European Journal of Pharmaceutical Sciences 66 (2015) 70–77
Contents lists available at ScienceDirect
European Journal of Pharmaceutical Sciences journal homepage: www.elsevier.com/locate/ejps
The impact of new partial AUC parameters for evaluating the bioequivalence of prolonged-release formulations Michaël Boily a,b, Catherine Dussault a, Julie Massicotte a, Pascal Guibord c, Marc Lefebvre a,⇑ a
Scientific and Regulatory Affairs, Algorithme Pharma Inc., 575 Armand-Frappier Blvd., Laval, QC H7V 4B3, Canada Département de Pharmacologie, Faculté de Médecine, Université de Montréal, 2900 Édouard-Montpetit Blvd., Montreal, QC, Canada c Biometrics, Algorithme Pharma Inc., 575 Armand-Frappier Blvd., Laval, QC H7V 4B3, Canada b
a r t i c l e
i n f o
Article history: Received 17 April 2014 Received in revised form 15 August 2014 Accepted 2 October 2014 Available online 13 October 2014 Keywords: Modified release formulations Bioequivalence Pharmacokinetics parameters Partial AUC Drug accumulation
a b s t r a c t To demonstrate bioequivalence (BE) between two prolonged-release (PR) drug formulations, single dose studies under fasting and fed state as well as at least one steady-state study are currently required by the European Medicines Agency (EMA). Recently, however, there have been debates regarding the relevance of steady-state studies. New requirements in single-dose investigations have also been suggested by the EMA to address the absence of a parameter that can adequately assess the equivalence of the shape of the curves. In the draft guideline issued in 2013, new partial area under the curve (pAUC) pharmacokinetic (PK) parameters were introduced to that effect. In light of these potential changes, there is a need of supportive clinical evidence to evaluate the impact of pAUCs on the evaluation of BE between PR formulations. In this retrospective analysis, it was investigated whether the newly defined parameters were associated with an increase in discriminatory ability or a change in variability compared to the conventional PK parameters. Among the single dose studies that met the requirements already in place, 20% were found unable to meet the EMA’s new requirements in regards to the pAUC PK parameters. When pairing fasting and fed studies for a same formulation, the failure rate increased to 40%. In some cases, due to the high variability of these parameters, an increase of the sample size would be required to prove BE. In other cases however, the pAUC parameters demonstrated a robust ability to detect differences between the shapes of the curves of PR formulations. The present analysis should help to better understand the impact of the upcoming changes in European regulations on PR formulations and in the design of future BE studies. Ó 2014 Elsevier B.V. All rights reserved.
1. Introduction In early 2013, the European Medicines Agency (EMA) issued a draft bioequivalence (BE) guideline for modified release (MR) drug products, providing recommendations for delayed-release and prolonged-release (PR) formulations (EMA, 2013). This guideline, when finalized, will replace its more than a decade old precursor. One of the major changes introduced in this document is a requirement to demonstrate the equivalence of the shape of the systemic concentration vs time curves between compared PR products. There are cases where a PR Test formulation, for which the 90% confidence intervals (CI) for the overall area under the curve (AUC) and the maximum plasma concentration (Cmax) fall within ⇑ Corresponding authors at: Scientific and Regulatory Affairs, Algorithme Pharma Inc., 575 Armand-Frappier Blvd., Laval, QC H7V 4B3, Canada. Tel.: +1 450 973 6077; fax: +1 450 973 2801. E-mail address: [email protected] (M. Lefebvre). http://dx.doi.org/10.1016/j.ejps.2014.10.001 0928-0987/Ó 2014 Elsevier B.V. All rights reserved.
the bioequivalence limits (80–125%) of the Reference PR product, presents a major difference in the early (Anschutz et al., 2010; Garcia-Arieta, 2013) or in the terminal segment of its curve (Garcia-Arieta, 2013). Detecting such patterns, which conventional parameters can fail to reveal, appears to be the rationale behind the introduction of new partial AUC (pAUC) parameters in the guideline. A pAUC is an AUC determined between given time points (Endrenyi and Tothfalusi, 2010). A parameter measuring the early segment of the curve, calculated as the AUC truncated at the tmax of the marketed reference product calculated for each study subject (AUCReftmax), was the first pAUC to be adopted by a regulatory agency (Health Canada, 1992). It is still currently used and recommended by Health Canada (Health Canada, 2012) when the early exposure of an immediate release drug product is important. The US Food and Drug Administration (FDA) introduced in 2003 the use of a partial AUC as an early exposure measure, defined as the
M. Boily et al. / European Journal of Pharmaceutical Sciences 66 (2015) 70–77
AUC truncated at the population median of tmax values for the reference formulation (FDA, 2003). The FDA is proposing to redefine this pAUC in their new draft guidance, where the time to truncate the partial area should be related to a clinically relevant pharmacodynamic measure (FDA, 2013). In line with the new draft guidance, the parameter AUC8–48 is now a requirement for mesalamine delayed-release tablets, a drug with a local action in the colon (FDA, 2012a). The FDA also discussed (FDA, 2010) and introduced product-specific pAUC parameters for at least two different multiphasic dosage forms (FDA, 2011, 2012b), and the EMA recommended a similar approach for the demonstration of BE for biphasic MR formulations (EMA, 2014). In its current guidance on MR formulations, effective since January 2000, the EMA requires BE for all PR formulations to be assessed in both single and multiple dose studies, using conventional parameters (AUC(0–t), AUC(0–1), Cmax; and AUCs, Cmax, Cmin, respectively) (EMEA, 1999). The new EMA draft guideline specifies that for PR formulations for which accumulation is likely after a single dose, BE demonstration in single dose studies will be based on the conventional parameters as well as an additional parameter identified as a pAUC (see Fig. 1). A steady-state study will also still be required for those formulations. In contrast, for PR drug products not likely to accumulate, a waiver for the conduct of a steady-state study may be obtained if acceptance limits are met for two newly added pAUC parameters (early and terminal) in the single dose fasting and fed state studies. Interestingly, the suggested truncation point is half of the dosage interval (s/2) (EMA, 2013), resulting in new parameters that have not been evaluated extensively, unlike other pAUCs such as AUCReftmax (Chen et al., 2011; Lionberger, 2010). The introduction of this decision tree and new pharmacokinetic (PK) parameters raises important questions on the outcome of future BE studies, as the new PK parameters may impact key study design factors that influence the cost of conducting the studies (e.g. sample size, number of studies, number of periods/sequence, sampling schedule, etc). As an example, the cost of the approval process would decrease significantly in the case where steady-state studies would no longer be required. To assess the impact of this draft guideline, a review of an internal database was conducted in order to recalculate different PK parameters including early and terminal pAUCs as well as the
71
conventional parameters from 53 PR single dose studies that met standard BE requirements. From this retrospective evaluation, the outcome of the new BE criteria was assessed. Furthermore, a secondary objective of this analysis was to characterize the variability of the new parameters. 2. Materials and methods 2.1. Inclusion of bioequivalence studies The database of Algorithme Pharma studies conducted between year 2007 and 2013 was screened for studies that met all of the following criteria: – Pivotal bioequivalence studies with a two-period, twosequence and two-treatment cross-over design; – Oral administration to healthy volunteers in a single dose study design; – Prolonged-release Test and Reference formulations (this criterion included extended-release, sustained-release and controlled-release formulations; and excluded delayedrelease formulations and multiphasic MR products for which different regulatory requirements are presented in the EMA draft guidance (EMA, 2013)); – 90% CI that falls within 80–125% for all the conventional PK parameters (as defined in the following section: 2.2 pharmacokinetics parameters). 2.2. Pharmacokinetic parameters The calculation of the different PK parameters was performed using Kinetic software (version 9.01; Algorithme Pharma Inc., Laval, Canada). AUCs were calculated using the linear trapezoidal method. The different PK parameters were classified as follows: (1) conventional; (2) early pAUCs; (3) terminal pAUCs; and (4) other parameters. The parameters in the conventional category are those commonly used for assessing BE in accordance with the current EMA guideline (CPMP/EWP/280/96) and are Cmax, AUC(0–t), and AUC(0–1). The early and terminal pAUCs include the new partial AUC parameters mentioned in the EMA draft guideline (CPMP/
Fig. 1. Interpretation of the EMA’s draft guidance’s new requirements concerning parameters to be used and studies to be conducted depending on the expected extent of accumulation of Test and Reference formulations. From our understanding of the EMA’s draft guidance, to demonstrate BE of two PR formulations, a fed and a fasting state study at single dose are minimally required. If the Test and the Reference formulations present a (AUC(0–s)/AUC(0–1)) ratio 90 %, they are considered likely to accumulate. If this ratio is > 90%, they are considered as unlikely to accumulate. The PK requirements corresponding to each possible case are shown (dashed line rectangle). The number of studies (n) among the 53 studies that met the inclusion criteria that corresponds to each case is also presented, whether or not the associated requirements were met.
72
M. Boily et al. / European Journal of Pharmaceutical Sciences 66 (2015) 70–77
EWP/280/96 Rev.1). Those two pAUCs (early and terminal) are separated by a time point which is usually the half of the dosage interval (s) of the investigated formulation (s/2) (EMA, 2013). Therefore, the early pAUC described in the draft guideline refers to AUC(0–s/2), being the AUC from the administration time (0) to the mid-interval time (s/2). For comparative and informational purposes, AUCReftmax was also included in the early pAUC category. The latter is a parameter recommended by Health Canada as a measure of early exposure, which is of interest when the time of onset of effect or the rate of absorption is important (Health Canada, 2012). AUCReftmax is defined as the area under the curve from time zero to the tmax of the Reference product calculated for each study subject. The terminal pAUC is described as the area under the curve from half of s (s/2) to an end point which is not defined in the EMA draft guideline. It was thus hypothesized that this end point could be either s, t (time of last observed quantifiable plasma concentration) or 1, resulting in 3 possible terminal pAUCs: AUC(s/2–s), AUC(s/2–t) and AUC(s/2–1). Finally, AUC(0–s) was also evaluated and added in the last category (other). To determine whether drugs are likely to accumulate, the AUC(0–s)/AUC(0–1) ratio (%) was also calculated for the Test and Reference product of each study. 2.3. Statistical analysis The natural logarithmic (ln) transformation of the parameters was used for all statistical inference. Parameters were statistically analyzed using an Analysis of Variance (ANOVA) model. The fixed factors included in this model were the subject effect (nested within sequence), the treatment received, the period at which it was given, as well as the sequence in which each treatment was received. The 90% CI for the exponential of the difference in LSmeans between the Test and Reference product was calculated for the ln-transformed parameters (Test to Reference ratio of geometric LSmeans). If a PK parameter could not be determined for one period in a subject, the corresponding subject was excluded from that particular statistical comparison. BE statistical analyses were generated with SASÒ software (version 9.2; SAS Institute Inc., Cary, NC, USA) using the GLM procedure. A study was considered to pass on a specific parameter if its 90% CI fell within the 80.00–125.00% range. For practical purposes, ratio estimates are presented rounded to the nearest unit. Fisher’s exact test was employed to compare proportions. Differences in intra-subject variability (ISCV) between 2 parameters are reported as the mean ± 95% CI. Sample size estimations were performed using the equation described by Julious et al. with a power of 80%, a type 1 error of 5% and using BE criteria of 20% (Julious, 2004). 3. Results
drugs and the dosage interval of the formulations found in these studies. The studies included 21 different drugs over 13 different therapeutic categories. According to the 2013 draft guideline, a low extent of accumulation will be expected if the AUC(0–s)/AUC(0–1) ratio is above 90% in the single dose study, as illustrated in Fig. 1. Moreover, for these products with no risk of accumulation, it will be possible to obtain a waiver for the steady-state study. Among the 53 studies selected, 49 (92.45%) were considered likely to accumulate vs only 4 (7.55%) unlikely to accumulate for both the Test and Reference products. As the 90% value is still under discussion, the impact of decreasing the AUC(0–s)/AUC(0–1) ratio to 80% was also assessed. In an outdated 1996 guideline for MR formulations, Health Canada recommended to consider a ratio of 80% as an indicator to characterize the likelihood of accumulation (Health Canada, 1996). With this 80% ratio, a total of 10 studies (18.86%) – 6 more than for the 90% ratio – would be considered unlikely to lead to accumulation. For formulations meeting the conventional BE standards and that are considered likely to accumulate, failure to meet the acceptance criterion for the new pAUC parameter would correspond in a failure to demonstrate BE. However, the pAUC proposed in the draft guideline, for both the fasting and fed single dose studies, is not clearly defined. It was thus hypothesized that this parameter was either an early or terminal pAUC with a truncation time point at s/2. This is essentially the same as either the early or the terminal pAUC proposed in the draft guideline for drugs unlikely to accumulate. Table 2 presents the number of studies that fail to show statistical equivalence for each early or terminal pAUC PK parameter as well as for other parameters. If AUC(0–s/2) was to be considered as the required pAUC parameter, 10 studies out of 49 (20.41%) would fail to demonstrate BE. The same number of studies would also fail to demonstrate BE if any of the proposed 3 terminal pAUCs was selected as an alternative. Table 2 also displays that there was a significantly higher proportion of fed studies that failed to meet the acceptance criterion as compared to the studies completed in a fasting state for both of the early pAUC PK parameters (33.33% vs 68.42% for AUCReftmax, and 10.00 vs 36.84% for AUC(0–s/2)). Such a difference was absent for other parameters. In the simplest scenario, a new generic PR formulation would meet acceptance requirements of the EMA if BE criteria are met in one single dose study in the fasting state and in one single dose study in the fed state. Among the 53 individual studies which met conventional PK parameters requirements, 12 pairs consisting of one fasting study and one fed study conducted with the same batch of comparison products were identified (for a total of 24 studies). Of those pairs, 41.6% (5/12) contained at least one study that did not meet the new criteria; the fasting or the fed study failed in four pairs whereas both studies failed in only one pair.
3.1. Impact of PK parameters on BE 3.2. Variability of the new PK parameters Among the studies evaluated, 53 met the criteria defined above (Section 2.1). Table 1 comprises a description of the categories of Table 1 Description of formulations in studies included in the analysis. 5 Most common drug class Identified by therapeutic categorya Analgesics Antidepressants Urologicals Anti-diabetics Vasodilatators Others Total a
Number of different drugs = 21.
Dosage interval %
(n)
33.96 18.87 9.43 7.55 5.66 24.53
(18) (10) (5) (4) (3) (13)
100.00
(53)
% 12 h 24 h
(n)
37.74 62.26
(20) (33)
100.00
(53)
The variability of the new pAUC parameters was assessed in order to determine if it accounted for the increase in the number of studies failing to meet BE standards. Cmax was the conventional parameter with the highest ISCV in most studies. In each study, the difference in ISCV between each AUC parameters and Cmax was compared. Fig. 2 presents the average difference in ISCV calculated for the 53 studies that met the conventional BE standards. As expected, the conventional parameters AUC(0–t), AUC(0–1) and AUC(0–s) were less variable than Cmax. In contrast, the early pAUC parameter AUC(0–s/2) appeared to be slightly more variable, presenting an average difference in ISCV of 0.76[0.81; 2.34] percentage points with Cmax. All the terminal pAUCs presented a significantly larger difference in ISCV ranging from 3.11[0.73; 5.50] to 3.94[1.43; 6.45] percentage points. AUCReftmax was the parameter
73
M. Boily et al. / European Journal of Pharmaceutical Sciences 66 (2015) 70–77 Table 2 Percentage of studies which 90% CI of individual PK parameters fall outside the acceptance limits in drugs likely to accumulate.
a
Parameter
All studies (n = 49)
Fasting (n = 30)
Fed (n = 19)
% fail
(n)
% fail
(n)
% fail
(n)
Early
AUCReftmax AUC(0–s/2)
46.94 20.41
(23) (10)
33.33 10.00
(10) (3)
68.42a 36.84a
(13) (7)
Terminal
AUC(s/2–s) AUC(s/2–t) AUC(s/2–1)
20.41 20.41 20.41
(10) (10) (10)
20.00 23.33 23.33
(6) (7) (7)
21.05 15.79 15.79
(4) (3) (3)
Other
AUCs
4.08
(2)
3.33
(1)
5.26
(1)
Significatively different from fasting state (p < 0.05).
Early pAUC
AUC(0−τ 2) AUCReftmax AUC(τ AUC(τ 2−t)
AUC(τ
Terminal pAUC
2−∞) 2−τ)
AUC(0−∞)
Studies All studies
AUC(0−t)
Fasting Fed
AUC(0−τ) Cmax
−10
0
10
20
30
Difference in ISCV (%) with Cmax Fig. 2. Difference between the ISCV of investigated parameters and the corresponding ISCV of Cmax in fasting and fed state studies. For each evaluated parameter, the mean difference (with the 95% CI) in ISCV of that parameter and Cmax is plotted for all studies (red circles, n = 53), fasting state studies (green triangles, n = 31) and fed state studies (blue squares, n = 22).
showing the highest variability. Thus, all the pAUC parameters were more variable than the conventional AUC parameters. The effect of food on the ISCV was also verified. As shown in Fig. 2, a difference between fasting and fed studies was observed for only the early pAUC parameters. Both AUCReftmax and AUC(0–s/2) were significantly more variable in the fed state than under fasting conditions. Fig. 3 presents the difference in ISCV with Cmax depending on whether studies met or not the acceptance criterion for the two pAUC parameters AUC(0–s/2) and AUC(s/2–s). In studies which CI did not fall within acceptance limits for AUC(0–s/2), there was a tendency for the parameter AUC(0–s/2) to be more variable than Cmax. A similar tendency was present in studies that failed to meet requirements for AUC(s/2–s) as in that case AUC(s/2–s) appeared to be more variable. Data for AUC(s/2–t) and AUC(s/2–1) also presented this pattern (data not shown). This suggests an association between failure on a pAUC parameter and the variability of that same parameter. 3.3. Impact of the new parameters on the Test to Reference ratio In Fig. 4A, the 53 studies that met the conventional BE criteria are plotted according to their respective geometric mean for the
Test/Reference (T/R) ratio and ISCV for AUC(0–s/2). One study with a T/R ratio of 268% and ISCV of 33% is not displayed in Fig. 4A. For studies that failed on this parameter, the 90% CI is illustrated. Studies were classified into 3 groups (1, 2, 3) based on their results. Group 1 includes the studies with ratios outside the 80–125% range; group 2 includes studies with ratios inside the 80–125% range but outside the 90–111% range, and group 3 includes studies with ratios inside the 90–111% range. Two out of 10 studies (20%) had T/R ratios out of the 80–125% BE acceptance limits (group 1) and 6 studies (60%) had ratios inside the latter limits but outside the 90–111% (group 2). Three of these 6 studies, had a higher ISCV for AUC(0–s/2) than for Cmax; however, in all cases, ISCVs were under 23%. Finally, the remaining 2 studies (20%) had a ratio located between 90 and 111% and this is associated with a large positive difference in ISCV with Cmax (group 3). An example plot of Test and Reference concentration over time is also given (Fig. 4A, bottom of panel) for a study belonging to group 1 and a study belonging to group 3. Fig. 4B is similar to Fig. 4A but displays data for AUC(s/2–s). The ratio of 5 studies out of 12 (42%) did not fall within the 80–125% limits (group 1). Another 5 studies (42%) could be classified as group 2. All their ISCVs were under 28% and 4 of these 5 studies had a higher variability for AUC(0–s/2) than for Cmax.
74
M. Boily et al. / European Journal of Pharmaceutical Sciences 66 (2015) 70–77
AUCReftmax
Early pAUC
AUC(0−τ 2) 2−τ)
Terminal pAUC
AUC(τ
Studies AUC(0−τ)
All studies Fail on early pAUC Fail on terminal pAUC No fail on pAUCs
Cmax
−10
0
10
20
30
Difference in ISCV (%) with Cmax Fig. 3. Difference between the ISCV of investigated parameters and the corresponding ISCV of Cmax in studies failing on a pAUC parameter. For parameters of early and terminal pAUC and comparison parameters, the mean difference in ISCV (with the 95% CI) of that parameter and Cmax is plotted for all studies (red circles, n = 53), for all studies that failed on AUC(0-s/2), the early pAUC (green triangles, n = 10), for all studies that failed on AUC(s/2-s), the terminal pAUC parameter (blue squares, n = 12) or studies meeting requirements for both of the 2 latter pAUC parameters (orange diamond shape, n = 34).
(A) 60
(B) 60
50
50
issue Fail
40
ii
i
ISCV (%)
ISCV (%)
40 30
Pass
i
30
group 1 A
20
20
10
10
0
0
2 B 3 C
ii 60
80
100
120
140
160
180
40
AUC 0−τ 2 T/R ratio (%)
6 12 18 24 30 36 42 48 54 60 66 72
6 12 18 24 30 36 42 48 54 60 66 72
Time (h)
Cmax
120
140
160
180
2−τ T/R ratio (%)
AUCt
AUC(0−τ 2)
AUC(τ
2−τ)
6 4 2 0
60
Time (h)
0
frequency (n)
(C)
100
80 10 0 12 0 14 0 60 80 10 0 12 0 14 0 60
ii
0
80
AUCτ
Concentration
i
60
80 10 0 12 0 14 0 60 80 10 0 12 0 14 0
40
T/R ratio (%) Fig. 4. ISCV in function of Test to Reference ratio for (A) AUC(0-s /2) or (B) AUC(s/2-s). For every studies (n = 53), the ISCV of the corresponding parameter is plotted in function of the T/R ratio. Studies that failed on the parameter (A) AUC(0-s/2) or (B) AUC(s/2-s) are plotted (red dots) with their 90% CI. Studies that met requirements for these parameters are also displayed (green triangles). Group 1 to 3 are highlighted according to a gray scale. Group 1 corresponds to studies with ratios outside the 80-125% acceptance limits (dark gray), which is also delimitated by dashed lines. Group 2 corresponds to ratios between 80-90% or 111-125% (light gray). Group 3 corresponds to ratios within the 90-111% limits (white). Two examples of studies not meeting acceptance criterion for AUC(0-s/2) are also plotted (i, ii). A plot of their concentration versus time is displayed for both the Test formulation (black line) and the Reference formulation (light gray). The first example did not meet requirements for both early and terminal pAUC (i). The timepoint s/2 is represented by a dashed line. (C) Comparison of the distribution of T/R ratios for various parameters. The bandwidth of the histograms corresponds to 1% T/R ratio.
M. Boily et al. / European Journal of Pharmaceutical Sciences 66 (2015) 70–77
Finally, two studies (17%) were included in group 3 and had a high variability, which was also at least twice that of Cmax. Fig. 4C depict the frequency of the T/R ratios found for two conventional parameters as well as for early (AUC(0–s/2)) and terminal (AUC(s/2–s)) pAUC parameters. A higher number of AUC(s/2–s) T/R ratios was outside the 80–125% limit than for the other parameters. Also, AUC(0–s/2) T/R ratios were more sparsely distributed resulting in less values around 100%.
4. Discussion While the addition of new pAUC parameters to the acceptance criteria should improve the comparison of the shape of the curves between PR formulations, more studies are expected to fail to demonstrate BE. As suggested by the results presented in the current analysis, a large majority of PR formulations will be categorized as products that are likely to accumulate with an increased difficulty to meet the new BE criteria. For the other drug products, for which a low extent of accumulation is expected, the number of studies evaluated for these formulations was probably too small to bring adequate conclusion. Nevertheless, this indicates that very few studies will meet the required conditions to obtain a waiver for the steady-state study. Therefore, further research may be of interest to evaluate the feasibility and impact of broadening the waiver qualification criteria provided that the safety considerations are not jeopardized. On the other hand, preventing the conduct of a steady-state study for even a small proportion of PR formulations could reduce the number of human subjects exposed in BE studies as well as decrease the development costs. This could be significant considering that a steady-state study is recommended by the current EMA guidance in nearly all circumstances (EMEA, 1999). In order to properly evaluate the impact of the new pAUC parameters, it was first necessary to identify why studies failed to meet acceptance criteria. A study may fail to demonstrate BE due to two main factors: (1) an intra-subject coefficient of variation greater than expected and (2) a larger-than-expected relative difference between the Test and Reference formulations; and both may translate into an insufficient sample size (Patterson and Jones, 2005). It is possible that in this analysis some studies failed solely due to an insufficient sample size given that the new pAUC parameters are more variable than the conventional parameters. Previous studies have shown that when an AUC is truncated in the first few hours post administration, it results in a higher ISCV; however, this variability decreases when the truncation is performed at a later time point. This pattern was observed in studies with both conventional (Midha et al., 1996) and MR (Lionberger, 2010) formulations. Therefore, the early pAUC parameters calculated in this study were expected to have a higher ISCV compared to the other AUC parameters, such as AUC(0–t) and the terminal pAUCs. In an analysis of FDA data on 95 MR formulations, the variability of AUCReftmax was shown to be high, particularly when the value of the reference tmax was low. The same analysis also revealed that about 40% of these 95 studies were unable to meet the acceptance criterion for AUCReftmax (Lionberger, 2010), a failure rate similar to what was found in the present analysis. In contrast, in another study focusing on a small dataset of in vivo BE studies submitted to the FDA (N = 17), the failure rate for the parameter was found to be 22% for PR formulations (11% if removing a study with multiple peaks) and 38% for IR formulations. Chen et al. explained this difference with simulation data suggesting that the discriminative power advantage of AUCReftmax over Cmax is smaller for PR than for IR formulations (Chen et al., 2011). These results confirm that the choice of AUCReftmax as a comparison parameter for the new pAUCs was appropriate in the present analysis as it has a
75
greater sensitivity in showing differences in the shape of the curve than the conventional parameters. An unexpected result was that the variability of AUC(0–s/2) was comparable to that of Cmax. The variability of this early pAUC parameter is however unveiled by separating the fed state studies – most variable –, from the fasting studies– less variable – (as shown in Fig. 2). A similar contrast also appears by separating studies according to their ability to show equivalence for AUC(0–s/2) (as shown in Fig. 3). The bioavailability (BA) of drugs that are administered orally is known to vary under fasting and fed conditions. Food can sometimes modify BA by several mechanisms including a delay in gastric emptying (FDA, 2002). It is noteworthy that the increased ISCVs observed in both early pAUC parameters in the fed studies seem to be explained by the intake of a high-fat meal. Most fed studies were associated with a longer time to peak concentration (data not shown). Thus it is possible that the increased variability in these studies is associated with a delay in the formulation drug delivery and/or drug absorption. Interestingly, the association between gastric emptying times and a high variability of pAUC parameters has previously been discussed (Chen et al., 2011). Additionally, the terminal pAUCs parameters presented a greater ISCV than the conventional parameters. These differences can possibly be explained by a smaller number of blood samples collected during the terminal portion of the AUC. In studies with a s of 24 h, the terminal portion represents a period where fewer blood samples were collected (sampling time between 12 and 24 h) in comparison to studies having a s of 12 h (sampling time between 6 and 12 h). This explanation could account for terminal pAUC parameters presenting a smaller difference in ISCV compared to Cmax in studies with a s of 12 h (data not shown). On the other hand, no difference in ISCV between AUC(s/2–s) and AUC(s/2–t) was found when s was 12; a circumstance where AUC(s/2–t) comprised the largest number of collected samples. This result suggests that the variability may not be entirely explained by the number of samples, but more likely by the frequency of sampling. In future studies, it may be possible to reduce the variability by adapting the sampling schedule to the particularities of the terminal pAUC parameter. T/R ratios played an important part in studies failure on new pAUC parameters. In all cases where AUC(0–s/2) or AUC(s/2–s) failed to meet the acceptance criterion, the T/R ratio of the parameter was worse than the ratio of Cmax (not shown). The ratios found in studies of group 1 (Fig. 4) suggest that the new pAUC parameters possess a higher sensitivity in detecting differences in the shape of the curve compared with the conventional parameters. Moreover, it is most likely that the acceptance criteria would not have been met for these formulations even if the studies had been designed to take into consideration the addition of the new pAUC parameters. Similarly, for studies in group 2 (Fig. 4), the variability may have contributed to but does not appear to underlie the failure seeing that the majority of these studies present low ISCVs for the specific parameter that did not meet the acceptance criteria (mean of 21%, range: 16–28%). The ratios are thus the principal cause of the failures for these studies. Considering that all the studies in group 3 (Fig. 4) have a high ISCV and that their ratios are not excessively far from 100%, it is reasonable to assume that BE could be demonstrated for these formulations by an adjustment of the sample size based on the expected variability of the pAUC parameter. It is possible also that some studies may comprise drug products considered as highly variable, which are defined by the EMA as presenting a Reference-to-Reference ISCV greater than 30% (EMA, 2010). Interestingly, the draft guideline would permit a reference-scaled widening of the acceptance interval of the new pAUC parameters for PR products (EMA, 2013) as it is done for Cmax of immediate
76
M. Boily et al. / European Journal of Pharmaceutical Sciences 66 (2015) 70–77
release products (EMA, 2010). This would be acceptable provided that the parameters are shown to be highly variable in a replicated design study. Since 2-way crossover studies were selected in our analysis, it was not possible to estimate the ISCV of the reference product for the pAUC parameters. Finally, there are possible biases in the selection of studies for the present analysis. The sample size of several pivotal studies included in this analysis was estimated based on specific BE requirements of the regulatory agency of submission. In some cases, the sample sizes may have been smaller due to less restrictive BE requirements for certain agencies. However, in all of these studies, the number of subjects included was verified to be sufficient to reach a power of 0.80 for the CI of Cmax with a T/R ratio of 95% (data not shown). This finding suggests that the selected studies did not skew the presented data in a way that might affect the conclusions drawn from the present analysis. Furthermore, there are molecule and drug class trends in the development of generic formulations. Thus, future BE studies may involve less or more variable drug products than those included in the present analysis. These trends may also have an impact on the proportion of drugs that are likely or unlikely to lead to accumulation. Finally, it is uncertain if the findings of the present analysis also apply to highly variable drugs as most of the studies included had ISCVs lower than 25% for the conventional parameters. As previously stated, one of the clear findings of the present evaluation was that very few drug formulations will qualify for a waiver for the steady-state study. Therefore, it would be of interest to evaluate if other PK parameters in single dose studies could be more predictive of the success or failure of steady-state studies. The predictive power of the parameter Cs has recently been evaluated; however, results from simulated data (Paixao et al., 2012) and clinical data (Garcia-Arieta et al., 2012) have been inconsistent. It was considered but not selected by the EMA (EMA, 2013). In all cases, the cost-benefits of the steady-state study waiver, including reducing the conduct of a number of large and expensive studies, should be weighed against the increased variability of the new pAUC parameters. 5. Conclusion The results of this analysis suggest that the changes regarding the pharmacokinetic evaluation of BE of PR formulations, proposed in the EMA draft guideline, will have an impact on future studies conducted for submission to European agencies. Although this draft guideline is unclear regarding the precise truncation time point of the newly introduced pAUC parameters, an increase in failed BE studies was observed following the evaluation of several possible pAUC parameters. Even if clinical relevance of the new pAUCs is not well known, it remains that differences in the concentration–time profile of the PR formulations will be captured. The ability to detect these differences could lead to an increase in patient safety, and for formulations where this could be clinically relevant, to a better formulation efficacy. On the other hand, studies failing to demonstrate BE due to their variability were also identified, which reinforces the need for new designs including more sampling time points and larger sample size. The present analysis was performed based on the authors’ interpretation of the draft guideline, which is likely to include modifications in its adopted version. Disclosure/conflict of interest Michael Boily, Catherine Dussault, Julie Massicotte, Pascal Guibord and Marc Lefebvre are employees of the contract research organisation Algorithme Pharma Inc.
Acknowledgements The authors would like to acknowledge Viviane Luu for her valuable input in the review of the draft manuscript. The collection of data, data analysis and the report of the results of this study were supported by Algorithme Pharma Inc. References Anschutz, M., Wonnemann, M., Schug, B., Toal, C., Donath, F., Pontius, A., Pauli, K., Brendel, E., Blume, H., 2010. Differences in bioavailability between 60 mg of nifedipine osmotic push-pull systems after fasted and fed administration. Int. J. Clin. Pharmacol. Ther. 48, 158–170. Chen, M.-L., Davit, B., Lionberger, R., Wahba, Z., Ahn, H.-Y., Yu, L., 2011. Using partial area for evaluation of bioavailability and bioequivalence. Pharm. Res. 28, 1939– 1947. EMA, 2010. Guideline on the Investigation of Bioequivalence (CPMP/EWP/QWP/ 1401/98 Rev.1/Corr⁄⁄). (Accessed 31.01.14). EMA, 2013. Guideline on the Pharmacokinetic and Clinical Evaluation of Modified Release Dosage Forms (EMA/CPMP/EWP/280/96 Corr1). (Accessed 31.01.14). EMA, 2014. Questions & Answers: Positions on Specific Questions Addressed to the Pharmacokinetics Working Party (EMA/618604/2008 Rev. 9). (Accessed 16.07.14). EMEA, 1999. Note for Guidance on Modified Release Oral and Transdermal Dosage Forms: Section II (CPMP/EWP/280/96). (Accessed 31.01.14). Endrenyi, L., Tothfalusi, L., 2010. Do regulatory bioequivalence requirements adequately reflect the therapeutic equivalence of modified-release drug products? J. Pharm. Pharm. Sci. 13, 107–113. FDA, 2002. Food-Effect Bioavailability and Fed Bioequivalence Studies. (Accessed 31.01.14). FDA, 2003. Bioavailability and Bioequivalence Studies for Orally Administered Drug Products—General Considerations: Guidance for Industry, FDA, Rockville, MD: Center for Drug Evaluation and Research. FDA, 2010. Background Information for the FDA Meeting of the Advisory Committee for Pharmaceutical Science and Clinical Pharmacology, Topic 2: Use of Partial Area Under the Curve (AUC) for Products with a Complex Pharmacokinetic (PK) Profile. Center for Drug Evaluation and Research. (Accessed 31.01.14. FDA, 2011. Guidance on Zolpidem. (Accessed 26.02.14). FDA, 2012a. Draft Guidance on Mesalamine. (Accessed 31.01.14). FDA, 2012b. Draft Guidance on Methylphenidate Hydrochloride. (Accessed 31.01.14). FDA, 2013. Bioequivalence Studies with Pharmacokinetic Endpoints for Drugs Submitted Under an ANDA: Guidance for industry (Draft Guidance). FDA, Rockville, MD: Center for Drug Evaluation and Research. Garcia-Arieta, A., 2013. Regulatory Requirements and Recommendations of the Guideline on the Pharmacokinetic and Clinical Evaluation of Modified Release Dosage Forms (EMA/CPMP/EWP/280/96 Corr1), Session III: Specific Issues for BE Assessment. EUFEPS BABP Network Open Discussion Forum. (Accessed 31.01.14. Garcia-Arieta, A., Morales-Alcelay, S., Herranz, M., de la Torre-Alvarado, J.M., Blazquez-Perez, A., Suarez-Gea, M.L., Alvarez, C., 2012. Investigation on the need of multiple dose bioequivalence studies for prolonged-release generic products. Int. J. Pharm. 423, 321–325. Health Canada, 1992. Report C: Report on Bioavailability of Oral Dosage Formulations, not in Modified Release Form, of Drugs Used for Systemic Effects, Having Complicated or Variable Pharmacokinetics, Health Canada, Ottawa, ON, Expert Advisory Committee on Bioavailability, Health Protection Branch. Health Canada, 1996. Guidance Document – Conduct and Analysis of Bioavailability and Bioequivalence Studies – Part B: Oral Modified Release Formulations. Health Canada, Ottawa, ON. Health Canada, 2012. Guidance Document – Comparative Bioavailability Standards: Formulations Used for Systemic Effects. (Accessed 31.01.14). Julious, S.A., 2004. Sample sizes for clinical trials with normal data. Stat. Med. 23, 1921–1986.
M. Boily et al. / European Journal of Pharmaceutical Sciences 66 (2015) 70–77 Lionberger, R.A., 2010. Topic 2: PK Profile Comparison for Modified Release Products. Advisory Committee for Pharmaceutical Science and Clinical Pharmacology. (Accessed 31.01.14). Midha, K.K., Hubbard, J.W., Rawson, M.J., 1996. Retrospective evaluation of relative extent of absorption by the use of partial areas under plasma concentration versus time curves in bioequivalence studies on conventional release products. Eur. J. Pharm. Sci. 4, 381–384.
77
Paixao, P., Gouveia, L.F., Morais, J.A., 2012. An alternative single dose parameter to avoid the need for steady-state studies on oral extended-release drug products. Eur. J. Pharm. Biopharm. 80, 410–417. Patterson, S., Jones, B., 2005. Dealing with Unexpected BE Challenges, Bioequivalence and Statistics in Clinical Pharmacology. Chapman & Hall/CRC, London, pp. 133–175.
Contents lists available at ScienceDirect
European Journal of Pharmaceutical Sciences journal homepage: www.elsevier.com/locate/ejps
The impact of new partial AUC parameters for evaluating the bioequivalence of prolonged-release formulations Michaël Boily a,b, Catherine Dussault a, Julie Massicotte a, Pascal Guibord c, Marc Lefebvre a,⇑ a
Scientific and Regulatory Affairs, Algorithme Pharma Inc., 575 Armand-Frappier Blvd., Laval, QC H7V 4B3, Canada Département de Pharmacologie, Faculté de Médecine, Université de Montréal, 2900 Édouard-Montpetit Blvd., Montreal, QC, Canada c Biometrics, Algorithme Pharma Inc., 575 Armand-Frappier Blvd., Laval, QC H7V 4B3, Canada b
a r t i c l e
i n f o
Article history: Received 17 April 2014 Received in revised form 15 August 2014 Accepted 2 October 2014 Available online 13 October 2014 Keywords: Modified release formulations Bioequivalence Pharmacokinetics parameters Partial AUC Drug accumulation
a b s t r a c t To demonstrate bioequivalence (BE) between two prolonged-release (PR) drug formulations, single dose studies under fasting and fed state as well as at least one steady-state study are currently required by the European Medicines Agency (EMA). Recently, however, there have been debates regarding the relevance of steady-state studies. New requirements in single-dose investigations have also been suggested by the EMA to address the absence of a parameter that can adequately assess the equivalence of the shape of the curves. In the draft guideline issued in 2013, new partial area under the curve (pAUC) pharmacokinetic (PK) parameters were introduced to that effect. In light of these potential changes, there is a need of supportive clinical evidence to evaluate the impact of pAUCs on the evaluation of BE between PR formulations. In this retrospective analysis, it was investigated whether the newly defined parameters were associated with an increase in discriminatory ability or a change in variability compared to the conventional PK parameters. Among the single dose studies that met the requirements already in place, 20% were found unable to meet the EMA’s new requirements in regards to the pAUC PK parameters. When pairing fasting and fed studies for a same formulation, the failure rate increased to 40%. In some cases, due to the high variability of these parameters, an increase of the sample size would be required to prove BE. In other cases however, the pAUC parameters demonstrated a robust ability to detect differences between the shapes of the curves of PR formulations. The present analysis should help to better understand the impact of the upcoming changes in European regulations on PR formulations and in the design of future BE studies. Ó 2014 Elsevier B.V. All rights reserved.
1. Introduction In early 2013, the European Medicines Agency (EMA) issued a draft bioequivalence (BE) guideline for modified release (MR) drug products, providing recommendations for delayed-release and prolonged-release (PR) formulations (EMA, 2013). This guideline, when finalized, will replace its more than a decade old precursor. One of the major changes introduced in this document is a requirement to demonstrate the equivalence of the shape of the systemic concentration vs time curves between compared PR products. There are cases where a PR Test formulation, for which the 90% confidence intervals (CI) for the overall area under the curve (AUC) and the maximum plasma concentration (Cmax) fall within ⇑ Corresponding authors at: Scientific and Regulatory Affairs, Algorithme Pharma Inc., 575 Armand-Frappier Blvd., Laval, QC H7V 4B3, Canada. Tel.: +1 450 973 6077; fax: +1 450 973 2801. E-mail address: [email protected] (M. Lefebvre). http://dx.doi.org/10.1016/j.ejps.2014.10.001 0928-0987/Ó 2014 Elsevier B.V. All rights reserved.
the bioequivalence limits (80–125%) of the Reference PR product, presents a major difference in the early (Anschutz et al., 2010; Garcia-Arieta, 2013) or in the terminal segment of its curve (Garcia-Arieta, 2013). Detecting such patterns, which conventional parameters can fail to reveal, appears to be the rationale behind the introduction of new partial AUC (pAUC) parameters in the guideline. A pAUC is an AUC determined between given time points (Endrenyi and Tothfalusi, 2010). A parameter measuring the early segment of the curve, calculated as the AUC truncated at the tmax of the marketed reference product calculated for each study subject (AUCReftmax), was the first pAUC to be adopted by a regulatory agency (Health Canada, 1992). It is still currently used and recommended by Health Canada (Health Canada, 2012) when the early exposure of an immediate release drug product is important. The US Food and Drug Administration (FDA) introduced in 2003 the use of a partial AUC as an early exposure measure, defined as the
M. Boily et al. / European Journal of Pharmaceutical Sciences 66 (2015) 70–77
AUC truncated at the population median of tmax values for the reference formulation (FDA, 2003). The FDA is proposing to redefine this pAUC in their new draft guidance, where the time to truncate the partial area should be related to a clinically relevant pharmacodynamic measure (FDA, 2013). In line with the new draft guidance, the parameter AUC8–48 is now a requirement for mesalamine delayed-release tablets, a drug with a local action in the colon (FDA, 2012a). The FDA also discussed (FDA, 2010) and introduced product-specific pAUC parameters for at least two different multiphasic dosage forms (FDA, 2011, 2012b), and the EMA recommended a similar approach for the demonstration of BE for biphasic MR formulations (EMA, 2014). In its current guidance on MR formulations, effective since January 2000, the EMA requires BE for all PR formulations to be assessed in both single and multiple dose studies, using conventional parameters (AUC(0–t), AUC(0–1), Cmax; and AUCs, Cmax, Cmin, respectively) (EMEA, 1999). The new EMA draft guideline specifies that for PR formulations for which accumulation is likely after a single dose, BE demonstration in single dose studies will be based on the conventional parameters as well as an additional parameter identified as a pAUC (see Fig. 1). A steady-state study will also still be required for those formulations. In contrast, for PR drug products not likely to accumulate, a waiver for the conduct of a steady-state study may be obtained if acceptance limits are met for two newly added pAUC parameters (early and terminal) in the single dose fasting and fed state studies. Interestingly, the suggested truncation point is half of the dosage interval (s/2) (EMA, 2013), resulting in new parameters that have not been evaluated extensively, unlike other pAUCs such as AUCReftmax (Chen et al., 2011; Lionberger, 2010). The introduction of this decision tree and new pharmacokinetic (PK) parameters raises important questions on the outcome of future BE studies, as the new PK parameters may impact key study design factors that influence the cost of conducting the studies (e.g. sample size, number of studies, number of periods/sequence, sampling schedule, etc). As an example, the cost of the approval process would decrease significantly in the case where steady-state studies would no longer be required. To assess the impact of this draft guideline, a review of an internal database was conducted in order to recalculate different PK parameters including early and terminal pAUCs as well as the
71
conventional parameters from 53 PR single dose studies that met standard BE requirements. From this retrospective evaluation, the outcome of the new BE criteria was assessed. Furthermore, a secondary objective of this analysis was to characterize the variability of the new parameters. 2. Materials and methods 2.1. Inclusion of bioequivalence studies The database of Algorithme Pharma studies conducted between year 2007 and 2013 was screened for studies that met all of the following criteria: – Pivotal bioequivalence studies with a two-period, twosequence and two-treatment cross-over design; – Oral administration to healthy volunteers in a single dose study design; – Prolonged-release Test and Reference formulations (this criterion included extended-release, sustained-release and controlled-release formulations; and excluded delayedrelease formulations and multiphasic MR products for which different regulatory requirements are presented in the EMA draft guidance (EMA, 2013)); – 90% CI that falls within 80–125% for all the conventional PK parameters (as defined in the following section: 2.2 pharmacokinetics parameters). 2.2. Pharmacokinetic parameters The calculation of the different PK parameters was performed using Kinetic software (version 9.01; Algorithme Pharma Inc., Laval, Canada). AUCs were calculated using the linear trapezoidal method. The different PK parameters were classified as follows: (1) conventional; (2) early pAUCs; (3) terminal pAUCs; and (4) other parameters. The parameters in the conventional category are those commonly used for assessing BE in accordance with the current EMA guideline (CPMP/EWP/280/96) and are Cmax, AUC(0–t), and AUC(0–1). The early and terminal pAUCs include the new partial AUC parameters mentioned in the EMA draft guideline (CPMP/
Fig. 1. Interpretation of the EMA’s draft guidance’s new requirements concerning parameters to be used and studies to be conducted depending on the expected extent of accumulation of Test and Reference formulations. From our understanding of the EMA’s draft guidance, to demonstrate BE of two PR formulations, a fed and a fasting state study at single dose are minimally required. If the Test and the Reference formulations present a (AUC(0–s)/AUC(0–1)) ratio 90 %, they are considered likely to accumulate. If this ratio is > 90%, they are considered as unlikely to accumulate. The PK requirements corresponding to each possible case are shown (dashed line rectangle). The number of studies (n) among the 53 studies that met the inclusion criteria that corresponds to each case is also presented, whether or not the associated requirements were met.
72
M. Boily et al. / European Journal of Pharmaceutical Sciences 66 (2015) 70–77
EWP/280/96 Rev.1). Those two pAUCs (early and terminal) are separated by a time point which is usually the half of the dosage interval (s) of the investigated formulation (s/2) (EMA, 2013). Therefore, the early pAUC described in the draft guideline refers to AUC(0–s/2), being the AUC from the administration time (0) to the mid-interval time (s/2). For comparative and informational purposes, AUCReftmax was also included in the early pAUC category. The latter is a parameter recommended by Health Canada as a measure of early exposure, which is of interest when the time of onset of effect or the rate of absorption is important (Health Canada, 2012). AUCReftmax is defined as the area under the curve from time zero to the tmax of the Reference product calculated for each study subject. The terminal pAUC is described as the area under the curve from half of s (s/2) to an end point which is not defined in the EMA draft guideline. It was thus hypothesized that this end point could be either s, t (time of last observed quantifiable plasma concentration) or 1, resulting in 3 possible terminal pAUCs: AUC(s/2–s), AUC(s/2–t) and AUC(s/2–1). Finally, AUC(0–s) was also evaluated and added in the last category (other). To determine whether drugs are likely to accumulate, the AUC(0–s)/AUC(0–1) ratio (%) was also calculated for the Test and Reference product of each study. 2.3. Statistical analysis The natural logarithmic (ln) transformation of the parameters was used for all statistical inference. Parameters were statistically analyzed using an Analysis of Variance (ANOVA) model. The fixed factors included in this model were the subject effect (nested within sequence), the treatment received, the period at which it was given, as well as the sequence in which each treatment was received. The 90% CI for the exponential of the difference in LSmeans between the Test and Reference product was calculated for the ln-transformed parameters (Test to Reference ratio of geometric LSmeans). If a PK parameter could not be determined for one period in a subject, the corresponding subject was excluded from that particular statistical comparison. BE statistical analyses were generated with SASÒ software (version 9.2; SAS Institute Inc., Cary, NC, USA) using the GLM procedure. A study was considered to pass on a specific parameter if its 90% CI fell within the 80.00–125.00% range. For practical purposes, ratio estimates are presented rounded to the nearest unit. Fisher’s exact test was employed to compare proportions. Differences in intra-subject variability (ISCV) between 2 parameters are reported as the mean ± 95% CI. Sample size estimations were performed using the equation described by Julious et al. with a power of 80%, a type 1 error of 5% and using BE criteria of 20% (Julious, 2004). 3. Results
drugs and the dosage interval of the formulations found in these studies. The studies included 21 different drugs over 13 different therapeutic categories. According to the 2013 draft guideline, a low extent of accumulation will be expected if the AUC(0–s)/AUC(0–1) ratio is above 90% in the single dose study, as illustrated in Fig. 1. Moreover, for these products with no risk of accumulation, it will be possible to obtain a waiver for the steady-state study. Among the 53 studies selected, 49 (92.45%) were considered likely to accumulate vs only 4 (7.55%) unlikely to accumulate for both the Test and Reference products. As the 90% value is still under discussion, the impact of decreasing the AUC(0–s)/AUC(0–1) ratio to 80% was also assessed. In an outdated 1996 guideline for MR formulations, Health Canada recommended to consider a ratio of 80% as an indicator to characterize the likelihood of accumulation (Health Canada, 1996). With this 80% ratio, a total of 10 studies (18.86%) – 6 more than for the 90% ratio – would be considered unlikely to lead to accumulation. For formulations meeting the conventional BE standards and that are considered likely to accumulate, failure to meet the acceptance criterion for the new pAUC parameter would correspond in a failure to demonstrate BE. However, the pAUC proposed in the draft guideline, for both the fasting and fed single dose studies, is not clearly defined. It was thus hypothesized that this parameter was either an early or terminal pAUC with a truncation time point at s/2. This is essentially the same as either the early or the terminal pAUC proposed in the draft guideline for drugs unlikely to accumulate. Table 2 presents the number of studies that fail to show statistical equivalence for each early or terminal pAUC PK parameter as well as for other parameters. If AUC(0–s/2) was to be considered as the required pAUC parameter, 10 studies out of 49 (20.41%) would fail to demonstrate BE. The same number of studies would also fail to demonstrate BE if any of the proposed 3 terminal pAUCs was selected as an alternative. Table 2 also displays that there was a significantly higher proportion of fed studies that failed to meet the acceptance criterion as compared to the studies completed in a fasting state for both of the early pAUC PK parameters (33.33% vs 68.42% for AUCReftmax, and 10.00 vs 36.84% for AUC(0–s/2)). Such a difference was absent for other parameters. In the simplest scenario, a new generic PR formulation would meet acceptance requirements of the EMA if BE criteria are met in one single dose study in the fasting state and in one single dose study in the fed state. Among the 53 individual studies which met conventional PK parameters requirements, 12 pairs consisting of one fasting study and one fed study conducted with the same batch of comparison products were identified (for a total of 24 studies). Of those pairs, 41.6% (5/12) contained at least one study that did not meet the new criteria; the fasting or the fed study failed in four pairs whereas both studies failed in only one pair.
3.1. Impact of PK parameters on BE 3.2. Variability of the new PK parameters Among the studies evaluated, 53 met the criteria defined above (Section 2.1). Table 1 comprises a description of the categories of Table 1 Description of formulations in studies included in the analysis. 5 Most common drug class Identified by therapeutic categorya Analgesics Antidepressants Urologicals Anti-diabetics Vasodilatators Others Total a
Number of different drugs = 21.
Dosage interval %
(n)
33.96 18.87 9.43 7.55 5.66 24.53
(18) (10) (5) (4) (3) (13)
100.00
(53)
% 12 h 24 h
(n)
37.74 62.26
(20) (33)
100.00
(53)
The variability of the new pAUC parameters was assessed in order to determine if it accounted for the increase in the number of studies failing to meet BE standards. Cmax was the conventional parameter with the highest ISCV in most studies. In each study, the difference in ISCV between each AUC parameters and Cmax was compared. Fig. 2 presents the average difference in ISCV calculated for the 53 studies that met the conventional BE standards. As expected, the conventional parameters AUC(0–t), AUC(0–1) and AUC(0–s) were less variable than Cmax. In contrast, the early pAUC parameter AUC(0–s/2) appeared to be slightly more variable, presenting an average difference in ISCV of 0.76[0.81; 2.34] percentage points with Cmax. All the terminal pAUCs presented a significantly larger difference in ISCV ranging from 3.11[0.73; 5.50] to 3.94[1.43; 6.45] percentage points. AUCReftmax was the parameter
73
M. Boily et al. / European Journal of Pharmaceutical Sciences 66 (2015) 70–77 Table 2 Percentage of studies which 90% CI of individual PK parameters fall outside the acceptance limits in drugs likely to accumulate.
a
Parameter
All studies (n = 49)
Fasting (n = 30)
Fed (n = 19)
% fail
(n)
% fail
(n)
% fail
(n)
Early
AUCReftmax AUC(0–s/2)
46.94 20.41
(23) (10)
33.33 10.00
(10) (3)
68.42a 36.84a
(13) (7)
Terminal
AUC(s/2–s) AUC(s/2–t) AUC(s/2–1)
20.41 20.41 20.41
(10) (10) (10)
20.00 23.33 23.33
(6) (7) (7)
21.05 15.79 15.79
(4) (3) (3)
Other
AUCs
4.08
(2)
3.33
(1)
5.26
(1)
Significatively different from fasting state (p < 0.05).
Early pAUC
AUC(0−τ 2) AUCReftmax AUC(τ AUC(τ 2−t)
AUC(τ
Terminal pAUC
2−∞) 2−τ)
AUC(0−∞)
Studies All studies
AUC(0−t)
Fasting Fed
AUC(0−τ) Cmax
−10
0
10
20
30
Difference in ISCV (%) with Cmax Fig. 2. Difference between the ISCV of investigated parameters and the corresponding ISCV of Cmax in fasting and fed state studies. For each evaluated parameter, the mean difference (with the 95% CI) in ISCV of that parameter and Cmax is plotted for all studies (red circles, n = 53), fasting state studies (green triangles, n = 31) and fed state studies (blue squares, n = 22).
showing the highest variability. Thus, all the pAUC parameters were more variable than the conventional AUC parameters. The effect of food on the ISCV was also verified. As shown in Fig. 2, a difference between fasting and fed studies was observed for only the early pAUC parameters. Both AUCReftmax and AUC(0–s/2) were significantly more variable in the fed state than under fasting conditions. Fig. 3 presents the difference in ISCV with Cmax depending on whether studies met or not the acceptance criterion for the two pAUC parameters AUC(0–s/2) and AUC(s/2–s). In studies which CI did not fall within acceptance limits for AUC(0–s/2), there was a tendency for the parameter AUC(0–s/2) to be more variable than Cmax. A similar tendency was present in studies that failed to meet requirements for AUC(s/2–s) as in that case AUC(s/2–s) appeared to be more variable. Data for AUC(s/2–t) and AUC(s/2–1) also presented this pattern (data not shown). This suggests an association between failure on a pAUC parameter and the variability of that same parameter. 3.3. Impact of the new parameters on the Test to Reference ratio In Fig. 4A, the 53 studies that met the conventional BE criteria are plotted according to their respective geometric mean for the
Test/Reference (T/R) ratio and ISCV for AUC(0–s/2). One study with a T/R ratio of 268% and ISCV of 33% is not displayed in Fig. 4A. For studies that failed on this parameter, the 90% CI is illustrated. Studies were classified into 3 groups (1, 2, 3) based on their results. Group 1 includes the studies with ratios outside the 80–125% range; group 2 includes studies with ratios inside the 80–125% range but outside the 90–111% range, and group 3 includes studies with ratios inside the 90–111% range. Two out of 10 studies (20%) had T/R ratios out of the 80–125% BE acceptance limits (group 1) and 6 studies (60%) had ratios inside the latter limits but outside the 90–111% (group 2). Three of these 6 studies, had a higher ISCV for AUC(0–s/2) than for Cmax; however, in all cases, ISCVs were under 23%. Finally, the remaining 2 studies (20%) had a ratio located between 90 and 111% and this is associated with a large positive difference in ISCV with Cmax (group 3). An example plot of Test and Reference concentration over time is also given (Fig. 4A, bottom of panel) for a study belonging to group 1 and a study belonging to group 3. Fig. 4B is similar to Fig. 4A but displays data for AUC(s/2–s). The ratio of 5 studies out of 12 (42%) did not fall within the 80–125% limits (group 1). Another 5 studies (42%) could be classified as group 2. All their ISCVs were under 28% and 4 of these 5 studies had a higher variability for AUC(0–s/2) than for Cmax.
74
M. Boily et al. / European Journal of Pharmaceutical Sciences 66 (2015) 70–77
AUCReftmax
Early pAUC
AUC(0−τ 2) 2−τ)
Terminal pAUC
AUC(τ
Studies AUC(0−τ)
All studies Fail on early pAUC Fail on terminal pAUC No fail on pAUCs
Cmax
−10
0
10
20
30
Difference in ISCV (%) with Cmax Fig. 3. Difference between the ISCV of investigated parameters and the corresponding ISCV of Cmax in studies failing on a pAUC parameter. For parameters of early and terminal pAUC and comparison parameters, the mean difference in ISCV (with the 95% CI) of that parameter and Cmax is plotted for all studies (red circles, n = 53), for all studies that failed on AUC(0-s/2), the early pAUC (green triangles, n = 10), for all studies that failed on AUC(s/2-s), the terminal pAUC parameter (blue squares, n = 12) or studies meeting requirements for both of the 2 latter pAUC parameters (orange diamond shape, n = 34).
(A) 60
(B) 60
50
50
issue Fail
40
ii
i
ISCV (%)
ISCV (%)
40 30
Pass
i
30
group 1 A
20
20
10
10
0
0
2 B 3 C
ii 60
80
100
120
140
160
180
40
AUC 0−τ 2 T/R ratio (%)
6 12 18 24 30 36 42 48 54 60 66 72
6 12 18 24 30 36 42 48 54 60 66 72
Time (h)
Cmax
120
140
160
180
2−τ T/R ratio (%)
AUCt
AUC(0−τ 2)
AUC(τ
2−τ)
6 4 2 0
60
Time (h)
0
frequency (n)
(C)
100
80 10 0 12 0 14 0 60 80 10 0 12 0 14 0 60
ii
0
80
AUCτ
Concentration
i
60
80 10 0 12 0 14 0 60 80 10 0 12 0 14 0
40
T/R ratio (%) Fig. 4. ISCV in function of Test to Reference ratio for (A) AUC(0-s /2) or (B) AUC(s/2-s). For every studies (n = 53), the ISCV of the corresponding parameter is plotted in function of the T/R ratio. Studies that failed on the parameter (A) AUC(0-s/2) or (B) AUC(s/2-s) are plotted (red dots) with their 90% CI. Studies that met requirements for these parameters are also displayed (green triangles). Group 1 to 3 are highlighted according to a gray scale. Group 1 corresponds to studies with ratios outside the 80-125% acceptance limits (dark gray), which is also delimitated by dashed lines. Group 2 corresponds to ratios between 80-90% or 111-125% (light gray). Group 3 corresponds to ratios within the 90-111% limits (white). Two examples of studies not meeting acceptance criterion for AUC(0-s/2) are also plotted (i, ii). A plot of their concentration versus time is displayed for both the Test formulation (black line) and the Reference formulation (light gray). The first example did not meet requirements for both early and terminal pAUC (i). The timepoint s/2 is represented by a dashed line. (C) Comparison of the distribution of T/R ratios for various parameters. The bandwidth of the histograms corresponds to 1% T/R ratio.
M. Boily et al. / European Journal of Pharmaceutical Sciences 66 (2015) 70–77
Finally, two studies (17%) were included in group 3 and had a high variability, which was also at least twice that of Cmax. Fig. 4C depict the frequency of the T/R ratios found for two conventional parameters as well as for early (AUC(0–s/2)) and terminal (AUC(s/2–s)) pAUC parameters. A higher number of AUC(s/2–s) T/R ratios was outside the 80–125% limit than for the other parameters. Also, AUC(0–s/2) T/R ratios were more sparsely distributed resulting in less values around 100%.
4. Discussion While the addition of new pAUC parameters to the acceptance criteria should improve the comparison of the shape of the curves between PR formulations, more studies are expected to fail to demonstrate BE. As suggested by the results presented in the current analysis, a large majority of PR formulations will be categorized as products that are likely to accumulate with an increased difficulty to meet the new BE criteria. For the other drug products, for which a low extent of accumulation is expected, the number of studies evaluated for these formulations was probably too small to bring adequate conclusion. Nevertheless, this indicates that very few studies will meet the required conditions to obtain a waiver for the steady-state study. Therefore, further research may be of interest to evaluate the feasibility and impact of broadening the waiver qualification criteria provided that the safety considerations are not jeopardized. On the other hand, preventing the conduct of a steady-state study for even a small proportion of PR formulations could reduce the number of human subjects exposed in BE studies as well as decrease the development costs. This could be significant considering that a steady-state study is recommended by the current EMA guidance in nearly all circumstances (EMEA, 1999). In order to properly evaluate the impact of the new pAUC parameters, it was first necessary to identify why studies failed to meet acceptance criteria. A study may fail to demonstrate BE due to two main factors: (1) an intra-subject coefficient of variation greater than expected and (2) a larger-than-expected relative difference between the Test and Reference formulations; and both may translate into an insufficient sample size (Patterson and Jones, 2005). It is possible that in this analysis some studies failed solely due to an insufficient sample size given that the new pAUC parameters are more variable than the conventional parameters. Previous studies have shown that when an AUC is truncated in the first few hours post administration, it results in a higher ISCV; however, this variability decreases when the truncation is performed at a later time point. This pattern was observed in studies with both conventional (Midha et al., 1996) and MR (Lionberger, 2010) formulations. Therefore, the early pAUC parameters calculated in this study were expected to have a higher ISCV compared to the other AUC parameters, such as AUC(0–t) and the terminal pAUCs. In an analysis of FDA data on 95 MR formulations, the variability of AUCReftmax was shown to be high, particularly when the value of the reference tmax was low. The same analysis also revealed that about 40% of these 95 studies were unable to meet the acceptance criterion for AUCReftmax (Lionberger, 2010), a failure rate similar to what was found in the present analysis. In contrast, in another study focusing on a small dataset of in vivo BE studies submitted to the FDA (N = 17), the failure rate for the parameter was found to be 22% for PR formulations (11% if removing a study with multiple peaks) and 38% for IR formulations. Chen et al. explained this difference with simulation data suggesting that the discriminative power advantage of AUCReftmax over Cmax is smaller for PR than for IR formulations (Chen et al., 2011). These results confirm that the choice of AUCReftmax as a comparison parameter for the new pAUCs was appropriate in the present analysis as it has a
75
greater sensitivity in showing differences in the shape of the curve than the conventional parameters. An unexpected result was that the variability of AUC(0–s/2) was comparable to that of Cmax. The variability of this early pAUC parameter is however unveiled by separating the fed state studies – most variable –, from the fasting studies– less variable – (as shown in Fig. 2). A similar contrast also appears by separating studies according to their ability to show equivalence for AUC(0–s/2) (as shown in Fig. 3). The bioavailability (BA) of drugs that are administered orally is known to vary under fasting and fed conditions. Food can sometimes modify BA by several mechanisms including a delay in gastric emptying (FDA, 2002). It is noteworthy that the increased ISCVs observed in both early pAUC parameters in the fed studies seem to be explained by the intake of a high-fat meal. Most fed studies were associated with a longer time to peak concentration (data not shown). Thus it is possible that the increased variability in these studies is associated with a delay in the formulation drug delivery and/or drug absorption. Interestingly, the association between gastric emptying times and a high variability of pAUC parameters has previously been discussed (Chen et al., 2011). Additionally, the terminal pAUCs parameters presented a greater ISCV than the conventional parameters. These differences can possibly be explained by a smaller number of blood samples collected during the terminal portion of the AUC. In studies with a s of 24 h, the terminal portion represents a period where fewer blood samples were collected (sampling time between 12 and 24 h) in comparison to studies having a s of 12 h (sampling time between 6 and 12 h). This explanation could account for terminal pAUC parameters presenting a smaller difference in ISCV compared to Cmax in studies with a s of 12 h (data not shown). On the other hand, no difference in ISCV between AUC(s/2–s) and AUC(s/2–t) was found when s was 12; a circumstance where AUC(s/2–t) comprised the largest number of collected samples. This result suggests that the variability may not be entirely explained by the number of samples, but more likely by the frequency of sampling. In future studies, it may be possible to reduce the variability by adapting the sampling schedule to the particularities of the terminal pAUC parameter. T/R ratios played an important part in studies failure on new pAUC parameters. In all cases where AUC(0–s/2) or AUC(s/2–s) failed to meet the acceptance criterion, the T/R ratio of the parameter was worse than the ratio of Cmax (not shown). The ratios found in studies of group 1 (Fig. 4) suggest that the new pAUC parameters possess a higher sensitivity in detecting differences in the shape of the curve compared with the conventional parameters. Moreover, it is most likely that the acceptance criteria would not have been met for these formulations even if the studies had been designed to take into consideration the addition of the new pAUC parameters. Similarly, for studies in group 2 (Fig. 4), the variability may have contributed to but does not appear to underlie the failure seeing that the majority of these studies present low ISCVs for the specific parameter that did not meet the acceptance criteria (mean of 21%, range: 16–28%). The ratios are thus the principal cause of the failures for these studies. Considering that all the studies in group 3 (Fig. 4) have a high ISCV and that their ratios are not excessively far from 100%, it is reasonable to assume that BE could be demonstrated for these formulations by an adjustment of the sample size based on the expected variability of the pAUC parameter. It is possible also that some studies may comprise drug products considered as highly variable, which are defined by the EMA as presenting a Reference-to-Reference ISCV greater than 30% (EMA, 2010). Interestingly, the draft guideline would permit a reference-scaled widening of the acceptance interval of the new pAUC parameters for PR products (EMA, 2013) as it is done for Cmax of immediate
76
M. Boily et al. / European Journal of Pharmaceutical Sciences 66 (2015) 70–77
release products (EMA, 2010). This would be acceptable provided that the parameters are shown to be highly variable in a replicated design study. Since 2-way crossover studies were selected in our analysis, it was not possible to estimate the ISCV of the reference product for the pAUC parameters. Finally, there are possible biases in the selection of studies for the present analysis. The sample size of several pivotal studies included in this analysis was estimated based on specific BE requirements of the regulatory agency of submission. In some cases, the sample sizes may have been smaller due to less restrictive BE requirements for certain agencies. However, in all of these studies, the number of subjects included was verified to be sufficient to reach a power of 0.80 for the CI of Cmax with a T/R ratio of 95% (data not shown). This finding suggests that the selected studies did not skew the presented data in a way that might affect the conclusions drawn from the present analysis. Furthermore, there are molecule and drug class trends in the development of generic formulations. Thus, future BE studies may involve less or more variable drug products than those included in the present analysis. These trends may also have an impact on the proportion of drugs that are likely or unlikely to lead to accumulation. Finally, it is uncertain if the findings of the present analysis also apply to highly variable drugs as most of the studies included had ISCVs lower than 25% for the conventional parameters. As previously stated, one of the clear findings of the present evaluation was that very few drug formulations will qualify for a waiver for the steady-state study. Therefore, it would be of interest to evaluate if other PK parameters in single dose studies could be more predictive of the success or failure of steady-state studies. The predictive power of the parameter Cs has recently been evaluated; however, results from simulated data (Paixao et al., 2012) and clinical data (Garcia-Arieta et al., 2012) have been inconsistent. It was considered but not selected by the EMA (EMA, 2013). In all cases, the cost-benefits of the steady-state study waiver, including reducing the conduct of a number of large and expensive studies, should be weighed against the increased variability of the new pAUC parameters. 5. Conclusion The results of this analysis suggest that the changes regarding the pharmacokinetic evaluation of BE of PR formulations, proposed in the EMA draft guideline, will have an impact on future studies conducted for submission to European agencies. Although this draft guideline is unclear regarding the precise truncation time point of the newly introduced pAUC parameters, an increase in failed BE studies was observed following the evaluation of several possible pAUC parameters. Even if clinical relevance of the new pAUCs is not well known, it remains that differences in the concentration–time profile of the PR formulations will be captured. The ability to detect these differences could lead to an increase in patient safety, and for formulations where this could be clinically relevant, to a better formulation efficacy. On the other hand, studies failing to demonstrate BE due to their variability were also identified, which reinforces the need for new designs including more sampling time points and larger sample size. The present analysis was performed based on the authors’ interpretation of the draft guideline, which is likely to include modifications in its adopted version. Disclosure/conflict of interest Michael Boily, Catherine Dussault, Julie Massicotte, Pascal Guibord and Marc Lefebvre are employees of the contract research organisation Algorithme Pharma Inc.
Acknowledgements The authors would like to acknowledge Viviane Luu for her valuable input in the review of the draft manuscript. The collection of data, data analysis and the report of the results of this study were supported by Algorithme Pharma Inc. References Anschutz, M., Wonnemann, M., Schug, B., Toal, C., Donath, F., Pontius, A., Pauli, K., Brendel, E., Blume, H., 2010. Differences in bioavailability between 60 mg of nifedipine osmotic push-pull systems after fasted and fed administration. Int. J. Clin. Pharmacol. Ther. 48, 158–170. Chen, M.-L., Davit, B., Lionberger, R., Wahba, Z., Ahn, H.-Y., Yu, L., 2011. Using partial area for evaluation of bioavailability and bioequivalence. Pharm. Res. 28, 1939– 1947. EMA, 2010. Guideline on the Investigation of Bioequivalence (CPMP/EWP/QWP/ 1401/98 Rev.1/Corr⁄⁄). (Accessed 31.01.14). EMA, 2013. Guideline on the Pharmacokinetic and Clinical Evaluation of Modified Release Dosage Forms (EMA/CPMP/EWP/280/96 Corr1). (Accessed 31.01.14). EMA, 2014. Questions & Answers: Positions on Specific Questions Addressed to the Pharmacokinetics Working Party (EMA/618604/2008 Rev. 9). (Accessed 16.07.14). EMEA, 1999. Note for Guidance on Modified Release Oral and Transdermal Dosage Forms: Section II (CPMP/EWP/280/96). (Accessed 31.01.14). Endrenyi, L., Tothfalusi, L., 2010. Do regulatory bioequivalence requirements adequately reflect the therapeutic equivalence of modified-release drug products? J. Pharm. Pharm. Sci. 13, 107–113. FDA, 2002. Food-Effect Bioavailability and Fed Bioequivalence Studies. (Accessed 31.01.14). FDA, 2003. Bioavailability and Bioequivalence Studies for Orally Administered Drug Products—General Considerations: Guidance for Industry, FDA, Rockville, MD: Center for Drug Evaluation and Research. FDA, 2010. Background Information for the FDA Meeting of the Advisory Committee for Pharmaceutical Science and Clinical Pharmacology, Topic 2: Use of Partial Area Under the Curve (AUC) for Products with a Complex Pharmacokinetic (PK) Profile. Center for Drug Evaluation and Research. (Accessed 31.01.14. FDA, 2011. Guidance on Zolpidem. (Accessed 26.02.14). FDA, 2012a. Draft Guidance on Mesalamine. (Accessed 31.01.14). FDA, 2012b. Draft Guidance on Methylphenidate Hydrochloride. (Accessed 31.01.14). FDA, 2013. Bioequivalence Studies with Pharmacokinetic Endpoints for Drugs Submitted Under an ANDA: Guidance for industry (Draft Guidance). FDA, Rockville, MD: Center for Drug Evaluation and Research. Garcia-Arieta, A., 2013. Regulatory Requirements and Recommendations of the Guideline on the Pharmacokinetic and Clinical Evaluation of Modified Release Dosage Forms (EMA/CPMP/EWP/280/96 Corr1), Session III: Specific Issues for BE Assessment. EUFEPS BABP Network Open Discussion Forum. (Accessed 31.01.14. Garcia-Arieta, A., Morales-Alcelay, S., Herranz, M., de la Torre-Alvarado, J.M., Blazquez-Perez, A., Suarez-Gea, M.L., Alvarez, C., 2012. Investigation on the need of multiple dose bioequivalence studies for prolonged-release generic products. Int. J. Pharm. 423, 321–325. Health Canada, 1992. Report C: Report on Bioavailability of Oral Dosage Formulations, not in Modified Release Form, of Drugs Used for Systemic Effects, Having Complicated or Variable Pharmacokinetics, Health Canada, Ottawa, ON, Expert Advisory Committee on Bioavailability, Health Protection Branch. Health Canada, 1996. Guidance Document – Conduct and Analysis of Bioavailability and Bioequivalence Studies – Part B: Oral Modified Release Formulations. Health Canada, Ottawa, ON. Health Canada, 2012. Guidance Document – Comparative Bioavailability Standards: Formulations Used for Systemic Effects. (Accessed 31.01.14). Julious, S.A., 2004. Sample sizes for clinical trials with normal data. Stat. Med. 23, 1921–1986.
M. Boily et al. / European Journal of Pharmaceutical Sciences 66 (2015) 70–77 Lionberger, R.A., 2010. Topic 2: PK Profile Comparison for Modified Release Products. Advisory Committee for Pharmaceutical Science and Clinical Pharmacology. (Accessed 31.01.14). Midha, K.K., Hubbard, J.W., Rawson, M.J., 1996. Retrospective evaluation of relative extent of absorption by the use of partial areas under plasma concentration versus time curves in bioequivalence studies on conventional release products. Eur. J. Pharm. Sci. 4, 381–384.
77
Paixao, P., Gouveia, L.F., Morais, J.A., 2012. An alternative single dose parameter to avoid the need for steady-state studies on oral extended-release drug products. Eur. J. Pharm. Biopharm. 80, 410–417. Patterson, S., Jones, B., 2005. Dealing with Unexpected BE Challenges, Bioequivalence and Statistics in Clinical Pharmacology. Chapman & Hall/CRC, London, pp. 133–175.
