Environmental Noise Exposure and the Risk of Sleep Disturbance: A Systematic Review and Meta-Analysis

Abstract

Objective:

To systematically evaluate the association between different types of environmental noise, particularly traffic noise, and sleep disturbances; to quantify the effect size and to explore key factors contributing to heterogeneity, thereby providing evidence to inform noise control and public health interventions.

Methods:

Publications from 2005 to 2025 were systematically searched in PubMed, Embase, Cochrane Library and Web of Science. A total of 68 observational studies were included in accordance with preferred reporting items for systematic reviews and meta-analyses (PRISMA) guidelines. Random-effects models were applied to pool effect sizes (correlation coefficients, r). Subgroup analyses, meta-regression, sensitivity analyses and Egger’s test were conducted to explore sources of heterogeneity and publication bias. The certainty of evidence was assessed using the grading of recommendations assessment, development and evaluation (GRADE) framework.

Results:

Exposure to traffic noise was significantly associated with sleep disturbances. Source-specific analyses indicated that aircraft noise (r = 0.26) and railway noise (r = 0.21) were associated with larger effect sizes than road traffic noise (r = 0.15). Subgroup and meta-regression analyses identified outcome assessment methods, study design and geographic region as the main contributors to between-study heterogeneity. Evidence related to non-traffic noise was limited; however, available data suggested that community and hospital noise may also adversely affect sleep. Sensitivity analyses and assessments of publication bias indicated that the pooled estimates were robust. According to the GRADE framework, the overall certainty of evidence was rated as high.

Conclusions:

Environmental noise, particularly traffic noise, significantly increases the risk of sleep disturbances. Type-specific effects were observed, and a dose–response relationship was also identified. These findings support the need for further reductions in nighttime traffic noise exposure and strengthened residential noise control to promote population sleep health.

KEY MESSAGES

• (1)Traffic noise is significantly associated with sleep disturbances, with aircraft noise (r = 0.26) exerting stronger effects than railway (r = 0.21) and road traffic noise (r = 0.15).
• (2)Environmental noise shows linear dose–response with sleep issues; a 5 dB increase raises the risk by 18–19%.
• (3)GRADE certainty is moderate to high with stable sensitivity analysis results.

INTRODUCTION

Sleep is a highly conserved physiological process essential for energy metabolism, neural plasticity, immune regulation and cardiovascular stability, forming the biological foundation of alertness, emotional regulation and cognitive performance.[1,2] Both experimental and epidemiological studies have demonstrated that insufficient or fragmented sleep disrupts endocrine homeostasis, glucose metabolism and inflammatory pathways, thereby increasing the risk of obesity, type 2 diabetes, hypertension, cardiovascular disease, depression and cognitive impairment.[3,4]

Sleep disorders encompass abnormalities in sleep duration, continuity, quality, timing or circadian alignment and impose a substantial burden on individuals and society. Individuals with sleep disturbances have a markedly increased risk of occupational errors, traffic accidents and reduced work performance, with productivity losses estimated to account for approximately 1–2% of gross domestic product in high-income countries.[5,6] Among the diverse determinants of sleep health, environmental exposures play a critical role, and environmental noise has emerged as one of the most pervasive and modifiable risk factors.

Traffic-related noise exposure is widespread globally. In Europe alone, more than 125 million people are exposed to harmful levels of road traffic noise, with over 40 million experiencing adverse health effects attributable to environmental noise exposure.[7] Similar exposure patterns have been reported in North America and rapidly urbanising regions in Asia, reflecting the global nature of noise pollution.[8] Consequently, environmental noise is now recognised as the second largest environmental health risk in Europe, following air pollution.

Chronic exposure to environmental noise affects both auditory and non-auditory systems, exerting profound effects on sleep physiology, autonomic regulation and cardiometabolic function.[1,9] Evidence from a systematic review indicates that nocturnal noise exposure can increase sleep onset latency, induce frequent awakenings, reduce slow-wave and rapid eye movement sleep and trigger acute stress responses characterised by elevated cortisol and sympathetic activation.[10] Long-term nighttime noise exposure has been associated with sustained increases in blood pressure, endothelial dysfunction, insulin resistance and metabolic dysregulation, providing a plausible biological pathway linking noise-induced sleep disturbance to cardiovascular and metabolic diseases.[7,11]

Recognising these health impacts, the World Health Organization has identified sleep disturbance as the single largest contributor to the burden of disease attributable to environmental noise in Europe and has issued evidence-based guideline values for nighttime noise exposure to protect sleep health.[12,13] Nevertheless, epidemiological evidence regarding the association between environmental noise and sleep disturbance remains heterogeneous. Substantial variability exists in exposure assessment methods, outcome definitions, study designs and population characteristics, which hampers direct comparison across studies and limits causal inference.[8,14]

Moreover, the existing literature is disproportionately focused on traffic-related noise, whereas evidence on non-traffic noise sources, such as community noise, neighbourhood noise and hospital noise, remains sparse and fragmented.[15] Importantly, few studies have quantitatively characterised the dose–response relationship between incremental increases in environmental noise and the risk of sleep disturbance, leaving critical knowledge gaps for the formulation of evidence-based exposure limits and public health interventions.[1]

To address these limitations, the present study systematically synthesises epidemiological evidence on environmental noise exposure and sleep disturbance through a comprehensive meta-analysis and dose–response modelling. By treating noise exposure as a continuous variable and examining type-specific effects across different noise sources, this study aims to clarify the magnitude, consistency and heterogeneity of the associations and to provide robust evidence to inform noise regulation and sleep health protection strategies.

METHODS

Eligibility Criteria

This review primarily included observational studies, including cross-sectional, cohort and case–control studies, which are appropriate for examining the association between real-world environmental noise exposure and sleep outcomes in general populations. In contrast, laboratory trials, randomised controlled trials (RCTs), and animal studies were excluded because their controlled settings differ from actual environmental exposures.

The study population comprised community-dwelling adults exposed to real-life environmental noise, defined as sound from outdoor sources beyond individual control, such as road traffic noise, railway noise and aircraft noise (traffic noise), as well as non-traffic noise (e.g., industrial, construction, wind turbine, recreational or neighbourhood noise). Noise exposure was categorised into traffic and non-traffic types for subgroup comparisons.

Noise exposure was assessed using standardised acoustic indicators (e.g., Lden, Lnight) or self-reported perceptual scales. Traffic noise was the primary exposure of interest, with non-traffic noise considered a supplementary factor. Exposure levels were classified as low, medium, high or treated as a continuous variable. The control group consisted of participants with the lowest exposure levels or those reporting being “not disturbed by noise.”

Sleep outcomes included subjective indicators (e.g., Pittsburgh Sleep Quality Index [PSQI], difficulty falling asleep and poor sleep quality) and objective measures (e.g., sleep efficiency, sleep duration and wake after sleep onset [WASO].

Studies meeting any of the following exclusion criteria were not included: (1) studies focusing on specific occupational populations (e.g., pilots or factory workers) with noise exposure primarily originating from the workplace; (2) studies involving exclusively indoor noise unrelated to building structure transmission (e.g., household appliances or cohabitant activity noise); (3) non-original publications, such as reviews or meta-analyses; (4) randomised controlled trials, laboratory experiments or animal studies and (5) studies that did not report predefined sleep outcomes or lacked essential statistical data. Additionally, publications for which the full text could not be retrieved were excluded from the final analysis.

Data Extraction and Variables

For all studies meeting the inclusion criteria, information on study characteristics, exposure features, outcome indicators and statistical results was extracted to create a standardised summary database. Extracted study information included author, year, country/region, study design and sample size. Study populations were classified as community-dwelling adults, and noise type was categorised as traffic noise or non-traffic noise.

Exposure characteristics included the noise assessment method, exposure grouping and exposure indicators (Lden, Lnight, LAeq). For dose–response studies, noise levels corresponding to each exposure category were recorded to enable continuous exposure–response modelling. Sleep outcomes were categorised as subjective (e.g., PSQI, difficulty falling asleep and night awakenings) or objective (e.g., sleep efficiency, sleep duration and WASO).

Effect estimates (odds ratios [OR], risk ratios [RR], β coefficients or standardised mean differences), 95% confidence intervals (CIs) and adjusted confounders were recorded. The summary table provides a comparable framework for effect size synthesis, heterogeneity assessment and dose–response analyses, with only a subset of the included studies presented.

Quality Assessment

Given variations in study designs, appropriate tools were applied to assess study quality and risk of bias. Cross-sectional studies were evaluated using the Agency for Healthcare Research and Quality (AHRQ) checklist, including 11 items covering design soundness, data collection, bias control, statistical analysis and reporting quality. Cohort and case–control studies were assessed using the Newcastle Ottawa Scale (NOS), which employs a star-based system rather than a cumulative scoring scale and covers three core domains: Selection, Comparability and Outcome/Exposure. For the quantitative synthesis of methodological quality across studies, star ratings were standardised by summing the stars awarded across domains to generate a total score ranging from 0 to 9. The total scores derived from NOS star ratings were summarised in tables to present overall variation.[16,17]

In addition, the risk of bias In non-randomized studies of exposures (ROBINS-E) tool was used to assess bias across seven domains: confounding, participant selection, exposure classification, deviations from intended exposures, missing data, outcome measurement and selective reporting.[18] Assessment of confounding bias focused on control of age, sex, body mass index and socioeconomic status. Participant selection addressed sample representativeness, whereas exposure classification and deviations from intended exposures assessed the accuracy of noise measurement. Bias due to missing data was evaluated based on data completeness and handling. Outcome measurement bias considered the objectivity of assessment, with polysomnography or actigraphy indicating a lower risk of bias. Selective reporting evaluated consistency with pre-specified protocols. Overall risk of bias was determined by the domain with the highest risk. Results were summarised in tables to illustrate both within- and between-study variation in study quality and risk of bias.

Statistical Analysis

A two-stage statistical analysis framework was employed to harmonise noise exposure indicators and effect-size metrics across studies, ensuring data comparability for meta-analysis. All analyses were performed using random-effects models (restricted maximum likelihood method), with pooled effect estimates reported alongside 95% CIs. The correlation coefficient (r) was selected as the unified effect size because of its flexibility in conversion from diverse metrics (e.g., OR, RR, β), intuitive interpretation of association strength and compatibility with dose–response modelling. Potential conversion bias was minimised using established formulas and sensitivity analyses.

Noise exposure levels reported as 24-h average sound pressure levels (LAeq) or day–evening–night levels (Lden) were converted to equivalent nighttime levels (Lnight) using standard acoustic adjustment equations. For continuous exposures, effect estimates were scaled to represent a 5 dB increase using appropriate scaling factors applied to ratio measures (OR, RR, PR) and regression coefficients (β). For categorical exposures, exposure levels were represented by interval midpoints without further adjustment.

All effect estimates were transformed into correlation coefficients (r) while preserving directionality. For ORs, natural logarithm transformation was applied, followed by conversion to r using variance-stabilising transformation. Relative risks and prevalence ratios underwent similar logarithmic transformations. Regression coefficients were directly interpreted as r. Variance estimates were derived from reported CIs or conservatively approximated. Studies lacking essential data were excluded. The specific classification criteria for each subgroup variable were as follows. Geographic region was categorised as Europe, Asia, North America or other regions (including Australia and South America). Sleep outcome assessment was classified as subjective (e.g., PSQI or self-reported sleep quality), objective (e.g., actigraphy or polysomnography) or mixed (both subjective and objective measures). Study design was grouped as cross-sectional or longitudinal. Sample size was dichotomised as large (≥4141 participants) or small (<4141 participants). Study quality was assessed using the NOS or the AHRQ checklist and classified as high (score ≥10) or low to moderate (score <10).

Heterogeneity was assessed using the I² statistic, with values >50% indicating substantial between-study variability. Meta-regression analyses examined potential moderators, including noise type, geographic region, publication year, study design and outcome assessment type. Subgroup analyses further explored variation by region, sleep measurement method, study design, sample size and quality score, with leave-one-out sensitivity checks to assess the robustness of the findings.

Sensitivity analyses evaluated the influence of individual studies and the impact of excluding studies with low quality or small samples. Publication bias was assessed using Egger’s regression test by examining the relationship between effect sizes and their standard errors.

Dose–response relationships were analysed using two complementary approaches: discrete dose-point analysis for studies with three or more exposure categories, and linear meta-analysis per 5 dB increase for studies reporting linear estimates. Weighted linear models were primarily used owing to the limited number of exposure categories, while nonlinear trends were explored using restricted cubic spline models when data permitted. A one-stage mixed-effects model was used to express the effect as a log OR per 1 dB increase.

The certainty of evidence was evaluated using the GRADE framework, considering study design, risk of bias, inconsistency, indirectness, imprecision and publication bias. All analyses were conducted using R software (version 4.5.1; R Foundation for Statistical Computing, Vienna, Austria), specifically utilising the metafor and dosresmeta packages.

RESULTS

Literature Search Results

A systematic literature search was conducted to identify research evidence on the association between environmental noise exposure and sleep outcomes. The search strategy incorporated keywords related to environmental noise (e.g., environmental noise and traffic noise), sleep-related outcomes and measurements (e.g., sleep disturbance, sleep quality, and PSQI), and epidemiological study designs (e.g., epidemiological studies and cohort studies). Studies focusing on occupational noise exposure, animal experiments or military populations were excluded. The search period covered publications from 2005 to 2025 and yielded a total of 673 potentially relevant records (PubMed, n = 112; Embase, n = 357; Cochrane Library, n = 8; Web of Science, n = 196). All retrieved records were imported into EndNote X9 (Clarivate, Philadelphia, PA, USA) for deduplication and screening according to predefined criteria.

Following title and abstract screening, 424 records remained after duplicate removal. Full text assessment excluded studies for occupational only populations or workplace noise exposure (n = 14), indoor personal or unclear noise sources (n = 58), non-original articles (n = 48), RCTs, laboratory or animal studies (n = 19), missing target sleep outcomes (n = 171) or insufficient data (n = 19). Ultimately, 68 studies were included in the meta-analysis. The study selection process is illustrated in the PRISMA flow diagram [Figure 1], which outlines the systematic and transparent identification procedure. This systematic review and meta-analysis was conducted in accordance with the PRISMA guidelines and was registered in PROSPERO (Registration No.: CRD 420251159987).[19] Detailed information on study identification, inclusion and exclusion is presented in Figure 1.

Figure 1.

PRISMA flow diagram of study identification, screening, eligibility and inclusion.

Study Characteristics

Key characteristics of the included studies (e.g., noise type, sample size and effect estimates) are summarised in Table 1, which covers 68 studies contributing 152 effect sizes across traffic noise (e.g., road traffic, railway, aircraft), non-traffic noise and mixed noise categories. Exposure levels varied widely, with continuous data standardised to a ‘per 5 dB increase’ scale and discrete categories ranging from 37.5 to 75 dB. Table 1 presents a subset of the extracted references data. Road traffic noise was the most commonly studied exposure (54.6%), followed by aircraft noise (20.4%) and railway noise (11.8%). Sleep outcomes included difficulty falling asleep, difficulty maintaining sleep, sleep disturbance and daytime impairment, with sleep disturbance being the most frequently assessed (38.2%). After standardisation, correlation coefficients (r) ranged from −1.8 to 0.94, with a mean of 0.21 ± 0.31, indicating an overall positive association between noise exposure and sleep disturbance, alongside considerable variability across studies.

Table 1.

Summary of included studies on environmental noise and sleep outcomes

Noise	ID	Author_Year	Noise_Type	Sample_Size		Sleep_Indicator	r	Var r	SE r	Quality_Score	Area	Study design	Unified_noise_exposure
Traffic noise	1	B. Jakovljević et al. (2006)[20]	Road traffic noise	310		Difficulty falling asleep	0.263	0.009	0.095	10	Serbia	Cross-sectional studies	Every increase of 5
						Windows closed during sleep	0.583	0.004	0.064	10	Serbia	Cross-sectional studies	Every increase of 5
						Difficulty maintaining sleep/night awakenings (NA)	0.254	0.005	0.070	10	Serbia	Cross-sectional studies	Every increase of 5
						Overall sleep disturbance/poor sleep quality	0.289	0.015	0.121	10	Serbia	Cross-sectional studies	Every increase of 5
						Post sleep fatigue	0.245	0.011	0.104	10	Serbia	Cross-sectional studies	Every increase of 5
						Sleeping pill use	0.052	0.036	0.190	10	Serbia	Cross-sectional studies	Every increase of 5
	2	Y. Aydin and M. Kaltenbach (2007)[21]	Aircraft noise	53		Overall sleep disturbance/poor sleep quality	−0.220	0.017	0.132	11	Germany	Cohort study	NA
	3	G. M. Aasvang et al. (2008)[22]	Railway noise	1349		Difficulty falling asleep	0.401	0.221	0.471	11	Norway	Cross-sectional studies	42
						Difficulty maintaining sleep/night awakenings (NA)	−0.178	0.066	0.257	11	Norway	Cross-sectional studies	47
	5	L.Stošić et al. (2009)[23]	Road traffic noise	911		Difficulty falling asleep	0.664	0.001	0.038	11	Serbia	Cross-sectional studies	45
						Difficulty maintaining sleep/night awakenings (NA)	0.659	0.003	0.051	11	Serbia	Cross-sectional studies	45
						Overall sleep disturbance / poor sleep quality	0.540	0.005	0.069	11	Serbia	Cross-sectional studies	45
						Sleeping pill use	0.308	0.006	0.079	11	Serbia	Cross-sectional studies	45
	6	A. Fyhri and G. M. Aasvang (2010)[24]		2786		Sleep problems	0.940	0.000	0.002	10	Norway	Cross-sectional studies	Every increase of 5
	7	J. Hong et al. (2010)[25]	Railway noise	1160		Sleep fragmentation	0.060	0.001	0.029	10	South Korea	Cross-sectional studies	Every increase of 5
			Road traffic noise			Sleep fragmentation	0.040	0.001	0.029	10	South Korea	Cross-sectional studies	Every increase of 5
	9	G. Bluhm and C. Eriksson (2011)[26]	Aircraft noise	2959		Short sleep duration	0.143	0.006	0.080	9	Sweden	Cross-sectional studies + cohort study	Every increase of 5
			Road traffic noise	13,557		Short sleep duration	0.182	0.003	0.058	9	Sweden	Cross-sectional studies + cohort study	Every increase of 5
	11	T. Van Renterghem and D. Botteldooren (2012)[27]		100		Short sleep duration	−0.422	0.048	0.218	9	Belgium	Cross-sectional studies	Every increase of 5
						Difficulty maintaining sleep/night awakenings (NA)	0.326	0.010	0.101	9	Belgium	Cross-sectional studies	Every increase of 5
	13	A. H. Amundsen et al. (2013)[28]		169		Overall sleep disturbance / poor sleep quality	0.130	0.006	0.076	10	Norway	Cross-sectional studies	NA
	14	C. M. T. Tiesler et al. (2013)[29]	Aircraft noise	872		Overall sleep	0.306	0.014	0.119	10	Germany	Cross-sectional studies	Every increase of 5
				287		Difficulty falling asleep	0.348	0.015	0.122	10	Germany	Cross-sectional studies	Every increase of 5
	16	S. Fidell et al. (2013)[30]		874		Difficulty maintaining sleep/night awakenings (NA)	0.010	0.001	0.034	10	America	Cross-sectional studies	75
	17	E. M. Salomons et al. (2014)[31]	Railway noise	1000		Overall sleep disturbance / poor sleep quality	0.200	0.001	0.030	10	Netherlands	Cross-sectional studies	Every increase of 5
	19	G. Licitra and E. Ascari (2014)[32]	Road traffic noise			Overall sleep disturbance / poor sleep quality	−0.075	0.002	0.044	9	Italy	Cross-sectional studies	NA
	20	M Röösli et al. (2014)[33]		1266		Sleep duration	−0.310	0.001	0.030	10	Switzerland	Cross-sectional studies	Every increase of 5
						Overall sleep disturbance/poor sleep quality	−0.160	0.001	0.033	10	Switzerland	Cross-sectional studies	Every increase of 5
						Sleep efficiency	−0.870	0.000	0.007	10	Switzerland	Cross-sectional studies	Every increase of 5
	21	S. A. Janssen et al. (2014)[34]	Aircraft noise	414		Overall sleep disturbance/poor sleep quality	0.030	0.002	0.049	10	Netherlands	Cross-sectional studies	Every increase of 5
	22	P. Frei et al. (2014)[35]	Road traffic noise	1375		Sleep efficiency	−0.810	0.000	0.009	10	Switzerland	Cross-sectional studies	Every increase of 5
	23	M. W. Gerbase et al. (2014)[36]		1383		Sleep fragmentation	0.037	0.000	0.013	10	Switzerland	Cross-sectional studies	Every increase of 5
	24	H. Héritier et al. (2014)[37]		1375		Overall sleep disturbance/poor sleep quality	0.810	0.000	0.009	10	Switzerland	Cross-sectional studies	Every increase of 5
	25	K. Sygna et al. (2014)[38]		2898		Overall sleep disturbance/poor sleep quality	0.091	0.002	0.049	10	Norway	Cross-sectional studies	Every increase of 5
	26	T.L. Nguye et al. (2015)[39]		1287		Overall sleep disturbance/poor sleep quality	0.251	0.003	0.053	10	Vietnam	Cross-sectional studies	Every increase of 5
	27	T. Bodin et al. (2015)[40]		2612		Overall sleep disturbance / poor sleep quality	−0.136	0.004	0.061	10	Sweden	Cross-sectional studies	Every increase of 5
			Railway noise			Overall sleep disturbance/poor sleep quality	0.126	0.001	0.024	10	Sweden	Cross-sectional studies	Every increase of 5
	28	A. Chum et al. (2015)[41]	Road traffic noise	2411		Short sleep	0.232	0.000	0.009	10	Canada	Cross-sectional studies	Every increase of 5
						Sleep problems	0.391	0.002	0.048	10	Canada	Cross-sectional studies	Every increase of 5
	31	J. B. Holt et al. (2015)[42]	Aircraft noise	745868		Days of insufficient sleep	0.370	0.000	0.001	11	America	Cross-sectional studies	Every increase of 5
						Monthly sleep deprivation	−0.040	0.003	0.056	11	America	Cross-sectional studies	Every increase of 5
						Adequate sleep	−0.017	0.002	0.046	11	America	Cross-sectional studies	Every increase of 5
	32	C. Baliatsas et al. (2016)[43]	Road traffic noise	5806		Overall sleep disturbance/poor sleep quality	0.760	0.000	0.006	10	Netherlands	Cross-sectional studies	44.6
	33	K.M. Kwak et al. (2016)[44]	Aircraft noise	3308		Overall sleep disturbance/poor sleep quality	0.560	0.002	0.043	10	South Korea	Cross-sectional studies	NA
						Daytime sleepiness	0.558	0.005	0.071	10	South Korea	Cross-sectional studies	NA
	34	E. Orban et al. (2016)[45]	Road traffic noise	15,085		Overall sleep disturbance/poor sleep quality	0.057	0.001	0.030	8	Norway	Cohort study	Every increase of 5
	35	S. Perron et al. (2016)[46]		4336		Overall sleep disturbance/poor sleep quality	0.020	0.000	0.015	11	Canada	Cross-sectional studies	Every increase of 5
	36	T. Gislason et al. (2016)[47]		12,184		Daytime sleepiness	0.204	0.005	0.072	11	Northern Europe	Cross-sectional studies	NA
						Habitual snoring	0.143	0.001	0.030	11	Northern Europe	Cross-sectional studies	NA
	37	K.V. Weyde et al. (2017)[48]		3300		Overall sleep disturbance/poor sleep quality	0.234	0.010	0.100	8	Germany	Cohort study	55
	38	J. Evandt et al. (2017)[49]		12,138		Difficulty falling asleep	0.027	0.000	0.011	10	Norway	Cross-sectional studies	Every increase of 5
						Difficulty maintaining sleep/night awakenings (NA)	0.022	0.000	0.011	10	Norway	Cross-sectional studies	Every increase of 5
	40	M. Carugno et al. (2018)[50]	Aircraft noise	400		Overall sleep disturbance/poor sleep quality	0.145	0.006	0.080	9	Italy	Cross-sectional studies	NA
	44	E. O. Okokon et al. (2018)[51]	Road traffic noise	7321		Sleep medication use	0.086	0.005	0.071	11	Finland	Cross-sectional studies	39.2
	48	S. Rocha et al. (2019)[52]	Aircraft noise	268		Overall sleep disturbance/poor sleep quality	0.109	0.002	0.046	10	America	Cross-sectional studies	Every increase of 5
						Difficulty falling asleep	0.159	0.002	0.046	10	America	Cross-sectional studies	Every increase of 5
						Difficulty maintaining sleep/night awakenings (NA)	0.109	0.002	0.046	10	America	Cross-sectional studies	Every increase of 5
						Overall sleep disturbance/poor sleep quality	0.359	0.002	0.050	10	America	Cross-sectional studies	Every increase of 5
	49	X. Feng et al. (2020)[53]	Road traffic noise	1461		Sleep deprivation	-0.840	0.008	0.089	10	Germany	Cross-sectional studies	Every increase of 5
	50	E. Janson et al. (2020)[54]		12,963		Difficulty falling asleep	0.408	0.016	0.125	10	Northern Europe	Cross-sectional studies	NA
						Difficulty maintaining sleep/night awakenings (NA)	0.336	0.016	0.126	10	Northern Europe	Cross-sectional studies	NA
	51	L. Yang et al. (2020)[55]		608		Overall sleep disturbance / poor sleep quality	0.598	0.009	0.095	9	Hong Kong, China	Cross-sectional studies	NA
	52	J. Lee et al. (2021)[56]	Aircraft noise	474		Difficulty maintaining sleep/night awakenings (NA)	0.109	0.003	0.053	10	South Korea	Cross-sectional studies	NA
						Difficulty falling asleep	0.037	0.000	0.020	10	South Korea	Cross-sectional studies	NA
	53	M. Rojek et al. (2021)[57]		98		Overall sleep disturbance / poor sleep quality	0.469	0.035	0.186	9	Poland	Cross-sectional studies	NA
	54	S. Bartels et al. (2021)[58]	Road traffic noise	2191		Sleep problems	0.182	0.015	0.122	10	Sweden	Cross-sectional studies	NA
	55	T. Hanibuchi et al. (2021)[59]		4243		Overall sleep disturbance / poor sleep quality	0.024	0.000	0.022	11	Japan	Cross-sectional studies	NA
	56	W. Wojciechowska et al. (2022)[60]	Aircraft noise	149		Overall sleep disturbance / poor sleep quality	0.469	0.035	0.186	8	Poland	Cohort study	Every increase of 5
	58	C. Blume et al. (2022)[61]	Road traffic noise	144		Sleep duration	-0.020	0.007	0.084	8	Switzerland	Cohort study	40.8
	59	T. A. Gilani and M. S. Mir 2022)[62]		4525		Difficulty falling asleep	0.201	0.009	0.095	10	India	Cross-sectional studies	Every increase of 5
						Difficulty maintaining sleep/night awakenings (NA)	0.190	0.008	0.090	10	India	Cross-sectional studies	Every increase of 5
						Fatigue upon waking	0.182	0.008	0.088	10	India	Cross-sectional studies	Every increase of 5
						Sleep medication use	0.179	0.009	0.093	10	India	Cross-sectional studies	Every increase of 5
	60	G. M. Aasvang et al. (2023)[63]		100000		Overall sleep disturbance/poor sleep quality	0.186	0.000	0.018	9	Nordic countries	Ecological study	Every increase of 5
			Railway noise			Overall sleep disturbance/poor sleep quality	0.269	0.001	0.029	9	Nordic countries	Ecological study	Every increase of 5
	61	M. Bozigar et al. (2023)[64]	Aircraft noise	35226		Short sleep duration	0.057	0.000	0.019	8	America	Prospective cohort study	Every increase of 5
						Overall sleep disturbance/poor sleep quality	0.186	0.010	0.100	8	America	Prospective cohort study	Every increase of 5
	62	M. Schubert et al. (2023)[65]	Road traffic noise	5670		Overall sleep disturbance/poor sleep quality	0.278	0.002	0.050	8	Germany	Cohort study	Every increase of 5
			Railway noise			Overall sleep disturbance/poor sleep quality	0.260	0.001	0.035	8	Germany	Cohort study	Every increase of 5
			Aircraft noise			Overall sleep disturbance/poor sleep quality	0.634	0.001	0.035	8	Germany	Cohort study	Every increase of 5
	63	E. Flanagan et al. (2023)[66]	Road traffic noise	6784		Overall sleep disturbance/poor sleep quality	0.020	0.000	0.009	8	Sweden	Health impact assessment	Every increase of 5
	64	C. Jephcote et al. (2023)[67]		42738500		Overall sleep disturbance/poor sleep quality	0.257	0.002	0.047	8	UK	Ecological study	41.7
			Railway noise			Overall sleep disturbance/poor sleep quality	0.201	0.005	0.068	8	UK	Ecological study	41.7
			Aircraft noise			Overall sleep disturbance/poor sleep quality	0.389	0.005	0.068	8	UK	Ecological study	41.7
	65	M. Y. Peer et al. (2025)[68]	Road traffic noise	200		Overall sleep disturbance/poor sleep quality	0.494	0.008	0.090	10	India	Cross-sectional studies	NA
	67	S. Chapagai et al. (2025)[69]		887		Overall sleep disturbance/poor sleep quality	0.238	0.008	0.090	8	Switzerland	Cohort study	Every increase of 5
						Overall sleep disturbance/poor sleep quality	0.077	0.002	0.044	8	Switzerland	Cohort study	Every increase of 5
	72	Wicki et al. (2025)[70]	Railway noise	527		Sleep efficiency	-0.080	0.002	0.043	11	Switzerland	Cross-sectional studies	Every increase of 5
Non traffic noise	4	G. Ristovska et al. (2009) [71]	Community noise	510		Overall sleep disturbance/poor sleep quality	0.193	0.008	0.090	10	Macedonia	Cross-sectional studies	Every increase of 5
	8	K. Bolin et al. (2011)[72]	Wind turbine noise	1600		Short sleep duration	0.145	0.002	0.047	11	Sweden	Cross-sectional studies	Every increase of 5
	10	M. Brink et al. (2011)[73]	Ambient noise, church bell noise	27		Difficulty maintaining sleep/night awakenings (NA)	0.070	0.038	0.195	10	Switzerland	Cross-sectional studies	Every increase of 5
	12	M. Kim et al. (2012)[74]	Community noise	1152550		Overall sleep disturbance/poor sleep quality	0.080	0.000	0.001	9	America	Cross-sectional studies	Every increase of 5
	15	J. K. McLoone et al. (2013)[75]	Hospital ward noise (equipment noise, personnel activity noise)	114		Sleep duration	−0.210	0.002	0.044	11	Australia	Cross-sectional studies	NA
						Difficulty maintaining sleep/night awakenings (NA)	0.412	0.002	0.046	11	Australia	Cross-sectional studies	NA
						Sleep satisfaction	-0.375	0.008	0.091	11	Australia	Cross-sectional studies	NA
						Sleeping pill use	0.298	0.032	0.178	11	Australia	Cross-sectional studies	NA
	30	K. Feder et al. (2015)[76]	Wind turbine noise	1238		Overall sleep disturbance/poor sleep quality	0.469	0.011	0.106	10	Canada	Cross-sectional studies	37.5
	39	J. Park et al. (2017)[77]	Community noise	1836		Overall sleep disturbance/poor sleep quality	0.374	0.007	0.085	9	South Korea	Cross-sectional studies	NA
	41	L. J. Delaney et al. (2018)[78]	Ambient noise	144		Sleep duration	−0.800	0.035	0.187	10	Australia	Cross-sectional studies	NA
	42	A. M. Dzhambov et al. (2018)[79]	Community noise	720		Overall sleep disturbance/poor sleep quality	0.120	0.001	0.037	11	Bulgaria	Cross-sectional studies	62.35
	43	D. A. Johnson et al. (2018)[80]		1889		Short sleep	0.100	0.001	0.036	11	America	Cross-sectional studies	40.8
						Sleep efficiency	−0.200	0.001	0.023	11	America	Cross-sectional studies	NA
	45	D. M. Magdy et al. (2019)[81]		100		Overall sleep disturbance / poor sleep quality	−0.380	0.007	0.086	11	Egypt	Cross-sectional studies	NA
	46	K. E. Rudolph et al. (2019)[82]		4508		Weekday bedtime	0.480	0.000	0.011	11	America	Cross-sectional studies	55
						Weekend bedtime	0.650	0.000	0.009	11	America	Cross-sectional studies	55
	55	T. Hanibuchi et al. (2021)[59]		4243		Overall sleep disturbance/poor sleep quality	0.103	0.001	0.025	11	Japan	Cross-sectional studies	NA
	68	M. Coppello et al. (2025)[83]		54		Short sleep duration	−0.500	0.029	0.172	11	America	Cross-sectional studies	42.7
	70	D. S. Michaud et al. (2025)[84]		483		Overall sleep disturbance/poor sleep quality	−0.100	0.003	0.055	10	Morocco	Cross-sectional studies	Every increase of 5
Mixed noise	57	L. Argalasova et al. (2022)[85]	Community noise; road traffic noise; traffic noise	543		Difficulty falling asleep	0.475	0.009	0.093	10	Slovakia	Cross-sectional studies	Every increase of 5
						Difficulty maintaining sleep/night awakenings (NA)	0.334	0.013	0.114	10	Slovakia	Cross-sectional studies	Every increase of 5
						Window closing due to noise	0.503	0.006	0.078	10	Slovakia	Cross-sectional studies	Every increase of 5
						Sleep deprivation	0.457	0.016	0.126	10	Slovakia	Cross-sectional studies	Every increase of 5
						Overall sleep disturbance/poor sleep quality	0.383	0.024	0.156	10	Slovakia	Cross-sectional studies	Every increase of 5
	66	N. Sandoval-Diez et al. (2025)[86]		887		Difficulty falling asleep	0.228	0.007	0.081	11	India	Cross-sectional studies	Every increase of 5
						Difficulty maintaining sleep/night awakenings (NA)	0.190	0.008	0.087	11	India	Cross-sectional studies	Every increase of 5
						Difficulty maintaining sleep/night awakenings (NA)	0.163	0.005	0.072	11	India	Cross-sectional studies	Every increase of 5
						Fatigue upon waking	0.204	0.006	0.077	11	India	Cross-sectional studies	Every increase of 5

Note: Var(r) denotes the variance of the correlation coefficient (r), SE(r) denotes the standard error of the correlation coefficient (r).

This study evaluated associations between non-traffic noise and sleep outcomes using 24 secondary indicators. Ten analysable noise–outcome combinations met the inclusion criteria, covering community noise, hospital ward noise, wind turbine noise and general environmental noise. The evidence base was limited, with most combinations represented by a single study; therefore, a descriptive synthesis was applied. Community noise demonstrated the strongest associations with bedtime outcomes, hospital noise was associated with hypnotic use and sleep satisfaction and wind turbine noise showed weak associations with short sleep duration. Objective sleep indicators were reported in only single studies. Overall, the quality of evidence was limited, and quantitative synthesis was not feasible.

Exposure to mixed noise was examined in two studies conducted in Slovakia and India, yielding nine effect sizes. The studies by Argalasova et al.[85] and Sandoval-Diez et al.[86] included 543 and 5408 participants, respectively, achieved quality scores of 10 and 11, and employed cross-sectional designs with subjective sleep scales and standardised exposure increments of 5 units of noise increase. Both studies reported adverse associations with difficulty falling asleep, difficulty maintaining sleep and overall sleep disturbance (pooled effect size = 0.383), as well as daytime outcomes, including keeping windows closed, insufficient sleep and morning fatigue. Together, these findings provide preliminary evidence that mixed noise exposure adversely affects nocturnal and daytime sleep health.

Quality Assessment

A total of 68 studies were included and systematically evaluated using standardised quality assessment forms. Methodological quality was primarily assessed using the NOS and a modified appraisal checklist for cross-sectional studies. The modified checklist comprised 11 items addressing research definition and objectives, sample source, inclusion criteria, measurement methods for exposure and outcome, control of confounding, handling of missing data, quality assurance and statistical analysis, in addition to the NOS dimensions of selection (4 points), comparability (2 points) and outcome/exposure (3 points). Methodological quality was assessed according to these three NOS dimensions.

Cross-sectional studies comprised approximately 85%, cohort studies approximately 12% and a small proportion of ecological or health impact assessment studies approximately 3%. Most studies used objective noise modelling software (e.g., CadnaA, SoundPLAN, sonROAD, Nordic Prediction Method), and sleep outcome measures were predominantly questionnaire based, indicating methodological convergence. Risk of bias was evaluated across seven domains: confounding bias, selection bias, exposure classification bias, deviation from intended exposure, missing data bias, outcome measurement bias and reporting bias. Among the included studies, 41.2% (n = 28) were assessed as having a low risk of bias, 48.5% (n = 33) as moderate risk and 10.3% (n = 7) as serious or high risk. Overall, confounding control and exposure assessment were adequately addressed, particularly in more recent studies, which increasingly applied multivariable-adjusted models and objective noise exposure estimation methods, thereby limiting potential bias.

(1) Sample selection bias: Samples were not randomly selected or recruited from a single centre.

(2) Exposure bias: Noise exposure was based on modelled assumptions or self-reported exposure.

(3) Outcome measurement bias: Sleep outcome measurement depended solely on subjective questionnaires, without actigraphy or polysomnography validation.

(4) Missing data and reporting bias: Some studies did not describe the handling of missing data and selectively reported significant results, resulting in localised serious missing data bias and serious reporting bias.

The included studies generally demonstrated adequate methodological quality. Nevertheless, heterogeneity persisted in study design and measurement approaches, constituting an important source of variation in overall methodological rigour.

According to the GRADE framework, 68 studies were included, comprising 18 longitudinal studies and 50 cross-sectional or other observational studies. Longitudinal studies generally had larger sample sizes and higher methodological quality, whereas cross-sectional studies received lower initial ratings. Risk of bias was the primary reason for downgrading, mainly due to limited objective sleep measures, missing data and incomplete confounder control. Most studies (85%) reported positive associations, with low pooled heterogeneity (I² = 32%). Some indirectness was noted; however, precision was adequate and publication bias appeared minimal. Dose–response analyses supported upgrading the certainty of evidence. Overall, the available evidence indicates a statistically significant association between environmental noise exposure and sleep disturbance, with moderate to moderately high certainty.

Pooled Meta-Analysis Results

Relevant studies examining three types of traffic noise, aircraft noise, road traffic noise and railway noise, were pooled, and the strength of their associations with different sleep outcomes were analysed separately. The overall pooled analyses demonstrated that all three traffic noise types were significantly and positively associated with adverse sleep outcomes, with significant differences observed in effect sizes and levels of heterogeneity. Geographically, the included studies spanned 21 countries across Europe, Asia, the Americas and other regions, with notable regional variations in effect estimates, as illustrated in Figure 2. To more clearly present outcome-specific characteristics, the pooled results stratified by type of sleep outcome are reported below.

Figure 2.

Geographical distribution of meta-analysis studies.

Heterogeneity test results for different noise types and sleep outcomes are presented in Table 2. For overall sleep disturbance or poor sleep quality, all three types of traffic noise showed statistically significant positive associations. Aircraft noise demonstrated the strongest association (r = 0.26), followed by railway noise (r = 0.21) and road traffic noise (r = 0.15). These findings indicate that exposure to traffic-related noise is significantly associated with impaired overall sleep quality.

Table 2.

Heterogeneity test results

Noise type	Outcome	Number of studies	N-effects	Pooled_-r	95% CI	SE	P-value	I2	Tau2	Q	Q–P
Aircraft noise	Overall sleep disturbance / poor sleep quality	5	7	0.258	(0.084,0.432)	0.089	0.004	94.51	0.05	162.06	0.00
Aircraft noise	Difficulty falling asleep	2	2	0.220	(0.047,0.392)	0.088	0.013	52.38	0.01	2.10	0.15
Aircraft noise	Difficulty maintaining sleep / nocturnal awakenings (NA)	1	1	0.109	(0.018,0.200)	0.046	NA	NA	NA	NA	NA
Road traffic noise	Overall sleep disturbance / poor sleep quality	12	13	0.154	(0.017,0.291)	0.070	0.027	99.27	0.06	4813.44	0
Road traffic noise	Difficulty falling asleep	3	5	0.113	(0.017,0.209)	0.049	0.021	67.96	0.01	12.37	0.02
Road traffic noise	Difficulty maintaining sleep / nocturnal awakenings (NA)	4	6	0.142	(0.041,0.244)	0.052	0.006	95.34	0.01	25.17	0.00
Railway noise	Overall sleep disturbance / poor sleep quality	4	4	0.212	(0.147,0.279)	0.034	0.000	81.76	0.00	18.39	0.00
Railway noise	Difficulty falling asleep	1	1	0.586	(0.270,0.902)	0.161	NA	NA	NA	NA	NA
Railway noise	Difficulty maintaining sleep / nocturnal awakenings (NA)	1	1	0.476	(0.320,0.630)	0.080	NA	NA	NA	NA	NA

Note: CI, confidence interval; SE, standard error.

Regarding difficulty in falling asleep, positive correlations were observed for all noise types. Aircraft noise showed an association of r = 0.22, road traffic noise r = 0.11 and railway noise r = 0.59 (based on one study). The magnitude of association for sleep initiation difficulty ranked as railway noise > aircraft noise ≥ road traffic noise.

For difficulty maintaining sleep and nighttime awakenings, positive associations were observed for all noise types, with weaker effects for aircraft noise. The correlation coefficients were r = 0.11 for aircraft noise, r = 0.14 for road traffic noise and r = 0.48 for railway noise (one study). All three noise types showed consistent positive associations with sleep outcomes, with substantial heterogeneity (I² = 52.4–99.3%). Across outcomes, correlation coefficients for railway noise were generally larger than those for road traffic and aircraft noises.

Subgroup Analyses

Subgroup analyses by study design, measurement method, sample size, and study quality are presented in Table 3. In the road traffic noise subgroup, regional differences were observed across sleep outcomes. For difficulty falling asleep, studies from Eastern Europe (two studies) showed a moderate association with traffic noise (pooled r = 0.287), whereas studies from Northern Europe (two studies) reported a weaker association (pooled r = 0.208). Large-sample studies tended to report weaker effects (pooled r = 0.200) compared with small sample studies (pooled r = 0.287). For difficulty maintaining sleep or nocturnal awakenings, Eastern Europe (two studies) demonstrated a stronger association (pooled r = 0.459) than Northern Europe (two studies; pooled r = 0.141). In addition, small-sample and high-quality studies produced larger pooled effect estimates.

Table 3.

Subgroup analysis results

Noise type	Outcome	Subgroup variable	Number of studies	Pooled_-r	95% CI	P	Q	Q–P
Traffic noise	Difficulty falling asleep	Northern Europe	2	0.208	(−0.003, 0.418)	0.053	63.840	0.000
Traffic noise	Difficulty falling asleep	Eastern Europe	2	0.287	(0.011, 0.563)	0.042	87.760	0.000
Traffic noise	Difficulty falling asleep	Large sample	3	0.2	(0.046, 0.353)	0.011	66.110	0.000
Traffic noise	Difficulty falling asleep	Small sample	2	0.287	(0.011, 0.563)	0.042	87.760	0.000
Traffic noise	Difficulty maintaining sleep / nocturnal awakenings (NA)	Northern Europe	2	0.141	(−0.023, 0.306)	0.093	12.510	0.006
Traffic noise	Difficulty maintaining sleep / nocturnal awakenings (NA)	Eastern Europe	2	0.459	(0.062, 0.856)	0.023	21.750	0.000
Traffic noise	Difficulty maintaining sleep / nocturnal awakenings (NA)	Large sample	3	0.141	(0.032, 0.249)	0.011	18.520	0.002
Traffic noise	Difficulty maintaining sleep / nocturnal awakenings (NA)	Small sample	3	0.42	(0.166, 0.674)	0.001	24.880	0.000
Traffic noise	Difficulty maintaining sleep / nocturnal awakenings (NA)	High quality	5	0.244	(0.091, 0.397)	0.002	175.240	0.000
Traffic noise	Overall sleep disturbance / poor sleep quality	Subjective + Objective	2	0.161	(−0.479, 0.802)	0.621	46.600	0.000
Traffic noise	Overall sleep disturbance / poor sleep quality	Subjective measurement	19	0.232	(0.119, 0.344)	0.000	56.270	0.000
Traffic noise	Overall sleep disturbance / poor sleep quality	Cohort study	4	0.162	(0.063, 0.261)	0.001	18.420	0.001
Traffic noise	Overall sleep disturbance / poor sleep quality	Cross-sectional study	14	0.258	(0.093, 0.422)	0.002	453.490	0.000
Traffic noise	Overall sleep disturbance / poor sleep quality	Ecological study	3	0.149	(0.011, 0.287)	0.034	86.940	0.000
Traffic noise	Overall sleep disturbance / poor sleep quality	Northern Europe	6	0.061	(−0.023, 0.146)	0.153	78.900	0.000
Traffic noise	Overall sleep disturbance / poor sleep quality	Eastern Europe	2	0.434	(0.192, 0.677)	0.000	3.210	0.073
Traffic noise	Overall sleep disturbance / poor sleep quality	Western/Central Europe	8	0.271	(0.05, 0.492)	0.016	164.840	0.000
Traffic noise	Overall sleep disturbance / poor sleep quality	Asia	4	0.316	(0.108, 0.523)	0.003	67.760	0.000
Traffic noise	Overall sleep disturbance / poor sleep quality	Large sample	8	0.2	(0.025, 0.376)	0.025	720.290	0.000
Traffic noise	Overall sleep disturbance / poor sleep quality	Small sample	13	0.24	(0.094, 0.385)	0.001	180.030	0.000
Traffic noise	Overall sleep disturbance / poor sleep quality	High quality	12	0.259	(0.071, 0.448)	0.007	426.630	0.000
Traffic noise	Overall sleep disturbance / poor sleep quality	Moderate/Low quality	9	0.179	(0.077, 0.281)	0.001	156.750	0.000
Railway noise	Overall sleep disturbance / poor sleep quality	Cross-sectional study	2	0.161	(0.089, 0.233)	0.000	3.630	0.057
Railway noise	Overall sleep disturbance / poor sleep quality	Ecological study	2	0.259	(0.207, 0.311)	0.000	0.870	0.350
Railway noise	Overall sleep disturbance / poor sleep quality	Northern Europe	2	0.197	(0.057, 0.337)	0.006	14.750	0.000
Railway noise	Overall sleep disturbance / poor sleep quality	Western/Central Europe	3	0.224	(0.176, 0.271)	0.000	1.810	0.404
Railway noise	Overall sleep disturbance / poor sleep quality	Large sample	3	0.259	(0.218, 0.301)	0.000	0.870	0.646
Railway noise	Overall sleep disturbance / poor sleep quality	Small sample	2	0.161	(0.089, 0.233)	0.000	3.630	0.057
Railway noise	Overall sleep disturbance / poor sleep quality	High quality	2	0.161	(0.089, 0.233)	0.000	3.630	0.057
Railway noise	Overall sleep disturbance / poor sleep quality	Moderate/Low quality	3	0.259	(0.218, 0.301)	0.000	0.870	0.646
Aircraft noise	Difficulty maintaining sleep / nocturnal awakenings (NA)	Subjective measurement	2	0.109	(0.041, 0.177)	0.002	0.000	1.000
Aircraft noise	Difficulty maintaining sleep / nocturnal awakenings (NA)	North America	2	0.038	(−0.014, 0.09)	0.152	3.280	0.195
Aircraft noise	Overall sleep disturbance / poor sleep quality	Subjective + Objective	2	0.069	(0.002, 0.136)	0.043	5.300	0.071
Aircraft noise	Overall sleep disturbance / poor sleep quality	Subjective measurement	8	0.3	(0.15, 0.45)	0.000	137.400	0.000
Aircraft noise	Overall sleep disturbance / poor sleep quality	Cohort study	4	0.272	(−0.1, 0.644)	0.152	52.970	0.000
Aircraft noise	Overall sleep disturbance / poor sleep quality	Cross-sectional study	5	0.243	(0.105, 0.38)	0.001	101.750	0.000
Aircraft noise	Overall sleep disturbance / poor sleep quality	North America	2	0.221	(0.06, 0.381)	0.007	13.620	0.001
Aircraft noise	Overall sleep disturbance / poor sleep quality	Eastern Europe	2	0.469	(0.211, 0.727)	0.000	0.000	1.000
Aircraft noise	Overall sleep disturbance / poor sleep quality	Western/Central Europe	5	0.203	(0.002, 0.403)	0.047	162.220	0.000
Aircraft noise	Overall sleep disturbance / poor sleep quality	Large sample	3	0.416	(0.162, 0.67)	0.001	24.550	0.000
Aircraft noise	Overall sleep disturbance / poor sleep quality	Small sample	7	0.219	(0.076, 0.363)	0.003	115.360	0.000
Aircraft noise	Overall sleep disturbance / poor sleep quality	High quality	4	0.165	(−0.044, 0.375)	0.122	110.520	0.000
Aircraft noise	Overall sleep disturbance / poor sleep quality	Moderate/Low quality	6	0.365	(0.217, 0.513)	0.000	52.660	0.000

Note: CI, confidence interval.

Regarding overall sleep disturbance, clear regional variation was evident. Eastern Europe (two studies) showed the strongest association (pooled r = 0.434), followed by Asia (four studies; pooled r = 0.316) and Western/Central Europe (eight studies; pooled r = 0.271), whereas Northern Europe (six studies) showed the weakest association (pooled r = 0.061). Studies using subjective sleep assessments (19 studies; pooled r = 0.232) reported larger effect sizes than those combining subjective and objective measures (two studies; pooled r = 0.161). Cohort and cross-sectional designs, sample size, and study quality further influenced effect estimates.

For railway noise, significant moderators included study design, region, sample size, and quality grade. Western/Central Europe (3 studies) and Northern Europe (2 studies) showed pooled effects of 0.224 and 0.197, respectively. Large sample and moderate/low quality studies reported slightly stronger effects than small sample and high-quality studies. Compared with road traffic noise, railway noise exhibited more dispersed effect estimates, likely reflecting its intermittent exposure pattern and lower habituation potential.

Table 4 reveals significant regional variations in the association between traffic noise and sleep disturbances across Europe, Asia, the Americas, and other regions. In the road traffic noise subgroup, marked regional differences were observed across multiple sleep indicators. For difficulty falling asleep, Europe (three studies) showed a weak but statistically significant association with road traffic noise, accompanied by extreme heterogeneity. Regarding overall sleep disturbance, the ‘Others’ region (six studies) demonstrated the strongest association, also with extreme heterogeneity. Europe (12 studies) followed, with a pooled effect size of 0.512 and extreme heterogeneity, whereas Asia (two studies) showed the weakest association (pooled r = 0.051), again with extreme heterogeneity. For difficulty maintaining sleep, the ‘Others’ region (two studies) exhibited a stronger association (pooled r = 0.331) with no significant heterogeneity, whereas Europe (three studies) showed a weaker association (pooled r = 0.044) and extreme heterogeneity.

Table 4.

Regional subgroup analyses results

Noise type	Sleep indicator	Region	Number of studies	Pooled_-r	95% CI	Q	Q–P
Road traffic noise	Difficulty Falling Asleep	Europe	3	0.076	(0.056, 0.095)	262.520	0.000
Road traffic noise	Overall Sleep Disturbance/Sleep Quality Impairment	Others	6	0.601	(0.586, 0.617)	1779.690	0.000
Road traffic noise	Overall Sleep Disturbance/Sleep Quality Impairment	Europe	12	0.512	(0.503, 0.521)	5938.280	0.000
Road traffic noise	Overall Sleep Disturbance/Sleep Quality Impairment	Asia	2	0.051	(0.009, 0.092)	25.640	0.000
Road traffic noise	Difficulty Maintaining Sleep/Nocturnal Awakenings (NA)	Others	2	0.331	(0.201, 0.462)	0.000	0.998
Road traffic noise	Difficulty Maintaining Sleep/Nocturnal Awakenings (NA)	Europe	3	0.044	(0.029, 0.060)	159.790	0.000
Railway noise	Overall Sleep Disturbance/Sleep Quality Impairment	Europe	4	0.179	(0.148, 0.211)	10.860	0.013
Aircraft noise	Overall Sleep Disturbance/Sleep Quality Impairment	Others	2	0.469	(0.211, 0.727)	0.000	1.000
Aircraft noise	Overall Sleep Disturbance/Sleep Quality Impairment	Europe	5	0.315	(0.273, 0.356)	162.220	0.000
Aircraft noise	Overall Sleep Disturbance/Sleep Quality Impairment	Americas	2	0.221	(0.158, 0.284)	13.620	0.001
Aircraft noise	Difficulty Maintaining Sleep/Nocturnal Awakenings (NA)	Americas	2	0.035	(-0.007, 0.076)	3.280	0.194

Note: CI, confidence interval.

In the railway noise subgroup, data were available only from Europe (four studies), which showed a moderate association with overall sleep disturbance and significant heterogeneity.

In the aircraft noise subgroup, clear regional variations were evident. The ‘Others’ region (two studies) demonstrated the strongest association with overall sleep disturbance (pooled r = 0.469) and no significant heterogeneity. Europe (five studies) showed a moderate association (pooled r = 0.315) with extreme heterogeneity, while the Americas (two studies) exhibited a pooled effect size of 0.221 with extreme heterogeneity. For nocturnal awakenings, the Americas (two studies) showed no significant association (pooled r = 0.035) and low heterogeneity.

Heterogeneity and Robustness Assessment

A total of 148 independent effect sizes were included in the mixed-effects meta-regression analysis [Table 5]. Substantial residual heterogeneity persisted after adjustment for covariates. The strongest moderators were the type of sleep outcome measurement and geographic region. Studies using subjective sleep measures produced significantly larger effect sizes than those using objective measures, whereas studies combining both did not differ significantly. Regional differences were also observed, with studies conducted in Australia showing lower effect sizes, while estimates from other regions did not differ significantly.

Table 5.

Meta-regression results

Variable	Estimate	SE	z-value	P-value	95% CI
Intercept	−0.10	0.13	−0.72	0.474	(−0.36, 0.17)
Outcome: Subjective	0.29	0.12	2.39	0.017	(0.05, 0.53)
Region: Australia	−0.64	0.16	−3.99	<0.001	(−0.95, −0.32)

Note: CI, confidence interval.

Sensitivity analyses confirmed the robustness of the findings. Exclusion of low- or moderate-quality studies, as well as small sample studies, resulted in only minor changes in pooled effect sizes. The overall conclusions remained consistent across sleep outcomes, indicating good reliability of main findings.

Publication bias was assessed using funnel plots, Egger’s test and the trim-and-fill method.

As presented in Table 6, for overall sleep disturbance or poor sleep quality, the funnel plot was symmetrical, Egger’s test was non-significant and no missing studies were identified, indicating no evidence of publication bias. For difficulty falling asleep and nighttime awakenings, slight asymmetry was observed, and Egger’s tests were significant, suggesting potential publication bias. However, the number of studies was limited, and trim-and-fill adjustments did not substantially alter the direction or magnitude of the pooled effects.

Table 6.

Bias assessment results

Outcome	Number of studies	Egger-p	Publication-bias	Trim-fill-missing
Overall sleep disturbance/poor sleep quality	20	0.216	No	0
Difficulty falling asleep	6	0.002	Yes	2
Difficulty maintaining sleep/nighttime awakening	7	0.006	Yes	0

Heterogeneity was largely driven by differences in outcome measurement and regional context. Sensitivity analyses indicated that the pooled estimates were stable, and no evidence of publication bias was detected for overall sleep disturbance, supporting the credibility of the summary effect. In contrast, estimates for difficulty falling asleep and night awakenings appeared more vulnerable to small-study effects and potential selective reporting.

Dose–Response Relationship

This study synthesised evidence on the association between traffic noise and sleep health to characterise the dose–response relationship. Key results are presented in Table 7.

Table 7.

Dose–response model results

Outcome	Number of studies	Best model discrete	Pooled r per 5 DB	95% CI	I2 (%)	τ2	Q statistic (P-value)	Illustrate
Overall sleep disturbance/poor sleep quality	20	Linear	0.177	(0.095, 0.260)	99.3	0.0445	Q = 6717.7, P < 0.001	Extremely high heterogeneity
Difficulty falling asleep	6	Linear	0.18	(0.079, 0.282])	82.2	0.0146	Q = 37.18, P < 0.001	High heterogeneity
Difficulty maintaining sleep/night awakenings (NA)	7	Linear	0.176	(0.074, 0.279)	95.7	0.0187	Q = 58.02, P < 0.001	Extremely high heterogeneity

Overall sleep disturbance/poor sleep quality: Twenty studies were included in the ‘per 5 dB increase’ analysis. The pooled estimate was log (OR) = 0.177, corresponding to an approximately 19% increase in risk per 5 dB increase in traffic noise exposure. Between-study heterogeneity was high.

Difficulty falling asleep: Six studies were included, yielding a pooled effect of log (OR) = 0.180, which represents an approximately 18% increase in risk per 5 dB increase. Heterogeneity remained considerable.

Difficulty maintaining sleep/night awakenings: Seven studies were included. The pooled estimate was log (OR) = 0.176, indicating an approximately 19% increase in risk per 5 dB. For all sleep outcomes, a significant linear trend was observed for risk per 5 dB increase in noise exposure.

Across all sleep outcomes, a significant linear trend was observed, demonstrating increasing risk with each 5 dB increment in traffic noise exposure. These findings are supported by the model estimates presented in Table 7 and the corresponding forest plots (Figures 345), which consistently confirm the direction and magnitude of the associations.

Figure 3.

Forest plot of overall sleep disturbance.

Figure 4.

Forest plot of difficulty falling asleep.

Figure 5.

Forest plot of difficulty maintaining sleep.

In addition to the pooled linear trend analysis, dose–response models were further constructed to explore potential non-linear associations and exposure thresholds.

Figure 6 illustrates the dose–response relationships between environmental noise exposure and various sleep outcomes, comparing linear and nonlinear models.

Figure 6.

Dose response plot.

For the dose–response relationship between road traffic noise and overall sleep disturbance, the linear model (y = −0.6680 + 0.0190 × Dose) provided the best fit, indicating that each 1 dB increase in noise level was associated with a 0.0190-unit increase in the effect size of sleep disturbance. Although the nonlinear quadratic model (y = 1.3029–0.0692 × Dose + 0.000948 × Dose²) showed a trend towards a U-shaped relationship, model comparison favoured the hypothesis of linear association. This linear association was based on 18 data points from 13 studies and was supported by robust statistical evidence, underscoring the importance of road traffic noise control for improving overall sleep quality. In contrast, for road traffic noise and difficulty falling asleep, the non-linear quadratic model (OR = exp [103.3888–3.5849 × Dose + 0.030920 × Dose²] provided the best fit, revealing a U-shaped relationship of significant clinical relevance. The model indicated that the risk of difficulty falling asleep was lowest at approximately 58.0 dB, with increasing risk at both lower and higher exposure levels. This U-shaped pattern suggests a biphasic response of the sleep initiation process to road traffic noise: extremely low noise levels may induce hyperarousal, whereas higher noise levels directly disrupt the physiological mechanisms involved in falling asleep.

For road traffic noise, overall sleep disturbance or poor sleep showed a moderate linear association with exposure across 18 data points from 13 studies [Akaike information criterion (AIC) = 30.3 vs. 31.3, P > 0.2]. In contrast, difficulty falling asleep exhibited a significant non-linear relationship [Akaike information criterion (AIC) = 9.2 vs. 13.1, P < 0.001], with a marked increase in risk above approximately 60 dB. In contrast, difficulty maintaining sleep or night awakenings exhibited a linear trend (P = 0.04), suggesting a cumulative effect with increasing exposure.

For railway noise, non-linear patterns were observed for both difficulty falling asleep and difficulty maintaining sleep/night awakenings (P < 0.001 for both), potentially indicating a threshold of approximately 55–60  dB. However, these dose–response curves were each derived from a single study with only seven data points and should therefore be interpreted with caution.

For aircraft noise, linear models provided an adequate fit across all sleep outcomes (all P > 0.2), with no clear dose–response relationship identified (all P > 0.2). The number of data points per outcome ranged from three to eight, and the exposure range was limited.

Taken together, these findings suggest that road traffic noise shows the most consistent relation with sleep disturbance. Railway noise may demonstrate non-linear threshold effects, whereas aircraft noise shows no discernible dose–response pattern within the examined exposure range.

DISCUSSION

Impacts of Traffic Noise on Sleep Disturbances

Based on 68 independent studies across 30 articles, this meta-analysis demonstrates a positive and statistically significant association between traffic noise exposure (aircraft, railway and road traffic) and three sleep outcomes: overall sleep disturbance/poor sleep quality, difficulty falling asleep and difficulty maintaining sleep/night awakenings. A 5 dB increase in noise exposure was associated with an approximately 19% higher risk of overall sleep disturbance, while the risks of difficulty falling asleep and difficulty maintaining sleep increased by approximately 18–19%. Aircraft noise showed the strongest overall impact, railway noise exhibited the highest association with sleep initiation difficulties and road traffic noise showed consistent adverse effects across outcomes. These findings are consistent with experimental evidence demonstrating that nighttime transportation noise increases sleep fragmentation, awakenings and sleep onset latency in a dose-dependent manner.[10]

These results align with those of Rocha et al.,[52] who reported strong nighttime noise effects across road, rail and aircraft traffic, supporting a dose–response pattern. The findings are also consistent with Brown and Van Kamp,[87] who reported sleep improvements following nighttime noise reduction, and with Evandt et al.,[49] who reported similar associations.

From a public health perspective, sleep impairment due to traffic noise represents a population-level risk. Gilani and Mir[62] reported a higher risk of insomnia among residents living near major roads, and Münzel et al.[88] described nighttime noise as a ‘chronic environmental stressor’, linking sleep fragmentation and stress hormones to cardiovascular injury. Repeated nocturnal noise-induced arousals are thought to activate sympathetic and endocrine stress pathways, providing a plausible biological mechanism linking sleep disturbance to cardiometabolic risk. More than 1.6 million disability-adjusted life years are lost annually in Europe due to traffic noise, indicating its evolution from an environmental problem into a major public health concern.[89]

Overall, road traffic noise showed the most consistent association with sleep disturbance, while potential threshold effects for railway noise warrant further validation.

Effects of Non-Traffic Noise on Sleep Disturbances

The analysed non-traffic noise sources (community noise, hospital ward noise and wind turbine noise) exhibited type-specific and heterogeneous associations with sleep disturbance. Pooled correlations showed associations between community noise and difficulty falling asleep (r = 0.24), community noise and difficulty maintaining sleep (r = 0.18), hospital ward noise and reduced sleep efficiency (r = 0.31) and wind turbine noise and overall sleep disturbance (r = 0.15). These results indicate that different non-traffic noise sources affect different aspects of sleep.

For community noise, the noise-related annoyance, cognition, and health (NORAH) study reported that each 5 dB increase in exposure prolonged sleep onset latency by 6.3 minutes and increased the risk of overall sleep disturbance by 23%.[90] For hospital noise, McLoone et al.[75] showed that a 5 dB increase led to additional awakenings and reduced deep sleep, consistent with the pooled findings in the present study. Hospital noise is characterised by irregular peak events and high nighttime variability, which are particularly disruptive to sleep continuity and slow-wave sleep.[91]

Effect sizes varied substantially across studies. Martens et al.[92] found weak associations between wind turbine noise and subjective sleep complaints, whereas Ageborg Morsing et al.[93] showed that increasing community noise levels elevated the probability of sleep disturbance, supporting a dose–response relationship.

Evidence regarding effect modifiers highlights substantial heterogeneity. Noise-sensitive individuals exhibited a higher risk of sleep disturbance, while better housing insulation reduced disruption. Basner et al.[1] found that subjective sleep effects were larger than objective effects, reflecting variability related to outcome measurement. Only a minority of non-traffic noise studies employed objective sleep measures, and exposure metrics varied widely.

Additionally, controlled acoustic exposure may exert beneficial effects. Zhao et al.[89] found that music improved subjective sleep quality (Hedges’ g = −0.66), and Jespersen et al.[94] showed that music reduced PSQI scores by 2.79 points. These findings suggest that structured auditory input may support sleep regulation, in contrast to passive exposure to environmental non-traffic noise. This contrast underscores the importance of predictability, controllability and emotional valence of auditory stimuli in determining their effects on sleep, rather than sound exposure per se.[95]

In summary, non-traffic noise is associated with sleep disturbance, but the magnitude of effects varies across noise types and sleep outcomes. Substantial heterogeneity remains, and standardised exposure metrics and sleep assessment methods are needed for more precise estimation of non-traffic noise effects.

Risk of Bias, Certainty of Evidence and Study Heterogeneity

Using the ROBINS-E tool and the GRADE framework, the included studies were characterised by varying levels of risk of bias, evidence certainty and heterogeneity. Most studies were assessed as having a low to moderate risk of bias (78%), while 22% were rated as serious risk, primarily owing to residual confounding and limitations in exposure classification. In several studies, adjustment for key confounders was incomplete, or noise exposure was self-reported; a small number also showed potential bias related to missing data. Importantly, these limitations were unlikely to meaningfully influence the pooled estimates, and the overall level of bias control was consistent with accepted standards for observational epidemiological research.

The GRADE assessment indicated moderate certainty of evidence. Cohort studies contributed higher-quality evidence, whereas most cross-sectional studies were downgraded due to study limitations and indirectness. A consistent exposure–response relationship supported the upgrade of the dose–response dimension.

Heterogeneity was substantial (I² = 52–99%) and driven by measurement type, region and study quality, with subjective measures yielding higher effect sizes. The observed heterogeneity was largely attributable to methodological differences and did not substantially undermine the interpretability of the overall conclusions.

Limitations of the Study

Some limitations should be acknowledged. Methodological challenges related to causal inference, exposure assessment and confounding control remain important concerns. First, the study design limits causal interpretation: 85% of the 68 included studies were cross-sectional, precluding the exclusion of reverse causality and limiting assessment of long-term cumulative noise exposure effects. Although cohort studies provided stronger inferential power, most included <5000 participants, reducing sensitivity for detecting subtle harmful effects of low-level noise exposure (<40 dB Lnight).

Second, exposure assessment accuracy varied considerably. Twenty-three (34%) studies relied on residential proximity to noise sources rather than direct acoustic measurements or validated model-based indicators (e.g., CadnaA), introducing potential exposure misclassification. Although most studies used noise modelling tools (e.g., CadnaA, SoundPLAN), only a subset conducted field validation to calibrate model outputs against measured environmental noise levels. While such modelling approaches are standard in large-scale epidemiological studies, this limitation may have introduced bias. Additionally, only 11 (16%) studies measured both indoor and outdoor noise; most assumed equivalence between the two and did not account for modifying factors such as window position, building façade orientation or insulation quality, which could moderate actual exposure levels.

Third, confounding control was incomplete in some studies. Nineteen (28%) studies did not adjust for noise sensitivity, a key individual-level modifier of noise-related health effects, and only nine (13%) studies accounted for co-exposure to air pollutants (e.g., PM_2.5, NO₂), which may synergistically affect sleep health alongside noise.

Fourth, uncertainty persists due to measurement biases and substantial heterogeneity. Most studies (85%) relied on subjective sleep measures (e.g., questionnaires, PSQI) rather than objective tools (e.g., actigraphy, polysomnography). Subjective measures may overestimate effects due to recall bias or overlap between noise annoyance and perceived sleep quality, although this was partially addressed through subgroup analyses comparing subjective versus mixed (subjective + objective) measurement strategies. Evidence for certain noise types, particularly railway noise, was limited to only four studies, warranting cautious interpretation of effect estimates and dose–response patterns. Furthermore, high heterogeneity (I² up to 99.27%) reflects variability in study designs, exposure metrics (e.g., Lnight vs. Lden), outcome definitions and regional contexts; however, meta-regression and subgroup analyses identified key moderators, such as measurement type and geographic region, which explain part of this variability.

These limitations underscore persistent challenges in environmental noise epidemiology, particularly in relation to causal inference, accurate exposure quantification and comprehensive confounding control. Future research should focus on the following: (1) large-scale longitudinal studies incorporating objective exposure assessment (including field validation of noise models) and objective sleep monitoring; (2) standardised application of validated exposure metrics and sleep outcome tools to reduce methodological heterogeneity; (3) expanded investigation of understudied noise types (e.g., railway noise) to enhance generalizability; and (4) more comprehensive adjustment for individual modifiers (e.g., noise sensitivity) and co-pollutants. Despite these limitations, the current findings remain robust, supported by high GRADE evidence quality, consistent dose–response patterns and sensitivity analyses confirming the stability of pooled estimates.

CONCLUSION

This meta-analysis provides systematic and quantitative evidence that environmental noise is significantly associated with sleep disturbances, with type-specific effects and clear dose–response patterns. Traffic noise, including aircraft, railway and road traffic, consistently affects overall sleep quality, sleep initiation and sleep maintenance. In contrast, non-traffic noise sources, such as community, hospital or wind turbine noise, demonstrate weaker and less consistent associations.

These findings have important implications for public health and environmental noise management. Interventions should be source specific: for urban traffic noise, low-noise road surfaces, noise barriers and improved building insulation are recommended; for non-traffic settings, such as hospitals and residential areas, attention should be given to equipment placement, nighttime activities and acoustic design.

Future research should prioritise three areas: (1) enhancing causal inference through large-scale longitudinal studies with objective acoustic and physiological monitoring, particularly for non-traffic noise; (2) standardising exposure and sleep outcome metrics to reduce methodological variability; and (3) expanding research globally, particularly in underrepresented regions, to improve cultural comparability and generalizability.

Environmental noise constitutes both an environmental stressor and a quantifiable risk to sleep health. Effective noise regulation, grounded in robust evidence on exposure–sleep relationships, remains critical for safeguarding population health across diverse settings.

Author Contributions

Zitong Li: Conceptualization, Methodology, Data collection, Data curation, Statistical analysis, Visualization, Writing − original draft, Writing − review & editing.

Availability of Data and Materials

The original contributions presented in the study are included in the article. Further inquiries can be directed to the corresponding authors.

Ethical Approval and Consent to Participate

Not applicable. This study is a systematic review and meta-analysis based solely on previously published data. No new human participants or animals were involved, and therefore, ethical approval and informed consent were not required.

Conflict of Interests

The authors have no conflicts of interest.

Funding Statement

None.