Limitations of Levothyroxine Bioequivalence Evaluation: Analysis of An Attempted Study

The relative bioavailability and therapeutic comparability of four levothyroxine products (two proprietary and two nonbranded) were determined in an open-label, block-randomized, four-way crossover trial with no washout periods in 24 presumed hypothyroid ambulatory care patients who were considered euthyroid while on 100 &mgr;g or 150 &mgr;g of oral levothyroxine daily for at least 3 months. Patients randomly received each of the four levothyroxine products for 6 weeks at the same dosage as their prestudy regimen. Area under the serum concentration versus time curve (AUC), maximum change in serum concentration (C(max)), and time to peak serum concentration (T(max)) were calculated for total thyroxine (TT(4)), total triiodothyronine (TT(3)), free thyroxine index (FT(4)I), and thyrotropin (TSH) for each product, with and without baseline correction and normalization for tablet potency differences. Ninety-percent confidence intervals were determined for each paired treatment comparison. Using the response variable of values uncorrected for baseline (Hour 0), 90% confidence intervals for paired treatment comparisons of nonlogarithmically transformed AUC and C(max) (TT(4), TT(3), and FT(4)I) were within the range of 80% to 120%. However, using the response variable of baseline corrected values, no confidence intervals for TT(4) or TT(3) AUC or TT(3) C(max) were within the range of 80% to 120%. Confidence intervals for TSH AUC and C(max) based on both baseline corrected and uncorrected values were outside the 80--120% range. Results were similar when AUC and C(max) were normalized for tablet potency. Thus, results both within and between response variables were conflicting. Further, results normalized for potency using either response variable also failed to demonstrate bioequivalence, pointing to the serious difficulties inherent in levothyroxine bioequivalence studies. Independent consultant review of study design, execution, and data analysis was used to identify critical issues in the conduct of future levothyroxine bioequivalence studies. Numerous flaws were identified, including many that could significantly impact the interpretation of results. These flaws included evaluation of a heterogeneous population of hypothyroid patients with variable residual thyroid function, failure to achieve steady-state serum thyroid hormone levels, use of levothyroxine preparations with variable and unmatched potency, calculations based on a response variable that requires linear pharmacokinetics, and application of bioequivalence criteria when required assumptions have not been met. By the criteria applied in this study and using a response variable based on uncorrected serum concentrations, the levothyroxine products studied, although appearing to be bioequivalent, were, in fact, therapeutically inequivalent. Yet, using an equally accepted and perhaps more valid response variable of baseline corrected values, this study failed to demonstrate bioequivalence between any two products. Finally, differences in tablet potency must be considered as contributing to the appearance of bioequivalence for the TT(4) comparisons. For these reasons, this study, as well as others based on similar models, cannot support a recommendation to substitute levothyroxine products without careful retesting and retitration. The basis for this position is that interchange of therapeutically inequivalent products may result in over- or under-replacement of thyroid hormone, with the known potential adverse consequences. From an economic perspective, product interchange may lead to additional medical costs that outweigh the initial savings obtained by the product switch.