Residential traffic noise exposure and headaches: Results from the population-based heinz nixdorf recall study.

CONTEXT AND AIM: The link between headaches and exposure to loud noise in occupational settings has been established. However, the effect of less intense but chronic residential traffic noise exposure on headache occurrence is less clear. SETTINGS AND
DESIGN: We included 3,025 participants from the Heinz Nixdorf Recall study in Germany for this cross-sectional analysis. METHODS AND MATERIAL: Residential road traffic noise exposure at the 2006-2008 address was modelled in A-weighted decibels (dB(A)) according to the European Noise Directive (2002/49/EC) for 24-hour (Lden) and night-time noise (22-6 h, Lnight). Indoor traffic noise exposure was obtained by modifying Lden and Lnight based on residence orientation, window type, and personal window opening habits. Traffic noise exposure below 55, 45 dB(A), 35 and 25 dB(A) were set as the reference for Lden, Lnight, Lden,indoor and Lnight,indoor, respectively. Average number of days with headache per month over the past three months was ascertained during the follow-up (2011-2015) medical interview. STATISTICAL ANALYSIS USED: Prevalence Odds Ratios (POR) of having eight or more headaches per month per 5 dB(A) increase in traffic noise exposure were calculated using logistic regression, adjusting for age, sex, sport, number of chronic conditions, years of education and smoking status.
RESULTS: The mean age of participants was 58.3. Mean Lden was 54 dB(A). Median monthly headache days was one. No association was seen between traffic noise exposure and having ≥8 headaches/month for all the examined traffic noise indicators. However, traffic noise was positively associated with traffic noise-annoyance and insomnia; and night-time traffic noise-annoyance and insomnia were positively associated with headache.
CONCLUSION: In conclusion, our data did not provide any evidence for an association between chronic traffic noise exposure and prevalence of headaches at this population's exposure levels. This should be explored in different populations given that this is the first study of its type and that noise exposure was generally low in our population.

INTRODUCTION

Traffic noise has been described by the World Health Organization (WHO) as an important environmental problem affecting the health and wellbeing of exposed individuals, especially in urban areas.[1] Sound evidence of the effects of traffic noise on annoyance, cardiovascular diseases and sleep disturbance is currently available.[23] In their most recent publication on guidelines for environmental noise, the WHO identified a need for longitudinal studies to evaluate the health effects of noise, taking into account relevant health outcomes.[4]

Headaches can be a major source of distress and disability, especially among chronic sufferers. Moreover, because headaches affect a relatively large part of the population including those in younger age groups, they lead to a considerable financial burden and loss in productivity. Globally, it is estimated that about 1.7%–4% of adults suffer from chronic headaches defined as headache on 15 days or more a month.[5] According to the Global Burden of Disease study updated in 2018, headaches were the second highest cause of years lost due to disability (YLD).[6]

Experiments have shown that noise can trigger headaches, especially in susceptible individuals.[7] High noise levels in occupational settings have also been reported to cause frequent headaches.[89] The effect of lower, but chronic environmental noise exposure on headaches has been a subject of interest recently, as large parts of the population are exposed to chronic noise throughout their lives. Noise was reported by episodic migraineurs and tension-type headache sufferers as one of many triggers of headaches.[101112] Also, women living with heavy snorers reported more frequent headaches compared to controls.[13] Within the context of the Large Analysis and Review of European housing and health status survey (LARES-survey), those who reported being severely annoyed by traffic noise were more than twice as likely to suffer from migraines.[14] In a survey-based study including 1,125 participants in Bejing, perceived road-noise pollution was found to be positively associated with headache of any frequency. Proximity to main roads partially explained the relationship.[15]

However, research on this topic is scarce and most studies have been carried out in cities with different traffic noise patterns than in Germany. For example, motorcycles were described in a few previous studies to be the main source of traffic noise. They also relied on perceived noise exposure or noise-annoyance alone, and lacked comparison groups.[161718] In order to detect those at risk of adverse health outcomes and plan interventions accordingly, it is important to use more objective measures of exposure.

The aim of this study is to examine the association between residential traffic noise exposure and headaches in participants of the population-based Heinz Nixdorf Recall study.

The objectives are:

To investigate the long-term association between various indicators of residential traffic noise exposure levels and the occurrence of headaches.

To explore the association between traffic noise and both traffic noise-annoyance and insomnia.

To explore the association between both traffic noise-annoyance and insomnia and frequent headaches.

To explore the effect of stratifying for traffic noise-annoyance and insomnia on any observed traffic noise-headache associations.

MATERIALS AND METHODS

Study population

Participants in this analysis come from the Heinz Nixdorf Recall (HNR) study which took place in the neighboring cities of Essen, Mülheim (Ruhr) and Bochum, located in the densely populated Ruhr Area in Western Germany. Recruitment took place between 2000 and 2003 by random selection from population registries. Individuals were eligible if they were between 45 and 75 years of age, had sufficient knowledge of the German language, and were not institutionalized, severely ill or otherwise unable to be medically examined and interviewed. Two follow-up examinations took place during 2006–2008 and 2011–2015 with yearly questionnaires sent to participants in-between. Details of the HNR study have been previously described.[19] The study was approved by the ethics committee of the University Hospitals of Essen and informed consent was obtained from all participants. The timeline of assessments is shown in Figure 1.

Figure 1

Timeline of data collection

Exposure assessment

Road traffic noise exposure was modelled for the year 2006 according to the European Parliament and Council Directive 2002/49/EC as weighted day-evening-night (24 hours, Lden) and night-time (10 p.m.- 6 a.m., Lnight) average noise levels in A-weighted decibels (dB(A)).[20] The noise models considered the following determinants: small-scale topography, dimensions of the buildings, noise barriers, street axis, vehicle type-specific traffic density, speed limit and type of road surface.[21] Noise data were provided as source-specific house façade values by the city administrations.

To define individual residential traffic noise exposure, we determined the maximum noise value (most exposed façade) within a 10-meter buffer around participants’ address coordinates for their home addresses at baseline and using the geographic information system ArcGIS (ArcInfo 10.0, ESRI, Redlands, California, USA). We assumed that average noise values are relatively stable over time.

In order to obtain values closer to true personal exposure at home, information about room use and orientation, type of windows and seasonal window opening habits were assessed via mailed questionnaire (“noise and dust” questionnaire) which was sent out between 2002 and 2006. The questionnaire was returned by 4,351participants (90.4%). According to the method described by Foraster et al.[22] and the “Good Practice Guide on Noise Exposure and Potential Health Effects” of the European Environment Agency,[23] we subtracted 20 dB from Lden/Lnight if the living room/bedroom did not face the postal address street side (faced a court yard or green space). If participants reported that they usually kept their windows closed, we further subtracted 30 and 40 dB for those who had single and double glazed or sound-proof windows, respectively. This was calculated separately for summer, defined as days with average temperature above 10°C (265 days), and winter, defined as days with average temperature below 10°C (100 days). These were combined to yield yearly average 24-hour and night-time indoor noise exposure (Lden-indoor: inside of living room, and Lnight-indoor: inside of bedroom) [Figure 2].

Figure 2

Derivation of indoor noise exposure

Due to the incomplete responses in the self-administered mailed questionnaire, there were 387 missing noise measures for Lden,indoor and 345 for Lnight,indoor, respectively [Table 1].

Table 1

Baseline characteristics of the analysis population (N = 3,025) by headache status at 10-year follow-up

All	Less than 8		8 or more
	N	(%)	N	(%)
	2969	98.15	56	1.85
Gender (missing=0)
Male	1,468	49.4	16	28.6
Years of formal education (missing=4)
<10	249	8.4	4	7.2
11-13	1,627	54.9	34	60.7
14-17	710	24.0	8	14.3
>=18	379	12.8	10	17.9
Employment status (missing=2)
Employed	1,402	47.3	23	41.1
Inactive or housewife	426	14.4	12	21.4
Retired	948	32.0	20	35.7
Unemployed	191	6.4	1	1.8
Comorbidities (missing=0)
0	1,303	43.9	22	39.3
1	1,013	34.1	19	33.9
>=2	653	23.0	15	26.8
Sports (missing=0)
Yes	1,197	40.3	24	42.9
Smoking (missing=0)
Never smoker	1,287	43.4	28	50.0
Former smoker	1,090	36.7	17	30.4
Current smoker	592	19.9	11	19.6
City of residence (missing=0)
Mülheim	1,115	37.5	17	30.4
Bochum	839	28.3	19	33.9
Essen	1,015	34.2	20	35.7
Moved during study period (missing=0)
Yes	459	15.5	16	28.6
Noise-annoyance day (missing=310)
Not at all	1452	54.4	26	55.32
Somewhat annoyed	775	29.1	10	21.28
Moderately annoyed	322	12.1	9	19.15
Strongly annoyed	119	4.5	2	4.26
Noise-annoyance night (missing=309)
Not at all	1980	74.2	34	72.3
Somewhat annoyed	491	18.4	7	14.9
Moderately annoyed	137	5.1	5	10.6
Strongly annoyed	61	2.3	1	2.1
Insomnia (missing=14)
Yes	312	20.0	11	10.6
	Mean	SD	Mean	SD
Body mass index (kg/m2) (missing=13)	27.53	4.34	27.23	4.28
Age (years) (missing=0)	58.34	7.34	57.73	7.07
Lden (dB)a (missing=0)	53.66	9.24	53.77	9.91
Lnight (dB)b (missing=352)	44.79	8.99	44.73	9.45
Ldenindoor (dB)c (missing=380)	34.28	15.02	33.07	15.76
Lnightindoor (dB)d (missing=345)	26.83	15.58	28.08	16.49

a24-hour mean road traffic noise at the most exposed facade.bNight-time mean traffic road noise at the most exposed façade.c24-hour mean road traffic noise in living room.dNight mean road traffic noise in bedroom.

Outcome assessment

Headache occurrence and frequency were ascertained during the 10-year follow-up computer-assisted medical interview (2011–2015) with the questions (1) “Have you ever had a headache?” (2) “Do you still have headaches on occasions?” and (3) “During the past three months, how many headache days have you had monthly on average?” The main outcome was reporting eight or more headache days per month on average during the past three months. This was chosen as clinically relevant cut-off since those who report ≥8 headache days per month are at higher risk of chronic migraines defined as 15 or more headache days per month. A cutoff of 15 headache days or more could not be used as only 18 participants had that outcome. Participants who answered “no” to either of the first two questions were assigned a headache frequency of zero per month.

Baseline characteristics, insomnia and noise-annoyance

Personal data were collected during a standardized computer-assisted interview at baseline (Figure 1). These included sex, age, years of formal education (≥10 years, 11–13 years, 14–17 years, ≥18 years), economic activity (employed, inactive/homemaker, retired, unemployed), smoking status (never smoker, former smoker at any time, current smoker), sport (any, none) and number of comorbidities (0, 1, ≥2 of the following: myocardial infarction, heart failure, stroke, diabetes, emphysema, asthma, cancer, arthritis, slipped disc). Insomnia was assessed in the same questionnaire using a standardized set of questions and defining insomnia as having experienced all of the following symptoms at least twice a week during the past four weeks: difficulty falling asleep, difficulty maintaining sleep or early arousal.[24] Traffic noise-annoyance was assessed in the “noise and dust questionnaire” by the questions “How strongly do you generally feel annoyed by traffic noise at home during the day?” and “How strongly do you generally feel annoyed by traffic noise at home during the night?”; with the response options “not at all, somewhat, moderately, strongly, or extremely”. Since 55% of participants reported not being annoyed at all and less than 5% were strongly or extremely annoyed, we dichotomized noise-annoyance into the two broad categories of not at all annoyed and annoyed to any degree (the latter four responses).

Statistical analysis

Of the 4,814 participants enrolled at baseline, 64.2% (n = 3,089) took part in the 10-year follow-up. Those with complete Lden at baseline addresses and headache data at the 10-year follow-up were included in the analysis where the exposure was Lden and Lnight (n = 3,025) [Figure 3]. For following analyses using indoor noise exposures, participants with complete questionnaire data allowing for the indoor noise derivation were included. For the stratified analysis, those with complete day-time and night-time traffic noise-annoyance and insomnia data (n = 2,428) were included. The numbers included for each exposure indicator are shown in the relevant results tables [Tables 2, 4, 5].

Figure 3

Study sample flowchart

Table 2

Prevalence odds ratios (PORs) and 95% confidence intervals of having eight or more headache days a month in the past three months per 5 dB(A) increase in traffic noise above noise exposure cut-off

Traffic noise indicator	Traffic noise cut-off dB(A)	N	POR	95% confidence intervals		POR	95% confidence intervals
				Lower limit	Upper limit		Lower limit	Upper limit
			Crude			Adjusted*
Lden a	<55	3025	1.07	0.87	1.30	1.08	0.88	1.31
Lnight b	<45	3025	1.03	0.85	1.25	1.04	0.85	1.26
Lden indoor c	<35	2545	0.97	0.83	1.14	0.97	0.82	1.15
Lnight indoor d	<25	2580	1.03	0.91	1.18	1.03	0.90	1.19

a24-hour mean road traffic noise at the most exposed façade.bNight-time road traffic noise at the most exposed façade.c24-hour mean road traffic noise in living room.dNight mean road traffic noise in bedroom.*Adjusted for: sex, age, years of formal education (≥10 years, 11–13 years, 14–17 years, ≥18 years), smoking status (never smoker, ex-smoker, current smoker), sport (yes, no) and number of comorbidities (0, 1, ≥2).

Table 4

Crude prevalence odd ratios (PORs) and their 95% confidence intervals for traffic noise-annoyance and insomnia per 5 dB(A) increase in traffic noise exposure

Traffic noise indicator	N	Reference dB(A)	POR	95% confidence interval
				Lower limit	Upper limit
Annoyance daye
Lden a	2715	<55	1.76	1.64	1.88
Lden indoor c	2537	<35	1.32	1.26	1.38
Annoyance nighte
Lden a	2716	<55	1.58	1.48	1.68
Lnight b	2716	<45	1.55	1.46	1.65
Lnight indoor d	2565	<25	1.22	1.17	1.27
Insomniaf
Lden a	3011	<55	1.05	0.97	1.15
Lnight b	3011	<45	1.06	0.98	1.16
Lnight indoor d	2667	<25	1.06	1.00	1.12

a24-hour mean road traffic noise at the most exposed façade.bNight-time mean traffic road noise at the most exposed façade.c24-hour mean road traffic noise in living room.dNight mean road traffic noise in bedroom.eAnswered with any of the following: somewhat, moderately, strongly, or extremely annoyed.fDifficulty falling asleep or maintaining sleep, or early arousal on at least two nights a week during the past four weeks.

Table 5

Crude prevalence odds ratios (PORs) and their 95% confidence intervals for having eight or more headache days a month in the past three months per 5 dB(A) increase in traffic noise exposure above the reference stratified by traffic noise-annoyance and insomnia

Traffic noise indicator	Reference dB(A)	POR	Lower limit	Upper limit	POR	Lower limit	Upper limit
Day time traffic noise-annoyance e
		Yes (n = 1,078)			No (n = 1,350)
Lden a	<55	1.03	0.72	1.46	0.97	0.65	1.46
Lden,indoor c	<35	0.97	0.76	1.24	0.93	0.70	1.23
Night time traffic noise-annoyancee
		Yes (n = 610)			No (n = 1,818)
Lden a	<55	1.11	0.73	1.66	0.88	0.61	1.25
Lnight b	<45	1.09	0.72	1.63	0.86	0.61	1.21
Lnight,indoor d	<25	1.06	0.82	1.36	1.02	0.85	1.22
Insomniaf
		Yes (n = 257)			No (n = 2,171)
Lden a	<55	1.36	0.79	2.36	0.89	0.66	1.19
Lnight d	<45	1.35	0.80	2.29	0.86	0.65	1.15
Lnight,indoor e	<25	0.97	0.67	1.40	1.04	0.89	1.21

a24-hour mean road traffic noise at the most exposed façade.bNight-time mean traffic road noise at the most exposed façade.c24-hour mean road traffic noise in living room.dNight mean road traffic noise in bedroom.eAnswered with any of the following: somewhat, moderately, strongly, or extremely annoyed.f Difficulty falling asleep or maintaining sleep, or early arousal on at least two nights a week during the past four weeks.

Noise exposure data were categorized into 5 dB(A) increments, which we considered to be the smallest meaningful change in noise level. Values of Lden below 55 dB(A) and Lnight below 45 dB(A) were set at 54 and 44 dB(A), respectively. This threshold was chosen based on the WHO Environmental Noise Guidelines for the European Region,4 which recommend that average traffic Lden is kept below 53 dB (level at which 10% of the population are highly annoyed by traffic noise), and that night-time traffic noise is kept below 45 dB (level at which 3% of the population are highly sleep disturbed). These cutoffs were similar to the median exposure in our study population.

In order to keep the analysis comparable between different noise indicators, and because there are no current recommendations for indoor traffic noise, cutoffs for Lden,indoor and Lnight,indoor were assigned which were also close to the median exposure of the study population. These were 35 and 25 dB(A), respectively.

Prevalence Odds Ratios (POR) of headaches per 5 dB(A) increase in traffic noise exposure were calculated with their corresponding 95% confidence intervals (95%CI) using logistic regression; adjusting for age, sex, years of formal education, smoking, sport and number of comorbidities.

The associations between traffic noise (exposure) and traffic noise-annoyance/insomnia (outcomes) on the one hand and between traffic noise-annoyance/insomnia (exposures) and headache (outcome) on the other were also calculated by logistic regression, adjusting for the above-mentioned variables.

To explore the effect of day and nighttime traffic noise-annoyance and insomnia on the noise-headache relationship, we repeated the first analysis while stratifying for each of the three variables (yes/no as defined above).

POR was used despite the longitudinal approach to exposure and outcome data collection because past noise exposure was considered to be a proxy for current exposure. Also, baseline prevalence of headaches was unknown therefore we could not assess the occurrence of new onset headaches.

Statistical analysis was carried out using SAS studio university 3.5, 2016, and SAS 9.4, SAS Institute Inc., Cary, NC, USA.

RESULTS

Characteristics of the analysis population

The mean age at baseline was 58.3 years (SD=7.3). 51% were female. 72.9% reported ever having headaches, and of those 56.9% were female. 45% and 26% reported any extent of day time and night time traffic noise-annoyance, respectively, and 10.7% had symptoms of insomnia. Characteristics of the analysis population are shown in Table 1 stratified by the outcome. Participants who moved during the study period and who did not suffer from insomnia were more likely to report having headaches. The frequency distributions of average number of headache days per month over the past three months and of Lden noise exposure at baseline are shown in Figures 4 and 5, respectively. Median headache frequency was 1 per month (25th percentile=0, 75th percentile=2, range 0–30). 1.9% of the analysis population had reported having eight or more headaches monthly on average over the past three months. The mean Lden at most exposed façade was 53.7 dB(A) (SD=9.2).

Figure 4

Frequency distribution of average number of headaches per month in the past three months (N=3,025)

Figure 5

Frequency distribution of Lden at baseline in 5 dB(A) categories

Those with an Lden ≥55 dB(A) at baseline home address had less years of education and were more likely to have moved during the study period (18.1% versus 14.3%) and to report traffic noise-annoyance. Other characteristics did not differ between the higher and lower exposure groups (Supplementary Table 2).

Supplementary table 2

Characteristics of included and excluded participants from the analysis (N = 4,814)

All	Excluded		Included
	N	(%)	N	(%)
	1789	37.2	3025	62.8
Gender (missing = 0)
Male	904	50.8	1484	49.1
Years of formal education (missing = 16)
<10	293	16.6	254	8.4
11–13	1010	57.1	1666	55.0
14–17	349	19.7	719	23.8
>=18	118	6.7	389	12.9
Employment status (missing = 15)
Employed	498	28.2	1430	47.2
Inactive or housewife	225	12.7	438	14.5
Retired	993	52.7	970	32.0
Unemployed	113	6.4	192	6.3
Comorbidities (missing = 0)
0	606	34.0	1329	43.8
1	585	32.8	1034	34.1
> = 2	591	33.2	669	22.1
Sports (missing = 1)
Yes	780	43.8	1807	59.6
Smoking (missing = 10)
Never smoker	698	39.4	1316	43.4
Former smoker	552	31.2	1110	36.6
Current smoker	552	29.5	606	20.0
City of residence (missing = 1)
Mülheim	620	34.8	1137	37.5
Bochum	540	30.3	860	28.4
Essen	621	34.9	1035	34.1
Moved during study period (missing = 0)
Yes	305	17.2	478	15.8
Noise-annoyance day (missing = 778)
Not at all	639	48.5	1483	54.5
Somewhat annoyed	405	30.8	786	28.9
Moderately annoyed	172	13.1	331	12.2
Strongly annoyed	83	6.3	101	3.7
Extremely annoyed	16	1.2	20	0.7
Noise-annoyance night (missing = 780)
Not at all	930	70.9	2021	74.2
Somewhat	annoyed	249	19.0	298	18.3
Moderately annoyed	94	7.2	142	5.2
Strongly annoyed	28	2.1	53	2.0
Extremely annoyed	10	0.8	9	0.3
Insomnia (missing = 42)
Yes	251	14.3	324	10.7
	Mean	SD	Mean	SD
Body mass index (kg/m2) (missing = 29)	28.61	5.00	27.52	4.34
Age (years) (missing = 0)	61.90	8.10	58.31	7.34
Lden (dB)a (missing = 41)	54.86	9.54	53.66	9.25
Lnight (dB)b (missing = 41)	45.91	9.27	44.79	9.00
Lden,indoor (dB)b (missing = 980)	38.32	15.81	34.26	15.03
Lnight,indoor (dB)d (missing = 931)	28.63	15.82	26.9	15.60

a24-hour mean road traffic noise at the most exposed facade. bNight-time mean traffic road noise at the most exposed façade. c24-hour mean road traffic noise in living room. dNight mean road traffic noise in bedroom.

Excluded participants were on average 3.5 years older, had a higher BMI, higher noise exposure for all measures, were more likely to be retired, have a lower educational level, report higher noise-annoyance (day and night-time) and insomnia, have two or more chronic morbidities, not practice sport regularly, and be a current smoker at the time of data collection (Supplementary Table 2).

Associations with noise exposure

No association was observed between headache and traffic noise exposure for all the noise exposure indicators in the crude and adjusted analysis [Table 2].

Insomnia was associated with having eigth or more headache days/month, with PORs: 2.12, 95% CI: 1.08–4.15 [Table 3]. There was also an association between traffic noise and both being annoyed by traffic noise, and to a lesser degree, reporting insomnia. The strongest association was between Lden and day-time traffic noise-annoyance, with POR of 1.76, 95% CI: 1.64–1.88 [Table 4]. The results did not change in the adjusted analysis (not shown).

Table 3

Prevalence odds ratios (PORs) and 95% confidence intervals of having eight or more headache days a month in the past three months for those expressing traffic noise-annoyance or insomnia

Predictor variable	N	POR	95% confidence intervals
			Lower limit	Upper limit
Day-time noise-annoyancea	2715	0.97	0.54	1.72
Night-time noise-annoyancea	2716	1.10	0.58	2.09
Insomniab	3011	2.12	1.08	4.15

aAnswered with any of the following: somewhat, moderately, strongly, or extremely annoyed.b Difficulty falling asleep or maintaining sleep, or early arousal on at least two nights a week during the past four weeks.

Stratified by traffic noise-annoyance and insomnia, the PORs for headache were generally higher in the group who reported annoyance or insomnia versus those who did not, but the confidence intervals overlapped [Table 5].

Sensitivity analyses

We ran the analyses using a 10 dB higher noise threshold; as well as excluding those who moved during the study period and those who have never had a headache or have not had any recently. We also limited all analysis to those with complete data on all noise exposure variables, annoyance, and insomnia (n = 2,428) in order to ensure always having the same individuals in the analysis. None of these sensitivity analyses changed the results but some yielded wide confidence intervals which made results difficult to interpret (not shown).

DISCUSSION

In this study, we analyzed the association between road traffic noise exposure and the occurrence of headaches in middle-aged to older residents of a high traffic density urban area in Western Germany. We also explored whether traffic noise-annoyance and insomnia may play a role in this relationship.

Our results did not show an association between any of the traffic noise indicators and headaches. However, associations were observed between traffic noise and each of traffic noise-annoyance and insomnia, as well as between insomnia and headache.

The effects of traffic noise on noise-annoyance and sleep disturbance are well recognized.[2526] Association between both noise-annoyance and sleep disturbance and headaches has also been established.[121427] Noise-annoyance and sleep deprivation could amplify the response to sound through stress mechanisms. However, we did not observe an association between traffic noise-annoyance and headache in this population. Reverse causation could also partially play a role in the observed association between insomnia and headache, as headache sufferers may have difficulty falling asleep or maintaining sleep during a headache.

The PORs of suffering headaches at higher noise levels did not notably differ between those who expressed traffic noise-annoyance or insomnia and those who did not. It has been suggested that noise has an independent and partially sub-cognitive effect on health which could last long after psychological and sleep adaptation to noise.[2628] This may explain the lack of difference in headache occurrence in the stratified analysis for this population with years of exposure behind them.

Strengths

This study includes a large sample of randomly selected participants from three different cities and uses address-based high-quality noise exposure modelling. We were also able to correct for room use and orientation, ventilation habits and window-type for indoor noise measures with the aim of yielding levels of exposure closer to reality. Especially since a weakness of Lden at the most exposed façade is not accounting for noise reduction measures taken by those at the highest exposure. This was evident in our study population as those who mostly kept their windows closed during summer had slightly higher Lden and Lnight exposure than those who usually kept them open.

Limitations

Recall of average number of headaches in the past three months could prove difficult but is unlikely to be differential by exposure and thus the error is expected to be random. Prospective studies with headache diaries would be helpful to overcome this in future studies.

Due to the relatively older age of our sample, it is possible that the observed effect is attenuated by more prominent health problems and age-related hearing loss. Age was consistently found to be negatively associated with the outcome in the adjusted analysis.

Noise data, along with noise-annoyance and participant characteristics were collected at an earlier time point (about 10 years) than the headache data. However, we do not expect traffic noise to change much over this period. This time lag could also have an added benefit of examining the effect after any adaptation had long taken place, which is more relevant for the larger part of the population. Those with higher exposure at baseline address were slightly more likely to move during the study period (3.8% more than those in lower exposure), which could have led to some exposure misclassification if they moved to a quieter area. However, excluding those who moved did not change the estimates.

Exclusion from the analysis was largely due to drop-out or death (96% of those excluded) and not to missing noise or headache frequency values. Thus, the differences in characteristics between those who were included and excluded (Supplementary Table 1) mostly reflect the older age and worse health status of the latter group. Nevertheless, excluded individuals also had less years of education and higher noise exposure on average. This could have caused some loss of participants with higher exposure and introduced selection bias affecting the generalizability of the results. The generally lower educational level of those with higher noise exposure (Supplementary Table 2) may reflect higher traffic noise exposure for those with a lower socioeconomic status who may be more likely to live in lower-demand dwellings, located on main roads or not have the choice of moving to a quieter place. The possible higher drop-out of participants of lower socioeconomic status and higher noise exposure could result in leaving out a more vulnerable subset.

Supplementary table 1

Characteristics of the analysis population with Lden exposure >=55 dB(A) and <55 db(A) at baseline

All	≤55 dB(A)		<55 dB(A)
	N	(%)	N	(%)
	1184	39.1	1848	60.9
Gender (missing = 0)
Male	572	48.3	918	49.68
Years of formal education (missing = 4)
<10	106	9.0	148	8.0
11–13	692	58.5	974	52.8
14–17	258	21.9	461	25.0
>=18	126	10.7	263	14.3
Employment status (missing = 2)
Employed	551	46.6	879	47.6
Inactive or housewife	167	14.1	271	14.7
Retired	371	31.4	599	32.4
Unemployed	93	7.9	99	5.4
Comorbidities (missing = 0)
0	498	42.1	831	45.0
1	403	34.0	631	34.2
>=2	283	23.9	386	20.9
Sports (missing = 0)
Yes	689	58.2	1118	60.5
Smoking (missing = 0)
Never smoker	496	41.9	820	44.4
Former smoker	429	36.2	681	36.9
Current smoker	259	21.9	347	18.8
City of residence (missing = 0)
Müllheim	441	37.3	696	37.7
Bochum	330	27.9	530	28.7
Essen	413	34.9	622	33.7
Moved during study period (missing = 0)
Yes	214	18.1	264	14.3
Noise-annoyance day (missing = 311)
Not at all	382	35.3	1101	67.2
Somewhat annoyed	387	34.9	408	24.9
Moderately annoyed	224	20.7	107	6.5
Strongly annoyed	98	9.1	23	1.4
Noise-annoyance night (missing = 309)
Not at all	567	61.2	1364	82.7
Somewhat annoyed	261	24.3	237	14.4
Moderately annoyed	104	9.7	38	2.3
Strongly annoyed	51	4.7	11	0.7
Insomnia (missing = 14)
Yes	1,040	88.29	1,654	89.89
	mean	SD	Mean	SD
Body mass index (missing = 13)	27.65	4.51	27.45	7.23
Age (missing = 0)	58.09	7.29	58.46	7.36

Other sources of noise at home (like neighbors) as well as noise exposure outside the home (e.g. at work) have not been taken into consideration. It is difficult to merge exposure from different noise sources into one meaningful measure, as not only noise level, but also character, information content of the noise, duration, timing and intermittency play a role in both the direct and indirect (through annoyance) effect on an individual. However, traffic noise seems to be the dominant source of annoyance in the general population, and apart from being the place where people spend most of their time, home is expected to be a place to rest and an environment which is within one’s own control.[16] Noise exposure while commuting or in the work place and during leisure time could all contribute to the burden and could be examined in future studies.Despite the large sample size, two thirds of the participants reported no longer having any occasional headaches (one quarter never had a headache), and only 1.9% had eight or more headache days a month. There were also relatively few participants who were exposed to higher noise levels. Due to this skewed distribution of headache frequencies and small numbers at higher levels of both exposures and outcome, it was not feasible to examine headache frequency as a continuous variable or with multiple categories.

Air pollution which has been addressed in some previous studies was not included as a potential confounder here as it was not associated with our outcome in this population. Past studies showed that noise and air pollution have largely independent effects on other health outcomes with different pathophysiological mechanisms (neuroendocrine versus inflammatory).[2229303132]

As this is an exploratory analysis using secondary data without considering sample size calculations, it serves a descriptive purpose which is useful for hypothesis generation. It also presents some valuable information about the nature of traffic noise exposure and headache frequency distributions in this population, and potential confounders to be sought in future studies.

CONCLUSION

Our results did not find an association between headaches and higher traffic-noise exposure at home but found an association between traffic noise exposure and traffic noise-annoyance and insomnia, and between expressing traffic noise-annoyance or insomnia and headache. Due to the ubiquitous exposure to noise and the high burden of headaches, studies in different populations are warranted, possibly focusing on those with chronic headaches, using headache diaries, and where more detailed analysis can be carried out to look for a dose-response effect.

Financial support and sponsorship

Nil.

Conflicts of interest

There are no conflicts of interest.