MAIN PAPER (wileyonlinelibrary.com) DOI: 10.1002/pst.1691
Published online 19 May 2015 in Wiley Online Library
Assessing switchability for biosimilar products: modelling approaches applied to children’s growth Rossella Belleli,a * Roland Fisch,a Didier Renard,a Heike Woehling,b and Sandro Gsteigera The present paper describes two statistical modelling approaches that have been developed to demonstrate switchability from the original recombinant human growth hormone (rhGH) formulation (Genotropinr / to a biosimilar product (Omnitroper / in children suffering from growth hormone deficiency. Demonstrating switchability between rhGH products is challenging because the process of growth varies with the age of the child and across children. The first modelling approach aims at predicting individual height measured at several time-points after switching to the biosimilar. The second modelling approach provides an estimate of the deviation from the overall growth rate after switching to the biosimilar, which can be regarded as an estimate of switchability. The results after applying these approaches to data from a randomized clinical trial are presented. The accuracy and precision of the predictions made using the first approach and the small deviation from switchability estimated with the second approach provide sufficient evidence to conclude that switching from Genotropinr to Omnitroper has a very small effect on growth, which is neither statistically significant nor clinically relevant. Copyright © 2015 John Wiley & Sons, Ltd. Keywords: Switchability; interchangeability; biosimilar; statistical modelling
1. INTRODUCTION
Pharmaceut. Statist. 2015, 14 341–349
a
Integrated Quantitative Sciences (IQS), Novartis Pharma AG, Basel, Switzerland
b
Hexal AG, Sandoz Biopharmaceuticals, Holzkirchen, Germany
*Correspondence to: Rossella Belleli, IQS (Integrated Quantitative Sciences) Novartis Pharma AG WSJ.27.3.25 Postfach CH-4002 Basel Switzerland E-mail: [email protected]
Copyright © 2015 John Wiley & Sons, Ltd.
341
The Biologics Price Competition and Innovation (BPCI) Act [1] issued in 2009 was the first US legislation aimed at regulating the development of biosimilar products. Biosimilarity was defined to mean ‘that the biological product is highly similar to the reference product notwithstanding minor differences in clinically inactive components’ and that ‘there are no clinically meaningful differences between the biological product and the reference product in terms of the safety, purity, and potency of the product’ [1]. It is worth underlying that biosimilarity is assessed through the totality of evidence and not based on the results of a single clinical trial. The same act introduced the concept of interchangeability, stating that for a new biological product to be accepted as interchangeable with a reference product, the sponsor has to demonstrate that ‘(i) the biological product is biosimilar to the reference product and (ii) can be expected to produce the same clinical result as the reference product in any given patient, and that the risk in terms of safety or diminished efficacy of alternating or switching between use of the biological product and the reference product is not greater than the risk of using the reference product without such alternation or switching’ [1,2]. The draft US Food and Drug Administration guidance ‘Scientific Considerations in Demonstrating Biosimilarity to a Reference Product’ was released in 2012 and finalized in April 2015 but did not clarify further the concept of interchangeability or switchability, nor did it provide clarification on the statistical methods to be employed in order to assess interchangeability or switchability [2,3]. It seems clear that interchangeability is an additional claim
to biosimilarity (at least in the USA), and that it concerns an assessment of the consequences on safety and efficacy of switching and alternating the reference and the biosimilar product in the same patients. The claim of ‘interchangeability’ implies in practice that ‘interchangeable products may be substituted for the reference product without the intervention of the prescribing healthcare provider’ [1,3]. Furthermore, the US Food and Drug Administration grants an exclusivity period for the first biological product that is demonstrated to be interchangeable [2–4]. Interchangeability therefore may represent an attractive claim for companies developing biosimilars for the US market. To our knowledge, there are no examples in the public domain of the statistical approaches used to support interchangeability. Chow et al. [5] proposed a switching index and an alternating index to assess interchangeability, based on the concept of reproducibility probability, which they then employed to calculate an overall biosimilarity index [6–8]. They also suggested study designs suitable to calculate these indexes. However, no applications of these methods have been published yet and, to our knowledge, no other statistical approaches have been proposed so far.
R. Belleli et al. In this paper, we present two statistical approaches to assess switchability (i.e. a single switch from originator to biosimilar) of recombinant human growth hormone (rhGH) products, using Omnitroper , a biosimilar for the reference product Genotropinr , as an example. The statistical methods that we employed are different, compared with those proposed by Chow et al.; in one case, we used modelling approaches for the purpose of predicting the individual height measurements after switching to the biosimilar, and in the other, we used them for the purpose of estimating the deviation from switchability. We share with Chow et al. [5,8] the same interpretation of the BPCI Act regarding the concept of switchability as ‘not only the switch from the Reference (R) to the Test (T) or “T to R” (narrow sense of switchability), but also from the Test to itself (“T to T”) and from the Reference to itself (“R to R”) (broader sense of switchability)’. In our case study, we assessed switchability using a study design employing only T to T and R to T switches. Furthermore, the BPCI Act statement that the reference and the biosimilar products are expected to produce the same clinical result as the reference product in any given patient led us to focus on individual patient responses to therapy, using statistical methods which account for intersubject differences. Our new approaches have been applied to show switchability between rhGH products. Growth hormone (GH) is commonly used to treat GH deficiency (GHD) in children [9]. Several originator rhGH products are commercially available, and two biosimilar products have been developed and approved in Europe [5,6,10,11], including Omnitroper , which is the focus of the statistical modelling approaches motivating the present paper. The data from a phase III clinical trial, which was included in the registration package submitted to demonstrate biosimilarity in
Europe, and used for the New Drug Application in the US, have been used to apply the new approaches [12–14]. In order to demonstrate switchability between rhGH biosimilar products, there are two challenges: firstly, the individual growth patterns have to be considered to assess the individual response to therapy; secondly, the comparison of the effect of two subsequent treatments on growth has to account for the fact that the rate of growth may be highly heterogeneous across individuals, and that it changes within the same individual over time. The two statistical approaches to assess switchability to Omnitroper that we developed attempted to address these challenges, accounting for the non-linearity in individual growth patterns and the heterogeneity in growth rate across children with random effects. The first approach focuses on prediction. It applies a non-linear mixed effects model to the height measured at several time-points during treatment with the originator product and predicts the individual height after the switching time-point, assuming identical effects before and after switch. The claim of switchability is then based on the level of accuracy and precision of the predictions for the R to T switch compared with the T to T switch and is evaluated through graphical presentations and summary statistics. The second approach focuses on estimation; it applies a non-linear mixed effects model for the entire set of individual height measurements (before and after the switching time-point) and estimates the difference in growth before and after the switching time-point for the R to T switch compared with the T to T switch. The paper is organized as follows: in Section 2, we introduce the case study describing the motivating example, based on a phase III study of Omnitroper [12–14]. We then explain the statistical methods in Section 3, and in Section 4, we present the
342
Figure 1. Phase III study design (adapted from Romer et al. [12,13]). A: original study design. B: Part of the study which was used for approach 1. C: part of the study which was used for approach 2.
Copyright © 2015 John Wiley & Sons, Ltd.
Pharmaceut. Statist. 2015, 14 341–349
R. Belleli et al. results of the analyses, applying the suggested methods to the case study data. The methods and results are then discussed in Section 5, leading to concluding remarks and suggestions for future applications in Section 6.
2. CASE STUDY The case study was a phase III randomized, controlled, multicentre clinical trial consisting of three parts (Figure 1A). Part 1, which comprised the first 9 months, had the objective of demonstrating that Omnitroper and Genotropinr are similar in terms of efficacy and safety. Part 2, which lasted for the following 6 months, had the objective of showing that two Omnitroper formulations (powder and solution) have similar efficacy and safety. The final part of the study (Part 3) continued for further 69 months with the objective of demonstrating long-term safety and efficacy of the Omnitrope solution. The study was submitted to the European and US health authorities for registration, and its results contributed to the approval of Omnitroper in 2006 as a biosimilar medicine in Europe [15] and for use ‘through the abbreviated pathway described by Section 505(b)(2) of the FDC Act’ in the USA [16]. Results of the study have already been presented in several publications [12–14] and quoted in a review discussing the regulatory approval of biosimilars using rhGH biosimilars as an example [17]. A total of 89 children with GHD were randomized in this study, and 86 were included in the analyses. Three patients, two in treatment group 1 and one in treatment group 2, were excluded from the analysis, because they provided only a single height measurement each. The children were 39 girls (45%) and 47 boys (55%)
and aged between 2 and 14 years (initial mean age: 8.1 years, standard deviation [SD] D 2.7 years),. Their initial height was on average three SD lower than the mean of the reference normal population for their age (height SD score [height SDS], min D 6.4, max D 1.7). Reference growth tables for each participating country were used. Summaries of age, height and height SDS by treatment and by gender are presented in Appendix 1, Tables 1 and 2, available online as Supporting Information. The children’s growth during the first 2 years of the study can be visualized by plotting their height against their age at different visits, as shown in Figure 2. These individual profiles correspond to the growth curves commonly drawn by pediatricians to monitor children’s growth. An alternative view of the same data is shown in Figure 3, where height is plotted against the time from the start of the study. This latter representation of the data helps to reveal the shape of the individual profiles; most children show an approximately linear growth in this relatively short period of time, although some of them, in particular, the smallest (i.e. youngest) ones present signs of non-linear growth over this period. Also, the main source of variability seems to derive from the initial height of the children, which is a function of age. Using the representation of height versus time from the start of the study, the differences in the individual growth rate over time appear to be less remarkable. The individual profiles of height SDS versus age and time from the start of the study (Appendix 1, Figures 1 and 2, available online as Supporting Information) showed more heterogeneity across children and more non-linearity with time. Therefore, we decided to use height and not height SDS to measure children’s growth over time in order to assess swichability from Genotropinr to Omnitroper .
Pharmaceut. Statist. 2015, 14 341–349
343
Figure 2. Individual profiles of height versus age in the case study.
Copyright © 2015 John Wiley & Sons, Ltd.
R. Belleli et al.
Figure 3. Individual profiles of height versus time from the start of the study.
Table I. Results from the analysis of the case study using approach 1 (height prediction): summary statistics of the differences between the predicted and observed height values nine months after the switch. Treatment group
Period %
Published online 19 May 2015 in Wiley Online Library
Assessing switchability for biosimilar products: modelling approaches applied to children’s growth Rossella Belleli,a * Roland Fisch,a Didier Renard,a Heike Woehling,b and Sandro Gsteigera The present paper describes two statistical modelling approaches that have been developed to demonstrate switchability from the original recombinant human growth hormone (rhGH) formulation (Genotropinr / to a biosimilar product (Omnitroper / in children suffering from growth hormone deficiency. Demonstrating switchability between rhGH products is challenging because the process of growth varies with the age of the child and across children. The first modelling approach aims at predicting individual height measured at several time-points after switching to the biosimilar. The second modelling approach provides an estimate of the deviation from the overall growth rate after switching to the biosimilar, which can be regarded as an estimate of switchability. The results after applying these approaches to data from a randomized clinical trial are presented. The accuracy and precision of the predictions made using the first approach and the small deviation from switchability estimated with the second approach provide sufficient evidence to conclude that switching from Genotropinr to Omnitroper has a very small effect on growth, which is neither statistically significant nor clinically relevant. Copyright © 2015 John Wiley & Sons, Ltd. Keywords: Switchability; interchangeability; biosimilar; statistical modelling
1. INTRODUCTION
Pharmaceut. Statist. 2015, 14 341–349
a
Integrated Quantitative Sciences (IQS), Novartis Pharma AG, Basel, Switzerland
b
Hexal AG, Sandoz Biopharmaceuticals, Holzkirchen, Germany
*Correspondence to: Rossella Belleli, IQS (Integrated Quantitative Sciences) Novartis Pharma AG WSJ.27.3.25 Postfach CH-4002 Basel Switzerland E-mail: [email protected]
Copyright © 2015 John Wiley & Sons, Ltd.
341
The Biologics Price Competition and Innovation (BPCI) Act [1] issued in 2009 was the first US legislation aimed at regulating the development of biosimilar products. Biosimilarity was defined to mean ‘that the biological product is highly similar to the reference product notwithstanding minor differences in clinically inactive components’ and that ‘there are no clinically meaningful differences between the biological product and the reference product in terms of the safety, purity, and potency of the product’ [1]. It is worth underlying that biosimilarity is assessed through the totality of evidence and not based on the results of a single clinical trial. The same act introduced the concept of interchangeability, stating that for a new biological product to be accepted as interchangeable with a reference product, the sponsor has to demonstrate that ‘(i) the biological product is biosimilar to the reference product and (ii) can be expected to produce the same clinical result as the reference product in any given patient, and that the risk in terms of safety or diminished efficacy of alternating or switching between use of the biological product and the reference product is not greater than the risk of using the reference product without such alternation or switching’ [1,2]. The draft US Food and Drug Administration guidance ‘Scientific Considerations in Demonstrating Biosimilarity to a Reference Product’ was released in 2012 and finalized in April 2015 but did not clarify further the concept of interchangeability or switchability, nor did it provide clarification on the statistical methods to be employed in order to assess interchangeability or switchability [2,3]. It seems clear that interchangeability is an additional claim
to biosimilarity (at least in the USA), and that it concerns an assessment of the consequences on safety and efficacy of switching and alternating the reference and the biosimilar product in the same patients. The claim of ‘interchangeability’ implies in practice that ‘interchangeable products may be substituted for the reference product without the intervention of the prescribing healthcare provider’ [1,3]. Furthermore, the US Food and Drug Administration grants an exclusivity period for the first biological product that is demonstrated to be interchangeable [2–4]. Interchangeability therefore may represent an attractive claim for companies developing biosimilars for the US market. To our knowledge, there are no examples in the public domain of the statistical approaches used to support interchangeability. Chow et al. [5] proposed a switching index and an alternating index to assess interchangeability, based on the concept of reproducibility probability, which they then employed to calculate an overall biosimilarity index [6–8]. They also suggested study designs suitable to calculate these indexes. However, no applications of these methods have been published yet and, to our knowledge, no other statistical approaches have been proposed so far.
R. Belleli et al. In this paper, we present two statistical approaches to assess switchability (i.e. a single switch from originator to biosimilar) of recombinant human growth hormone (rhGH) products, using Omnitroper , a biosimilar for the reference product Genotropinr , as an example. The statistical methods that we employed are different, compared with those proposed by Chow et al.; in one case, we used modelling approaches for the purpose of predicting the individual height measurements after switching to the biosimilar, and in the other, we used them for the purpose of estimating the deviation from switchability. We share with Chow et al. [5,8] the same interpretation of the BPCI Act regarding the concept of switchability as ‘not only the switch from the Reference (R) to the Test (T) or “T to R” (narrow sense of switchability), but also from the Test to itself (“T to T”) and from the Reference to itself (“R to R”) (broader sense of switchability)’. In our case study, we assessed switchability using a study design employing only T to T and R to T switches. Furthermore, the BPCI Act statement that the reference and the biosimilar products are expected to produce the same clinical result as the reference product in any given patient led us to focus on individual patient responses to therapy, using statistical methods which account for intersubject differences. Our new approaches have been applied to show switchability between rhGH products. Growth hormone (GH) is commonly used to treat GH deficiency (GHD) in children [9]. Several originator rhGH products are commercially available, and two biosimilar products have been developed and approved in Europe [5,6,10,11], including Omnitroper , which is the focus of the statistical modelling approaches motivating the present paper. The data from a phase III clinical trial, which was included in the registration package submitted to demonstrate biosimilarity in
Europe, and used for the New Drug Application in the US, have been used to apply the new approaches [12–14]. In order to demonstrate switchability between rhGH biosimilar products, there are two challenges: firstly, the individual growth patterns have to be considered to assess the individual response to therapy; secondly, the comparison of the effect of two subsequent treatments on growth has to account for the fact that the rate of growth may be highly heterogeneous across individuals, and that it changes within the same individual over time. The two statistical approaches to assess switchability to Omnitroper that we developed attempted to address these challenges, accounting for the non-linearity in individual growth patterns and the heterogeneity in growth rate across children with random effects. The first approach focuses on prediction. It applies a non-linear mixed effects model to the height measured at several time-points during treatment with the originator product and predicts the individual height after the switching time-point, assuming identical effects before and after switch. The claim of switchability is then based on the level of accuracy and precision of the predictions for the R to T switch compared with the T to T switch and is evaluated through graphical presentations and summary statistics. The second approach focuses on estimation; it applies a non-linear mixed effects model for the entire set of individual height measurements (before and after the switching time-point) and estimates the difference in growth before and after the switching time-point for the R to T switch compared with the T to T switch. The paper is organized as follows: in Section 2, we introduce the case study describing the motivating example, based on a phase III study of Omnitroper [12–14]. We then explain the statistical methods in Section 3, and in Section 4, we present the
342
Figure 1. Phase III study design (adapted from Romer et al. [12,13]). A: original study design. B: Part of the study which was used for approach 1. C: part of the study which was used for approach 2.
Copyright © 2015 John Wiley & Sons, Ltd.
Pharmaceut. Statist. 2015, 14 341–349
R. Belleli et al. results of the analyses, applying the suggested methods to the case study data. The methods and results are then discussed in Section 5, leading to concluding remarks and suggestions for future applications in Section 6.
2. CASE STUDY The case study was a phase III randomized, controlled, multicentre clinical trial consisting of three parts (Figure 1A). Part 1, which comprised the first 9 months, had the objective of demonstrating that Omnitroper and Genotropinr are similar in terms of efficacy and safety. Part 2, which lasted for the following 6 months, had the objective of showing that two Omnitroper formulations (powder and solution) have similar efficacy and safety. The final part of the study (Part 3) continued for further 69 months with the objective of demonstrating long-term safety and efficacy of the Omnitrope solution. The study was submitted to the European and US health authorities for registration, and its results contributed to the approval of Omnitroper in 2006 as a biosimilar medicine in Europe [15] and for use ‘through the abbreviated pathway described by Section 505(b)(2) of the FDC Act’ in the USA [16]. Results of the study have already been presented in several publications [12–14] and quoted in a review discussing the regulatory approval of biosimilars using rhGH biosimilars as an example [17]. A total of 89 children with GHD were randomized in this study, and 86 were included in the analyses. Three patients, two in treatment group 1 and one in treatment group 2, were excluded from the analysis, because they provided only a single height measurement each. The children were 39 girls (45%) and 47 boys (55%)
and aged between 2 and 14 years (initial mean age: 8.1 years, standard deviation [SD] D 2.7 years),. Their initial height was on average three SD lower than the mean of the reference normal population for their age (height SD score [height SDS], min D 6.4, max D 1.7). Reference growth tables for each participating country were used. Summaries of age, height and height SDS by treatment and by gender are presented in Appendix 1, Tables 1 and 2, available online as Supporting Information. The children’s growth during the first 2 years of the study can be visualized by plotting their height against their age at different visits, as shown in Figure 2. These individual profiles correspond to the growth curves commonly drawn by pediatricians to monitor children’s growth. An alternative view of the same data is shown in Figure 3, where height is plotted against the time from the start of the study. This latter representation of the data helps to reveal the shape of the individual profiles; most children show an approximately linear growth in this relatively short period of time, although some of them, in particular, the smallest (i.e. youngest) ones present signs of non-linear growth over this period. Also, the main source of variability seems to derive from the initial height of the children, which is a function of age. Using the representation of height versus time from the start of the study, the differences in the individual growth rate over time appear to be less remarkable. The individual profiles of height SDS versus age and time from the start of the study (Appendix 1, Figures 1 and 2, available online as Supporting Information) showed more heterogeneity across children and more non-linearity with time. Therefore, we decided to use height and not height SDS to measure children’s growth over time in order to assess swichability from Genotropinr to Omnitroper .
Pharmaceut. Statist. 2015, 14 341–349
343
Figure 2. Individual profiles of height versus age in the case study.
Copyright © 2015 John Wiley & Sons, Ltd.
R. Belleli et al.
Figure 3. Individual profiles of height versus time from the start of the study.
Table I. Results from the analysis of the case study using approach 1 (height prediction): summary statistics of the differences between the predicted and observed height values nine months after the switch. Treatment group
Period %
