Occupational and environmental noise exposure during pregnancy and rare health outcomes of offspring: a scoping review focusing on congenital anomalies and perinatal mortality

Introduction

Noise is a ubiquitous health hazard with demonstrated effects on hearing, hypertension, ischemic heart disease, annoyance, sleep disturbance, and school performance [1], [2], [3], [4]. In Europe, it is estimated that more than 13 million people (20% of the population) are adversely affected by long-term road traffic noise – i.e., exposed to levels considered harmful for health according to the WHO [5]. Noise is recognized as the second most harmful environmental hazard in Europe after fine particulate matter [5, 6]. The number of people exposed are underestimated as the estimate is only based on large urban areas and major road, airport, and railway exposures.

Noise is also a widespread occupational hazard. In Europe, about a third of the work force reported being exposed to such high noise levels that they were not able to speak in a normal voice for at least 25% of their working time [7]. Men are predominantly exposed to very high workplace noise levels in the mining industry, shipyards, and construction sites, whereas women are predominantly exposed to very high noise levels in the textile and food industries and at preschools and schools where the need for communication and caregiving often prevents hearing protection [8]. Compared to male workers, little is known about how occupational noise impacts the long-term health of female workers. Working during pregnancy is increasingly more common in high-income countries. Consequently, women of childbearing age are exposed to noise levels during pregnancy as high as the rest of the population in Western societies.

Congenital anomalies, structural or functional anomalies that occur during intrauterine life, are leading causes of infant mortality in high-income countries [9]. Although these anomalies are rare (approximately 6,000 cases per 100,000 babies in the Nordic countries in 2013) [10], they are a major cause of infant mortality, childhood morbidity, and long-term disability with high individual and societal costs. The prevalence of these anomalies has increased during the last 25 years due to better prenatal diagnosis, increased maternal age, and perhaps environmental factors [11]. The aetiology of these anomalies is overall poorly understood, especially considering the heterogeneity of anomalies and the complex interconnection between exposures. The literature often relates congenital anomalies to genetics, micronutrient deficiencies, and environmental factors, including infections [9], chemical exposure, air pollution, and potentially noise exposure [12, 13].

Perinatal mortality refers to the number of stillbirths and deaths in the first week of life. The perinatal period starts at 22 completed weeks of gestation and ends 7 completed days after birth [14]. The perinatal mortality is an important indicator of maternal and foetal health and reflects the quality of obstetric and paediatric care available as well. In high-income countries, perinatal mortality accounts for almost all infant mortality. In these countries, the causes of perinatal mortality vary and include congenital anomalies, placental and umbilical cord problems, and infections [14]. There is also a high proportion of unexplained deaths [15].

Evidence suggests that noise is a general stressor that leads to a heightened activity of the sympathetic-adrenal medullary system (SAM) and hypothalamic-pituitary-adrenal (HPA) axis, triggering the release of neuro-hormones and up-regulation of key stress hormones [16, 17]. Maternal stress can affect the foetus stress regulation [18, 19]. Although laboratory studies of pregnant women generally show dampened physiological reactions to stressors [20], large individual variability is evident. However, a slight elevation of cortisol can significantly affect baseline cortisol levels of the foetus [21]. Neurohormones and stress hormones could directly affect cellular processes (e.g., apoptosis and cellular differentiation), angiogenesis, and immune responses, critical processes for peri-conception, embryogenesis, and overall foetus development [22, 23]. Furthermore, previous studies have associated other stressful life situations (e.g., divorce, job loss, or death of a close relative or friend) with congenital anomalies [24]. The use of corticosteroids, synthetic cortisol-like substances, have also been associated with congenital anomalies, especially orofacial clefts [25, 26].

Some animal studies support the effects of noise exposure on congenital anomalies and perinatal mortality in rats [27, 28]. For example, Ward et al. [27] reported dwarfed hind limbs and tail defects in rats exposed to a noise generator (82–85 dB intermittent exposure for 60–75% of each hour for 5–18 h per day). In addition, noise exposure during pregnancy induced hearing dysfunction in the offspring of guinea pigs [29] and sheep [30, 31]. Even though some animal studies found no support for the effects of noise on established congenital anomalies, they did find an association between noise and foetal death or poor embryogenic and foetal development [28].

Review studies including the one in the WHO environmental noise guidelines have summarised the association between environmental or occupational noise exposures and reproductive outcomes [12, 19, 32], [33], [34]. These reviews reported few, low quality published studies on congenital anomalies or perinatal mortality in relation to maternal noise exposure [12, 32]. However, these reviews included studies published through 2014, justifying the need for an update. Most of these reviews included several outcomes (e.g., low birth weight, pre-term birth, perinatal mortality, and congenital anomalies) with an untargeted discussion on how research is conducted, and what are the key factors relevant for future research on congenital anomalies and perinatal mortality. The search strategies in previous reviews focused on search terms that potentially lead to the exclusion of studies on specific anomalies such as congenital hearing dysfunction and polydactyly. Finally, previous reviews have often assessed and discussed studies either on occupational or environmental exposures. The possibilities of co-exposures are relevant not only for increased noise dose but also for lack of at-home restoration, which might result in increased overall physical and mental load that may negatively affect both maternal and foetal health.

Objectives

The main purpose of this scoping review is to collect, summarise, and discuss the existing scientific research about the relationship between noise exposure during pregnancy and congenital anomalies in offspring and perinatal mortality.

Methods

Results

Figure 1 presents a flow chart of the selection process. The database searches resulted in 3,656 records in the first search (from 1970 to October 2019) and 198 in the second search (from November 2019 to March 2021). We identified an additional seven records from other sources (i.e., reference list of papers). After deduplication, 2,603 records remained. Of these, we judged 35 papers as eligible based on title and abstract. After assessing the full text of these 35 papers, we excluded 19 papers for the following reasons: 16 papers were not focused on either congenital anomalies or perinatal mortality and three papers were not original research papers. Studies with noise as secondary exposure were also included considering the scarcity of research in the area. This left 16 papers for inclusion in the final review: seven focused on congenital anomalies, three focused on perinatal mortality, and two focused on both congenital anomalies and perinatal mortality. We also assessed four studies on congenital hearing dysfunction using the definition of congenital anomalies (including functional anomalies) and the biological plausibility of this association.

Figure 1:

Flow chart of the literature review process according to PRISMA-P recommendations.

Table 2 summarises the characteristics of the included studies. Of the 16 publications, five reported results from cohort studies [36], [37], [38], [39], [40], six from case-control studies [41], [42], [43], [44], [45], [46], four from cross-sectional studies [45, 47], [48], [49], one from an ecological study [50], and one from a time-series study [51]. Because Edmonds et al. [45] used different study designs – one cross-sectional and one case-control study – for different outcomes, they are presented separately in the table. In addition, Hartikainen et al. [36] and Zhang et al. [43] investigated both congenital anomalies and perinatal mortality; although each was counted as one study (using the same design and sample), it appears twice in the table.

Most studies were conducted in high-income countries (USA, UK, Canada, Nordic countries, and Spain), but some studies were conducted in Turkey [48], China [40, 41, 43], and Brazil [46]. Two studies were conducted in the 1970s, two in the 1980s, three in the 1990s, two in the 2000s, six in the 2010s, and one in 2020. For environmental noise exposures and congenital anomalies, there is a significant gap between the 2017 paper from Pedersen et al. [37] and the other two papers, which were published in the late 1970s [45, 50]. For occupational noise exposures, studies were distributed more evenly since the 1980s.

Most studies on occupational noise relied on occupation or industry-based exposure assessments while some used self-report assessment – two used a combination of self-report and occupational assessment [39, 44] and two relied on self-reported exposure only [43, 48]. Among the studies that used self-report assessment, Kurppa et al. [44] estimated that 2% were misclassified, although others did not report an estimated misclassification [39, 43, 48]. The estimation by Magann et al. [39] was based on a combination of documented noise levels and self-report exposure time per day and per pregnancy week.

For environmental exposures, studies modelled the effects of road and air traffic noise but not railway noise. Only Arroyo et al. [51] used measured exposure. For the environmental exposures, some authors used a continuous exposure [37, 47] and some [45, 50] applied a cut-off for the exposed vs. non-exposed mothers. Thresholds for the definition of high environmental exposure levels were 65 dBA, except for the ecological study, which had a threshold of 90 dBA [50].

Outcomes were defined objectively based on medical records, medical assessment, or diagnostic procedures although authors included different selections of congenital anomalies based on the research questions and data availability. Some authors focused on specific groups of outcomes such as congenital heart defects [40], neural tube defects [45], and polydactyly [41]. Pedersen et al. [37] included a more comprehensive list of outcomes as they used registry data (ICD-10 codes) and following the EUROCAT classification of major and minor anomalies and the subgroups of anomalies [52]. Perinatal mortality definitions have changed overtime, so a consensual definition was not applied. The five selected studies on perinatal mortality [36, 39, 43, 47, 51] covered different periods of mortality: some included miscarriages and late neonatal mortality and others excluded the early neonatal period. Studies on hearing dysfunction were based on different assessments of the outcomes; some authors used audiometric assessment [46, 48, 49], but Selander et al. [38] used registry data on hearing dysfunction.

Statistical methods were not comparable between studies. Some performed only descriptive analysis comparing means or proportions while others performed regression analysis with adjusted assessments by including confounders and matching samples. The most common adjustment was in relation to maternal age and parity. In occupational settings, authors often discussed other work-related exposures, such as exposure to chemicals, vibration, and heat, although the analysis did not account for multiple exposures. However, Magann et al. analysed the interactions between occupational exposures – standing, lifting, and noise [39]. In the environmental exposure studies, air pollution and seasonality were often included but not modelled together, the exception being Pedersen et al. [37]. Adjustments to known determinants of health (e.g., life-style factors, infections, and use of drugs and medications) were often lacking.

Overall, results were not statistically significant for congenital anomalies or perinatal mortality in adjusted analyses; however, Smith et al. [47] reported a positive association between road traffic noise and stillbirths and Selander et al. [38] reported a borderline statistically significant association between occupational noise exposure and hearing dysfunction in offspring.

Shi et al. [41] performed a case-control study and Gong et al. [40] a cohort study. Both described statistically significant associations between occupational noise exposure and specific congenital anomalies (i.e., polydactyly [41] and congenital heart disease [40], respectively), although neither reported adjusted analysis. Furthermore, the exposure assessment was not sufficiently clear in both. In another case-control study, Zhang et al. [43] reported a small increase in the risk of congenital anomalies as a group in relation to self-reported exposure to noise, but the estimates were not statistically significant in adjusted analysis (OR 1.3, 95% CI 0.8–2.2). Other authors [36, 42] did not find any associations between occupational noise exposure and any congenital anomalies in adjusted analysis, including a prospective cohort analysis on chromosomal abnormalities [42]. Kurppa et al. [44] reported a tendency of an effect of occupational noise exposure ≥80 dBA on oral clefts (OR 1.1, 95% CI 0.8–1.6) and congenital heart defects (OR 1.3, 95% CI 0.7–2.3).

Pedersen et al. [37] investigated the associations between the exposure to road traffic noise during prenatal development and subgroups of congenital anomalies. Associations were not statistically significant, except for a borderline effect on orofacial clefts (OR 1.38 per 10dB(A) L _den, 95% CI 1.00–1.70) [37]. The other two studies on environmental noise exposure are from the 1970s and focused on air traffic noise [45, 50]. Edmonds et al. [45] was not able to show any adverse effects among the birth outcomes studied for those residing at noise levels above 65 dBA as compared to below the threshold value. In an ecological study representing the most extreme noise exposures, Jones & Tauscher [50] reported higher rates of congenital anomalies in a black population exposed to air traffic noise >90 dBA.

No studies found an association between occupational noise exposure and perinatal mortality. Two small cohort studies [36, 39] report lack of an association. Magann et al. [39] focused on multiple occupational exposures and performed an analysis not adjusted for sociodemographic factors. Furthermore, the sample of women exposed only to noise was not large enough to draw conclusions on the estimates for noise [39]. Hartikainen et al. [36] found only one case of perinatal mortality due to intrauterine infection and therefore no conclusions could be drawn. Zhang et al. [43] in a case-control study provides estimates that suggest a potential although non-significant effect of self-reported exposure to noise on antepartum mortality (OR 1.9, 95% CI 0.8–4.7). However, the authors’ adjustment of their models to maternal chronic illness may have masked the effect of noise as maternal illness may mediate perinatal mortality [43].

In a cross-sectional study, Smith et al. [47] demonstrate an association between road traffic noise (environmental noise exposure) and stillbirths (perinatal mortality). Although a cross-sectional study, this study has a good sample and a detailed noise exposure assessment. Arroyo et al. [51], using a time series analysis, found no support for an effect of road traffic noise on stillbirths. They included late stillbirths and deaths during the first 24 h of life, which does not follow the WHO definition of perinatal mortality and is not directly comparable to the outcome used by Smith et al. [47].

Studies on maternal noise exposure during pregnancy and congenital hearing dysfunction in the offspring suggest there might be an effect of maternal occupational noise exposure. Selander et al. [38] performed a cohort study of children born between 1986 and 2008 who had been diagnosed between 2003 and 2008 with a sensorineural hearing disorder (selected ICD-10 codes). They show an association between maternal occupational noise exposure ≥85 dBA compared to <75 dBA and sensorineural hearing dysfunction in offspring (HR 1.27, 95% CI 0.99–1.64). The other three papers on congenital hearing dysfunction were all based on small studies. Güven et al. [48], in a cross-sectional study using the neonatal screening for hearing dysfunction, found no associations between maternal noise exposure and the outcome. Rocha et al. [46], in a case control study using otoacoustic emissions, also reported no associations. In a cross-sectional study, Lalande et al. [49] demonstrated high frequency hearing loss in children of mothers who were occupationally exposed to noise ≥85 dBA.

Discussion

We found few studies that investigated the effects of occupational or environmental noise exposure during pregnancy on congenital anomalies (including congenital hearing dysfunction) and/or perinatal mortality. Yet, the number of papers focusing on this topic have increased recently; however, combined effects of occupational and environmental noise exposure have not been studied. Furthermore, our findings indicate relevant overarching issues for the research topic concerning (i) the definitions and assessments of occupational noise exposure, (ii) the use and choice of adjustments, and (iii) the sample sizes for studies on this topic. In addition, we present specific points related to each of the outcomes and discuss congenital hearing disorders and other congenital anomalies separately.

Overall, occupational noise exposure was assessed according to the literature, either by self-reported exposure or by occupational/industry-based exposure assessments. Previous studies found that self-reported noise exposure is overall a reliable and suitable noise exposure assessment tool in occupational settings comparable to the job exposure matrix (JEM) with an exposure limit set at 80 dBA [53]. However, the questions used for self-reported noise exposure and the limit to establish exposed vs. non-exposed populations were not standardised in the included studies and in one study not even outlined. In addition, the use of an exposure gradient, which was used in some of the included studies [38, 44, 49], could strengthen the evidence of potential associations through the demonstration of exposure-response functions [54]. The use of the JEM (as in Selander et al. [38]) is a useful tool that can allow a more standardised exposure assessment in occupational settings, particularly where there is limited information on noise exposure [55]. As most studies were retrospective, there is a risk for misclassification of exposure considering the exposure time (e.g., pregnancy period, hours per day, and absenteeism). Based on the above, we consider that the risk of exposure misclassification for the studies using occupational and/or a combination of occupational and self-report assessments may be modest to high, whereas there is a high risk for misclassification for the studies using only self-reported exposures, especially as the questions were not outlined [43] and posed post-partum [43, 48].

The studies that used a cut-off level for the classification of occupationally exposed and non-exposed mothers applied different cut-off points for this classification, varying from 78 to 85 dBA. These discrepancies might relate to specific working regulations and recommendations in different countries during different periods. Nevertheless, it seems that all the authors tried to follow the noise directives used in each country at a given time. Currently, the noise directives in Europe point to the need for action if occupational noise exposure reaches 80 dB L _AEq,8h and the ceiling threshold for acceptable noise exposure is 85 dB L _AEq,8h [56]. These thresholds are a general recommendation based on the noise levels that can potentially damage hearing and not necessarily related to levels that can harm pregnant women or their children. Yet, specific pregnancy protection recommendations might exist in some countries, for example, the reassignment of mothers to less noisy work duties or even paid leave if reassignment is not possible.

The assessment and discussion about the co-exposure between occupational and environmental noise was one of our goals with this review, but none of the papers assessed such co-exposure. We acknowledge that the co-exposure assessment might be troublesome considering the need to reach good distribution among the different levels of occupational and environmental noise exposures – e.g., individuals who are exposed to high levels of occupational noise as well as high or low levels of environmental noise.

Other covariates are also missing in several of the studies. In occupational settings, psychosocial stress and other exposures (e.g., standing, lifting, and exposure to chemicals, vibration, heat, and physical strain) were rarely included in the analysis, either as co-exposures, effect modifiers or confounders. These were discussed in some papers [41, 46] but rarely analysed, except for Hartikainen et al. [36] and Magann et al. [39]. Kurppa et al. [44] adjusted their models to co-exposure (i.e., exposure to solvents in the first trimester) without further modelling their potential synergistic effects, which could explain their statistically insignificant estimates in relation to noise. In addition, the amount of time that a mother is exposed to a given sound level (hours/day as well as days of pregnancy) should be relevant for the exposure assessment but was overlooked in several studies, except for Selander et al. [38]. Importantly, Selander et al. showed that mothers exposed for longer periods (full-time workers with few sick leave days) were at greater risk for having children with hearing dysfunction [38].

Based on previous knowledge and plausibility of effects, the most prominent environmental covariates to be acknowledged in noise and health studies are air pollution and access to restorative places such as green and blue areas. Several authors [37, 47, 51] included air pollution estimates in their analysis. Arroyo et al. [51] and Smith et al. [47] included both exposures in the same model, but Pedersen et al. [37] further tested interactions between air pollution and noise, although with statistically non-significant results. Other environmental covariates included temperature [51] and seasonality [37, 47]. Access to restorative places were not included in any of the papers although this factor has been found relevant for other outcomes connected to noise exposure e.g., [57].

At the individual level, maternal age, smoking, education, race, socioeconomic status, parity, and maternal chronic diseases are factors previous reviews have found relevant in studies targeting reproductive outcomes in relation to occupational and environmental health exposures [12, 19]. However, some of the studies in the present review have not adjusted their models for these factors. Others have included other potentially relevant covariates that could affect the outcomes such as alcohol use, drug and medication use, signs and symptoms of prenatal infections, use of vitamins (e.g., folic acid in relation to neural tube defects) [37], paternal age, and sex of the child. While it is still unclear which factors are most relevant, the choice of covariates and the statistical model strategies need to be carefully discussed considering the complex interplay of exposures and covariates.

Although some studies had large sample sizes, the outcomes in focus are rare, especially if the study used subgroup analysis and the estimated population exposed was small. This might explain the lack of statistical significance of some of the estimates presented in the included papers. To detect an effect size of at least 20%, a sample size of over 9,000 individuals with 5 controls for each case is needed for a case-control study. For cohort studies, a sample of over 24,000 individuals is needed. These sample size calculations were made assuming an exposure of 20%, incidence of congenital anomalies of 6%, and statistical power of 80. For subgroup analysis, the sample size should be even larger depending on the incidence of the specific anomaly. If the percentage of the population exposed is lower than anticipated for the sample size calculations as we observed in some of the studies in this review [45, 50], a larger sample size is also required. Therefore, to establish an association and the size of the risk, studies focusing on these outcomes should strive for a good sample size, good representation of the outcome and exposure, and a study design commensurate with the analytical modelling strategy. A current suggestion is to pool standardised and high-quality data from several countries, as proposed by EUROCAT for congenital anomalies [52], and use adequate sampling strategies and exposure assessments. This pooled data can facilitate the identification of exposures with potential to cause congenital anomalies [11].

Perinatal mortality

Studies on perinatal mortality were very scarce and used different definitions of the outcome, resulting in insufficient evidence for drawing conclusions. Health authorities and the research community should strive for a common and practical definition of perinatal mortality. Furthermore, countries should work to implement internationally comparable registry practices [14].

We would like to stress the need to discuss further the mechanisms from prenatal noise exposure on perinatal mortality. In theory, this association can be mediated by several factors such as congenital anomalies and several aspects of maternal morbidity, including maternal obesity, hypertension (including pre-eclampsia and eclampsia), and diabetes (including gestational diabetes) [59], [60], [61], [62]. All factors related to the placentation and vascular function of the placenta could be relevant to perinatal mortality [19]. Low birth weight and prematurity have been associated with prenatal noise exposure [12, 19] and therefore can be mediators as perinatal mortality includes the first week of life, which is critical for pre-term and low birth weight babies. Figure 2 provides a schematic representation of possible pathways and the biological basis for these. This scheme is relevant for the choice of confounders to include and for other modelling strategies.

Figure 2:

Schematic representation of possible pathways and biological mechanisms from noise to congenital anomalies and perinatal mortality. Biological mechanisms are adapted and extended from [19].

Conclusions

Few studies found an association between occupational or environmental noise and congenital anomalies or perinatal mortality. Our scoping review suggests that high quality studies with larger sample sizes, good representation of cases, better exposure assessment, and thoughtful study design and statistical modelling strategies are necessary. Furthermore, outcome definitions and registry practices might need to be improved if we are to test confidently the hypotheses connecting maternal exposure to noise and these rare outcomes in offspring.