Review

Introduction

An acute chemical incident may be defined as an unexpected event with a clear onset during which substance(s) are released into the environment and pose a (potential) hazard to humans, animals, or the environment. Such events include fires, explosions, leakages, and accidental or deliberate release of toxic substances that may cause illness, injury, disability, or death (WHO Regional Office for Europe, 1997; Cullinan, 2002). A chemical incident may be divided into three phases. The planning and preparedness phase takes place before incidents happen. The early or response phase begins after the onset of the incident is recognized and ends when rapid interventions are no longer conducted. The late or follow-up phase starts after the termination of the rapid response activities and health studies may be conducted during this phase (WHO Regional Office for Europe, 1997). Experiences from the early and late phases may be implemented in the management and planning processes of the planning and preparedness phase, because this phase is a continuous process with periodic updates and revisions. Reviewing exposure assessments and epidemiological health studies makes it possible to address exposure assessment in the design of contingency plans and disaster management.

Exposure assessment within the framework of chemical incidents has three basic objectives, namely,

1. Risk assessment (based on existing knowledge of the exposure–response relation)
2. Definition of the population or groups at risk
3. Providing (individual) exposure estimates for epidemiological studies

The primary goal of exposure assessment taking place during the first hours of incidents usually is assessment of the situation to be able to decide on measures aimed at controlling and minimizing potential health risks during the incident including advice on evacuation and stay in place. Only limited time is available during and directly after a chemical incident and it may be a challenge to collect information that is essential for all three objectives. During the aftermath, data collected for the initial risk assessment may be used to decide whether a health follow-up study addressing the health effects caused by the released chemical(s) is required or not and the same data may be used to create exposure measures for health studies. Both the risk assessment during a chemical incident and the epidemiological health study that may follow afterward need data on exposure, but data collected for one purpose may not suit the other. For example, environmental sampling for risk assessment often takes place close to the source to estimate worst case scenarios, and these measurements may not accurately represent the exposure of (different) populations. The possibility of long-term health studies should be taken into account during the early phase, and additional information on exposure should be collected accordingly. This additional information may also strengthen the initial risk assessment.

Not every incident calls for a sophisticated individual exposure assessment to confirm or rule out an association between exposure to chemicals and observed health effects. It is useful to know when extensive (personal) exposure assessment is needed and when crude exposure indices based on location of individuals and dispersion of the chemical(s) is sufficient. Some studies may reach conclusive results with a limited amount of data, whereas inadequate exposure assessment may lead to inconclusive studies in other cases.

It is difficult to predict in advance the extent of exposure assessment that will be needed. Therefore, we studied previous health studies and determine to what degree collected data contributed to answering research questions. Several health studies performed after chemical incidents were used to help answering questions about exposure assessment needed for health studies following a chemical incident. In addition to defining data considered important from a conceptual point of view, this literature review is meant to identify problems that may obstruct meaningful exposure assessments in the future. Discussion points and recommendations about exposure assessment described in reviewed health studies were examined as they may give insight on the need to collect exposure data specifically for epidemiological studies during the early stages after a chemical incident.

In the last decade, the Netherlands have been startled by several chemical incidents. After most of these incidents, no clear relation could be observed between exposure to substances released during the incidents and health effects attributed to the exposure (RIVM Project Team Health Research Firework Disaster Enschede, R, 2001; Slottje et al., 2005a), although exposure data collection for some health studies started years after the incident (Boin et al., 2001). To minimize public unrest caused by uncertainty about possible exposure, a political and scientific debate started on whether early measurements could prevent uncertainty concerning possible exposure. In the Netherlands, the Ministry of Health founded the National Expert Center for Health Impact Assessment of Disasters. One of the tasks of the Center is to prepare for rapid exposure assessment after chemical incidents and disasters, and the purpose of this review is to contribute to the preparedness of the Center.

The specific research questions that form the basis of this review include assessment of the type of exposure data that were collected, the role of exposure data in the decision to conduct a health impact assessment, which exposure data were actually used in health studies (i.e., which exposure variables were generated), and whether potentially relevant information was not used and, if so, why not. This review aims to evaluate how information on previous health studies may be used to develop tools and strategies for exposure assessment during and after chemical incidents that include all three objectives (risk assessment, definition of population(s) at risk, and providing exposure measures) to facilitate prospective health studies. This paper focuses on exposure assessment needed for health studies. Exposure assessment for risk assessment including decisions on evacuation, stay in place, recovery, and restoration are beyond the scope of this paper.

Methods

A general search was conducted in Medline for bibliographical files published between 1975 and July 2007. The search criteria were specified to include the keywords epidemiology and expos^*, one of the following four keywords: environmental, industrial, chemical, or hazardous, and one of the following nine keywords: disaster, calamity, catastrophe, chemical incident*, chemical accident*, chemical release, spill, explosion, or fire. Results containing the key word natural disaster* were excluded. The search resulted in 272 hits. Titles and abstracts were screened for acute chemical incidents after which an epidemiological study had been conducted. In all, 28 incidents were identified including 6 accidental industrial chemical releases or spills, 6 chemical fires, 5 oil tanker spills, 3 collisions or derailments of chemical freight trains, 2 explosions in a warehouse or depot, 2 plane crashes, and 4 different other incidents (illegal dump of toxic waste, leak of fire extinguisher gas, truck leak, and chemical spill in hospital).

From these 28 cases, well-known chemical incidents and smaller, less well-known incidents were selected to form a small, but diverse collection of cases. In a few cases when chemical incidents with the same substances involved (e.g., crude oil from oil tanker spills) were encountered, one incident was picked at random. Even though each chemical incident is unique, incidents with the same type of released substances share a certain amount of similarities such as expected health effects, exposure route, and dispersion media. A conscious choice on the number of reviewed cases had to be made, as it was not possible to review all 28 cases within the available time frame. The selected incidents may not form a complete list of interesting cases, but they were selected with confidence that together they cover all research questions. The goal of this review was to perform a qualitative instead of a quantitative study on how health studies were conducted and the exposure assessment needs of health studies.

After the selection of incidents, additional information on related epidemiological studies was obtained through PubMed with a search containing keywords specific to the incidents. Additional known reports on Dutch incidents were added. Health studies related to the selected incidents were reviewed and special attention went in determining what type of data on exposure was collected and when, how exposure measures were established, and how the population at risk was identified.

Results

A chemical incident may be divided into three phases. The planning and preparedness phase takes place before incidents happen. The early phase begins after the onset of the incident is recognized and ends when rapid interventions are no longer conducted. The late phase starts after the termination of the rapid response activities and health studies are conducted during this phase (WHO Regional Office for Europe, 1997). Literature indicates that a considerable part of the information on exposure needed for health studies conducted in the late phase must be collected during the early phase. Several options were observed about how data that were collected during the early phase were used for health studies in the late phase. These observations have been characterized in four types of scenarios (Table 1) based on if or why exposure data were collected in the early phase and whether it was used for health studies in the late phase. The selected chemical incidents were subdivided in the following scenarios (Tables 2, 3, 4 and 5).

Table 1 - Different scenario types concerning the relation between exposure data collected in the early phase and those used for health studies in the late phase.

Full table

Table 2 - Examples of scenario type1 and the information used for risk assessment, defining the population at risk and the exposure estimates.

Full table

Table 3 - Examples of scenario type 2 and the information used for risk assessment, defining the population at risk and the exposure estimates.

Full table

Table 4 - Examples of scenario type 3 and the information used for risk assessment, defining the population at risk and the exposure estimates.

Full table

Table 5 - Examples of scenario type 4 and the information used for risk assessment, defining the population at risk and the exposure estimates.

Full table

Type 1

For incidents classified in the type 1 category (see Table 2 for examples), no exposure data are collected in the early phase due to different reasons. The magnitude of the incident may overwhelm the authorities as was the case with the release of methyl isocyanate (MIC) at the Union Carbide Corporation Plant in Bhopal (India). The release of 27 tons of MIC led to approximately 2500 deaths and several thousands injured within days (Dhara et al., 2002; Sharma, 2002). The absence of timely available funding made it impossible to collect biological samples within a relevant time frame after the hydrofluoric acid (HF) spill in Texas City (TX, USA) (Dayal et al., 1992). No data on exposure to fire extinguisher gas were collected after a fire was extinguished in Taipei (Taiwan), because the extinguisher gas leaked to a residential area by accident and people were caught by surprise (Lo et al., 2006).

The lack of data collected during the early phase indicated that health studies have to rely on data collected during the late phase, which may not be the most important data for chemically reactive species. Often valuable data on exposure that were not collected during the early phase could not be collected during the late phase as was the case with the Bhopal MIC disaster and the Texas City HF spill. In both cases, a mathematical dispersion model was created, but both models were not considered adequate enough for health studies. Insufficient meteorological data and the absence of measured environmental concentrations of the released chemical(s) that should be used to validate the models formed the basis of the inadequacy of the models (Dayal et al., 1992, 1994a, 1994b; Dhara et al., 2002).

The mathematical model created after the fire extinguisher gas leak in Taipei was more successful even though there were no measured concentrations of gas to validate the model (Lo et al., 2006). The small scale of the incident and the geographically uncomplicated area may have contributed to its success. The estimated concentrations of extinguisher gas from the model corresponded with the data collected from a door-to-door interview and observed symptoms (Lo et al., 2006).

The absence of data from environmental sampling led to uncertainties about the exposure dose in Bhopal and Texas City, and, without environmental samples, it was impossible to determine with certainty whether the population of Bhopal was exposed to compounds other than MIC. Other means had to be utilized to create exposure measures when there were no data available from the early phase and mathematical modeling was not possible. Distance to the Union Carbide Plant, exposure duration, and activities during the Bhopal MIC disaster were used in one health study to create exposure indices (Dhara et al., 2002), an approach that was more successful in linking health effects to exposure than other studies that based exposure on distance only. Many health studies conducted after the Bhopal MIC disaster ended with inconclusive results due to poor exposure assessment (Dhara et al., 2002). In addition, lack of funding, expertise and resources, an unfavorable political climate, small sample sizes, and inadequate study designs contributed to the inconclusive results (Dhara, 2002; Crabb, 2004).

After the Texas City HF spill, exposure indices were based on self-reported exposure and the self-report of three key symptoms related to HF exposure collected 1 year after the incident (Dayal et al., 1992). A mathematical dispersion model and a plume map describing the geographical area exposed to HF were discarded when areas of high exposure indicated by the model and map did not correspond with the residence of individuals hospitalized after HF exposure and who were considered to be highly exposed. Adverse health symptoms reported 2 years after the incident were linked to the exposure indices. Data on self-reported use of medication collected 2 years after the incident were applied to validate the exposure indices based on self-reported data (Dayal et al., 1994a, 1994b), but the method remains sensitive to recall and report bias.

The Taipei gas leak distinguishes itself from the other two cases by its smaller scale and the timely collection of biological samples. Blood samples were collected 1 week after the incident and gave impartial results on health effects. A questionnaire focused on health effects caused by the gas exposure included symptoms unrelated to fire extinguisher gas exposure and indicated a 6% potential report bias (Lo et al., 2006).

In conclusion, no data are actively collected during the early phase of a type 1 scenario, which leads to a limited amount of exposure data to study the relation between exposure and subsequent health effects. During the late phase, it might not be possible to collect essential data on exposure that were not collected during the early phase. No environmental sampling took place after the Bhopal MIC exposure and the Texas City HF spill, and meteorological data were insufficient to support mathematical dispersion models. Consequently, other methods had to be used to quantify dose of the released chemicals. Crude estimations of exposure were based on distance to the source and the self-report of exposure, symptoms of exposure, and activity during exposure. These methods are sensitive to bias and inaccuracies led to inconclusive results in several studies concerning the Bhopal MIC disaster and the Texas City HF spill, but the same methods were satisfactory for the health study conducted after the Taipei gas exposure. Sufficient meteorological data, individual exposure data, and measured concentrations of released chemicals were considered essential information in the first two cases, but are exactly the type of data that are lost when it is not properly collected during the early phase or at the beginning of the late phase.

Type 2

Exposure data are collected in the early phase of a type 2 scenario (see Table 3 for examples) for the purpose of risk assessment, but these data are not used for health studies that follow the chemical incident due to various reasons described below.

When air sampling took place during a fire at a storehouse containing chemicals in Schweizerhalle (Switzerland), it was unclear what kinds of substances were involved. The air samples were incomplete or collected too late to give an accurate estimation of exposure. An experimental recreation of the incident showed that over 15,000 substances could have been released at Schweizerhalle (Ackermann-Liebrich et al., 1992). Epidemiological health studies that followed focused on symptoms, morbidity, and mortality in the exposed population with a crude indication of exposure based on location and time. The exposed area was defined by the direction the plume of smoke was carried by the wind and the report of a foul odor. Hair, urine, and blood samples were collected of which the first two were examined for mercury, a compound that was stored in the warehouse, but no increased levels of mercury were found (Ackermann-Liebrich et al., 1992). To our knowledge, the results on the analysis of the blood samples were not published in scientific journals. The study population of an ongoing study focused on respiratory symptoms of preschool children was included in the health study assessing the impact the fire had on the health of the exposed population in the nearby town Basel. The small study population was extended with parents of the children and members of the exposed general population, but the necessary government permission to examine additional subjects from unexposed locations was not given (Ackermann-Liebrich et al., 1992). Blood samples were collected from occupationally exposed firemen and their unexposed colleagues within 2 weeks after the Schweizerhalle fire. Increased levels of serum iron and -glutamyl transpeptidase were found in exposed firemen, but no further interpretations could be made, because other symptoms were not included in the study (Ackermann-Liebrich et al., 1992). On the basis of the study results and the characteristics of the chemicals involved, authorities concluded that short-term respiratory effects could have been caused by the fire, but no long-term health effects were to be expected (Ackermann-Liebrich et al., 1992).

During the 3 days of a metam sodium spill into the Sacramento River caused by a train derailment six miles above Dunsmuir in Northern California (CA, USA), only water samples were collected, because the equipment needed for air sampling was operational after the third day of the incident. Metam sodium degrades into several gases including methylisothiocyanate (MITC) when mixed with water. The water samples were used to create a mathematical model to estimate the potential MITC exposure, but estimated peak exposure exceeded the reference exposure level for lethality, whereas no deaths occurred after the MITC exposure (Alexeeff et al., 1994). Modeling of peak exposure was complicated by the uncertainty in choice of water-to-air ratio, metam sodium decomposition, wind dispersion, and the complex terrain involved (Alexeeff et al., 1994; Cone et al., 1994). The health studies that were conducted based exposure measures on distance to the river, contact with contaminated water, and self-reported duration of exposure instead and the model was disregarded (Cone et al., 1994; Koo et al., 1995).

After a cargo plane crashed in a residential area called the Bijlmermeer in Amsterdam, routine measurements were performed by the fire department to check for dangerous levels of gasses, such as carbon monoxide, sulfur dioxide, and cyanide, and radiation. Actual concentrations were never reported, but no excess of threshold values were measured (Uijt de Haag et al., 2000). The contaminated soil was removed and only increased concentrations of hydrocarbons, such as kerosene, were measured in the soil and ground water (Uijt de Haag et al., 2000). The aftercare focused on minimizing psychological problems caused by traumatic experiences and material loss (Boin et al., 2001), but possible health effects due to released substances were disregarded. Public unrest was fueled by uncertainties surrounding the cargo of the crashed plane, possible exposure to harmful substances, and an increase of medically unexplained physical symptoms in the residents of the affected area. A parliamentary inquiry was launched 6 years after the incident and issued the start of several health studies. Because no proper environmental sampling took place during the early phase, exposure was based on crude indicators such as distance to the incident site, self-reported exposure based on performed tasks, and residents' own perception of exposure. Many health complaints were attributed to the incident by the residents, although GPs indicated that there was only a relation between the incident and a small part of the reported symptoms (Donker et al., 2002). A retrospective risk analysis consisted of three mathematical models that were created to estimate the exposure to uranium oxide. The models made use of meteorological data from Schiphol Airport, located 12 km from the incident site, but there were no measurements of uranium oxide available to validate the models. No long-term health effects caused by exposure to uranium oxide were expected based on the worst case scenario estimations (Uijt de Haag et al., 2000). Occupationally exposed relief workers and hangar workers handling the wreck of the plane reported more autoimmune symptoms, but no difference in the incidence of autoantibodies was found between the exposed and their unexposed colleagues (Slottje et al., 2005a, 2005b). A similar study among residents was canceled due to low response rates (Slottje et al., 2005a, 2005b).

In short, the data collected during the early phases of the three described incidents were considered insufficient, and crude exposure measures were based on location and self-reported exposure. Data that were considered important to create more than crude exposure measures were not collected during the early phase and could not be collected in the late phase. Local authorities were unprepared for the metam sodium spill near Dunsmuir, which resulted in confusion about the released chemical and delayed proper disaster management (Koehler and Van Ness, 1993). As a result, air sampling started after peak exposure had already occurred and could not be used to validate the model or refine the exposure indices based on location and self-reported exposure. Health studies conducted after the Schweizerhalle fire could not make the best use of available exposure data due to problems with study design and acquiring necessary permission. The absence of data on symptoms in firemen for comparison with the other health study and a unexposed control group from the general population were considered flaws that could have been avoided if there had been better communication between different study groups and more cooperation from the government (Ackermann-Liebrich et al., 1992). The local authorities of Amsterdam did not consider to study the health effects of possible exposure to harmful substances and focused on minimizing psychological health problems instead. Because exposure to harmful substances was not ruled out, public unrest increased and unexplained symptoms were attributed to the incident (Boin et al., 2001; Donker et al., 2002). Six years after the incident, when authorities were forced by the public to conduct health studies, it was too late to collect sufficient data on exposure. These three incidents indicate that insufficient exposure data collection during the early phase may hamper health studies.

Type 3

In a scenario of the third type (see Table 4 for examples), exposure data are collected in the early phase for risk assessment, and the data are used in the late phase for health studies. Additional data on exposure may or may not be collected in the late phase.

Environmental sampling took place after the Seveso 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) exposure in Italy (di Domenico et al., 1980) and the Hoechst chemical spill in Germany, which affected the suburb Schwanheim near Frankfurt (Traupe et al., 1997). In both cases, evacuation orders were based on measured concentrations and the affected areas were decontaminated. Urine samples were collected a few days after the Hoechst chemical spill to assess the potential health risk and showed increased concentrations of o-nitrophenol, the metabolite of o-nitroanisole that was released during the incident, in the exposed population when compared with the control population (Heudorf and Peters, 1994). Health studies started after these two incidents based exposure measures on the measured concentrations of the released chemicals (Traupe et al., 1997; Eskenazi et al., 2001).

The health study that followed the Hoechst chemical spill selected exposed and unexposed children from the decontaminated inner zone and the outer zone of Schwanheim, respectively. Reference subjects were selected from nearby areas in Frankfurt and South Hessen. Additional data on exposure were collected with a questionnaire (Traupe et al., 1997). Because a long political debate about the necessity of a health study preceded the study, it was not possible to collect urine samples from the study subjects within the time frame o-nitrophenol can be measured in urine after o-nitroanisole exposure. It is unclear why a study including the individuals from whom urine samples were collected to assess the potential health risk was not conducted.

Studies focusing on short-term health effects of TCDD exposure in Seveso based exposure on the zone of residence that reflected different soil TCDD concentrations, but except for chloracne, no conclusive evidence was found to support the link between short-term health effects and exposure to TCDD (Bertazzi et al., 1998). Small sample size and improper control groups in the early health studies are more likely to have contributed to the inconclusive results than the initial exposure assessment (Bertazzi, 1991; Bertazzi et al., 1998). Fortunately, serum samples collected shortly after the Seveso incident for clinical laboratory tests were properly stored (Mocarelli et al., 1991) and could be used to determine individual exposure when in 1987, 11 years after the incident, the analytic method to measure low concentrations of TCDD in small serum samples was developed (Patterson et al., 1987). Even though later soil sampling with improved technology indicated that the initial soil concentrations were underestimated (Bertazzi, 1991), the initial delineation of exposed areas corresponded with individual TCDD serum concentrations (Landi et al., 1998; Eskenazi et al., 2001), but not in all cases (Warner et al., 2004). Additional data on exposure were collected through questionnaires and proved to be accurate when compared with individual TCDD serum concentrations (Eskenazi et al., 2004). The Seveso Women's Health Study (SWHS), which focuses on long-term effects of TCDD exposure on reproductive health, made use of zone of residence, individual TCDD plasma concentrations, and additional information from questionnaires to create exposure measures, which led to the confirmation or exclusion of a relation between certain health effects and TCDD exposure (Eskenazi et al., 2000, 2001, 2002, 2003, 2004; Warner et al., 2002, 2004). A population database was built after the incident, but contained many inconsistencies due to the hectic social climate and political pressure that prevailed shortly after the incident, definitions lacking scientific certainty, time pressure, and the shortage of technical resources (Bertazzi, 1991). The address information from municipality records proved to be a more reliable tool for long-term longitudinal and mortality studies (Bertazzi et al., 2001).

Extensive environmental sampling took place during a 3-day fire at a large plastics recycling plant in Hamilton (Canada), and led to a 12-h evacuation order. Air, soil, vegetation, and runoff water were screened for substances likely to be released during the combustion of polyvinylchloride plastics (PVCs). Two mathematical models were used to calculate the airborne exposure to dioxins. Elevated levels of chemicals above acceptable levels were measured and estimated with the models, but elevated levels occurred only briefly and were not considered to pose a significant long-term risk (Upshur et al., 2001). The most predominant chemical released was hydrogen chloride, which may cause respiratory problems and irritation to skin, eyes, and throat, but the media and the community were more concerned with exposure to dioxins. Directly after the incident, a registry was formed to record all cases of chloracne, a symptom of dioxin exposure, but no cases of chloracne were reported (Upshur et al., 2001). Short-term health effects, such as respiratory problems and irritation of skin, eyes, and throat, were expected in the exposed population but no increase was noticed in the number of hospital visits and admissions between 14 days before the incident, 12 days after the incident, and the same period of the preceding year. A health survey focused on short-term health effects in the evacuated population took place between 4 and 6 days after the start of the incident. Exposure was based on residence in the evacuated area, but there was concern that the initial risk assessment was insufficient and people outside the evacuated area might have been exposed. Attempts were made to place the health survey results in general context by comparing the results with a previous and unrelated health study, because no control group was included. The comparison was possible for respiratory symptoms, but not for the occurrence of headache and abdominal symptoms (Upshur et al., 2001). Another disadvantage was the preparation time of 3 days that was too short to validate the questionnaire and survey method (Upshur et al., 2001). The study results and the measured and estimated concentrations of released substances led to the conclusion that the population was not exposed to serious short-term health risks. There was concern for intermediate-term health effects, such as prolonged respiratory problems, but the study was not designed to capture intermediate-term health effects (Upshur et al., 2001).

The exposure data collected in the early phase of the described incidents were used to delineate the contaminated area during the late phase and formed a solid base for exposure measures in combination with additional data on exposure acquired through questionnaires and biological sampling.

Type 4

During the early phase of the fourth scenario (see Table 5 for examples), the possibility of a health study following a chemical incident is taken into account from the beginning of the incident. Data are collected during the early phase for risk assessment and to be able to answer questions about exposure that may arise afterward. During the late phase, available data from the early phase may or may not be used for health studies and additional exposure data may be collected. Concerns about the health of the potentially exposed population are often the reason for the collection of additional data to clearly confirm or rule out a relation between exposure and health effects to minimize public concern. There is a possibility that the collected data indicates a low health risk for the exposed population and no health study is initiated.

Uncertainty about the exact nature and quantities of the substances stored at a warehouse that caught fire in St. Bastile Le Grand (Canada), in combination with the public health authorities being unprepared for the incident, resulted in a risk assessment based on worst case assumptions instead of measured data, and the area was evacuated 2 days after the start of the fire. The evacuation lasted for 18 days and the collection of blood samples from the evacuated population started on the third day of the incident (Aubin et al., 1994). Because polychlorinated biphenyls (PCBs) were stored at the warehouse, blood samples were screened for PCB contamination, but no concentrations above detection level were observed. The timely collection of biological samples after the fire at St. Bastile Le Grand compensated for the absence of environmental monitoring during the early phase of the incident and led to the conclusion that no adverse health effects related to PCB exposure after the fire were to be expected. Further studies focused on evaluating the utilization of offered health care to predict the health-care needs during similar incidents in the future (Aubin et al., 1994). The evacuation status was used as exposure measure in the health study.

After the oil tanker the Prestige leaked its cargo and tons of crude oil contaminated the Spanish Coast, the Spanish Government issued the creation of a registry consisting of all people who had been in contact with the oil during cleanup activities to facilitate prospective health studies (Bosch, 2003). The registry contained basic information on the people involved in the cleanup operation, and several studies used the registry as a sample frame to select study populations. Individuals were registered as a certain type of cleaner based on background (paid workers, volunteers, and seamen) and type of cleaning activities (bird cleaners), which was used to create crude exposure indices and additional information on exposure was collected through questionnaires (Suarez et al., 2005; Carrasco et al., 2006; Perez-Cadahia et al., 2007). The incompleteness of the data collected in the register was regretted, because it might have biased the sample frame. For example, about 50% of seamen were excluded from the sample frame due to incomplete data (Suarez et al., 2005). One study focused on long-term health effects in seamen did not make use of the register and contacted exposed and unexposed individuals directly through fishermen cooperatives. Exposure and health effects were based on the results of a questionnaire (Zock et al., 2007). This method of the self-report of both exposure and symptoms is sensitive to bias, but exclusion of individuals who expressed anxiety or attributed their health problems to exposure to crude oil probably minimized bias (Zock et al., 2007). Another study made use of concentrations of volatile organic compounds measured during cleanup activities. Blood samples were collected to refine the exposure indices, and to determine genotoxicity and endocrine disruption (Perez-Cadahia et al., 2007). Although health studies conducted after the Prestige oil spill were considered adequate, small sample size and the lack of a control group were considered methodological disadvantages (Suarez et al., 2005; Carrasco et al., 2006; Perez-Cadahia et al., 2007).

The firework disaster at Enschede (the Netherlands), when an explosion at a firework storage depot and the resulting fire that lasted for 3 days destroyed a residential area of the city, showed that collecting additional exposure data during the early phase is not always an easy task. The Environmental Incident Service, a special unit similar to a HAZMAT team carrying equipment for environmental sampling, was not allowed at the scene of the incident for over an hour, because the situation was considered too dangerous and to avoid members of the team getting in the way of relief workers (Mennen et al., 2001). Air sampling started several hours after the start of the incident and continued for 12 days, whereas grass and dust samples were collected at the same time to assess the impact the incident would have had on human health and the environment. Together with exposure data from the fire department and meteorological data collected by the nearby air base, Twente, a mathematical dispersion model was developed. Measured concentrations and results from the dispersion model led to the conclusion that short-term health effects could be expected, but the exposure to harmful substances was too low to cause long-term health effects (Mennen et al., 2001). Nevertheless, a health study was conducted 3 weeks after the incident during which blood and urine samples were collected. Analysis of the blood and urine samples showed no raised concentrations of harmful substances that may be linked to exposure during the incident (RIVM Project Team Health Research Firework Disaster Enschede, 2001; Roorda et al., 2004), and further health studies were focused on the psychological health effects of the incident and health-care needs. Dust, grass, blood, and urine samples are stored for 10 years should any new information on possible exposure and health effects warrant further investigation.

Environmental sampling took place shortly after the explosion in a warehouse storing ammonium nitrate at the AZF fertilizer factory in Toulouse (France) (Lang et al., 2007). The company owning the factory and the fire department provided information on the emitted substances. Air and water sampling was performed by the regular local air quality-monitoring network and by official institutions (Lang et al., 2007). Data collected by meteorological experts were used to create a mathematical model to estimate the exposure of the population to harmful substances, and possible exposure through soil contamination was estimated with mathematical modeling as well. The exposure assessment based on the measurements collected in the early phase and the estimations from the mathematical models indicated that no potential health effects caused by exposure during or after the explosion were to be expected (Lang et al., 2007). Nevertheless, morbidity possibly related to the incident was monitored to investigate in detail any possible toxic effects of the incident to avoid attribution of unexplained symptoms to the disaster (Lang et al., 2007). A high incidence of ear injuries and mental health problems were noticed, and further health studies focused on ear injury and mental health effects.

The magnitude of the World Trace Center (WTC) attack in New York (NY, USA) made it a complicated case for epidemiological studies. Several thousand people were killed when the impact of two jetliners and subsequent aviation fuel fires completely destroyed the WTC towers and several adjacent buildings. Large quantities of particulate matter and organic vapors were released in a dust cloud and a smoke plume. For approximately 2 months, fires continued to emit combustion products and debris removal resuspended settled dust particles (Lioy et al., 2006). Estimated 250,000–400,000 people in lower Manhattan were exposed to varying degrees over time with the highest being only the first 12–48 h to polluted air that contained airborne asbestos, pulverized concrete, glass fibers, polycyclic aromatic hydrocarbons, polychlorinated furans, and dioxins (Moline et al., 2006). Exposure could be divided into distinct categories based on temporal scales, levels, types of gasses and particles emitted, exposure groups, and location (outdoor or indoor) (Lioy and Georgopoulos, 2006).

Gasses could not be measured during the first hours of the incident, because the measuring devices were not available (Lioy et al., 2006). Concentrations of fine (<2.5 m diameter) and course (2.5–10 m diameter) dust particles were measured as early as 14 September, but analysis of the settled dust indicated over 90% of the dust particles consisted of supercoarse particles (>10 m diameter) (Lioy et al., 2006), which were not measured during the weeks that followed. Blood and urine were collected 5 months after the attack from bystanding pregnant women (Wolff et al., 2005). Sputum was collected from firemen 10 months after 9/11 (Fireman et al., 2004). The samples from exposed groups were compared with their unexposed control groups.

Ground-level environmental air pollution levels were estimated with dispersion and fluid dynamic models. Meteorological data, satellite and aerial images, and information on the major air pollutants were used as model input. Results from the models were used to calculate exposure measures for pregnant women (Wolff et al., 2005). Other studies used job tasks, location, and self-reported duration of exposure to estimate exposure (CDC, 2004; Salzman et al., 2004; Johnson et al., 2005; Herbert et al., 2006). This was confined to post 12 h, as there were no measurements at the beginning and modeling could only reasonably deal with the plume of smoke. Disaster responder health data were collected in a registry. Individual health data were collected with the Syndromic Surveillance of the New York Hospital Emergency Department (Das et al., 2003). In general, respiratory health was found to be stronger affected for firemen involved in the early response and residents who lived close by to the WTC and were in the initial collapse dust event. Several long-term studies started in recent years to determine long-term health effects of the WTC disaster (Moline et al., 2006). In short, valuable exposure data from the early phase are missing. Dust samples helped to reconstruct part of the initial exposure and led to increased use of respiratory protection, but this did not change relief workers and volunteers who barely used respiratory protection during the first 12–24 h.

In conclusion, the register created during the early phase after the Prestige oil spill helped to identify the exposed population and provided data on exposure, but additional data were collected to improve the exposure assessment. Extensive data collection after the incidents in Enschede and Toulouse helped to rule out the possibility of health effects related to exposure to harmful substances. One of the goals in both cases was to avoid attribution of health problems to possible exposure similar to what was observed after the Amsterdam air disaster (Donker et al., 2002). The WTC disaster shows that even though environmental sampling started early, the collected data from the early phase were incomplete due to the absence of the right equipment or the use of standard sampling protocols that were not adequate for the situation.

Discussion

Chemical incidents are as diverse as the number of different harmful substances that may be released during them. It is impossible to prepare specifically for all possible chemical incidents, but certain recommendations may be made to include the needs of epidemiological studies in the design of contingency plans.

Upon reviewing the health studies conducted after several chemical incidents, three basic variables were considered essential to know when conducting health studies after chemical incidents. Those three essential variables are as follows:

1. The identification of released substances
2. The concentrations of released substances
3. The dispersion of released substances

They are important for assessing the potential health risk, defining the populations at risk, and creating exposure measures for epidemiological studies. In many of the reviewed cases, exposure data collected for initial risk assessment covered these three basic elements, but were still insufficient for creating adequate exposure measures for health studies.

The Identification of Released Substances

Identifying a chemical reveals its characteristics that determine the adverse effects it will have on human health. Therefore, knowing what kind of substances a population is exposed to and, subsequently, knowing what kind of health effects are to be expected is essential when designing a health study. This makes identification of the substances in the late phase essential, but identification of released substances is even more crucial in the early phase when the right sampling methods need to be applied. Measurements conducted during the Schweizerhalle fire might have given an incomplete image of the released chemicals, because the involved substances were unknown when environmental sampling took place (Ackermann-Liebrich et al., 1992). After the Enschede firework disaster, it was not possible to measure barium in blood due to high barium concentrations in the blood test tubes (RIVM Project Team Health Research Firework Disaster Enschede, 2001) and standard sampling procedures were not adequate enough to collect all necessary exposure data after the WTC disaster (Lioy et al., 2006). A timely identification of the released substances should prevent the use of inappropriate sampling equipment and methods. It may be the difference between a type 2 scenario in which exposure data from the early phase are unsuitable for health studies and a type 3 scenario in which exposure data from the early phase may be used for creating exposure measures for health studies. Although it is impossible to anticipate all substances that may be released during a chemical incident, the design of contingency plans and regulations concerning chemicals may make information about substances involved in spills and fires rapidly available. This may prevent incomplete exposure data limited in their use for health studies as was the case after the metam sodium spill in Dunsmuir (Alexeeff et al., 1994; Cone et al., 1994) and the fire in Schweizerhalle (Ackermann-Liebrich et al., 1992).

The Concentrations of Released Substances

The concentrations of the released chemicals may be obtained from environmental, personal, and biological sampling, but also from environmental and personal modeling. It may be used for risk assessment, definition of the population at risk, and the creation of exposure measures needed for epidemiological studies. Measurements of peak exposures may be used for risk analysis. For example, when the highest measured concentrations are within acceptable range, health risks are low.

The Dispersion of Released Substances

Information on the dispersion of released substances may be used to identify the population at risk and create crude exposure measures. In case of evacuation, the size of the evacuated area depends on the dispersion of harmful chemicals. Crude exposure measures (exposed and unexposed) may be based on the area affected by the released substances.

Methods of Exposure Data Collection

The collection of accurate and comprehensive exposure data during the early and late phase is necessary to create reliable exposure measures and is, therefore, invaluable for health studies. A type 1 scenario with no data collection during the early phase should be avoided whenever possible. The same applies to the collection of data that may be used for risk assessment, but are unsuitable for health studies, because they are inaccurate, incomplete, or irrelevant for health studies as is the case with a type 2 scenario. In the unfavorable circumstance, no (useful) exposure data from the early phase are available, information on exposure should be obtained as early as possible in the late phase, because collecting data needed for an adequate exposure assessment may become difficult or impossible in the late phase. Exposure measures may be based on data from crude indicators of exposure, interviews, questionnaires and registries, environmental sampling and modeling, personal sampling and modeling, and biomonitoring. These different methods usually provide either qualitative data or quantitative data. Combining data collected with different methods may minimize bias and inaccuracies.

Crude Indicators of Exposure

Crude indicators of exposure such as self-reported exposure and exposure duration, distance from the source, and wind direction may be used to create exposure measures. In a type 1 scenario, this is often the only data that may be collected. This type of information was sufficient to create adequate exposure zones after the Taipei gas leak (Lo et al., 2006), but the same type of information was not enough to create reliable exposure measures after the incidents at Bhopal (Dhara et al., 2002) and Texas City (Dayal et al., 1992). Areas that have been evacuated or decontaminated may also function as crude exposure measures. When an evacuation order is based on environmental sampling in the early phase, such as after the Seveso TCDD exposure and the PVC fire in Hamilton, the collected data answer the three objectives of exposure assessment. There is an indication of potential health risk and the population at risk, whereas crude exposure measures may be based on evacuation status. Nonetheless, considering the pressure (of time) under which evacuation orders are given, it is possible that these orders are based on insufficient information and are sensitive to human error. For example, after the PVC fire in Hamilton, it was thought that people outside the evacuated area might have been exposed (Upshur et al., 2001), and additional soil samples collected with advanced technology several decades after the Seveso TCDD exposure indicated that the initial soil concentrations of TCDD were underestimated (Bertazzi, 1991). Additional exposure data collected during the late phase should minimize misclassification of study subjects.

Questionnaires, Interviews, and Registries

Questionnaires and interviews, for example, focusing on (experienced) exposure, activities, and location (inside or outside) during the incident, consumption of (home grown) produce, and animal mortality are valuable tools to collect data that may supplement the exposure assessment as was the case with the Seveso TCDD exposure and the Prestige oil spill, but unlike these two cases no additional data could be collected to support the exposure assessment after the Bhopal MIC exposure, the Texas City HF spill, the Taipei gas leak, and the Amsterdam air disaster. Self-reported exposure is also sensitive to bias. For example, both the exposed and the unexposed study groups in the Texas City HF spill study reported a decrease of health complaints 2 years after the incident. This was attributed to misclassification of the study subjects (Dayal et al., 1992), but may also be caused by report bias. The questionnaire used in the study after the Taipei gas exposure indicated a 6% potential report bias (Lo et al., 2006). Nevertheless, questionnaires are valuable tools for collecting data on exposure. Especially after major events, a majority of the affected population will remember the incident very well.

Registries may provide information on exposure as well, if they include more than the minimum of information needed to identify individuals. The basic cleaning activities of individuals described in the Prestige oil spill registry were used to create crude exposure indices, but the incompleteness of the registry may have resulted in selection bias (Suarez et al., 2005). The early creation of a register for (potentially) exposed individuals may assist prospective health studies with the identification of the exposed population. They need to be carefully compiled, kept, and updated or they may produce bias or even be rendered useless as was the case with the registers created after the Prestige oil spill (Suarez et al., 2005) and the Seveso TCDD exposure (Bertazzi et al., 2001). It is important to keep in mind that in certain countries, laws protecting privacy may limit the use of registries or make them unavailable for health studies. Additional data on (personal) exposure are recommended to supplement exposure data from questionnaires and registries to reduce subject misclassification and correct for bias.

Environmental Sampling

Environmental sampling may not only directly provide exposure measures and assist risk assessment, but may also support the creation and validation of mathematical dispersion models. If not all released substances are known, environmental sampling may help to identify them during or after a chemical incident. Even today, it is unresolved whether the populations in Bhopal and Dunsmuir were exposed to substances other than MIC and MITC, respectively, or not, and it has limited adequate treatment of the exposed. Environmental sampling may also be used to confirm or rule out exposure to harmful substances, which happened after the incidents at Enschede and Toulouse. Through these measurements, public concern might have been decreased, whereas the lack of conclusive evidence that the population in the Bijlmermeer (Amsterdam) was not exposed to harmful substances was reason for public unrest and health symptoms were attributed to exposure during the plane crash in Amsterdam (Boin et al., 2001).

It is essential that environmental sampling takes place as early as possible, because released substances may only be present for a limited amount of time. Being prepared to perform environmental sampling is a key issue. The safety of people performing data collection, however, outweighs the need for data collection and may lead to loss of exposure data, as was the case during the Enschede fireworks disaster. Handheld and portable sampling equipment may provide opportunities to collect exposure data without additional human risk, and possibilities of how these may be combined with other techniques, such as drones or robots, can be explored. Handheld or portable sampling equipment may also provide flexibility and fast application during the early stages of incidents and disasters. Prolonged environmental sampling during recovery and restoration events may facilitate health studies concerning cleanup workers.

Personal Sampling, Personal Modeling, and Biomonitoring

In many reviewed studies, crude exposure measures like low, medium, and high exposure were used, because it was not possible to determine individual exposure. The lack of individual exposure assessment does not have to lead to inconclusive results, but in certain cases a more refined exposure assessment was regretted. In an ideal situation, every member of the potentially exposed population would be carrying personal samplers during a chemical incident to provide data on individual exposure. This is of course impossible to obtain. The next option would be to equip relief workers with sampling equipment, but it may conflict with their main task, which is to get a (potentially) dangerous situation under control and rescue people in danger. Subsequently, relief workers might be reluctant to concern themselves with equipment that does not directly contribute to the safety of victims or themselves. In addition, it is impossible to anticipate all substances that will be released during an incident and equip relief workers accordingly. It remains difficult to determine individual exposure during the early phase, but biological sampling and modeling are good methods to obtain individual exposure data in the early and late phase.

Individual concentrations measured in biological samples or modeled with the aid of biological samples and/or questionnaires may be suitable for creating individual exposure measures. In most cases, biological sampling will take place during the late phase, because the collection of biological samples requires a certain amount of organization and logistics. Nevertheless, the collection of biological samples should start as early as possible. At the start of the health studies after the Texas City HF spill (Dayal et al., 1992) and the Hoechst chemical spill (Traupe et al., 1997), it was already too late for biological sampling in the study populations, but biological samples were collected in a timely manner after the fires at Schweizerhalle (Ackermann-Liebrich et al., 1992), St. Bastile Le Grand (Aubin et al., 1994), and Enschede (Roorda et al., 2004) and indicated low exposure in these three cases. Induced sputum samples collected from New York firemen working at the WTC site during the WTC disaster (Fireman et al., 2004) were able to further characterize the actual dust exposure after the limitations of the used environmental sampling methods were discovered (Moline et al., 2006).

If a chemical has a high half-life period, it may even be possible to model the initial body burden with the aid of biological samples obtained a substantial period after the incident. Serum samples collected shortly after the Seveso TCDD exposure were not available for all exposed individuals, but it was possible to back-extrapolate initial TCDD concentrations with the aid of blood samples obtained 20 years later (Eskenazi et al., 2004).

When there are uncertainties about the involved chemicals, it may be difficult to apply the right sampling technique and equipment. Preservatives, sampling volume, and storage containers may limit the analytical procedures as was the case when blood tubes used in the Enschede firework disaster health study turned out to contain barium, eliminating the possibility of screening blood samples for that substance (RIVM Project Team Health Research Firework Disaster Enschede, R, 2001). The recommendation is made to collect several different kinds of biological samples, especially when not all released key substances are identified at the moment of sampling. Some consider it better to collect too many biological samples and to collect them "too early," than miss the opportunity to gather data that may be irretrievably lost later and be faced with insufficient exposure data (Cullinan, 2002). In contrast, the proper collection, storage, and analysis of biological samples is costly and time consuming, it requires medical ethical approval, and it may be difficult to convince involved authorities of the necessity. A careful decision has to be made between costs and effort on one side and the need for adequate exposure assessment on the other. Despite the fact that local authorities of Enschede did not support collection of biological samples and argued that no toxic substances were involved, the Dutch Ministry of Health carried out a health study to rule out exposure and avoid public unrest (Boin et al., 2001). Should new information on possible incident-related exposure be presented later on, stored biological samples may aid to rule out or confirm this exposure and avoid unnecessary public unrest.

Environmental Modeling

Although biological and environmental sampling may take place in the late phase, peak exposure and concentrations of substances with a short half-life may only be measured in the early phase, and it is not always possible to collect all necessary data during the early and late phase. Therefore, environmental modeling based on data from the early and late phase may provide additional information on the dispersion of the substances and the concentration of substances at different locations and time intervals. With the results from models, the population at risk may be identified and exposure measures may be created. Modeling makes it possible to estimate concentrations of released chemicals for periods of which no measured data are available, which, for example, may be at the beginning of the early phase before environmental sampling takes place, as was the case with the Enschede firework disaster. When using mathematical models to define dispersion, it is important to keep in mind that general wind direction and speed may influence the general dispersion, but winds at ground level may deviate from the general wind direction and speed. This may cause differences in concentrations that may not be visible when using only general wind characteristics and may explain why the mathematical model created after the Texas City HF spill did not correspond with other data on high-exposure areas. Certain mathematical models, for example computational fluid dynamics models, may give estimates for concentrations at complex terrains, such as the WTC site (Wolff et al., 2005), where substances do not disperse uniformly, provided there is sufficient (meteorological) data available for modeling and validation, which was not the case after the Bhopal MIC disaster.

Meteorological data are essential for environmental modeling, but measured concentrations of released substances may function as additional input and to validate the model. The MITC dispersion model created after the metam sodium spill in Dunsmuir was based on measured ratios of air–water concentrations (Alexeeff et al., 1994), and the model created after the Enschede firework disaster was adjusted with measured concentrations (Mennen et al., 2001). Possibly, the small scale of the Taipei gas leak contributed to a functional model without validation with measured concentrations (Lo et al., 2006). Nevertheless, adequate environmental sampling during the early phase for risk assessment and prospective health studies remains important.

In short, the reviewed cases showed that data on the identities, concentration, and dispersion of released substances are essential for adequate exposure assessment needed for health studies. Different methods are available to collect data on exposure, ranging from sampling and modeling to questionnaires and registers. Some methods are sensitive to bias, whereas others have a small window in which they may be applied. Supplementing exposure data with data collected with another method may minimize bias and inaccuracy. What most data collection methods have in common is that they need to start as early as possible. Quick analysis of the collected data may reveal limitations and leave room for additional data collection (early) in the late phase as compensation.

Although proper exposure assessment is essential to conclusive health studies, literature showed that shortcomings other than poor exposure assessment may trouble health studies. The lack of timely available funding was named as one of the obstacles that obstructed the health study after the Texas City HF spill (Dayal et al., 1992, 1994a). Funding was also an obstacle for the health studies conducted after the Bhopal MIC exposure and led to the termination of cohort studies before health effects with long latency periods could be studied. The inconclusiveness of several health studies conducted at Bhopal are also attributed to the unfavorable political climate and the lack of expertise, which are not completely unrelated to India's status as a developing country (Dhara, 2002; Crabb, 2004). Poor communication between different study groups was mentioned to hamper investigations at Bhopal (Dhara, 2002), Seveso (Bertazzi, 1991), and Schweizerhalle (Ackermann-Liebrich et al., 1992). Small sample size of the study population and the absence of control groups are criticized several times by different study groups as was the short period of time available to design health studies. The latter forced the hasty creation of questionnaires that were used untested or were missing important questions as was the case with health studies conducted after the Hamilton PVC fire (Upshur et al., 2001) and the Enschede firework disaster (RIVM Project Team Health Research Firework Disaster Enschede, 2001). Logistics in combination with a short preparation time led to a 6% loss of blood samples in the Enschede health study, and there was no time to collect blood and urine samples from a control group (RIVM Project Team Health Research Firework Disaster Enschede, 2001) and no additional exposure data could be collected during the first hour a special team for taking measurements was present at the scene during the Enschede fireworks disaster (Mennen et al., 2001). The hasty design of an epidemiological study after the Hoechst chemical spill caused an ethnic difference between the study population and the control group, which was noticed afterward (Traupe et al., 1997). Training may lead to better coordination and situational awareness.

Although the type 4 scenario is the preferred situation in case a chemical incident or disaster takes place, the reviewed cases show that there is still room for improvement. Exposure data collection for health studies after such incidents should be an integral part of disaster management to obtain a better understanding of the possible health risks. In addition to collecting as much valuable data as possible, thought should be given in advance on how to evaluate as fast as possible what kind of data is important and how to collect it. Not all chemical incidents will require massive data collection. In some cases, the standard environmental sampling procedures in combination with questionnaires will be sufficient, whereas in other cases, each available technique that is not applied may represent a missed opportunity. Rapid exposure assessment methods and protocols are needed to support quick decisions on tailored exposure assessment needs and strategies.

Conclusion

This literature review showed that proper exposure assessment after chemical incidents is crucial for studying associations between exposure and health effects.

A type 4 scenario, in which exposure data collection for the purpose of conducting a health study is part of the early response, is preferred from an epidemiological point of view. Although extensive data collection may be a costly procedure and conclusive health studies have been conducted with available data from type 1, 2, and 3 scenarios, it is important not to underestimate the value of proper exposure assessment. It may be the difference between a study with conclusive results and uncertainties leading to public unrest.

During the early phase, emphasis should be on collecting data regarding identification, concentrations, and dispersion of released chemicals. Information regarding these three aspects should be collected during the early phase, because it might not be possible to collect all necessary data in the late phase.

Exposure data that are not collected during the early phase should be collected as early as possible in the late phase to supplement the initial exposure assessment. This applies, in particular, to exposure data from questionnaires and biological sampling that are often difficult to collect in the early phase, due the time-consuming nature of these methods. Dispersion models may be created in the late phase as long as sufficient data have been collected in the early phase to validate the exposure models.

Not only poor exposure assessments troubled the reviewed health studies, but it is also impossible to anticipate all possible circumstances that might occur during a chemical incident and prepare accordingly. However, many problems that may trouble health studies including poor exposure assessment may be prevented when the general design and needs of health studies are taken into account when designing contingency plans as part of (inter)national disaster management programs.

Together with steps that will help facilitate funding, design, and coordination of health studies, measures that will lead to useful exposure assessment during and after a chemical incident should be prepared for. Those measures should include methods that lead to a swift identification of released substances, and determination of concentrations and dispersion of released substances. Questionnaires may provide valuable information to support exposure assessment and a basic questionnaire outline should be designed in advance and leave room for quick adaptation to the unique characters of future chemical incidents. In addition, protocols or methods that can support rapid decisions on the usefulness and necessity of employing biological sampling are needed.

References

1. Ackermann-Liebrich U.A., Braun C., and Rapp R.C. Epidemiologic analysis of an environmental disaster: the Schweizerhalle experience. Environ Res 1992: 58(1): 1–14. | Article | PubMed | ChemPort |
2. Alexeeff G.V., Shusterman D.J., Howd R.A., and Jackson R.J. Dose–response assessment of airborne methyl isothiocyanate (MITC) following a metam sodium spill. Risk Anal 1994: 14(2): 191–198. | Article | PubMed | ChemPort |
3. Aubin J., Potvin L., Beland F., and Pineault R. Utilization of a specialized clinic following an ecological accident. Med Care 1994: 32(1): 1–14. | Article | PubMed | ChemPort |
4. Bertazzi P.A. Long-term effects of chemical disasters. Lessons and results from Seveso. Sci Total Environ 1991: 106(1–2): 5–20. | Article | PubMed | ChemPort |
5. Bertazzi P.A., Bernucci I., Brambilla G., Consonni D., and Pesatori A.C. The Seveso studies on early and long-term effects of dioxin exposure: a review. Environ Health Perspect 1998: 106(Suppl. 2): 625–633. | Article | PubMed | ChemPort |
6. Bertazzi P.A., Consonni D., Bachetti S., Rubagotti M., Baccarelli A., and Zocchetti C., et al. Health effects of dioxin exposure: a 20-year mortality study. Am J Epidemiol 2001: 153(11): 1031–1044. | Article | PubMed | ChemPort |
7. Boin A., van Duin M., and Heyse L. Toxic fear: the management of uncertainty in the wake of the Amsterdam air crash. J Hazard Mater 2001: 88(2–3): 213–234. | Article | PubMed | ChemPort |
8. Bosch X. Exposure to oil spill has detrimental effect on clean-up workers' health. Lancet 2003: 361(9352): 147. | Article | PubMed |
9. Carrasco J.M., Lope V., Perez-Gomez B., Aragones N., Suarez B., and Lopez-Abente G., et al. Association between health information, use of protective devices and occurrence of acute health problems in the Prestige oil spill clean-up in Asturias and Cantabria (Spain): a cross-sectional study. BMC Public Health 2006: 6: 1. | Article | PubMed |
10. Centers for Disease Control and Prevention (CDC). Preliminary report 2,3,7,8-tetrachlorodibenzo-p-dioxin exposure to humans-seveso, Italy. MMWR Morb Mortal Wkly Rep 1988: 37(48): 733–736.
11. Centers for Disease Control and Prevention (CDC). Physical health status of World Trade Center rescue and recovery workers and volunteers — New York City, July 2002–August 2004. MMWR Morb Mortal Wkly Rep 2004: 53(35): 807–812.
12. Cone J.E., Wugofski L., Balmes J.R., Das R., Bowler R., and Alexeeff G., et al. Persistent respiratory health effects after a metam sodium pesticide spill. Chest 1994: 106(2): 500–508. | Article | PubMed | ChemPort |
13. Crabb C. Revisiting the Bhopal tragedy. Science 2004: 306(5702): 1670–1671. | Article | PubMed | ChemPort |
14. Cullinan P. Epidemiological assessment of health effects from chemical incidents. Occup Environ Med 2002: 59(8): 568–572; quiz 572,516. | Article | PubMed |
15. Das D., Weiss D., Mostashari F., Treadwell T., McQuiston J., and Hutwagner L., et al. Enhanced drop-in syndromic surveillance in New York City following September 11, 2001. J Urban Health 2003: 80(2 Suppl. 1): i76–i88. | PubMed |
16. Dayal H.H., Baranowski T., Li Y.H., and Morris R. Hazardous chemicals: psychological dimensions of the health sequelae of a community exposure in Texas. J Epidemiol Community Health 1994a: 48(6): 560–568. | ChemPort |
17. Dayal H.H., Brodwick M., Morris R., Baranowski T., Trieff N., and Harrison J.A., et al. A community-based epidemiologic study of health sequelae of exposure to hydrofluoric acid. Ann Epidemiol 1992: 2(3): 213–230. | PubMed | ChemPort |
18. Dayal H.H., Li Y.H., Dayal V., Mittal C.K., and Snodgrass W. Use of medication data to validate an association in community-based symptom prevalence studies. Arch Environ Health 1994b: 49(2): 93–97. | ChemPort |
19. Dhara V.R. What ails the Bhopal disaster investigations? (And is there a cure?). Int J Occup Environ Health 2002: 8(4): 371–379. | PubMed |
20. Dhara V.R., and Dhara R. The Union Carbide disaster in Bhopal: a review of health effects. Arch Environ Health 2002: 57(5): 391–404. | PubMed |
21. Dhara V.R., Dhara R., Acquilla S.D., and Cullinan P. Personal exposure and long-term health effects in survivors of the union carbide disaster at Bhopal. Environ Health Perspect 2002: 110(5): 487–500. | PubMed | ChemPort |
22. di Domenico A., Silano V., Viviano G., and Zapponi G. Accidental release of 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) at Seveso, Italy. II. TCDD distribution in the soil surface layer. Ecotoxicol Environ Saf 1980: 4(3): 298–320. | Article | PubMed | ChemPort |
23. Donker G.A., Yzermans C.J., Spreeuwenberg P., and van der Zee J. Symptom attribution after a plane crash: comparison between self-reported symptoms and GP records. Br J Gen Pract 2002: 52(484): 917–922. | PubMed | ChemPort |
24. Eskenazi B., Mocarelli P., Warner M., Chee W.Y., Gerthoux P.M., and Samuels S., et al. Maternal serum dioxin levels and birth outcomes in women of Seveso, Italy. Environ Health Perspect 2003: 111(7): 947–953. | PubMed | ChemPort |
25. Eskenazi B., Mocarelli P., Warner M., Needham L., Patterson Jr. D.G., and Samuels S., et al. Relationship of serum TCDD concentrations and age at exposure of female residents of Seveso, Italy. Environ Health Perspect 2004: 112(1): 22–27. | PubMed | ChemPort |
26. Eskenazi B., Mocarelli P., Warner M., Samuels S., Needham L., and Patterson D., et al. Seveso Women's Health Study: does zone of residence predict individual TCDD exposure? Chemosphere 2001: 43(4–7): 937–942. | Article | PubMed | ChemPort |
27. Eskenazi B., Mocarelli P., Warner M., Samuels S., Vercellini P., and Olive D., et al. Seveso Women's Health Study: a study of the effects of 2,3,7,8-tetrachlorodibenzo-p-dioxin on reproductive health. Chemosphere 2000: 40(9–11): 1247–1253. | Article | PubMed | ChemPort |
28. Eskenazi B., Mocarelli P., Warner M., Samuels S., Vercellini P., and Olive D., et al. Serum dioxin concentrations and endometriosis: a cohort study in Seveso, Italy. Environ Health Perspect 2002: 110(7): 629–634. | PubMed | ChemPort |
29. Fireman E.M., Lerman Y., Ganor E., Greif J., Fireman-Shoresh S., and Lioy P.J., et al. Induced sputum assessment in New York City firefighters exposed to World Trade Center dust. Environ Health Perspect 2004: 112(15): 1564–1569. | PubMed | ChemPort |
30. Herbert R., Moline J., Skloot G., Metzger K., Baron S., and Luft B., et al. The World Trade Center disaster and the health of workers: five-year assessment of a unique medical screening program. Environ Health Perspect 2006: 114(12): 1853–1858. | PubMed |
31. Heudorf U., and Peters M. [ Human biomonitoring after a severe chemical accident — results of a study after the Hoechst AG chemical spill 22 February 1993] . Gesundheitswesen 1994: 56(10): 558–562. | PubMed | ChemPort |
32. Johnson S.B., Langlieb A.M., Teret S.P., Gross R., Schwab M., and Massa J., et al. Rethinking first response: effects of the clean up and recovery effort on workers at the world trade center disaster site. J Occup Environ Med 2005: 47(4): 386–391. | Article | PubMed |
33. Koehler G.A., and Van Ness C. The emergency medical response to the Cantara hazardous materials incident. Prehosp Disaster Med 1993: 8(4): 359–365. | PubMed | ChemPort |
34. Koo D., Goldman L., and Baron R. Irritant dermatitis among workers cleaning up a pesticide spill: California 1991. Am J Ind Med 1995: 27(4): 545–553. | Article | PubMed | ChemPort |
35. Landi M.T., Consonni D., Patterson Jr. D.G., Needham L.L., Lucier G., and Brambilla P., et al. 2,3,7,8-Tetrachlorodibenzo-p-dioxin plasma levels in Seveso 20 years after the accident. Environ Health Perspect 1998: 106(5): 273–277. | Article | PubMed | ChemPort |
36. Lang T., Schwoebel V., Diene E., Bauvin E., Garrigue E., and Lapierre-Duval K., et al. Assessing post-disaster consequences for health at the population level: experience from the AZF factory explosion in Toulouse. J Epidemiol Community Health 2007: 61(2): 103–107. | Article | PubMed |
37. Lioy P.J., and Georgopoulos P. The anatomy of the exposures that occurred around the World Trade Center site: 9/11 and beyond. Ann NY Acad Sci 2006: 1076: 54–79. | Article | PubMed |
38. Lioy P.J., Pellizzari E., and Prezant D. The World Trade Center aftermath and its effects on health: understanding and learning through human-exposure science. Environ Sci Technol 2006: 40(22): 6876–6885. | PubMed | ChemPort |
39. Lo S.H., Chan C.C., Chen W.C., and Wang J.D. Grand rounds: outbreak of hematologic abnormalities in a community of people exposed to leakage of fire extinguisher gas. Environ Health Perspect 2006: 114(11): 1713–1717. | PubMed | ChemPort |
40. Mennen M.G., Kliest J.J.G., and van Bruggen M. Vuurwerkramp Enschede: Metingen van concentraties, verspreiding en depositie van schadelijke stoffen: [ Fireworks Tragedy in Enschede, The Netherlands: Measurements for Concentrations, Dispersion and Deposition of Harmful Substances: Report of the Environment Study] . RIVM, Bilthoven, 2001.
41. Mocarelli P., Needham L.L., Marocchi A., Patterson Jr. D.G., Brambilla P., and Gerthoux P.M., et al. Serum concentrations of 2,3,7,8-tetrachlorodibenzo-p-dioxin and test results from selected residents of Seveso, Italy. J Toxicol Environ Health 1991: 32(4): 357–366. | PubMed | ISI | ChemPort |
42. Moline J., Herbert R., and Nguyen N. Health consequences of the September 11 World Trade Center attacks: a review. Cancer Invest 2006: 24(3): 294–301. | Article | PubMed |
43. Patterson Jr. D.G., Hampton L., Lapeza Jr. C.R., Belser W.T., Green V., and Alexander L., et al. High-resolution gas chromatographic/high-resolution mass spectrometric analysis of human serum on a whole-weight and lipid basis for 2,3,7,8-tetrachlorodibenzo-p-dioxin. Anal Chem 1987: 59(15): 2000–2005. | Article | PubMed | ChemPort |
44. Perez-Cadahia B., Lafuente A., Cabaleiro T., Pasaro E., Mendez J., and Laffon B. Initial study on the effects of Prestige oil on human health. Environ Int 2007: 33(2): 176–185. | Article | PubMed | ChemPort |
45. RIVM Project Team Health Research Firework Disaster Enschede. Vuurwerkramp Enschede: Stoffen in bloed en urine; rapportage van het gezondheidsonderzoek [ Firework Disaster Enschede: Measurements of Elements in Blood and Urine; Health Impact Assessment] . RIVM, Bilthoven, 2001.