Effect of welding fume on heart rate variability among workers with respirators in a shipyard

Welding fume exposure is associated with heart rate variability (HRV) reduction. It is still unknown whether respirator can reduce effect of welding fume on HRV among welding workers in a shipyard. We recruited 68 welding workers with respirator and 52 welding workers without respirator to measure HRV indices, including standard deviation of normal-to-normal intervals (SDNN) and square root of the mean squared differences of successive intervals (r-MSSD) by ambulatory electrocardiographic (ECG). Personal exposure to particulate matter less than or equal to 2.5 μm in diameter (PM2.5) was measured by a dust monitor. The association between 5-minute mean PM2.5 and log10-transformed HRV indices was analyzed by mixed-effects models. We found 5-minute mean PM2.5 was associated with 8.9% and 10.3% decreases in SDNN and r-MSSD. Effect of PM2.5 on HRV indices was greatest among workers without respirator {SDNN: 12.4% (95% confidence interval = −18.8–−6.9); r-MSSD: 14.7% (95% confidence interval = −20.8–−8.6)}. Workers with respirator showed slight decreases in HRV indices {SDNN: 2.2% (95% confidence interval = −6.3–−1.9); r-MSSD: 4.0% (95% confidence interval = −6.4–−1.6)}. We conclude that respirator use reduces the effect of PM2.5 exposure on HRV among workers performing welding in a shipyard.

particles and HRV changes are still inconsistent. Moreover, it is still unknown whether respirator can reduce adverse effect of welding fume on HRV.
In order to answer whether occupational exposure to welding fume is associated with impaired cardiac autonomic control and the effectiveness of respirator on protecting workers from welding fume, we recruited a panel of welding workers in Taiwan to clarify these scientific questions.

Results
Summarized statistics for 120 participants' characteristics, heart rate variability monitoring, PM 2.5 exposure and climate conditions are presented in Table 1. The study participants were adult workers with a mean age of 51.1 years {standard deviation (SD) = 10.3 years}, 37.5% were current smokers, and 56.7% used respirator while working. There were no female workers among our population. No significant different was observed in age, BMI and seniority between workers with and without respirators. For PM 2.5 , temperature and humidity monitoring, no significant difference was observed in PM 2.5 exposure and climate conditions between workers with and without respirators during the study period. For HRV monitoring, the workers with respirators had significantly higher values of HRV indices (Log 10 SDNN and Log 10 r-MSSD) than did the workers without respirators. The results of associations between concurrent PM 2.5 and HRV indices estimated by the mixed-effects models are shown in Tables 2. With age, BMI, smoking, temperature and relative humidity being adjusted in all mixed-effects models, we found that PM 2.5 exposure was associated with HRV changes for welding workers in general. It is noteworthy that PM 2.5 exposure significantly decreased HRV indices for workers without respirators. Log 10 SDNN decreased by 12.4% per IQR of PM 2.5 at 5-minute mean among workers without respirators. No significant association between PM2.5 and Log 10 SDNN was observed among workers with respirators. Likewise, r-MSSD decreased by 14.7% per IQR of PM 2.5 at 5-minute mean among workers without respirators but only decreased by 4.0% among workers with respirators. Table 3 shows the results of associations of concurrent PM 2.5 with HRV indices estimated by the mixed-effects models stratified by smoking. With age, BMI, temperature and relative humidity being adjusted in all mixed-effects models, we found that PM 2.5 exposure was associated with HRV reduction for both smoking and non-smoking workers without respirators. No significant association of PM 2.5 with HRV reduction was observed among workers with respirators except Log10 r-MSSD in smoking workers.
We further assessed effect modification in mixed-effects models and found an effect modification of PM 2.5 by respirator usage (Table 4). Workers without respirators showed a decrease of 7.4% in Log 10 SDNN, which was associated with an increased 5-minute mean PM 2.5 . In contrast, workers with respirators showed no significant change in Log 10 SDNN. Likewise, workers without respirators showed a decrease of 10.2% in Log 10 r-MSSD by PM 2.5 exposure, but workers with respirators only showed a mild decrease of 3.5% in Log 10 r-MSSD.

Discussion
Welding fume has been commonly measured as total dust or inhalable particles in past studies. However, fine particles, which can reach the alveoli, are more specific with regard to cardiopulmonary diseases and systemic particles exposure from welding fume 14 . Previous studies demonstrate the association of environmental exposure to PM 2.5 with decreased HRV in human subjects [5][6][7] . Recent studies further report the association between occupational exposure to PM 2.5 , mainly from welding, and HRV changes in workers [9][10][11][12] . The present study confirms that occupational exposure to PM 2.5 has comparable magnitudes of effects on decreasing HRV indices in workers. Magari et al. 9 showed that SDNN decreased by 1.4% per 1 mg/m 3 of concurrent PM 2.5 (− 0.77 msec) among 39 boilermakers 9 . Cavallari et al. 10 reported that total PM 2.5 exposure was associated with decline in night r-MSSD by 18.8% (− 0.006 msec/μ g/m 3 ) among 26 boilermakers 10 .
In addition to the PM 2.5 effects on HRV indices, the present study shows respirator could modify the association between PM 2.5 and HRV. As the results of mixed-effects model with effect modifier showed (Table 3), workers without respirators had higher effects of PM 2.5 on HRV reduction compared to those in workers with respirators. This finding is the first one to demonstrate the protective effect of respirator on reducing PM 2.5 exposure and cardiovascular adverse effects. Our finding also provides further evidence to support that respirator can be the  Table 2. Percentage changes a in heart rate variability indices for an interquartile range change of 5-minute mean PM 2.5 in mixed-effects models. PM 2.5 , particulate matter less than or equal to 2.5 μ m in diameter; r-MSSD, square root of the mean squared differences of successive intervals; SDNN, standard deviation of normal-to-normal intervals. a The values are presented as percentage changes and 95% confidence interval for interquartile range changes after adjusting for age, body mass index, smoking, temperature and relative humidity in all models.  last resort in controlling welding fume exposure to worker in the aspect of occupational safety and health management system 15,16 . The possible pathophysiological mechanism between PM 2.5 and autonomic alteration has been well discussed in previous studies. In brief, inhaled PM 2.5 may exacerbate the autonomic function of the heart via induced inflammation in lung and proinflammatory cytokine expression in cardiac macrophages 17 . The time course of PM 2.5 on HRV in workers at 5-minute mean is in agreement with the findings of previous studies among workers 18 , young adults 19 and general population 20 . It has been reported that fine particles can affect both sympathetic and parasympathetic nervous systems immediately. The biological mechanism is PM-induced activation of pulmonary neural reflexes secondary to autonomic alteration after PM inhalation [21][22][23] .
Several limitations should be addressed in the present study. First, the measurement of PM 2.5 might not be representative of welding fume exposure in this study because it was monitored by only personal dust monitor. There might be unmeasured gaseous pollutants that could contribute to the observed HRV reduction. However, if the present findings were driven by gaseous pollutants, there would be no effect modification by respirator status. Therefore, the observed HRV reduction was less likely due to gaseous pollutants as gases travel freely through respirators. Second, the unmeasured meteorological variables, such as wind speed and wind direction, were likely to influence the level of personal PM 2.5 exposure 24 . Such limitation might lead to exposure misclassification and bias the pollution effects on HRV reduction toward null 25 . Third, we could not exclude the respiration's confounding effects on the association of personal PM 2.5 exposure with HRV reduction because our participants' physical activities and breathing patterns were not measured in the present study 26 . Lastly, short-term reduction of HRV has not been associated with higher risk of cardiovascular disease clinically. Cohort study is needed to elucidate the effect of respirator on the relationship between welding fume and cardiovascular health.

Conclusion
The present study found that personal exposure to welding fume was significantly associated with decreased HRV. Such adverse cardiovascular effect was lower among welders with respirators than among welders without respirators. The use of a respirator is therefore beneficial for occupational health management and welder health improvement. Therefore, we encourage employers to provide suitable respirators and adequate training for employees.

Methods
Ethics. The Ethics Committee of the Taipei Medical University-Joint Institutional Review Board approved the study protocol. The methods were carried out in accordance with the approved guidelines. All subjects received written and oral information prior to inclusion and provided informed consent.

Study participants.
We recruited 120 healthy male welding workers from a shipyard located in Taiwan as our study participants for this panel study in July and August in 2013 and 2014. The main duty of these welding workers was to weld all types of steel plates of varying sizes and shapes with electric welder. The company was obligated to provide respirators and training courses for all welding workers before they performed arc welding. The type of respirator was disposable activated-carbon facemask. We assigned a well-trained assistant to conduct eight-hour continuous personal PM 2.5 exposure monitoring and HRV monitoring for each participant while working (from 09:00 am to 17:00 pm). Each participant's age, body mass index (BMI), smoking, work history, current health status, respirator usage and work-activity pattern, including sitting, walking, talking, welding, dining, etc. were obtained from a questionnaire by a professionally trained nurse. None of participants had hyperthyroidism, hypoxemia, acute cardiopulmonary failure and paced cardiac rhythm. They were informed of the objective of this study, the benefits and possible danger to health. The ethic committee of the Taipei Medical University Joint Institutional Review Board (Taipei, Taiwan) approved this study (TMU-JIRB No.: 201306041). An informed consent was obtained from each participant before the study embarked and the individuals were allowed to give up at any moment.  Table 4. Effect modification of association of heart rate variability with 5-minute mean PM 2.5 among 120 study participants. BMI, body mass index; PM 2.5 , particulate matter less than or equal to 2.5 μ m in diameter; r-MSSD, square root of the mean squared differences of successive intervals; SDNN, standard deviation of normal-to-normal intervals. a The values are presented as percentage changes and 95% confidence interval for interquartile range changes after adjusting for age, body mass index, smoking, temperature and relative humidity in all models. Statistical analysis. We applied mixed-effects models to examine the association of PM 2.5 with SDNN and r-MSSD for our study participants by running R statistical software version 2.15.1. The SDNN and r-MSSD were log 10 -transformed to improve normality and stabilize the variance. We treated each participant's age, BMI, smoking (Yes vs. No), respirator usage as time-invariant variables, while 5-minute mean PM 2.5 , 5-minute mean temperature, 5-minute mean relative humidity, log 10 SDNN and log 10 r-MSSD as time-varying variables in our data analysis. The outcome variables were log 10 SDNN and log 10 r-MSSD, and the exposure variables were 5-minute mean PM 2.5 . We treated each participant as a random effect in our mixed-effects models. In further analyses, we investigated if PM 2.5 effects differed among smokers and non-smokers using mixed-effects models stratified by smoking. Effect modification by respirators usage (Yes vs. No) was assessed in a separate mixed model by including interaction terms between PM 2.5 exposure and respirators usage among all workers. Minimizing Akaike's Information Criterion (AIC) 13 was used as the criteria of model selection. The pollution effects are expressed as percent changes in HRV by an interquartile range (IQR) change in pollution levels for PM 2.5 .