning of the terms sound and noise. We added four 2007; Dratva et al. 2012; Selander et al. 2009; Sorensen articles identified by snowballing and expert consultation. et al. 2011, 2012a, b). As noted in Tables 1 and 2, seven Of these 20 articles, a final 11 studies were discarded out of the nine studies reviewed reported validation infor- because it was impossible to assess the impact of the mation on their air pollution model (Beelen et al. 2009; confounding variables of interest (Fig. 1). We did not find Clark et al. 2012; Dratva et al. 2012; Gan et al. 2012; any document in the gray literature that could benefit this Sorensen et al. 2011, 2012a, b). The Clark et al. study review. reported an average difference between the estimated and Nine studies evaluating the relationship between CV the measured pollutant levels of 2.4 ppb. The remaining outcomes, noise and traffic-related air pollution met our studies presented R2 that ranged from 0.52 to 0.75. All criterion (Beelen et al. 2009; Clark et al. 2012; de Klui- articles presented average yearly outdoor exposure for both zenaar et al. 2007; Dratva et al. 2012; Gan et al. 2012; noise and air pollutant levels as a proxy of individual Selander et al. 2009; Sorensen et al. 2011, 2012a, b). Of exposure. The correlations between traffic noise and air those articles, one assessed CV disease mortalities in pollutants in the studies reviewed range from 0.16 to 0.72. general (Beelen et al. 2009), four assessed ischemic heart It is surprising that studies where both noise and air pol- diseases (Beelen et al. 2009; Gan et al. 2012; Selander et al. lutants were modeled, relying on some of the same 2009; Sorensen et al. 2012a), two assessed cerebrovascular variables (e.g., distance to traffic source), could generate diseases (Beelen et al. 2009; Sorensen et al. 2011), four low and disparate correlations. 123 652 L.-F. Te´treault et al.

Fig. 1 Stages of the selection of studies for analysis Potentially relevant studies identified with key words (n=221) Commentaries, editorials, review articles, studies not performed in humans, not related to road traffic, not in French or English (n=177) Potentially relevant studies identified with key words and screened for retrieval (n=44) Studies with no exposure to both noise and air pollutants simultaneously (n=15) Potentially relevant studies identified with key words and screened for retrieval (n=29)

Studies without cardiovascular assessment (n=13)

Potentially relevant studies identified with key words and screened for retrieval (n=16) Studies retrieved from snowballing and consultation with experts (n=4)

Studies identified to be included in the analysis (n=20) Studies not reporting the impact of the confounding variables (n=11) Studies analysed (n=9)

Quality of studies while controlling for air pollutants. Gan et al. (2012) reported an association between death from coronary heart

Table 3 presents the assessment of the quality of the disease and a 10 dB(A) elevation of the Lden level of 1.13 studies by LFT and SP. All studies, with the exception of (95 % CI 1.06, 1.21) before and 1.09 (95 % CI 1.01, 1.18)

Selander et al. (2009) and Dratva et al. (2012), only con- after controlling for NO2,PM2.5 and black carbon. Models sidered the exposure at home or at work/school, which we adjusted for PM2.5 and NO2 only in Gan et al. (2012), considered as a minor classification bias likely to be non- produced small modifications of the point estimate (Gan differential. Two studies also used medical administrative et al. 2012). The IRR estimated by Sorensen et al. (2012a) databases to identify the cause of death, potentially leading for the association between the incidence of myocardial to another minor classification bias. Four studies reported infarction and Lden was 1.10 (95 % CI 1.03, 1.19) per response rates between 30 and 60 % which could lead to a increase of 10 dB(A); this IRR increased to 1.12 (95 % CI selection bias and one did not report the response rate. 1.02, 1.22) after adjustment for NOx. Selander et al. (2009) Finally, four studies used cross-sectional designs to assess only reported an association between myocardial infarction associations with cardiovascular outcomes reported in this and Leq,24h after controlling for NO2 (1.12 (95 % CI 0.95, review. Most studies contained less than four minor biases. 1.33) with the reference category \50 dB[A]). While Se- We did not identify any major bias in the studies assessed. lander et al. (2009) did not present point estimates before and after adjusting for noise, the authors reported a 7 % Noise effects controlled for air pollution effects change of the crude coefficient when compared with the model adjusted for air pollution. The percentage changes in As illustrated in Table 1, four studies assessed the associ- the point estimates in the studies mentioned above were all ation between noise levels and ischemic heart diseases below 10 % (ranging between 0 and 7 %). Beelen et al.

123 Table 1 Estimation of the effects of exposure to noise levels on cardiovascular mortality and morbidity while controlling for air pollution (Sweden 653 [1992–1994], Denmark [1993–2006], confounded mutually associations are noise: and pollution air trafﬁc-related health, Cardiovascular Canada [1994–2002], Switzerland [2002–2003], Netherlands [1997–1998] and, [1987–1996])

Studies Characteristics of the Traffic noise exposure Air pollution exposure (air Exposure levels Health outcome Main findings: (95 % Percentage of studya (noise indicator) pollution indicator) (definition) confidence interval) change in the [validation of the model] [validation of the model] estimate

de Kluizenaar Groningen Cross-sectional Dispersion model : The Dispersion models: Average (SD) Lden in the Self-reported OR per 10 dB(A) increase Full sample et al. sample: N: 40,856 Standaart Kartering (i) Local trafﬁc contribution Groningen sample: antihypertensive Full sample 1.98 % (2007) Method 2 implemented from the model CAR II No AHT: 53.3 (6.9) dB(A) medication intake Subgroup 45–55 28–75 years ORA 1.01 (0.96–1.06) in Urbis (Lden ) [no (ii) The Gaussian dispersion AHT: 54.6 (7.0) dB(A) years 10.19 % Groningen, Netherlands information on ORP 1.03 (0.96–1.11) model ‘‘Pluim’’ 1997–1998 validation] Median (5–95 percentile) PM10 Subgroup 45–55 years (PM10) level in the Groningen ORA 1.08 (0.97–1.20) [no information on sample: validation] No AHT: 33.5 (32.8–37.5) ORP 1.19 (1.02–1.40) PREVENT Cross-sectional lg/m3 Hypertension: use of OR per 10 dB(A) increase Full sample cohort N: 8,592 AHT: 33.6 (32.9–37.6) lg/m3 antihypertensive Full sample 0.93 % sub medication (pharmacy Subgroup 45–55 28–75 years ORA 1.07 (0.98–1.18) sample record) or systolic years 9.45 % Groningen, Netherlands blood pressure C140 ORP 1.08 (0.95–1.23) 1997–1998 and diastolic blood Subgroup 45–55 years pressure C90 (mean of the last 2 ORA 1.27 (1.08–1.49) measurements from ORP 1.39 (1.08–1.77) the 2 visits)

Beelen et al. Cohort Dispersion model :Empara Land use regression Average Lden level (SD) : Mortality from: ischemic RR compared to a reference Overall (2009) N: 117,528 (Lden ) (black smoke) 52 (7) dB(A) heart disease, category of B50 dB(A) cardiovascular cardiovascular mortality 55–69 years [Measured vs. Estimated: [R2 = 0.59] Black smoke average level: Overall cardiovascular on average 13.9 (2.2) lg/m3 disease, mortality 6.40 % Netherlands \2–3 dB(A)] 3 cerebrovascular Ischemic heart NO2 average level: 30 mg/m RRA 1.25 (1.01–1.53) 1987–1996 disease, heart failure disease and cardiac RRP 1.17 (0.94–1.45) mortality dysrhythmia (ICD 9 Ischemic heart disease 12.17 % for 1986–1995 and mortality ICD 10 for 1996) Cerebrovascular RRA 1.15 (0.86–1.53) mortality 7.95 % RRP 1.01 (0.74–1.36) Cerebrovascular mortality Heart failure mortality RRA 0.88 (0.52–1.50) 4.52 % RRP 0.95 (0.55–1.66) Cardiac Heart failure mortality dysrhythmia mortality RRA 1.99 (1.05–3.79) 0.00 % RRP 1.90 (0.96–3.78) Cardiac dysrhythmia mortality

RRA 1.23 (0.50–3.01)

RRP 1.23 (0.48–3.13) 123 654 123

Table 1 continued

Selander et al. Case–control Dispersion model: Dispersion model: Gaussian LA,eq,24h :NA Myocardial infarction OR compared to a reference 7% (2009) simpliﬁed Nordic Air Quality Dispersion 3 (coronary records, category of B50 dB N: 2,095 (controls) ? NO2: median level 12.9 lg/m 1,571 (cases) prediction method model (NO2) for controls hospital discharge Full sample (LA,eq,24h ) [No information on register and the 45–70 years Median level for cases: NA ORP 1.12 (0.95–1.33) [No information on validation] National Cause of Stockholm country, validation] Death at statistic Sweden Sweden) 1992–1994

Sorensen et al. Cohort Dispersion model: Sound Dispersion model: Median Lden (5–95 percentile) : Stroke (hospital IRR per 10 dB(A) increase 3.64 % (2011) N: 51,485 plan with the Nordic Danish AirGis \64.5 years 57.8 (NA) discharge register ICD Full sample prediction method, 2 dB(A) and C64.5 years 58.2 8 and 10) (NOX)[R = 0.75] 55–64 years DANSIM and INM 3 (NA) dB(A) IRRA 1.10 (1.03–1.18) Copenhagen or Aarthus, (Lden ) IRRP 1.14 (1.03–1.25) NOX 5–95 percentile) median Denmark [Measured vs. Estimated: levels at 1993–2006 on average 0.2 dB] Lden B60 dB 18.5 (14.1–28.3) 3 lg/m and Lden [60 dB 34.3 (16.9–137) lg/m3

Sorensen et al. Cohort Dispersion model: Sound Dispersion model: Median Lden (5–95 percentile) Myocardial infarction IRR per 10 dB(A) increase 1.82 % (2012a) N: 50,614 plan with the Nordic Danish AirGis 56.4 (48.5–70.1) dB(A) (ICD 10) Full sample (per 10 dB[A]) prediction method, 2 (NOX)[R = 0.75] Median NOX (5–95 percentile) 55–64 years IRRcrude 1.10 (1.03–1.19) DANSIM and INM 3 20.8 (14.4–88.0) lg/m3 Copenhagen or Aarthus, (Lden ) IRRAdj 1.12 (1.02–1.22) Denmark [Measured vs. Estimated: 1993–2006 on average 0.2 dB]

Gan et al. Cohort Dispersion model: CadnaA Land use regression Average Lden (SD): 63.4 Ischemic heart diseases RR per increase of 10 dB(A) PM2.5 (2012) base model using the (5.0) dB(A) mortalities N: 445,868 (NO2,PM2.5 and black PM2.5 0.00 % EMME/2 for trafﬁc carbon) PM2.5 average level (SD): (ICD-9 and ICD-10) 45–85 years 3 RRA 1.13 (1.06–1.21) NO2 ? PM2.5 volume 2 4.10 (1.64) lg/m [NO2: R = 0.56 PM2.5 Vancouver Canada RR P 1.13 (1.06–1.21) 0.88 % (Lden ) 2 R = 0.52 Black carbon NO2 average level (SD): 1994–2002 [No information on NA] 32.3 (8.1) lg/m3 NO2 ? PM2.5 Black carbon ? NO2 ? PM2.5 validation] RRA 1.13 (1.06–1.21) NOX average level (SD): 3 3.54 % 32.2 (12.0) lg/m RR P 1.12 (1.05–1.21)

Black carbon average level Black carbon ? NO2 ? PM2.5 (SD): 1.50 (1.1) 10-5 /m RRA 1.13 (1.06–1.21)

RR P 1.09 (1.01–1.18) Te L.-F. ´ ral tal. et treault Table 1 continued 655 confounded mutually associations are noise: and pollution air trafﬁc-related health, Cardiovascular

Dratva et al. Cross-sectional Dispersion model: Dispersion model: PolluMap Average (SD) Lday : Blood pressure Increase in BP per 10 dB (A) Nighttime (2012) N: 6,450 SONABASE Gaussian dispersion model 50.5 (7.2) dB(A) (measured by the Night time systolic BP systolic BP Riva-Rocci method by 1,600 % (LDay and Lnight ) (NO2) Average (SD) Lnight : 28–72 years trained ﬁeld workers) bA: -0.01 (-0.6 to 0.59) [Measured vs. Estimated: [R2 = 0.72] 38.7 (7.8) dB(A) Nighttime Switzerland bAP : 0.15 (-0.48 to 0.77) on average ± 2.6 Average (SD) levels for: diastolic BP 2002–2003 3 Nighttime diastolic BP 200 % dB(A) (day) ± 3.1 NO2: 23.0 (9.9) lg/m dB(A) (night)] 3 bA: -0.05 (-0.41 to 0.30) Daytime systolic PM10: 21.3 (7.1) lg/m BP 145.45 % bAP : -0.15 (-0.36 to 0.39) Daytime systolic BP Daytime diastolic BP 60 % bA: -0.11 (-0.68 to 0.47)

bAP : 0.05 (-0.56 to 0.07) Daytime diastolic BP

bA: -0.10 (-0.44 to 0.24)

bAP : -0.04 (-0.40 to 0.33) a Final sample sizes used for analysis OR odds ratio, NA not available, IRR incident rate ratio, RR risk ratio, A adjusted for some of the following potential confounding factors: age, education, employment, marital status, study area, mean pulse, hearing impairment, noise at work, crowding, home ownership, mother’s educational level, language spoken at home, parental support for schoolwork, classroom window glazing, body mass index, cuff size, room temperature, birth weight, parental high blood pressure, prematurity, smoking status, family history of CVD, physical inactivity smoking intensity, intake of fruits, intake of vegetables, intake of coffee, alcohol intake diabetes, antihypertensive medication, high blood pressure, long-standing illness and other comorbidity (see original article for details), P adjusted for some potential confounding factors and air pollution levels, AHT antihypertensive treatment, SD standard deviation, b regression coefficient, PM2.5 particles with a diameter of 2.5 lm or smaller, PM10 particles with a diameter of larger than 2.5 lm, but smaller than 10 lm, NO2 Nitrogen dioxide, NOX Nitrogen oxide, dB(A) A-weighted decibels, Lday integrated A-weighted sound level over 16 h (0600–2200), Lnight integrated A-weighted sound level over 8 h (2200–0600), LA,eq,24h integrated A-weighted sound level over 24 h, Lden integrated A-weighted sound level over 24 h (day, evening and night) in which sound levels during the evening (1900–2300 hours) are increased by 5 dB(A) and those during the night (2300–0700 hours) by 10 dB(A) 123 656 123 Table 2 Estimation of effects of exposure to ambient air pollutants on cardiovascular mortality and morbidity while controlling for noise (Netherlands [1987–1996], Denmark [2000–2002] and UK [2001–2003]) Studies Characteristics Traffic noise exposure Air pollution Exposure levels Health outcome (definition) Main findings : air pollutions Percentage of of the studya (noise indicator) exposure (air effects adjusted for noise effects change in the [validation of the model] pollution indicator) estimate [validation of the model]

3 Beelen et al. Cohort Dispersion model Land use regression Average Lden Mortality from: ischemic heart RR for an increase of 10 lg/m of Overall (2009) N 117,528 :Empara (Lden ) (black smoke) level (SD) : disease, cardiovascular disease, black smoke and adjusted for cardiovascular 2 52 (7) dB(A) cerebrovascular disease, heart trafﬁc intensity mortality 55–69 years [Measured vs. Estimated: [R = 0.59] on average Black smoke failure and cardiac dysrhythmia Overall cardiovascular mortality 0.00 % Netherlands [0.00 %] \2–3 dB(A)] average level: (ICD 9 for 1986–1995 and ICD 10 RRA 1.11 (0.96–1.28) RRPT 1.11 1987–1996 13.9 (2.2) for 1996) (0.95–1.28) [RRA 1.01 (1.00–1.02) Ischemic heart 3 lg/m RRPT 1.01 (0.99–1.02)] disease mortality NO2 average Ischemic heart disease mortality 0.00 % level : RRA 1.01 (0.83–1.22) RRPT 1.01 3 [0.00 %] 30 mg/m (0.83–1.22) [RRA 1.00 (0.98–1.02)

RRPT 1.00 (0.98–1.02)] Cerebrovascular mortality Cerebrovascular mortality RRA 1.44 % 1.39 (0.99–1.94) RRPT 1.41 [0.14 %] (1.01–1.97) [RRA 1.03 (1.00–1.07)

RRPT 1.03 (1.00–1.07)] Heart failure mortality Heart failure mortality RRA 1.75 0.57 % (1.00–3.05) RRPT 1.76 (1.01–3.08) [0.06 %] [RRA 1.06 (1.00–1.12) RRPT 1.06 (1.00–1.12)] Cardiac dysrhythmia Cardiac dysrhythmia mortality RRA mortality 0.96 (0.51–1.79) RRPT 0.94 2.08 % (0.50–1.76) [RRA 1.00 (0.99–1.06) [0.21 %] RRPT 0.99 (0.93–1.06)] Sorensen Cohort (cross- Dispersion model: Sound Dispersion model: Median baseline Difference in BP (mmHg) Regression coefficient for a Systolic BP et al. sectional to plan with the Nordic Danish AirGis Lden 5–95 doubling in NOX level 35.90 % (2012b) assess the prediction method, 2 percentile) 56.3 (NOX)[R = 0.75] Systolic BP BP results) DANSIM and INM 3 (48.4–70.0) b : -0.39 (-0.64; -0.13) dB(A) A N: 44,436 (Lden ) bP: -0.53 (-0.88; -0.19) 55–64 years [Measured vs. Estimated: Median baseline NO (5–95 Copenhagen or on average 0.2 dB] X percentile) Aarthus, Denmark 20.2 (14.3–86.8) Te L.-F. lg/m3 2000–2002 ´ ral tal. et treault adoaclrhat,tafi-eae i olto n os:aeascain uulycnone 657 confounded mutually associations are noise: and pollution air traffic-related health, Cardiovascular Table 2 continued Studies Characteristics Traffic noise exposure Air pollution Exposure levels Health outcome (definition) Main findings : air pollutions Percentage of of the studya (noise indicator) exposure (air effects adjusted for noise effects change in the [validation of the model] pollution indicator) estimate [validation of the model]

Clark et al. Cross-sectional Dispersion model: Dispersion model: Average Lday : Blood pressure measured using Regression coefficient for an Systolic BP 3 (2012) N: 719 simplified form of the King’s College 50 dB(A) automatic blood pressure meters increase of 1 lg/m NO2 increase 20.69 % UK standard London Emissions (OMORON 711) 9–10 years NO2 average Systolic BP Diastolic BP calculation of road Toolkit level b : 0.058 (-0.092 to 0.210) 166.67 % UK traffic noise 3 A (NO2) 42.73 lg/m 2001–2003 bP: 0.070 (-0.120 to 0.259) (LDay ) [Measured vs. Diastolic BP [No information on Estimated: on validation] average 2.4 ppb] bA: 0.033 (-0.084 to 0.151) bP: 0.088 (-0.059 to 0.236)

RR Risk ratio, SD standard deviation, b regression coefﬁcient, NO2 Nitrogen dioxide, NOX Nitrogen oxide, dB(A) A-weighted decibels, Lday Integrated A-weighted sound level over 16 h (0600–2200), Lden Integrated A-weighted sound level over 24 h (day, evening and night) in which sound levels during the evening (1900–2300 hours) are increased by 5 dB(A) and those during the night (2300–0700 hours) by 10 dB(A), A adjusted for some of the following potential confounding factors: age, education, employment, marital status, study area, mean pulse, hearing impairment, noise at work, crowding, home ownership, mother’s educational level, language spoken at home, parental support for schoolwork, classroom window glazing, body mass index, cuff size, room temperature, birth weight, parental high blood pressure, prematurity, smoking status, family history of CVD, physical inactivity smoking intensity, intake of fruits, intake of vegetables, intake of coffee, alcohol intake diabetes, antihypertensive medication, high blood pressure, long-standing illness and other comorbidity, P adjusted for potential confounding factors including air pollution levels a Final sample sizes used for analysis Results for an increase of 1 lg/m3 123 658 L.-F. Te´treault et al.

a a a (2009) also reported a reduction of the RR between

ischemic heart disease mortality and annual Lden level after adjusting for black smoke and trafﬁc intensity (from 1.15 (95 % CI 0.86, 1.53) to 1.01 (95 % CI 0.74, 1.36) with the Case cohort Cross-sectional Cohort

3], Netherlands reference category \50 dB[A]). However, this study was the only one that presented a variation of [10 % in the point estimate (12.17 %) following adjustment for air pollution. For cerebrovascular diseases, Sorensen et al. (2011) published a positive association between hospitalization for

stroke and Lden before (1.18 (95 % CI 1.11, 1.26) per 10 dB[A]) and after (1.14 (95 % CI 1.03, 1.25) per

10 dB[A]) controlling for NOx. The crude association between Lden level and cerebrovascular mortalities found in Beelen et al. (2009) was 0.88 (95 % CI 0.52, 1.50) and moved toward unity 0.95 (95 % CI 0.55, 1.66) after controlling for black smoke. Both studies reported a percentage change in their estimate of\10 % (respectively 3.64 and 7.95 %). Two studies assessed the effect of noise on blood Exposure assessed with the residential address only databases Exposure assessed with the residential address only home address Cause ofadministrative death databases based Exposure on assessed non-validatedresidential with medico address the only pressure. In the ﬁrst (de Kluizenaar et al. 2007), reported

associations between Lden and self-reported antihypertensive medication intake or hypertension were, respectively, 1.01 (95 % CI 0.96, 1.06) and 1.07 (0.98; 1.18) before

controlling for PM10. Controlling for air pollutants for both outcomes resulted in a small change in the odds ratios, respectively, of 1.03 (95 % CI 0.96, 1.11) and 1.08 (95 %

CI 0.95, 1.23). So adjusting for PM10 produced a CIE of 0.93 % for hypertension and 1.96 % for self-reported antihypertensive medication intake. Once stratified by age in both samples (the Groningen sample and the prevent cohort subsample), the only subgroup presenting significant associations was composed of individuals between 45 and 55 years old. In this age group, the percentage CIE was near our cutoff point for confounding effects (9.45 % in the prevent cohort and 10.19 % in the Groningen sample). The second study assessing blood pressure (Dratva et al. 2012) presented no significant association between road traffic

noise (Lnight and Lday) and either systolic or diastolic blood pressure before and after adjustment for NO2. The regression coefficient did, however, vary extensively before and after adjustment for air pollutants resulting in CIE ranging from 60 to 1,600 %. The effects of noise on overall cardiovascular diseases, heart failure and cardiac dysrhythmia were reported only in Beelen et al. (2009). All associations Major Minorin the Major final Minor model were reduced or were identical after ) None Response rate between 30 and 60 %an adjustment None The air pollution indicator was not specific to road traffic for black smoke. The percentage CIE ) None Response rate between 30 and 60 % None Exposure assessed with the residential address only Cross-sectional ) None Response rate between 30 and 60 % None Exposure assessed with the residential address only Cohort ) None Response rate between 30 and 60 % None Exposure assessed with the residential address only Cohort ) None None None None Case control 2007 ) None Approximately, 85 % of the population at baseline had no paid job. None Input data from 2000 for the noise model paired with the 1986

) None No direct information on response ratewas \10 %. None None Cross-sectional ) None Exclusion of 7 of the 9 school because of missing air pollution exposure None Exposure assessed at school only Cross-sectional 2011 2012a 2012b ) None None None Cause of death based on non-validated medico administrative 2009 The correlation between road trafﬁc noise and air pol- 2009 2012 2012

2012 lution reported in the studies on noise effects described Quality assessment of the studies reviewed (UK [2001–2003], Sweden [1992–1994], Denmark [1993–2006], Canada [1994–2002], Switzerland [2002–200 above does not seem to influence the CIE produced by adjusting for air pollution levels. Studies that presented Design used for the CV outcome of interest de Kluizenaar et al. ( Dratva et al. ( Sorensen et al. ( Sorensen et al. ( Sorensen et al. ( Selander et al. ( Table 3 Author(s)Beelen et al. ( Selection biasesGan et al. ( Clark et al. ( a Classification biases Study design [1997–1998] and, [1987–1996]) weak and high correlations (see ‘‘Appendix’’) were both 123 Cardiovascular health, traffic-related air pollution and noise: are associations mutually confounded 659 subject to large CIE. The quality of the approach used to Interaction estimate the confounder exposure levels (i.e., air pollution) does not appear to impact the CIE either. As presented in Regarding studies that assessed the interaction between air Table 1, the largest CIE were observed neither in studies pollutant levels and noise on cardiovascular outcomes, only with the small R2 nor in those with the large R2. Though two were identified. Selander et al. (2009) did not report a CIE does not seem to be linked to the quality of the study significant interaction between annual NO2 levels and (quantity of biases), cohort studies appear to generally Leq,24h. Gan et al. (2012) did not find a statistically sig- report smaller CIE than studies using case–control or cross- nificant interaction between black carbon and noise levels sectional designs. (Lden) for ischemic heart diseases.

Air pollution controlled for noise Discussion As shown in Table 2, three studies assessed the association between air pollutant levels and cardiovascular diseases This review aimed to assess the confounding effects of one while controlling for noise levels. Beelen et al. (2009) traffic-related exposure (either noise or air pollutants) on reported associations between black smoke and mortality the association between its counterpart and cardiovascular from overall cardiovascular diseases, heart failure ischemic outcomes. In general, the results of the nine studies diseases, cardiac dysrhythmia and cerebrovascular dis- reviewed here showed that when associations between eases. The percentage changes in the estimates for the noise and CV diseases were adjusted for air pollutants, associations reported in Beelen et al. (2009) ranged from modifications of the point estimates for cardiovascular 0.00 to 2.08 %, well below our predefined cutoff point for diseases were \10 %, after controlling for the air pollu- confounding effects. The two remaining studies reported tants, with the exception of the studies by de Kluizenaar associations with blood pressure. Sorensen et al. (2012b) et al. (2007), Beelen et al. (2009) and Dratva et al. (2012) presented a negative association between NOx levels and where the CIE was higher than our cutoff point for con- systolic BP (-0.39 (95 % CI -0.64, -0.13) for doubling founding effects. The Beelen et al. (2009) study reported a the 1 year concentrations). This association was stronger marked decrease of the strength of the association after after adjustment for Lden (-0.53 (95 % CI -0.88, -0.19) adjustment for air pollution and traffic intensity. Yet, the (CI obtained from a personal communication with Mette simultaneous adjustment for traffic intensity and black Sorensen 01-06-2012). Adjusting for noise levels led to a smoke makes the evaluation of confounding by black 35.90 % change in the regression coefficient. In Clark smoke difficult in this study. Nonetheless, no association et al. (2012), the regression coefficient representing the between road traffic and CV outcomes (before and after association between diastolic BP and NO2 varied from adjustment) was found in both the Dratva et al. (2012) and 0.033 (95 % CI -0.084, 0.151) to 0,088 (95 % CI - the Beelen et al. (2009) studies, rendering the CIE mean- 3 0.059, 0.236) per one point increase of NO2 (lg/m ); ingless. By its definition, a confounder must modify the before and after adjusting for noise levels (Lday). The association between the exposure and the outcome. To regression coefficient for the association between systolic confound, such associations must be present at least before

BP and NO2 increased from 0.058 (95 % CI -0.092, or after adjustment. de Kluizenaar et al. (2007) reported 0.210) to 0.070 (95 % CI -0.120, 0.259) per one point CIE of\2 % in both the Groningen sample and the prevent 3 increase of NO2 (lg/m ), before and after adjusting for cohort. However, CIEs nearing 10 % were observed in the noise levels (Lday). The modification of the regression 45–55 years subgroup, which could indicate the presence coefficient for diastolic and systolic BP was, respectively, of a small confounding between the two exposures in this 166.67 and 20.69 %. particular subgroup. Similar findings were found for asso- The correlation between road traffic noise and air pol- ciations between air pollutant levels and CV diseases, lution reported in the studies on traffic-related pollutants although the number of studies was limited (N = 3): con- described above does not seem to influence the CIE pro- trolling for noise levels either changed the point estimates duced by controlling for air pollution levels. The CIE also for CV diseases by \10 % or the study did not present an appears to be independent of the number of biases. On the association between the exposure and the outcome (before other hand, studies with cross-sectional design presented and after adjustment). Only the Sorensen et al. (2012b) higher CIE than the case–control study. Since none of the study presented an indication of confounding by traffic studies had validation information on noise exposure esti- noise in the association between NOx levels and blood mates, we could not assess the impact of the quality of the pressure. Nonetheless, overall these findings suggest an approach used to estimate the confounder exposure levels independent effect of noise and air pollution on CV dis- (i.e., noise) in these studies. eases, particularly ischemic disease for which there were a 123 660 L.-F. Te´treault et al. greater number of studies. The review also points to the gray literature and reviewing some non-English publications, absence of comparability between studies. Most studies but this bias cannot be excluded. We also tried to minimize the were difficult to compare because different noise or air selective reporting bias, which could be a major one in this pollution indicators were used, the pollution levels were review. Indeed, the principal objective of most of the studies assessed using different techniques and very few studies reviewed was not to assess the confounding effect between assessed comparable health outcomes. noise and air pollutants from road traffic. As shown in Fig. 1, In this review, we also tried to verify if the impacts of noise more than half of the studies presenting the basic character- and air pollutants on CV were subject to the same interactive istics to be included did not report the impact of the co- effects. Though both studies (Gan et al. 2012; Selander et al. variables of interest or the regression coefficients. It is also 2009) that assessed interaction effects did not find any effect, possible that some authors assessing the effect of either noise one cannot conclude that there was an absence of interaction or air pollutant levels on CV diseases found that the corre- effect between noise and air pollutant levels with so few sponding co-variable was not significant and did not report it. studies. This is particularly true in the light of the point raised This omission would, however, strengthen our results, as it by Selander et al. (2009) that the interaction analysis might would seem to indicate the independence of the effects of both have lacked power. It should also be noted that both studies exposures. Finally, the strategy used to assess the confounding used different noise indicators, the air pollutant or the car- effect in studies could also be criticized. The 10 % variation in diovascular outcome, to identify possible interaction effects. the estimate is an arbitrary cutoff point that does not neces- Those results do suggest, though, that if a multiplicative sarily rule out confounding or inform on the statistical interaction exists, it is likely to be small. variability. Nonetheless, this cutoff point is widely used in the Due to limitations of the literature, we cannot conclu- literature and was identified as the least biased in a simulation sively ascertain the independence of the effects of the two presented in Maldonado and Greenland (1993), in the absence risks on any CV health outcome. Nonetheless, the results of prior knowledge of confounders. Another potential limi- reported tend to indicate that the impacts of traffic noise tation in our assessment of confounding is that the 10 % cutoff and air pollution on cardiovascular outcome are distinct, or implies that effects are linear. In the case of nonlinear effect at least that they are not completely dependent on one estimates, the absence of CIE does not preclude confounding another. Furthermore, the correlation between noise and air (Janes et al. 2010). This might be the case if noise had a pollutant levels does not seem to influence the confounding threshold effect, and future studies should address this. effects. A wide range of correlations between noise and air To assess clearly the presence of confounding, future pollutants were reported in the studies reviewed and this studies should use coherent noise and air pollution indi- could be partially explained by differences in the urban cators. Those indicators should be chosen according to the structure at each location (building height, distance of effects examined. It would be better to use maximum noise buildings to sidewalks, street width, traffic intensity and levels (Lmax) or equivalent noise over 1 h (Leq,1h) to assess distance to major road). This would suggest that con- acute effects, and Leq,24h to assess chronic effects. The use founding between traffic-related noise and air pollution is a of Lden or LDN as the noise metrics in the studies included study-dependant issue. However, high correlations between in this review could be contested. Those noise metrics noise and pollutant levels were not associated with greater created to assess annoyance increase the weight noise confounding effects in the studies reviewed. Additionally, levels occurring in evening or at night and are therefore not the quality of the exposure assessment of the confounding representative of the actual sound exposure. The source of variables and the quality of the studies (number of biases) the exposure is also an important factor to take into account do not seem to influence the confounding effects. However, while choosing an indicator. For a surrogate of all pollu- the reported CIE seemed higher in studies that used a cross- tants emitted by on-road traffic, one should use individual sectional design. This may be because these studies mainly pollutants such as black carbon, NOX and ultrafine partic- used linear regressions and presented regression coeffi- ulate that are more source specific. More studies on the cients. In fact, the approach that we used to assess relationship between noise outdoor exposure levels and confounding effects was developed for risk ratios and may personal exposure levels should be conducted. This is not be applicable for linear regression. Nonetheless, this important given that this relationship could differ between approach has been suggested by some authors in linear noise and air pollution. Furthermore, future studies should regression text books (Vittinghoff et al. 2012). present adjusted results for both noise and air pollution This review is subjected to a few limitations. First, only modeled separately, so that the impact of the pollutant that a small number of studies were available to be reviewed, was controlled for can be assessed. As shown in Tables 1 which reduced the strength of our findings. Secondly, with and 2, only few studies provided validation information for any systematic review, the possibility of publication bias is the exposure models used in their study. Theabsence of present. We tried to minimize this bias by searching the such information preventsthe reader from judging the 123 Cardiovascular health, traffic-related air pollution and noise: are associations mutually confounded 661 quality ofthe exposure assessment and therebyprecluding Appendix the reader to judge from assessing the quality of the adjustments made. Finally, more studies are needed to find Search strategy out if the confounding effect is specific to subcategories of 1. cardiovascular.mp. (906078) CV outcomes. Ideally, those studies would need an epi- 2. hypertension.mp. (887805) demiologic design enabling them to assess the chronic 3. arterial tension.mp. (560) effects of both traffic-related pollutants. We also recom- 4. blood pressure.mp. (753655) mend that those studies focus their research on ischemic 5. arrythmia.mp. (783) heart diseases, hypertension or the fluctuation in blood 6. myocardial infraction.mp. (811) pressure, for which mechanisms should yield more con- 7. stroke.mp. (382573) clusive results. 8. vasoconstriction.mp. (79992) 9. ischemia.mp. (461795) Conclusion 10. heart.mp. (2590525) 11. coagulation.mp. (227773) Results from this review suggest that confounding of car- 12. arteries.mp. (364558) diovascular effects by noise or air pollutants is low, on 13. blood flow.mp. (467842) average, though heterogeneity across studies and areas 14. electrocardiogram.mp. (112029) within studies has been reported. The quality of exposure 15. cardioa.mp. (1557392) assessment of the confounding variable, the quality of 16. 1 or 2 or 3 or 4 or 5 or 6 or 7 or 8 or 9 or 10 or 11 or studies as well as the correlation between noise and pol- 12 or 13 or 14 or 15 (5132805) lutant levels do not seem to influence the confounding 17. noise.mp. (184815) effect, though with further improvements in exposure 18. sound.mp. (142837) assessment, the situation may change. More studies using 19. 17 or 18 (309938) air pollution indicators specific to road traffic are needed to 20. 16 and 19 (31528) properly assess if road noise and pollutant effects on CV 21. air pollution.mp. (104657) outcomes are subjected to the confounding effect of one 22. particles.mp. (301041) another. 23. particulate.mp. (84570) 24. nitrogen oxide.mp. (9147) Acknowledgments This study was made possible through a finan- 25. ozone.mp. (39435) cial contribution from the Public Health Agency of Canada through the National Collaborating Centre for Environmental Health. We 26. diesel.mp. (12621) would like to thank Hugh Davies, Helen Ward and Allan Brand for 27. motor vehicles.mp. (6278) their helpful comments on the writing of this manuscript. 28. 21 or 22 or 23 or 24 or 25 or 26 or 27 (493806) 29. 20 and 28 (328) Open Access This article is distributed under the terms of the Creative Commons Attribution License which permits any use, dis- 30. remove duplicates from 29 (221) tribution, and reproduction in any medium, provided the original author(s) and the source are credited. See Table 4.

123 662 123 Table 4 Correlations between sound levels and air pollution levels in urban areas (UK [2001–2003], Sweden [1992–1994] and [2004–2005], Denmark [1993–2006], Canada [1994–2002] and [2001], Switzerland [2002–2003], Spain [1995–2000] and [2008], Belgium [2009], Germany [2008], USA [2006–2007], Netherlands [1997–1998], [1987–1996] and [2006])

Studies Characteristics Noise Air pollution Correlation between noise and air pollution levels

Noise exposure Method Air pollution Method NO2 NOx PM10 PM2.5 UFP Black O3 Black (noise indicator) exposure (air smoke carbon pollutants)

Dratva et al. Cross-sectional LDay and Lnight Dispersion model: NO2 PM10 PolluMap Gaussian 0.28–0.29 0.16–0.17 M M (2012) N: 6,450 SONABASE dispersion 2002–2003 Switzerland

Clark et al. Cross-sectional LDay Dispersion model: NO2 King’s college London 0.46 M (2012) N: 719 simpliﬁed form of Emissions Toolkit the UK standard 2001–2003 calculation of UK road trafﬁc noiseM

Gan et al. N: 445,868 Lden Annual average of NO2 NOX Annual average levels 0.33 0.39 – 0.14 – – – 0.44 (2012) 1994–2002 equivalent SPL PM2.5 (land-use regression) for 24 h (CadnaA Vancouver Canada model using the EMME/2 for trafﬁc volume)M

Sorensen et al. N: 51,485 Lden Annual average of NOx Annual average levels – 0.62 – – – – – – (2011) 1995, 2000 and 2005 equivalent SPL (Danish AirGis for 24 h (sound modelling system)M Copenhagen or plan model with Aarthus, the nordic Denmark prediction method, DANSIM and INM)M

Foraster et al. 77 sites LA,eq,24h Average of NO2 Annual average 0.62 – – – – – – – (2011) 2008 equivalent SPL estimated using NO2 for 24 h (Girona monthly Girona, Spain trafﬁc noise measurementsM model)M

Gan et al. 1 site LA,eq,15min 15 min average of NO2 NOX 15 min average 0.29 0.46 – – 0.38 – – – (2010) 2009 equivalent SPL UFP concentration (integrated multi gas Wolfstraat, Belgium measurement platform [Airpointer, Recordum Austria])

Boogaard 264 sites (132 LA,eq,1min 1 min average SPL PM2.5 One minute average – – – 0.009 - - – - et al. routes) (measurement concentration .F Te L.-F. (2009) 2006 from 10:00 to (condensation particle 16:00) counter) Netherlands ´ ral tal. et treault adoaclrhat,taﬁ-eae i olto n os:aeascain uulycnone 663 confounded mutually associations are noise: and pollution air trafﬁc-related health, Cardiovascular Table 4 continued

Studies Characteristics Noise Air pollution Correlation between noise and air pollution levels

Noise exposure Method Air pollution Method NO2 NOx PM10 PM2.5 UFP Black O3 Black (noise indicator) exposure (air smoke carbon pollutants)

Weber (2009) 50 sites Leq,20s Equivalent SPL for PM2.5 Average of air pollution – – – 0.41–0.81 – – – – 2008 20 s level at the beginning and the end of the Essen, Germany noise monitoring period values (condensational particulate counter 1s)

Selander et al. N: 3,666 LA,eq,24h Annual average of NO2 Annual average levels 0.6 – – – – – – – (2009) 1970 to 1992–1994 equivalent SPL (Stockholm county for 24 h dispersion model)M Stockholm county, (simpliﬁed Sweden version of the nordic prediction method)M

Beelen et al. N: 117,528 Lden Annual average of black smoke Annual levels (land use – – – – – 0.24 – – (2009) 1986 home address equivalent SPL and regression model with 2000–2001 for 24 h (empara (sum of regional, noise model)M urban and local trafﬁc noise imput data M and 1987–1996 contributions)) air pollutants input data Netherlands a a a Allen et al. 105 sites LA,eq,5min Equivalent SPL for NO2,NOx 14 days average (Ogawa 0.16–0.62 0.49–0.62 – – 0.22–0.41 ––– (2009) 2006–2007 5 min and UFP sampler) (measurement Chicago and from 10:00 to Riverside county 16:00) USA

Davies et al. 103 sites LA,eq,5min Equivalent SPL for NO2 and NOx 14 days average (passive 0.53 0.64 – – – – – – (2009) 2001 5 min sampler) (measurement at Vancouver, Canada 103 sites from 8:00 to 18:00)

de Kluizenaar N: 40,856 Lden Annual average of PM10 Annual average levels – – 0.72 – – – – – et al. 1997–1998 equivalent SPL (Netherlands’ (2007) for 24 h (Urbis standard Dutch Sweden Groningen, model [standard models for local air Netherlands Karterings pollution calculations: Method 2])M the street model CAR I and dispersion model ‘‘Pluim’’)M 123 664 123 Table 4 continued

Studies Characteristics Noise Air pollution Correlation between noise and air pollution levels

Noise exposure Method Air pollution Method NO2 NOx PM10 PM2.5 UFP Black O3 Black (noise indicator) exposure (air smoke carbon pollutants)

Persson et al. N: 2,856 LA,eq,24h Annual average of NOX Modiﬁed Gaussian 0.50 (2007) 2004–2005 equivalent SPL dispersion model for 24 h (ENVIMAN)M Scania, Sweden (simpliﬁed version of the nordic prediction method)M a a ALNAP- N:NA Lden Annual average of NO2, and Annual average (network 0.12–0.48 – 0.09–0.39 –– ––– 1 (2006) NA equivalent SPL PM10 emission model for 24 h [NEMO])M Tyrol, Austria (harmonoise model)M

Linares et al. 6 sites LA,eq,24h Equivalent SPL for NO2,SO2,O3, Daily average (from the 0.138 0.206 0.089 – – – -0.275 – (2006) 1995–2000 24 h PM10 and Madrid city air NOx pollution network) Madrid, Spain

Ising et al. 25 sites Lnight Equivalent SPL from NO2 Exposure from 58–93 h 0.836 – – – – – – – (2004) NA 22:00 to 6:00 (palmes tube) Germany

Tobias et al 6 sites Leq,24h Equivalent SPL for NO2,SO2,O3 Daily average (from the 0.32 0.35 – – – – -0.42 – (2001) 1995–1997 24 h and NOx Madrid city air pollution network) Madrid Spain except for O3 (1 h maximum value)

Klæboe et al. N: 1,028, 1,140, LA,eq,24h Annual average of NO2,PM10, 3 month average of 0.48 – 0.34 0.39 – – – – (2000) 1,097 equivalent SPL PM2.5 hourly estimations 1987, 1994 and 1996 for 24 h (nordic (dispersion model prediction EPISODE)M Oslo, Norway method)M b BEG-(1998) N:NA Ldn Annual average of NO2 and PM10 NA (Gaussian 0.63 – 0.61 – – – – – M NA equivalent SPL propagation model) for 24 h (sound Tyrol, Austria plan software)M a Lowest and highest correlation of all the sampling sessions M Exposure asses using models b The description of those studies were available only in Lercher et al. 2011 (those studies are either not publish in English or French or are not publish at all) N ﬁnal sample size used for analysis .F Te L.-F. PM2.5 particles with a diameter of 2.5 lm or smaller, PM10 particles with a diameter of larger than 2.5 lm but smaller than 10 lm, NO2 nitrogen dioxide, NOX nitrogen oxide, O3 Ozone, dB(A) a-weighted decibels, LA,eq,24h integrated A-weighted sound level over 24 h, Lday integrated sound level over 16 h (0600–2200), Lnight integrated A-weighted sound level over 8 h ´

(2200–0600), LA,eq,20s integrated A-weighted sound level over 20 s, LA,eq,1min integrated A-weighted sound level over 1 min, LA,eq,5min integrated A-weighted sound level over 5 min, LA,eq,15min al. et treault integrated A-weighted sound level over 15 min, Ldn integrated A-weighted sound level over 24 h (day, night) in which sound levels during the night (22h00–07h00) are increased by 10 dB(A), Lden integrated A-weighted sound level over 24 h (day, evening and night) in which sound levels during the evening (1900–2300 hours) are increased by 5 dB(A) and those during the night (2300–0700 hours) by 10 dB(A), Last validation 12th April 2012 Cardiovascular health, trafﬁc-related air pollution and noise: are associations mutually confounded 665

References Clark C, Crombie R, Head J, van Kamp I, van Kempen E, Stansfeld SA (2012) Does traffic-related air pollution explain associations Adar SD, Sheppard L, Vedal S, Polak JF, Sampson PD, Diez Roux of aircraft and road traffic noise exposure on children’s health AV, Budoff M, Jacobs DR Jr, Barr RG, Watson K, Kaufman JD and cognition? A secondary analysis of the United Kingdom (2013) Fine particulate air pollution and the progression of sample from the RANCH project. Am J Epidemiol carotid intima-medial thickness: a prospective cohort study from 176(4):327–337. doi:10.1093/aje/kws012 the multi-ethnic study of atherosclerosis and air pollution. PLoS Davies HW, Vlaanderen JJ, Henderson SB, Brauer M (2009) Med 10(4):e1001430 Correlation between co-exposures to noise and air pollution Allen RW, Adar SD (2011) Are both air pollution and noise driving from traffic sources. Occup Environ Med 66(5):347–350 adverse cardiovascular health effects from motor vehicles? de Kluizenaar Y, Gansevoort RT, Miedema HM, de Jong PE (2007) Environ Res 111(1):184–185 Hypertension and road traffic noise exposure. J Occup Environ Allen RW, Davies H, Cohen MA, Mallach G, Kaufman JD, Adar SD Med 49(5):484–492. doi:10.1097/JOM.0b013e318058a9ff (2009) The spatial relationship between traffic-generated air Dratva J, Phuleria HC, Foraster M, Gaspoz JM, Keidel D, Kunzli N, pollution and noise in 2 US cities. Environ Res 109(3):334–342 Liu LJ, Pons M, Zemp E, Gerbase MW, Schindler C (2012) Amato MC, Giordano C, Galia M, Criscimanna A, Vitabile S, Midiri Transportation noise and blood pressure in a population-based M, Galluzzo A (2010) Visceral Adiposity Index: a reliable sample of adults. Environ Health Perspect 120(1):50–55. doi:10. indicator of visceral fat function associated with cardiometabolic 1289/ehp.1103448 risk. Diabetes Care 33(4):920–922. doi:10.2337/dc09-1825 European Environment Agency (EEA) (2010) Good practice guide on Babisch W (2006) Transportation noise and cardiovascular risk: noise exposure and potential health effects. European Environ- updated review and synthesis of epidemiological studies indicate ment Agency (EEA), Copenhagen that the evidence has increased. Noise Health 8(30):1–29 Finkelstein MM, Jerrett M, Sears MR (2004) Traffic air pollution and Babisch W, Kamp I (2009) Exposure–response relationship of the mortality rate advancement periods. Am J Epidemiol association between aircraft noise and the risk of hypertension. 160(2):173–177. doi:10.1093/aje/kwh181 Noise Health 11(44):161–168. doi:10.4103/1463-1741.53363 Foraster M, Deltell A, Basaganã X, Medina-Ramoń M, Aguilera I, Beelen R, Hoek G, van den Brandt PA, Goldbohm RA, Fischer P, Bouso L, Grau M, Phuleria HC, Rivera M, Slama R, Sunyer J, Schouten LJ, Armstrong B, Brunekreef B (2008) Long-term Targa J, Ku¨nzli N (2011) Local determinants of road traffic noise exposure to traffic-related air pollution and lung cancer risk. levels versus determinants of air pollution levels in a Mediter- Epidemiology 19(5):702–710. doi:10.1097/EDE.0b013e318181 ranean city. Environ Res 111(1):177–183. ISSN 0013-9351. b3ca doi:10.1016/j.envres.2010.10.013. http://www.sciencedirect.com/ Beelen R, Hoek G, Houthuijs D, Van Den Brandt PA, Goldbohm RA, science/article/pii/S0013935110001878 Fischer P, Schouten LJ, Armstrong B, Brunekreef B (2009) The Gan WQ, Tamburic L, Davies HW, Demers PA, Koehoorn M, Brauer joint association of air pollution and noise from road traffic with M (2010) Changes in residential proximity to road traffic and the cardiovascular mortality in a cohort study. Occup Environ Med risk of death from coronary heart disease. Epidemiology 66(4):243–250 21(5):642–649 Bliss R, Weinberg J, Webster T, Vieira V (2012) Determining the Gan WQ, Davies HW, Koehoorn M, Brauer M (2012) Association of probability distribution and evaluating sensitivity and false long-term exposure to community noise and traffic-related air positive rate of a confounder detection method applied to logistic pollution with coronary heart disease mortality. Am J Epidemiol regression. J Biom Biostat 3(4):142. doi:10.4172/2155-6180. 175(9):898–906. doi:10.1093/aje/kwr424 1000142 Haralabidis AS, Dimakopoulou K, Velonaki V, Barbaglia G, Mussin Boogaard H, Borgman F, Kamminga J, Hoek G (2009) Exposure to M, Giampaolo M, Selander J, Pershagen G, Dudley M-L, Babisch ultrafine and fine particles and noise during cycling and driving W, Swart W, Katsouyanni K, Jarup L, HYENA Consortium in 11 Dutch cities. Atmos Environ 43(27):4234–4242 (2011) Can exposure to noise affect the 24 h blood pressure Brook RD, Franklin B, Cascio W, Hong Y, Howard G, Lipsett M, profile? Results from the HYENA study. J Epidemiol Community Luepker R, Mittleman M, Samet J, Smith SC Jr, Tager I (2004) Health 65(6):535–541. doi:10.1136/jech.2009.102954 Air pollution and cardiovascular disease: a statement for HEI Panel on the Health Effects of Traffic-Related Air Pollution healthcare professionals from the expert panel on population (2010) Traffic-related air pollution: a critical review of the and prevention science of the American Heart Association. literature on emissions, exposure, and health effects vol special Circulation 109(21):2655–2671. doi:10.1161/01.CIR.0000128587. report 17. Health Effects Institute, Boston, MA 30041.C8 Hoek G, Brunekreef B, Goldbohm S, Fischer P, van den Brandt PA Brook RD, Rajagopalan S, Pope CA, Brook JR, Bhatnagar A, Diez- (2002) Association between mortality and indicators of traffic- Roux AV, Holguin F, Hong Y, Luepker RV, Mittleman MA, related air pollution in the Netherlands: a cohort study. Lancet Peters A, Siscovick D, Smith SC, Whitsel L, Kaufman JD, 360(9341):1203–1209 American Heart Association Council on Epidemiology and Hoffmann B, Moebus S, Stang A, Beck EM, Dragano N, Mo¨hlenk- Prevention, Council on the Kidney in Cardiovascular Disease, amp S, Schmermund A, Memmesheimer M, Mann K, Erbel R, and Council on Nutrition, Physical Activity and Metabolism Jo¨ckel KH, Heinz Nixdorf RECALL Study Investigative Group (2010) Particulate matter air pollution and cardiovascular (2006) Residence close to high traffic and prevalence of coronary disease. Circulation 121(21):2331–2378. doi:10.1161/CIR.0b013 heart disease. Eur Heart J 27(22):2696–2702. doi:10.1093/ e3181dbece1 eurheartj/ehl278 Burgan O, Smargiassi A, Perron S, Kosatsky T (2010) Cardiovascular Hoffmann B, Moebus S, Dragano N, Mo¨hlenkamp S, Memmesheimer effects of sub-daily levels of ambient fine particles: a systematic M, Erbel R, Jo¨ckel KH, Heinz Nixdorf Recall Investigative review. Environ Health 9:26. doi:10.1186/1476-069X-9-26 Group (2009) Residential traffic exposure and coronary heart Chen H, Goldberg MS, Burnett RT, Jerrett M, Wheeler AJ, disease: results from the Heinz Nixdorf Recall Study. Biomark- Villeneuve PJ (2013) Long-term exposure to traffic-related air ers 14(s1):74–78. doi:10.1080/13547500902965096 pollution and cardiovascular mortality. Epidemiology 24(1):35– Ising H, Lange-Asschenfeldt H, Moriske HJ, Born J, Eilts M (2004) 43. doi:10.1097/EDE.0b013e318276c005 Low frequency noise and stress: bronchitis and cortisol in

123 666 L.-F. Te´treault et al.

children exposed chronically to traffic noise and exhaust fumes. Selander J, Nilsson ME, Bluhm G, Rosenlund M, Lindqvist M, Nise Noise Health 6(23):21–28 G, Pershagen G (2009) Long-term exposure to road traffic noise Janes H, Dominici F, Zeger S (2010) On quantifying the magnitude and myocardial infarction. Epidemiology 20(2):272–279 of confounding. Biostatistics 11(3):572–582. doi:10.1093/ Sorensen M, Hvidberg M, Andersen ZJ, Nordsborg RB, Lillelund biostatistics/kxq007 KG, Jakobsen J, Tjonneland A, Overvad K, Raaschou-Nielsen O Klæboe R, Kolbenstvedt M, Clench-Aas J, Bartonova A (2000) Oslo (2011) Road traffic noise and stroke: a prospective cohort study. traffic study: part 1: an integrated approach to assess the Eur Heart J. 32(6):737–744 combined effects of noise and air pollution on annoyance. Atmos Sorensen M, Andersen ZJ, Nordsborg RB, Jensen SS, Lillelund KG, Environ 34(27):4727–4736 Beelen R, Schmidt EB, Tjonneland A, Overvad K, Raaschou- Lercher P, Botteldooren D, Widmann U, Uhrner U, Kammeringer E Nielsen O (2012a) Road traffic noise and incident myocardial (2011) Cardiovascular effects of environmental noise: research infarction: a prospective cohort study. PLoS One 7(6):e39283. in Austria. Noise Health 13(52):234–250 doi:10.1371/journal.pone.0039283 Liao D, Shaffer ML, He F, Rodriguez-Colon S, Wu R, Whitsel EA, Sorensen M, Hoffmann B, Hvidberg M, Ketzel M, Jensen SS, Bixler EO, Cascio WE (2010) Fine particulate air pollution is Andersen ZJ, Tjonneland A, Overvad K, Raaschou-Nielsen O associated with higher vulnerability to atrial fibrillation—the (2012b) Long-term exposure to traffic-related air pollution APACR study. J Toxicol Environ Health, Part A 74(11):693– associated with blood pressure and self-reported hypertension 705. doi:10.1080/15287394.2011.556056 in a Danish cohort. Environ Health Perspect 120(3):418–424. Linares C, Diaz J, Tobias A, De Miguel JM, Otero A (2006) Impact of doi:10.1289/ehp.1103631 urban air pollutants and noise levels over daily hospital Taheri S, Lin L, Austin D, Young T, Mignot E (2004) Short sleep admissions in children in Madrid: a time series analysis. Int duration is associated with reduced leptin, elevated ghrelin, and Arch Occup Environ Health 79(2):143–152. doi:10.1007/ increased body mass index. PLoS Med 1(3):e62 s00420-005-0032-0 Tobias A, Diaz J, Saez M, Carlos Alberdi J (2001) Use of poisson Maheswaran R, Elliott P (2003) Stroke mortality associated with regression and Box-Jenkins models to evaluate the short-term living near main roads in England and Wales. Stroke effects of environmental noise levels on daily emergency 34(12):2776–2780. doi:10.1161/01.str.0000101750.77547.11 admissions in Madrid, Spain. Eur J Epidemiol 17(8):765–771 Maldonado G, Greenland S (1993) Simulation study of confounder- van Kempen E, Babisch W (2012) The quantitative relationship selection strategies. Am J Epidemiol 138(11):923–936 between road traffic noise and hypertension: a meta-analysis. Mills NL, Tornqvist H, Robinson SD, Gonzalez MC, Soderberg S, J Hypertens 30(6):1075–1086. doi:10.1097/HJH.0b013e328352 Sandstrom T, Blomberg A, Newby DE, Donaldson K (2007) Air ac54 pollution and atherothrombosis. Inhal Toxicol 19(Suppl Vardeny O, Peppard PE, Finn LA, Faraco JH, Mignot E, Hla KM 1):81–89. doi:10.1080/08958370701495170 (2011) [beta]2 adrenergic receptor polymorphisms and nocturnal Nuvolone D, Balzi D, Chini M, Scala D, Giovannini F, Barchielli A blood pressure dipping status in the Wisconsin Sleep Cohort (2011) Short-term association between ambient air pollution and Study. J Am Soc Hypertens 5(2):114–122 risk of hospitalization for acute myocardial infarction: results of Vittinghoff E, Glidden DV, Shiboski SC, McCulloch CE (2012) the cardiovascular risk and air pollution in tuscany (RISCAT) Regression methods in biostatistics. Linear, logistic, survival, study. Am J Epidemiol 174(1):63–71. doi:10.1093/aje/kwr046 and repeated measures models, 2nd edn. Springer Science?Busi- Perron S, Tetreault LF, King N, Plante C, Smargiassi A (2012) Review ness Media, Boston of the effect of aircraft noise on sleep disturbance in adults. Noise Weber S (2009) Spatio-temporal covariation of urban particle number Health 14(57):58–67. doi:10.4103/1463-1741.95133 concentration and ambient noise. Atmos Environ 43(34):5518– Persson R, Bjork J, Ardo J, Albin M, Jakobsson K (2007) Trait 5525 anxiety and modeled exposure as determinants of self-reported Westman JC, Walters JR (1981) Noise and stress: a comprehensive annoyance to sound, air pollution and other environmental approach. Environ Health Perspect 41:291–309 factors in the home. Int Arch Occup Environ Health World Health Organisation (2011) Burden of disease from environ- 81(2):179–191. doi:10.1007/s00420-007-0204-1 mental noise. WHO Regional Office for Europe, Copenhagen Pirrera S, De Valck E, Cluydts R (2010) Nocturnal road traffic noise: World Health Organisation Regional Office for Europe (2011) Burden a review on its assessment and consequences on sleep and health. of disease from environmental noise. World Health Organisa- Environ Int 36(5):492–498 tion, Copenhagen Rosenlund M, Berglind N, Pershagen G, Hallqvist J, Jonson T, Bellander T (2006) Long-term exposure to urban air pollution and myocardial infarction. Epidemiology 17(4):383–390

123 Article Annoyance from Road Trafﬁc, Trains, Airplanes and from Total Environmental Noise Levels

Martina S. Ragettli 1,2,3,4, Sophie Goudreau 2, Céline Plante 2, Stéphane Perron 2, Michel Fournier 2 and Audrey Smargiassi 1,5,6,*

Received: 12 July 2015; Accepted: 21 December 2015; Published: 29 December 2015 Academic Editors: Ronny Klaeboe, Peter Lercher and Mariola Sliwinska-Kowalska

1 Department of Environmental Health and Occupational Health, School of Public Health, University of Montreal, Montreal, QC H3C 3J7, Canada; [email protected] 2 Public Health Department of Montreal, Montreal, QC H2L 1M3, Canada; [email protected] (S.G.); [email protected] (C.P.); [email protected] (S.P.); [email protected] (M.F.) 3 Swiss Tropical and Public Health Institute, Basel 4002, Switzerland 4 University of Basel, Basel 4003, Switzerland 5 National Institute of Public Health of Quebec, Montreal, QC H3C 2B9, Canada 6 Public Health Research Institute, University of Montreal, QC H3C 3J7, Canada * Correspondence: [email protected]; Tel.: +1-514-343-6111 (ext. 38528)

Abstract: There is a lack of studies assessing the exposure-response relationship between transportation noise and annoyance in North America. Our aims were to investigate the prevalence of noise annoyance induced by road traffic, trains and airplanes in relation to distance to transportation noise sources, and to total environmental noise levels in Montreal, Canada; annoyance was assessed as noise-induced disturbance. A telephone-based survey among 4336 persons aged >18 years was conducted. Exposure to total environmental noise (A-weighted outdoor noise levels—LAeq24h and day-evening-night equivalent noise levels—Lden) for each study participant was determined using a statistical noise model (land use regression—LUR) that is based on actual outdoor noise measurements. The proportion of the population annoyed by road traffic, airplane and train noise was 20.1%, 13.0% and 6.1%, respectively. As the distance to major roads, railways and the Montreal International Airport increased, the percentage of people disturbed and highly disturbed due to the corresponding traffic noise significantly decreased. When applying the statistical noise model we found a relationship between noise levels and disturbance from road traffic and total environmental noise, with Prevalence Proportion Ratios (PPR) for highly disturbed people of 1.10 (95% CI: 1.07–1.13) and 1.04 (1.02–1.06) per 1 dB(A) Lden, respectively. Our study provides the first comprehensive information on the relationship between transportation noise levels and disturbance in a Canadian city. LUR models are still in development and further studies on transportation noise induced annoyance are consequently needed, especially for sources other than road traffic.

Keywords: environmental noise; exposure; annoyance; transportation; Canada

1. Introduction Annoyance is the most prevalent health effect in a population exposed to environmental noise. It can result from noise disrupting people during daily activities, sleep or rest and may cause a variety of negative responses such as anger, distraction, depression, anxiety, exhaustion and stress-related symptoms [1,2]. Noise annoyance has further been shown to be associated with reduced quality of life [3,4] and well-being [2]. Trafﬁc noise is one of the main sources of annoyance. Several studies have shown positive exposure-response relationships between annoyance and increasing environmental noise levels induced by road trafﬁc, trains and airplane movements [4–7].

Int. J. Environ. Res. Public Health 2016, 13, 90; doi:10.3390/ijerph13010090 www.mdpi.com/journal/ijerph Int. J. Environ. Res. Public Health 2016, 13, 90 2 of 13

The exposure to environmental noise from transportation sources and its related health burden is of increasing concern. In urban areas, road traffic is the most widespread noise source. According to the World Health Organization (WHO) Europe, exposures to A-weighted day-evening-night equivalent sound pressure level (Lden) of road traffic noise that exceed 55 dB(A) pose a serious risk to health [1,8]. In 2011, it was estimated that about 50% of the population in large agglomerations (>250,000 inhabitants) of the European Union (EU) was exposed to road traffic noise levels (Lden) above 55 dB(A) [8]. While sensitivity and exposure to noise may differ between regions of the globe (e.g., due to differences in house construction, window opening habits), the relation between environmental noise levels and annoyance has mostly been studied in Europe [2,9]. For example, the EU has implemented the Environmental Noise Directive 2002/49/CE [10] to assess and reduce population-wide exposure to noise from transportation sources. Similar national initiatives in other countries in the world are rare. In Canada and the United Sates (US) for example, no standardized noise estimation approaches exists [11,12]. Furthermore, thus far, the relation between environmental noise levels and annoyance has only been established with modeled environmental noise levels that are not based on actual noise measurements but simulated with noise propagation models [4,5,8]. Such numerical models predict noise for a specific source by means of physical rules of noise propagation and attenuation [13]. Recent noise monitoring campaigns in the US [11], and in the three Canadian cities Halifax [14], Toronto [15] and Montreal [16], have shown that measured noise levels tend to exceed the 55 dB(A) threshold that has been associated with negative health impacts mainly in European studies using noise estimates from propagation noise models [1,8]. Furthermore, according to a national survey performed in Canada in 2005, 6.7% of Canadians reported being highly annoyed by road traffic noise, with generally higher percentages found in larger cities than in smaller communities [17]. In view of the ever-growing urban populations and public health concerns about excessive transportation noise in North American cities, studies investigating exposure-response relationships and potential noise-related health effects are warranted. The current study presents data of a telephone-based survey on noise annoyance carried out in 2014 on the Montreal Island (Canada). Our aims were to investigate the prevalence of annoyance induced by road traffic, trains and airplanes in relation to (a) distance to transportation noise sources; and (b) to total environmental noise levels estimated with a land use regression model that is based on an extensive noise measurement campaign carried out in Montreal [18].

2. Methods

2.1. Study Population There were about 1.9 million people living on the Island of Montreal in 2011 [19], on an area of approximately 500 km2 (Figure1). Montreal Island includes 14 municipalities among which belongs the city of Montreal—that includes most of the population of the Island—and which is the second largest city in Canada. A significant fraction of the population lives close to major roads such as arteries and national highways (e.g., Trans-Canada Highway). Some residences are also located in proximity to railway tracks and/or in the vicinity and in the flight paths of the Montreal International Airport. Hereafter, the zone around the airport is referred to as the Noise Exposure Forecast zone 25 (NEF25). Defined by the Airport Authority and used for urban planning purposes, it describes an area where annoyance is likely to occur in the surrounding of the airport [20]. Int. J. Environ. Res. Public Health 2016, 13 3

than in smaller communities [17]. In view of the ever-growing urban populations and public health concerns about excessive transportation noise in North American cities, studies investigating exposure-response relationships and potential noise-related health effects are warranted. The current study presents data of a telephone-based survey on noise annoyance carried out in 2014 on the Montreal Island (Canada). Our aims were to investigate the prevalence of annoyance induced by road traffic, trains and airplanes in relation to (a) distance to transportation noise sources; and (b) to total environmental noise levels estimated with a land use regression model that is based on an extensive noise measurement campaign carried out in Montreal [18].

2. Methods

2.1. Study Population

There were about 1.9 million people living on the Island of Montreal in 2011 [19], on an area of approximately 500 km2 (Figure 1). Montreal Island includes 14 municipalities among which belongs the city of Montreal—that includes most of the population of the Island—and which is the second largest city in Canada. A significant fraction of the population lives close to major roads such as arteries and national highways (e.g., Trans-Canada Highway). Some residences are also located in proximity to railway tracks and/or in the vicinity and in the flight paths of the Montreal International Airport. Hereafter, the zone around the airport is referred to as the Noise Exposure Forecast zone 25 (NEF25). Int.Defined J. Environ. by Res. the Public Airport Health Aut2016hority, 13, 90 and used for urban planning purposes, it describes an area where3 of 13 annoyance is likely to occur in the surrounding of the airport [20].

FigureFigure 1. 1. Map of the studystudy area withwith potentialpotential transportationtransportation noisenoise sources.sources.

To ensure a sufficient number of subjects exposed to each transportation noise source, survey respondents >18 years of age were randomly selected from pre-defined survey strata (road, rail, NEF25, not-exposed ). To build these strata, the population on the island was first divided in seven transportation noise exposure categories by means of ArcGIS 9.2 (ESRI, Redlands, CA, USA) using a digital road network (Addresses Québec 2013), a railway network (CanMapr Rail V2010.1), and the geo-referenced NEF25 map of the Montreal International Airport [16,21]. The exposure categories were defined based on the results of previous noise exposure studies in Montreal reporting noise levels in the vicinity of transportation noise sources [16,22,23] and to ensure sufficient numbers of respondents in each strata as follows: people living within 100 m of an artery or highway (road), within 150 m of a main railway line or main line of a railroad shunting yard (rail), within 1 km of the NEF25 zone of the Montreal International Airport (NEF25), as well as the combinations of the road, rail and NEF25 categories (i.e., road and rail, road and NEF25, rail and NEF25, road and rail and NEF25). People living within those categories are considered to be most exposed to the respective noise sources. The categories were then combined to four survey strata based on their population share: (1) road only; (2) rail (rail and road and rail); (3) NEF25 (NEF25, road and NEF25, rail and NEF25, road and rail and NEF25); and (4) non-exposed (an area not directly exposed to transportation noise sources). In each stratum, six digit residential postal codes were identified and linked to a local telephone number registry and then selected randomly for interviews. Each six-character postal code usually represents a street block or a large apartment complex in the core of the city [24]. We aimed to obtain in total, approximately 4500 respondents (i.e., 1125 respondents for each survey strata).

2.2. Survey and Noise Annoyance Data The purpose of the survey was to evaluate perceived annoyance and sleep disturbance from various noise sources in the population living on the island of Montreal. The questionnaire was adapted from the European LARES-survey (Large Analysis and Review of European housing and health Status) [25]. Questions on socio-demographic factors (age, gender, household income, education, number of people per household, time lived at the place of residence) and self-estimated noise-sensitivity were also included. Postal codes that allow to geographically locating the respondents were veriﬁed during the interview. The survey was carried out as telephone interviews between the 10 April and the 20 June 2014. Int. J. Environ. Res. Public Health 2016, 13, 90 4 of 13

This article addresses only the results regarding annoyance due to transportation noise sources and the total (outdoor) environmental noise. Annoyance was assessed as noise-induced disturbance [26] in a five-point category scale for eight outdoor noise sources: (1) traffic; (2) trains; (3) airplanes; (4) parking lots; (5) neighborhood (bars, discos, demonstrations); (6) animals or birds (outdoors); (7) shopping centers, industrial or construction zones; and (8) schools and parks. The question was phrased as follows: “In the past 12 months, have you been not at all, a little, somewhat, quite a bit or a lot disturbed at home by the following sources of noise?” Note that we did not include the words “bothered” or “annoyed” in addition to “disturbed” in our question. In the presentation of the results, the “annoyed” category consists of those that were somewhat, quite a bit or a lot disturbed by a specific noise source, and the “highly annoyed” category included those who reported being quite a bit or a lot disturbed. Non-applicable answers were considered as not annoyed. High annoyance to total environmental noise was defined as being quite a bit or a lot disturbed by at least one of the outdoor noise sources (1 to 8).

2.3. Noise Levels and Distance to Noise Sources

A-weighted outdoor noise levels (LAeq24h) and day-evening-night equivalent noise levels (Lden) for 2014 were estimated at the geographic coordinate of the six digit postal code of each subject. Lden includes a 5 dB(A) penalty for the evening (19:00–22:59) and a 10 dB(A) penalty for the night-time hours (23:00–6:59). Total noise estimates were derived from land use regression (LUR) noise models developed by Ragettli et al. [18]. In brief, the LUR models for LAeq24h and Lden were built based on 204 noise samples collected during two periods: a two-week sampling period in the summer 2010 [16], and a five-week sampling campaign in the spring 2014. Measurements were performed at all sites for at least one week. Noise levels were measured continuously in 2-min intervals (recording the 2-min averages) with the Type-II Sound Level Meter Data Logger Noise Sentry (Convergence Instruments, Sherbrooke, QC, Canada). LUR models were developed by establishing a statistical relationship between the noise measurements and surrounding determinants of the built and natural environment. The models explained 68% and 69% of the variability in environmental noise levels for LAeq24h and Lden, respectively. Among others, the models included predictor variables related to all traffic noise sources: road (length of major roads within 50 m, annual average traffic counts at the nearest road, number of intersections within 200 m, distance to highways), rail (presence of a railway within 150 m, presence of a railroad shunting yard within 100 m) and air traffic (1 km or less from the NEF25 contour). Main predictors of measured noise levels were road traffic and vegetation variables. A map of the estimated noise levels for Montreal can be found in [18]. In addition, for each subject, the distance of the residential six digit postal code to the nearest transportation noise source such as major roads, railways, and the NEF25 contour was computed using the open software PostGreSQL 9.1 (PostgreSQL Global Development Group, Berkeley, CA, USA). The distances to the closest major road and railway (which also includes the main railway lines of railroad shunting yards) were divided in 50 m categories based on the distributions [7]. The distance to the NEF25 contour was categorized by 1000 m (1–1000, 1001–2000, >2000), including a category within the NEF25 contour [22].

2.4. Survey Weights and Statistical Analysis In order for noise annoyance estimates produced from the survey to be representative of the total population of Montreal (ě18 years), survey weights were computed for each record considering the survey strata (road, rail, NEF25 and non-exposed), and age, gender and education. First, we computed the sampling weights for each stratum by multiplying the reciprocals of the proportions of respondents by strata and the corresponding proportion of the total population aged 18 and more (obtained from the 2011 Census [19]). These sampling weights were then “raked” so that weighted totals would match census totals for sex, and nine classes of age and education (i.e., three classes of education ˆ three classes of age) [27]. Int. J. Environ. Res. Public Health 2016, 13, 90 5 of 13

The association between the percentage of “highly annoyed” (%HA) and “annoyed” persons (%A) (with a dichotomous response variable) due to single environmental noise sources (i.e., trafﬁc, trains, airplanes) and to total environmental noise sources, and continuous noise levels (i.e., LAeq24h and Lden) were analyzed using log-binomial regression models. We used log-binomial regression models to obtain covariate-adjusted Prevalence Proportion Ratios (PPR) of noise annoyance in relation to noise levels as suggested by Barros et al., 2003 [28]. Marginal proportions for annoyed and highly annoyed people in 5 dB(A) noise categories were derived from these regression models using the STATA command “margins”. Similarly, the annoyance due to trafﬁc, trains and airplanes was studied in relation to distance (in categories) to the respective transportation source. All models were adjusted for age, sex and education (in categories described in Table1). Statistical analyses were conducted in STATA SE version 13.0 (StataCorp, College Station, TX, USA).

3. Results In total, 15,697 randomly selected telephone numbers were dialed (of which 21.5% were out of service or not valid) to achieve 4500 respondents. Using the American Association for Public Opinion Research (AAPOR) standards [29], this corresponds to a response rate of 46.8%. After excluding ineligible observations (missing data for age and/or education for generating survey weights), 4336 observations were available for computing the survey weights and carrying out the analyses. The distribution of the respondents among the four survey strata road, rail, air and the non-exposed was 25.0%, 25.1%, 24.6% and 25.3%, respectively. The socio-demographic characteristics of the study sample with complete data, as well as the noise exposure categories in relation to distance to transportation noise source and estimated LAeq24h noise levels are presented in Table1 (corresponding population-weighted results are presented in the supplemental material, Table S1). The crude proportions of individuals of the study sample that have been living <5 years, 6–10 years and >10 years at their place of residence was 32.3%, 21.0% and 46.7%, respectively. Estimated environmental noise levels at postal codes ranged between 50.1–76.1 for LAeq24h and between 55.0–78.7 dB(A) for Lden. The majority of the study participants (48%) were exposed to noise levels (Lden) between 61 and 65 dB(A) (see Table1 for LA eq24h). The weighted proportion of the population that was annoyed and highly annoyed by at least one environmental noise source (i.e., with outdoor origin) was 42.3% and 23.1%, respectively. Figure2 shows the population-weighted percentage of highly annoyed (%HA) and percentage of annoyed (%A) to the various noise sources. In Montreal, 20.1% of the population was annoyed by road traffic noise, followed by airplane noise with 13.0%. Noise from trains ranked seventh with 6.1% of the population. Similarly, the %HA was highest for road traffic (8.8%), followed by aircraft noise (5.7%). In the weighted sample, 18.1% of the population considered themselves to be quite a bit or a lot sensitive to noise. We did not detect a statistically significant difference in noise sensitivity between survey strata.

Table 1. Characteristics of the survey respondents (n = 4336). Results are not population-weighted.

Characteristics % Age categories 18–29 6.9 30–39 13.5 40–49 19.1 50–59 23.3 60–69 19.9 70–80 11.6 >80 5.7 Int. J. Environ. Res. Public Health 2016, 13 7

Table 1. Cont.

Characteristics % Int. J. Environ. Res. Public Health 2016 13 , , 90 Sex 6 of 13 Women 48.2 Men Table 1. Cont. 51.8 Educational level Characteristics % No diploma or elementary school 8.0 Sex High school 17.5 Women 48.2 MenCollege 51.826.2 EducationalUniversity level degree 48.3 DistanceNo of diploma residential or elementarypostal code school to noise source (in m) median 8.0 (range) High school 17.5 Major road 148 (1–2043) College 26.2 UniversityRailway degree 722 48.3(5–4531) DistanceNEF25 of residential contour postal of Montreal code to International noise source (inAirport m) median 5887 (0–25,743) (range) Major road 148 (1–2043) Exposure byRailway transportation noise source 722 (5–4531) NEF25 contourRoad of (within Montreal 100 International m of a major Airport road) 5887 (0–25,743) 39.1 ExposureAirplanes by transportation (within 1000 noisem of sourceNEF zone) 24.4 Road (withinRail (within 100 m 150 of a m major of a railway) road) 39.1 19.1 Airplanes (within 1000 m of NEF zone) 24.4 Rail (within 150Not-exposed m of a railway) 19.129.7 EstimatedNot-exposed LAeq24 noise level * 29.7

Estimated LAloweq24 (<55noise dBA) level * 7.9 lowmedium (<55 dBA)(56–60 dBA) 7.9 45.6 medium (56–60 dBA) 45.6 highhigh (61–65 (61–65 dBA) dBA) 34.8 34.8 veryvery high high (>65 (>65 dBA) dBA) 11.7 11.7 Note:Note: * A *LandA Land Use Use Regression Regression model waswas used used to to estimate estimate the the noise noise levels levels for the for year the 2014. year 2014.

Figure 2. The proportion annoyed (somewhat, quite a bit, a lot) and highly annoyed (quite a bit, a lot) Figuredue to 2. eight The outdoor proportion noise sourcesannoyed in the(somewhat, weighted study quite sample. a bit, a lot) and highly annoyed (quite a bit, a lot) due to eight outdoor noise sources in the weighted study sample. We found a strong relationship between the distance to the noise source and the prevalence of annoyance from all transportation noise sources. Figure3 shows the adjusted proportions with 95% conﬁdence intervals of highly annoyed and annoyed persons predicted at various distance categories from the noise sources (log-binomial regression models are presented in the Supplement Tables S2–S7). For example, the adjusted %HA due to road trafﬁc noise was 22% within 50 m, 10% within 51–100 m, Int. J. Environ. Res. Public Health 2016, 13 8

We found a strong relationship between the distance to the noise source and the prevalence of annoyance from all transportation noise sources. Figure 3 shows the adjusted proportions with 95% confidence intervals of highly annoyed and annoyed persons predicted at various distance categories from the noise sources (log-binomial regression models are presented in the Supplement Tables S2–S7). For example,Int. J. Environ. the Res. adjusted Public Health %HA2016 ,due13, 90 to road traffic noise was 22% within 50 m, 10% within 751–100 of 13 m, and below 10% at categories of 100 and more meters away from major roads. Thus, a sharp decrease of %HAand and below %A due 10% to at road categories traffic of noise 100 and was more observed meters awayfor the from first major 150 roads.m. The Thus, %HA a sharpand %A decrease due to of noise from %HAtrains and rapidly %A due decreased to road trafﬁc when noise moving was observed away from for the the ﬁrst railway 150 m. Thetracks. %HA We and observed %A due tothat a noise from trains rapidly decreased when moving away from the railway tracks. We observed that a considerable proportion of the population was disturbed by aircraft noise, even outside of the NEF25 considerable proportion of the population was disturbed by aircraft noise, even outside of the NEF25 zone. zone.In the In residential the residential areas areas between between 1 1and and 2 2 km km aw awayay fromfrom the the NEF25 NEF25 zone, zone, the adjustedthe adjusted predictions predictions of annoyanceof annoyance and high and high annoyance annoyance was was still still 39% 39% and and 23%, 23%, respectively.

Figure 3. Estimated marginal proportions of annoyed and highly annoyed persons (with 95% CI) from Figureroad 3. traffic Estimated (A); train (marginalB) and airplane proportions (C) noise by of distance annoyed to the and transportation highly annoyed sources (in persons categories), (with 95%adjusted CI) from for age, road education traffic and ( sex.A); train (B) and airplane (C) noise by distance to the transportation sources (in categories), adjusted for age, education and sex. Figure4 illustrates the exposure-response relationships for %HA and %A from road traffic, trains, Figureairplanes 4 illustrates and for annoyance the exposure-res from totalponse environmental relationships noise for in relation%HA and to modeled %A from LA roadeq24h (left)traffic, and trains, Lden (right) in the total study population. The adjusted marginal proportions of annoyance were airplanes and for annoyance from total environmental noise in relation to modeled LAeq24h (left) and Lden predicted for 5 dB(A) noise categories. For %HA, we found a Prevalence Proportion Ratios of 1.10 (right) in the total study population. The adjusted marginal proportions of annoyance were predicted for (95% CI: 1.07–1.13) and 1.04 (1.02–1.06) per 1 dB(A) increase in Lden levels, for annoyance due to 5 dB(A)traffic noise noise categories. and to total For environmental %HA, we found noise, a Prevalence respectively Proportion (the log-binomial Ratios regression of 1.10 (95% models CI: for 1.07–1.13) %A and 1.04and (1.02–1.06) %HA related per to L1den dB(A)and LAincreaseeq24 are in provided Lden levels, in the for supplemental annoyance material, due to traffic Tables S8–S23).noise and The to total adjusted proportion of the population highly annoyed by road traffic noise was 4% at an Lden of 55 environmental noise, respectively (the log-binomial regression models for %A and %HA related to Lden dB(A) and reached 24% at an Lden of 75 dB(A). While we observed increasing proportions of %HA and and LAeq24 are provided in the supplemental material, Tables S8–S23). The adjusted proportion of the %A from road traffic noise and the total environmental noise with both increasing LAeq24h and Lden populationlevels, highly no relationship annoyed between by road estimated traffic noise environmental was 4% at noise an L levelsden of and 55 annoyancedB(A) and from reached trains 24% and at an airplane noise was found. Comparing the adjusted marginal proportion of highly annoyed people from total environmental noise in the noise sensitive (a lot, quite a bit) and non-sensitive (not at all, a bit, somewhat) subgroups showed that noise sensitivity increases annoyance independently by noise exposure level (Table S24). Int. J. Environ. Res. Public Health 2016, 13 9

Lden of 75 dB(A). While we observed increasing proportions of %HA and %A from road traffic noise and the total environmental noise with both increasing LAeq24h and Lden levels, no relationship between estimated environmental noise levels and annoyance from trains and airplane noise was found. Comparing the adjusted marginal proportion of highly annoyed people from total environmental noise in the noise sensitive (a lot, quite a bit) and non-sensitive (not at all, a bit, somewhat) subgroups showed Int. J. Environ. Res. Public Health 2016, 13, 90 8 of 13 that noise sensitivity increases annoyance independently by noise exposure level (Table S24).

Figure 4. Estimated marginal proportions of annoyed (a lot, quite a bit, somewhat) and highly annoyed Figure 4. Estimated marginal proportions of annoyed (a lot, quite a bit, somewhat) and (a lot, quite a bit) persons in the total study population (with 95% CI) from trafﬁc, airplanes, trains and totalhighly environmental annoyed (a noiselot, quite (includes a bit) noise persons from in transport, the total neighborhood, study population industrial (with and95% commercial CI) from

sources,traffic, schools,airplanes, parks, trains animals and and total birds) environmental as a function of noise LAeq24h (inc(leftludes) and noise Lden ( rightfrom), adjustedtransport, for age,neighborhood, education and industrial sex. and commercial sources, schools, parks, animals and birds) as a

function of LAeq24h (left) and Lden (right), adjusted for age, education and sex. 4. Discussion 4. DiscussionWe assessed the prevalence of noise annoyance in Montreal, Canada, in relation to distance to transportation noise sources and to estimated environmental noise levels. The most common outdoor noiseWe sourcesassessed of the annoyance prevalence in the of populationnoise annoyance were road in Montreal, traffic noise, Canada, followed in re bylation airplane to distance noise. We to transportationobserved a clear noise relationship sources and between to estimated the distance environmental to transportation noise levels. noise sourcesThe most and common the prevalence outdoor noiseof noise sources annoyance of annoyance caused in by the road population traffic, trainswere ro andad traffic airplanes. noise, As followed the distance by airplane to major noise. roads, We observedrailways a and clear NEF25 relationship zone of between the Montreal the distance International to transportati airporton increased, noise sources %HA and and the %A prevalence due to the of noisecorresponding annoyance traffic caused noise by road gradually traffic, decreased. trains and However,airplanes. weAs notedthe distance a relationship to major with roads, total railways noise andlevels NEF25 (i.e., LAzoneeq24h of theand Montreal Lden) only International with annoyance airport increased, from road %HA traffic and and %A from due total to th environmentale corresponding noise sources. Our results are consistent with other studies that showed that the prevalence of annoyance and high annoyance is higher close to major roads, railways and airports than further away [7,17,30,31]. For instance, a national Canadian study on road traffic noise annoyance in 2005 found that people living within a self-reported proximity of 30 m to a heavily traveled road were more likely to indicate that they were highly annoyed by traffic noise than those who lived 30–500 and >500 m away. Compared to people living >500 m away from busy roads, respondents living next to a busy road were 6.0 times more likely to be bothered by road traffic noise [17]. The %HA and %A in the close proximity to railways and major roads was lower in Montreal than in previous studies conducted elsewhere in the world. For instance, in the Swedish municipality of Lerum, the %A within 101–150 m of a railway was 50% while it was 30% in Montreal. Possible explanations for the higher prevalence in the latter study Int. J. Environ. Res. Public Health 2016, 13, 90 9 of 13 compared to Montreal are the higher train frequencies in Lerum and the fact that all study participants lived relatively close to the railway (0–800 m in Lerrum versus 5–4531 m in Montreal) [7]. In addition, differences in %HA and %A may also arise due to differences in train type frequencies (passenger trains and freight trains), railway infrastructure and train speed [32,33]. Similarly to previous studies, we found a significant exposure-response relationship between noise levels and annoyance induced by road traffic noise. However, comparing our exposure-response curves to similar produced functions of %HA and %A as for example in [4–6,34], we generally observed lower percentages of annoyance for the same level of Lden in Montreal. For instance, applying the established equation for %HA due to road traffic noise by Miedema and Oudshoorn [5] that is based on several studies, the %HA at a Lden of 55 dB(A) and 75 dB(A) were 6.4% and 36.7%, respectively. In Montreal, the corresponding percentages were 3.6% and 23.9%. In fact, the noise level in Lden at which more than 10% of the population was highly annoyed by road traffic noise was 5 dB(A) lower (and outside the 95% CI) in Montreal than presented in the synthesis curves by Miedema and Oudshoorn [5]. Unlike other studies, we did not find a clear relationship between noise levels and annoyance from trains and airplanes (although associations were found with distance from these sources). There are a number of considerations related to the methods and the characteristics of the built environment that may explain the observed differences in the exposure-response relationship between annoyance and noise levels in our study compared to previous studies. First, we used LUR models developed based on the association between noise measurements of the total outdoor sound environment and land use characteristics of the built and natural environment. Previous studies mostly assessed the dose-response relationship for annoyance with numerical models that predict noise for a specific source by means of physical rules of noise propagation and attenuation [4,5,7,35,36]. LUR models may be unable to assess small-scale variations of noise levels due to physical characteristics of noise (e.g., sound reflection from the built environment). In fact, LUR modeling represents a rather new method to model noise in epidemiological studies. The LUR models used in this study explained close to 70% of the spatial variability of environmental noise levels in Montreal [18]. They could potentially be improved by including more detailed information on the road and railway networks as well as airplane and train frequencies. Nonetheless, LUR models have been suggested as a promising tool for exposure assessment in epidemiological studies, especially for areas where high quality numerical models are not available. A recent comparison of three noise LUR models—that were built with 20-min short-term noise measurements to estimates within-city variability of road traffic noise in three European cities—to standard noise models showed no systematic differences in the spatial patterns between LUR estimates and noise estimates from standard noise models. However, LUR estimates tended to be higher than those of standard models, especially at low noise levels [37]. Additionally, propagation noise models may have limitations in accurately representing the exposure situation encountered by populations. Comparisons of such noise propagation estimates with continuous noise measurements have shown systematic underestimation of modeled noise levels, partly as models neglect local noise sources such as for example parking cars [11,38]. Further comparisons between noise propagation models and LUR models for various transportation noise sources, and potential differences when using them in health assessments are warranted. Second, it needs to be considered that the noise estimates used in our study are from noise LUR models which themselves present limitations [16,37]. For example, noise levels were measured using Type-II devices which do not accurately measure noise levels below 40 dB(A) and although corrected, sound reflection from buildings cannot be ruled out. Most importantly, the models were built based on 2-min average noise levels. It has been shown that there can be large differences in maximum noise levels between road traffic and railway noise, even if the total sound levels are the same [7]. Thus, in contrast to the rather constant noise from road traffic, the sampling interval may be too large to capture the intermittent noise produced by trains or airplanes passing by. This may explain why we did not find a relationship between noise levels (LAeq24h and Lden) and annoyance from trains and airplanes. Furthermore, it may be that instead of total environmental noise, only noise from railway and aircrafts, Int. J. Environ. Res. Public Health 2016, 13, 90 10 of 13 which we did not model, is related to annoyance from these sources. Alternatively, road traffic noise that contributes to environmental noise in proximity to railways and airports may contribute more to LAeq24h than noise from these sources and may also partly explain why we did not find a clear relationship between total environmental noise levels and annoyance from railway and aircraft noise. To our knowledge, the noise LUR models by Ragettli et al. [18] are the first aiming to include also noise from rail and air traffic noise. To better understand the insignificant relationship between noise levels and annoyance from trains and airplanes found in this study, further investigations are needed that formally compare noise exposure estimates close to railways and airports from LUR models to those of propagation models using noise measurements of high temporal resolution. Additionally, it is recommended that further studies on transportation noise induced annoyance take into account differences in acoustic properties between the noise sources. Finally, it should be emphasized that most of the studies on transportation noise induced annoyance were conducted in Europe, where cities are frequently configured differently than in North America. For example, the distance of the buildings relative to the street and the street width is generally larger in Montreal than in many European cities, in particular in residential areas. This may also contribute to the lower prevalence of annoyance in close proximity to major roads observed in Montreal compared to other cities [7,30]. In addition, due to the cold winter temperatures in Montreal, houses are isolated well and windows—that are often double or triple glazed—are normally closed during winter months, which likely also contributes to lower perceived disturbance from traffic noise during the year [2,9]. However, a cold winter climate may not solely lead to lower noise annoyance prevalence as for example a Norwegian study [34] found stronger exposure-response relationships between road-traffic noise and indoor noise annoyance than reported in our study and in Miedema and Oudshoorn [5]. Despite the methodological differences, this would imply that the noise annoyance prevalence and the exposure-response relationship are variable between cities and continents and are dependent on local characteristics such as the built environment and the characteristics of the traffic noise sources. This is contrary to the hypothesis from Miedema and Oudshoorn [5] who hypothesized that there are no important differences between countries in the reaction of the population to similar noise exposures and future work is needed to better characterize noise exposure differences between cities, countries and continents. To our knowledge, this is the first comprehensive study assessing the prevalence of annoyance to various transportation noise sources in a North American city, where policies regulating noise are less prevalent and noise-related health effects are less studied than in European countries [11,12]. Unlike previous studies establishing exposure-response curves for noise annoyance, we related annoyance to noise levels that are based on actual noise measurements. A limitation of our study is the lack of information on additional exposure modifiers and exposure misclassification. Noise exposure was assessed by modeled noise levels and distances to transportation noise sources based on residential postal codes only. Thus, possible misclassification of exposure cannot be ruled out as information on the position of the main living quarters relative to the road or railway and on façade insulation could not be considered [2,9,39]. It is important to note though that our focus was on the community response, not on individuals. As we assessed prevalence of annoyance, we did not have the information on whether people moved to a quieter area of the city due to noise annoyance. The fact that the prevalence of people reporting to be a lot or quite a bit sensitive to noise is similar in all survey strata suggests that people easily annoyed by noise did not necessarily live in quieter areas [40].

5. Conclusions Our study provides the ﬁrst comprehensive information on the relation between transportation noise sources, total environmental noise levels and prevalence of noise annoyance in Montreal. Our study clearly shows that Montreal residents living near busy roads, main railway lines, as well as within and close to the NEF25 zone of the Montreal International airport are annoyed by transportation noise. Int. J. Environ. Res. Public Health 2016, 13, 90 11 of 13

There is increasing evidence that exposure to environmental noise is related—in addition to annoyance—to various negative health outcomes such sleep problems, impaired cognitive performance, hypertension and cardiovascular disease [1]. Since national approaches to assess noise exposure currently do not exist in Canada, initiatives aiming to reduce and monitor noise exposure in Canadian cities are warranted at various levels of government. Adoption of such directives would require the speciﬁcation, for example, of allowed construction zones around trafﬁc sources and acceptable methods for noise monitoring and modelling.

Acknowledgments: This work was financially supported by the Direction de Santé Publique de l’Agence de la Santé et des Services Sociaux de Montréal and by the Institut National de Santé Publique du Québec. Author Contributions: Martina S. Ragettli conducted the analysis, interpreted the results and wrote the first draft of the manuscript. Sophie Goudreau was involved in the data collection and analyses. Céline Plante and Michel Fournier contributed to the statistical analysis and interpretation of results. Stéphane Perron and Audrey Smargiassi conceived the study, contributed to the design of the study and writing of the manuscript. All authors reviewed and approved the final manuscript. Conflicts of Interest: The authors declare no conflict of interest.

References

1. Basner, M.; Babisch, W.; Davis, A.; Brink, M.; Clark, C.; Janssen, S.; Stansfeld, S. Auditory and non-auditory effects of noise on health. Lancet 2014, 383, 1325–1332. [CrossRef] 2. Öhrström, E.; Skånberg, A.; Svensson, H.; Gidlöf-Gunnarsson, A. Effects of road traffic noise and the benefit of access to quietness. J. Sound Vib. 2006, 295, 40–59. [CrossRef] 3. Dratva, J.; Zemp, E.; Dietrich, D.F.; Bridevaux, P.-O.; Rochat, T.; Schindler, C.; Gerbase, M.W. Impact of road traffic noise annoyance on health-related quality of life: Results from a population-based study. Qual. Life Res. 2010, 19, 37–46. [CrossRef][PubMed] 4. Heritier, H.; Vienneau, D.; Frei, P.; Eze, I.C.; Brink, M.; Probst-Hensch, N.; Roeoesli, M. The association between road traffic noise exposure, annoyance and health-related quality of life (HRQOL). Int. J. Environ. Res. Public Health 2014, 11, 12652–12667. [CrossRef][PubMed] 5. Miedema, H.M.E.; Oudshoorn, C.G.M. Annoyance from transportation noise: Relationships with exposure metrics DNL and DENL and their confidence intervals. Environ. Health Perspect. 2001, 109, 409–416. [CrossRef][PubMed] 6. European Commission. Position paper on Dose Response Relationships between Transportation Noise and Annoyance; Office for Official Publications of the European Communities: Luxembourg, Belgium, 2002; pp. 1–24. 7. Öhrstrom, E.; Barregård, L.; Andersson, E.; Skånberg, A.; Svensson, H.; Ängerheim, P. Annoyance due to single and combined sound exposure from railway and road traffic. J. Acoust. Soc. Am. 2007, 122, 2642–2652. [CrossRef][PubMed] 8. WHO. Burden of disease from environmental noise. In Quantification of Healthy Life Years Lost in Europe; WHO Regional Office for Europe: Copenhagen, Denmark, 2011; pp. 1–106. 9. Babisch, W.; Swart, W.; Houthuijs, D.; Selander, J.; Bluhm, G.; Pershagen, G.; Dimakopoulou, K.; Haralabidis, A.S.; Katsouyanni, K.; Davou, E. Exposure modifiers of the relationships of transportation noise with high blood pressure and noise annoyance. J. Acoust. Soc. Am. 2012, 132, 3788–3808. [CrossRef] [PubMed] 10. European Commission. Directive 2002/49/EC of the European Parliment and of the Council of 25 June 2001 relating to the assessment and management of environmental noise. Off. J. Eur. Commun. 2002, 117, 6–7. 11. Lee, E.Y.; Jerrett, M.; Ross, Z.; Coogan, P.F.; Seto, E.Y.W. Assessment of traffic-related noise in three cities in the United States. Environ. Res. 2014, 132, 182–189. [CrossRef][PubMed] 12. Gan, W.Q.; McLean, K.; Brauer, M.; Chiarello, S.A.; Davies, H.W. Modeling population exposure to community noise and air pollution in a large metropolitan area. Environ. Res. 2012, 116, 11–16. [CrossRef] [PubMed] 13. Steele, C. A critical review of some traffic noise prediction models. Appl. Acoust. 2001, 62, 271–287. [CrossRef] 14. King, G.; Roland-Mieszkowski, M.; Jason, T.; Rainham, D.G. Noise levels associated with urban land use. J. Urban. Health Bull. N. Y. Acad. Med. 2012, 89, 1017–1030. [CrossRef][PubMed] Int. J. Environ. Res. Public Health 2016, 13, 90 12 of 13

15. Zuo, F.; Li, Y.; Johnson, S.; Johnson, J.; Varughese, S.; Copes, R.; Liu, F.; Wu, H.J.; Hou, R.; Chen, H. Temporal and spatial variability of traffic-related noise in the city of Toronto, Canada. Sci. Total Environ. 2014, 472, 1100–1107. [CrossRef][PubMed] 16. Goudreau, S.; Plante, C.; Fournier, M.; Brand, A.; Roche, Y.; Smargiassi, A. Estimation of spatial variations in urban noise levels with a land use regression model. Environ. Pollut. 2014, 3, 48–58. [CrossRef] 17. Michaud, D.S.; Keith, S.E.; McMurchy, D. Annoyance and disturbance of daily activities from road traffic noise in Canada. J. Acoust. Soc. Am. 2008, 123, 784–792. [CrossRef][PubMed] 18. Ragettli, M.S.; Goudreau, S.; Plante, C.; Fournier, M.; Hatzopoulou, M.; Perron, S.; Smargiassi, A. Statistical modeling of the spatial variability of environmental noise levels in Montreal, Canada, using noise measurements and land use characteristics. J. Expos. Sci. Environ. Epidemiol. 2015, in press. 19. Statistics Canada. Population and Dwelling Counts, for Canada and Economic Regions, 2011 and 2006 Censuses. Available online: https://www12.statcan.gc.ca/census-recensement/2011/dp-pd/ hlt-fst/pd-pl/Table-Tableau.cfm?LANG=Eng&T=1401&SR=1&S=9&O=D&RPP=25 (accessed on 22 December 2015). 20. Transport Canada. Noise Exposure Forecast and Related Programs. https://www.tc.gc.ca/eng/ civilaviation/standards/aerodromeairnav-standards-noise-nef-924.htm (accessed on 22 December 2015). 21. BOING. Courbe D’ambiance Sonore. Available online: http://www.boeing.com/commercial/noise/montreal 2011.pdf (accessed on 11 November 2011). 22. Dale, L.M.; Debia, M.; Mudaheranwa, O.C.; Plante, C.; Smargiassi, A. An exploration of transportation source contribution to noise levels near an airport. Environ. Pollut. 2013, 3.[CrossRef] 23. Price, K.; Perron, S. Avis de Santé Publique Concernant Les Impacts Sanitaires du Bruit Engendré Par Les Activités Ferroviaires de la Compagnie CN à Pointe-Saint-Charles; Agence de la santé et des services sociaux de Montréal: Montreal, QC, Canada, 2013; pp. 1–19. 24. Statistics Canada. Postal Code Conversion File (PCCF), Reference Guide. Available online: http://geodepot. statcan.ca/2006/Reference/Freepub/92-153-GWE/2007002/using.htm (accessed on 22 December 2015). 25. Niemann, H.; Bonnefoy, X.; Braubach, M.; Hecht, K.; Maschke, C.; Rodrigues, C.; Robbel, N. Noise-induced annoyance and morbidity results from the pan-European LARES study. Noise Health 2006, 8, 63–79. [CrossRef][PubMed] 26. Guski, R.; Felscher-Suhr, U.; Schuemer, R. The concept of noise annoyance: How international experts see it. J. Sound Vib. 1999, 223, 513–527. [CrossRef] 27. Frankel, M. Sampling theory. In Handbook of Survey Research, 2nd ed.; Emerald Group Publishing Limited: Bingley, UK, 2010. 28. Barros, A.J.; Hirakata, V.N. Alternatives for logistic regression in cross-sectional studies: An empirical comparison of models that directly estimate the prevalence ratio. BMC Med. Res. Methodol. 2003, 3. [CrossRef][PubMed] 29. American Association for Public Opinion Research (AAPOR). Response Rates—An Overview. Available online: http://www.aapor.org/AAPORKentico/Education-Resources/For-Researchers/Poll-Survey-FAQ/ Response-Rates-An-Overview.aspx (accessed on 10 June 2015). 30. Di, G.; Liu, X.; Lin, Q.; Zheng, Y.; He, L. The relationship between urban combined traffic noise and annoyance: An investigation in Dalian, north of China. Sci. Total Environ. 2012, 432, 189–194. [CrossRef] [PubMed] 31. Morihara, T.; Sato, T.; Yano, T. Comparison of dose-response relationships between railway and road traffic noises: The moderating effect of distance. J. Sound Vib. 2004, 277, 559–565. [CrossRef] 32. Pennig, S.; Quehl, J.; Mueller, U.; Rolny, V.; Maass, H.; Basner, M.; Elmenhorst, E.-M. Annoyance and self-reported sleep disturbance due to night-time railway noise examined in the field. J. Acoust. Soc. Am. 2012, 132, 3109–3117. [CrossRef][PubMed] 33. Gidlof-Gunnarsson, A.; Ogren, M.; Jerson, T.; Ohrstrom, E. Railway noise annoyance and the importance of number of trains, ground vibration, and building situational factors. Noise Health 2012, 14, 190–201. [CrossRef][PubMed] 34. Klæboe, R.; Amundsen, A.; Fyhri, A.; Solberg, S. Road traffic noise—The relationship between noise exposure and noise annoyance in Norway. Appl. Acoust. 2004, 65, 893–912. [CrossRef] Int. J. Environ. Res. Public Health 2016, 13, 90 13 of 13

35. Janssen, S.A.; Vos, H.; van Kempen, E.E.M.M.; Breugelmans, O.R.P.; Miedema, H.M.E. Trends in aircraft noise annoyance: The role of study and sample characteristics. J. Acoust. Soc. Am. 2011, 129, 1953–1962. [CrossRef][PubMed] 36. Klaeboe, R. Noise and health: Annoyance and interference. In Encyclopedia of Environmental Health; Elsevier: Philadelphia, PA, USA, 2011; pp. 152–163. 37. Aguilera, I.; Foraster, M.; Basagaña, X.; Corradi, E.; Deltell, A.; Morelli, X.; Phuleria, H.C.; Ragettli, M.S.; Rivera, M.; Thomasson, A.; et al. Application of land use regression modelling to assess the spatial distribution of road traffic noise in three European cities. J. Expo. Sci. Environ. Epidemiol. 2015, 25, 97–105. [CrossRef][PubMed] 38. Mioduszewski, P.; Ejsmont, J.A.; Grabowski, J.; Karpiński,D. Noise map validation by continuous noise monitoring. Appl. Acoust. 2011, 72, 582–589. [CrossRef] 39. Eriksson, C.; Nilsson, M.E.; Stenkvist, D.; Bellander, T.; Pershagen, G. Residential traffic noise exposure assessment: Application and evaluation of European Environmental Noise Directive maps. J. Expo. Sci. Environ. Epidemiol. 2013, 23, 531–538. [CrossRef][PubMed] 40. Okokon, E.O.; Turunen, A.W.; Ung-Lanki, S.; Vartiainen, A.-K.; Tiittanen, P.; Lanki, T. Road-traffic noise: Annoyance, risk perception, and noise sensitivity in the Finnish adult population. Int. J. Environ. Res. Public Health 2015, 12, 5712–5734. [CrossRef][PubMed]

© 2015 by the authors; licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons by Attribution (CC-BY) license (http://creativecommons.org/licenses/by/4.0/).