Exploring the effects of occupational noise on cardiovascular and cerebrovascular diseases: A Mendelian randomization study

Abstract

Existing observational studies have demonstrated a potential association between exposure to noisy workplaces and the occurrence of cardiovascular and cerebrovascular diseases. However, current research is still inconclusive and inconsistent. Therefore, we attempted to delineate the causal correlation between noisy workplace exposure and both cardiovascular and cerebrovascular diseases in the European population by using Mendelian randomization analysis. The necessary data on cardiovascular and cerebrovascular diseases, as well as noisy workplaces, were all accessible from the genome-wide association studies database. The inverse variance weighted analysis served as the primary method, supplemented by a few sensitivity analysis methods to evaluate the robustness of our findings. Genetically inverse variance weighted methods revealed that suffering from occupational noise may be linked to a significantly increased incidence of hypertension (P = .005; odds ratio (OR): 1.049; 95% CI: 1.015–1.084), heart failure (P = .014; OR: 1.325; 95% CI: 1.058–1.660), and stroke (P = .019; OR: 1.010; 95% CI: 1.002–1.018). However, there was no causal relation between occupational noise and angina pectoris (P = .249; OR: 1.006; 95% CI: 0.996–1.017), acute myocardial infarction (P = .112; OR: 1.008; 95% CI: 0.998–1.018) as well as atrial fibrillation and flutter (P = .560; OR: 1.002; 95% CI: 0.995–1.009). In addition, sensitivity analysis techniques, including Cochran’s Q test, leave-one-out analysis, as well as MR-Egger intercept, were employed to guarantee that our findings were more reliable and robust. The Mendelian randomization study suggested a genetically predicted causal involvement of working noise in the increased pathogenesis of hypertension, heart failure, and stroke, instead of acute myocardial infarction, angina pectoris, atrial fibrillation, and flutter.

1. Introduction

Noise pollution is a significant environmental concern and a major issue in many countries around the world, particularly in the workplace.^[1] Workplace noise is primarily generated by the operation of machinery and equipment due to technical limitations.^[2] In addition, it arises from the operational activities of the workers. According to European surveys, about 1/3 of factory workers are subjected to noisy environments during working hours.^[3] Similarly, it is reported that no <80 million Chinese workers in industrial fields and workplaces are experiencing hazardous and harmful noise.^[4]

Numerous observational studies have indicated that working in a noisy environment may contribute to a wide range of health issues, such as hearing loss, poor sleep quality, and cardiovascular diseases, thereby adversely affecting the well-being of workers.^[5,6] In addition, a recent study has revealed that workers accessible to working noise might be at a heightened risk of developing high blood pressure in southern China.^[2] However, the aforementioned observational studies are inherently biased and have insufficient and limited sample sizes. And to some extent, they also suffer from reverse causation. The implementation of randomized controlled trials has been restricted due to ethical concerns and high costs, resulting in a scarcity of available data. Besides, it is reported that there are approximately 523 million patients with cardiovascular diseases worldwide (including coronary heart disease, stroke, heart failure (HF), etc), accounting for 6.7% of the global population. Given the substantial impact of CVD on public health, it is imperative to investigate and understand the potential link between noisy workplaces and CVD at present. To address this, we attempted to assess the correlation between the likelihood of developing cardiovascular and cerebrovascular diseases and noise in the workplace by utilizing Mendelian randomization (MR).

MR refers to a very valuable and useful tool for understanding the connection between exposure and outcome.^[7] This approach can effectively help mitigate residual confounding, bias, and reverse causality^[8] that frequently show up in conventional observational studies. The lack of randomized clinical trials (RCTs) means that the MR method serves as a surrogate strategy for causal inference at present. Gene variants in meiosis are isolated at random, which mimics RCTs.^[9]

To this day, there has been no study utilizing MR methods to examine the potential adverse effects of working noise on both cardiovascular and cerebrovascular diseases. Thus, we obtained summary statistics for occupational noise and cardiovascular and cerebrovascular diseases from the genome-wide association studies (GWAS). These analyses will be instrumental in gaining deeper insights into the cardiovascular risks associated with occupational noise, potentially leading to significant clinical and public health implications in the near future.

2. Materials and methods

2.1. Study design

This research examined the genetic causal link between a noisy workplace and cardiovascular and cerebrovascular diseases. The noisy workplace served as the exposure, while 6 kinds of cardiovascular and cerebrovascular diseases were respectively as the outcomes in our analysis. MR operates on 3 fundamental assumptions (as depicted in Fig. 1), which the chosen instrumental variables (IVs) must satisfy. The research flow of this study is shown in Figure 2.

Figure 1.

Three fundamental postulates of Mendelian randomization (MR). IVs = instrumental variables, MR = Mendelian randomization, SNPs = single-nucleotide polymorphisms.

Figure 2.

Illustration of the 2-sample Mendelian randomization (MR) analysis process. MR = Mendelian randomization.

2.2. Data sources

All the necessary data were derived and collected from the GWAS database (https://gwas.mrcieu.ac.uk/). Noisy workplaces represent occupational noise. The data we used were the most recent, with a relatively large sample size (n > 100,000). The number of single-nucleotide polymorphisms (SNPs) was more than 1 million, and information was relatively complete. The sample size of the noisy workplace phenotype is 151,624 (GWAS ID: ukb-b-2091). The GWAS database was adopted to obtain data on 6 types of cardiovascular and cerebrovascular diseases. More detailed information regarding these data is depicted in Table 1.

Table 1.

More detailed introduction of necessary data.

Diseases	ID	Cases	Controls	Year	Number of SNPs
Hypertension	ukb-b-12493	54,358	408,652	2018	9,851,867
Heart failure	ebi-a-GCST009541	47,309	930,014	2020	7,773,021
Stroke	ukb-b-8714	7055	454,825	2018	9,851,867
Acute myocardial infarction	ukb-a-533	3927	333,272	2017	10,894,596
Angina pectoris	ukb-a-532	4837	332,362	2017	10,894,596
Atrial fibrillation and flutter	ukb-a-536	3818	333,381	2017	10,894,596

SNPs = single-nucleotide polymorphisms.

2.3. Instrumental variables selection

We attained SNPs of the noisy workplace, reaching a genome-wide association (P < 1 × 10⁻⁶). To minimize the effect of linkage disequilibrium, we set a threshold at R² < 0.001 and genetic distance = 10,000 kb.^[10] Furthermore, we got rid of SNPs relevant to known confounders from the LDlink database. We did not utilize the traditional power calculations. On the contrary, to mitigate the potential bias due to weak IVs in our study, we used the F-statistic to obtain strongly correlated IVs.^[11] The equation of F-statistic is: F = R²(n − k − 1)/k (1 − R²) and R² = β²(1 − EAF) × 2EAF.^[12] When the F-value was more than 10, the SNP was regarded as compliant and strong.^[13] After careful calculations, we found that the F-values of the 18 SNPs we obtained ranged from 20 to 35. For more detailed information on the 18 SNPs, see the Supplemental Content, Supplemental Digital Content, https://links.lww.com/MD/R284.

2.4. Statistical analysis

Three MR methods were adopted to explore the causal impact of the noisy workplace on cardiovascular and cerebrovascular diseases. Inverse variance weighted (IVW) is primarily applied to explore and estimate causal effects in MR studies and is known for its stronger ability to detect causality.^[14] However, IVW requires that IVs only have an influence on outcomes through exposure in the study, and there is no horizontal pleiotropy.^[15] If these conditions are not met, then the hypothesis may not be reliable. Hence, we used complementary analysis approaches, including MR-Egger and the weighted median estimate. When horizontal pleiotropy exists, the MR-Egger method is more appropriate than others.^[16] The weighted median estimate not only can offer a similar estimation of causal effects but also is less biased than the IVW method when invalid IVs are present, as long as at least 50% are derived from valid IVs.^[17] However, each method has its limitations in a way; therefore, we adopted a variety of sensitivity analysis methods for the purpose of improving the reliability of our results.

2.5. Sensitivity analysis

Three MR approaches were employed to ensure the robustness of our findings. With the help of Cochran’s Q test, heterogeneity can be evaluated.^[18] Significant heterogeneity was observed to exist when Pcochran’s Q was <0.05.^[19] Although Pcochrane’s Q < .05, it did not affect the results of the IVW method. In the meanwhile, the MR-Egger intercept approach was utilized to detect whether pleiotropy existed.^[20] If the P-value is <.05, there might be a bias in the IVW estimates.^[16] Finally, a leave-one-out analysis was adopted by us to test and clarify the effect of each SNP. The leave-one-out test is an approach that involves the systematic removal of an SNP at a time.^[21,22] When the result alters greatly after removing an SNP, it indicates that the presence of an SNP has a great impact on the outcome.^[23,24]

3. Results

3.1. Causal impact of occupational noise on diseases

We applied 3 MR approaches to explore the causality of the noisy workplace in both cardiovascular and cerebrovascular diseases. As shown in Table 2, significant associations were found in essential hypertension (HP), HF, and stroke, instead of acute myocardial infarction (AMI), atrial fibrillation and flutter (AFF), and angina pectoris (AP).

Table 2.

Risk estimates of the noisy workplace for 6 diseases.

Outcome	Methods	Odds ratios (95% CI)	P-values
Hypertension	Inverse variance weighted	1.049 (1.015, 1.084)	.005
	Weighted median	1.043 (1.012, 1.074)	.005
	MR-Egger	1.008 (0.921, 1.102)	.869
Heart failure	Inverse variance weighted	1.325 (1.058, 1.660)	.014
	Weighted median	1.262 (0.942, 1.692)	.135
	MR-Egger	1.931 (0.994, 3.750)	.076
Stroke	Inverse variance weighted	1.010 (1.002, 1.018)	.019
	Weighted median	1.008 (0.998, 1.020)	.135
	MR-Egger	1.009 (0.911, 1.118)	.862
Angina pectoris	Inverse variance weighted	1.006 (0.996, 1.017)	.249
	Weighted median	1.009 (0.997, 1.020)	.154
	MR-Egger	1.007 (0.979, 1.037)	.627
Acute myocardial infarction	Inverse variance weighted	1.008 (0.998, 1.018)	.112
	Weighted median	1.008 (0.997, 1.018)	.148
	MR-Egger	1.028 (1.003, 1.054)	.044
Atrial fibrillation and flutter	Inverse variance weighted	1.002 (0.995, 1.009)	.560
	Weighted median	0.999 (0.991, 1.008)	.890
	MR-Egger	1.004 (0.986, 1.022)	.667

CI = confidence interval, MR = Mendelian randomization.

The IVW method revealed a positive and potential association between a noisy workplace and HP, HF, and stroke, instead of AFF, AMI, and AP. The odds ratios of HP, stroke, and HF were 1.049, 1.010, and 1.325. Although the statistical significance of the causal impact of a noisy workplace on HP/HF/stroke was less pronounced in the MR-Egger or weighted median methods, the trends of odds ratio in these methods were similar to those of the primary IVW method. Therefore, we considered these results to be reliable and effective.

3.2. Sensitivity analysis

In our analysis, we utilized Cochran’s Q test to identify whether heterogeneity exists. What is more, the MR-Egger intercept method was employed to detect the horizontal pleiotropy. As presented in Table 3, our findings confirmed that horizontal pleiotropy was not present in this analysis. According to Cochran’s Q test, we detected heterogeneity in HP, AP, and AMI, whereas no heterogeneity was observed in other diseases.

Table 3.

Heterogeneity and horizontal pleiotropy tests.

Outcome	Pcochran’s Q	Pmr-Egger intercept
Hypertension	1.94 × 10−6	.359
Heart failure	.876	.260
Stroke	.720	.995
Angina pectoris	.020	.931
Acute myocardial infarction	.011	.109
Atrial fibrillation and flutter	.897	.814

In the “Leave-one-out” method we conducted, we found that no single SNP impacted the overall estimation of causal association, which to some extent validates the reliability and robustness. See Figures 3–8 below.

Figure 3.

“Leave-one-out” analysis for noisy workplace on hypertension. MR = Mendelian randomization.

Figure 8.

“Leave-one-out” analysis for noisy workplace on atrial fibrillation and flutter. MR = Mendelian randomization.

Figure 4.

“Leave-one-out” analysis for a noisy workplace on heart failure. MR = Mendelian randomization.

Figure 5.

“Leave-one-out” analysis for a noisy workplace on stroke. MR = Mendelian randomization,

Figure 6.

“Leave-one-out” analysis for noisy workplace on angina pectoris. MR = Mendelian randomization.

Figure 7.

“Leave-one-out” analysis for noisy workplace on acute myocardial infarction. MR = Mendelian randomization.

4. Discussion

There has been no study utilizing MR methods to examine the potential adverse effects of working noise on both cardiovascular and cerebrovascular diseases to now. Hence, by employing MR analysis, we sought to explore the probable causality and susceptibility between exposure to a noisy workplace and specific cardiovascular diseases. By utilizing the IEU GWAS pooled database on occupational noise and 6 kinds of diseases, our results revealed a notable potential association between genetic predisposition to occupational noise and HF, and stroke, as well as a heightened risk of HP instead of AMI, AFF, and AP.

Accumulating observational studies and evidence have found a strong correlation between working noise and cardiovascular diseases.^[25] Our research aligns with these previous studies and findings. For instance, a study whose data were obtained from the UK Biobank indicated that prolonged access to road traffic noise was greatly linked with an increased risk of incident HF.^[26] In Canada, the higher incidence of congestive HF is due to long-term exposure to road traffic noise.^[27] What is more, in a retrospective study, a higher frequency of working noise exposure was most strongly relevant to HP.^[28] Another study highlighted that there was a notably elevated risk of HP among individuals suffering from both aircraft and road traffic noise at or above 45 dB.^[29] However, some observational studies have yielded different conclusions. For example, a prospective study found that there was no elevated risk of HP when individuals suffered from noise ranging from 85 to 90 dB(A).^[30] In addition, Tessier-Sherman reported that the hazard ratio of HP never varies greatly among workers when using classified noise exposure metrics in a 6-year study.^[31] Likewise, no indication of increased chronic HF risk was observed in the study by Hahad’s study after 5 years of follow-up.^[32] Upon comparison and analysis, the conflicting results of these published studies might stem from differences in populations, races, study design, sample sizes, duration of noise exposure, and confounding factors in various studies.

The potential biological mechanisms of HP, stroke, and HF caused by noise have not been fully interpreted and elucidated at present. Most experts posit that noise may perturb various physiological systems to promote the Hermon secretion of catecholamines, cortisol, as well as angiotensin II, which could give rise to HP.^[33] In addition, noise could stimulate a variety of inflammatory reactions, as well as oxidative stress responses,^[34] which may further lead to vascular endothelial cell dysfunction and decreased vascular NO bioavailability, ultimately leading to HP, stroke, and HF.^[35,36]

In comparison to MR analysis, it is important to note that observational studies may have more bias and limited sample sizes, and there is a possibility of reverse causality. Moreover, due to ethical considerations, time constraints, and high costs, randomized controlled trials have not been widely conducted and are not easily accessible at present. The MR approach is a suitable strategy for causal reasoning. Genes were randomly separated, which is equivalent to a RCT at the gene level.^[37] In addition, when an RCT is not feasible, MR has the highest evidence-based rating, ranking first with a higher level of evidence than traditional observational studies.^[38] Hence, our study’s strengths include the minimization of unmeasured confounders and avoidance of reverse causation, as well as the reduction of time consumption and high costs. Additionally, to guarantee the reliability of the results, we utilized a few types of sensitivity tests and assessed SNP strength, such as the F-statistic and removal of weak IVs. Consequently, the results obtained through MR analysis are more dependable and impactful compared to those of traditional observational studies.

Our study has its limitations. To begin with, since our study sample only originated from European populations, it is suggested that the results may not be universally applicable to some extent, especially to populations and races in East Asian countries. Secondly, although we examined the genetic aspects of the relationship, the underlying mechanisms remain unknown and require further investigation. Additionally, our research is confined to the genetic level and does not explore specific phenotypes such as the severity of HP, sex differences, HF, or acute exacerbations. Finally, MR analysis could not determine the effects of clinical interventions on occupational noise exposure.

5. Conclusion

In conclusion, our study utilizing MR has validated that occupational noise may heighten susceptibility to HP, stroke, and HF. In consideration of our research findings, we advocate that it is imperative and urgent to cut down noise exposure in the workplace to mitigate the incidence of HP, stroke, as well as HF among workers.

Acknowledgments

We express our heartfelt thanks to all the people who were involved in the mentioned data for making their GWAS data accessible and available.

Author contributions

Conceptualization: Siyu Peng.

Project administration: Siyu Peng.

Software: Siyu Peng.

Writing – original draft: Siyu Peng.

Writing – review & editing: Siyu Peng, Jingbo Hou.

Visualization: Xingtao Huang.

Supervision: Chao Fu.

Validation: Chao Fu.

Methodology: Qi Liu.

Supplementary Material

medi-105-e47331-s001.xlsx ^{(11.5KB, xlsx)}