| Abstract
| - Serum sex hormones may be related to the risk of several diseases in postmeno pausal women including osteoporosis, heart disease, and breast and endometrial cancer. For assessment of the relation of sex hormones to disease, the measurements should be reliable, valid, and practical. In this paper, the authors evaluated the short-term (4-week) and long-term (2-year) reliability of serum sex hormones and interrelations among serum sex hormones in white postmenopausal women recruited in Pittsburgh, Pennsylvania, 1981-1986. For comparison, the authors simultaneously evaluated the short- and long-term reliability of other commonly measured risk factors, i.e., lipids, lipoproteins, and blood pressure. Serum concentrations of estrone, estradiol, testosterone, and androstenedione were measured by extraction, column chromatography, and radioimmurloassay. Reliability was estimated by calculating the intraclass correlation coefficients (Ř) and their 95% confidence interval. About 50% of the estradiol levels were below the sensitivity of the assay and, therefore, these results should be inter preted with some caution. The intraclass correlation coefficient for testosterone was 0.92 (95% confidence interval 1.0-0.82), suggesting that a single measure may be reliable in characterizing women for epidemiologic research. Over 4 weeks, estrone could be measured more reliably (Ř = 0.72) than over 2 years (Ř = 0.56), but the variability over the long term was similar to that observed for other biologic variables, suggesting that, in situations where the relation between estrone and disease is fairly substantial, a single measure may be used. For estradiol and androstenedione, the intraclass correlations were small, indicating poor reproducibility and the need for more measurements. Estrone concentrations were 11 pg/ml or 46% higher in women with measurable estradiol. Estrone was also positively related to androstenedione concen trations (r = 0.33, p< 0.001). Concentrations of estradiol are extremely low in postmenopausal women, and accordingly, there is a greater possibility of laboratory error. Since the data suggest that estrone levels can be more reliably measured and are, in fact, related to estradiol levels, it is possible that estrone levels may be used to indicate the total estrogen status of postmenopausal women. Am J Epidemiol 1991 ;133:50-7.
|
