Impact of a pollution breach at a coke oven factory on asthma control in nearby vulnerable adults

Abstract

Background:

Previous studies have related sulfur dioxide (SO₂) exposure to asthma exacerbations. We utilized the University of Pittsburgh Asthma Institute registry to study associations of asthma exacerbations between 2 geographically distinct populations of adults with asthma.

Objective:

Our objective was to examine whether asthma symptoms worsened following a significant fire event that destroyed pollution control equipment at the largest coke works in the United States.

Methods:

Two groups of patients with asthma, namely, those residing within 10 miles of the coke works fire (the proximal group [n = 39]) and those residing beyond that range (the control group [n = 44]), were geocoded by residential address. Concentrations of ambient air SO₂ were generated by using local University of Pittsburgh Asthma Institute registry air monitoring data. Factory emissions were also evaluated. Data from a patient historical acute exposure survey and in-person follow-up data were evaluated. Inferential statistics were used to compare the groups.

Results:

In the immediate postfire period (6–8 weeks), the level of emissions of SO₂ from the factory emissions increased to 25 times more than the typical level. Following the pollution control breach, the proximal cohort self-reported an increase in medication use (risk ratio = 1.76; 95% CI = 1.1–2.8; P < .01) and more exacerbations. In a small subset of the follow-up cohort of those who completed the acute exposure survey only, asthma control metrics improved.

Conclusions:

Real-world exposure to a marked increase in ambient levels of SO₂ from a pollution control breach was associated with worsened asthma control in patients proximal to the event, with the worsened control improving following repair of the controls. Improved spatial resolution of air pollutant measurements would enable better examination of exposures and subsequent health impacts.

Patients with asthma tend to be more sensitive to development of symptoms from poor air quality than healthy populations. Overwhelming evidence that particulate matter (at levels of 10 ppb and at levels of 2.5 ppb)^1,2 and gases such as nitrogen dioxide^3,4 and ozone^5,6 can worsen asthma control and lead to exacerbations has accumulated. Although the mechanisms remain inadequately defined, experimental exposure to various air pollutants increases oxidative stress, which can contribute to greater airway reactivity and bronchial inflammation.^7–9 Although much of the recent work has focused on traffic-related emissions common in urban settings,^10,11 major industrial point sources remain in this country, including coal-fired power plants and various manufacturing plants, as well as plants related to the steel industry.

Significant industrial point sources of emissions in Allegheny County, Pennsylvania (home to Pittsburgh), that are related to its legacy as a steel industry center continue to exist. Steel manufacturing requires the use of highly refined coal, or coke, as fuel for blast furnaces to produce steel. Coke is generated from extreme heating of coal under anaerobic conditions to volatilize and remove impurities, of which hydrogen sulfide (H₂S) and, to a lesser extent, sulfur dioxide (SO₂) are major components.^12,13 Although the primary form of sulfur is in reduced form as H₂S, fugitive emissions of H₂S can spontaneously oxidize to SO₂ in the air or following deliberate flaring of coke oven gas, thereby contributing to release of SO₂ into the atmosphere. Although pollution control equipment collects and eliminates most H₂S emissions, disruption of these controls, as was seen following a catastrophic fire at the US Steel Clairton Coke Works plant (the largest in the United States) can (and did) render this equipment inoperable. Consequently, the coke works burned a large volume of coke oven gas to eliminate the noxious H₂S and released an unprecedented amount of SO₂ into the environment over the 102 days before repairs to the damaged pollution control equipment were completed.

SO₂ is a recognized air pollutant that is monitored and regulated by the US Environmental Protection Agency (EPA). SO₂ is biochemically transformed in the atmosphere, and on the airway mucosa, SO₂ can oxidize to sulfurous acid and sulfuric acid, which act as powerful lung irritants.^14–17 Also formed are sulfites and bisulfites, which have been reported to increase mucus production in bronchial airways.¹⁸ These known toxic effects led to studies that associated SO₂ with both asthma exacerbations and worsened lung function, including lower lung function in children.^10,19,20 However, other studies have failed to show such an effect.²¹

On the basis of this convergence of acute large-scale increases in a known toxicant in a defined locale, prior controversies regarding SO₂ and asthma, and our institute’s ability to rapidly engage patients with asthma in the regions affected by the fire, we hypothesized that compared with individuals living further away from the coke works, patients living within close proximity to it would acutely demonstrate a measurably higher asthma symptomology. We correlated these results with local air quality (in particular, SO₂ levels) during and after the fire.

METHODS

The participants (n = 83) were recruited from The University of Pittsburgh Asthma Institute registry (AIR). This registry was established in 2007 and contains approximately 2200 patients with physician-diagnosed asthma. All registrants had asthma questionnaire data and agreed to be contacted in the future for additional studies, and most of them had baseline pulmonary function testing from time of enrollment. Immediately after the fire, a 13-question environmental health survey (EHS) (see Data File E1 in this article’s Online Repository at www.jacionline.org) was developed for distribution to existing patients in the AIR who had been recruited to the study. A Consolidated Standards of Reporting Trials diagram outlining recruitment is presented in Fig 1, and a map of the study area and patient locations within that area is presented in Fig E1 (available in this article’s Online Repository at www.jacionline.org). The buffer was set to reflect the municipalities listed as “potentially affected” by the health department, in addition to the natural break in the distribution of patients in the AIR who resided around the Pittsburgh area. Recruitment began with the identification of all patients in the AIR who resided within a 10-mile radius of the coke works fire (CWF) (n = 177). Second, individuals included in the AIR (n = 1920) were identified as potential controls if they lived distal to the CWF (ie, those residing >10 miles from the plant). Participants were ordered on the basis of their most recent involvement with the asthma institute and then contacted. Because of the time-sensitive nature of the event, recruitment was stopped after slightly more than 200 people had been contacted,leading to more than 40 participants in both the proximal and distal groups from whom rapid consent was obtained by completed phone surveys. To obtain a snapshot of their health during the immediate aftermath of the CWF, the patients with asthma were asked to recall information specific to the 4 weeks immediately before the date of the CWF.

FIG 1.

Consolidated Standards of Reporting Trials flow diagram outlining study recruitment. SW, Southwest.

Following recruitment, baseline demographic, clinical, and physiologic data were pulled from the historical AIR. An asthma severity designation of mild or moderate-to-severe was assigned to each participant by using a combination of FEV₁ percent predicted, prescribed inhaled dosage of corticosteroids and/or oral steroids, and symptom type and frequency. The initial cohort recruitment and questionnaire administration occurred between February 1 and 20, 2019, which represented the period corresponding to the pollution control failure and high emissions.

After the pollution control equipment was operational (ie, 2 months later), participants were invited to return to the asthma institute in person to repeat the EHS; complete an updated asthma control questionnaire; and undergo additional pulmonary function testing and determination of exhaled nitric oxide value to assess asthma severity, control, and inflammation.²² The University of Pittsburgh’s institutional review board approved all studies.

SO₂ emission and regulatory air monitor data

Daily SO₂ emission estimates from the coke works corresponding to the first 6 months of 2019 were obtained from the Allegheny County Health Department. The emission data are reported as tons per day and include the amount of SO₂ released into the atmosphere directly through a variety of processes (eg, fugitive, flaring of coke gas), as well as through secondary conversion from reduced H₂S. To determine whether elevated emissions during the time of the CWF translated into changes in ambient air quality, we compared readings at EPA regulatory air monitors proximal and distal to the CWF.

Statistical analysis

Allegheny County Health Department air monitoring data were analyzed by using the nonparametric Kruskal-Wallis test followed by the Dunn multiple comparisons test. EHS responses were set up in contingency tables, and risk ratios (RRs) were calculated by using the following formula: RR = Rate [1]/Rate [2], where Rate is the proportion in the group with the condition present; CIs accompanied the ratios. Chi-square tests of association and the Fisher exact probability test were used in combination with the RR to determine statistical significance between group responses. Continuous data (emission, age, and lung function) are reported as means plus or minus SDs. Mann-Whitney U tests were performed on continuous nonparametric data. The acute survey results (before the control room repair) were compared with the followup responses (after the control room repair) by using paired t tests. P values less than .05 were considered significant. Data analysis and graphs were generated by using JMP Pro14 software (SAS Inc, Cary, NC) and GraphPad Prism 8 (GraphPad Software, San Diego, Calif).

RESULTS

In the immediate aftermath of the fire, SO₂ emissions from the coke works averaged between 40 and 50 tons per day. At times (especially in the months of March and April), daily emissions exceeded 50 tons per day (which was 25 times higher than the typical levels). Immediately following repair of the pollution control room, daily emissions of SO₂ declined dramatically to levels that were less than 5% of those reported during the breach (Fig 2).

FIG 2.

Each point represents a single day of coke works factory emissions. The larger black dot indicates the date on which the pollution control equipment was damaged and rendered inoperable. The gray triangle indicates a second fire that damaged pollution controls for a single day. The horizontal error bars indicate the initial baseline study and follow-up periods.

In 2019 (coinciding with the large emissions of SO₂ from the coke works during the pollution control breach), daily maximal SO₂ readings and SO₂-dependent air quality index values were elevated at monitor 1 (the monitor closest to the coke works) as compared with in the previous 2 years (Fig 3). In contrast, time-dependent changes in measurements of SO₂ in ambient air were not observed at either of the other 2 monitors (see Table E1 in this article’s Online Repository at www.jacionline.org). Thus, enhanced emission(s) of SO₂ during the pollution control breach were associated with measurable changes of SO₂ content in ambient air in close proximity to the coke works.

FIG 3.

A and B, The 2 proximal regulatory air monitors located within 10 miles of the CWF and 1 distal control monitor beyond the 10-mile range, as well as relative location relative to the Clairton Coke Works. Daily ambient SO₂ expressed as median maximum () and air quality index values (B) for 2019 and after the CWF compared with the historical data in 2017 and 2018 for reference. AQ, Air quality; NE, northeast; NW, northwest

Impact during the acute exposure (during loss of pollution control system)

To determine the impact of these increased emissions on asthma control, we recruited and surveyed proximal and distal (control) cohorts of patients with asthma both before and after the pollution controls had been repaired. Historical registry data (collected 1–8 years before the fire) indicated that both the proximal and distal groups were generally well matched with regard to a variety of demographic factors, including age, race, sex, and type of health insurance (Table I). Both educational attainment and marital status tended toward being or were statistically different across the groups (P = .07 for education and P = .04 for marital status), supporting the idea of higher socioeconomic status (SES) in the distal/control group. Despite this, the groups were well matched from the standpoint of known confounders of asthma such as exposure to or history of smoking and body mass index. Notably, there was a significant difference in historical FEV_₁ percent predicted and asthma-related emergency department visits compared with the values for the control group (Table I).

TABLE I.

Baseline comparison of the control and proximal cohorts using Asthma Institute Registry data

Characteristic	Distal control (n = 44)	Proximal group (n = 39)	P value
Age (y), mean ± SD	45.1 ± 14.6	45.6 ± 14.6	.41
Female sex, no. (%)	39 (84.0)	32 (86.0)	.76
Nonwhite, no. (%)	16 (36)	16 (41)	.82
Body mass index, mean ± SD	31.9 ± 11	33.1 ± 10.5	.41
Asthma indicators
Mean FEV1 (% predicted), no. (SEM)	84.8 (7.1)	77.2 (12.1)	.04*
Moderate-to-severe asthma, no. (%)	18 (41)	19 (49)	.51
Health insurance, no. (%)
Medicaid	12 (27)	14 (36)	.47
Other	32 (72)	25 (64)
Employment status, no. (%)
Outside of home	31 (70)	24 (62)	.36
Education, no. (%)
High school graduate or less	4(9)	10 (26)	.04
Some college or college graduate	40 (90)	29 (74)
Marital status, no. (%)
Single or other	22 (50)	27 (69)	.07
Married	22 (50)	11 (28)
Smoking status, no. (%)
History of smoking	13 (29)	9 (23)	.62
Secondhand smoke	8(18)	11 (28)	.31

Data represent registry metrics at time of enrollment into AIR (N = 83).

^*P < .05.

Given the need for rapid assessment of a potentially “at-risk” population, the first patient questionnaires given to these 2 cohorts were administered by telephone. Not surprisingly (and consistent with theSO₂ data),more patients in the proximal group than in the distal/control group reported the presence of rotten egg smells near their home and more were aware of the CWF event (Fig 4). During the period corresponding to the pollution control breach, more individuals in the proximal group than in the distal control group also self-reported an increase in asthma exacerbations (question 5) (RR = 1.7; 95% CI = 1.1–2.8; P < .01), as well as increased medication use (RR = 1.4; 95% CI 1.0–2.0; P < .05) (Fig 4). Nearly all of the patients in the proximal group (77%), as opposed to 48% of the distal controls, reported that their asthma had worsened and attributed the worsening to air pollution.

FIG 4.

Graph depicting difference by percentage of respondents (n = 83) who answered yes between the control and proximal groups. *P < .05; **P < .005; ***P < .0005. Rate is the proportion in each group with responses of yes. RR = Rate [1]/Rate [2] in the proximal versus the distal groups. Survey questions are listed below the graph and correspond to the question number along the y-axis. ER, Emergency room; ID, identifier; Q, question.

To account for boundary effects of potential “edge bias”²³ to our 10-mile buffer zone that defined the proximal group, we examined whether there were differences in those residing less than 5 miles (n = 13 [33%]) from the CWF and those residing 5 to 10 miles away (n = 26 [67%]). As expected, those living closer to the factory (<5 miles away) were more likely to notice industry smells (question 5, P = .01; question 6, P = .009) and were more aware of the CWF (question 8, P = .0002) than were those in the 5- to 10-mile group (see Table E2 in this article’s Online Repository at www.jacionline.org). However, there were no differences in asthma control between the near and far proximal groups in the aftermath of the fire.

Lastly, because general news coverage of the CWF may have selectively influenced survey responses, we compared responses within the proximal group that was based on their expressed knowledge of the CWF (54% aware vs 46% unaware). No differences in survey responses regarding medication and/or asthma-related exacerbations based on stated awareness were observed. Of note, those reporting less awareness of the CWF tended to be more likely to report having Medicaid (53% [P = .16]) as their primary insurance, and they were less likely to have a college degree (71% [P = .07]) than those who were aware of the CWF, suggesting lower SES.

Follow-up (after repair)

To assess the potential impact of changing exposure to SO₂ more directly on measures of asthma control and to determine the relationship of location to current pulmonary function and airway inflammation, 2 months after repair of the pollution control room we recontacted those patients who had completed the initial survey to arrange an in-person visit. Of the original cohort, 57% (22 from the proximal group and 25 from the distal control group) returned to the asthma institute after resolution of the breach for an in-person evaluation. There were baseline differences between those who completed only the phone survey and those who completed both the phone and in-person surveys, particularly within the proximal cohort (see Table E3 in this article’s Online Repository at www.jacionline.org).

Those in the proximal group who agreed to the follow-up visit tended to report better asthma control at the time of the phone survey than did those who answered only the initial phone survey (see Table E3). Although there were no differences in asthma severity or FEV₁ value, those who returned reported a higher percentage of employment outside the home and a lower percentage of having Medicaid insurance than did those who did not return for the in-person survey, which is consistent with potential differences in SES. In contrast, within the distal/control group, the initial contact–only and follow-up subgroups differed solely in terms of secondhand smoke exposure (see Table E2).

Paired testing of those who completed both the phone and in-person questionnaires was used for initial and follow-up comparisons of the questionnaire responses. Our analysis of the proximal group indicated persistent industry smells (question 6) and improved asthma control (questions 9 and 11) (Fig 5, A), which coincided with the improved factory emissions. There were no longitudinal changes in responses to the asthma control questions in the distal/control group. However, a heightened perception that adverse air quality may sometimes affect their asthma control was observed in the members of the distal/control group (question 7), perhaps on account of the repeat nature of the survey (Fig 5, B). Lastly, all participants who came to the asthma institute were evaluated by determination of fraction of exhaled nitric oxide, FEV₁, and asthma control questionnaire results. Overall, the proximal group had worse asthma control and more inflammation than the distal controls did, although the small sample size limited significance (Table II). A paired analysis of historical and follow-up mean FEV₁% predicted value over a 2.5-year period revealed no change in the control group (P = .7) but values of 73.7 versus 76.1 (P = .2) within the proximal group.

FIG 5.

A and B, Comparison of the responses of yes with the responses to our survey at baseline following the CWF relative to the follow-up responses after repair of the factory. See Fig 4 for the survey questions. *P < .05. Q, Question.

TABLE II.

Physiologic results at follow-up visit between the control and proximal groups (n = 47) after the Coke Works’ pollution controls were repaired

Patient characteristic after factory repair	Distal control	Proximal group	P value
Feno, mean ± SD	20.3 ± 12.6	27.1 ± 19	.2
Asthma control, mean ± SD*	1.32 ± .95	1.61 ± 1.2	.3
Mean FEV1 (% predicted), no. (%)	82.4 (16.7)	76.1 (19.6)	.3

^*According to the air quality questionnaire.

DISCUSSION

This study provided the opportunity to examine short-term asthma outcomes relative to an individual’s proximity to a major industrial event that resulted in emission of 20 times more SO₂ than is typically discharged per day over a period of several months. Key findings include measurable differences detected in self-reported asthma control between those living proximal to and distal from the CWF that corresponded to large changes in factory emissions and air quality. Despite the persistently poor air quality in this region, historical air monitoring data support our assertion that air quality during our study period was exceptionally poor and likely contributed to acute negative effects on asthma outcomes in our small cohort of individuals with asthma. To our knowledge, this study is the first of its kind to describe the short-term impact of real-world factory emissions from a coke works plant on well-characterized patients with asthma and to identify decrements in air quality–related health outcomes within a population of adults with asthma.

Short-term controlled (chamber) exposures to SO₂ have been shown to augment bronchoconstriction in patients with asthma.²⁴ Although the mechanisms of its effects are unclear, SO₂ can act as both an oxidizing agent and a reducing agent, depending on its environment. Exposure to SO₂ has been reported to increase mucus production and lead to bronchoconstriction through unclear molecular pathways.²⁵ Exposure has also been linked to exacerbations and reductions in pulmonary function (in particular, FEV₁ value).^25,26 In adults, higher SO₂ exposure during the winter months was associated with lower lung function in individuals with poorly controlled asthma.²⁷ Similarly, SO₂ levels were high in the winter months of our study and may have affected lung function. Other studies, however, failed to demonstrate consistent associations.²⁸ This may be attributable to the spectrum of study designs in the literature, combined with different outcomes. Furthermore, some individuals with asthma may respond to SO₂, whereas others may not.²⁹ This suggests a rationale for deeper phenotyping of these patients to identify potential SO₂-hyperreactive patients with asthma. Lastly, a recent study has implicated ambient SO₂ exposure to increased emergency department visits overextended lag periods.³⁰ These studies suggest that a multifaceted approach is needed to create a holistic picture of the impact that SO₂ has on adults with asthma.

Although a direct measure of SO₂ was not obtained at the homes of the participants, data from regulatory-grade air monitors in close proximity to the plant at which the CWF occurred were publicly available. Despite persistently high levels of factory SO₂ emissions (over 102 days), the most proximal monitor indicated that the EPA threshold (75 ppb) was exceeded on only 9 days. This is likely due to a combination of meteorologic and topographic factors that in addition to emissions, influence ground-level SO₂ concentrations. The highest SO₂ concentrations occur during meteorologic inversions, which are periods when vertical mixing in the atmosphere is limited and emissions are concentrated near the Earth’s surface. The net result is that the actual ground-level SO₂ concentrations are a complex function of multiple factors influenced by both emissions and meteorology that may not be reflected in monitor measurements.

Five of the days on which the thresholds were exceeded occurred during the time frame of the acute survey following the CWF (ie, from January 1 through February 20, 2019 [the baseline survey period]) (see Fig E2 in this article’s Online Repository at www.jacionline.org). Importantly, these instances of the threshold being exceeded were reported only at the most proximal monitor, suggesting that the location of monitoring equipment is very important. Communities that rely on a few citywide air monitors are likely unable to model a population’s true exposure. This is especially true in communities with varying topography and/or with major point sources that can have disproportionate impacts on areas immediately downwind. In-home monitoring may be needed to better reflect the more localized impact of point source pollution.

The acute health effects measured throughout this study trended with air quality and emission changes. First, those living close to the CWF reported worse outcomes than did those who lived further away during the same period, corresponding to elevated emissions. This minimizes the chance that weather or prevalence of seasonal respiratory infections contributed the differences. Supporting this, longitudinal data on those patients who completed both phases of the study indicated that asthma control improved from the immediate postfire period in the proximal group, when pollution controls were operational and air quality improved.

To control for information bias, we examined whether participants who knew about the CWF answered their questions differently from participants who did not know about it. Responses did not differ between those with and those without knowledge of the fire. However, those who were less likely to be aware (nearly 50% of the proximal group) of the CWF also reported a lower SES.

This study demonstrates the benefits of having a preexisting interactive asthma registry of well-characterized patients with asthma. Participant consent gave us the ability to quickly gain access to patients following an environmental disaster. This study also revealed some potential reasons why some patients may be more likely to follow up or participate in environmental health research than others. Patients with poorer SES indicators and less education may have transportation issues or work in occupations with limited paid time off, and therefore, such patients are less likely to visit the asthma institute. Because participant dropout is an unfortunate reality of research studies, our aim was to better understand the population that did not complete our study. The asthma registry data allowed us to compare various SESs and other metrics between groups. Participants who did not return to the asthma institute for the follow-up visit tended to have poorer asthma control, and according to the baseline survey, these participants were more likely to report a clinic, emergency department, or ambulance visit because of their asthma. Additionally, this subgroup tended to have poorer SES indicators. These factors may have influenced communication and transportation efforts within this subgroup. Unfortunately, because this subgroup did not complete the follow-up questionnaire, we were unable to discern how their asthma control may have changed with improved air quality.

Some incidental findings of our study alluded to the potential for more chronic asthma–related issues within this cohort based on their residential proximity to the coke works. When considering the historical registry data, we identified significant baseline differences in mean lung function between those proximal to the coke works and those distal from it. Although multiple factors could have contributed to these differences, the 2 groups were generally well matched for baseline demographics, as well as for smoking status and SES. This suggests that chronic exposures may adversely affect lung health, and it warrants further study in larger patient surveys.

This study focused specifically on emissions associated with an operational failure at a single point source near Pittsburgh. Excess emissions after the CWF, modified by meteorology and topography, affected ground-level air pollution as well as the resultant exposure and health impacts cataloged here. Exposure assessment was limited because the sparse network of air pollutant monitors did not enable neighborhood-level air quality assessment. Communities in other cities are also affected by nearby point sources, and exposure estimates in those locations may be similarly affected by a lack of nearby monitors.

Study limitations

As with all human and environmental studies, the potential for confounders exists. The largest concern is the small cohort, in which confounders may have influenced the observed differences. Overall, however, the 2 groups were well matched, and in the proximal group, differences improved in concert with improved air quality when other factors were constant. Our study population was not a random population; rather, it was derived from an existing asthma registry. However, given the historical information on these patients, including the results of standard lung function testing, it is very likely that these patients did indeed have asthma, as compared with individuals contacted through use of random dialing approaches.³¹ Importantly, only 50% of patients returned to the asthma institute for full evaluation after the repairs had been completed, and consequently, they returned in a different season (winter vs spring). There were substantial differences between those who returned and those who did not, which may explain the resolution of all clinical differences between the 2 groups at the later follow-up.

Conclusion

Data from this study reveal an association between acute exposures to increased ambient levels of SO₂ and worsened asthma control in a potentially vulnerable population living close to the source of the pollution. However, it also supports the loss of acute effects when EPA-established air quality levels are reestablished. Although our data additionally suggest that chronic long-term exposure may negatively influence lung function, further study is needed.

Clinical implications:

The study data confirm an association between adverse asthma outcomes and acute increases in ambient SO₂ level. Clinicians should be aware of the potential for environmental pollutants such as SO₂ to influence asthma control.

Abbreviations used

AIR

Asthma Institute Registry

CWF

Coke works fire

EHS

Environmental health survey

EPA

US Environmental Protection Agency

H₂S

Hydrogen sulfide

RR

Risk ratio

SES

Socioeconomic status

SO₂

Sulfur dioxide