We experienced many challenges and developed corresponding strategies in conducting these epidemiologic studies after the WTC disaster. The first problem we encountered was recruitment difficulty because a significant number of residents moved out of the affected area after 9/11 and centralized rosters were not available to track WTC responders. Some individuals refused to answer questions related to the disaster that provoked an emotional reaction, or were overwhelmed with requests for participation from multiple agencies.
Voluntary response could introduce selection bias, i.e., people who experienced symptoms and lived or worked in the affected areas were more likely to participate than those who did not. This bias may have affected the representativeness of our studies and could also have led to overestimation of disease incidence in exposed residents and responders. We minimized this bias by: (1) emphasizing the importance of participation regardless of respiratory issues and using general terms such as “breathing problem”; (2) employing similar recruitment efforts in residential areas or among all WTC responders to reduce differential motivation; (3) using exclusion criteria to limit misclassification due to mobility and occupational exposure; (4) obtaining similar response rates and comparable demographics between exposed and control populations; and (5) estimating the change in exposure-disease association for areas targeted and non-targeted for additional recruitment.
Another important challenge is recall or reporting bias, i.e., affected-area participants or exposed responders might have recalled and reported more symptoms than the controls. To minimize this bias, we asked for specific time frames, severity, and frequency of symptoms, and lists of prescription medications. We also assessed memorable events such as new diagnoses, emergency department visits, and hospitalizations, as they are less prone to recall bias. To estimate reporting bias, we also queried conditions unrelated to WTC and excluded individuals who responded affirmatively to all health conditions. Finally, in some sub-groups, we conducted pulmonary function tests, methacholine challenge tests, and used fractional exhaled nitric oxide (FENO) as a biomarker to validate self-reported symptoms.
Another challenge we faced was a lack of objective exposure measurements. A semi-quantitative exposure assessment was conducted which categorized participants by different patterns of exposure. The algorithm incorporated dust and smoke exposure, time period, and duration on-site, location of work, personal protective equipment used, and quantitative data from air monitoring in pertinent locations.10