Methods

The Institutional Review Board at the Harvard School of Public Health approved the study protocol, and written informed consent was obtained from each adult prior to participation. From 1999 to 2006, we recruited apprentice boilermaker construction workers or ‘boilermakers’ from a union to participate in a paired work and non-workday panel study. Participants were monitored by ambulatory ECG at a training school over the duration of a work shift and while at home for up to 24 h. For the current study, 18 participants who were monitored for a minimum of 22 h starting between 8:00 and 11:00 were eligible for inclusion in the analysis.

Over a work shift, boilermakers received instruction and practised shielded metal arc (stick) and gas metal arc welding, most commonly using base metals of mild steel and stainless steel. Personal, integrated, gravimetric particulate matter with an aerodynamic diameter of <2.5 μm (PM_2.5) samples were collected over the duration of the work shift as previously described in detail.7 Like workday monitoring, non-workday monitoring occurred at the union hall when participants were not welding and while at home over a 24-h period. Non-workday monitoring occurred within 5 months of workday monitoring, although 86% occurred within the same week. While the majority (64%) of workers were monitored once, four workers were monitored twice during different years.

A self-administered questionnaire was used to collect information on medical history, medication use, and smoking history. Participants were fitted with a standard 5-lead ECG Holter monitor each workday or non-workday morning. Each cardiac tape was sent to Raytel Cardiac Services (Haddonfield, NJ, USA) for processing and analysis using a DelMar Avionic (Irvine, CA, USA) Model Strata Scan 563. Trained technicians, blinded to the work and non-work periods, used standard criteria to accept or reject all normal or abnormal findings. The duration between consecutive R waves of the ECG (RR interval) was identified and only beats with an RR interval between 0.6 and 1.5 s and ratio of 0.8–1.2 were included in the analysis. Tapes were analysed in the time domain, and the SD of all normal-to-normal intervals (SDNN) was calculated for all 5-min segments. The SDNN index (SDNN_i), calculated from the mean of the 5-min SDNN, was summarised hourly.

Since declines in HRV are observed up to 14 h following welding,6 the entire 24-h period from the start of the work shift was considered to be the work period. To account for correlated outcomes among workers who participated over multiple years, linear mixed-effects regression models with random intercepts for individuals, an indicator variable for work/non-work period and unstructured covariance, were used to examine the difference between work and non-work SDNN_i within hourly individual models. Using a repeated, balanced study design, each participant served as their own control, which eliminates confounding by individual factors that effect HRV, such as age or smoking status.

Each hourly measurement was pooled into one model to assess disruption in the circadian pattern and to identify the mean difference in SDNN_i between work and non-work over the entire measurement period. A linear mixed-effects model with random intercepts for each individual, an indicator variable for work period, and the first and second harmonic regression coefficients for time [sin (2π(t)/24); cos(2π(t)/24); sin (2π(t)/12); and cos(2π(t)/12), where t is the time of day] were used with an interaction term between the four harmonic regression variables and work to investigate whether the circadian pattern of HRV varied over time. Model fit for harmonic and interaction terms was assessed using the likelihood ratio test and χ² distributions. Statistical significance was assessed at the α=0.05 level. Analyses were performed using SAS version 9.1.