An Integrated Bioinformatics and Quantitative Modeling Approach to Investigate Potential Claims of Oral Generic Drug Product Bioinequivalence: Introduction.

Abstract

It has been reported that 88% of prescriptions filled in the United States are for generic drugs, and this has saved the US health system $1.68 trillion from 2005 to 2014.1 Over the same time period, the number of approved Abbreviated New Drug Applications for generic drugs increased by 44%.2 Generic drugs are considered safe, effective, and substitutable for the reference-listed drug under all clinical use conditions. This is because the Food and Drug Administration (FDA) mandates that they be pharmaceutically equivalent and bioequivalent through vigorous testing; therefore, they are considered to be therapeutically equivalent to the reference-listed drug (or brand-name) product.3, 4 However, the FDA's Office of Generic Drugs will occasionally receive complaints from patients and/or healthcare providers that a generic drug was either not as effective or safe as the brand name product that they were taking or prescribing previously. Such was the case with a generic version of bupropion.5 Under these circumstances, because many diseases are difficult to treat, and these reports are lacking controls, it is extremely challenging for the FDA to determine if complaints about benefits or risks of generic drugs are real or due to other factors related to disease progression or to some other factors. A related point is that the FDA interprets drug safety as a benefit-to-risk ratio. A drug is safe if its benefits favorably outweigh its risks. If patients receiving a bioinequivalent generic drug product fail to gain the benefits provided by a brand-name product, then its benefit-to-risk ratio is overstated, and failure to provide benefits may be considered a safety issue. Therefore, it is imperative that complaints about generic drug substitution be investigated thoroughly to determine if the generic drug is actually meeting bioequivalence standards and if there is a rational, mechanistic explanation for the purported reduction in benefits or increase in risks attributed to the generic drug. Through this process of investigation, the regulatory science supporting the approval criteria for all drugs, including both generic and brand-name products, continues to evolve and remain rigorous and comprehensive. In the commitment letter of the Generic Drug User Fee Act (GDUFA) of 2012, the FDA committed to consult with industry and the public to create an annual list of regulatory science initiatives specific to research on generic drugs. The research studies conducted under these initiatives will advance the public health by providing access to safe and effective generic drugs. The regulatory science results will provide new tools for the FDA to evaluate generic drug equivalence and for industry to efficiently develop new generic products in all product categories. The Office of Generic Drugs implements the GDUFA Regulatory Science Research Program by collaborating within the FDA as well as externally through grants or contracts. GDUFA's authorization of support for regulatory science research activities illustrates the importance of regulatory science innovation in the generic drug program.6 With public input, postmarket evaluation of generic drugs has been identified as a GDUFA research priority.7 Furthermore, GDUFA regulatory science priority areas include quantitative modeling and simulation tools, such as physiologically based absorption and pharmacokinetic models (PBPK) and population pharmacokinetic-pharmacodynamic (PK/PD) models as well as bioinformatics approaches for benefit-to-risk ratio assessments.8, 9 The University of Florida Center for Pharmacometrics and Systems Pharmacology and FDA's Office of Generic Drugs have collaborated on a research project to develop a mechanism- and risk-based strategy to evaluate reported postmarketing complaints about orally administered generic drugs. This strategy, starting with a patient or provider complaint about a generic drug, has 3 integrated components: (1) bioinformatics tools, (2) PBPK models, and (3) PK/PD models. Figure 1 illustrates the approach graphically that can be used to evaluate postmarketing complaints. The first step in enabling this integrated approach is to use software tools for bioinformatics that analyze the underlying molecular mechanism of the purported safety signal or reduction in effectiveness of the generic drug. This is achieved, for example, with adverse events (AEs) by using bioinformatics software tools to interrogate AE: Drug pairs from the public, online FDA Adverse Event Reporting System using bioinformatics tools such as SAS® Platform (SAS Institute, Cary, North Carolina), or the Molecular Analysis of Side Effects platform (Molecular Health, Heidelberg, Germany). These analyses use disproportionality analysis measures (eg, proportional reporting ratio) to identify a positive risk or safety signal along with a drug target–AE association. These results are supplemented by examination of a web-based deidentified, individual-level healthcare claims data warehouse that links drug prescriptions with clinical outcomes (Truven Health Analytics Marketscan® Research Database, Ann Arbor, Michigan). These databases, in turn, are complemented by additional software, EvidexTM Web-based Platform (Advera Health Analytics, Santa Rosa, California) to categorize and interrogate AEs in the FDA Adverse Event Reporting System using a proprietary computer algorithm in order to generate a rank order of AEs and associated downstream costs for a given drug that are either on-label or off-label AEs. This extensive bioinformatics approach provides the user with a pharmacological hypothesis for an AE-drug pair that is then used to evaluate the likelihood that the generic drug can plausibly cause the purported AE. This approach provides a more deterministic way for the FDA to prioritize, stratify, and categorize purported generic drug reduced efficacy or AEs into those with biological plausibility that are likely to be caused by a generic drug (ie, a signal) vs those that are unlikely to be caused by the generic drug (ie, noise). The AE-drug hypothesis generated by the bioinformatics approach is then evaluated from a bottom-up and top-down approach using a PBPK and PK/PD models, respectively, to evaluate the biological, physiological, and drug- and/or formulation-related causes of the AE-drug pair report. The second step of the integrated approach is to build a PBPK model that simulates PK profiles from, for example, oral immediate- or extended-release test and reference products and provide in silico results of virtual bioequivalence studies among different formulations. These simulations are inputted into the PK/PD model in step 3 and also are used to conduct a PBPK-based sensitivity analysis that may differentially influence the bioavailability of a generic vs brand-name drug product. PBPK modeling allows various physical, chemical, and physiological factors that influence the in vitro dissolution and in vivo absorption to be included in the sensitivity analysis. These include particle size, excipients, solubility-permeability attributes (biopharmaceutical classification system), pH-dependent solubility of the active ingredient, pH-dependent dissolution of the formulation, effect of dose, and variations in gut transit time and motility. PBPK models that incorporate key physical-chemical properties of the drug and its formulation and their associated variability may then be used to investigate possible in vitro–in vivo correlations as well as to compare in vitro dissolution profiles of test and reference drug products. This may support evaluation of new mechanistic model-based parameters to predict bioequivalence by comparing dissolution curves for differences or similarity as an alternative to the empirical factors f1 and f2 used by regulatory agencies. The third step of the integrated strategy is to build a quantitative PK/PD model of the generic drug to investigate in silico the impact of variability or differences in PK associated with potentially bioinequivalent drug products on the PD of the drug. The clinical impact of PK variability depends on the respective shapes of the PK/PD relationship for benefit and risk. Simulations of PK profiles for different degrees of differences in bioavailability between test (generic drug) and reference (brand-name drug) products generated by the related PBPK models are used to determine how different test and reference products would have to be in order for their area under the concentration-time curves and peak concentrations to render the AE (or loss of efficacy) that was the focus of the bioinformatics approach. The significance of this research project is that an integrated bioinformatics and pharmacometrics model-based strategy allows for a thorough risk-based evaluation of purported claims of bioinequivalence of generic drugs in the postapproval marketplace. The bioinformatics approach enables the identification of pharmacological pathways, including targets, that lead to AE-drug relationships and provide compelling evidence that the generic drug can or cannot cause the reduced effectiveness or AE reported by patients and/or healthcare providers. The PBPK modeling process allows one to deconstruct the PK profiles of the genetic and brand-name products and identify formulation-related differences between the products in terms of a sensitivity analysis of drug and formulation factors that have the highest likelihood of providing insights into relative bioavailability of generic and brand-name products. On confirmation that the generic drug can, in fact, cause a reported complaint, the PK/PD modeling process provides insight into the extent to which a generic drug must differ in terms of drug exposure (peak concentration and/or area under the curve), from its brand-name counterpart in order to yield the reported complaint. The bioinformatics and pharmacometrics model-based processes to evaluate marketplace performance and questions related to substitution of approved generic drug products for brand-name products was evaluated with several exemplar research projects that are reported in more detail in the individual manuscripts in this journal. These include an assessment of potential bioinequivalence of high-risk (ie, biopharmaceutical classification system type II drugs) generic products such as oral, immediate-release, and modified-release antiepileptic drugs and extended-release metoprolol, and an assessment of potential liability of bioinequivalence of generic products not yet off patent, using new oral anticoagulants as an example. Although the objective of this research is to establish a rigorous mechanistic workflow to assess complaints about generic drug bioinequivalence (ie, substitutability), this same process is equally useful for comparing brand-name drugs used in pivotal clinical trials preapproval to their to-be-marketed market formulations (ie, prescribability). The integrated approach can also be applied to other situations, such as (1) to identify and prioritize the selection of postmarketing surveillance of specific generic products that are thought to have the greater probability of bioinequivalence or whose bioinequivalence would have the greatest impact on clinical outcomes, and (2) to provide new hypotheses for future GDUFA regulatory science priority projects. Furthermore, the integrated strategy of this research can be extended to include assessment of potential food effects and bioinequivalence and factors influencing comparative bioavailability of generic vs brand-name drugs among healthy volunteers and intended patients. The authors thank Dr. Yehua Xie (Office of Generic Drugs, CDER, FDA) for his project management and administrative assistance with this article. FDA Grant No. 1U01FD005210-01 funded this study.