Logistic Regression Analysis of the Association Between VOC Exposure Metrics and SBS Symptoms
Multivariate logistic regressions were used to assess the relationships between VOC exposure metrics and SBS symptoms. The final models were adjusted for other risk or protective factors and contained terms for demographic variables, activities that might affect VOC exposure, building ventilation type, temperature, humidity, and individual sensitivity. Each symptom was tested with and without the exposure metric and the results were compared to determine if the exposure metric provided improved prediction of a symptom. This odds ratio (OR) was used as the measure of effect. Most of the metrics including TVOC, VOCi and the irritancy-weighted sum were not effective in predicting SBS irritancy symptoms. The "carpet/building materials" and the "cleaning products & water-based paint" emission vectors were found to be the most influential of the four principal components of the irritancy/PC metric. The compounds with the highest loadings in these two vectors were styrene, 2-butoxyethanol and isopropyl alcohol. Thus, the irritancy/PC metric was defined as the sum of these two components. This redefined metric was effective and statistically significant in the prediction of several SBS symptoms including eye, irritant, sleepiness, and irritated mucous membrane symptoms. The success of this metric suggests that it variously accounted for the irritant potencies of individual VOCs, the highly correlated nature of indoor VOC mixtures and the probable presence of potent, but unmeasured VOCs that are associated with the detected compounds.

Further Application of the Irritancy/Principal Components VOC Metric
The VOC irritant/PC metric that was successfully used in the CHBS to explore the relationship between VOC exposures and SBS symptoms among office workers was subsequently applied to a subset of buildings from the U.S. EPA BASE study (Apte and Daisey, 1999). Data collected in 1995 and 1996 from 28 buildings in 11 states were utilized. Thirteen VOCs collected in canisters and analyzed by GC/MS plus carbon monoxide were included. Principal component (PC) analysis was used to derive an exposure metric as previously described by Ten Brinke et al. Four emission vectors were defined. The sources were attributed to photocopiers, motor vehicles outdoors, environmental tobacco smoke (ETS) and water-based paint. The "photocopier" vector with high loadings of styrene, 2-butanone, ethyl acetate and aromatic hydrocarbons was found to be significantly associated with mucous membrane irritant symptoms. The "paint" vector with high loadings of methyl isobutyl ketone and butyl acetate was found to be significantly associated with sore throat. Significant "protective" associations with a number of symptoms were observed for the ETS vector. A probable cause for this observation is uncertain.