Abstract
Robust assessments of the nonclinical safety profile of biopharmaceuticals are best developed on a scientifically justified, case-by-case basis, with consideration of the therapeutic molecule, molecular target, and differences/similarities between nonclinical species and humans (ICH S6). Significant experience has been gained in the 10 years ensuing since publication of the ICH S6 guidance. In a PhRMA-FDA–sponsored workshop, “Nonclinical Aspects of Biopharmaceutical Development,” industry and US regulatory representatives engaged in exploration of current scientific and regulatory issues relating to the nonclinical development of biopharmaceuticals in order to share scientific learning and experience and to work towards establishing consistency in application of general principles and approaches. The proceedings and discussions of this workshop confirm general alignment of strategy and tactics in development of biopharmaceuticals with regard to such areas as species selection, selection of high doses in toxicology studies, selection of clinical doses, the conduct of developmental and reproductive toxicity (DART) studies, and assessment of carcinogenic potential. However, several important aspects, including, for example, appropriate use of homologues, nonhuman primates, and/or in vitro models in the assessment of risk for potential developmental and carcinogenic effects, were identified as requiring further scientific exploration and discussion.
Before any drug can proceed to testing in humans, a nonclinical risk assessment evaluating the pharmacologic, drug dispositional, and toxicologic profile must be submitted to regulatory agencies. Current paradigms for nonclinical safety assessment are, in large part, built upon traditional experience with small molecules (see ICH M3 (R1) 2000). However, biotechnology-derived pharmaceutical products (biopharmaceuticals), which are increasingly being developed as therapies for a variety of clinical indications, are generally large, complex proteins with some important differences in safety issues compared to traditional small molecule pharmaceutical drugs (e.g., immunogenicity). Inappropriate application of small-molecule paradigms to large molecules can result in compromised science, regulation, and risk assessment. Currently, robust assessments of the safety of biopharmaceuticals are best developed on a scientifically justified, case-by-case basis, with consideration of the therapeutic molecule, molecular target, and differences/similarities between nonclinical species and humans.
The ICH S6 guidance document (1997) discusses requirements for animal studies for the safety assessment of biopharmaceuticals, emphasizing a case-by-case approach because of the unique properties of these drugs and their targets. Biologic activity is, by design, specific and target related, so that off-target toxicity is rare (Tabrizi, Tseng, and Roskos 2006). Unlike small molecules, metabolism of a biopharmaceutical does not occur by enzymatic biotransformation processes (e.g., P450 pathways) but rather by simple protein catabolism. Biopharmaceuticals are not chemically reactive with physiologic macromolecules, so that genetic toxicology studies are generally not applicable. The size of a biopharmaceutical may preclude its ability to enter a cell. The human ether-a-go-go-related gene (hERG) assay is generally not usefully employed for biopharmaceuticals that have a very low potential to interact with the extracellular or intracellular (pore) domains on the hERG channel (Vargas et al. accepted). With regard to nonclinical developmental and reproductive toxicity (DART) evaluations, substances of very high molecular weight do not usually traverse the placenta, but there are a few exceptions such as immunoglobulin G (IgG) via Fc-mediated transport. The placental transport of maternal antibodies to the human embryo/fetus occurs primarily in the third trimester (Simister 2003), after organogenesis, whereas it appears that IgG can cross into the embryo/fetus earlier in rats (Roberts, Guenthert, and Rodewald 1990). More research is needed to understand how the timing and extent of transfer in animal models relates to prediction of human developmental risk.
Safety assessment experience with biopharmaceuticals accumulated over the last 10 years in conjunction with the development of new regional regulatory guidance has prompted discussions regarding an update of information contained in the current ICH S6 guidance. In addition, the US Food and Drug Administration (FDA) review and evaluation of certain biopharmaceuticals was transferred from the Center for Biologic Evaluation and Research to multiple therapeutic divisions within the Center for Drug Evaluation and Research in 2003, and that transfer had several implications regarding FDA reviewer expectations for biopharmaceutical development (Schwieterman 2006). These factors emphasize the need for increased consistency in application of general principles and approaches for nonclinical development of modern biopharmaceuticals.
MATERIALS AND METHODS
A meeting to discuss “Nonclinical Aspects of Biopharmaceutical Development” was sponsored by the Pharmaceutical Research Manufacturer’s Association (PhRMA) and FDA in April 2007. The goals of the forum were to engage industry and US regulatory scientists in exploration of current scientific and regulatory issues relating to the nonclinical development of biopharmaceuticals, to enable shared scientific learning and experience, and to work towards establishing realistic expectations for development needs. This goal was addressed through a combination of (1) platform presentations, (2) case studies discussed in breakout sessions, and (3) full group presentation/discussion of the case studies. Participants were divided into three groups of approximately 30 scientists each from industry and the FDA. Each group circulated through three of four case studies that were fictionalized versions of actual biopharmaceutical safety assessment programs. Group discussion was focused around specific scenarios and questions posed regarding nonclinical safety assessment plans, e.g., species selection, study designs, and assessment of developmental toxicity and carcinogenic potential. This paper summarizes the case studies and points of discussion in the ensuing breakout groups.
RESULTS
Case A: A Recombinant Human Protein with Immunomodulatory Activity for the Treatment of Cancer
This case involves a recombinant human protein of approximately 20 kDa, an immune modulator/stimulator that is being developed for treatment of patients with advanced cancer and which may be used in combination with other oncology drugs. Pharmacodynamic markers of activity (increased monocytes, natural killer cells, and large unstained cells in peripheral blood) observed in nonclinical species were also confirmed in humans; thus, the activity of the protein therapeutic is monitorable in patients.
The protein is relatively conserved in nonhuman primates (NHPs) and humans (96% homologue in macaques; 100% in chimpanzee); however, it is not conserved in lower species, hence a murine homologue of this protein was developed. This murine homologue was used in all the mouse studies. The initial clinical schedule, intravenous administration for a single cycle of daily × 5, was supported by safety pharmacology and single-cycle (daily × 5) toxicology studies in the mouse and cynomolgus monkey. The toxicology studies also included evaluation of reversibility. Repeated-dose toxicology studies in mice and cynomolgus monkeys were conducted to support chronic administration of the drug in patients in daily × 5 cycles repeated every 2 weeks. Because of strong antigenic responses in cynomolgus monkeys, the product was evaluated in alternative NHP species; therefore, preliminary studies in rhesus monkeys and chimpanzees were additionally conducted.
The human and murine constructs were highly immunogenic in all species tested. This may have been due to the nature of the product (an immune modulator), thus enhancing antibody formation against the product. Anaphylaxis was observed in all high-dose monkeys and in mice upon readministration of the drug, as early as cycle 2. In addition, neutralizing and/or clearing antibodies were seen in both monkeys and mice at most doses evaluated. Because of the anaphylactic reactions and neutralizing and/or clearing antibody formation in mice and monkeys, it was concluded that meaningful chronic safety testing of the drug in the animals could not be conducted. A combined female fertility, early embryonic development, and embryofetal toxicology study in mice (combined Segments 1 and 2) was performed in windows of approximately daily × 7 (i.e., Days − 5 premating to + 3 of mating, Gestation Days [GDs] 0 to 7, and GDs 8 to 15) with separate groups of mice to avoid anaphylactic reactions that were seen in dose range-finding studies. The drug did not cause developmental or reproductive effects in mice.
Questions and Discussion
What factors influence the evaluation of multiple NHP species and/or support use of surrogate molecules for nonclinical toxicity studies? In light of the serious adverse reactions reported for TGN1412 (Suntharalingam et al. 2006), there was some concern regarding the drug’s stimulating effect on the immune system. Thus, studies in more than one species were favored for agents with immunostimulatory activity. However, considering that rhesus and cynomolgus monkeys are generally similar in their response to drugs, toxicology studies in one species of NHPs, i.e., cynomolgus monkeys, were deemed sufficient. Studies with chimpanzees were not favored due to ethical concerns. Consistency in pharmacologic response supported the use of a murine homologue for evaluation of developmental and reproductive toxicity (see below).
What factors should be considered when considering first-inhuman doses based on definitive toxicology studies with this immunomodulatory protein? Although a more conservative approach, i.e., using a minimally anticipated biological effect level (MABEL) to estimate initial clinical doses, may be generally desirable for immune-stimulating agents (such as TGN1412) in a population of healthy human volunteers, dose selection based on MABEL for an oncology trial will likely expose a number of cancer patients to suboptimal doses. Several dose escalations would be needed to reach a pharmacologically active dose. Given life-threatening conditions such as advanced/refractory cancer, the MABEL-based approach thus appeared overly conservative. The risk in this patient population can be minimized by the standard practice of sequentially entering cancer patients into the trial, as is usually done for most cancer drugs, rather than dosing multiple study subjects in a short period of time.
Establishment of a maximum tolerated dose (MTD) in toxicology studies of proteins was thought not to be essential. Demonstration of dose-dependent toxicity with identification of end-organ toxicities and identification of a no observed adverse effect level (NOAEL) would be sufficient for selection of the starting clinical dose. Thus, elicitation of an MTD in the nonclinical studies may not add value. In fact, too high of a dose may result in formation of protein aggregates and protein overload resulting in nonrelevant findings, for example proteinuria or nephropathy (Bugelski et al. 1992; Zoja, Benigni, and Remuzzi 2004). Therefore, for cases when the drug is non-toxic, the high dose in the toxicology studies may be based on a suitable multiple (e.g., 10× per FDA guidance, 2005) of the highest anticipated human dose or exposure. Finally, considering that the drug in Case Study A is small (approximately 20 kDa) and will not be confined to the bloodstream, human dose selection based on body surface area (mg/m2) was deemed to be most appropriate.
Is the DART study with the murine homologue in mice sufficient? Risk-benefit considerations regarding this indication (oncology/advanced cancer) would likely preclude a real need for reproductive toxicology studies. Apart from this, because consistent pharmacologic effects (e.g., comparable potency for increases in monocytes and natural killer cells in the blood) were observed in mice, NHPs, and humans, the study in mice with the homologue could be considered applicable, whereas the conduct of additional studies in NHPs was deemed unnecessary. Difficulties in designing and conducting reproductive toxicity studies in NHPs were discussed, e.g., limited availability of reproductively mature animals, limited number of offspring (usually one per pregnancy), difficulty in concurrent timing of pregnancies, small group sizes (e.g., 8 to 10 NHPs per group versus 20 to 25 for rodents), the inherent low fertility of NHPs, and the high rate of spontaneous pregnancy failures.
Despite the support for the study in mice versus NHPs, it was recognized that a weight-of-evidence approach that included an interrogation of the literature would best inform the reproductive safety assessment. For example, it was noted that this drug could likely increase interferon levels, and interferon reproductive toxicity is well described (principally abortifacient effects). In this case, then, based on this known class effect, a negative result in the mouse study would not necessarily provide reassurance regarding reproductive risk.
Are Nonclinical Studies to Support Combination Oncology Clinical Trials Necessary? Pharmacology studies of a combination should provide information regarding the potential for pharmacodynamic interactions, e.g., understanding the potential for overlapping or contradicting pathways if both drugs have immunomodulatory activities. Although inclusion of toxicity end points into the design of combination pharmacology studies is desirable, this approach may not always be feasible for immune modulators. For example, some pharmacology studies are conducted in immune-compromised mice, which do not represent a feasible model for assessing pharmacodynamic end points of immune modulators. Finally, the toxicity and pharmacology of each separate product is usually well characterized before clinical studies are performed with an investigational product combined with an approved product. Combination toxicology studies would not be necessary when the investigational product will be used with an approved anticancer product, unless there is a cause for concern, e.g., there is a potential for pharmacodynamic interactions that would increase risk.
Case Study B: A Human IgG1 Antibody for the Treatment of a Central Nervous System Disease
HuSafe-001 is a human IgG1 monoclonal antibody (mAb) targeting a cell-associated protein. It is being developed for the treatment of a central nervous system (CNS) disease that is slow in progression and highly disabling at later stages in life. Published data indicate that the target protein is expressed centrally and peripherally but in measurable levels in the CNS only in the proposed disease indication. A good association between disease manifestation and target protein deposits in brain tissues has been reported. Additionally, target-deficient knockout mice are reported to show a generally benign phenotype and are disease resistant.
The proposed mechanism of action for HuSafe-001 is Fc-mediated phagocytosis of protein deposits in the CNS. The phagocytosis of protein-coated beads by human peripheral blood monocytes has been demonstrated in vitro, and the engagement of Fcγ receptors is believed to be essential for antibody efficacy properties in vivo. Additionally, antibody penetration into the CNS is necessary to exert its pharmacological effect. The ratio of cerebral spinal fluid:serum partitioning of the antibody has been determined be >0.1% in animal models and patients (Pardridge 2007; Rubenstein 2003).
Affinity determination using recombinant target protein revealed that HuSafe-001 binds with a K d = 300, 3000, and 300 pM to mouse, cynomolgus monkey, and human protein, respectively; no binding to the recombinant marmoset protein was detected. As part of the development activities, a murinized IgG1 (murine framework and constant regions along with human constant domain region (CDR) region), MuSafe-001 was developed. MuSafe-001 was assumed to have an affinity similar to HuSafe-001, as the CDR region was not changed. Additionally, tissue cross-reactivity studies with HuSafe-001 were conducted with tissues from mouse, cynomolgus monkey, and marmoset. Immunohistochemistry studies in tissues from humans and healthy animals of all four species showed no binding in the CNS and a variable cell membrane binding pattern in other tissues including major organs and vascular endothelium. The immunohistochemistry studies in tissues obtained from diseased mice revealed strong binding in CNS tissues with HuSafe-001.
Pharmacology studies were conducted in a murine disease model; mice were exposed for 1 month to MuSafe-001 or a murine IgG1 control antibody. MuSafe-001 was administered at intravenous (IV) doses of 1, 3, 10, and 30 mg/kg/week for 4 weeks. Efficacy end points, such as reduction in target protein deposits in the CNS, were not met, and a longer-term study was initiated. Immunohistochemistry of target brain tissue in the disease model using an anti-idiotypic antibody to the MuSafe-001 after in vivo dosing indicated dose-dependent staining of the target protein by MuSafe-001 and no staining (binding of antibody to brain) in control animals detecting binding of the antibody to brain. No safety issues were noted in the study.
Questions and Discussion
What are the most relevant animal species for the conduct of HuSafe-001 safety studies? Discussions reinforced the primary requirement that a toxicologically relevant animal species be used that demonstrates appropriate target binding and function of the monoclonal antibody (Tabrizi and Roskos 2007). Important criteria include: evaluation of the target protein with respect to sequence and structural properties; the binding epitope; conservation of the sequence; cross-species homologue; and mAb affinity for and functional modulation of the target antigen. Comparable binding of antibody to antigen in relevant tissues in the nonclinical species and humans can be another important step in determination of the toxicologically relevant species (Chapman et al. 2007).
The affinity of the antibody for human and mouse protein were similar (300 pM); however, antibody binding affinity in the cynomolgus monkey protein was 10-fold lower (3000 pM). It was noted that the affinity differences across species will need to be considered when designing the safety studies and can potentially be overcome by dose. Both the cynomolgus monkey and the mouse should be considered as relevant species for the conduct of safety studies. As no binding to marmoset protein was detected, the marmoset was not considered a toxicologically relevant species.
The affinity data are supported by the relative strength of the staining observed in immunohistochemistry studies in tissues obtained from healthy human, cynomolgus monkey, and mouse. The lack of binding to CNS tissue in healthy animals is consistent with the lack of constitutive expression of the protein in healthy animals. On the other hand, immunohistochemistry data in marmoset that demonstrate some tissue binding (thus suggesting the possibility of off-target binding) make clear interpretation of membrane binding in other (cross-reactive) species more challenging; in other words, it is not clear whether membrane binding in a cross-reactive species is off-target or target specific, as similar binding is observed in the non–cross-reactive species. The availability of the target protein–deficient knockout (KO) mice provides the opportunity for addressing any off-target binding concern, in that comparison of tissue binding data in the KO mice, which do not express the target protein, with that of the wild-type mice, in which membrane staining in various tissues has been observed previously, would allow clear interpretation of the immunohistochemistry data.
Is there any value in assessing the safety of the antibody in the murine disease model? Safety assessment studies are not usually performed in animal models of disease due to the lack of appropriate historical data and background information. A disease model is best employed when there is a discreet parameter to target based on a particular hypothesis regarding safety. For this indication, the toxicity of antibodies could specifically arise from the mAb interactions with the target protein; therefore, the murine disease model can provide a potential opportunity for evaluation of target-specific CNS toxicity of HuSafe-001. The protein was expressed in the nonclinical disease model, and dose-dependent binding was also demonstrated. The ability of the mAb to cross the blood-brain barrier could be of concern, as antibodies generally have low partitioning into the CNS (Pardridge 2007). However, data from studies conducted in the murine disease model indicate that partitioning of the antibody does occur in the brain compartment of the diseased animals.
Which antibody construct (HuSafe-001 or MuSafe-001) is most relevant for evaluation of the antibody safety and efficacy? The lead candidate HuSafe-001 is a human IgG1 antibody. Human IgG1 are efficient in interacting with human effector cells to induce effector functions (complement-dependent cytotoxicity; antibody-dependent cell-mediated cytotoxicity; phagocytosis). Human IgG1 antibodies can interact with murine effector cells albeit with a lower affinity as compared to the murine equivalent construct; however, they are highly immunogenic in mice. The mouse equivalent to human IgG1 is the murine IgG2a and not murine IgG1 (Desjarlais 2007). As MuSafe-001 is a murinized IgG1 construct, it cannot be considered a suitable molecule for studying the effects of target modulation since it lacks the effector functions observed with human IgG1. A mouse IgG2a version of MuSafe-001 should be produced, which is representative of human IgG1 effector function; the affinity of the IgG2a murine construct will need to be determined. As human antibodies are highly immunogenic in mice, studies in the disease model could be conducted with the murinized IgG2a construct to minimize immunogenicity and to engage the effector functions. The lack of efficacy in light of the binding data in the murine disease model might reflect the inappropriateness of the murinized IgG1 antibody construct.
Dose range–finding and 4-week safety assessment studies were conducted in mouse and cynomolgus monkey using HuSafe-001 intravenous doses of 0 (vehicle control), 3, 10, and 30 mg/kg/week. The NOAEL in cynomolgus monkey was 30 mg/kg. In the 4-week safety study in mice, in the low-and mid-dose groups, many animals exhibited lethargy and a hunched posture shortly after dosing following the third and fourth doses. No unusual clinical signs were observed in the 30 mg/kg group, and no other findings were noted in the mouse. No pharmacokinetic or immunogenicity (murine anti-human antibody) data were provided from the mouse study. Additionally, single-dose pharmacokinetic studies were conducted in mice, marmoset, and cynomolgus monkeys for both HuSafe-001 and the murinized antibody, MuSafe-001. Pharmacokinetic properties of the antibodies were linear and similar to endogenous IgG1 in those species. The safety margins in cynomolgus monkeys and mice were estimated using an allometric scaling approach based on body weight and predicted human exposure (FDA 2005).
A 3-month study in mice and cynomolgus monkeys at doses of 3, 10, and 30 mg/kg/week was proposed to support a phase 1 clinical single-ascending-dose study in patients over a dose range of 0.1 to 10 mg/kg (starting dose at 0.1 mg/kg with half-log ascending schedule at 0.3, 1, 3, and 10 mg/kg in patients), with a subsequent 3-month multiple-ascending-dose study over the same dose range (0.1, 0.3, 1, 3, 10 mg/kg every 2 weeks × 12 weeks in patients).
Are the predicted safety margins in monkeys and mice based on the predicted human exposure acceptable? Fc receptors (FcRn) function as a salvage pathway in regulation of antibody clearance (Tabrizi, Tseng, and Roskos 2006). As human IgG antibodies have a higher affinity for murine FcRn receptors than for human FcRn receptors (Ober et al. 2001), IgG antibodies demonstrate a much lower clearance and longer half-life (t 1/2) in the mouse that is not predictive of human clearance. Because the mouse pharmacokinetic data with HuSafe are included in the allometric scaling approach, the prediction of human clearance (14 ml/day/kg) is overestimated, resulting in lower predicted exposure (serum area under the curve [AUC]) in humans and thus overestimation of the predicted safety margins in monkeys. With the exception of mouse clearance, clearance values for all species are as expected for human IgG1 (with no indication of an effect of antigen on clearance). Therefore, it can be assumed that the typical clearance of endogenous IgG1 in humans is a valid estimation for determining human exposure. For IgG1 antibodies in humans, clearance should be approximately 2.5 ml/day/kg (2.5-fold slower than that for the monkey), so clearance in the cynomolgus monkey can be used to obtain a reasonable estimate of human clearance. Based on this approach, then, the predicted exposure in humans and safety margins in the cynomolgus monkey should be recalculated with inclusion of adjustments for affinity differences between human and monkey protein.
Is the proposed safety package for HuSafe-001 adequate to support the phase 1 single- and multiple-dose studies? Is the starting dose and dose range proposed for phase 1 acceptable? Based on the discussions presented above, a 1-month dose range–finding study in normal mice with MuSafe-001 IgG2a antibody could be considered in order to determine safety, pharmacokinetics, and immunogenicity. Although overcome with higher doses, clinical signs consistent with immunogenicity (lethargy and a hunched posture) were observed in mice shortly after dosing following the third and fourth doses of HuSafe-001. Therefore, the use of the surrogate MuSafe-001 IgG2a in the longer-term safety and efficacy studies in mice is justified. Studies in healthy mice, which do not possess the target protein, could inform the consequences of nonspecific membrane binding and might also facilitate the design of a study in the murine disease model with respect to dose and duration. Because a murine disease model was available, a safety study in the murine disease model with MuSafe-001 could be conducted in lieu of a study in wild-type mice and should provide the additional opportunity for evaluation of target-specific toxicity in the CNS (especially relevant to the phase 1 trials, which will be in patients). The single-dose study in humans could be supported by the 4-week monkey study after correction for affinity differences of HuSafe-001 for monkey antigen. A 3-month multi-dose toxicity study in monkeys, which should mimic the clinical dosing regimen after correction for the pharmacokinetic differences, was recommended to support the multiple-dose clinical trial.
With respect to the proposed starting dose in clinical studies, more information was desired. Data from the murine disease model with MuSafe-001 were considered crucial for determination of a MABEL from the relevant nonclinical efficacy and safety studies. Studies in monkeys with appropriate doses (with respect to exposure, affinity, and duration) were also recognized as important for determination of all potential toxicities.
Case Study C: A Monoclonal Antibody that Neutralizes a Circulating Protein for the Treatment of Cardiorenal Disease
Case study C focused on a mAb that binds and neutralizes a circulating protein which plays a key role in regulation of cellular and extracellular processes. Both the target protein and its pharmacology are conserved across species (rodent, monkey, human). Nonclinical pharmacology models for efficacy were conducted in rodents using a murine antibody.
To support the initial single-dose and multiple-dose safety studies in humans, an extended toxicology study with safety pharmacology and immunotoxicology end points and a 1-month reversibility phase for control and high-dose animals was conducted in cynomolgus monkeys using five weekly doses. Exposure decreased over the course of treatment at the high dose. As drug levels decreased, high titers of antidrug antibodies were detected during the recovery period for high-dose monkeys. A target organ of toxicity was not identified at the highest dose tested in monkeys, which represented 100- and 9-fold multiples relative to predicted clinical starting and stopping doses, respectively. Tissue cross-reactivity studies were conducted with tissues from humans and cynomolgus monkeys.
Based on a regulatory request, a 9- (versus 6-) month study in cynomolgus monkeys is planned to support phase 2 clinical trials and market registration. Reproductive and developmental toxicity studies are tentatively planned.
Questions and Discussion
Given that the target is more than 94% to 100% conserved across species (mouse, monkey, human), but a significant antidrug antibody response (high titers in most monkeys) in the monkey to humanized monoclonal antibody, what is the development path forward? Discussion focused on the need for testing in species additional to cynomolgus monkeys and the best way to accomplish this, if needed. Although toxicity testing in the monkey only should be adequate because the target was conserved across species, short-term toxicity studies using the clinical candidate in the rodent should be considered to assess the feasibility of conducting rodent studies. A tissue cross-reactivity study using rodent tissue might provide additional evidence that the monkey is the only relevant species. The use of either an additional monkey species or a rodent surrogate was not generally recommended.
If toxicity had been observed in monkeys, would a surrogate study in rodents still add value? If a surrogate were to be developed, What characterization of the rodent surrogate is needed to ensure comparability to the clinical candidate? A rodent surrogate was not considered to add value if toxicity had been demonstrated in monkeys. Apart from recognition of the significant development effort associated with appropriate characterization of a surrogate antibody, the relevance of a surrogate was questioned based on its potential differences from the clinical candidate. A surrogate molecule should have target affinity and selectivity similar to the clinical candidate. The potency of the surrogate and clinical candidate should be comparable in a functional assay, e.g., the immunoglobulin subclass selected for surrogate studies should possess similar functionality as the intended therapeutic human subclass of monoclonal antibody. The impurity and aggregation profiles should be comparable. A tissue cross-reactivity study could provide additional validation of the rodent surrogate in providing a comparison of binding in various tissues across species. Differences in pharmacokinetic profiles of the surrogate and the clinical candidate would be acceptable if doses could be adjusted to yield similar exposure.
Because the initial toxicity study in monkeys showed no significant toxicity, how should the chronic study in monkey be designed? The use of a high dose that elicits some toxicity was considered optimal. If toxicity could not be achieved, then administration of the maximum feasible subcutaneous dose should be considered. However, as advances in technology provide antibodies with increased solubility, the value of a maximum feasible dose is unclear and achieving this can result in a significant development/resource challenge. Because intravenous dosing is generally less immunogenic, slow bolus intravenous dosing (but not long continuous infusions) could be considered. The maximum dose in the chronic study should provide at least a 10-fold safety margin over the maximum clinical dose based on a multiple of the expected target saturation.
How should the reproductive and developmental toxicity studies be designed? Although the clinical candidate would be expected to be immunogenic in rodents, a segmented dosing approach covering different periods of organogenesis (ICH S5(R2) 2005) may be employed to avoid decreased exposure due to antibodies. At the least, exposure of the dams to the clinical candidate should be demonstrated. If the clinical candidate could not be used, a rodent surrogate approach could be considered. The use of nonhuman primates for reproductive and developmental toxicity studies was the least preferred option due to difficulties in design and interpretation of these studies (aforementioned in Case Study A).
Literature on the knockout of the target might clearly indicate an adverse effect of the clinical candidate on embryofetal development (e.g., based on knockout data). Although a category C label would likely ensue, regulators may still request the conduct of an embryofetal development study, if possible, to obtain some data with the clinical candidate.
Should carcinogenicity be assessed if the target has a well-established role in carcinogenesis? What if the role of the target in carcinogenesis is not well understood? No additional studies would be needed if the role of the target in carcinogenesis is well established. Labeling of the product would convey this information to the physician and patient. If the role of the target in carcinogenesis is not well understood, a hypothesis-driven strategy to assess carcinogenic risk may be considered. Approaches might include the use of in vitro cell-based assays, xenograft models, and tumor-promotion models. A careful evaluation for any proliferative signals in chronic rodent and monkey studies may aid in determining the need for further studies, e.g., a 2-year carcinogenicity study if a rodent surrogate was available. However, the response in the rodent may not be representative of a pleiotropic immunomodulatory response to the target in the human.
Case Study D: Assessment of Carcinogenic Potential of D1, a Biotechnology-Derived Product
D1 is a recombinant human hormone analog of an endogenous protein D* having several amino acid substitutions. D1 is an antagonist to the endogenous receptor complex D*R, thus blocking signal transduction and decreasing circulating insulin growth factor-1 (IGF-1). D1 is intended for use in treating an abnormal growth disorder. D1 has marked species differences in its ability to displace D* (human ~ monkey > dog > mouse > rat), with rat D* being approximately 100-fold less displaced than human. In vivo studies showed D1 decreased plasma IGF-1 in all tested species, in general proportion to dose level/chronicity and D*R affinity. D1 has no affinity for the prolactin receptor.
Although not required under ICH S6, genotoxicity studies were performed with D1. D1 was negative in the Ames bacterial reverse mutation assay and human lymphocyte chromosomal aberration assay. Acute, 14- or 28-day, and 6-month chronic toxicology studies were conducted in rats and monkeys. In the rat 6-month study, there was dose-related exposure, suggesting no appreciable antibody response, and the MTD was below doses that elicit decreased IGF-1 in rats. Treatment with D1 caused chronic inflammation at injection sites. Decreased alkaline phosphatase and lymphocytes, vacuolated macrophages in spleen and lymph node, and equivocal proteinuria and nephropathy were observed at the MTD in female rats. The NOAEL was ~2 to 5 times human exposure at the clinical dose. In the monkey 6-month study, D1 decreased IGF-1 during the dosing phase, decreased alkaline phosphatase, and was weakly immunogenic. The NOAEL was less than 1× the human clinical exposure. Finally, D1 did not demonstrate growth promoting potential in various D*-responsive tumor cell lines in xenograft mice.
Questions and Discussion
Was there sufficient information available for evaluation of carcinogenic potential? The genotoxicity studies were not considered useful for this assessment. As previously mentioned, these studies are generally not relevant for biopharmaceutical molecules due to their inherent lack of chemical reactivity. The necessity of conducting a 2-year bioassay was questioned in view of D1’s antagonistic activity regarding growth potential (i.e., decreasing IGF-1) and its demonstrated down-regulation of proliferation in vitro mouse xenograft studies. Further, D1 did not cause proliferative or inflammatory effects other than at the site of injection in the chronic in vivo studies. The lack of a signal in the chronic toxicology studies or in the in vitro growth-promotion studies minimizes concern for a carcinogenic effect of D1.
Several limitations of in vitro growth-promotion models in transformed cells were recognized: the artificial nature of these systems, the fact that cell lines are transformed and may have lost apoptotic capabilities, and the nondefinitive nature of these models. It was also noted that, although quantitative measures of cell proliferation in vivo and in vitro may have potential value, investigation as to the relevance of the binding affinity data, which was done in hepatocytes, to binding in other tissues and the tumor cell lines could be warranted.
In addition to a diminished cause for concern regarding carcinogenic potential, the actual relevance of a rat carcinogenicity bioassay is diminished in view of the low rodent potency of D1. Although the binding affinity in mice was slightly greater than that in rats, it was still low in comparison to human, which also brought into question the utility of knockout mouse models. Even so, there was a theoretical concern regarding the long residence time of the protein analogue in the body, and the complexity of hormonal signaling pathways (with potential compensatory up-regulation in related pathways) also generated uncertainty. Some participants thought that knockout models such as the hRas or p53 model may provide some assessment of tumorigenic potential.
Finally, decreases in lymphocyte counts were observed in the 6-month rat study. Immunotoxicity studies could be conducted to understand if this signals an immunosuppressive effect, which could increase concern regarding enhanced susceptibility to carcinogenicity. However, it was also noted that, in the presence of immunomodulation, carcinogenicity studies are not particularly helpful because a negative result would not diminish concern on this point.
To what extent would the above considerations change if D1’s affinity to D*R was similar across species, so that, in the rat study, exposures were comparable to clinical exposures and were limited by pharmacodynamic effects? If affinities were similar across species, a rodent carcinogenicity bioassay becomes potentially more relevant (although there may be a question as to whether the binding affinity data, which was determined in vitro in hepatocytes, was truly reflective of the in vivo affinity). Likewise, there may be increased value in the use of transgenic mouse models.
What are the implications of having polyethylene glycol (PEG) moieties bound to D1 in order to increase circulating half-life? PEG modification raises the possibilities of elicitation of injection site reactions with many repeated injections, which might compromise the ability to conduct a long-term bioassay, and the potential for renal effects, e.g., the equivocal proteinuria and nephropathy noted above. Bendele and coworkers (1998) noted that PEG-linked proteins could induce renal tubular vacuolation at high doses; however, the change was not associated with alteration of clinical pathology or functional markers. Although the PEG moiety itself does not represent any kind of genotoxic alert, the method of attachment could involve organic linkers, in which case the genetic toxicity studies might be more relevant. If an organic linker were used, genetic toxicology studies might be conducted with the entire molecule or just the linker itself.
Would a change in intended indication, which is for a severe and debilitating disease, to one that was less life-threatening, alter considerations around evaluation of carcinogenic potential? Carcinogenicity bioassays remain of uncertain utility due to the relatively low rodent binding affinity. For the reasons discussed above, overall concern of carcinogenic potential is low because neither the full-length protein nor its catalysis products are likely to be mutagenic, and D1 was shown not to possess growth-promoting activity in vitro or cause proliferation or inflammation in vivo. However, the risk-benefit ratio is potentially increased in the case of broader use situations of longer durations or where there is potential use in younger populations. Some participants felt that long-term pediatric use may be of special concern, as adolescents are thought to be more sensitive to growth signals. Therefore, any literature on the growth promotion potential involving this pathway and experience with other growth modulators in nonclinical studies was thought to be potentially informative.
DISCUSSION
Paradigms for nonclinical safety assessments of pharmaceuticals culminated in the 1997 ICH M3 (R1) and ICH S6 guidelines for small molecules and biopharmaceuticals, respectively. At the time of adoption in 1997, these guidelines were heavily influenced by more than 50 years of nonclinical and clinical experience with small molecule pharmaceuticals, but only an initial decade or so of experience with biopharmaceuticals. The ICH S6 guidelines provided a general framework for considerations in the design of nonclinical safety assessment programs for biopharmaceuticals. Compared with guidance for pharmaceuticals, the ICH S6 approach was appropriately much more flexible and individualized, i.e., “case-by-case.” Flexibility is necessary to allow for the design of rational, scientifically based programs for unique molecules, but it is clear that industry and regulatory scientists can benefit from the doubling of accumulated experience with biopharmaceutical development over the last 10 years. In particular, expectations around defining relevant species, the use of homologues, immunogenicity in nonclinical species, DART testing, and carcinogenic assessment were considered of specific interest, and the case studies described herein were designed to gain some agreement or appreciation for differences in perspective around these subjects. Emphasis was placed on explicit articulation of the scientific rationale for the need (or lack thereof) and design of a study.
The following concepts for nonclinical biopharmaceutical development emerged from discussions of the case studies:
The use of chimpanzees and other great ape species is not recommended for nonclinical studies because of scientific, logistical and ethical issues, including the fact that these species are classified as endangered by IUCN/USESA.
In summary, although the proceedings and discussions of this workshop confirm substantial overlap of strategy and tactics in development of biopharmaceuticals and small molecules, they also expose important aspects that are not shared. Small molecule development is based upon a progression of studies applying nonclinical species and tissues, leading to a ‘proof of principle’ in early clinical trials, and reflecting the assumption that one of the nonclinical species/models will ‘predict’ the human outcome. Biopharmaceuticals, on the other hand, are human proteins (e.g., human growth hormone) and protein constructs (e.g., humanized monoclonal antibodies) designed to interact pharmacologically with specific human targets with minimal immunogenic effect. For biopharmaceuticals, humans (ethics aside) are the only appropriate species for efficacy and safety evaluations, and any other species, at best, a lesser approximation. As is demonstrated by consideration of the four case studies discussed herein, nonclinical development of biopharmaceuticals must be considered largely on a case-by-case basis driven by an understanding of the science and considerations of human safety.
Footnotes
Acknowledgements
The authors appreciate the following scientists for supporting the concept and conduct of the workshop and for their critical review of the manuscript: Drs. Andrew Dahlem (Eli Lilly & Company), David Jacobson-Kram (FDA), Ruth Lightfoot-Dunn (Amgen), Jeanine Bussiere (Amgen), Hanan Ghantous (FDA), Timothy MacLachlan (Genzyme), John Vahle (Eli Lilly & Company), Bill Breslin (Eli Lilly & Company), and George Treacy (Centocor).
Disclaimer: The views expressed in this document are those of the authors and do not necessarily reflect the views or policies of the U.S. Food and Drug Administration. Also, no official support or endorsement by this agency is intended or should be inferred.