Adjusted indirect comparisons to assess bioequivalence between generic clopidogrel products in Serbia

Abstract

Aims

Generic products can be regarded as therapeutically equivalent and switchable with the reference product. However, switchability between generics is unknown, as direct comparisons between generics are not performed. The aim of this study was to investigate the bioequivalence between generic clopidogrel products by means of adjusted indirect comparisons (AICs).

Methods

AICs were conducted to assess bioequivalence between 4 generic clopidogrel products that are authorised in Serbia. Generics are considered equivalent to the reference if the 90% confidence intervals (CIs) for the ratios test/reference of the maximum concentration (C_max) and area under the curve up to the last measurable concentration (AUC_0–t) fall within the acceptance range 80.00–125.00%. However, for AICs between generics, the Canadian acceptance criterion for C_max was employed, where only the point estimate of C_max needs to be within 80.00–125.00%.

Results

The 90% CIs of the AICs demonstrated bioequivalence within 80.00–125.00% for all AUC_0–t comparisons. The point estimates of C_max in all AICs were also within this range.

Conclusion

This study demonstrates that the bioavailability of these 4 generic clopidogrel products authorised in Serbia is very similar. Despite the limited power of AICs, bioequivalence was demonstrated for all 90% CIs of AUC_0–t and all 90% CIs of C_max comparisons were within or very close to the acceptance range, being able to comply with the acceptance criterion employed in Canada for C_max. Therefore, these 4 generic clopidogrel products authorised in Serbia can be considered switchable with each other in clinical practice based on the adjusted indirect comparisons.

What is already known about this subject

• Clopidogrel generics are bioequivalent and switchable with the reference product, but not necessarily with other clopidogrel generics.

• Enhanced or reduced clopidogrel bioavailability in different generic products might cause potentially serious clinical consequences if generics are switched between them.

• Adjusted indirect comparisons are suitable to assess bioequivalence between generics.

What this study adds

• This is the first time that adjusted indirect comparisons between clopidogrel generics are performed.

• The bioavailability of the investigated clopidogrel generics is very similar.

• The tested generic clopidogrel products can be considered switchable with each other in clinical practice.

1. INTRODUCTION

Generic products need to prove bioequivalence with the innovator (reference) product to obtain a marketing authorisation. Often, several generic products, all bioequivalent to the same reference product, are available in the same market. Each of them could be regarded as therapeutically equivalent and switchable with the reference product, but not necessarily with each other, since the generic products have not been mutually compared and bioequivalence between them has not been proven. When switching from 1 generic to another, a so‐called generic shift or drift may appear, meaning that generic formulations that are bioequivalent to the innovator drug may not be bioequivalent to each other, leading to possible change in the therapeutic and/or adverse effects.1, 2

In practice, patients may be switched from 1 generic product to another, although for clinicians and pharmacists it is difficult to assume whether 1 generic product could be safely substituted with another. However, regulators strictly assess the generics with an acceptance range of 20%, i.e. 80.00–125.00%,3, 4, 5 even when the generic companies justify that larger differences are clinically irrelevant, in order to avoid large differences between generics and to ensure that generics will be switchable between themselves.

Clopidogrel, a platelet aggregation inhibitor, is a prodrug that requires oxidation, first to its intermediate metabolite 2‐oxo‐clopidogrel, and further to the pharmacologically active thiol metabolite.6 An additional inactive metabolite, clopidogrel carboxylic acid, represents the main circulating metabolite formed by parallel pathway with plasma levels up to 2000‐fold higher than those of clopidogrel. Previously, clopidogrel carboxylic acid measurements were acceptable for the establishment of the clopidogrel bioequivalence as bioanalytical methodology for the reliable determination of the plasma clopidogrel concentrations were lacking.7 However, current scientific guidelines recommend the use of clopidogrel concentrations for bioequivalence studies since the parent drug is the most sensitive analyte to detect differences in extent and/or rate of absorption between formulations.7, 8

Clopidogrel was chosen for this investigation as the use of generic products have been claimed to cause acute and subacute stent thrombosis9 and reduction in ADP‐induced platelet aggregation.10 However, the validity of these conclusions might be questioned because the first study was a case–control study where 3 historical controls with a defined reason for stent thrombosis were excluded to reach statistical significance and a stent thrombosis‐predictive algorithm revealed no difference in the likelihood of stent thrombosis between patients receiving generic clopidogrel and historic controls.9 In the second study, the reference product was administered after the generic products and the results may be biased by the existence of a period effect.10 Importantly, the daily maintenance dose is 75 mg irrespective of CYP2C19 metabolic status, sex or body weight, and large intersubject variability has been described in the pharmacokinetic parameters of the reference product11 without the need of drug monitoring, which would be complicated also by its high intrasubject variability. The high intersubject variability in pharmacokinetic parameters is responsible for the patient‐specific and variable response to clopidogrel between patients after the treatment with the reference product.12, 13, 14 Furthermore, the reference product can be taken with or without food, although a high‐fat meal increased clopidogrel C_max and area under the curve (AUC) extrapolated to infinity by 6.1‐fold and 9.2‐fold, respectively.15 For all these reasons, the European Medicine Agency does not consider clopidogrel to be a narrow therapeutic index drug, but it is considered a highly variable drug where the widening of the acceptance range is not possible because wider differences in maximum plasma concentrations are clinically relevant.7

Direct comparisons between generics are not performed as there is no regulatory requirement for conducting these studies since they are not practical.16 Therefore, the adjusted indirect comparisons (AICs) are the most suitable method to assess bioequivalence between generics.16, 17, 18, 19, 20, 21 To our knowledge, generic products of clopidogrel have not been compared between themselves, neither indirectly nor directly. Hence, the aim of our study was to investigate the bioequivalence between generic clopidogrel products of the Serbian market, whose bioequivalence data are available in the public domain, using the method of AICs.

2. METHODS

2.1. Generic products and the original bioequivalence studies

Four generic clopidogrel products authorised in Serbia were considered for this analysis because the data of the bioequivalence studies are available in the Public Assessment Reports issued in the European Union.22, 23, 24, 25 In all these studies, the same reference product (Plavix, Sanofi) was administered at a single dose of 75 mg with the same 2×2 cross‐over design in healthy volunteers in the fasting state. The plasma concentration of the parent drug was measured in all studies, while in 3 out of 4 studies, clopidogrel carboxylic acid was measured as supportive data.

The generic products from the corresponding bioequivalence studies included in this analysis are given in Table 1.

Table 1.

Generic products included in the analysis

	Name in the EU	EU procedure number (reference)	Name in Serbia
Generic 1	Agregex 75 mg film‐coated tablets (Actavis)	DK/H/1624/001/DC22	Antiagrex
Generic 2	Clopidogrel Teva 75 mg film‐coated tablets (Teva)	EMEA/H/C/00105324	Angiclod
Generic 3	Perclod 75 mg film‐coated tablets (Pharmathen)	DK/H/1601/001/DC23	Clopidogrel Kwizda
Generic 4	Zyllt 75 mg film‐coated tablets (Krka)	EMEA/H/C/105825	Zyllt

The results of the original bioequivalence studies for each generic product are given in Table 2. In all cases, the 90% confidence interval (CI) for the ratio test/reference of the maximum concentration (C_max) and area under the curve up to the last measurable concentration (AUC_0–t) was within the acceptance range of 80.00–125.00%, which is the conventional acceptance criterion for bioequivalence in the European Union.

Table 2.

Results of the bioequivalence studies submitted for marketing authorisation for clopidogrel generic products22, 23, 24, 25

	n	Cmax	AUC0–t	Cmax CV	AUC0–t CV
Generic 1	53	103.42 (92.29–115.89)	96.30 (87.98–105.42)	35.87%	28.16%
Generic 2	97	97.22 (87.43–105.10)	97.44 (88.72–107.02)	47.00%	41.00%
Generic 3	62	97.00 (86.00–110.00)	96.00 (88.00–105.00)	44.10%	30.90%
Generic 4	91	102.00 (89.00–117.00)	95.00 (84.00–107.00)	58.80%	52.20%

C_max: maximum concentration; AUC_0–t: area under the curve up to the last measurable concentration.

CV: intrasubject coefficient of variation; n: sample size of the study.

2.2. Adjusted indirect comparison between generic products

Clopidogrel generic products were compared using the method of AICs that uses the results from the individual bioequivalence studies for each generic product vs the reference product. The comparison is performed between the pairs of generic products in relative terms since the reference represents the 100% value in all bioequivalence studies, which adjusts the study results.16, 17, 18, 19, 20 In this study, for each pair of generic products, adjusted indirect comparison is performed in 1 direction, e.g. G1 vs G2, and the inverse comparison, i.e. G2 vs G1, can be deduced simply by calculation the inverse (e.g. 1.25 is the inverse of 0.8).

The point estimate of adjusted indirect comparison (adjPE _G1,G2) is calculated from the difference between the point estimates of 2 bioequivalence studies. As all bioequivalence study results are expressed as the percentage of the ratio test/reference, the interstudy comparisons are always adjusted. The point estimate of a bioequivalence study is the difference between generic 1 and reference product in the log scale (d _G1–R  = G1 − R). Therefore, the difference between the point estimates of 2 bioequivalence studies gives the adjusted difference between generic 1 and generic 2 in the log scale [d _G1−G2 = d _G1–R  − d _G2–R  = (G1 − R) − (G2 − R) = G1–G2], where G1 and G2 are the pharmacokinetic parameters of interest of generic products 1 and 2, respectively.17, 26

The width of the 90% CIs of the AICs was calculated using 3 methods: assuming homogeneous variance (homoscedastic method), heterogeneous variances (heteroscedastic method) and assuming heterogeneous variances but approximate degrees of freedom (pragmatic approach) as described elsewhere.17, 19

Homoscedastic method: This method assumes small sample sizes with homogeneous variances whose difference follows Student t‐test distribution (t_0.9,d.f.), where the variability of the pair of studies under comparison is weighted to obtain a common estimation of variability ${\mathit{SD}}_{\mathit{\text{pooled}}}^2=\frac{\left(n_1-1\right){\mathit{SD}}_1^2+\left(n_2-1\right){\mathit{SD}}_2^2}{n_1+n_2-2}$, where the degrees of freedom are (n₁ + n₂–2), where n₁ and n₂ are the number of subjects in study 1 for generic 1 and in study 2 for generic 2, respectively, SD is the standard deviation, and the subscript 1 or 2 indicates the study number. The SD of each study is calculated by $\mathit{SD}=\frac{2·{\mathit{SE}}_{\left(d\right)}}{\sqrt{\frac{1}{n_1}+\frac{1}{n_2}}}$, where n₁ and n₂ represents the number of subjects in sequence 1 and sequence 2, respectively, since these are direct comparisons with cross‐over designs. Once the SD _pooled is calculated, the SE _d for the indirect comparison is calculated by ${\mathit{SE}}_{\left(d\right)}={\mathit{SD}}_{\mathit{\text{pooled}}}\sqrt{\frac{1}{n_1}+\frac{1}{n_2}}$, where the subscript 1 or 2 indicates the study number, since it is no longer a crossover comparison, but an interstudy indirect comparison. Heteroscedastic method: This method assumes small samples sizes (t_0.9,d.f.) with heterogeneous variances which relaxes the underlying assumptions for the variance estimation (SE _d ²  = SE ₁ ²  + SE ₂ ²). It must be noted that a test that fails to reject the null hypothesis of homogeneous variances should not be considered as proof of homogenous variances, especially when the number of studies under comparison is low. However, the heteroscedastic approach requires the calculation of the degrees of freedom according to Welch,27 $d.f.=\frac{{\left(\frac{{\mathit{SD}}_1^2}{n_1}+\frac{{\mathit{SD}}_2^2}{n_2}\right)}^2}{\frac{{\left(\frac{{\mathit{SD}}_1^2}{n_1}\right)}^2}{n_1-1}+\frac{{\left(\frac{{\mathit{SD}}_2^2}{n_2}\right)}^2}{n_2-1}}$, in order to solve the Behrens–Fisher problem.

Pragmatic method: This method does not require the assumption of homogeneity of variances (SE _d ²  = SE ₁ ²  + SE ₂ ²), since it is unlikely verifiable, between studies with small sample sizes that follow Student t‐test distribution (t_0.9,d.f.), whose degrees of freedom are approximated for simplicity as if the variances were homogeneous (n₁ + n₂–2).

Data analysis was done using Microsoft Office Excel 2016. The generic products were considered bioequivalent if the 90% CI for the ratio test/reference of AUC_0–t obtained by adjusted indirect comparison fall within the conventional acceptance range of 80.00–125.00%. However, for the AICs of C_max the Canadian acceptance criterion for C_max was employed, where only the point estimate of C_max needs to be within 80.00–125.00%. This permissive criterion was employed because C_max of clopidogrel is highly variable and AICs lack enough power to conclude equivalence since the original studies were not powered to conduct AICs, but simply to show equivalence with the reference product.16

2.3. Nomenclature of targets and ligands

Key protein targets and ligands in this article are hyperlinked to corresponding entries in http://www.guidetopharmacology.org, the common portal for data from the IUPHAR/BPS Guide to PHARMACOLOGY.28

3. RESULTS

The results of the 6 AICs using previously described methods for the pharmacokinetic parameters C_max and AUC_0–t for clopidogrel are shown in Table 3. Figure 1 represents the comparisons between all original bioequivalence studies and AICs using the homoscedastic method.

Table 3.

Point estimates for C_max and AUC_0–t of clopidogrel calculated by different methods of AIC of clopidogrel generic products

	Method	Cmax, PE (90%CI)	AUC0–t, PE (90%CI)
G1 vs G2	A	106.38 (90.15–125.53)	98.83 (85.69–113.99)
	B	106.38 (91.15–124.15	98.83 (86.85–112.48)
	C	106.38 (91.16–124.14)	98.83 (86.65–112.47)
G1 vs G3	A	106.33 (89.86–125.82)	100.19 (88.34–113.63)
	B	106.33 (90.05–125.56)	100.19 (88.39–113.56)
	C	106.33 (90.05–125.56)	100.19 (88.39–113.56)
G1 vs G4	A	101.35 (83.12–123.57)	101.58 (85.52–120.67)
	B	101.35 (84.93–120.93)	101.58 (87.43–118.03)
	C	101.35 (84.93–120.93)	101.58 (87.43–118.03)
G2 vs G3	A	99.95 (84.81–117.80)	101.37 (88.46–116.16)
	B	99.95 (85.05–117.47)	101.37 (89.19–115.21)
	C	99.95 (85.06–117.45)	101.37 (89.19–115.21)
G2 vs G4	A	104.97 (88.47–124.54)	97.29 (83.64–113.18)
	B	104.97 (88.36–124.70)	97.29 (83.55–113.30)
	C	104.97 (88.36–124.69)	97.29 (83.55–113.30)
G3 vs G4	A	104.92 (86.47–127.29)	98.63 (83.80–116.08)
	B	104.92 (87.39–125.96)	98.63 (84.98–114.46)
	C	104.92 (87.39–125.96)	98.63 (84.98–114.46)

C_max: maximum concentration; AUC_0–t: area under the curve up to the last measurable concentration; AIC: adjusted indirect comparison; PE: point estimate; bold: results deviating from the 80.00–125.00% range.

Methods: A: homoscedastic; B: heteroscedastic; C: pragmatic.

Figure 1.

Point estimate (PE) and 90% confidence interval (CI) of: maximum concentration (C_max; A, B) and area under the curve up to the last measurable concentration (AUC_0–t; C, D) of each indirect comparison (A, C) and original bioequivalence study (B, D)

The 90% CI of the ratio between pairs of generics for AUC_0–t was within the range 80.00–125.00% in all AICs. The point estimates of the C_max ratios in all these AICs were also within this range. Moreover, the 90% CIs for the C_max ratios between pairs of generics were within the 80.00–125.00% for 3 pairs of generics (G1/G4, G2/G3, G2/G4), and were slightly outside this range for G1/G2 (125.53%), G1/G3 (125.82%) and G3/G4 (127.29%).

4. DISCUSSION

This study demonstrates that the bioavailability of these 4 generic clopidogrel products authorised in Serbia is very similar. Despite the limited power of the AICs, bioequivalence was demonstrated for the 90% CIs of AUC_0–t in all comparisons. Similarly, the 90% CIs of all C_max comparisons were within or very close to the acceptance range, being able to comply with the acceptance criterion employed in Canada for C_max. 5 Therefore, these 4 generic clopidogrel products authorised in Serbia can be considered switchable with each other in clinical practice.

AICs have been adopted as a useful tool for assessing bioequivalence between generic products and their switchability,16, 17, 18, 19, 20, 21 as well as for comparing originator products from different markets that have previously been shown to be bioequivalent with the same generic product.29 To our knowledge, this is the first study to compare clopidogrel generic products. AICs method utilises the publicly available summarized results from bioequivalence studies and, consequently, it is the simplest and the most appropriate method when only 2 interventions are to be compared indirectly.30 The homoscedastic, heteroscedastic and pragmatic approaches have been used to show that the conclusions are similar irrespective of the method (Table 3). Indeed, similar results were obtained using the 3 methods, with the results of the homoscedastic method being slightly worse for the conclusion of bioequivalence and therefore the most conservative. If a method based on the Z‐distribution had been used, the 90% CIs would have been narrower and biased in favour of the bioequivalence conclusion, since the sample size of bioequivalence studies is not large.

The most important limitation of these AICs is the imprecision of the estimations since the variability of the original bioequivalence studies is combined to obtain a pooled estimation for the calculation of the 90% CI of the AICs between generics. Despite this limitation, the results of this study are reassuring since the 90% CI of AUC were within 80.00–125.00% in all cases and those of C_max deviated only marginally in 3 cases. These minor deviations cannot be considered a public health concern since C_max is considered less clinically relevant than AUC. In Canada, the acceptance criterion for the C_max of clopidogrel generics when compared directly with the reference is based on the point estimate only.5 We do not propose the use of the Canadian criterion for the comparison between generics and the reference product because it is necessary to be strict in the assessment of the direct comparisons between generics and the reference product in order to ensure that generics would not be notably different between themselves. However, in our opinion it is a reasonable approach for indirect comparisons since C_max is usually more variable than AUC and the AICs for C_max are notably underpowered. Similarly, other authors have proposed to use a widened acceptance range for the 90% CI of C_max due to this limitation.16 Since the AICs are less precise than the direct comparisons, the 90% CIs obtained in indirect comparison tend to be wider than those that would have been obtained if an in vivo comparison between generics had been performed.31, 32 The results of the AICs for C_max of this study would be included in any widened acceptance range (e.g. 75.00–133.33%) since the deviations are only marginal (127.29% in the worst case of the homoscedastic method). It is unrealistic to expect that the AICs for C_max would be able to comply with conventional 80.00–125.00% acceptance range, especially in the case of a highly variable drug such as clopidogrel and taking into account that the sample size of the bioequivalence studies was calculated to show equivalence in the direct comparisons and not in the indirect ones. Another limitation of the present study is that bioequivalence between these 4 generic clopidogrel products was demonstrated only in fasted state in accordance with the bioequivalence requirements of the European Union, in contrast to the requirements of the US Food and Drug Administration, where bioequivalence has to be demonstrated in fasted and fed state for products that can be taken with or without meals.8 In the European Union, it is assumed that the bioequivalence study in the fasted state is the most discriminative to detect differences between formulations. Consequently, study in the fed state is not required for the approval of generic products.3

The validity of AICs is based on the assumption that the variation in the observed results for subjects treated with the common reference will account for differences between studies in terms of methodology (e.g. bioanalytical methods, sampling times), subjects' baseline characteristics (e.g. metabolic status, race, sex, dose).31, 33 This assumption is reasonable in our investigation since the studies estimate the same effect, and there is no important difference in the design of the bioequivalence studies under comparison in aspects that could bias the formulation effect. All studies had the same design, i.e. single dose cross‐over study in fasted state, and the studies had the same objective and outcome measurements, i.e. to demonstrate bioequivalence in the same pharmacokinetic parameters, C_max and AUC, of the parent drug clopidogrel in plasma. Furthermore, it is recognised worldwide that differences between formulations in bioequivalence studies do not depend on factors such as sex or race of the healthy volunteers. It is also taken for granted that there is no subject‐by‐formulation interaction or, in other words, the magnitude of the formulation differences does not depend on the covariates defining subgroups of patients.16

Some limitations that might affect AICs of efficacy trials do not affect those of bioequivalence studies because these studies are conducted in healthy volunteers with a cross‐over design. Therefore, differences in disease state are not a concern since bioequivalence studies usually recruit healthy volunteers and this method controls for differences in baseline characteristics between treatment groups.33

Generic substitution has received close debate and review lately without consensus. Consequently, policies among European countries are diverse. Generally, generics are allowed to be switched during dispensing in the pharmacy, with the exception of some therapeutic groups, such as antiepileptic and immunosuppressant drugs.34, 35 In Serbia, the substitution policy is very cautious because medicines are prescribed by their brand names and generic substitution during dispensing is allowed only if the prescribed medicine is not available, provided that the patient agrees to the substitution.36 In other countries, the substitution is automatic for the cheapest generic without the need of the patient's consent. To rely on safe switchability between generics AICs between generics is a useful tool able to reassure that the substitution is not a potential risk for the public health. Enhanced or reduced clopidogrel bioavailability in different generic products may cause potentially serious clinical consequences leading to bleeding in case of enhanced bioavailability or inadequate prevention of atherothrombotic and thromboembolic events if the bioavailability is reduced.6 These concerns can be discarded based on the results of the present study.

5. CONCLUSION

The bioavailability of these 4 generic clopidogrel products authorised in Serbia is very similar. Despite the limited power of AICs, bioequivalence was demonstrated for the 90% CIs of AUC_0–t in all comparisons. The 90% CIs of all C_max comparisons were within or very close to the conventional acceptance range and these comparisons were able to comply with the acceptance criterion employed in Canada for C_max. Therefore, these 4 generic clopidogrel products can be considered switchable with each other in clinical practice based on the AICs.

COMPETING INTERESTS

There are no competing interests to declare.

CONTRIBUTORS

Z.P. performed the analysis, interpreted the results, wrote the manuscript; K.V. performed the analysis, wrote the manuscript; A.G.A. interpreted the results, revised the manuscript; B.M. interpreted the results and revised the manuscript.

Pejčić Z, Vučićević K, García‐Arieta A, Miljković B. Adjusted indirect comparisons to assess bioequivalence between generic clopidogrel products in Serbia. Br J Clin Pharmacol. 2019;85:2059–2065. 10.1111/bcp.13997

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available in The Heads of Medicines Agencies and European Medicine Agency at http://www.hma.eu/ and https://www.ema.europa.eu/, reference number DK/H/1624/001/DC, DK/H/1601/001/DC, EMEA/H/C/001053 and EMEA/H/C/1058. These data were derived from the following resources available in the public domain: http://www.hma.eu/fileadmin/dateien/pipar/dk1624/parmod5_dk1624agregex.pdf, http://www.hma.eu/fileadmin/dateien/pipar/dk1601/parmod5_dk1601perclod.pdf, https://www.ema.europa.eu/documents/assessment‐report/clopidogrel‐teva‐hydrogen‐sulphate‐epar‐public‐assessment‐report_en.pdf, and http://www.ema.europa.eu/docs/en_GB/document_library/EPAR_‐_Public_assessment_report/human/001058/WC500054069.pdf