All samples

were also tested for specific IgE to common aeroallergens (house dust mite, cat, dog, grass, or birch pollen) (Doekes et al. 1996). Analytical results were dichotomized and IgE (work-related or common allergens) was considered elevated if above 0.35 kU/L. Subjects were classified atopic if they had elevated IgE in response to at least one of the common aeroallergens. click here symptoms Respiratory symptoms PKC inhibitor and skin symptoms were reported on a self-completed questionnaire derived from the International Union Against Tuberculosis and Lung Disease (IUATLD) and the Medical Research Council—European Community of Coal and Steel (MRC-ECCS) for the bakery workers, and from the British Medical Research Council (BMRC) respiratory questionnaire for auto body shop workers (Burney et al. 1989; van der Lende and Orie 1972; Medical Research Council on the Aetiology of Chronic Bronchitis 1960). Information on cough, phlegm, wheeze, chest tightness, shortness of breath, and self-reported asthma was included. A variable describing asthma-like symptoms (wheezing, chest tightness, current/previous asthma) was constructed using the individual symptom

responses. Skin itch and dry skin were reported on the questionnaire; a dichotomous Ruxolitinib variable describing the presence of either itchy or dry skin was constructed. Work-related symptoms were explicit items on the questionnaire. Subjects were asked directly whether they have itchy skin at work and whether they experience asthma-like symptoms at work. No work-related symptom variables were constructed post hoc. Additional O-methylated flavonoid variables Age, sex, smoking (current and historical) as well as years working were self-reported on the questionnaire. Analyses Iterative non-parametric regression models (smoothing splines) with generalized additive models (PROC GAM) were first used to explore the shape of the exposure–response relationships for skin outcomes at the

population level. These models were used to explore unadjusted non-linear relationships between estimated exposure and symptoms outcomes. Generalized cross-validation (GCV) was used to select the smoothing parameter degrees of freedom (df); the df selected were limited to four to avoid large fluctuations that are likely not biologically relevant (Hastie 1990). Generalized linear models (SAS PROC GENMOD) with a log function were used to estimate unadjusted and adjusted prevalence ratios (PR) for the associations between exposure, atopy, specific sensitization, and symptoms. Adjusted models included atopy, work-related specific IgE sensitization, age, and sex; respiratory symptom models were additionally adjusted for smoking status. Sensitivity analyses were completed to explore whether atopy and specific sensitization were modifying the exposure–response relationships. Exposure–response relationships were investigated in models where atopic and specific sensitized subjects were excluded.