A Comparison of Methods for Sex Steroid Hormone Adjustment, National Health and Nutrition Examination Survey (NHANES III)

Introduction: Differences in testosterone (T), 17-â estradiol (E) and sex hormone binding globulin (SHBG) have been hypothesized to explain racial disparities in chronic disease rates, yet epidemiologic results have been mixed. Since T and E are transported by SHBG, it is considered essential to estimate free T and E by adjusting total T and E levels by SHBG. Therefore, we compared standardization (E/SHBG or T/SHBG) and different statistical adjustment approaches of T and E for SHBG. Methods: We investigated data of 1,506 men, >=17 years from the Third National Health and Nutrition Examination Survey (NHANES III). To examine linearity between sex hormones and SHBG we used plots and regression models. We included age, body mass index (BMI), E, and race in models to determine if the statistical importance of these variables changed by the method of SHBG adjustment. Results: We found that T and E were moderate and weakly correlated with SHBG, respectively (r-squared, T 0.25, E 0.04). Based on model residual plots and r-squared, the categorical model performed better than the linear models. To address non-linearity by using SHBG in quintiles adjusting T, the associations of T with specific levels of age, race, BMI, and E changed by more than 10% compared to both linear models. Conclusion: Choosing an appropriate adjustment for model variables is important to prevent bias in analyses, and confusion across studies. Linearity between variables should not be assumed automatically when adjusting sex steroid hormones by SHBG, however, this needs to be determined in each study.