|
|
|
|
|
|
||
Evaluation of Residual Respiratory and Other Health Effects from a Chlorine Release
Draft Final Report for Public Comments
Public Comments Period: December 1, 1999 - January 31, 2000
Send comments to:
ATSDR
1600 Clifton Road, E-31
Atlanta, GA 30333
Attention: Catherine Johnson
November, 1999
Division of Health Studies
Agency for Toxic Substances and Disease Registry
U.S. Department of Health and Human Services
Atlanta, Georgia 30333
ABSTRACT
The Agency for Toxic Substances and Disease Registry (ATSDR) conducted a follow-up study to evaluate residual respiratory, dermatological, ocular and neurological effects in community members exposed acutely to chlorine, as the result of a train derailment which occurred April 11, 1996. A total of 121 individuals who participated in phase 1 of this study were considered to be the exposed group, and 99 individuals selected from a nearby community served as a control group. Study participants completed a questionnaire, had blood drawn to measure IgE levels, underwent skin and eye examination by a physician, and had breathing tests (pulmonary function tests and methacholine challenge test) administered. The exposed and control groups were compared with regard to various self-reported symptoms as well as skin, eye, and pulmonary health.
Exposed subjects tended to be older and of lower socioeconomic status (less education and lower income) than control participants. Exposed subjects were more likely than control subjects to be current smokers, to have smoked 20 to 40 pack years, and to have been exposed to passive smoke. Exposed participants also appeared to have had more occupational dust exposure overall, particularly exposure to sand/silica dust and insulating materials. Occupational volatile organic compound exposure overall was similar among exposed and control subjects.
This study found an increased prevalence of nearly all self-reported health problems among exposed compared to control participants when they were asked about occurrence of the problem since the train accident. This included gastrointestinal symptoms, fatigue and sleepiness, aching muscles, dermatological problems, and ocular problems. Although more control than exposed participants reported "Doctor diagnosed" asthma and other lung disease, more exposed than control participants reported ever having experienced various respiratory symptoms, having experienced the various respiratory symptoms only after the accident, and having various respiratory symptoms worsen after the accident, if they had also been present before. A higher percentage of exposed than control subjects also answered "Yes" to each of 16 neurological screening questions, including one question which is supposed to assess response bias. Known difficulties in interpreting self-reported data, such as recall bias, are why the physician examination and pulmonary function tests were also conducted in this study.
The results of the physician examination suggest increased eye and skin problems among the exposed compared to the control subjects. More exposed than control participants had a current rash (adjusted OR = 2.4, 95% CI = (1.2 - 4.8)). This rash was most often described as dry skin, keratinization, flat macules with color changes and papules, and was located mostly on the face and arms. Most eye problems were noted more often among the exposed than the control participants; those eye problems associated with exposure were eyelid margin inflammation (adjusted OR = 5.1, 95% CI = (1.0, 26.9)) and meibomian gland inflammation (adjusted OR = 5.8, 95% CI = (0.7, 50.8)). Of the eye or skin abnormalities noted, the examining physician was more likely to think they were possibly due to chlorine exposure in the exposed than the control subjects.
This study found some evidence of poorer performance among certain subgroups of the exposed participants on the breathing tests, particularly among women, non-smoking participants, and those without preexisting respiratory symptoms. Among participants without preexisting respiratory symptoms, exposed subjects had lower mean FEV1 than control subjects (3.4 L versus 3.7 L), and a similar association was observed among those without preexisting asthma (3.4 L versus 3.6 L). When the pulmonary function test results were dichotomized into normal versus abnormal using Lower Limits of Normal, approximately 8% of exposed and 10% of control subjects were noted to have abnormal FEV1. Exposure was associated with abnormal FEV1 among non-smokers (adjusted OR = 6.3, 95% CI = (0.3, 148.0), and also somewhat among those without preexisting respiratory symptoms or asthma. Similar types of results were seen for FVC.
Approximately two-thirds of the exposed participants and three-fourths of the control participants completed the methacholine challenge testing; of these, about 36% of the exposed and 27% of the control subjects were defined as having abnormal methacholine challenge testing results. Exposure was associated with an abnormal test result among those without preexisting respiratory symptoms (unadjusted OR = 2.8, 95% CI = (0.8, 10.4)) and those with preexisting asthma (100% of the exposed and 36% of the control subjects had abnormal test results). The mean methacholine challenge dose-response slope was higher among exposed than control participants (7.1 versus 6.0), indicating a greater, although non-significant, response to a given dose of methacholine. Differences in mean slope between exposed and control subjects were greatest among non-smokers (6.1 versus 2.8), those without preexisting respiratory symptoms (5.0 versus 2.5) and those with preexisting asthma (55.8 versus 16.8).
The results of this study suggest that an acute exposure to chlorine following a train accident may have had chronic health effects on exposed persons. Various study limitations, including poor participation rates among eligible exposed and control subjects, issues of recall bias in the analysis and interpretation of self-reported symptoms, uncontrolled confounding, and difficulties in exposure assessment, make cautious interpretation of the study results necessary.
INTRODUCTION
On April 11, 1996, at about 4:30 am, 18 of the 71 cars in a freight train derailed about two miles west of Alberton, Montana. The crash punctured a tanker car, which released an estimated total of 64.8 tons of chlorine. Eighty-five pounds of dry sodium chlorate prills and 17,000 gallons of potassium cresylate solution also spilled from two other cars (1).
Prevailing winds carried the chlorine cloud 2 miles east along the river valley into Alberton, a small town with a population of about 350 (2). Local volunteer firemen and townspeople notified local residents about the accident and most residents evacuated first to the Frenchtown High School and then to Missoula, which is located approximately 30 miles east of Alberton. Residents were evacuated from Alberton and from an approximate four mile radius around the derailment (1). Most of the evacuees remained in motels or with friends or relatives for two weeks until the leakage of chlorine was brought under control. Initially, approximately 1000 people were evacuated from their homes. A relatively small number of people in the vicinity who were never evacuated may have also been exposed to chlorine.
The acute adverse health impact on people from the chlorine release was substantial. In addition to the death of one man from respiratory failure, approximately 300 people sought care from local emergency rooms. Approximately 40 people required treatment and 11 were hospitalized. Anecdotal reports of continuing health problems in the affected community have been received by ATSDR from a number of sources.
The chlorine that remained in the tanker car continued to leak despite efforts to control it. Chlorine concentrations were measured by the U.S. Environmental Protection Agency (EPA) and a Montana Rail Link (MRL) contractor, Marine Environmental, in approximately eighteen different locations at the derailment site and in the surrounding area. Monitoring in the area immediately adjacent to the wreckage yielded periodic chlorine readings in excess of the Occupational Safety and Health Administration (OSHA) ceiling exposure limit of 1 part per million (ppm), as well as the National Institute of Occupational Safety and Health (NIOSH) Immediately Dangerous to Life and Health (IDLH) level of 10 ppm. Readings at the derailment and along the interstate highway were recorded above the IDLH limits and occasionally reached as high as 1000 ppm between April 11 and April 27, 1996 (1).
Beginning on April 13, 1996, when wind conditions were favorable, residents were permitted to return to their homes for two hours in the morning and two hours in the evening to care for their animals and get personal belongings. Air monitoring was done under EPA supervision each time residents were allowed in the area. Results of the monitoring showed undetectable levels of chlorine except for one measurement of 0.1 ppm made on April 15, 1996 (1).
According to an ATSDR Health Consultation, which reviewed two reports with information on environmental contamination resulting from the derailment, based on the findings of clinical examinations of persons who had been examined due to symptoms experienced immediately after the derailment, the chlorine gas released during the derailment had migrated off site and affected human health (3). The consultation also concluded that available data did not indicate that other chemicals spilled in the accident or any of their reaction products (such as chlorinated phenol compounds produced from a chemical reaction between chlorine and potassium cresylate) migrated off-site (3).
The Phase I Study
Several days after the accident, the Montana Department of Public Health and Human Services (MDPHHS) contacted ATSDR and requested assistance in evaluating the impact of the chlorine release on the residents' health status. The purpose of this evaluation was to assist MDPHHS in describing the exposure and its acute health effects on the evacuated residents. A second objective was to identify a cohort of residents that could be followed and assessed for chronic health effects resulting from the spill.
All individuals who were at home at the time of the spill, and evacuated from their homes, were eligible to participate in the survey. Out of approximately 1000 people meeting these criteria, 682 were identified and interviewed (4). Despite having no complete list of evacuees, we were able to obtain a listing of guests in a number of motels in Missoula whose bills were the responsibility of Montana Rail Link (MRL) and contacted these residents. In addition, we made announcements about the survey at the daily community update meetings, and we announced a telephone number where interviewers could be contacted. Data collected included demographic information, smoking history, the time of evacuation, and information about health conditions. The questionnaire was directed toward the occurrence and duration of health symptoms experienced after the release, and about the residents' health status, including pre-existing diseases, which might increase their susceptibility to the exposure. The 682 participants included people of all ages. Interviews were conducted between April 23 and April 27, 1996.
The leading health complaints reported by respondents as attributable to the release were burning eyes (50%), throat irritation (50%), headache (52%), and coughing (56%). The proportions of those reporting difficulty breathing, burning chest pain, or wheezing were 33%, 24%, and 20%, respectively. More detailed information about the Phase I study is available in the Phase I study report (4).
Literature Review
Because chlorine is a gas, it most frequently comes in contact with, and affects, the upper and lower respiratory system, the eyes, and the skin. It is an extremely reactive chemical and readily combines with many elements. The mechanism of chlorine's irritant effect is postulated to be due to its reaction with water resulting in the liberation of hydrogen chloride and an oxygen free radical. Tissue damage is caused by either chlorine itself or one of its by-products, hydrochloric or hypochlorous acid (5).
The specific portion of the respiratory tract affected by an irritant gas depends on the dose, the duration of the exposure, and the solubility of the gas in water. Gases with a high water solubility react with the mucous membranes in the nose and throat while less soluble gases are more likely to affect lower portions of the respiratory tract. Chlorine has an intermediate solubility in water and, depending on the extent of the exposure, can produce effects throughout the respiratory tract (6). In a low dose exposure, chlorine tends to affect only the upper airways, while a more intense exposure will also damage the lower airways where air exchange occurs (the alveoli). In the upper portion of the respiratory tract, chlorine may cause injury to the mucosal cells and exaggerate physiologic responses, such as cough and mucus secretion. When chlorine reaches the lower airway, endothelial and type one epithelial cells are damaged.
Exposure to chlorine has both immediate and long-term health consequences (5-7). The severity of the adverse consequences on the respiratory system depends on the level of the chlorine dose. Doses as low as 1-2 ppm may cause the mucous membranes of the upper respiratory tract to become irritated and cause burning and tearing of the eyes, and burning of the nose and mouth (7). Cough, shortness of breath, headache, and substernal pain are frequent acute symptoms. Laryngospasm and bronchospasm may occur. If the exposure is severe, pulmonary function tests (PFT) will show airway obstruction and hypoxia due to cellular injury and obstruction of the small airways. Several hours after the exposure, a chemical pneumonitis may develop. With exposure to very high concentrations of chlorine, the inflammation and leakage of fluid in the lower respiratory tract may lead to pulmonary edema, adult respiratory distress syndrome (ARDS), and respiratory failure.
After an acute exposure, the symptoms and pulmonary abnormalities will improve and often resolve over time, usually within a month (5,7). However, chronic diseases of the airways, including chronic obstructive pulmonary disease and persistent asthma, have been reported to occur after an acute inhalation exposure to chlorine (5,6). A number of studies have shown that chronic obstructive changes (decreased FEV1 and FEV1/FVC) can occur in workers who had been exposed previously to high levels of chlorine (8-12), and one of these studies suggested an interaction with cigarette smoking (8). Changes in airway reactivity (measured by challenge testing, such as methacholine challenge testing) among workers exposed to chlorine have also been assessed (9,10). These studies suggest that bronchial hyperresponsiveness appears to be a chronic effect of chlorine among exposed workers.
In addition to studies of workers exposed to chlorine, where there may be more than one exposure over a period of time, there have been reports of individuals with one acute exposure to chlorine who appear to be suffering chronic respiratory effects from the exposure (13-15). Schonhofer et al followed-up three exposed individuals (13), and noted bronchial hyperresponsiveness and reactive airways dysfunction syndrome more than 2.5 years post exposure. Schwartz et al (14), after an average of 8.5 years follow-up, noted airflow obstruction in exposed individuals, but suggested it was probably not due to exposure. However, they also noted airway hyperreactivity in 5 of 13 individuals, 12 years after exposure, that appeared to be directly related to the degree of airflow obstruction and air trapping observed immediately after exposure, and hence may be related to exposure. Boulet (15) reported on two cases of exposed individuals, one exposed to hydrochloric acid and the other to a bleaching agent which contained chlorine, with bronchial hyperresponsiveness present one and six years after exposure, respectively. The presence of mild preexisting asthma in the first case may have exacerbated the effects of exposure.
Besides the research mentioned above, there has been some work which does not suggest chronic effects of chlorine exposure (16-19). The most recent report was by Leroyer et al in 1998 (16), and their 4-year follow-up of 13 workers with accidental chlorine exposure, showed complete recovery in three months for the individual who had decreased FEV1 and two individuals with decreased PC20. Two of the studies (17,19) involved persons exposed to chlorine due to a train derailment. Neither found any chronic spirometric abnormalities that couldn't be explained by smoking or preexisting respiratory health. The study by Hasan et al (18) found improvement in respiratory symptoms, FVC, and FEV1 within 5 months. The latter three studies did not assess bronchial hyperresponsiveness.
In addition to the pulmonary effects of chlorine, which have received the most attention in the literature, chlorine exposure is well known to be irritating or damaging to the eyes and skin (7). Courteau et al (20), in a study of construction workers repeatedly exposed to chlorine, noted throat and eye irritation as well as flu-like symptoms. In general, these non-respiratory symptoms are considered to be short-term only.
There is very limited evidence in the literature for chronic adverse neurological effects following acute chlorine exposure. Kilburn evaluated seven persons with acute chlorine exposure, referred to an environmental clinic specializing in neurotoxicology (21). Effects observed included memory loss, hearing loss, impaired balance, decreased vibration sensitivity, decreased color discrimination, and decreased verbal recall. Cerebral hemorrhages have been observed in one fatal (22) and one non-fatal (23) case of chlorine gassing; these lesions were likely secondary to hypoxia from widespread pulmonary damage. In general, it is not accepted that chlorine is directly neurotoxic.
Objectives
The specific objective of this study was to evaluate whether there is an increased prevalence of chronic respiratory, dermatological, ocular, or neurological health problems in the exposed community as compared to the non-exposed control participants.
METHODS
Study Design Rationale
This study, conducted in October of 1997, followed up Alberton area residents who were evacuated from their homes on the morning of the train derailment and interviewed as part of the Phase I Study . The purpose of this Phase II study was to assess the chronic health effects of an acute chlorine release. The primary health endpoint of interest was respiratory health (continued asthmatic symptoms, bronchial hyperresponsiveness, and spirometry decrements), other endpoints were chronic ocular, dermal, and neurological effects. Regarding the latter, although there is very limited evidence for adverse neurological effects as a result of chlorine exposure, concerns and anecdotal reports regarding neurological problems had been received from the community, and ATSDR felt it necessary to address these issues.
Exposure Assessment
The primary data source for exposure assessment was the chlorine air monitoring results, none of which had been from Alberton. Initially ATSDR attempted to use modeling to evaluate the chlorine emissions from the derailment (24,25) in order to identify the most highly exposed Phase I study participants, based on geography and temporality, and to allow for the creation of exposure categories in analysis. The numerous uncertainties and difficulties in this modeling, as well as uncertainties in individual risks for exposure (evacuation information), led to selecting Phase I participants for the Phase II study, based primarily on the location of their geocoded address with regard to the derailment. It was determined that those Alberton Phase I study participants who were likely to have had the greatest chlorine exposure were those residing within about 8 miles downwind (east of the spill, down the river valley) of the train derailment, hence exposed participants were selected from this group. Most of the selected participants lived closer than 8 miles downwind. There is a high likelihood that those residents receiving the highest exposure were included in the selection, based on both the ATSDR health consultation (3) and the conclusion in the ATSDR Record of Activity (AROA) (25). Despite the difficulties in modeling chlorine emissions, it was felt that the concentration of chlorine present in the ambient air of the community of Alberton could have reached levels of health concern and such exposures would have been more prevalent for persons living downwind of the derailment.
Study Population
Among the 682 persons interviewed as part of the Phase I Study, only those persons between the ages of 24 and 65, for whom there was substantial evidence of acute chlorine exposure effect were eligible for enrollment in Phase II. The purpose of the age limitation was to minimize the affect of age on pulmonary test outcome, and also to avoid performing these tests (see survey instruments section, below) on children. Available exposed persons included 104 men and 115 women.
The state and county health departments helped identify potential comparison populations. The comparison population used was selected from the residents of West Riverside and Piltsville, two small communities similar to Alberton located about 40 miles away. Residents were identified by a door-to-door survey of the two neighborhoods. An age and sex stratified sampling strategy was employed in order to enhance comparability between the exposed and unexposed study groups. As in the exposure group, only persons between the ages of 24 and 65 were eligible. The available control subject pool consisted of 112 males and 116 females.
Informed Consent
Informed consent was obtained from each study participant when they reported for their appointment, and prior to questionnaire administration, examination, pulmonary testing, and blood testing. The consent forms include descriptions of study procedures, risks, benefits, results notification, and time cost reimbursement. Any questions asked by prospective participants were answered prior to beginning an aspect of the study.
Survey Instruments
All study participants were administered a computer assisted questionnaire, in person (Appendix A). It included questions addressing demographics, including income and education, preexisting medical conditions, smoking status, occupational and domestic exposures, and health complaints (the latter reflecting those asked about in the Phase I questionnaire). Other questions concerned respiratory symptoms, adapted from a questionnaire validated in field trials by Abramson, et al. (26). A series of 16 questions developed by Hogstedt, et al. (27,28) was used to screen for possible neurological problems. The latter series of questions, validated by Hogstedt in field trials against the results of neurobehavioral test batteries, were initially used for those exposed to solvents, have also been used for those exposed to lead (29,30), and have been successfully employed by ATSDR in a similar community based study.
Spirometry was performed by most participants (1 exposed and 1 control individual did not perform the pulmonary function tests) using a dry rolling seal spirometer. Spirometry followed American Thoracic Society guidelines (31) and was administered by a certified spirometrist. Spirometrics included FEV1, FVC, and FEV1/FVC.
Factors that may alter an individual's response to pulmonary function testing include exposure to cigarette smoke and other irritants, upper respiratory tract infections, and asthma medications. When scheduling the testing, participants were asked to avoid these factors, if possible. As avoidance of these factors may not have always been possible, participants were questioned about these factors immediately before testing. The pre-spirometry questionnaire is included in Appendix B.
The methacholine challenge test (MCT) was completed by 81 exposed and 74 control participants. In order to minimize risks associated with the methacholine challenge, and based on the most recent guidelines of the European Respiratory Society (32), the test was not conducted on participants with any of the following contraindications: use of a beta-blocker medication; cardiovascular disease, including aneurysm, active angina, and recent myocardial infarction or stroke (within three months); pulse below 60, above 100, or irregular; history of problems taking inhaled medication; pregnancy; prostate illness resulting in symptoms of urinary slowing or change in frequency; inability to understand the procedures and the implications of a challenge test; or abnormal lung function at baseline testing (FEV1 more than 1.2 liters below predicted). The methacholine challenge test procedure was based on the abbreviated protocol described by Chatham, et al. (33), as modified by the experience of the NIOSH test personnel over a period of several years. The following is the order in which a number of breaths of a particular methacholine concentration was administered: 1 breath at 0.1 mg/ml; 4 breaths at 0.1 mg/ml; 1 breath at 1.0 mg/ml; 4 breaths at 1.0 mg/ml; 1 breath at 5.0 mg/ml; 4 breaths at 5.0 mg/ml; 1 breath at 25.0 mg/ml; and 4 breaths at 25.0 mg/ml. Each subject completing the methacholine challenge testing was administered an increasing methacholine dose until their FEV1 decreased to 15% or more below their initial FEV1, at which point a bronchodilator was administered if necessary, or until they had completed the above protocol.
Each participant was examined by a physician for potentially chlorine-related ocular irritation or damage and dermal effects. The ocular evaluation was visual only. The form used by the physician for recording his/her observations is included in Appendix C.
Lastly, each participant was evaluated for atopy on the basis of a blood analysis for IgE. Atopy, a type 1 allergic reaction associated with the IgE antibody, is found in a majority of persons with allergic asthma and was to be assessed as a confounder in this study.
Data Management/Quality Control
Interviewers were trained to use the questionnaire through role playing and observation of interviews. Each questionnaire was checked for completeness and accuracy and questions were referred back to the interviewer for resolution. Internal range and consistency checks were used to identify and correct coding and data entry errors. Spirometry followed American Thoracic Society (31) guidelines and was performed by a certified spirometrist. Study participants were counseled to avoid revealing their exposure status to their examining spirometrist or physician.
Participant Notification
The individual results (breathing tests, physician examination and IgE level), interpretations, and recommended actions were sent by letter to each participant. A sample letter outline is included as Appendix D. The name of a contact person at ATSDR was included in the letter should a study participant or their health care provider have further questions regarding the study results or other issues. The test results may have been sent to the subject's physician or health care provider, but only with the written permission of the participant.
Data Analysis
The various data (questionnaire, skin and eye exam, IgE, and pulmonary function and methacholine tests) were cleaned and merged together for the analysis.
Creation of Variables
Various variables were created to use in multivariate analyses. Age was categorized into less than or equal to 35, 36 to 45, 46 to 55 and 56 to 65 years. Although the allowable age range for participation was supposed to be 24-65, one 23 year old exposed individual participated and was included in all analyses for several reasons: this individual was exposed, this individual completed all of the parts of the study, and this individual was only a few months from being 24 years old. Education was grouped into high school graduate or less, and more than a high school education (which included junior college or technical training as well as four-year college and graduate school). Family income was divided into less than $20,000, $20,000 or more, and unknown.
Cigarette smoking categories were never smoked, current smoker, and former smoker. Total pack years smoked was created by multiplying years smoked by number of cigarettes smoked per day, and the mean number of cigarettes smoked was assigned for those missing number of cigarettes per day. Exposure to occupational dust was defined for those who answered yes to any of the following occupational exposures: cotton dust; grain dust or other farming or other agricultural dusts; wood or saw dust; sand or silica dust; metal dust; and insulating materials such as asbestos or fiberglass. Occupational volatile organic compound (VOC) exposure was defined for those who answered yes to any of the following occupational exposures: solvents, cleaning agents or degreasers; fumes from plastics, adhesives or glues; paints, paint removers or paint thinners; and gasoline, kerosene or lubrication oils.
Preexisting asthma was defined based on the three questions asked about asthma: have you ever had an attack of asthma, and if so did this happen before and/or after the train derailment; have you ever taken any medicines, pills, or inhalers for asthma, and if so did you take them before and/or after the train accident; and has a doctor or other health care provider ever told you that you had asthma, and if so, at what age were you first told that. Using these questions, a variable was created indicating whether the participant ever had asthma -- by a positive answer to asthma attack, asthma medications or asthma diagnosis. If there was a positive answer to any of these questions, the asthma was further designated as to whether it had existed before the train accident.
A participant was considered to have preexisting respiratory symptoms if they answered yes to the presence of any of the 12 respiratory symptoms before the accident. These 12 symptoms include chest wheezing/whistling, chest tightness (in the morning; with dust exposure), shortness of breath (with or without strenuous activity; woken at night by; with dust exposure), coughing (woken at night by; first thing in the morning), bringing up phlegm, having an asthma attack, and taking medications for asthma.
Statistical Analyses, Demographics and Symptoms
All statistical analyses were completed using the software of the Statistical Analysis System (SAS). The exposed and control groups were compared with regard to demographics, lifestyle variables such as smoking, and occupational exposures. Similarly, exposed and control groups were compared on self-reported health problems (including gastrointestinal symptoms, ocular symptoms, respiratory symptoms and dermatological symptoms) which were asked about as had been done in the Phase I study. The phrasing of these questions was "Since the train accident, are you having this health problem more than usual?", and "If yes, are you having this problem now?" Exposed and control groups were also compared with regard to self-reported respiratory symptoms such as chest wheezing/whistling. The phrasing of these questions was "Have you ever had ...?", "Did this happen before the train accident in April 1996?", "Did this happen after the train accident in April 1996?", and "If both before and after, was it worse or did it happen more often after the train accident?" Also available for comparing exposed and control subjects were reports of doctor or other health care provider diagnosed medical conditions: asthma; chronic bronchitis, emphysema, or chronic obstructive pulmonary disease (COPD); other lung or pulmonary disease; and hayfever or other allergic diseases.
Statistical Analyses, Neurological Symptoms
To analyze the neurological symptoms, the reporting of each of 16 neurological symptoms was compared between exposed and control persons and crude and adjusted odds ratios were computed using logistic regression. Multivariate logistic regression models adjusted for age, sex, education, income, cigarette smoking, cotton dust exposure, grain and other agricultural dust exposure, metal dust exposure and occupational exposure to various VOCs: solvents, cleaning agents, degreasers; plastic, adhesives or glue fumes; and paint, paint remover or paint thinner. The total number of neurological screening questions answered positively out of 16 was then summed. A dichotomous variable was created, indicating less than 6, or greater than or equal to, 6 positive answers, and exposed and control groups were again compared using logistic regression with regard to this variable. A cut point of 6 positive answers was chosen based on Hogstedt (27).
Statistical Analyses, Skin and Eye Examination
Analyses of the physician examination was conducted similarly to the analyses described above, comparing exposed and control subjects with regard to the physician's observations. Crude and adjusted odds ratios were computed using logistic regression where enough positive and negative responses allowed.
Statistical Analyses, Spirometrics
The first part of the analysis of the pulmonary function test results was computation of means and standard deviations of various test results: adjusted FEV1, adjusted FVC, FEV1/FVC, FEV1 percent of predicted, FVC percent of predicted, and FEV1/FVC percent of predicted. The adjusted values were standardized for height by dividing an individual's value for a particular pulmonary function test by height2 and multiplying by mean height2, which was calculated separately for males and females (34). The predicted values were supplied by NIOSH, and the formula from which they were created came from Knudson (35). Means and standard deviations were also computed by sex, cigarette smoking, preexisting respiratory symptoms, and preexisting asthma. The mean PFT values among exposed and control participants were compared using a T test, and resulting p-values are two-tailed.
To further analyze possible differences in pulmonary function test results between exposed and control subjects, and between exposed and control subjects by various strata (sex, cigarette smoking, preexisting respiratory symptoms, and preexisting asthma), general linear modeling was used. This type of modeling allows determination of whether the mean value of the pulmonary function test varies significantly by exposure while simultaneously controlling for numerous confounding factors. The covariates included in the adjusted general linear models were sex, age, age2, height (not included in models for adjusted FEV1 or adjusted FVC, since the adjustment is a height standardization), pack years smoked for current and former smokers, packs per day for current smokers (all of these chosen based on Dockery, 1988 (34)), preexisting asthma, education, income, and occupational dust exposure. Examination of these variables showed them to be important risk factors for the outcomes in the data. For the general linear modeling an F-test and p-value are presented for the exposure variable using Type III sums of squares. This test compares the difference in mean pulmonary function test results between exposed and control subjects while adjusting for the other possible confounding variables included in the model.
The pulmonary function test results were dichotomized by creating Lower Limits of Normal (LLN) (34,36). A subject's LLN was created by taking their predicted value for a particular PFT minus 1.645 times the standard deviation around the regression line for that PFT; the predicted values and standard deviation came from Knudson (35). A subject was considered to have an abnormal result for a particular pulmonary function test if their value for that test was below their LLN for that test. Logistic regression was used to compare exposed and control groups with regard to abnormal versus normal results for FEV1, FVC, and FEV1/FVC. These comparisons were also done stratifying on the variables sex, cigarette smoking, preexisting respiratory symptoms, and preexisting asthma.
Multivariate logistic regression was used to determine whether the risk of having an abnormal PFT outcome was related to exposure while adjusting for various variables. Models adjusted for age, sex, education, income, preexisting asthma, smoking and occupational dust exposure. These confounders were chosen because they were individually strong risk factors for the outcomes and/or they changed the OR (Odds Ratio) for exposure by 10% or more, when included in a model with exposure (for at least one of the outcomes).
Statistical Analyses, Methacholine Challenge Test
The methacholine challenge test results were analyzed in two different ways. The first method involved creation of a dichotomous variable, which indicated a normal or abnormal methacholine challenge test result for each participant who completed the test. An abnormal result was indicated if an individual's FEV1 dropped 15% or more below their baseline FEV1 in response to a dose of methacholine. Determination of an abnormal result was based on an examination of each individual's graphed results, which plotted the percentage decrease in FEV1 for each dose of methacholine administered. Logistic regression was used to determine whether the risk of having an abnormal test result varied by exposure status in an unadjusted model, as well as in a model controlling for the following confounders: best FEV1, passive smoke exposure, preexisting asthma, and cigarette smoking (these covariates were chosen because they were associated with the outcome and changed the odds ratio for exposure by 10% or more when included in a model individually with exposure), as well as sex, age, education and income. One participant who completed the methacholine challenge testing had taken albuterol before testing; the analyses were repeated excluding this individual. Unadjusted and adjusted logistic models were also run, stratifying on the variables sex, cigarette smoking, preexisting respiratory symptoms, and preexisting asthma.
The second method for analyzing the methacholine challenge test results involved creation of a methacholine dose-response slope for each individual who completed the test, based on instructions from NIOSH, in order to assess differences in dose-response between exposed and control subjects (37). NIOSH engineers involved in the nebulizer testing determined a rough average mass concentration of 164mg/M3 over a flow range of 12.5 to 35 liters per minute, which should cover most of the range of flows that the subjects would use, inhaling a vital capacity of 3 to 6 liters in 10 seconds. An approximation of the dose administered was calculated as follows: dose administered = 164 mg/M3 * (FVC in liters/breath/1000liters/M3) * (methacholine concentration in mg/ml solution/1000 mg/ml solution) * (number of breaths at that concentration). The final dose was the sum of all cumulative doses for an individual. Dose response slope was the initial FEV1 minus final FEV1, divided by initial FEV1, then divided by the final dose. A larger slope would indicate a greater response to methacholine.
Means and standard deviations of methacholine dose response slope were computed by sex, cigarette smoking, preexisting respiratory symptoms, and preexisting asthma. The mean PFT values were compared between exposed and control participants using a T test, and resulting p-values are two-tailed.
To further analyze the differences in methacholine challenge test results between exposed and control subjects, general linear modeling was used. The covariates included in the adjusted general linear models were sex, age, preexisting asthma, education, had a recent cold, recent vaccination, passive smoke exposure, cigarette smoking (these confounders were chosen because they changed the parameter estimate for exposure by 10% or more when included in a model individually with exposure) and income. Relevant results from the general linear models procedure, as noted above, are the probability of the F-value for the Type III sums of squares.
It was noted that approximately one-third of the exposed and one-fourth of the control subjects did not complete the methacholine challenge testing (with reasons ranging from refusal to pregnancy to physician declined), and there were concerns that this may affect the interpretation of the methacholine challenge test results. Those who did and did not complete the methacholine challenge testing were compared on various characteristics, including demographics, preexisting asthma, other exposures, pulmonary function test results, and possible reasons why the methacholine challenge testing was not done.
RESULTS
Sample Selection and Participation
Among the 682 persons interviewed as part of the Phase I Study, only those persons between the ages of 24 and 65, for whom there was substantial evidence of acute chlorine exposure, were eligible for enrollment in Phase II. Available exposed persons included 104 men and 115 women, of these, 60 men and 72 women made appointments, and 55 men and 66 women were actually seen, for a total of 121 exposed participants available for analysis. Of the eligible exposed participants, 55% participated. Although the age range of participants for this study was supposed to be 24-65, one exposed participant, aged 23, participated, and this subject has been included in the analyses for reasons mentioned previously.
A comparison population was selected from the residents of West Riverside and Piltsville, two small communities similar to Alberton, located about 40 miles away. An age and sex stratified sampling strategy was employed in order to enhance comparability between the exposed and unexposed study groups. As in the exposure group, only persons between the ages of 24 and 65 were eligible. The available control subject pool consisted of 112 males and 116 females. Of these persons, 48 men and 67 women made appointments, and 44 men and 57 women were actually seen, for a total of 101 control participants. Of the eligible control participants, 44% participated. Two of these control participants were excluded from all analyses, one because the subject reported being involved in clean-up at the site of the train accident and the second because the participant's questionnaire data was completely overwritten on the computer by another participant's data. Hence, a total of 99 control participants were available for analysis.
All subjects completed the questionnaire and were examined by a physician. One exposed and one control participant did not complete the pulmonary function tests, and approximately one-third of the exposed subjects and one-quarter of the control subjects did not complete the methacholine challenge testing.
Description of the Study Population
A description of the study participants is presented in Table 1. Exposed subjects tended to be older and of lower socioeconomic status (less education and lower income) than control participants. More exposed than control subjects were current smokers, were exposed to passive smoke, and had smoked more pack-years. Exposed participants appeared to have had more occupational dust exposure overall, particularly exposure to sand/silica dust and insulating materials. Occupational VOC exposure overall was similar among exposed and control subjects, although exposed participants had somewhat more exposure to plastic/adhesive/glue fumes and gasoline/kerosene/lubrication oils, while control participants were more likely to have had exposure to paints/paint remover/paint thinner. Exposed participants were more likely than control participants to have used a wood stove or fireplace at least once a week in the last cold season, less likely to have a humidifier built into their heating system, and more likely to have reported a mold or mildew problem inside the house in the past year. A slightly higher percentage of control than exposed participants, who had blood drawn, had high IgE levels, which were defined as being over 400 IU/ml (7% of control subjects versus 4% of exposed subjects). The median IgE level was 35 IU/ml among exposed subjects and 27 IU/ml among control subjects, and the mean IgE level was 100.2 IU/ml (SD=214.5) for exposed subjects and 121.2 IU/ml (SD=305.8) for control subjects. The adult normal range is 3 - 423 IU/ml.
More exposed than control subjects reported feeling that they are more sensitive to chemicals than most people, especially after the train accident. Exposed participants also reported more concern that the chemicals spilled in the train wreck have harmed their health or may do so in the future, but the level of concern about something in the neighborhood environment that may be harming health was similar among exposed and control participants.
Self-reported Health Problems
Participants were asked about various health problems, similar to those which had been asked about in the first study in Alberton. They were asked if they were having a particular health problem, more than usual, since the train accident, and if yes, whether they were having the particular problem 'now'. Health problems addressed this way included gastrointestinal symptoms, fatigue and sleepiness, aching muscles, dermatological and ocular problems, and various respiratory symptoms. The prevalence of each self-reported health problem since the train accident was higher among exposed than control subjects (Table 2).
The results of the second part of this question, 'are you having the problem now', are also presented in Table 2.
Respiratory Symptoms
A higher percentage of exposed than control subjects reported ever having experienced most of the respiratory symptoms (Table 3). For all of the respiratory symptoms, a larger percentage of exposed than control subjects reported presence of the symptom only after the accident. Among those reporting the symptoms, both before and after the accident, more exposed than control subjects reported that the symptoms were worse after, rather than before, the train accident.
Of participants who completed the pulmonary function testing, 75 (62.5%) of 120 exposed subjects and 63 (64.3%) of 98 control subjects had preexisting respiratory symptoms. The presence or absence of preexisting respiratory symptoms is used to stratify in later analyses. Two of the 12 questions are considered to refer more to chronic bronchitis (morning cough and morning phlegm), while the others are considered to represent asthma (26). Removing these 2 bronchitis questions and rerunning the stratified analyses did not change the results. Because a better variable to define preexisting asthma is available (described below), the sum of 10 asthma symptoms was not used.
Based on three questions addressing asthma - use of asthma medications, experiencing an asthma attack, and age at which asthma was diagnosed, a variable which addressed timing of asthma in relation to the accident, was created. Approximately equal numbers of exposed and control subjects ever had asthma based on this variable, more control participants had asthma only before the accident or both before and after the accident, and more exposed subjects had asthma only after the accident. Those subjects who were defined as having had asthma before the accident, based on these questions, were considered to have preexisting asthma. Thirteen exposed participants (10.7%) and 21 control participants (21.2%) were considered to have preexisting asthma.
Regarding doctor diagnosed respiratory conditions, more control than exposed subjects reported ever having been diagnosed with a lung disease, other than chronic bronchitis/emphysema/COPD, or ever having been diagnosed with asthma.
Neurological Symptoms
A higher percentage of exposed than control subjects answered yes to each of 16 neurological screening questions (Table 4), including the question about problems with buttoning, which is supposed to assess response bias (Hogstedt 1980). Odds ratios, both unadjusted and adjusted, were elevated for all of the screening questions. The confidence intervals for these odds ratios included 1.0, and hence did not achieve statistical significance, only for difficulty of falling asleep (both unadjusted and adjusted models), and feeling lightheaded or dizzy (the adjusted model only). When the total number of neurological screening questions answered positively out of the 16 was summed, 51.2% of the exposed and 13.1% of the control subjects had 6 or more positive answers. The adjusted odds ratio for the association between exposure and 6 or more positive answers to the neurological screening questions was 7.3, with a 95% confidence interval (CI) of 3.3-16.1.
Physician Examination of Eyes and Skin
A greater percentage of exposed than control subjects reported various skin troubles to the examining physician: persistent skin rash since April, 1996, rash duration of two or more months, and an intermittent rash (Table 5). The unadjusted and adjusted odds ratios for these skin effects were elevated and confidence intervals did not include 1.0, hence the results may be considered to be statistically significant. The examining physician noted the presence of a rash at the time of the examination (current rash) in 33.9% of the exposed group and 19.2% of the control group. The exposed subjects were more than twice as likely as the control subjects to have a current rash even after adjusting for age, sex, SES, smoking, occupational dust exposure and occupational VOC exposure (adjusted odds ratio (OR) = 2.4, 95% confidence interval (CI) = (1.2-4.8)). The current rashes were described mostly as dry skin, keratinization, flat macules with color changes and papules, and were located mostly on the face and arms.
The occurrence of most of the various eye problems was relatively rare (Table 6). For some of the problems, there were either zero exposed subjects or zero control subjects with the problem, so adjusted logistic analyses could not be done. For example, zero controls were noted to have eyelid edema and eyelid margin infection, and zero exposed individuals were noted to have cloudy cornea or corneal scarring. A higher percentage of exposed than control subjects were noted by the physician to have the following problems: eyelid margin inflammation, eyelid margin infection, meibomian gland discharge, meibomian gland inflammation, conjunctival inflammation, conjunctival discharge, dry eyes, and pterygia pinquecula. Adjusted odds ratios were elevated for all of these problems except for the latter one, but only for eyelid margin inflammation did the confidence interval exclude 1.0, and hence achieve statistical significance (adjusted OR = 5.1, 95% CI = (1.0-26.9)). The results which most suggest differences among exposed and control subjects, are eyelid margin inflammation and meibomian gland inflammation (adjusted OR = 5.8, 95% CI = (0.7-50.8)).
The physicians were asked to answer the following questions after examining each individual: "Should this person be referred for follow-up care?"; "Is this for follow-up of an eye/skin abnormality?"; and "May the abnormality be due to chlorine exposure?" Overall, the physicians recommended that 26.4% of the exposed and 11.1% of the control subjects be referred for follow-up care. This question was missing for 6.6% of the exposed and 6.1% of the control subjects. For 18 of the exposed and 6 of the control individuals, the physician indicated the follow-up was for an eye abnormality, and for 24 of the exposed and 7 of the control individuals, the physician indicated the follow-up was for a skin abnormality. The physician's answer to whether the eye abnormality may be due to chlorine exposure was yes/maybe for 15 exposed and 2 control subjects, and the answer to whether the skin abnormality may be due to chlorine exposure was yes/maybe for 11 exposed and 1 control subject.
Pulmonary Function Tests
Mean pulmonary function test results among exposed and control subjects were compared. The hypotheses were that the exposed group would have lower FEV1 and FEV1/FVC results than the control group (2-tailed p-values are presented, however), and although it was not hypothesized that the FVC would be affected by exposure, exposed and control subjects were also compared on this value.
Mean adjusted FEV1 was similar among the exposed and the control groups (3.3 Liters (L) and 3.5 L) (Table 7). When exposed and control groups were compared by various strata such as sex (male, female) and preexisting asthma (yes, no), statistically significant or nearly significant differences between exposed and control participants were noted for females (2.9 L in the exposed versus 3.1 L in the controls), subjects without preexisting respiratory symptoms (3.4 L versus 3.7 L), and those without preexisting asthma (3.4 L versus 3.6 L).
Mean adjusted FVC among the exposed group was not very different from that among the control group (4.3 L versus 4.4 L) (Table 8). As with FEV1, there was a difference in mean FVC between exposed and control females (3.6 L versus 3.8 L). There appeared to be other, smaller, differences between exposed and control subjects, such as among those without preexisting asthma (4.3 L versus 4.4 L, p=0.04 after multivariate adjustment).
The mean FEV1/FVC % among exposed and control subjects were similar (78.8 and 78.4) (Table 9). There was not much difference between exposed and control mean FEV1/FVC % by various categories. The only statistically significant difference was among those with preexisting respiratory symptoms, in the opposite direction from that expected.
The results for mean percent of predicted FEV1 were similar to those for adjusted FEV1, in that the mean values were similar among the exposed and control groups (98.9 and 101.9) and the differences in the means among exposed and control participants appeared greatest for females, those without preexisting respiratory symptoms and those without preexisting asthma (Table 10).
The mean values of percent of predicted FVC for exposed and control groups (103.9 and 107.2) were not statistically different (Table 11). The only statistically significant differences were among females (106.5 in the exposed versus 113.2 in the controls) and subjects without preexisting asthma (103.8 versus 108.9).
The mean percent of predicted FEV1/FVC % did not differ between exposed and control groups (about 95% for both groups) (Table 12). Exposed and control subjects were similar on the various categories, except possibly for the difference between exposed and control subjects with preexisting respiratory symptoms (94.6 versus 92.2), again in the opposite direction from that expected.
When the pulmonary function test results were dichotomized into normal/abnormal, the following was observed (Table 13): abnormal FEV1 was noted in approximately 8% of exposed and 10% of control subjects; abnormal FVC was seen in nearly 6% of exposed and about 5% of control participants; and approximately 13% of exposed and 16% of control subjects were defined as having abnormal FEV1/FVC. Unadjusted and adjusted odds ratios and 95% confidence intervals showed essentially no differences in risk of an abnormal PFT by exposure status.
Exposure was also considered as a risk factor for abnormal PFTs by strata of the variables sex (male, female), smoking (never, current, former), preexisting respiratory symptoms (yes, no), and preexisting asthma (yes, no). Due to small numbers, logistic regression could not be used for some of the subgroup analyses, and most confidence intervals were very wide.
For abnormal FEV1 (Table 14), the only confidence interval not including 1.0, and hence achieving statistical significance, was among those subjects with preexisting respiratory symptoms, of whom about 9% of exposed and 16% of control participants had abnormal FEV1 (adjusted OR = 0.2, 95% CI = (0.1-0.8)). The risk of abnormal FEV1 also varied somewhat in the smoking strata (among never smokers, more exposed than control participants were defined as having abnormal FEV1) and in the respiratory symptoms and asthma strata (among those without preexisting respiratory symptoms or asthma, more exposed than control participants were defined as having abnormal FEV1).
Considering exposure as a risk for abnormal FVC (Table 15), all of the strata for which odds ratios could be calculated had wide 95% confidence intervals which included 1.0, and hence did not achieve statistical significance. The risk of abnormal FVC appeared greatest for exposed individuals who had no preexisting respiratory symptoms or were never smokers.
The results for normal/abnormal FEV1/FVC (Table 16) tend to echo what was observed for abnormal FEV1 and abnormal FVC, and for those associations for which odds ratios were calculated, all of the 95% confidence intervals included one.
Methacholine Challenge Test
Of the 120 exposed participants who completed the pulmonary function testing, 81 (67%) also completed the methacholine challenge testing. Of the 98 control participants who completed the pulmonary function testing, 74 (76%) completed the methacholine challenge testing. The reasons noted on site for not completing the methacholine challenge testing were physician declined (less than 10%), possibly pregnant (less than 5%) and refusal (about 90%), with a specific reason for the refusal generally not noted.
Approximately 36% of the exposed participants and 27% of the control participants were defined as having an abnormal methacholine challenge testing result because their FEV1 dropped 15% or more below their baseline FEV1 in response to a dose of methacholine (Table 17). Exposure status was not, however, significantly associated with an abnormal test result in an unadjusted model (OR = 1.5, 95% CI = (0.8-3.0)). In a model adjusted for best FEV1, passive smoke exposure, preexisting asthma, smoking, sex, age, education, and income (IgE was not shown to be a confounder in our data), the odds ratio decreased (OR = 1.3, 95% CI = (0.6-3.2)). If the one individual (exposed, with an abnormal methacholine challenge test result) who took albuterol before testing was excluded, the results essentially did not change, resulting in an adjusted odds ratio of 1.3 (95% CI = (0.6-3.1)).
The stratified analyses showed some differences in the risk of abnormal methacholine challenge test result among various strata. The most remarkable differences were among males, those without respiratory symptoms before the accident, and those with preexisting asthma. About 27% of the exposed males and 13% of the control males had an abnormal test result. The adjusted odds ratio for this comparison was 3.1 (95% CI = (0.4-21.7)). About 31% of exposed subjects and 14% of control subjects without respiratory symptoms before the accident had an abnormal methacholine challenge test result (unadjusted odds ratio 2.8, 95% CI = (0.8-10.4)). Of those with preexisting asthma, 100% of the exposed and about 36% of the control subjects had abnormal methacholine challenge test results.
A methacholine dose-response slope was calculated for all who completed the methacholine challenge test, based on information and a formula supplied by NIOSH. The mean slope was higher among exposed participants than among control participants (7.1 versus 6.0), indicating a greater response to a given dose of methacholine, but the differences between these two means was not statistically significant (all statistical testing was done using log-transformed slope), with an unadjusted p-value of 0.13. (Table 18). Mean slope values were also computed by various strata such as sex (male, female) and smoking (never, current, former).
Generalized linear modeling was used to evaluate the differences in mean methacholine dose-response slope between exposed and control participants while adjusting for relevant confounders. Considering the p-values for the T test to be unadjusted values, the p-values for the differences in means for most of the comparisons increased after adjustment (Table 18). Differences in mean slope between exposed and control subjects were greatest among males (4.6 versus 5.0), non-smokers (6.1 versus 2.8), those without preexisting respiratory symptoms (5.0 versus 2.5), and those with preexisting asthma (55.8 versus 16.8). The comparison of means for relevant categories was repeated after removing the participant who took albuterol before testing, and the p-values changed very little.
The last analysis was a comparison of exposed and control subjects who completed and did not complete the methacholine challenge testing (Table 19). These comparisons were done to evaluate whether there were any differences between those who completed and those who did not complete the methacholine challenge testing. Overall, it appears that those who did not complete the testing tended to have worse pulmonary function test results, were older, had lower SES, were more likely to be current smokers and to have cardiovascular problems or breathing problems. Some of these differences were not of equal impact among exposed and control subjects; for example, cardiovascular problems were reported by 25.6% of the exposed who did not complete the testing and 8.3% of the controls who did not complete the testing.
DISCUSSION
This investigation was designed to determine whether persons suffered chronic effects after an acute exposure to chlorine. Chlorine is most noted for its effects on the respiratory system, which is what this study emphasizes, as well as its effects on the skin and eyes. A neurological screening questionnaire was also administered to participants, due to concerns about reported neurological symptoms in the exposed community.
There were some basic notable differences between the exposed and control participants. The exposed group tended to be less well-educated, have lower income, and smoke more than the control group.
The prevalence of each self-reported health problem since the train accident was greater among exposed participants than the control participants (Table 2). Making any conclusions about this is difficult, however, because the way in which the questions were asked - 'Since the train accident, are you having ___ more than usual?' - provides a very strong point of reference for exposed participants but not for control participants. The train accident, based on all accounts, was a very memorable event to the exposed participants. Although the control participants likely knew of the accident, it is unlikely that it affected their lives as much. Questions framed this way may bias the exposed group toward over-reporting or better-remembering of their health problems compared to the control group. This is suggested by a higher prevalence of every single self-reported health problem among the exposed compared to the control participants. For nearly all respiratory symptoms, also, more exposed than control participants reported having the symptoms only after the accident, or, if they had it both before and after the accident, reported that it was worse after (Table 3). It is, perhaps, notable that this pattern did not hold true for the question about swollen feet, which was located in the midst of the questions about respiratory symptoms. This may argue against over-reporting. However, the placement of this question makes its utility somewhat uncertain.
The questions about respiratory symptoms and conditions do show that control participants appear to have more underlying respiratory disease than the exposed participants: the control group reported more doctor diagnosed lung disease other than chronic bronchitis / emphysema / COPD and more doctor diagnosed asthma. Based on the questions about asthma attacks, use of asthma medications, doctor diagnosed asthma and age at diagnosis, more control than exposed participants had asthma both before and after the accident. Inadvertent recruitment of a control group with poorer respiratory health than expected may have occurred. It is also possible that control recruitment may have attracted persons with preexisting respiratory illness, particularly asthma, maybe because of the tests being conducted. At the same time, as stated earlier in the Report, the control group exhibited a higher socioeconomic status than the exposed group. Therefore, the control participants might have a more comprehensive medical insurance which might result in a more realistic knowledge of their underlying respiratory disease.
Reporting of neurological symptoms was greater among exposed than control subjects for each symptom, including difficulty buttoning buttons, which is supposed to assess response bias. The questionnaire which was used to screen for neurological symptoms has been used and validated on workers exposed to solvents (27,28) and has also been used on workers with lead exposure (29,30). The questionnaire's use in this study was primarily to addresses residents' concerns about neurological symptoms, as it has not been shown that chlorine is neurotoxic, nor has the presence of any other chemicals besides chlorine from the derailment been documented off-site. It is unclear how the results of this screening questionnaire should be interpreted for this study. There may be over-reporting going on, as indicated by an increased prevalence in difficulty buttoning buttons among the exposed group, or these results may be an actual indication of neuropsychiatric problems in the exposed participants. Given the lower socioeconomic status of exposed participants compared to control participants and the stress on exposed participants due to the derailment and its after-effects, an increased prevalence of neuropsychiatric illness in the exposed subjects before and/or after the derailment would not be surprising. Other possible and unmeasured causes of neuropsychiatric illness, which might explain some of the findings, include other exposures (adjustment for some occupational exposures, particularly VOC exposure, was possible generally, but adjustment for specific solvents, lead, or mercury, for example, was not possible), alcoholism and arteriosclerosis.
The difficulty in interpreting all the measures described above is that they are self-reported symptoms referring to a specific period of time, which only really happened to, and hence is likely significant, only for the exposed participants. Because of this, it is not surprising that the prevalence of nearly every symptom in our questionnaire was higher among exposed than control subjects after the derailment, and nearly every symptom already present worsened after the derailment in the exposed participants. This may be due to actual health differences among exposed and control groups, or it may be due to recall bias, as has been demonstrated previously in this sort of situation (38), resulting either in over-reporting or better-remembering of symptoms by the exposed group. Considering the known difficulties in the use of self-reported symptoms to compare exposed and control groups, more objective measures of the long term effects of an acute exposure to chlorine were collected: the physician exam, pulmonary function tests, and the methacholine challenge test.
The physician examined the skin and eyes of participants. The first four questions of the skin effects exam (Any skin rash since April 1996? Rash duration at least 2 months? Intermittent rash? and Date of onset of rash?) appear to be self reports of skin rashes since the derailment. The question on the existence of a current rash and the questions thereafter appear to be based on actual physician examination. The exposed subjects were more than twice as likely as the control subjects to have a current rash, even after adjustment for variables such as age, sex, smoking and occupation (Table 5). These rashes were described mostly as dry skin, papules, flat macules with color changes, and keratinization and were located primarily on the face and arms. These descriptions do not seem to be what would be expected from chlorine exposure (chloracne), and it is possible there is some uncontrolled confounding, such as time spent outdoors and other environmental exposures. However, the examining physicians did indicate that the skin abnormality may be due to chlorine exposure for 11 exposed and only 1 control participant, so possibly the rashes seen in the exposed participants are the long term sequelae of an acute exposure to chlorine.
Nearly all of the eye abnormalities for which the physicians examined were found more frequently in exposed participants (Table 6). The two abnormalities most strongly associated with exposure were eyelid margin and meibomian gland inflammation, although adjustment for other variables did lessen the associations. Unadjusted confounding may account for some of the differences, although again, the examining physicians seemed to feel that the eye abnormalities may be associated more with chlorine exposed than control participants (15 versus 2).
Based on the physician examination of participants, it is possible that chlorine may have caused damage to the skin and eyes in some of our exposed subjects which, even one and one half years after exposure, is evidenced in increased skin problems (rashes, particularly dry skin, papules, flat macules with color changes, and keratinization) and eye problems (eyelid margin and meibomian gland inflammation). It should be noted that the study would have benefit if the examining physician had known status of the skin and eyes of exposed and control participants before the accident. Also, there is not much previous research on the long term effects of an acute chlorine exposure on the skin and eyes. Based on these findings, more study may be warranted.
Pulmonary function tests were administered to assess physiologic differences between exposed and control subjects which may be due to chlorine exposure. It was expected that FEV1 would be the most likely to show an effect of exposure, as well as FEV1/FVC. Some minor differences were noted, such as with FVC (4.3L in the exposed versus 4.4L in the controls, Table 8), which may approach statistical significance, but given an estimated technical between subject variation of 3% (36), these differences are unlikely to be meaningful. The largest differences in FEV1 between exposed and control participants in the expected direction were found among females, subjects without preexisting respiratory symptoms, and subjects without preexisting asthma (Table 7, Table 10). This suggests possible persistent obstructive lung changes due chlorine exposure in these groups of individuals. The largest differences in FVC between exposed and control participants in the expected direction were found among females and participants without preexisting asthma (Table 8, Table 11). This suggests possible persistent restrictive lung changes due to chlorine exposure in these groups of individuals. There was not much evidence of differences in mean FEV1/FVC between exposed and control subjects, except for those subjects with preexisting respiratory symptoms and asthma, which were in the opposite direction of what was expected. (Table 9, Table 12).
The analysis of PFT results as dichotomous outcomes, abnormal or normal FEV1, FVC, and FEV1/FVC, essentially agreed with the analyses of linear PFT results. Chlorine exposure appeared to be more of a risk for abnormal FEV1 among females, those without preexisting respiratory symptoms or asthma, and non-smokers (Table 14). The relationship between exposure and abnormal FVC appeared strongest and in the expected direction among females, those without preexisting respiratory symptoms or asthma, and non-smokers (Table 15). Only among those without preexisting respiratory symptoms did there appear to be a relationship between chlorine exposure and abnormal FEV1/FVC (Table 16).
In general, these findings are not what was expected, in that it would have been predicted that subjects with preexisting respiratory abnormalities or a history of cigarette smoking would be more likely to have lasting effects of chlorine exposure reflected particularly in decreased FEV1. There are several explanations for the findings that those without respiratory abnormalities and non-smokers, showed the greatest effect of exposure when compared with control participants. The controls may have had poorer preexisting respiratory health than the exposed group (also suggested by the increased prevalence of preexisting asthma in control compared to exposed participants), and it is only in removing these subjects that the effects of chlorine on the pulmonary function of exposed subjects are suggested. It is also possible, for the findings among non-smokers, that current or former smoking so overwhelms the respiratory system that the effects of the exposure can only be seen in the non-smokers. Another possibility is that the levels of exposure to chlorine actually experienced by the exposed group are quite variable, and the inclusion of those with much less exposure may diminish the effects which would be seen if only highly exposed individuals were included in the study. This study attempted to select those individuals with greatest potential exposure to chlorine, but given the difficulties in modeling the chlorine exposure it was only possible to use distance and direction from the spill in the selection process. Therefore it is possible that some individuals without much chlorine exposure are included in the exposed group.
The results of methacholine challenge testing indicated that about 36% of exposed and 27% of control participants had an abnormal methacholine challenge test result (Table 17). The 95% confidence intervals for unadjusted and adjusted odds ratios included 1.0, and hence was not considered to be statistically significant, and adjustment lowered the odds ratio (1.5 decreased to 1.3). The interpretation of the stratified analyses is somewhat difficult due to small numbers and resulting very wide confidence intervals, all of which include 1.0. The most interesting result is the nearly three times increased risk of abnormal methacholine challenge test result among exposed versus control participants without preexisting respiratory symptoms.
Additionally, methacholine dose-response slope was created, and the mean slope of exposed and control subjects was compared. There appeared to be some difference in mean slope by exposure, with exposed subjects having a larger mean slope and hence a greater response to methacholine, but after multivariate adjustment, the p-value increased from 0.13 to 0.51 (Table 18). The mean slope of exposed and control subjects was also compared after stratification on sex, cigarette smoking, preexisting respiratory symptoms and preexisting asthma, and it appeared that a difference between exposed and control subjects in the expected direction (larger slope among exposed subjects) was present among "never smokers", those without preexisting respiratory symptoms, and those with preexisting asthma.
The findings from the analysis of the methacholine challenge test results echo those of the spirometry results, which are not what were expected, in that an effect of exposure was seen among those without preexisting respiratory symptoms. This may be due to the reasons noted above for the similar findings with spirometrics.
An exception to not finding what was expected in the methacholine challenge testing analysis is the difference in mean slope and abnormal methacholine challenge results among exposed and control subjects with preexisting asthma. The results of these analyses suggest an effect of exposure on airway hyperresponsiveness among those with preexisting asthma. One question about this sub-analysis concerns the expectation that persons who currently have asthma would have abnormal methacholine challenge test results. That all control subjects with preexisting asthma do not have abnormal test results, may reflect those who had asthma only before the accident, such as asthma in childhood, which has since resolved. This is suggested by the following: of the four exposed subjects with preexisting asthma in the subanalysis, one (25%) had asthma only before the accident; and of the eleven control subjects with preexisting asthma in the subanalysis, six (55%) had asthma only before the accident.
Another issue, involving all of the pulmonary testing as well as the physician examination, is the lack of baseline test results or medical histories. Certainly, in such a situation as that in Alberton, it is obvious that baseline test results will not be available. The presence of small numbers, particularly in the stratified analyses, is also an area of concern. Small numbers are of particular concern in the stratified analyses of methacholine challenge test results, because there may have been differential participation among exposed and control subjects. For example, possible differential participation among exposed and control subjects in the methacholine challenge testing due to abnormal lung function, which is a reason for not administering the methacholine challenge test, could bias the results. This is demonstrated in the following: of 13 exposed participants with preexisting asthma, 4 (31%) completed the methacholine challenge testing, while of 21 control participants with preexisting asthma, 11 (52%) completed the methacholine challenge testing (Table 19). Additionally, the apparent inadvertent recruitment of a control group with poorer respiratory health than expected, possibly related to selection bias issues associated with poorer participation rates, is an additional concern applying to all of the breathing test results, as is the issue of multiple statistical testing.
CONCLUSION
The results of this study suggest that an acute exposure to chlorine following a train accident may have resulted in chronic health effects in exposed persons. This study found increased reporting of nearly each symptom asked about on the questionnaire, including respiratory, gastrointestinal and neurological symptoms. Also noted in the physician examination was evidence of increased eye and skin problems among the exposed. Most importantly, this study found some evidence of poorer performance among the exposed participants in the pulmonary function and methacholine challenge tests, especially among certain subgroups of participants, such as non-smokers.
The difficulty of exposure assessment, particularly modeling to evaluate the chlorine emissions from the derailment, and their migration from the site, resulted in the use of distance and direction from the site as the indicator of exposure. It is possible that assessing exposure in this simple way may have resulted in the inclusion of exposed participants with varying levels of exposure rather than solely those with the greatest exposure. This may account for the lack of strong respiratory health effects findings in the expected strata (such as those with preexisting asthma) in this study.
Given the occupational literature and suggestion of effects noted in this study, if a similar situation were to arise, another study of this sort should be attempted. Hopefully, better air monitoring data and less uncertainty about the toxin release and migration would enable a more accurate exposure assessment.
RECOMMENDATIONS
Based on this study, further work on the chronic effects of chlorine exposure, particularly in non-occupationally exposed populations, would be recommended. This would be particularly true if the exposure were well monitored.
The Division of Health Education and Promotion at ATSDR and the Association of Occupational and Environmental Clinics (AOEC) are currently working on a clinical evaluation plan for exposed persons in Alberton, Montana. It is hoped that this further follow-up will address residents' remaining health concerns about their chlorine exposure, due to the derailment.
Follow-up of the children and adolescents who were exposed with an emphasis on monitoring health care utilization should be considered.
REFERENCES
| Table 1. Description of study participants. | ||
|
Demographics |
Exposed (N=121) |
Controls (N=99) |
| N (%) | N (%) | |
| Age | ||
| 23-35 | 26 (21.5) | 34 (34.3) |
| 36-45 | 37 (30.6) | 22 (22.2) |
| 46-55 | 40 (33.1) | 31 (31.3) |
| 56-65 | 18 (14.9) | 12 (12.1) |
| Sex | ||
| Female | 66 (54.6) | 56 (56.6) |
| Male | 55 (45.5) | 43 (43.4) |
| Education | ||
| Less than 12 years | 9 (7.4) | 2 (2.0) |
| High school grad | 51 (42.1) | 30 (30.3) |
| Tech/trade school | 8 (6.6) | 5 (5.1) |
| Jr/community college | 5 (4.1) | 1 (1.0) |
| 4-yr college, some | 29 (24.0) | 19 (19.2) |
| 4-yr college grad | 15 (12.4) | 24 (24.2) |
| Grad school | 4 (3.3) | 18 (18.2) |
| Family income | ||
| < $20,000 | 33 (27.3) | 19 (19.2) |
| >= $20,000 | 84 (69.4) | 77 (77.8) |
| $20,000-<30,000 | 24 (19.8) | 17 (17.2) |
| $30,000-<40,000 | 22 (18.2) | 20 (20.2) |
| $40,000-<50,000 | 22 (18.2) | 10 (10.1) |
| $50,000 + | 14 (11.6) | 28 (28.3) |
| Refused/missing | 2 (1.7) | 2 (2.0) |
| Refused | 0 | 2 (2.0) |
| Missing | 4 (3.3) | 1 (1.0) |
| Cigarette smoking | ||
| Never | 47 (38.8) | 54 (54.6) |
| Current | 38 (31.4) | 17 (17.2) |
| Former | 36 (29.8) | 28 (28.3) |
| Pack years smoked | ||
| Never smoked | 47 (38.8) | 54 (54.6) |
| <= 5 | 13 (10.7) | 10 (10.1) |
| > 5 - <= 20 | 17 (14.0) | 15 (15.2) |
| > 20 - <= 40 | 28 (23.1) | 8 (8.1) |
| > 40 | 16 (13.2) | 12 (12.1) |
| Passive smoke exposure | ||
| Yes | 58 (47.9) | 25 (25.3) |
| No | 62 (51.2) | 74 (74.8) |
| Unknown | 1 (0.8) | 0 |
| Occupational dust exposure | ||
| Yes | 52 (43.0) | 35 (35.4) |
| No | 69 (57.0) | 64 (64.7) |
| Cotton dust | ||
| Yes | 3 (2.5) | 4 (4.0) |
| No | 118 (97.5) | 95 (96.0) |
| Grain/farming dust | ||
| Yes | 14 (11.6) | 12 (12.1) |
| No | 107 (88.4) | 87 (87.9) |
| Wood/saw dust | ||
| Yes | 37 (30.6) | 28 (28.3) |
| No | 84 (69.4) | 71 (71.7) |
| Sand/silica dust | ||
| Yes | 12 (9.9) | 4 (4.0) |
| No | 108 (89.3) | 95 (96.0) |
| missing | 1 (0.8) | 0 |
| Insulating materials | ||
| Yes | 9 (7.4) | 1 (1.0) |
| No | 112 (92.6) | 98 (99.0) |
| Metal dust | ||
| Yes | 9 (7.4) | 5 (5.1) |
| No | 112 (92.6) | 94 (95.0) |
| Occupational VOC exposure | ||
| Yes | 69 (57.0) | 54 (54.6) |
| No | 52 (43.0) | 45 (45.5) |
| Solvents/cleaning agents/degreasers | ||
| Yes | 51 (42.2) | 41 (41.4) |
| No | 70 (57.9) | 58 (58.6) |
| Plastic/adhesive/glue fumes | ||
| Yes | 16 (13.2) | 6 (6.1) |
| No | 105 (86.8) | 93 (93.9) |
| Paints/paint remover/paint thinner | ||
| Yes | 12 (9.9) | 16 (16.2) |
| No | 109 (90.1) | 83 (83.8) |
| Gasoline/kerosene/lubrication oils | ||
| Yes | 36 (29.8) | 22 (22.2) |
| No | 85 (70.3) | 77 (77.8) |
| Pets (dogs, cats, birds) in home | ||
| Yes | 88 (72.7) | 76 (76.8) |
| No | 33 (27.3) | 23 (23.2) |
| Use of the following at least once a week in the last cold season | ||
| Wood stove/fireplace | ||
| Yes | 61 (50.4) | 26 (26.3) |
| No | 60 (49.6) | 73 (73.7) |
| Coal stove | ||
| Yes | 1 (0.8) | 0 |
| No | 120 (99.2) | 99 (100.0) |
| Kerosene or gas heater, unvented | ||
| Yes | 2 (1.7) | 2 (2.0) |
| No | 119 (98.4) | 97 (98.0) |
| Use of central air conditioning | ||
| Yes | 2 (1.7) | 3 (3.0) |
| No | 119 (98.4) | 96 (97.0) |
| Mold or mildew problem inside house in past year | ||
| Yes | 20 (16.5) | 8 (8.1) |
| No | 99 (81.8) | 91 (91.9) |
| Unknown | 2 (1.7) | 0 |
| Humidifier built into heating system | ||
| Yes | 5 (4.1) | 8 (8.1) |
| No | 113 (93.4) | 89 (89.9) |
| Unknown | 3 (2.5) | 2 (2.0) |
| Humidifier used regularly | ||
| Yes | 26 (21.5) | 22 (22.2) |
| No | 95 (78.5) | 77 (77.8) |
| IgE level | ||
| Low (< 400) | 107 (88.4) | 88 (88.9) |
| High (>= 400) | 5 (4.1) | 7 (7.1) |
| Missing | 9 (7.4) | 4 (4.0) |
| Do you feel you are more sensitive to chemicals than most people? | ||
| Yes | 53 (43.8) | 24 (24.2) |
| No | 65 (53.7) | 74 (74.8) |
| Missing | 3 (2.5) | 1 (1.0) |
| more sensitive before | 8 (6.6) | 24 (24.2) |
| more sensitive after | 53 (43.8) | 23 (23.2) |
| before and after, worse after | 3 (2.5) | 1 (1.0) |
| How concerned are you that the chemicals spilled in the train wreck have harmed your health? | ||
| Not at all | 39 (32.2) | 85 (85.9) |
| A little | 41 (33.9) | 9 (9.1) |
| Very | 41 (33.9) | 3 (3.0) |
| Missing | 0 | 2 (2.0) |
| How concerned are you that the chemicals spilled in the train wreck may harm your health in the future? | ||
| Not at all | 34 (28.1) | 68 (68.7) |
| A little | 37 (30.6) | 24 (24.2) |
| Very | 49 (40.5) | 6 (6.1) |
| Missing | 1 (0.8) | 1 (1.0) |
| How concerned are you that there is something in your neighborhood environment that may be harming your health? | ||
| Not at all | 56 (46.3) | 40 (40.4) |
| A little | 35 (28.9) | 33 (33.3) |
| Very | 30 (24.8) | 25 (25.3) |
| Missing | 0 | 1 (1.0) |
| Table 2. Various self-reported health problems. | ||||
|
Health Problem* |
Since The Train Accident, Are You Having This Health Problem More Than Usual? |
If Yes, Are You Having This Problem Now? | ||
| Yes | No | Yes | No | |
| N (%) | N (%) | N (%) | N (%) | |
| Nausea or vomiting | ||||
| Exposed | 17 (14.0) | 104 (86.0) | 5 (4.1) | 12 (9.9) |
| Controls | 4 (4.0) | 95 (96.0) | 1 (1.0) | 3 (3.0) |
| Diarrhea | ||||
| Exposed | 25 (20.7) | 95 (78.5) | 10 (8.3) | 15 (12.4) |
| Controls | 1 (1.0) | 97 (98.0) | 0 | 1 (1.0) |
| Unusual fatigue or tiredness | ||||
| Exposed | 54 (44.6) | 66 (54.5) | 36 (29.8) | 18 (14.9) |
| Controls | 6 (6.1) | 93 (93.9) | 2 (2.0) | 3 (3.0) |
| Sleepiness or drowsiness | ||||
| Exposed | 42 (34.7) | 79 (65.3) | 26 (21.5) | 16 (13.2) |
| Controls | 5 (5.1) | 94 (94.9) | 3 (3.0) | 1 (1.0) |
| Aching muscles or joints | ||||
| Exposed | 57 (47.1) | 63 (52.1) | 41 (33.9) | 16 (13.2) |
| Controls | 8 (8.1) | 91 (91.9) | 5 (5.1) | 3 (3.0) |
| Dry or itchy skin | ||||
| Exposed | 43 (35.5) | 78 (64.5) | 33 (27.3) | 10 (8.3) |
| Controls | 6 (6.1) | 93 (93.9) | 6 (6.1) | 0 |
| Skin rash | ||||
| Exposed | 23 (19.0) | 98 (81.0) | 13 (10.7) | 10 (8.3) |
| Controls | 3 (3.0) | 95 (96.0) | 2 (2.0) | 1 (1.0) |
| Dry or itching eyes | ||||
| Exposed | 55 (45.5) | 66 (54.5) | 31 (25.6) | 24 (19.8) |
| Controls | 9 (9.1) | 90 (90.9) | 5 (5.1) | 4 (4.0) |
| Sore/strained eyes | ||||
| Exposed | 46 (38.0) | 73 (60.3) | 24 (19.8) | 22 (18.2) |
| Controls | 9 (9.1) | 90 (90.9) | 6 (6.1) | 3 (3.0) |
| Blurry/double vision | ||||
| Exposed | 37 (30.6) | 84 (69.4) | 16 (13.2) | 21 (17.4) |
| Controls | 3 (3.0) | 96 (97.0) | 1 (1.0) | 2 (2.0) |
| Burning eyes | ||||
| Exposed | 42 (34.7) | 79 (65.3) | 20 (16.5) | 22 (18.2) |
| Controls | 5 (5.1) | 94 (95.0) | 1 (1.0) | 4 (4.0) |
| Contact lens trouble | ||||
| Exposed | 8 (6.6) | 113 (93.4) | 4 (3.3) | 4 (3.3) |
| Controls | 5 (5.1) | 94 (95.0) | 2 (2.0) | 3 (3.0) |
| Tearing eyes | ||||
| Exposed | 29 (24.0) | 92 (76.0) | 9 (7.4) | 20 (16.5) |
| Controls | 6 (6.1) | 93 (93.9) | 2 (2.0) | 4 (4.0) |
| Red eyes or conjunctivitis | ||||
| Exposed | 33 (27.3) | 87 (71.9) | 10 (8.3) | 23 (19.0) |
| Controls | 3 (3.0) | 96 (97.0) | 1 (1.0) | 2 (2.0) |
| Puffy eyes | ||||
| Exposed | 21 (17.4) | 99 (81.8) | 9 (7.4) | 12 (9.9) |
| Controls | 5 (5.1) | 94 (95.0) | 3 (3.0) | 2 (2.0) |
| Difficulty breathing | ||||
| Exposed | 41 (33.9) | 79 (65.3) | 13 (10.7) | 28 (23.1) |
| Controls | 6 (6.1) | 93 (93.9) | 2 (2.0) | 4 (4.0) |
| Pain when breathing | ||||
| Exposed | 11 (9.1) | 110 (90.9) | 5 (4.1) | 6 (5.0) |
| Controls | 3 (3.0) | 96 (97.0) | 1 (1.0) | 2 (2.0) |
| Burning throat | ||||
| Exposed | 25 (20.7) | 96 (79.3) | 8 (6.6) | 17 (14.0) |
| Controls | 2 (2.0) | 97 (98.0) | 1 (1.0) | 1 (1.0) |
| Burning chest pain | ||||
| Exposed | 12 (9.9) | 108 (89.3) | 4 (3.3) | 8 (6.6) |
| Controls | 1 (1.0) | 97 (98.0) | 0 | 1 (1.0) |
| Cold or respiratory infection | ||||
| Exposed | 38 (31.4) | 81 (66.9) | 14 (11.6) | 24 (19.8) |
| Controls | 7 (7.1) | 92 (92.9) | 2 (2.0) | 5 (5.1) |
| Stuffy nose or sinus congestion | ||||
| Exposed | 58 (47.9) | 63 (52.1) | 37 (30.6) | 21 (17.4) |
| Controls | 11 (11.1) | 87 (87.9) | 4 (4.0) | 6 (6.1) |
| Nose irritation or bleeding | ||||
| Exposed | 37 (30.6) | 83 (68.6) | 19 (15.7) | 18 (14.9) |
| Controls | 3 (3.0) | 96 (97.0) | 2 (2.0) | 1 (1.0) |
| Throat irritation or hoarseness | ||||
| Exposed | 40 (33.1) | 81 (66.9) | 17 (14.0) | 23 (19.0) |
| Controls | 8 (8.1) | 90 (90.9) | 3 (3.0) | 4 (4.0) |
|
* There were 121 exposed and 99 control participants. Numbers may not add to totals due to missing values. | ||||
| Table 3. Respiratory symptoms and conditions. | ||||||
| Symptom* | Presence of symptom | Presence of symptom in relation to the date of the train accident, 4/11/96 | ||||
| Ever | Never | Only Before | Only After | Both Before and After | If Both, Worse After | |
| N (%) | N (%) | N(%) | N (%) | N (%) | N (%) | |
| Wheezing/whistling | ||||||
| Exposed | 54 (44.6) | 67 (55.4) | 3 (2.5) | 21 (17.4) | 29 (24.0) | 14 (11.6) |
| Controls | 35 (35.4) | 64 (64.6) | 8 (8.1) | 3 (3.0) | 24 (24.2) | 2 (2.0) |
| Chest tightness | ||||||
| Exposed | 41 (33.9) | 80 (66.1) | 3 (2.5) | 23 (19.0) | 14 (11.6) | 5 (4.1) |
| Controls | 16 (16.2) | 83 (83.8) | 4 (4.0) | 0 (0.0) | 12 (12.1) | 1 (1.0) |
| Shortness of breath with non-strenuous activity | ||||||
| Exposed | 34 (28.1) | 87 (71.9) | 0 (0.0) | 24 (19.8) | 10 (8.3) | 4 (3.3) |
| Controls | 16 (16.2) | 83 (83.8) | 5 (5.1) | 0 (0.0) | 10 (10.1) | 0 (0.0) |
| Shortness of breath after strenuous activity | ||||||
| Exposed | 44 (36.4) | 76 (62.8) | 2 (1.7) | 22 (18.2) | 20 (16.5) | 10 (8.3) |
| Controls | 24 (24.2) | 75 (75.8) | 5 (5.1) | 3 (3.0) | 15 (15.2) | 1 (1.0) |
| Woken at night by shortness of breath | ||||||
| Exposed | 21 (17.4) | 99 (81.8) | 1 (0.8) | 15 (12.4) | 4 (3.3) | 1 (0.8) |
| Controls | 14 (14.1) | 85 (85.9) | 5 (5.1) | 2 (2.0) | 7 (7.1) | 0 (0.0) |
| Woken at night by coughing | ||||||
| Exposed | 67 (55.4) | 54 (44.6) | 4 (3.3) | 27 (22.3) | 33 (27.3) | 11 (9.1) |
| Controls | 41 (41.4) | 58 (58.6) | 12 (12.1) | 0 (0.0) | 29 (29.3) | 1 (1.0) |
| Cough in the morning | ||||||
| Exposed | 32 (26.4) | 89 (73.6) | 0 (0.0) | 7 (5.8) | 25 (20.7) | 12 (9.9) |
| Controls | 16 (16.2) | 83 (83.8) | 1 (1.0) | 1 (1.0) | 14 (14.1) | 1 (1.0) |
| Phlegm in the morning | ||||||
| Exposed | 34 (28.1) | 87 (71.9) | 0 (0.0) | 11 (9.1) | 22 (18.2) | 11 (9.1) |
| Controls | 16 (16.2) | 82 (82.8) | 2 (2.0) | 1 (1.0) | 13 (13.1) | 1 (1.0) | <