Background: Assessment of dosage form performance in delivering endogenous compounds, such as hormones, in vivo requires a specific approach.
Objectives: Assessment of relative bioavailability of levothyroxine sodium (L-T4) from eight solid preparations, compared with a liquid formulation, by using pharmacological doses, and critical evaluation of trial methodology based on the pooled analysis of individual data.
Design: Eight open-label, randomised, single-dose, crossover phase I studies using eight solid L-T4 dosage forms (25, 50, 75, 100, 125, 150, 175, 200 microg per tablet; administered total doses 600, 625 or 700 microg) and a liquid formulation; assessment of relative bioavailability by 90% confidence intervals for the relative area under the concentration-time curve (AUC) of total thyroxine (TT4), i.e. protein-bound plus free thyroxine, calculated by using the recommended log AUC four-way analysis of variance models for crossover designs. For the pooled analysis, general linear models were applied to assess the validity of model assumptions, to identify potential sources of effect modification, to discuss alternative modelling approaches with respect to endogenous hormone secretion and to give recommendations for future designs and sample sizes.
Participants: One hundred and sixty-nine healthy males; 29 of these individuals participating in two studies.
Interventions: Single oral doses of L-T4 tablets and the liquid formulation administered after fasting, separated by at least 6 weeks; a total of 396 drug exposures.
Main outcome measures: TT4 AUC from 0 to 48 hours and peak plasma concentration with and without baseline correction.
Results: Each study demonstrated equivalence of the tablets to the drinking solution, independent of the chosen analysis model. Sequence effects that could devalidate the chosen crossover approach were not found. Period effects with changing directions that could best be explained by seasonal variation were detected. While the pre-specified method of baseline correction of simply subtracting individual time-zero TT4 values was disadvantageous, the analysis of total AUC could be improved considerably by covariate adjustment for baseline TT4. With this approach, sample sizes could have been substantially reduced or, alternatively, the recommended equivalence ranges could be reduced to +/-6%.
Conclusion: Using a single pharmacological dose of L-T4 in two-period crossover designs is a safe and reliable procedure to assess L-T4 dosage form performance. With an adequate statistical modelling approach, the design is efficient and allows general conclusions with moderate sample sizes.