A generalized estimating equation (GEE) approach is used to estimate the sensitivity and specificity of five non-culture screening tests for detecting Chlamydia trachomatis in endocervical specimens from women attending family planning clinics. Since the estimates of these parameters can be very close to the upper limit of one, confidence interval construction based on the traditional approaches, such as the delta method or a back-transformation approach, may not have the stated coverage. We compare different approaches to calculating confidence intervals for the sensitivity and specificity from the results of the GEE approach. These methods include the delta method, a back-transformation approach, and a bootstrap approach. In addition, we use simulations to investigate the estimated coverage rates of the delta method and the back-transformation approach. Published in 2001 by John Wiley & Sons, Ltd.