Introduction

Auto repair shops, which often provide both mechanical and autobody repair, are commonplace – employing 1% of the US workforce1 Click or tap here to enter text.. In these shops, workers are exposed on an almost-continuous basis to volatile organic compounds (VOCs) from common activities (e.g., cleaning brakes). VOCs are a class of thousands of chemicals that can volatilize or evaporate or quickly at normal temperatures (boiling point 50–260 °C) and can be inhaled. After evaporating, they can react with other chemicals to form new ones (e.g., ozone). VOCs come from a range of sources, from gasoline to nail polish remover, and include such chemicals like acetone or ethanol.

VOCs have well-documented, often irreversible health effects, including asthma2,3, cardiovascular disease4, cancer5,6, and cognitive and neurological symptoms7,8. Despite potential health risks and the substantial exposed population, most studies have focused only on benzene9,10 or isocyanate exposures in auto body painting11, with no studies have examined representative auto shops exposures, aside from one or two shops over limited time and activities.

Compounding this lack of data, exposures are likely to be more acute in small businesses (< 100 employees), as they are less likely to hire industrial hygiene (IH) consultants12,13. Of all auto repair workers, 56% are in shops with < 20 employees1, yet the Occupational Health and Safety Administration (OSHA) does not have capacity to regularly assess this many small operations12. Additionally, Latino workers are over-represented in the small business workforce14 and in low-wage jobs with increased risk of occupational injury15. Meanwhile, they are less likely to trust government agencies to ask for assistance16, and there are few linguistically and culturally appropriate occupational health materials17. To better understand VOC exposures to this vulnerable yet ubiquitous worker population, we investigated how personal and shop VOC exposures during entire shifts related to worker activities and shop characteristics in small auto repair shops in southern metropolitan Tucson, AZ with primarily Spanish-speaking workers. In addition to detailing types and magnitudes of VOC exposures in a marginalized population, our study may help inform intervention strategies and direct more targeted studies of exposure factors.

Methods

Recruitment

As described previously, participants at local auto repair shops with < 25 employees were recruited door-to-door by Sonora Environmental Research Institute, Inc. (SERI) promotoras or community health workers18,19 in south metropolitan Tucson, AZ in July, August, and November of 2018. To obtain representative exposures during entire shifts while minimizing participant burden, promotoras signed up workers or owners who expected to be at the shop most of the day. To assess personal VOC exposures, the study team monitored four work shifts per shop, which could include any combination of single or multiple shifts for participants. Upon consenting, promotoras administered participant demographic and background surveys. All verbal and written communications were completed in the participant’s language of choice (i.e., Spanish or English), and site visits for VOC monitoring were scheduled at the business’ convenience. The University of Arizona Human Subjects Protection Program approved all human-subjects materials related to the study (#1709821542).

Worker activities and other exposure characteristics

Worker activities, room, ventilation, nearby activity, and any other relevant notes for exposure assessment were recorded in real-time by study personnel from the University of Arizona18,20. Since certain specific activity and ventilation types had very few observations, specific activities were combined into broader categories (Table S11). Broad activities included: 1) administration, 2) body work, 3) taking a break, 4) cleaning, 5) engine work, 6) fluids work, 7) tire work, and 8) unknown (Table S1). Ventilation conditions were categorized sequentially by availability of outdoor air: 1) outside, 2) open bay/garage door, 3) fan, 4) swamp cooler, 5) AC, and 6) other (Table S2).

VOC measurements

As detailed previously18, personal total VOC exposure was measured using a calibrated ppbRAE 3000 monitor (RAE Systems, Inc., San Jose, CA) with tubing affixed in the worker’s breathing zone. The monitors were placed on the participant on a belt, in a sling bag, or near them as they worked. Monitors logged 20-s interval averages, and concentrations are reported in isobutylene units (parts per billion [ppb]). To identify specific VOCs present in the workplace, Summa canisters (Restek, Bellefonte, PA) were placed in the workspace on each sampling day. Specific VOCs from each canister were measured using the US Environmental Protection Agency TO-15 analysis method21 at the TestAmerica Lab (Phoenix, AZ), along with additional tentatively identified compounds. Time-weighted average concentrations of each VOC were calculated. Canisters were deployed in the main work area, or where the most worker activity was expected. In two shops, samples were taken in the main work area on two consecutive days; in another shop, duplicate samples were taken both in the main repair area and in the paint booth on the same day.

Total VOCs analysis

Concentrations below the limit of detection (LOD) of 1 ppb were replaced with the LOD/√2 ≈0.707 ppb22. Because measurements were correlated in time, these data were combined into time weighted averages for each sequential change in activity-ventilation combination in a shift. These data were natural log-transformed to reduce the skew of the distribution.

Specific VOC analysis

Ambient concentrations for > 70 specific VOCs were measured throughout the shift using Summa canisters. A hazard score for each canister was calculated based on US EPA Inhalation Reference Concentration (RfC) or inhalation cancer unit risk factor if there was no RfC for each specific VOC. Barring these, the oral reference dose was used. Briefly, the specific VOC concentration was divided by the corresponding VOC reference concentration, then the values were summed for all specific VOCs for the canister. Additional detail can be found in previous publications18,23.

Statistical analysis

To examine the relationship between total VOCs and work activities, ventilation types, shops, and shifts, we fit a linear mixed effects model of log-transformed VOCs on the fixed effects of broad activity and broad ventilation and the random effects of shop and shift within shop. Shifts were placed within shops in the model to account for their nested structure, and each shop and shift within a shop was allowed to have a different intercept. To investigate the size of the direct of fixed predictors, we partitioned the variance into conditional (fixed and random effects) and marginal (only fixed effects) R2 values using the partR2 package24 following the Nakagawa method25. A p-value < 0.05 was assumed to be statistically significant. Data cleaning and statistical analyses were performed using R Statistical Software (R version 4.4.2 (2024–06-14). For data manipulation and linear mixed effects models, we used the tidyverse26 and lme4 packages27, respectively. All packages are listed in Supplemental Material.

Results

Participant demographics

Community health workers enrolled 11/23 (43%) auto shops they visited. One shop completed initial recruiting paperwork but opted out of the study before any sampling occurred. Another shop did not allow study staff to view the work area, resulting in data with severely limited information on activity, ventilation and other exposure characteristics. Among recruited shops, 6/10 (60%) of eligible owners and 16/24 (75%) workers participated. Additional information on recruitment has been reported previously18. Of the nine shops, 2 were sampled in July, 6 in August, and 1 in November, 2018. While 29 participants completed the demographic survey, only 22 agreed to wear monitors. The participants were between the ages of 19 and 70, and were mostly male (21/22 = 95.5%), White (21/22 = 95.5%), Latino (16/22 = 72.7%), and employees (16/22 = 72.7%) (Table 1).

Table 1 Participant demographics for monitored participants from nine auto shops.
Full size table

Total VOC concentrations

Usable total VOCs samples were collected from 22 participants over the course of 35 shifts at nine auto shops. In one shop (007), the owner barred the study team from sampling on a second day (i.e., shifts 3 and 4). As described previously20, observations that occurred during flow faults (i.e., ppbRAE monitor airflow through tubing was impeded) or equipment malfunctions were removed from further analysis. After data cleaning, 36.5% of observations were below the limit of detection (LOD). Frequencies of activity-ventilation combination (e.g., cleaning parts with bay door open, administrative work with fan on) spans ranged from 4 to 55 combinations per shift, with a median of 27 (Table 2), with shop 010 having the highest frequencies (two shifts with 55 combinations each). Once the data were aggregated over the activity-ventilation span, 17.5% of the observations were below the LOD.

Table 2 Total activity-ventilation combinations by shop and shift number.
Full size table

The activities with the highest average VOC exposures were bodywork, fluids, and tire work (geometric mean [GM] = 419 ppb, 295 ppb, 192 ppb, respectively), while administration, taking a break, and unknown activities were lowest (GM = 50 ppb, 45 ppb, 48 ppb, respectively) (Fig. 1). While cleaning activities were not the highest on average, they had the highest maximum value, reaching 396,801 ppb, above body repair’s maximum (180,393 ppb).

Fig. 1
figure 1

Total VOC concentrations stratified by activity and colored by shop. The first quartile (Q1), median, and third quartile (Q3) are shown by black horizontal lines; a black diamond shows the geometric mean. Each data point represents the arithmetic mean for an activity-ventilation span, resulting from aggregating over each broad activity-broad ventilation span in sequence.

Full size image

Exposure patterns were characterized by episodic peaks, often an order of magnitude higher than the background, as exemplified in Fig. 2. The highest peaks (> 2 orders of magnitude above background) were evenly divided among the spraying brake cleaner and body work activities in 22/35 shifts (Figure S1).

Fig. 2
figure 2

Real-time total VOC concentration colored by activity with peaks labeled with specific activities for one shift.

Full size image

Unlike activities, which had a relatively even distribution of observations across types, ventilation characteristics were almost exclusively open garage or bay door or outside (Fig. 3), which also had the lowest mean values (GM = 90 ppb and 85 ppb, respectively).

Fig. 3
figure 3

Total VOC concentrations stratified by ventilation and colored by shop. The first quartile (Q1), median, and third quartile (Q3) are shown by black horizontal lines; a black diamond shows the geometric mean. Each data point represents the arithmetic mean for an activity-ventilation span, resulting from aggregating over each broad activity-broad ventilation span in sequence.

Full size image

Relationships between total VOCs and activity, ventilation, shop, and shift

In a mixed effects model, activity and ventilation types with limited observations (i.e., the tire activity and AC, fan, swamp cooler, and unknown ventilation conditions) were excluded (Tables S3 and S4). A model including these produced similar results. In comparing variances of random effects, that for shift within shop was 2.4 times larger than that of between shops (Table 3). The model itself only accounted for about 50% of the variance in total VOC concentrations (35% and 15% for shift within shop and between shops, respectively) (Table 3).

Table 3 Variances and standard deviations from the model of log total VOCs on the fixed effects of activity and ventilation and the random effects of shop and shift within shop.
Full size table

Activity was significantly associated with total VOC concentrations (p = 2.90 × 10–9), while ventilation was not (p = 0.32) (Table 4). Despite activity’s statistical significance, its direct effect was weak, accounting for 3% of the variation in VOC concentrations, compared to about 50% explained by random effects of shop and shift within shop (Table S5). Additional sensitivity analyses for shop characteristics and sample timing showed no associations with total VOCs.

Table 4 Analysis of variance (ANOVA) for the model of log total VOCs on the fixed effects of activity and ventilation and the random effects of shop and shift within shop with the Kenward-Roger degrees of freedom method.
Full size table

Specific VOC concentrations

Summa canisters were deployed in 9 shops (011 was excluded due to costs), with 2 duplicate samples taken in shop 003 and repeat measures in shops 006 and 009 (Fig. 4). Among Summa canister samples for specific VOCs, 30 VOCs were detected in any sample across all shops, with acetone and toluene found in all and benzene, ethylbenzene, and xylene in most (9–11 out of 14 samples; all but one shop). Hazard scores ranged from 0.01 to 9.9, with a median of 1.1, with the maximum value driven by 1,3-butadiene exposure. Other higher scores were brought on by other contaminants, including chloromethane (hazard score = 7.5) and tetrachloroethene (hazard score = 6.0). Duplicate samples in shop 003 generally had mixed agreement when it came to VOCs detected. For specific VOCs detected in at least one of the duplicates, main repair area and paint booth samples matched 45% and 33% of the time, respectively. The relative percent difference in hazard scores for the main repair area was 180% and 6% for the paint booth. When comparing samples taken in the same location on different days, shops 006 and 009 had 50% and 18% of the same VOCs detected, respectively, and relative percent differences of 49% and 173%.

Fig. 4
figure 4

Ambient concentrations of specific VOCs measured in at least one shop and calculated hazard score (last row) for each Summa canister by shop, date, placement location, and duplicate status (dup).

Full size image

Discussion

In this study, we measured real-time total VOC exposures for 22 workers over their entire shifts, often over multiple days, while recording detailed activities and ventilation conditions. We measured specific VOCs in 9 auto repair shops. For both total and specific VOCs, concentrations were highly variable. While total VOC exposures were related to shift within each shop and the shop itself, neither activity nor ventilation explained much variation. Overall, average total VOC exposures were an order of magnitude lower for activities not associated with VOC sources (e.g., administration GM = 50 ppb, taking a break GM = 45 ppb), compared to higher exposure work like body repair (GM = 419 ppb) or cleaning (GM = 165 ppb). While body work had the highest average exposures, cleaning had a higher maximum exposure (396,801 ppb).

In our study, workers were often exposed to high total VOC peaks along with relatively low background concentrations throughout the shift. The highest peaks (> 2 orders of magnitude greater than the background) occurred during spraying brake cleaner and painting activities in 22/35 shifts, with many resembling Fig. 3. While a similar study28 investigated several individual VOCs common in auto shops during a shift, the peaks were less than an order of magnitude above background (Figure S1). This pattern of a concentration spike followed by a decline to background within minutes may not be surprising given that most observed work was done outside or with an open bay door, providing constant mixing with VOC-free ambient air, resulting in no measurable VOC concentrations. In another study using the same sampling device, the mechanic was exposed to peaks one to two orders of magnitude lower than seen in our study during brake cleaning in a shop with open bay doors29. This study also noted the effectiveness of reducing exposures with bay doors open, with freely flowing wind from outside. With doors open, the amount of brake cleaner sprayed did not change exposures. Our results highlight a quandary in exposure science noted by Zartarian and colleagues30: What is the most representative way to characterize exposure?

Even though there are occupational limits that incorporate maximum concentration values (e.g., National Institutes of Occupational Safety and Health’s ceiling value) as well as time-weighted averages, these may not appropriately characterize the exposure’s health impact. Toluene, most notably as it was found in all shops, can have different neurological impacts in acute vs. chronic exposure scenarios in rat models31,32. Further, acute exposures can cause inflammation and tissue damage to both the lungs and brain in mice33 and psychological aftereffects 24 h after a single acute exposure34. While health impacts from varying exposure duration and concentration change with the contaminant35, further work (i.e., controlled exposures and toxicological or health effect assessment) should investigate the most appropriate exposure characterization. While this is beyond our scope, it may offer some explanation as to why activity did not explain much total VOC variation: the way we calculated total VOCs from a time-weighted average over an activity-ventilation span was dominated by low background concentrations, not peaks.

In modeling total VOC exposures, there was substantial variation in total VOC concentrations explained by shifts within shops, and to a lesser degree between shops. Meanwhile, fixed effects of activity and ventilation explained virtually no variance. As stated before, this could partly result from the effect of peaks of an activity, no matter how dramatic, being averaged out or diluted over the course of an entire activity-ventilation span. Future analyses could analyze activity as a function of peak magnitude and frequency to better characterize exposures that matter most in terms of health impacts. Alternatively, how activities are done, even by the same person in the same shop, may simply vary that much, even on a day-to-day basis. In our study, of the 35 total shifts, 26 of those were from monitoring the same participant on two different days, either on back-to-back days or in the same week, indicating that exposures for the same person may not be comparable on different shifts with similar activity, however this was not formally assessed. Another study illustrated how aerosol VOC exposure had no relationship to task, individual, or shop36. A paper by Egeghy et al. focused on benzene exposure found that the shift, not the worker, explained most of the variability10. The particulars of the same task (e.g., kind of degreaser used) may have also varied, blunting any relationship between total VOC exposure and activity type.

While we attempted to note product type, this was often impossible due to the products being kept in unlabeled containers or having their label obscured by residue on the outside of the container. However, it would be expected that workers in the same shop would use the same product and equipment (e.g., paint gun), though because of the difficulty of conducting timely and accurate product and equipment inventories in such shops, this could not be confirmed. Further, differences between shops only accounted for 15% of the total VOC variance. This suggests that shop-level product or process differences were important, but not as much as the variation in work between shifts, which was similarly illustrated by specific VOC results.

Among duplicate Summa canister samples taken in the main repair floor area and in the paint booth, chemicals detected matched less than half the time. While this disagreement is likely in part due to relatively low VOC concentrations (often < 10 ppb), there was often turbulent air (i.e., outdoors or by open bay doors) for potential differences in VOCs in such a small area. Deploying additional samplers on multiple shifts would likely put this variability into context. Similarly, in shops sampled on consecutive days in the same location (e.g., main repair area), few detected VOCs matched. This may reinforce how varied activities were for each day, including who is working where in the shop. Not surprisingly, the hazard scores were substantially different, especially those on the main repair floor, which has far more air mixing. However, the paint booth duplicates had nearly identical scores, albeit from different contaminants, likely resulting from adequate mixing throughout the chamber. Despite this general variability, two chemicals were found in all samples: acetone and toluene.

While toluene is expected, given it is part of the chemical ‘signature’ of gasoline of benzene, toluene, ethylbenzene, and xylenes (BTEX), the other constituents of the ‘signature’ were found in nearly every shop but in only 2/3 of samples. However, both toluene and acetone may be from degreasers, notably brake cleaner, as seen in a similar study36. This is of particular interest, as acetone can amplify the severity and duration of neurotoxic effects of n-hexane8, which was found in nearly every shop. While hexane has been largely removed from degreaser and paint thinner formulations, it has other sources (e.g., sensor cleaner, paint catalyst). Among the highest peak total VOC exposures (2–3 orders of magnitude above background average), 22/35 occurred during spraying brake cleaner or in painting operations. While just a single example with three chemicals in common tasks, this highlights how variable, complex, and potentially dangerous exposures can be in these auto repair shops, which are grossly understudied despite the large share of workers they employ.

A major strength of our project is the amount and level of detail on total and specific VOCs we collected among so many workers for full shifts, especially from such small shops in a very underserved and under-resourced community. Our study corroborated what others have found: worker VOC exposures tend to be episodic peaks that are highly variable among workers, shifts, activities, and shops. Despite the amount of data we collected, our analyses still may be limited by numbers of observations. We may also not be fully accounting for episodic peak exposures in total VOCs. Future analyses may examine alternative ways of conceptualizing and analyzing exposure to potentially elucidate relationships not shown in our current analyses. In addition, other factors that may impact personal exposures, notably nearby activities or product type used, could not be reliably recorded without disrupting workers.

In summary, after taking extensively detailed sampling and exposure factor data for entire work shifts in auto repair shops, we found substantial variability in total and specific VOC concentrations, especially between shifts in each shop and among shops themselves. Activities and ventilation did not explain much of the variation in the total VOC exposures, with shift and shop explaining half of the variation. Meanwhile, specific VOCs varied substantially in duplicate samples and in the same shops on consecutive days. Taken together, our findings show how complex the exposure environment is for auto repair workers. Future studies could hone data collection methods to account for more exposure factors (e.g., type of product, nearby activity) or monitor workers for more shifts over time to better model variation of VOC exposures. However, any additional data gathering should be done with feasible interventions with worker input in mind37, tailoring as closely as possible to both worker and shop alike, maximizing workplace safety improvement for 1 in 100 workers in the US.