Weight and height were ascertained at W3, and BMI was calculated as: [weight (pounds) / height (inches)2] × 703, rounded to the nearest tenth. The CDC’s BMI guidelines for ages ≥ 15 years were used to exclude persons with biologically implausible values (Centers for Disease Control and Prevention, 2011a). Overweight was defined as a BMI between 25 and 29.9, and obese was a BMI of 30 or greater. find protocol We also considered respondent history of hypertension, which was queried at

all three waves, and history of high cholesterol, assessed at W3 only. To define 9/11-related exposure, we used a 12-item index, based on a tool created by Adams et al. (2006) and later modified based on Registry data by Brackbill et al. (2013). This scale included information on an enrollee’s exposures on 9/11 and during the subsequent recovery and cleanup effort, loss of loved Selleckchem PI3K inhibitor ones or coworkers, job loss due to 9/11, and damage to or loss of property or a home. The number of disaster-related events or conditions experienced was summed, and enrollees were categorized as having had none/low (0–1 experiences), medium (2–3), high (4–5), or very high (6 or more) exposure. Of 71,434 Registry enrollees, we included participants who completed the W3 follow-up

survey (n = 43,134). We excluded enrollees who were < 18 years of age at 9/11 (n = 739), enrollees who reported having been diagnosed with diabetes before Registry enrollment (i.e., prevalent cases; n = 2479), and those missing a history Farnesyltransferase of diabetes (n = 456). After removing those who were missing demographic or exposure data, 36,899 participants were included in this analysis. The frequencies of sociodemographic and 9/11-exposure characteristics of persons with diabetes were compared with those of persons without diabetes in bivariate analyses. We also compared characteristics of the study population to W1-only participants who

were not included in this analysis to assess possible bias from loss to follow-up. Logistic regression was used to calculate odds ratios (ORs) and 95% confidence intervals (CI) for the association between PTSD at W1 and new-onset diabetes. Multiple logistic regression models were adjusted for covariates that were significant in the bivariate analysis and that are commonly associated with diabetes, including age, sex, race/ethnicity, educational status at W1, hypertension, high cholesterol, and BMI at W3. Models that included smoking status at W1 and eligibility group were evaluated, but as the adjusted ORs (AORs) did not change substantially, these variables were not included in the final model. The 9/11 exposure index was no longer significant in the multivariable model and thus was not included in the final model. We tested for interactions between PTSD and other variables and found none. Model fit was assessed with the Hosmer–Lemeshow goodness of fit χ2 test. Analyses used SAS version 9.2 (SAS Institute Inc.