All air samples during bridge painting were collected using personal samplers: 3M Organic Vapor Monitors (OVM 3500). Outdoor paint is applied to steel at air temperatures between 4°C and 32°C and relative humidity of <85% for their optimal performance. Thus, most samples in this study were collected in summer and fall, the two main working seasons for bridge painting. The study protocol for collection of personal air samples was approved by the Institutional Review Board for Human Subjects from the University of Medicine and Dentistry of New Jersey. All participants signed informed consent forms.
There were two types of air samples collected in this study: activity-specific air samples (N
57, from 16 sampling days) and daily workday air samples (N
30, from 10 subjects) (). The activity-specific samples measured personal air concentrations during specific painting activities (1–4 h), which were expected to be the peak exposures to a painter. The sampling duration for the activity-specific sample was the time period that a painter engaged in one specific application method. The daily workday air samples represent the cumulative personal solvent exposure during a full work shift (~8 h), including both painting activities and non-painting activities. The break time for lunch was recorded but not included in the time duration when calculating the air concentration for both activity-specific samples and daily workday samples. No record of whether shorter breaks were taken was documented.
Comparison of two types of air samples from bridge painting
Sample analysis and quality controls.
A standard analytical procedure (NIOSH 1500/1501) was used to analyze the air samples within 1 month of their receipt. Briefly, the organic chemicals were desorbed from the charcoal pads of badges using a mixture of acetone:CS2 (2:1 in volume) and analyzed by Gas Chromatography - Mass Spectrometry (GC-MS, HP6890) operated in the split mode. Twenty-seven major chromatogram peaks were identified in both activity-specific samples and daily samples based on a comparison of retention times and mass spectra to peaks from a calibration standard, Japanese Indoor Air Standards Mix (Supelco, 47537-U). The chemical compounds identified in the air samples were grouped into four classes of compounds: aromatic compounds, esters, alkanes, and ketones plus alcohols represented as a ketone group. A chemical with a concentration lower than its method detection limits (MDL) was assigned ½ MDL as its concentration.
One field blank was collected within each field trip for activity-specific samples and one field blank was collected along with the three daily samples returned by each subject by mail (). All the field blanks were analyzed following the same procedure as the air samples. To monitor any potential loss during the mailing process, 5 ng of each target compound was spiked onto five OVM badges in the laboratory. These spiked badges were sent out and mailed back within 1 week. The recovery rates of the spiked standards from these five badges, which were considered as positive controls, ranged from 71 to 90%, so no correction for losses of compounds during collection or storage of badges was applied to the samples.
Auxiliary information collected
For the activity-specific sampling, study personnel recorded information on application methods, working environments, paint brands, and paint coatings. For the daily solvent exposure monitoring, each participant completed a simple log sheet describing the paint brands, application methods, overall sampling duration, and duration for each painting activity. The completed log sheets and the collected air samples were returned to us by mail. Meteorological conditions for both activity-specific samples and daily samples were time-averaged values and obtained from the National Climatic Data Center.
Three common application methods, spraying, rolling, and brushing, are used for bridge painting. Current regulations require spraying to be conducted in a confined working space to minimize environmental release of toxic compounds. Wearing protective equipment, both half-face mask and appropriate clothing, is mandated during spraying operation. Rolling and brushing are more often conducted in an open working space. Rolling is a more efficient and even painting method than brushing. On the other hand, brushing is good to use on uneven surfaces or the spots hard to reach (e.g. corner or very narrow areas).
A three-coat system (primer, intermediate, and finish coatings) using a single paint brand is currently the most widely used system for bridge painting (National Steel Bridge Alliance
, 2010). The major chemicals present in different coating layers vary due to their specific function in the painting process. But the chemical compositions in each coating layer are quite similar across different paint brands. Thus, rather than paint brands, the influence of paint coatings on the VOC emissions from painting was assessed.
Activity-specific exposure samples were used to explore the associations between the air concentrations of four groups of compounds and the potential influencing factors, including painting method, paint coating type, meteorological condition, and if the painting was done in a confinement (SigmaPlot for Windows version 10.0). To account for any potential interaction among these factors, a multivariate regression model then was constructed based on activity-specific samples. Since not all compounds were present in all samples, there were too few data points to fit a multivariate linear regression model separately to each of the four groups of compounds. Thus, multivariate analysis was only applied to the total volatile organic compound (TVOC) concentrations (SAS for Windows version 9.1). There were between three and five activity-specific exposure samples collected on a single sampling day. To account for between-day variability, a mixed-effect model was constructed with sampling day as the grouping factor.
A three-step model selection procedure was performed in this study based on Akaike’s Information Criterion (AIC). AIC score, which takes into account both the goodness of fit and the complexity of an estimated statistical model, was used as an evaluation criterion for model selection. Briefly, we first constructed simple mixed models for each individual factor separately. Then, a multivariate mixed model was constructed including the factors that were statistically significant from the individual models. Finally, the multivariate mixed model was compared with the individual models with significant variables and a full mixed model with all factors to select the best model based on the evaluation criterion. Lower values of AIC indicate the preferred model, i.e. the one with the fewest parameters that still provided an adequate fit to the data.
The prediction ability of the mixed model was assessed using an independent dataset, the daily exposure samples. The daily exposure samples included the solvent exposure from painting activities and non-painting activities within one work shift, collected from different painters, and on different days than the activity-specific samples. The predicted exposure level for each painting activity in daily samples was estimated from the mixed models separately, which assumed independent contributions to the solvent exposure from sequential painting activities. The total exposure levels for each daily sample with multiple painting tasks were calculated as the sum of the time-weighted predicted values for each painting task from the model.