RESUMEN
Tanner stage indices have been frequently used in research studies to characterize the level of sexual maturation among adolescents. The Cooperative Study of Sickle Cell Disease (CSSCD) is a large multi-center study in which the assessment of Tanner stage indices has been performed to describe delays in sexual maturation and correlates of delays. Rigorous training, however, in rating Tanner stages was not provided for the clinicians in the study. Data from the CSSCD were analyzed to characterize the reliability of the assessment of sexual maturation in this study, and demonstrate the potential for increasing the reliability by more careful training. In lieu of a gold standard, a maximum likelihood algorithm was employed to examine the longitudinal consistency of Tanner stage assessments. Based on the assumption that the underlying process of sexual maturation is progressive, models were developed that allowed one to estimate jointly the incidence distribution for the attainment of Tanner stages and the error rates of assessment. Estimated false-positive rates for assessing the attainment of Tanner stages from the CSSCD ranged from 2% to 19%. Estimated false-negative rates ranged from 1% to 18%. The occurrence of these errors was found not to vary markedly with the subject's age. Error rates of this magnitude can introduce considerable bias to the estimation of the distribution of Tanner stages and result in serious losses in relative efficiency. While the validity of Tanner stage indices can be demonstrated from study data, the loss in statistical efficiency and power associated with the study error rates was severe. This underscores the utility of careful training in pubertal rating for research studies.