Confidence Interval Criteria for Assessment of Dose Proportionality

• Published in: Pharmaceutical research Vol. 17; no. 10; pp. 1278 - 1283
• Main Authors: Smith, Brian P., Vandenhende, Francois R., DeSante, Karl A., Farid, Nagy A., Welch, Pamela A., Callaghan, John T., Forgue, S. Thomas
• Format: Journal Article
• Language: English
• Published: New York, NY Springer 01.10.2000; Springer Nature B.V
• Subjects: Analysis of Variance; Bioequivalence; Biological and medical sciences; Confidence Intervals; Dose-Response Relationship, Drug; Drug dosages; Expected values; General pharmacology; Hypotheses; Hypothesis testing; Linear Models; Mathematical Computing; Medical sciences; Models, Biological; Pharmacokinetics; Pharmacokinetics. Pharmacogenetics. Drug-receptor interactions; Pharmacology. Drug treatments; Therapeutic Equivalency; Variance analysis; Increasing dose; Absorption; Bioequivalence; Estimation; Elimination; Mathematical model; Pharmacokinetics; Dose; Confidence interval
• ISSN: 0724-8741; 1573-904X
• DOI: 10.1023/A:1026451721686

The aim of this work was a pragmatic, statistically sound and clinically relevant approach to dose-proportionality analyses that is compatible with common study designs. Statistical estimation is used to derive a (1-alpha)% confidence interval (CI) for the ratio of dose-normalized, geometric mean values (Rdnm) of a pharmacokinetic variable (PK). An acceptance interval for Rdnm defining the clinically relevant, dose-proportional region is established a priori. Proportionality is declared if the CI for Rdnm is completely contained within the critical region. The approach is illustrated with mixed-effects models based on a power function of the form PK = beta0 x Dose(beta1); however, the logic holds for other functional forms. It was observed that the dose-proportional region delineated by a power model depends only on the dose ratio. Furthermore, a dose ratio (rho1) can be calculated such that the CI lies entirely within the pre-specified critical region. A larger ratio (rho2) may exist such that the CI lies completely outside that region. The approach supports inferences about the PK response that are not constrained to the exact dose levels studied. The proposed method enhances the information from a clinical dose-proportionality study and helps to standardize decision rules.