Drug Development and Potential Regulatory Paths for Insulin Biosimilars

Abstract

Under the Biologics Price Competition and Innovation Act (BPCI Act), a biological product may be demonstrated to be “biosimilar” if data show that, among other things, the product is “highly similar” to an already-approved biological product. Biosimilar insulins have the potential to reduce ever growing costs associated with insulin treatment by allowing competition. In this article, we describe the current drug development and regulatory paths for biosimilar insulins. Most likely basis of market approval for biosimilar insulins by the US Food and Drug Administration (FDA) and guidance for developing insulin biosimilars by European Medicines Agency (EMA) are discussed in detail. Currently, no product specific biosimilar FDA guidance for insulin biosimilarity assessment exists. We propose efficient and cost-effective drug development and potential regulatory paths based on scientific justification. In addition, novel trial designs for demonstrating interchangeability between the biosimilar and the reference insulin products are presented.

Section 351(i) of the Public Health Service (PHS) Act defines biosimilarity as “that the biological product is highly similar to the reference product not withstanding 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 Insulin biosimilar will enable market competition. According to a recent survey, global insulin sales touched $16.7 billion in 2011 ² and are expected to reach $32 billion in 2018. ³ Biosimilar insulins have the potential to reduce ever growing costs associated with insulin treatment. Structurally, even though, a biosimilar will have the same amino acid sequence as its reference insulin product, it may differ in its efficacy and safety profiles and/or immunogenicity. Whether such a difference is meaningful or not is a regulatory concern.

In this article, we propose the drug development and regulatory paths for biosimilar insulins. Most likely basis of market approval for biosimilar insulins by the US Food and Drug Administration (FDA) and guidance on insulin biosimilar development by European Medicines Agency (EMA) are discussed in detail. We also describe alternative drug development paths based on scientific justification. In addition novel trial designs for demonstrating interchangeability between the biosimilar and the reference insulin products are also presented.

Drug Development and Current Regulatory Paths for Insulin Biosimilars

Given the information provided by the FDA in the public domain, basis of approval through 505(b)(2) pathway is mostly likely to be based on extensive physicochemical characterization, PKPD clamp studies and demonstration of efficacy based on HbA1c criteria and acceptable immunogenicity potential as that of the reference insulin product which is approved for use in the United States. The primary goal for addition of section 505(b)(2) to the Federal Food Drug and Cosmetics Act in 1984 was to avoid unnecessary duplication of preclinical and certain human studies. ⁴

The European medical agency has provided product specific guidelines for developing insulin biosimilars. The Marketing Authorisation (MA) application for biosimilar insulin should contain in vitro characterization, clamp studies, and 12 month safety (immunogenicity) data. The PK goal is to demonstrate that the 90% confidence interval of the ratio test/reference of the mean area under the curve (AUC) and maximum concentration (C_max) should lie within 80% to 125%. The primary glycemic control comparison is based on the AUC of the glucose-infusion rate over time (GIR_AUC) and maximum GIR (GIR_max). The time to reach maximum GIR (TGIR_max) and time to reach half-maximal GIR (T_GIR50%) should be used as secondary PD endpoints. Table 1 provides information comparing the basis for approval and types of studies required for respective agencies. Listed below are the key steps with regard to drug development program for an insulin biosimilar that are adapted from FDA and EMA guidelines to demonstrate biosimilarity.^5,6

Table 1.

Basis for Approval and Types of Studies, Clinical Endpoints Required for Respective Agencies.

Category	FDA	EMA
Study type	Efficacy and safety trial	Clamp studies
Clinical endpoint	Change in HbA1c at 26 or 52 weeks	GIR and AUC/Cmax test/reference ratios
Testing	Noninferiority (margin of 0.4% HbA1c)	Equivalence (80%-125%)
Patient population	Type 1 and type 2 diabetes patients	Type 1 extrapolation to type 2 patients
Sample size	Minimum 500 total	Depends on the within-subject coefficient of variation (% CV) from published literature or trials for reference product

Comparative Physicochemical Characterization

No product specific biosimilar FDA guidance for insulin biosimilarity assessment exists. The FDA’s guidance for demonstrating biosimilarity recommends that sponsors should consider all relevant characteristics of the proposed product (eg, the primary, secondary, tertiary, and quaternary structure; posttranslational modifications; and biological activities) to demonstrate that the proposed product is highly similar to the reference product not withstanding minor differences in clinically inactive components. ⁷ In addition the guidance also states that extensive structural characterization should be conducted in multiple representative lots of the proposed product and the reference product to understand the lot-to-lot variability of both drug substances in the manufacturing processes. Lots used for the analysis should support the biosimilarity of both the clinical material used in registration clinical trials and the to-be-marketed proposed product. Differences in formulation between the proposed product and the reference product are among the factors that may affect the extent and nature of subsequent animal or clinical testing. In addition, comparative in vitro insulin release profiles and content and stability assays may give an early indication for the differences (if any) that could be expected in the time course of systemic levels.

Comparative Functional Assays

Biological methods are available to characterize the receptor affinity and biological activity of insulin in vitro and in vivo.

In Vitro Studies

To assess any differences in properties between the biosimilar and the reference product, comparative studies such as in vitro bioassays for affinity and insulin- and IGF-1-receptor binding assays, as well as tests for intrinsic activity should be performed. It is important that assays used for comparability testing are demonstrated to have appropriate sensitivity to detect minute differences and that experiments are based on a sufficient number of dilutions per curve to characterize the whole concentration-response relationship. The available information about these assays, including sensitivity, specificity, and extent of validation, can affect the amount and type of additional animal or clinical data that may be needed to establish biosimilarity. ⁶

Animal Studies

In general, nonclinical safety pharmacology, reproductive and developmental toxicity, and carcinogenicity studies are not warranted when the proposed product and reference product have been demonstrated to be highly similar through extensive structural and functional characterization and animal toxicity studies. If there are specific safety concerns based on the clinical use of the reference product, some of or all such additional animal studies with the proposed product may be warranted.

EMA guidance for development of insulin biosimilars states that “comparative study(ies) of PD effects would not be anticipated to be sensitive enough to detect differences not identified by in vitro assays, and are normally not required as part of the comparability exercise.” FDA guidance, however, mentions that “under certain circumstances, a single-dose study in animals comparing the proposed product and reference product using PK and PD measures may contribute to the totality of evidence that supports a demonstration of biosimilarity.” Generally, separate repeated dose toxicity studies are not required. In specific cases, for example, when novel or less well-studied excipients are introduced, the need for additional toxicology studies should be considered. ⁷

Clinical Studies

Human PKPD studies and safety–efficacy studies comparing the proposed product to the reference product are major contributors in supporting a demonstration of biosimilarity.

Clinical Pharmacology Studies: Clamp Studies

Euglycaemic or isoglycaemic hyperinsulinaemic clamp technique is the most common method and is considered as a gold standard for the measurement of insulin action. ⁵ In these clamp experiments, a constant-rate insulin infusion is given which results in insulin concentrations higher than the physiological baseline. Blood glucose is measured frequently and maintained within the euglycaemic range (90 mg/dl) by glucose infusions at variable rates. Due to the exogenous insulin infusion, hepatic glucose production is assumed to be completely inhibited; hence, the glucose infusion rate (GIR) is an indirect measure of the glucose utilization rate (or insulin action). Effects of exogenous insulin are often measured by comparing the areas under the GIR versus time curve after various doses of insulin. Different clamp methods and feedback algorithms for maintaining blood glucose levels exist. Clamp studies can be performed manually or using an automated procedure, for example, the Biostator (blood glucose concentration is measured every minute), and the glucose infusion rates are calculated in a computerized manner by means of a negative feedback algorithm. Manual clamps, on the other hand, are associated with higher blood loss when blood glucose measurements are performed with standard laboratory methods (typical measurement intervals of 5 to 10 min) and have a considerable demand for manpower. Both techniques require substantial experience. However, both methods have been reported to provide similar and reproducible results. ⁶

Efficacy and Safety Studies

No guidelines are issued by FDA assessing insulin biosimilar drug products. FDA guidance for development of new antihyperglycemic drugs states that final demonstration of efficacy should be based on reduction in HbA1c (ie, primary endpoint), which will support an indication of glycemic control. ⁸ Typically a new insulin product is shown to be noninferior to an approved insulin product. These studies should be directed at achieving actual reductions in glycemia (as opposed to simple maintenance of pretrial levels of control) from baseline to end of study. Test and comparator groups should be treated to similar goals. Similar (or better) degrees of glycemic control should be achieved so that comparisons among groups in frequency and severity of hypoglycemia will be interpretable in ultimate risk–benefit assessments.

Although FDA requires noninferiority or superiority based on HbA1c endpoint; EMA on the other hand considers this end point to be insensitive for the purpose of showing biosimilarity of two insulins and hence uses clamp studies as there bases of approval. ⁶ Demonstration of similar PK and PD profiles is considered the mainstay of proof of similar efficacy of the biosimilar and the reference insulin. For this purpose, crossover studies, preferably double-blind insulin clamp studies using single subcutaneous doses of the test and reference agents and performed at an interval of a few days to a few weeks, are considered suitable. The time-concentration and time-action profiles may be studied separately or, preferably, simultaneously (in the same clamp study). The EMA approach is scientifically justifiable. Typically registration trials for insulin include a titration phase; hence any minor or moderate differences between the reference and test products are leveled out during the titration phase. Placing major emphasis on the pivotal trial results potentially increases the false-positive rate (ie, approval of truly nonbiosmilar products).

Clinical Safety

Immunogenicity is one of the most important safety concern related to biological product such as insulin. According to EMA guidance for biosimilar insulin, immunogenicity studies should always include a reasonable number of patients with type 1 diabetes. ⁶ If a mixed population is included, stratification for type of diabetes and preexisting anti-insulin antibodies is necessary. Subjects should be exposed to the test for at least 12 months, including a comparative phase of six months. The primary outcome measure should be the incidence and titers of antibodies to the test and reference medicinal products but there is no need to power the study to formally demonstrate noninferiority regarding immunogenicity. The potential impact of antibodies, if detected, on glycemic control, insulin requirements and safety, especially local and systemic hypersensitivity reactions, should be investigated, and the necessity for further characterization, for example, with regard to their neutralizing potential, considered. EMA guidance suggests that if the immune response to a reference product is rare, both pre- and postmarketing immunogenicity studies will be required. The premarket study will be powered to detect major differences between the two products, and the postmarket study will be designed to detect more subtle differences in immunogenicity.

The FDA has emphasized that it will take a “totality-of-evidence” approach to evaluating biosimilars and that companies should closely consult with the FDA while developing their clinical testing programs. ⁷ The FDA has not yet published its recommendations for the pharmacovigilance program for insulins.

Study Design Considerations

Design of Clamp Studies

Study Population

EMA guidance on biosimilar insulin development states that the study population should be homogenous and insulin-sensitive to best detect potential product related differences and may consist of normal-weight healthy volunteers or patients with type 1 diabetes. ⁶ In addition to faster recruitment, healthy volunteers have the advantage of relatively consistent fasting blood glucose levels. The only disadvantage being the presence of endogenous insulin which cannot be distinguished from exogenously administered insulin by the available assays, except for some insulin analogues. Corrective measures for suppressing endogenous insulin or normalizing the endogenous insulin levels should be adopted. In patients with type 1 diabetes, serum C-peptide concentration should be measured and accounted for to ensure absence of relevant remaining endogenous insulin secretion.

Dose

The subcutaneously administered dose of the test and reference insulin should reflect commonly used therapeutic doses. For rapid-/short-acting insulins doses of 0.2 to 0.3 IU/kg body weight and for intermediate-/long-acting insulins doses of 0.3 to 0.4 IU/ kg body weight are frequently used. The middle physiological range of hyperinsulinaemia (60-70 mIU/L) should be targeted as it has been shown to correspond to the steepest part of the dose response curve of insulin and can thus be expected to be most sensitive to detect potential differences in the time-action profiles of two insulins. All injections should preferably be performed by the same experienced investigator to ensure a reproducible subcutaneous injection. To minimize between occasion variability, it is also preferred that the site of injection (for the second visit), known to potentially influence the rate of absorption of insulin, should also be the same.

Duration

The duration of the clamp studies needs to take into account the known duration of action of the investigated insulin preparation and its dose-dependency. The duration of action in glucose clamp studies may be defined as the time from insulin injection to GIR returning to baseline or, in patients with diabetes, of blood glucose values exceeding a predefined threshold, for example, 150 mg/dl. Clamp durations of 8 to 10 hours for rapid- and short-acting insulins and of 24 hours and more for long-acting insulins have been reported for healthy volunteers or patients with type 1 diabetes when using therapeutic doses. The EMA guidance for development of biosimilar insulin states that a rationale for the selection of the clamp duration should be provided.

Endpoints/Statistical Analyses

Comprehensive comparative data should be provided on the time-concentration profiles of the biosimilar and the reference insulin with AUC and C_max as the primary and T_ma and half-life (t_1/2) as secondary PK endpoints. The area under the glucose-infusion rate (GIR_AUC) and maximum GIR (GIR_max) should be measured as primary and time to GIR_max (T_GIRmax) and time to half-maximal GIR (T_GIR50%) as secondary PD endpoints.

Sample Size

Sample size depends on the magnitude of variability and the design of the study. Variance estimates to determine the number of subjects for a specific insulin product can be obtained from literature or published studies of the reference product. ⁹ FDA guidance on statistical approaches in establishing bioequivalence provides sample size determination considerations for 80% and 90% power using the specified study design, given a selection of within-subject standard deviations (natural log scale), between-subject standard deviations (natural log scale), and subject-by-formulation interaction, as appropriate. ¹⁰ In addition, we came across a useful free excel based (iterative process) web resource that can calculate sample size for various bioequivalence study designs after feeding in the respective within subject variability estimates for desired power. ¹¹

Design of Clinical Efficacy and Safety Trial

To date, no product has been approved as an insulin biosimilar; we speculate the study design for FDA approval to follow a noninferiority trial design based on HbA1c clinical endpoint. As an example, study design for a noninferiority trial comparing insulin detemir with Lantus in patients with type 2 diabetes is described here. ¹²

Study Population and Inclusion/Exclusion Criteria

Insulin naïve men and women aged >18 years who had a diagnosis of type 2 diabetes for >12 months, HbAlc ranging from 7.0% to 10.5% at screening, and had been receiving any oral hypoglycemic regimen with or without insulin for >4 months were randomized to either insulin detemir or Lantus arms.

Dose and Dose Titration

Basal insulin was initiated once daily in the evening at a dose of 12 IU and titrated according to a structured treatment algorithm. Evening insulin doses were titrated throughout the trial to a fasting plasma glucose (FPG) target ≤6.0 mmol/L in the absence of hypoglycemia. Dose adjustments were to be based on the average of three self-measurements before breakfast. During the first 12 weeks, participants had weekly investigator contact. A titration committee monitored the algorithm for insulin dose optimization and reviewed prescribed insulin doses periodically.

Primary Endpoint/Sample Size

The primary endpoint was baseline adjusted HbA1c at end of treatment. The sample size was based on noninferiority of insulin detemir relative to insulin glargine for HbA1c after 52 weeks. To achieve a power of 95% with an expected SD for change in HbA1c of 1.1%, a dropout rate of 15%, and keeping in mind the safety and regulatory requirements for FDA and EMA, 566 patients were recruited in the study. Noninferiority was accepted if the upper limit of the two-sided 95% CI for the difference in HbA1c (detemir–glargine) was less than 0.4%-units.

Immunogenicity

Although not mentioned in this study specifically, the primary outcome measure should be the incidence and titers of antibodies to the test and reference medicinal products, but there is no need to power the study to formally demonstrate noninferiority regarding immunogenicity.

Future Perspectives

Efficient Development Paths for Insulin Biosimilars for FDA Approval

Following are the potential key drug-development questions when seeking market approval for biosimilar insulin:

1. What is the most conservative drug-development path for insulin biosimilar?

2. Can efficacy and safety be extrapolated between type 1 or type 2 diabetes patients?

3. Can clamp studies be used for demonstrating efficacy?

4. What is the optimal design to demonstrate interchangeability? Should it be based on HbA1c or FPG endpoint?

We try to answer these questions by proposing most likely and competing drug-development paths. Several scenarios are considered that are based on either EMA policies and/or sound clinical and statistical reasoning.

Scenario A: One Clamp Study and Two Efficacy Trials (Based on HbA1c Endpoint)

This is considered as the starting point as it is similar to the drug development paths for any other branded insulin product wherein one or two clamp studies are carried out in early phase, followed by two efficacy trials; one each in type 1 and type 2 diabetes patient populations. These efficacy (noninferiority in prescribability) trials (duration 26 or 52 weeks) are designed to demonstrate noninferiority in change from baseline in HbA1c, to closest active comparator with regard to the duration of action. Typical total sample size ranges from 450 to 550 subjects.^12-14 These studies employ a sample size of at least 250 subjects per product for assessing immunogenicity potential.

Scenario B: One Clamp Study (Type 1) and One Efficacy Trial (Type 2)

Although scenario A would be considered as a very conservative route for establishing insulin biosimilarity, we present potential alternate drug development paths which still can answer the key questions and are more efficient. If noninferiority is demonstrated in type 2 patient population (for example), assuming that the reference product is approved for both type 1 and type 2 diabetics, is it really necessary to conduct an additional trial in type 1 diabetics? As 90%-95% of diabetic cases are type 2, it’s preferable to have an efficacy trial in type 2 patients that can also serve as the bridge to efficacy in type 1. Furthermore, the clamp study could be conducted in type 1 diabetics as a surrogate for efficacy trial. This strategy has already been adopted by the EMA, where the basis for approval is bioequivalence in PK and glycemic control (FPG and not HbA1c). In addition, EMA requires a separate safety trial to demonstrate immunogenicity potential of the test insulin biosimilar product, similar to the one stated above, only difference being a comparative phase of 6 months is required in a 12-month trial only for the test product.

Scenario C: One Clamp Study (Type 1) and One Abbreviated Efficacy Trial (Type 2)

Even a more optimistic scenario could be to conduct one clamp study in type 1 diabetics to demonstrate efficacy and one abbreviated safety trial (immunogenicity) with 250 type 2 diabetic subjects exposed to the test biosimilar compound.

Scenario D: One Clamp Study (Type 1) and Efficacy Plus Interchangeability Trial (HbA1c)

The primary focus for new insulin products has been prescribability. On the lines of generic drug products, patients and prescribers can benefit from knowing whether a biosimilar product is only prescribable or interchangeable too with the reference product. This is important from a public health point of view, given that patients receive insulin for their lifetime. Naturally prescribers and patients would want to ascertain if two biosimilar insulins are indeed interchangeable or not. Hence we urge regulatory agencies to encourage companies to assess this important property routinely.

According to the US Biologics Price Competition and Innovation Act of 2009, a biologic is considered interchangeable if it “can be expected to produce the same clinical result as the reference product in any given patient.” Biosimilar products are considered interchangeable if switching between them and a reference product poses no additional risk beyond that of the reference product alone. It is believed that the physicians will be more comfortable in prescribing an insulin biosimilar due to proven interchangeability. The question is what should the interchangeability assessment comprise of? The design should randomize patients to the reference and test products. After the titration period and at the end of the maintenance phase, responders can be rerandomized to either test or reference. The primary endpoint for interchangeability is proposed to be the proportion of responders (patients that attain target levels <6.5% HbA1c) after the switch within a specific noninferiority margin of reference. Pharmacometric modeling and simulation can play a pivotal role in deriving the noninferiority margin. Calculation of this noninferiority margin is beyond the scope of this manuscript but briefly, using semimechanistic PKPD modeling and simulation, time course of FPG-HbA1c and the corresponding proportion of responders before and after the switch (assuming no difference between test and reference) can be calculated.^15,16 This selected margin can then be used for estimating the sample size to adequately power the study to show that the test is within the noninferiority margin for interchangeability (proportion of responders that still remain responders after the switch are no worse than the reference itself).

Scenario E: One Replicated Design Crossover-Clamp Study and One Safety Trial

Interchangeability criteria described in scenario D is based on a population level. Individual bioequivalence or replicated design crossover studies are known to capture within subject variability for both test and reference products, in addition to the subject-formulation interaction term. However, these studies will not be feasible with an efficacy endpoint of HbA1c. Clamp studies are much better suited for this kind of a trial design. Primarily because of the duration of the study and high correlation of FPG and HbA1c levels. It is important to appreciate that biosimilarity assessment if fundamentally different from efficacy demonstration for a new product. FPG is better suited to be a surrogate once that new product is shown to be efficacious based on HbA1c change. ¹⁷ Once individual bioequivalence is demonstrated for test/reference then a case could be made for an interchangeability claim based on FPG, alleviating a more complex design based on HbA1c as discussed in scenario D earlier.

These alternative drug development paths are presented based on sound clinical and statistical reasoning, and it is expected that, along with early discussion with the regulatory agency, they might aid in drastically reducing the development cost and gaining an interchangeability claim for a biosimilar of one of the most studied antidiabetic drug.