and for the opportunity to respond to their comments. Our overall conclusion matches that of Drs. Scialli and Gibb, which is that additional study of the risk of congenital heart defects (CHDs) from trichloroethylene (TCE) is warranted. We believe the language of the article's conclusion, "that the observation significantly adds to the overall body of literature that suggests (emphasis added) TCE is a cardiac teratogen," was appropriately selected and that it is consistent with this overall conclusion. Notably, we did not state that the specific distance used in our study constitutes a risk factor.
As we previously noted, and as Drs. Scialli and Gibb reiterated, several methodologic limitations, including the use of a surrogate exposure variable, merit consideration. The correspondents are concerned that exposure may not have occurred to mothers during their pregnancies or even occurred at all. While there is always some uncertainty when surrogate variables are used, Drs. Scialli and Gibb are incorrect about the degree of uncertainty regarding emissions. The uncertainty factor is limited to 1 year, not 4 years, as they imply. For each baby, the distance from the mother's residence to an emitting site was calculated only for those sites with emissions reported within the year of the fifth week of gestation of that individual baby. Unfortunately, data available through regulatory agencies are annualized reports; therefore, more specific time assessments are not possible. In addition, such data are self-reported estimates. Thus, the quantitative discriminating ability of the amounts reported is not certain. Of interest, the majority of Milwaukee pregnant women have blood and/or urine evidence of TCE exposure (Ma, Garland, Herzog, McCarver, unpublished results). Thus, the question is not whether women are exposed to TCE at all during pregnancy, but rather what level of exposure constitutes risk. Identifying the source of exposure is a secondary question.
We also noted in the discussion that weather could be a complicating factor; however, a full discussion of weather would have been highly speculative. Data for changes in wind velocity, humidity, and temperature at each residence and emitting site were not available; thus, it could not be modeled. Prior to data analysis, the authors surveyed wind information available from the National Oceanic and Atmospheric Administration, but the data were too detailed and did not lend themselves to data reduction. Wind speed data are generated at multiple observational points on an hourly basis, and the wind in the Milwaukee area is highly variable. Thus, true adjustment for this confounder is not practical because the exact date and time of emission is not known, as