PURPOSE: The aim of this study was to explore relationships between the Children's Depression Rating Scale-Revised (CDRS-R) and Hamilton Rating Scale for Depression (HAM-D) in adolescent bipolar disorder. We hypothesized that CDRS-R and HAM-D scores would be significantly correlated and both scales would be sensitive to change resulting from treatment. METHODS: Data from a randomized, double-blind, placebo-controlled study of quetiapine with divalproex versus divalproex alone (n = 30) and an open-label study of lithium (n = 27) were used. Assessments using the CDRS-R and HAM-D were completed separately without any dependence on ratings from the alternate rating scale or specific order. All raters for symptom rating scales were blinded to the study for which they were completing a patient assessment and the intent to conduct these post hoc analyses and were different from those who performed diagnostic interviews. Relationships between these measures were assessed using the Pearson correlation coefficient, and receiver operator characteristic curves evaluated the ability of CDRS-R and HAM-D scores to predict the level of global improvement. RESULTS: The mean (+/- standard deviation) age of subjects was 15.0 +/- 1.6 years. Baseline CDRS-R and HAM-D scores were moderately correlated (r = 0.63; p < 0.001), while week 6 scores showed stronger correlation (r = 0.88; p < 0.001). Both measures showed significant main effects for time on treatment (CDRS-R: F(3.4,191.2) = 39.4, p < 0.001; HAM-D: F(3.9,217.9) = 38.5, p < 0.001). A CDRS-R score of 30.5 and HAM-D score of 7.5 represent the highest sensitivity and specificity in classifying responders and nonresponders. A 42.2% and 60.6% reduction in baseline CDRS-R and HAM-D scores, respectively, provided the highest sensitivity and specificity. CONCLUSIONS: Our preliminary results demonstrated that CDRS-R and HAM-D scores were significantly correlated, suggesting that either measure may be used to assess depressive symptoms in studies of adolescent bipolar disorder. Limitations to this study included a relatively small pooled sample, variation in clinician judgment, and study design not specific for the comparison of rating scales. Studies with methods specific to evaluate and compare these rating scales and larger patient samples from multiple treatment sites are needed to confirm these findings.