Are Bioequivalence Studies of Levothyroxine Sodium Formulations in Euthyroid Volunteers Reliable?

Transcription

1 THYROID Volume 14, Number 3, 2004 Mary Ann Liebert, Inc. Are Bioequivalence Studies of Levothyroxine Sodium Formulations in Euthyroid Volunteers Reliable? Vicky Blakesley, 1 Walid Awni, 1 Charles Locke, 1 Thomas Ludden, 2 G. Richard Granneman, 1 and Lewis E. Braverman 3 Levothyroxine (LT 4 ) has a narrow therapeutic index. Consequently, precise standards for assessing the bioequivalence of different LT 4 products are vital. We examined the methodology that the Food and Drug Administration (FDA) recommends for comparing the bioavailability of LT 4 products, as well as three modifications to correct for endogenous, thyroxine (T 4 ) levels, to determine if the methodology could distinguish LT 4 products that differ by 12.5%, 25%, or 33%. With no baseline correction for the endogenous T 4 pool, differences in administered LT 4 doses that differed by 25% 33% could not be detected (450 mg and 400 mg doses versus 600 mg dose, respectively). The three mathematical correction methods could distinguish the doses that differed by 25% and 33%. None of the correction methods could distinguish dosage strengths that differed by 12.5% (450 mg versus 400 mg). Dose differences within this range are known to result in clinically relevant differences in safety and effectiveness. Methods of analysis of bioequivalence data that do not consider endogenous T 4 concentrations confound accurate quantitation and interpretation of LT 4 bioavailability. As a result, products inappropriately deemed bioequivalent may put patients at risk for iatrogenic hyperthyroidism or hypothyroidism. More precise methods for defining bioequivalence are required in order to ensure that LT 4 products accepted as bioequivalent will perform equivalently in patients without the need for further monitoring and retitration of their dose. Introduction HYROXINE (T T 4 ), AN ENDOGENOUS HORMONE secreted from the thyroid gland, is subject to complex biologic regulation. T 4 administered therapeutically to patients with hypothyroidism and various forms of thyroid neoplasia has two characteristics that make determination of the bioequivalence among T 4 products challenging. First, the synthetic compound levothyroxine sodium is biochemically and physiologically indistinguishable from endogenously produced T 4, the production of which is tightly controlled by the hypothalamic-pituitary-thyroid axis (1). Second, LT 4 has a narrow therapeutic index with the potential for putting patients at risk for iatrogenic hyperthyroidism or hypothyroidism at doses only 25% less or greater than optimal, based on patients serum thyrotropin (TSH) concentration (2). As a result, safe and effective use of LT 4 requires careful titration and close clinical follow-up. Data from large public health investigations (3,4) and studies of general clinical practices (5) suggest that 15% 29% of patients currently receive inadequate doses of LT 4 and 18% 24% receive excessive doses based on having serum TSH levels outside the reference range. It is vital that different marketed LT 4 formulations be bioequivalent. To be bioequivalent, two products must be pharmaceutically equivalent and have equivalent therapeutic effects. To be pharmaceutically equivalent, products must have the same active ingredient, be of equal strength, and have undergone similar, high-quality manufacturing processes. The most recent clinical practice guidelines from the American Association of Clinical Endocrinologists (AACE) (6) and the American Thyroid Association (7) recognize that the various brands of LT 4 have not been proven bioequivalent and recommend that the patient s brand not be changed during therapy. The clinical issue of therapeutic equivalence of LT 4 preparations is directly related to the method of determining their bioequivalence. Bioequivalence of two drug products is generally defined as the relative bioavailabilities of the active ingredient of the two products. This, in turn, is related to the rate and extent of absorption of the active ingredient. Pharmacokinetic studies designed to assess bioequivalence for 1 Abbott Laboratories, Global Pharmaceutical Research and Development, Abbott Park, Illinois. 2 GloboMax LLC, Pharmacometric Research and Development, Washington, DC. 3 Section of Endocrinology, Diabetes and Nutrition, Boston University School of Medicine, Boston, Massachusetts. 191

2 192 the majority of drugs are performed in healthy subjects and are typically not complicated, but this is not true for LT 4. This is because T 4 an endogenously produced hormone secreted continuously by the thyroid gland is indistinguishable from exogenously administered LT 4, both in its biochemical characteristics and physiologic effects. For pharmacokinetic studies designed to measure the bioavailability of LT 4 formulations, the Food and Drug Administration (FDA) (8) recommends that a single dose be administered to healthy subjects at a strength several times the normal therapeutic dose. The objective is to raise serum concentrations of the hormone sufficiently above endogenous baseline levels to achieve meaningful pharmacokinetic measurements. Even so, bioavailability studies in healthy subjects have indicated that when LT 4 doses as high as 600 mg are used, endogenous T 4 levels continue to contribute significantly to the total area under the curve (AUC). Concern over the confounding effect of this baseline contribution of endogenous T 4 led us to question the validity of the recommended methodology for assessing bioequivalence of LT 4 preparations. In an attempt to achieve an improved protocol, we designed a study that examined three different methods of mathematical correction that are intended to account for the contributions of endogenous T 4, each of which is based upon known physiologic principles of thyroid hormone economy. Methods Subjects Thirty-six volunteers (18 women and 18 men) between 19 and 50 years of age, inclusive, were enrolled. Subjects were clinically and biochemically euthyroid and in general good health based on the results of medical history, physical examination, 12-lead electrocardiogram, and routine laboratory tests. None of the women were pregnant or at risk for pregnancy (i.e., all were postmenopausal, sterile, or practicing an acceptable method of birth control other than contraceptive drugs); none were taking contraceptive drugs or breastfeeding. The study was conducted after Institutional Review Board approval in accordance with the principles of the Declaration of Helsinki. After the nature and procedures of the study had been explained to each participant, signed informed consent was obtained. The study was conducted at the Abbott Clinical Pharmacology Research Unit, Waukegan, Illinois. Healthy volunteers were recruited through media advertising and direct phone contacts using the Local Institutional Review Board-approved processes. Study design This single-dose, open-label pharmacokinetic study of LT 4 in healthy volunteers had a three-period, randomized crossover design. The administered dose of LT 4 (Synthroid, Abbott Laboratories, Abbott Park, IL) in the three study arms was 600 mg, 450 mg, and 400 mg, respectively. On the morning of study day 1 of each period (8:30 AM), after fasting for 10 hours, subjects received appropriate sets of 50-mg LT 4 tablets (12, 9, and 8 tablets, respectively; all from the same lot). A washout interval of at least 44 days separated each of the three study periods. Subjects were confined to the study site for 2 days before dosing and remained there until the 96-hour blood sample after dosing had been collected (study day 5). No food or beverages other than water to quench thirst were allowed between 10:00 PM on study day 22 and collection of the 4- hour blood sample at 12:30 PM the next day (study day 21) and between 10:00 PM on study day 21 and collection of the 4-hour blood sample at 12:30 PM after dosing the next day (study day 1). The meal content was identical on study day 21 and study day 1 in each period. Except for these two fasting periods, subjects received a standardized diet; no caffeine-containing beverages were permitted. No strenuous activity was allowed during confinement. Blood samples sufficient to yield approximately 2 ml of serum were collected by venipuncture into 5-mL evacuated siliconized collection tubes for assay of total T 4, total triiodothyronine (T 3 ), and TSH. On study day 21 of each period, samples were drawn at 0 hours (approximately 8:30 AM) and at 0.5, 1, 1.5, 2, 2.5, 3, 4, 6, 8, 10, 12 and 18 hours. On study day 1, three samples were drawn before dosing, at 230 minutes, 215 minutes, and immediately before dosing (time zero). Thereafter, samples were collected at 0.5, 1, 1.5, 2, 2.5, 3, 4, 6, 8, 10, 12, 18, 24, 36, 48, 72, and 96 hours after dosing. Serum concentrations of T 4 and T 3 were determined by use of validated radioimmunoassay methods (Coat-A- Count, Diagnostics Product Corporation, Los Angeles, CA). The serum TSH concentrations were determined by a validated two-site immunoradiometric assay method (TSH IRMA, Diagnostics Product Corporation). The assays were specifically validated for this study over the following concentration ranges: mg/dl for T 4, ng/dl for T 3, and miu/l for TSH. The in-study calibration for T 4 contained six standards ranging from approximately mg/dl. All calibration curves had correlation coefficient (r) values greater than or equal to The average back-calculated calibration standards had coefficient of variations (CV) that ranged from 1.27% 2.63%, with percent differences from theoretical ranging from The interassay CV of the quality controls (theoretical concentrations of 2.00, 6.00, and 18.0 mg/dl) from the analytical runs ranged from 7.49% 13.3%, with percent differences from theoretical ranging from Analytes from each subject for all dosing regimens were measured in duplicate in the same analytical assay. Pharmacokinetics and statistical methods BLAKESLEY ET AL. The pharmacokinetic parameters for T 4 were determined using noncompartmental methods; these included maximum serum concentration (C max ), time to C max (T max ), and area under the serum concentration-time curve (AUC) from 0 48 hours (AUC 48 ), 0 72 hours (AUC 72 ), and 0 96 hours (AUC 96 ). For T 4, the values for these parameters were determined both with and without correction for endogenous T 4 levels. Correction for endogenous T 4 levels was done using each of the three methods described below. The corrected 0-hour concentrations were set at zero. Correction Method 1. For each subject and period the mean of the three T 4 values at 20.5, 20.25, and 0 hours before dosing was subtracted from each concentration after dosing. Correction Method 2. For each subject and period, each T 4 concentration after dosing was corrected for the hypo-

3 LEVOTHYROXINE BIOEQUIVALENCE STUDIES 193 thetical decay of endogenous T 4 with a 7-day half-life, beginning with the level obtained immediately before dosing (with the 0-hour concentration assumed to be the average of the three concentrations at 20.5, and 0 hours before dosing). Correction Method 3. For each subject and period, each T 4 concentration after dosing was adjusted by subtracting the T 4 concentration measured at the analogous time-point on the day prior to administration of the LT 4 dose (study day 21) of each period. For uncorrected and corrected T 4, an analysis of variance (ANOVA) was performed for T max and the natural logarithms of C max, AUC 48, AUC 72, and AUC 96, with fixed effects for gender, sequence, gender-by-sequence interaction, period, regimen, the interaction of gender with each of period and regimen, and random effects for subjects nested within gender-by-sequence combination. A significance level of 0.05 was used for all tests. For uncorrected T 4 concentration and for each version of corrected T 4 concentration, bioequivalence assessments were carried out. For each pair of doses (i.e., 450 mg versus 600 mg, 400 mg versus 600 mg, and 450 mg versus 400 mg), a bioequivalence assessment was performed by the two onesided tests procedure (9) via 90% confidence intervals for the ratio of central values obtained within the framework of the ANOVA for the natural logarithms of AUC 48 and C max. The endpoints of the 90% confidence interval were obtained by back-transforming the endpoints of the confidence interval for the difference of means for the logarithmically transformed variable. The FDA and other regulatory agencies require that bioequivalence assessment be based on an analysis of the logarithm of AUC and the logarithm of C max. This is because it has been found that for virtually all drugs the logarithmic transformations of AUC and C max have more nearly symmetric probability distributions than the untransformed variables. Because pharmacokineticists think in terms of ratios (relative bioavailability) or percentage differences, many think that the logarithmic transformation should be used for this reason as a difference in logarithms corresponds directly to a ratio. Note that the individual concentrations are not transformed. The variables AUC and C max are transformed. Bioequivalence was concluded if the two 90% confidence intervals from the analyses of the natural logarithms of AUC 48 and C max were both within the range. This is the criterion used by the FDA and other regulatory agencies for determining whether two products can be declared bioequivalent. There is a statistical basis for this. It can be shown that if two products are not bioequivalent (i.e., the true ratio of central values is less than 0.80 or greater than 1.25 for either AUC or C max ), then the probability that the criterion will be met is at most There is at most a 5% chance that two inequivalent products will be incorrectly declared bioequivalent. The same analyses were done using each of AUC 72 and AUC 96 in place of AUC 48. A repeated measures analysis was performed on the T 4 data of day 21 (the 24 hours before dosing) of each period to address the question of diurnal variation. For this analysis, an eight-component vector of observations for a subject was defined by: the average of the measurements at 0, 0.5, 1, 1.5, and 2 hours; the average of the measurements at 2.5, 3, and 4 hours; the individual measurements at 6, 8, 10, 12, and 18 hours; and the average of the three measurements obtained in the half hour prior to the LT 4 dose on the morning of day 1. The eight components of the vector were logarithmically transformed for the analysis. In addition to having an effect for time as defined by the components of this vector, the model had effects for sequence, gender, sequence*gender interaction, and gender*time interaction. No conditions were imposed on the covariance matrix of the vector except that it was assumed to be same for all the combinations of sequence and gender. An analysis was performed to address the question of whether there were unequal carryover effects for T 4. For this analysis the sequence factor was removed from the ANOVA model described above, and terms for first-order and second-order carryover effects were added. For TSH, an ANOVA was performed on the logarithm of the reduction in 24-hour AUC from before dosing to the first 24 hours after dosing (equivalently, the logarithm of the AUC for the difference in TSH concentration, with the study day 1 concentration subtracted from the study day 21 concentration). The effects in the model were the same as those for the analysis of postdose T 4 concentration variables with no terms for carryover effects. Within the framework of this model, the dose level effects were compared pair wise. Results The mean age of the 18 men and 18 women in the study was 32.9 years, and all were of normal weight for height. The data for three subjects were excluded from pharmacokinetic analyses because they received the study drug in only one period. In the 33 remaining subjects with pharmacokinetic data (16 men, 17 women), the mean age was 33.1 years. Two of the 33 subjects did not receive study drug in period 3; one because of a positive pregnancy test and the other because of a positive drug screen. These two subjects were to have received 600 mg in period 3. FIG. 1. Mean thyroxine (T 4 ) concentration-time profiles on study day 1 after single-dose administration of levothyroxine sodium uncorrected for endogenous T 4 baseline concentrations.

4 194 BLAKESLEY ET AL. Uncorrected serum T 4 data TABLE 1. MEAN STANDARD DEVIATIO N PHARMACOKINETIC PARAMETERS OF LEVOTHYROXINE WITHOUT CORRECTING FOR ENDOGENOUS THYROXINE BASELINE CONCENTRATIONS The mean T 4 serum concentration-time profiles after administration of each of the 400, 450, and 600 mg LT 4 doses are shown in Figure 1. Mean T 4 concentrations before dosing were approximately 7.5 mg/dl and reached a maximum of 13 to 14 mg/dl before declining. By 96 hours, mean T 4 concentrations decreased to approximately 9 mg/dl. There were no statistically significant differences between males and females for the analyses performed without correction for endogenous levels. The point estimate, a measurement of the ratio of the AUC 48 central value of the study drug to the AUC 48 central value of the reference drug, should be 1.00 if the two products have identical bioavailabilities. Theoretically the point estimates should be 0.75 and 0.67 for 450 mg and 400 mg, respectively, when compared to 600 mg. The point estimates for AUC 48 were and for the 450 mg and 400 mg comparisons to 600 mg, respectively. When no correction was Regimens Pharmacokinetic 600-mg dose 450-mg dose 400-mg dose parameters (units) (n 5 31) (n 5 33) (n 5 33) T max (h) C max (mg/dl) a a AUC 48 (mg? h/dl) a a AUC 72 (mg? h/dl) a a,b AUC 96 (mg? h/dl) a,b a Statistically significantly different from 600 mg (ANOVA, p, 0.05). b Statistically significantly different from 450 mg (ANOVA, p, 0.05). T max, time to maximum serum concentration; C max, maximum serum concentration; AUC 48, area under the serum concentration-time curve from 0 48 hours; AUC 72 area under the serum concentration-time curve from 0 72 hours; AUC 96, area under the serum concentration-time curve from 0 96 hours; ANOVA, analysis of variance. performed for endogenous T 4 levels, both the 450 mg and 400 mg doses were bioequivalent to the 600 mg dose because the 90% confidence intervals for evaluating bioequivalence were contained within the range (Table 1 and 2). The 450 mg and 400 mg doses were also bioequivalent to each other because the 90% confidence intervals fell within this range. In addition, the point estimate for the AUC 48 was 1.026, which differed from the predicted point estimate of (450 mg versus 400 mg). Thus, the use of baseline uncorrected T 4 C max, AUC 48, AUC 72, and AUC 96 values resulted in a finding of bioequivalence of LT 4 doses that were 25% and 33% lower than the reference dose (450 mg and 400 mg versus 600 mg, respectively). Corrected T 4 data The concentration-time profiles for each method are depicted in Figures 2, 3, and 4 for correction Methods 1, 2, and TABLE 2. BIOEQUIVALENCE AND RELATIVE BIOAVAILABILITY: UNCORRECTED LEVOTHYROXINE Relative bioavailability Central value a Regimens test vs. Pharmacokinetic 90% confidence reference parameter Test Reference Point estimate b interval 450 mg vs. 600 mg C max AUC AUC AUC mg vs. 600 mg C max AUC AUC AUC mg vs. 400 mg C max AUC AUC AUC a Antilogarithm of the least squares means for logarithms. b Antilogarithm of the difference (test minus reference) of the least squares means for logarithms. C max, maximum serum concentration; AUC 48, area under the serum concentration-time curve from 0 48 hours; AUC 72 area under the serum concentration-time curve from 0 72 hours; AUC 96, area under the serum concentration-time curve from 0 96 hours.

5 LEVOTHYROXINE BIOEQUIVALENCE STUDIES 195 FIG. 2. Mean thyroxine (T 4 ) concentration-time profiles after correction for endogenous baseline levels of T 4 using correction method 1. 3, respectively. The pharmacokinetic parameters of T 4 after correction for endogenous T 4 baseline concentrations are shown in Table 3. The results for bioequivalence data assessment are shown in Tables 4, 5, and 6. Method 1 (average of the three predose measurements subtracted from each postdose T 4 measurement). Throughout the 96-hour sampling period after dosing, the mean serum T 4 concentrations after correction for endogenous baseline levels were higher after administration of the 600 mg dose than after the 450 mg and 400 mg doses (Fig. 2). By 96 hours after dosing, the mean baseline corrected T 4 concentration remained at approximately 1 to 2 mg/dl. Neither of the lower doses was bioequivalent to the 600 mg dose because the 90% confidence intervals were not within the range for C max, AUC 48, AUC 72, and AUC 96 (Table 4). However, the mean corrected T 4 concentrations for the 450 mg and 400 mg doses were comparable throughout the 96-hour sampling period after dosing; the two doses were FIG. 3. Mean thyroxine (T 4 ) concentration-time profiles after correction for endogenous baseline levels of T 4 using correction method 2. FIG. 4. Mean thyroxine (T 4 ) concentration-time profiles after correction for endogenous baseline levels of T 4 using correction method 3. bioequivalent to each other since their 90% confidence intervals ( and for C max and AUC 48, respectively) were within the specified range of Thus, the use of this correction method allowed distinction of doses that differed by 25% and 33% from the 600 mg dose, but 450 mg was bioequivalent to 400 mg. Method 2 (T 4 concentrations corrected assuming a 7-day half-life). Throughout the 96-hour sampling period after dosing, the mean serum T 4 concentrations after correction for endogenous baseline levels were higher after administration of the 600 mg dose than after the 450 mg and 400 mg doses (Fig. 3). By 96 hours after dosing, the mean baseline corrected T 4 concentration remained at approximately 3 4 mg/dl. Neither of the lower doses was bioequivalent to the 600 mg dose because the 90% confidence intervals were not within the range for C max, AUC 48, AUC 72, and AUC 96 (Table 5). However, the mean corrected T 4 concentrations for the 450 mg and 400 mg doses were comparable throughout the 96-hour sampling period after dosing; the two doses were bioequivalent to each other because their 90% confidence intervals ( and for C max and AUC 48, respectively) were within the specified range of This correction method also distinguished the 450 and 400 mg doses from the 600 mg reference dose; however 450 mg was bioequivalent to 400 mg. Method 3 (correction by subtraction of measurement from analogous time-point on day 21). Throughout the 96-hour sampling period after dosing, the mean serum T 4 concentrations after correction were higher after administration of the 600 mg dose than after the 450 mg and 400 mg doses (Fig. 4). By 96 hours after dosing, the mean baseline corrected T 4 concentration remained at approximately 1 2 mg/dl. Neither of the lower doses was bioequivalent to the 600 mg dose because the 90% confidence intervals were not within the range for C max, AUC 48, AUC 72, and AUC 96 (Table 6). However, the mean corrected T 4 concentrations for the 450 mg and 400 mg doses were comparable throughout the 96-hour sampling period after dosing; with the exception of

6 196 BLAKESLEY ET AL. TABLE 3. MEAN STANDARD DEVIATIO N PHARMACOKINETIC PARAMETERS OF LEVOTHYROXINE AFTER CORRECTING FOR ENDOGENOUS THYROXINE BASELINE CONCENTRATIONS the use of AUC 72, the two doses were bioequivalent to each other because their 90% confidence intervals ( and for C max and AUC 48, respectively) were within the specified range of As with the other two Regimens Pharmacokinetic 600-mg dose 450-mg dose 400-mg dose parameters (units) (n 5 31) (n 5 33) (n 5 33) Correction method 1 T max (h) C max (mg/dl) a a AUC 48 (mg? h/dl) a a AUC 72 (mg? h/dl) a a AUC 96 (mg? h/dl) a a Correction method 2 T max (h) C max (mg/dl) a a AUC 48 (mg? h/dl) a a AUC 72 (mg? h/dl) a a AUC 96 (mg? h/dl) a a Correction method 3 T max (h) C max (mg/dl) a a AUC 48 (mg? h/dl) a a AUC 72 (mg? h/dl) a a,b AUC 96 (mg? h/dl) a a a Statistically significantly different from 600 mg (ANOVA, p, 0.05). b Statistically significantly different from 450 mg (ANOVA, p, 0.05). T max, time to maximum serum concentration; C max, maximum serum concentration; AUC 48, area under the serum concentration-time curve from 0 48 hours; AUC 72 area under the serum concentration-time curve from 0 72 hours; AUC 96, area under the serum concentration-time curve from 0 96 hours; ANOVA, analysis of variance. correction methods, this method could not distinguish the two doses that differed by 12.5% (450 mg versus 400 mg) but when doses differed by 25% and 33%, they would not be declared bioequivalent. TABLE 4. BIOEQUIVALENCE AND RELATIVE BIOAVAILABILITY FOR LEVOTHYROXINE (CORRECTION METHOD 1) Relative bioavailability Central value a Regimens test vs. Pharmacokinetic 90% confidence reference parameter Test Reference Point estimate b interval 450 mg vs. 600 mg C max AUC AUC AUC mg vs. 600 mg C max AUC AUC AUC mg vs. 400 mg C max AUC AUC AUC a Antilogarithm of the least squares means for logarithms. b Antilogarithm of the difference (test minus reference) of the least squares means for logarithms. C max, maximum serum concentration; AUC 48, area under the serum concentration-time curve from 0 48 hours; AUC 72 area under the serum concentration-time curve from 0 72 hours; AUC 96, area under the serum concentration-time curve from 0 96 hours.

7 LEVOTHYROXINE BIOEQUIVALENCE STUDIES 197 Relative bioavailability Central value a Regimens test vs. Pharmacokinetic 90% confidence reference parameter Test Reference Point estimate b interval Baseline T 4 before dosing (study day 21) The mean serum concentration-time plots for baseline T 4 on study day 21 of each period prior to dosing demonstrated diurnal variation throughout a 24-hour day with lowest levels at 2:00 AM (Fig. 5). Changes during the course of the 24 hours preceding dosing were statistically significant for each period (p, 0.001). TSH TABLE 5. BIOEQUIVALENCE AND RELATIVE BIOAVAILABILITY FOR LEVOTHYROXINE (CORRECTION METHOD 2) 450 mg vs. 600 mg C max AUC AUC AUC mg vs. 600 mg C max AUC AUC AUC mg vs. 400 mg C max AUC AUC AUC a Antilogarithm of the least squares means for logarithms. b Antilogarithm of the difference (test minus reference) of the least squares means for logarithms. C max, maximum serum concentration; AUC 48, area under the serum concentration-time curve from 0 48 hours; AUC 72 area under the serum concentration-time curve from 0 72 hours; AUC 96, area under the serum concentration-time curve from 0 96 hours. The mean serum concentration-time plots for TSH before and after dosing showed clear diurnal variation prior to LT 4 administration (Fig. 6). During the 24-hour predose period, the TSH concentrations were highest after midnight and then declined during the morning hours, reaching their lowest levels at noon and then gradually returning to maximum values over the next 14 hours (2:00 AM on the morning of study day 1, before dosing). Administration of LT 4 at any of the three doses significantly, but incompletely, suppressed TSH serum concentrations throughout the 24-hour period after dosing on study day 1. When peak nocturnal values for TSH were compared in the predose and postdose period, the postdose value was lower in amplitude but occurred at approximately the same clock time. Concentrations remained low throughout the 4- day postdose sampling period, failing to return to baseline values even after 96 hours. The rank order of suppression of the serum TSH concentrations was consistent with the rank TABLE 6. BIOEQUIVALENCE AND RELATIVE BIOAVAILABILITY FOR LEVOTHYROXINE (CORRECTION METHOD 3) Relative bioavailability Central value a Regimens test vs. Pharmacokinetic 90% confidence reference parameter Test Reference Point estimate b interval 450 mg vs. 600 mg C max AUC AUC AUC mg vs. 600 mg C max AUC AUC AUC mg vs. 400 mg C max AUC AUC AUC a Antilogarithm of the least squares means for logarithms. b Antilogarithm of the difference (test minus reference) of the least squares means for logarithms. C max, maximum serum concentration; AUC 48, area under the serum concentration-time curve from 0 48 hours; AUC 72 area under the serum concentration-time curve from 0 72 hours; AUC 96, area under the serum concentration-time curve from 0 96 hours.

8 198 FIG. 5. Mean thyroxine (T 4 ) concentration-time profiles on study day 21 prior to dosing with levothyroxine sodium by period. order of the size of the LT 4 dose, with the greatest TSH suppression occurring in association with administration of the largest dose (i.e., 600 mg). A statistical analysis was performed on the logarithm of the reduction in 24-hour AUC from the 24 hours prior to dosing to the first 24 hours after dosing. The difference between the 600 mg dose and each of the smaller doses was significant (p, and p ), while the p value for the comparison of the 450 mg and 400 mg doses was Carryover effects of T 4 In addition to the above mentioned dose-dependent changes in TSH, the data gave evidence of altered homeostasis as a result of the supraphysiologic exposures from the large LT 4 doses of the study. Statistical analysis revealed FIG. 6. Mean thyrotropin (TSH) concentration-time profiles for the 24 hours prior to (study day 21) and for the 96 hours after administration of levothyroxine sodium on study day 1. highly significant period effects for uncorrected C max and AUC (p, 0.001) with the central values decreasing with time. Further analysis revealed differential biologic carryover effects by the three dose levels across periods (p # 0.004), with the carryover effect increasing with dose (e.g., the central value for a 450 mg dose in period 2 is larger if a 600 mg dose was administered in Period 1 than if a 400 mg dose was administered in period 1). Likewise, for all three methods of correction there were statistically significant period effects for both C max and AUC 48, with the p value ranging from for the six analyses. The central values decreased with time for corrected AUC 48, and for corrected C max the period 3 central value was always lowest. Further analysis revealed statistically significant unequal carryover effects for C max (p value ranging from for the several analyses), but not for AUC 48. The nonsignificant statistics for AUC 48 could be partly the result of the limited power of the test for unequal carryover effects. As was the case for uncorrected C max, so also for corrected C max the 600 mg dose had the largest carryover effect. These results indicate there were carryover effects even with washouts of 44 and 53 days between doses and that inequivalent products have unequal carryover effects. In the standard statistical analysis for a bioequivalence study with a crossover design, unequal carryover effects result in biased estimates and make the products appear to be more similar in performance than they really are. The unequal carryover effects on corrected C max are one reason the point estimates for the ratio of both the 450 and 400 mg central values to the 600 mg central value are larger than would be expected from the ratio of doses (Tables 4 6). T 3 concentrations The mean T 3 concentrations for the 24-hour period prior to dosing and throughout the 96-hour postdose period were in the narrow range of ng/dl for all three LT 4 dose regimens (data not shown). Little diurnal variation in T 3 concentration was observed. Discussion BLAKESLEY ET AL. The therapeutic equivalence of LT 4 preparations (i.e., the equivalence of their clinical effects) is directly related to their bioequivalence (relative bioavailability). Bioavailability, in turn, is related to the rate and extent of absorption and metabolism of the hormone. While most pharmacokinetic studies of bioequivalence in normal subjects are straightforward, this is not true of LT 4 studies, because the hormone is secreted continuously by the thyroid gland and the endogenous hormone is indistinguishable from exogenously administered LT 4 in its biochemical characteristics and physiologic effects. Because T 4 is naturally present in the blood at levels ranging from 5 12 mg/dl, the FDA (8) recommends that a single dose of LT 4 be administered to healthy subjects at a strength several times the normal therapeutic dose (e.g., 600 mg), with the goal of reducing the influence of endogenous hormone on pharmacokinetic findings. Results from several bioavailability studies and a stochastic simulation study of LT 4 products have suggested that, given certain reasonable assumptions about endogenous T 4 behavior in healthy subjects, the use of baseline uncorrected C max and AUC 48 values can result in two products being declared bioequivalent when they actually dif-

9 LEVOTHYROXINE BIOEQUIVALENCE STUDIES 199 fer by as much as 35% (results of the simulation study were submitted to the FDA). In this study, the application of criteria for bioequivalence without accounting for endogenous T 4 levels resulted in failure to identify differences between LT 4 products varying by as much as 25% 33% in dosage strength. When we applied mathematical corrections to compensate for the contribution of endogenous T 4 and for hypothalamic-pituitary-thyroid suppression of LT 4 production, we reduced the chances of two LT 4 products being declared bioequivalent when they differed in dosage strength by 25%. However, the use of correction methods did not eliminate the chance that two products differing by 12.5% would be declared bioequivalent. We compared three methods of correction. The first method of correction was based on the assumption that endogenous T 4 levels remain constant after a single dose of LT 4, although this is not actually the case. TSH levels are definitely lowered after the exogenous LT 4 doses. The suppression of TSH is expected to result in reduced thyroidal T 4 secretion thereby lowering the endogenous contribution to the AUC. Moreover, the method relies on data from only three samples obtained during a 30-minute interval just prior to dosing, excluding the influence of the diurnal variation in hormone levels. On the basis of randomness alone, the impact of an unreliable mean value derived from only three measurements could be significant. Thus, the corrected AUC value obtained by this method may be imprecise. The second correction method assumed complete suppression of T 4 production via central feedback suppression of TSH, resulting in declining levels of T 4 at an average rate calculated on the basis of the 7-day half-life of the compound. This method, too, has drawbacks. First, a reasonable correction is possible only if endogenous T 4 production is completely and abruptly inhibited after administration of the study drug and does not resume over the sampling period. Even if this unlikely assumption were valid, the results would be in error, with the size of the error depending on how much the elimination half-life differed from 7 days in a given subject. Because TSH levels are not completely suppressed by exogenous LT 4, it seems unlikely that endogenous T 4 production would be reduced to zero, even with an accompanying 7-day half-life. Moreover, the use of a single value for half-life disregards possible individual variations such as thyroid status, gender, race, age, and iodine intake. The currently recommended study design for healthy volunteers does not allow estimation of T 4 half-life in individual subjects. Finally, this correction method has the same sampling disadvantage as Method 1 in its reliance on the average of three serum concentrations measured over a 30- minute interval immediately before dosing. The third method of correction involved the adjustment of T 4 concentration at each time of postdose sampling based on the corresponding predose concentration 24 hours earlier, thus offering some advantages over the first two methods. Baseline data obtained from numerous samples collected over the 24-hour period prior to dosing make it possible to adjust data obtained at the same clock time after dosing in each period, thus taking into account diurnal variations in hormone concentrations. This would appear to improve the reliability of correction. Method 3 does not assume that endogenous T 4 production stops abruptly after LT 4 administration nor that its half-life is constant across subjects. Similar to Method 1, it assumes that endogenous production of T 4 is not suppressed after administration of a single dose of LT 4 in healthy volunteers. It also assumes that the circadian pattern of endogenous T 4 production is unaltered by a single large dose of exogenous LT 4. Because LT 4 has a narrow therapeutic range, its safe and effective use requires careful titration and close clinical follow-up. If therapy is not carefully monitored, the patient is frequently at risk for iatrogenic hyperthyroidism or hypothyroidism (3,4). In a recent epidemiologic health-fair survey of thyroid function (n 5 25,862), nearly 40% of approximately 1500 individuals who reported taking thyroid medication had serum TSH levels outside the therapeutic range (3) and could be classified as having subclinical thyroid dysfunction. Even in the presence of mild ( subclinical ) hypothyroidism, there is a potential for cardiovascular problems such as hypercholesterolemia (10), increased fibrinolytic activity (11), systolic and diastolic dysfunction (12,13), atherosclerosis, and myocardial infarction (14,15). On the other hand, mild thyrotoxicosis increases the likelihood of arrhythmias, ventricular dysfunction, reduced exercise performance, and possible cardiovascular death (5,15,16). In the elderly, iatrogenic hyperthyroidism has been associated with the development of osteoporosis (17) and approximately 600,000 elderly patients may be at risk for bone demineralization because of LT 4 overdose (18). The sensitivity and precision of bioequivalence assessments for LT 4 products need to be improved so that the designation of bioequivalence ensures therapeutic equivalence. One solution to this problem might be the adjustment of statistical criteria to a narrower range than the currently recommended range of of the reference product. However, even if the target was narrowed to and correction Method 1 was applied, the 450 mg dose would have passed bioequivalence versus the 400 mg dose although these doses actually differ by 12.5%. In the present study, the application of either correction methods 2 or 3 and a narrowing of the statistical criteria range to would allow differentiation of the two doses that differed by 12.5% (450 mg versus 400 mg), based on the 90% confidence intervals for AUC 48. While the use of a narrower target range may be useful in appropriate settings, our data demonstrated that the method must be well validated before it is adopted as part of the bioequivalence criteria. Not only does sensitivity need to be optimized so that clinically relevant differences in doses can be distinguished, but the method and criteria must have adequate specificity to ensure that truly bio-equivalent products are declared bioequivalent by the methodology. A potentially superior model for future bioequivalence trials could be based on pharmacokinetic studies in athyreotic subjects (19). This would essentially eliminate the confounding effects of endogenous T 4. Presumably such studies would involve multiple doses of LT 4 administered over a sufficient period to account for carryover effects. The absence of endogenous T 4 production would remove the need for a method of baseline correction. Last, the pharmacodynamic marker TSH could be used as an additional measure of bioequivalence because TSH is considered the gold standard for monitoring LT 4 therapy. Accordingly, should claims of bioequivalence be allowed when differential TSH perturbations suggest otherwise? Granted, TSH is probably too sensitive to allow application of the con-

10 200 ventional regulatory range of , but should more liberal yet still discriminating criteria be considered? To date no studies have been done to test the precision and sensitivity of TSH measurements in euthyroid volunteers. Thus, it is unlikely that subtle changes in serum TSH concentrations in euthyroid volunteers after LT 4 administration would be sufficiently accurate to determine bioequivalence. Ideally future studies would incorporate appropriate study populations, controls, and study design. It is the recommendation of these authors that any proposed criteria be tested and validated prior to adoption. In conclusion, this study reveals important flaws in the current design and analysis of single-dose crossover studies currently used to assess the bioequivalence of LT 4 products. The use of criteria that do not adjust for endogenous T 4 levels in healthy study subjects may result in two products being declared bioequivalent when they differ in dosage strength by as much as 25% 33%, a situation posing substantial risk to patients treated with LT 4, particularly in cases where patients are switched from one brand to another at the same labeled dosage without notification of both the patient and the physician. Achieving optimal safety and efficacy in the use of drugs with a narrow therapeutic index, such as LT 4, places special responsibilities on physicians, patients, dispensing pharmacists, pharmaceutical manufacturers, and regulatory authorities. Clearly, improved methods for determining bioequivalence are needed as a step in this direction. Acknowledgment This study was supported by Abbott Laboratories, Abbott Park, Illinois. References 1. Scanlon MF, Toft AD 2000 Regulation of thyrotropin secretion. In: Braverman LE, Utiger RD (eds) Werner & Ingbar s The Thyroid: A Fundamental and Clinical Text, 8th ed. Lippincott Williams and Wilkins, Philadelphia, pp Carr D, McLeod DT, Parry G, Thornes HM 1988 Fine adjustment of thyroxine replacement dosage: Comparison of the thyrotropin releasing hormone test using a sensitive thyrotropin assay with measurement of free thyroid hormones and clinical assessment. Clin Endocrinol (Oxf) 28: Canaris GJ, Manowitz NR, Mayor GM, Ridgway EC 2000 The Colorado thyroid disease prevalence study. Arch Intern Med 160: Hollowell JG, Staehling NW, Flanders WD, Hannon WH, Gunter EW, Spencer CA, Braverman LE 2002 Serum TSH, T 4, and thyroid antibodies in the United States Population (1988 to 1994): National Health and Nutritional Examination Survey (NHANES III). J Clin Endocrinol Metab 87: Parle J, Franklyn J, Cross K, Jones C 1991 Prevalence and follow-up of abnormal thyrotropin concentrations in the elderly in the United Kingdom. Clin Endocrinol (Oxf) 34: AACE Task Force 2002 American Association of Clinical Endocrinologists medical guidelines for clinical practice for the evaluation and treatment of hyperthyroidism and hypothyroidism. Endocr Pract 8: BLAKESLEY ET AL. 7. Singer PA, Cooper DS, Levy EG, Landenson PW, Braverman LE, Daniels G, Greenspan FS, McDougall IR, Nikolai TF 1995 Treatment guidelines for patients with hyperthyroidism and hypothyroidism: Standards of Care Committee, American Thyroid Association. JAMA 273: U.S. Department of Health and Human Services, Food and Drug Administration, Center for Drug Evaluation and Research 2000 Guidance for Industry: Levothyroxine Sodium Tablets In Vivo Pharmacokinetic and Bioavailability Studies and In Vitro Dissolution Testing. 9. Schuirman DJ 1987 A comparison of the two one-sided tests procedure and the power approach for assessing the equivalence of average bioavailability. J Pharmacokinetics Biopharm 15: Danese MD, Ladenson PW, Meinert CL, Powe NR 2000 Effect of thyroxine therapy on serum lipoproteins in patients with mild thyroid failure: A quanitative review of the literature. J Clin Endocrinol Metab 85: Chadarevian R, Bruckert E, Leenhardt L, Giral P, Ankri A, Turpin G 2001 Components of the fibrinolytic system are differently altered in moderate and severe hypothyroidism. J Clin Endocrinol Metab 86: Monzani F, Di Bello V, Caraccio N, Bertini A, Giorgio D, Giusti C, Ferrannini E 2001 Effect of levothyroxine on cardiac function and structure in subclinical hypothyroidism: A double blind, placebo-controlled study. J Clin Endocrinol Metab 86: Brenta G, Mutti LA, Schnitman M, Fretes O, Perrone A, Matute ML 2003 Assessment of left ventricular diastolic function by radionuclide ventriculography at rest and exercise in subclinical hypothyroidism, and its response to L- thyroxine therapy. Am J Cardiol 91: Perk M, O Neill BJ 1997 The effect of thyroid hormone therapy on angiographic coronary artery disase progression. Can J Cardiol 13: Biondi B, Palmieri EA, Lombardi G, Fazio S 2002 Effects of subclinical thyroid dysfunction on the heart. Ann Intern Med 137: Sawin CT, Geller A, Wolf PA, Belanger AJ, Baker E, Bacharach P, Wilson PW, Benjamin EJ, D Agostino RB 1994 Low serum thyrotropin concentrations as a risk factor for atrial fibrillation in older persons. N Engl J Med 331: Toft AD 2001 Clinical practice. Subclinical hyperthyroidism. N Engl J Med 345: Cooper DS, Ridgway EC 2002 Thoughts on prevention of thyroid disease in the United States. Thyroid 12: Wartofsky L 2002 Levothyroxine: Therapeutic use and regulatory issues related to bioequivalence. Expert Opin Pharmacother 3: Address reprint requests to: Vicky Blakesley, M.D., Ph.D. Synthroid Global Project Team Global Pharmaceutical Research and Development D-R4DM AP Abbott Park Road Abbott Park, IL vicky.blakesley@abbott.com