Abstract
Introduction
The administration of the first dose of a novel pharmaceutical compound to humans is the culmination of many years of efforts by many people. There is often a feeling of relief that the compound has made it this far and a feeling of great optimism and anticipation. However, it also can be a time of great trepidation because it is the first opportunity to determine the accuracy of the projections of the compound’s efficacy, ADME (absorption, distribution, metabolism, and elimination) properties, and most important, safety in humans. The administration of the first dose to humans usually results in a celebration by the development project team, although the first few dose levels of the initial clinical trial are usually uneventful and not much can be learned regarding the compound’s efficacy, ADME properties, and safety due to the low doses used in these initial dose cohorts. This is particularly true for nononcology compounds that are typically evaluated first in normal healthy volunteers (NHV) who would be expected to derive no benefit from the compound and thus generally receive a first-in-human (FIH) dose that is not anticipated to have pharmacologic activity.
Unfortunately, in some instances, things do not go as anticipated. Sometimes it is learned from these early dose cohorts that the projections of the pharmacokinetics (PK) of the compound are wrong, or that the compound will not be as efficacious in humans as in animals. Sometimes the compound is found to have a side-effect profile in humans that was not predicted from the animal toxicology studies. Usually such side effects are relatively benign, consisting of headaches, nausea, drowsiness and other nuisance events which would have been difficult to appreciate in animal studies. On very rare occasions, the side effects may be more severe and, even more rarely, life-threatening. One such example of the latter was the “TeGenero Incident” (Expert Group on Phase One Clinical Trials 2006; Gambrill 2006). Serious, life-threatening adverse events were observed in all dosed subjects after the administration of TGN1412, a new class of monoclonal antibody (mAb) with a stimulatory mode of action on a subset of T cells known as regulatory T cells (Tregs; Expert Group on Phase One Clinical Trials 2006). This antibody was a humanized IgG4κ mAb that was derived from the precursor mouse mAb 5.11A1 by grafting of the complementarity-determining regions (CDR). TGN1412 binds to CD28 and activates T cells without need for T cell receptor (TCR) preactivation, resulting in polyclonal T cell expansion and activation and concentration-dependent IL-2 production. As such, TGN1412 was termed a “superagonist.” The intent was to develop TGN1412 as a treatment for B cell chronic lymphocytic leukemia, in which T cells are deficient, and for autoimmune diseases, such as rheumatoid arthritis, in which Treg cell expansion might be beneficial.
The TeGenero incident spurred a subsequent investigation by an expert scientific group (ESG) on Phase 1 Clinical Trials, appointed by the Medicines and Healthcare Products Regulatory Agency (MHRA) and under the leadership of Professor Gordon Duff, which culminated in a report known as “the Duff Report.”
The objective of this article is to review the TeGenero incident and the Duff Report and to comment on whether the TeGenero incident was predictable and/or avoidable. A separate article discusses the likely effects of the “Guideline on Strategies to Identify and Mitigate Risks for First-In-Human Clinical Trials with Investigational Medicinal Products,” which was released by the European Union (EU) European Medicines Agency (EMEA) Committee for Medicinal Products for Human Use (CHMP; Milton and Horvath 2009). The authors have both had primary and shared responsibility for the preclinical development of several mAbs directed against T cell surface antigens and for determining the FIH doses and their anticipated effects in Phase 1 trials conducted in NHVs. It is the opinion of the authors that sufficient information about cytokine release as a dose-limiting toxicologic effect of other agonistic anti–T cell mAbs was available in the public domain to have strongly cautioned against dosing NHVs with such a product, especially when designed as a superagonist. Furthermore, historical experience with other anti–T cell mAbs and data available to TeGenero prior to the FIH trial suggested that first-dose reactions were likely to occur. The opinions expressed in this article are those of the authors and should not be considered representative of their employers, past or present.
The TeGenero Incident
At 08:00 on March 13, 2006, the first healthy volunteer was administered TGN1412 as a three– to six–minute intravenous infusion at the PAREXEL Clinical Pharmacology Unit within leased space at the Northwick Park Hospital in North London, UK. The dose administered was 0.1 mg/kg. Subsequently, other subjects were administered TGN1412 or placebo at ten–minute intervals. By 09:10, six volunteers had been administered TGN1412, and two volunteers had been administered a placebo. Within ninety minutes, those who had received TGN1412 had a systemic inflammatory response that was characterized by a rapid induction of proinflammatory cytokines and accompanied by headache, myalgia, nausea, diarrhea, erythema, vasodilatation, and hypotension (Suntharalingam et al. 2006). Between twelve and sixteen hours postdose, they became critically ill, with pulmonary infiltrates and lung injury, renal failure, and disseminated intravascular coagulation (DIC). Severe and unexpected depletion of lymphocytes and monocytes occurred within eight hours, reaching a nadir at twenty-four hours. By midnight, they had been transferred to an intensive care unit, where they received intensive cardio-pulmonary support (including dialysis), high-dose methylprednisolone, and an anti–interleukin–2 receptor antagonist antibody. Prolonged cardiovascular shock and acute respiratory distress syndrome developed in two patients, who required intensive organ support for eight and sixteen days (Suntharalingam et al. 2006). It was soon concluded that TGN1412 had caused a “cytokine storm.” Although all six volunteers survived, the long-term prognosis for these subjects is unknown. One of these patients has since had all of his toes and the tips of several fingers amputated (Gibb 2008).
The incident was reported to the MHRA on the afternoon of March 14, 2006 (MHRA 2006b). The MHRA reacted rapidly, immediately suspending the Clinical Trial Authorization (CTA), confirming that TGN1412 was not in use in any other trial anywhere in the world, alerting international pharmaceutical regulatory authorities of the events in case any similar product of this class might be in use, and sending a team of inspectors to the unit in Northwick Park to secure documents, samples, and other evidence (in liaison with the Metropolitan Police). The news of these tragic events was broken by U.K. print and broadcast media and was soon known throughout the world. The British Broadcasting Corporation (BBC) filed Freedom of Information (FOI) request(s) (NB122/6) on March 17, 2006. In response, the MHRA moved rapidly and issued an interim report on the incident on April 6, 2006 (MHRA 2006b). They also made many documents available to the scientific community and the public (MHRA 2006a). These documents included the TGN1412 clinical trial protocol, Investigator Brochure (IB), Investigational Medicinal Product Dossier (IMPD), and Assessment Report. Within these documents, some of the biological rationale, planned assays on samples, references, laboratories, and other names—including those of assessors—were blanked out by reason of commercial or other confidentiality. Very rarely, if ever, are these documents made available to the public, and their availability allowed a very clear insight into the material available to the regulatory authorities at the time that they approved the clinical trial protocol. It should be noted that while preclinical data were summarized within sponsor-generated documents, individual nonclinical study reports were not available to the MHRA as there is currently no requirement to submit them. Because it is expected that the IMPD will not be a large document, the nonclinical information was summarized in a few pages. Similarly, the nonclinical information in the Investigator’s Brochure (IB) was presented in summarized form only (International Conference on Harmonisation [ICH] 1996). In contrast, the U.S. Food and Drug Administration (FDA) requires all study reports to be submitted as part of an Investigational New Drug (IND) application. Reliance on a sponsor-generated summary of nonclinical data may be a weakness in the clinical trial approval process in the EU.
The MHRA quickly identified and evaluated potential causes of the adverse reactions. These included a dosing error, an error in the formulation or dilution of the mAb, contamination of the mAb with some other substance, and a previously unknown biological effect on humans that did not arise in any of the animal-testing phases. Inspections were rapidly performed, and it was concluded that the adverse incidents did not involve contaminants or errors in the manufacture of TGN1412 or in its formulation, dilution, or administration to trial participants. The MHRA therefore initially concluded that an unpredicted biological action of the mAb in humans was the most likely cause of the adverse reactions in the trial participants. They then established an ESG under the leadership of Professor Gordon Duff (professor of molecular medicine at Sheffield University). The terms of reference of the ESG were to consider what may be necessary in the transition from preclinical to FIH Phase 1 studies and in the design of these trials, with specific reference to biological molecules with novel mechanisms of action; new agents with a highly species-specific action; and new products directed toward immune system targets. The MHRA also adopted (May 1, 2006) a pre-cautionary approach to FIH (Phase 1) trials of any mAb (regardless of intended target) or other new molecules that target the immune system via a novel mechanism. They mandated that such trials would not be sanctioned by the United Kingdom’s Competent Authority without having had additional expert opinion from the Commission on Human Medicines (CHM) on whether a cytokine storm might be caused by the novel agent. In parallel with the ESG, a working party chaired by Professor Stephen Senn was established by the Royal Statistical Society (RSS) to review statistical aspects of FIH studies. By the end of July 2006, the ESG’s interim report and supporting papers were released for public consultation, and the final report and supporting papers were published on December 7, 2006 (Expert Group on Phase One Clinical Trials 2006). The final report included stakeholder submissions from the RSS and the Association of the British Pharmaceutical Industry (ABPI)/BioIndustry Association (BIA) Joint Early Stage Clinical Trial Taskforce (Early Stage Clinical Trial Taskforce 2007; Senn et al. 2007). A separate report was issued by the Royal Statistical Society (Senn et al. 2007; RSS 2007).
The Duff Report concluded that the preclinical development studies that were performed with TGN1412 did not predict a safe dose for use in humans, even though current regulatory requirements were met. The RSS advocated for a greater role for statistical analysis in the nonclinical evaluation of novel products, even to the standards laid out in ICH E9 (ICH 1998). The ABPI/BIA advocated for adoption of a pharmacologically based method of establishing the FIH starting dose, rather than traditional toxicology-based algorithms derived from the no observed (toxicologic) effect (dose) level (NOEL) or the no observable adverse effect (dose) level (NOAEL) in animal toxicology studies. The proposed method involves estimation of the minimum anticipated biological effect (dose) level (MABEL) for humans and selecting an FIH dose that does not exceed it.
An Opinion on the “Predictiveness” of the Nonclinical Data for TGN1412
In an excellent review, Tabrizi and Roskos point out that appropriate preclinical safety testing of mAbs is usually highly predictive of human clinical PK, PD, and safety (Tabrizi and Roskos 2007). They review the clinical adverse reactions associated with fourteen approved mAbs, ten of which carry in their labels warnings of the potential for infusion-related reactions. While some of these infusion reactions are related to immunogenicity and occur on secondary exposures, for several of these mAbs (e.g., alemtuzumab, muromonomab-CD3, and rituximab), the infusion reactions are generally most severe on the first exposure and are directly related to the mechanism of action and desired pharmacologic effect (Wing et al. 1996; Winkler et al. 1999; Gaston et al. 1991). Yet much has been written regarding the “nonpredictiveness” of the TGN1412 preclinical data (Expert Group on Phase One Clinical Trials 2006; St. Clair 2008), and in fact, the Duff Report concluded that the “pre-clinical development studies that were performed with TGN1412 did not predict a safe dose for use in humans, even though current regulatory requirements were met.” Within the TeGenero IMPD, it was stated that although there was “no sign of a first-dose cytokine release syndrome … in pre-clinical studies … it cannot be excluded that a first-dose effect on cytokine release may occur in humans.” In addition, within the IB, it was stated that “activation of T lymphocytes by … the anti-CD3 monoclonal antibody OKT3, elicits a ‘cytokine storm’ characterized by an increase in systemic inflammatory mediators.” Clearly, then, TeGenero had considered T cell activation as a potential outcome, but had they designed their preclinical studies to detect it?
OKT3® (muromonomab–CD3), a mouse IgG2a mAb directed against human CD3, is the prototypical anti–T cell mAb and is approved for use as an immunosuppressive agent in renal transplant patients. OKT3 is primarily recognized for its T cell depleting properties, which are mediated in part, if not primarily, through lymphocytolysis attributable to antibody-dependent cell-mediated cytolysis (ADCC) and complement-dependent cytolysis (CDC; Alegre et al. 1990), both of which are dependent on the Fc portion. Other mechanisms of depletion of circulating T cells may include cell coating leading to negative regulatory signals, cross-linking of surface antigens leading to apoptosis, cross-linking of target cells with cytolytic effector cells and/or margination, and emigration from the vascular compartment in response to endothelial activation and/or tissue inflammatory signals (Hale et al. 1983). With the emphasis on clinical use of OKT3 as an immunosuppressive, it is often forgotten that OKT3 is a potent T cell mitogen (Hale et al. 1983) and can activate T cells without costimulation of CD28, resulting in lymphocyte activation, T cell proliferation, and cytokine (IL-2) release. These attributes are also dependent on the Fc portion and its interactions with Fc receptors (FcRs) on monocytes (Van Wauwe, De Mey, and Goossens 1980). OKT3 is routinely used as a costimulatory agent (similar to activation of CD28 by TGN1412) for in vitro activation of T cells in mixed lymphocyte reactions (MLR), where immobilization of the Fc portion is required for optimal activity. Thus, while the desired effect of OKT3 is T cell depletion, it is the dose-dependent, Fc-mediated cytokine release that characterizes the “first-dose reaction” (Gaston et al. 1991), establishes dose-limiting toxicity (DLT), and has resulted in a “black-box” warning of the possibility of cytokine release syndrome (CRS) on its label. At the 5 mg clinical dose (~0.07 mg/kg) for OKT3, virtually all patients experience CRS (Chatenoud, Ferran, and Bach 1990). This dose is comparable to the FIH dose (0.1 mg/kg) that was used for TGN1412.
Interestingly, cytokine release and T cell depletion were not observed when OKT3 was tested in rhesus monkeys in toxicology studies supportive of marketing approval. This is because OKT3 is not pharmacologically active (does not bind to CD3) in this species. It should be remembered that the approval of OKT3 in 1986 and the submitted nonclinical studies with OKT3 predate by almost a decade the implementation of the ICH S6 requirements for testing of biologics in a pharmacologically relevant species. Several years after approval, when administered to chimpanzees, a pharmacologically relevant species, OKT3 was demonstrated to induce T cell depletion, accompanied by fever and cytokine release, but not overt CRS (Rao et al. 1991). T cell depletion occurred within two hours; circulating T cell counts returned to baseline values after one week and were mildly elevated (~2x) after two weeks. Thus, for OKT3, the preclinical results that supported marketing approval were not predictive of the clinical results (because the rhesus monkeys were not relevant), but testing in a relevant species (chimpanzees) would have allowed “predictiveness.”
The Fc-dependent CRS accompanying OKT3 dosing is a toxicologic liability that has fostered numerous attempts to develop non–FcR–binding (Fc–mutated) variants of OKT3 (Alegre et al. 1992) and other “kinder, gentler” anti-CD3 mAbs. The most relevant of these for comparison to TGN1412 is visilizumab (HuM291 or Nuvion®), a humanized anti-CD3 IgG2 mAb with an Fc mutation intended to reduce FcR binding. Visilizumab displays reduced mitogenicity in vitro relative to OKT3 but maintains immunosuppressive (T cell depleting) properties (Cole et al. 1999). It is of interest that visilizumab was described at the time of initial development (ca. 1999) as a T cell “superantibody,” much as TGN1412 was described as a superagonist. Visilizumab cross-reacts only with chimpanzees and was tested in this species at doses up to ~0.17 mg/kg (10 mg/animal), which was the NOAEL (Hsu et al. 1999; Klingbeil and Hsu 1999). It is considered unacceptable to induce overt toxicity in chimpanzees, and so the maximum dose tested was low. Even so, this dose was sufficient to produce the desired pharmacologic effect (complete depletion of circulating T cells) and was accompanied by mild cytokine release, without CRS. The minimum pharmacologically active dose (PAD) in chimpanzees was ≤ ~0.0017 mg/kg, the lowest dose tested. At this dose, T cell depletion occurred within one hour; circulating T cell counts returned to baseline values after one week and were mildly elevated after two weeks. The FIH dose that was selected for HuM291 was 0.000015 mg/kg, which represented ~1/10,000th of the NOAEL or ~1/100th of the PAD for chimpanzees. This dose was administered to renal allograft patients and produced very little T cell depletion. At 0.0015 mg/kg, complete T cell depletion occurred within two hours. Dose escalation was completed up to 0.015 mg/kg, at which point DLT was reached in the form of CRS accompanied by T cell depletion that returned to baseline levels by about two weeks (Norman et al. 2000). Thus, DLT was observed in humans at a dose (on a mg/kg basis) that was ~1/10th of the NOAEL dose in chimpanzees, suggesting that the relative potency of HuM291 in chimpanzees was ≤ ~1/ 10th of that for humans. That is, for HuM291, the preclinical testing was predictive of the clinical results but not for the doses at which they would occur. After these results, the development of visilizumab was discontinued, with CRS being cited as the reason for such. It should be noted that the FIH dose for TGN1412 (0.1 mg/kg) was ~7000-fold higher than the FIH dose for HuM291.
For the preclinical testing of TGN1412 to have been capable of predicting the human results, the safety studies should have been designed to evaluate the consequences of the anticipated clinical events. This required selection of pharmacologically relevant species and appropriate end points and time points. As clearly demonstrated by both OKT3 and HuM291, testing of TGN1412 in a relevant nonhuman primate species should be expected to reveal T cell activation and moderate cytokine release, but not CRS, accompanied by early, rapid T cell depletion followed by recovery and then elevation of T cell counts by about two weeks. This effect should be an Fc-dependent phenomenon. Indeed, immobilization (fixation) of the Fc portion of TGN1412 was subsequently demonstrated to be an essential requirement for robust in vitro activation, proliferation, and T cell cytokine release from human T cells (Stebbings et al. 2007). Since publication of this finding, many sponsors have reported being requested by regulatory agencies to assess whether their biotherapeutic compounds are capable of eliciting in vitro cytokine release, and this topic has been featured in recent industry and FDA meetings (BioSafe General Membership Meeting 2008; FDA CDER Immunotoxicology Subcommittee Meeting 2008). Although there is not yet consensus on the preferred methods of testing or interpretation of the results, there seemed to be general agreement that such testing should be considered on a case-by-case basis.
TGN1412 was demonstrated by TeGenero to bind to cynomolgus monkey CD28 in vitro and in vivo and pharmacologic relevancy of this species was thus inferred, although binding of the Fc portion to FcRs in monkeys was not evaluated. Thus, it is possible that differences in Fc-FcR interactions could render results in monkeys unreliable. If so, one would anticipate absence of T cell activation, depletion, cytokine release, and/ or proliferation. When TGN1412 was administered to cynomolgus monkeys in the repeated-dose toxicology study at four weekly doses of 5 and 50 mg/kg, T cell numbers and cytokine concentrations were monitored. TGN1412 induced minimal cytokine release in monkeys, which is consistent with in vitro results with monkey T cells even after Fc immobilization (Stebbings et al. 2007) and with the experience with OKT3 and HuM291 in chimpanzees. Other investigators confirmed this finding with T cells from rhesus as well as cynomolgus monkeys (Waibler et al. 2008). Thus, nonhuman primates appear not to “predict” cytokine release.
If cytokine release is not a reliable predictor in nonhuman primates, then that leaves T cell depletion and rebound proliferation as potential indicators of the adverse events observed in man. Unfortunately, TeGenero’s studies were not designed to reveal T cell depletion because T cell counts were evaluated only once weekly prior to each day of dosing, so that one cannot assess whether rapid T cell depletion occurred after the first dose. In actuality, for both doses of TGN1412, T cell counts were less than baseline values at one week and peaked at modest (~2-fold) increases at two weeks, despite continued dosing and minimal immunogenicity. Nearly identical results had been observed in a dose-escalating study of 5, 10, 25, and 50 mg/kg TGN1412 administered to monkeys at weekly intervals. Thus, TGN1412 displayed no differential dose response, no incremental responses to dose escalation and only modest lymphoproliferation. The increase in T cell counts after two weeks was only ~1/10th of that seen after three days with JJ316 in rats (see below). Collectively, these results suggest that TGN1412 lacked CD28 superagonist properties in monkeys but do not allow us to conclude that TGN1412 had no effect on monkey T cells. In fact, the lack of cytokine release and delayed, modest increases in T cell counts are consistent with OKT3 and HuM291 in chimpanzees, suggesting the possibility that early, rapid T cell depletion (and thus T cell activation) might have also occurred with TGN1412. If one compares circulating T cell counts from monkeys receiving TGN1412 (Figure 42 in the IMPD) to those for subjects in the Phase 1 trial (Figure 3B in Suntharalingam et al. 2006), it becomes apparent that data from common time points is virtually identical. That is, after rapid, complete T cell depletion, the healthy volunteers had modest (~2-fold) elevations in T cell counts at two weeks. Unfortunately, when TGN1412 was administered to cynomolgus monkeys at doses from 0.1 to 50 mg/kg in a follow-up investigation (Stebbings et al. 2007), the early effects on T cell counts, if evaluated, were not reported.
There are other data to suggest that T cell activation and depletion was likely to occur with TGN1412. TeGenero had conducted numerous pharmacology studies in rats with a mouse antirat CD28 IgG1 mAb, JJ316, which also had Fc-dependent superagonist properties (Tacke et al. 1997; Lin and Hunig 2003; Beyersdorf 2006). As is common practice for such studies, T cell counts were not evaluated until several days after dosing, when T cell proliferation was demonstrated. In fact, within three days, T cell counts were increased ~20-fold, accompanied by remarkable splenomegaly (~6-fold increase in spleen size). These results are in marked contrast to the modest, delayed increases in T cell counts seen with TGN1412 in monkeys (and humans). The lymphoproliferative effects of JJ316 were therefore consistent with the putative mechanism of action of a superagonist and with the expectations of those developing TGN1412. However, JJ316 was subsequently shown to induce complete depletion of circulating T cells almost instantaneously (within 2 minutes) after IV administration (Müller et al. 2008). This change was attributed to arrest of T cells within lymphoid organs, where they became strongly activated but did not produce elevated plasma cytokine concentrations. Even stronger evidence for T cell activation and T cell depletion was available for mAb 5.11A1, the mouse IgG2 mAb from which TGN1412 was derived by grafting of the CDR (Legrand et al. 2006). This mAb was administered to H2d Rag2−/−γc −/− mice that had been irradiated and reconstituted with human CD34+ fetal liver (stem) cells to allow development of human immune systems (HIS [Balb-Rag/γ] mice) characterized by the presence of all major human lymphoid cellular compartments, including T cells expressing human CD28 and myeloid cells expressing human FcRs. Doses of 0.01 to 0.03 mg/mouse (~0.3–10 mg/kg) induced proliferation of human Treg cells in the thymus after six days, as was emphasized by the authors. But less prominently discussed is the fact that this precursor of TGN1412 also induced rapid, profound depletion of circulating T cells that persisted through sixty days. The finding of T cell depletion was not described in the TeGenero IMPD, which was dated December 19, 2005, although this article was submitted for review on January 17, 2006, accepted for publication on February 17, 2006, and published on March 2, 2006. Eleven days later, TGN1412 was administered to normal healthy volunteers. It has been suggested that regulatory authorities may have reconsidered the wisdom of dosing NHVs with TGN1412 if they were aware of this information (St. Clair 2008).
In the opinion of the present authors, the historical experience with other anti–T cell mAbs and data available to TeGenero prior to the Phase 1 trial suggested that first-dose reactions were likely to occur at the FIH dose. Admittedly, the preclinical studies conducted for TGN1412 did not predict a cytokine storm, as was also true for OKT3 and HuM291, both of which cause intense cytokine release in humans. The absence of CRS-like findings in monkeys and rats should not be unexpected, as there is also considerable phylogenetic variation in species sensitivity to cytokine-releasing stimuli. For example, while chimpanzees and humans are highly sensitive to endotoxin or bacterial challenge, baboons, monkeys, and mice are highly resistant (Smirnova et al. 2000), related in part to differential expression and function of toll-like receptors (TLRs). Other mechanisms may explain differences in T cell activation and cytokine release, such as differential expression of immune regulatory molecules (e.g., Siglecs) between humans and nonhuman primates, including chimpanzees (Nguyen et al. 2006). It should also be recognized that many of the adverse events associated with cytokine release (e.g., headache, nausea, myalgia) are correctly termed symptoms (feelings reported by the subject) and that in animal studies, it is only possible to identify clinical signs (changes recorded by the observer). Because of the absence of CRS in nonhuman primates, careful monitoring of T cell counts can be useful to predict T cell activation and T cell depletion that will occur in humans. For TGN1412, there was ample preclinical evidence available to predict that such would occur in humans. As such, the Phase 1 clinical trial should have been designed to emphasize caution and a concern for the safety of the healthy volunteers rather than to facilitate efficient dosing and sample collection. Thus, in the opinion of these authors, the TeGenero incident was an avoidable, rather than unforeseeable, event.
Duff Report Recommendations
The Duff Report provided twenty-two recommendations, several of which either directly affect the role of the toxicologist or are of interest to the pharmacologists, toxicologists, and pathologists responsible for establishing safety and proposing FIH doses. We shall review some of those recommendations and discuss how they might affect preclinical groups in preparing for FIH clinical studies.
Conduct of Preclinical Development Programs
1. Decisions on the strategy for preclinical development of a new medicine and the experimental approaches used to assemble information relevant to the safety of first-in-man clinical trials must be science-based, made, and justified case-by-case by individuals with appropriate training.
While acknowledging the role of ICH S6 and M3(R1) in pre-clinical development, the ESG questioned their applicability to “high-risk” products and in particular to FIH clinical trials. The concept of high-risk products and the eventual removal of this term from the Guideline on Strategies to Identify and Mitigate Risks for First-In-Human Clinical Trials with Investigational Medicinal Products has been discussed by Milton and Horvath (2009). Whereas the draft guideline emphasized high-risk products, the final guideline does not single such products out. In essence, the logical argument was that some products can be assumed a priori to be inherently riskier, and others inherently less risky, but it is only after FIH dosing that one can reliably confirm the degree of risk. Therefore, no product should be assumed to be “low risk,” and all products should be assessed with equal concern for subject safety.
The Duff Report also called for a new guideline related to high-risk medicines to be created at the EU level, with particular respect to FIH studies, and for a subsequent guidance to be created at an international level (presumably by ICH). The CHMP has indeed created such a guideline (Expert Group on Phase One Clinical Trials 2006), which is discussed in a companion article to this current article (Milton and Horvath 2009), but there has not yet been any clear initiative by the ICH on this topic. However, at the ICH meeting that was held between May 31 and June 5, 2008, there was an informal meeting to discuss the need to update the ICH S6 Guideline (ICH 2008b). All of the preclinical experts were of the view that the main principles of the existing guideline that are based on a flexible, case-by-case, science-based approach still hold. However, it was concluded that significant scientific development over the past several years has made an update of the guideline necessary. The Steering Committee therefore agreed to create an expert working group to address the scientific advancements in the field as they relate to the current ICH guideline.
The acknowledgement that preclinical development should be case by case accentuates the central tenant of ICH S6. For those experienced in the design of preclinical development programs, particularly for biotherapeutics, this is welcomed as confirmation of their approach to preclinical development. However, it is unclear how it will be judged that an individual has appropriate training or not and whether this caveat applies to both the sponsor’s toxicologist and to the regulatory reviewer. Presumably, it would be best if both possessed equivalent expertise, as that would result in the greatest likelihood of potential safety concerns being appropriately considered. The principles and evidence that support the design and conduct of preclinical studies in a manner that permits effective and safe FIH studies of potential new medicines in humans were reviewed by Greaves, Williams, and Eve (2004) several years before the TeGenero incident. Based on the historical data available for other mAbs directed against T cell surface antigens, one could argue (see above) that some degree of inexperience or unfamiliarity with this class of products played a role in the TeGenero event. It should be noted that the ICH M3(R1) document that was cited in the Duff Report is currently being updated (ICH M3(R2)) and was recently released as a Step 3 document for review (ICH 2000, 2008a).
2. The regulatory process for first-in-man trials of higher risk agents and advanced medicinal products based on innovative technologies should be subject to regular review.
The ESG additionally suggested that it is important that the regulation of medicines is science-based and that the regulatory process keeps pace with, and is appropriate for, changes in science and technology. These seem like reasonable goals, although it is not clear what mechanism(s) might be used for periodic review of regulatory processes to ensure that they remain current. For at least the FDA, staying current with rapid changes in technology and our understanding of complex biological processes is an ongoing challenge (FDA Science Board 2007). One might have some concerns that changes in regulatory processes during conduct of the IND– or FIH–enabling nonclinical studies might affect a preclinical program, potentially leading to delays.
3. Developers of medicines, research funding bodies and regulatory authorities should expedite the collection of information from unpublished preclinical studies relevant to the safety of human exposure. As a first step this should focus on preclinical reports of results that signal potential danger to humans with the use of a higher risk medicine or group of medicines. This should provide a platform for information sharing between regulators at EU and international level, e.g. in the form of a confidential database. The submission of such data to a database by the investigators is strongly encouraged.
It was recommended that the ultimate goal should be an open-access database, although to avoid the risk of delay in sharing critical safety information, such a database may initially be confidential. Regular review of the reason(s) for delaying open access to this preclinical information was suggested. This recommendation appears to be intended to put in place a formal process for informing regulatory reviewers of potential concerns for high-risk products—essentially an institutional means of pooling experiences. In principle, this seems useful, although in practice, no database will likely supplant the wisdom or instinct gained through applied experience on the part of reviewers. It is further unclear how data generated by pharmaceutical companies on proprietary products would be handled or who would organize such a database. At its extreme, this recommendation could be taken as suggesting that data from compounds that never enter development should be submitted and that such information would ultimately make its way into the public domain.
4. Regulatory authorities should consider ways to expedite the sharing of safety information on phase one clinical trials between regulators within the EU and worldwide. This should certainly include information on first-in-man experience with higher risk medicines. Trials with negative safety outcomes should be included. This database might be widened to include products that may not currently be perceived as high risk, or trials conducted later in development, that suggest a strong warning for first-in-man use of similar products.
In the EU, the suggestion was to build on the existing (since 2004) European Drug Regulatory Authorities (EUDRA) databases, EudraCT for clinical trials, and EudraVigilance for reporting of suspected unexpected serious adverse reactions (SUSARs; European Commission 2001). If this were to be implemented, it would likely cover all phases of clinical development and not just Phase 1. Again, the logistics and timing of submission and eventual public disclosure of this proprietary information would be of significant concern to sponsors. Such a practice for products in early development might understandably cause more concern than the practice of making similar information available for marketed products. On the topic of SUSARs, it is questionable whether the cytokine storm of the TeGenero incident would qualify as a “suspected unexpected” event (though it was no doubt serious), because the TeGenero IB and IMPD documents clearly describe a cytokine storm as a possible outcome of exuberant T cell activation. Indeed, CRS associated with mild flu-like to serious life-threatening reactions and even death is common in patients dosed with mAbs directed against lymphocyte surface antigens such as CD3, CD20, and CD52 (Liedert et al. 2007). It is difficult to believe that an international database would successfully help to avoid such an adverse event if more than twenty-five years of collective development experience with anti–T cell mAbs is publicly available yet largely ignored.
5. More communication is strongly recommended between developers and the regulator at an earlier stage before an application is filed, especially for higher risk agents, to ensure that there is time for an appropriate consideration of any safety concerns without introducing undue delay to product development. Ways to increase communication between the regulator and research ethics committees should also be considered.
This may seem to be an obvious suggestion, especially as such communications often occur between sponsors and the FDA in the form of pre-pre-IND meetings and pre-IND meetings, and this was formerly the advocated position of the FDA Center for Biologics Evaluation and Research (CBER). After the reorganization of the CBER and assignment of biologics programs to Center for Drugs Evaluation and Research (CDER), many sponsors have found the willingness of different FDA divisions to engage in these early discussions to be quite variable. This is consistent with predictions of decreased opportunities for direct communication between sponsor development teams and FDA review teams (Schwieterman 2006). Communication is at least as difficult within the EU, where the process for meeting with regulators (obtaining scientific advice) is markedly different, varies from country to country, and is lengthy and expensive (EMEA 2007). Perhaps the recommendation from the Duff Report will allow discussion as to why early and open communication with regulatory agencies is in the best interest of all parties and, most importantly, FIH subjects.
In lieu of direct dialogue, the ESG recommended that the review time for the FIH Clinical Trial Application for high-risk medicines be extended and that the sponsor inform the regulators six weeks in advance of such a submission, if no presubmission meeting has been held. This additional time was proposed to allow the reviewer to implement two more recommendations:
6. For appraisal of applications for trials of higher risk agents, as defined by the nature of the agent, its degree of novelty, its intended pharmacological target, and its intended recipient, the regulator should have access to additional opinion from independent, specialist experts with research knowledge of their fields.
7. An Expert Advisory Group (EAG) of the Commission on Human Medicines, or a similar body, might undertake this role with a core membership of appropriate experts and the ability to co-opt additional expertise as the need dictates.
The methods by which these additional opinions and or reviews would occur and the timelines they would require were not discussed. However, sensitive to the fact that they would likely prolong the review process, a final recommendation was made:
8. Consideration should be given to introducing some flexibility in the time-scale of clinical trial appraisal in exceptional cases of unusual complexity.
Selection of the FIH Dose for Clinical Trials
One aspect of the Duff Report that should be of keen interest to the toxicologist is the recommendations regarding calculation of the starting dose for the FIH study. To put these recommendations in context, it is important to understand industry practices for establishing FIH doses. Calculation of the FIH dose has traditionally been based on toxicologic testing in animals, followed by application of an empirically derived algorithm to the NOEL. One of the earliest methods suggested the FIH dose, on a mg/kg basis (no allometric scaling), should be either 1/10th of the rodent NOEL or 1/6th of the nonrodent NOEL from four- or thirteen-week studies, whichever was lower (AFDO 1959). A variation on this method used the same “safety factors” but relied on allometric scaling (mg/m2 basis) after determining the severely toxic dose (STD) for rodents and the highest nonseverely toxic dose (HNSTD) for nonrodents (DeGeorge 1997). More recently, FDA/CDER had focused on the NOAEL and suggested the maximum recommended or safe starting dose (MRSD or SSD) to be 1/10th of the NOAEL dose in the most sensitive species (rodent or nonrodent), on a mg/m2 basis (FDA 2005a). In this guidance document, it is stated that animal NOAEL doses could be converted to human equivalent doses (HED) by normalizing to body surface area, the implication being that on a mg/m2 basis, the dose that has no adverse effects in animals will have no adverse effects in humans. These NOEL- and NOAEL-based algorithms have their origin in small molecule (drug) toxicity testing, where toxicity is often associated with metabolites and “off-target” effects. They tend not to emphasize pharmacologic effects, so that toxicology is not placed in the context of pharmacology.
Some efforts to incorporate pharmacologic effects in FIH dose selection have been made. As early as 1977, the ABPI and BIA suggested that the FIH dose should not exceed the human projected therapeutic dose (ABPI/BIA 1977). While not addressing dose selection, ICH S6 clearly states that for biotherapeutics, it is important that the animal species be pharmacologically relevant because “toxicology” may reflect “on-target” effects and/or exaggerated pharmacology. The case for pharmacological relevance of the test species is furthered in the FDA/CDER guidance on safe starting dose (FDA 2005a), where it is stated that the MRSD should be modified (potentially lowered) based on calculation of a pharmacologically active dose (PAD); that is, a pharmacologic HED based on animal data. The degree to which this method would be “predictive” is of course dependent on the degree to which the pharmacology in animals matches that in humans. Unfortunately, discussion of the method for “selection of PAD [was] beyond the scope of this guidance,” as was discussion of a method for estimating the pharmacologic effects of the FIH dose. It is curious then that several guidance documents have been issued for FIH doses that are not pharmacologically active. The FDA/CDER proposed that the FIH dose for a trial intended to evaluate a pharmacologically inactive “micro-dose” could be 1/100th of the dose calculated to yield a pharmacologic effect, or 1/100th of the NOAEL (FDA 2006). For pharmacologically relevant doses, the FIH dose could be 1/50th of the rodent or nonrodent NOAEL from a two-week study, whichever is lower. The EMEA/CPMP also proposed for microdosing a safety factor of 1/100th of the dose calculated to yield a pharmacologic effect, on a mg/m2 basis, and suggested that such a dose should be 1/1000th of the minimal toxic effect dose (MTED) from a single-dose toxicology study (EMEA 2004). Even more recently, the draft regulatory guidance, ICHM3(R2), attempts to address the issue of the selection of the safe starting dose with several different approaches but an emphasis on the NOAEL (ICH 2008a). For any of these methods, it should be noted that use of the terms safe or recommended with respect to starting dose, or designation of a micro-dose as pharmacologically “active” or “inactive,” should not be taken literally. Predictions are being made based on the best available data using certain assumptions, particularly regarding the toxicological consequences of the pharmacological action of the drug. Whether the FIH dose for a compound is truly inactive, active, or safe cannot be known until after that dose has been administered to humans.
Clearly, there are multiple, not necessarily conflicting, methods for establishing FIH doses, including others not discussed here. ICHM3(R2) is intended to recommend international standards for, and promote harmonization of, the nonclinical safety studies recommended to support human clinical trials of a given scope and duration and marketing authorization. Whether this guidance, in its finalized form, will provide harmonization or clarity remains to be seen, as regional guidelines may not be in strict alignment with the finalized ICH guidance. Regardless of the method chosen, no algorithm for selecting the FIH dose should be blindly accepted; the dosing rationale must be justified for every product.
Against this background of existing practices for FIH dose selection, the recommendations of the Duff Report for FIH dose selection were as follows:
9. Special consideration should be given to new agents for which the primary pharmacological action, for the proposed therapeutic effect, cannot be demonstrated in an animal model. The underpinning case for dose selection in any trial of this kind should include a clear rationale for the proposed mechanism of action and also for the safety and efficacy of the substance in its intended clinical use.
10. A broader approach to dose calculation, beyond reliance on [NOEL] or [NOAEL] in animal studies, should be taken. The calculation of starting dose should utilize all relevant information. Factors to be taken into account include the novelty of the agent, its biological potency and its mechanism of action, the degree of species-specificity of the agent, the dose-response curves of biological effects in human and animal cells, dose-response data from in vivo animal studies, pharmacokinetic and pharmacodynamic modeling, the calculation of target occupancy versus concentration and the calculated exposure of targets or target cells in humans in vivo.
11. If different methods give different estimates of a safe dose in humans, the lowest value should be taken as the starting point in first-in-man trials and a margin of safety introduced in calculation of the actual starting doses in man.
12. When it is likely that preclinical information, for any reason, may be a poor guide to human responses in vivo, the starting doses in first-in-man trials should be calculated to err on the side of caution. Further dose increases should proceed with caution since the initial dose may have been particularly low and there may be a steep dose-response curve.
Thus, the Duff Report recommended that the calculation of starting dose should utilize all relevant information. The report further stated that the MABEL (minimum anticipated biological effect [dose] level) approach is a useful model for achieving this. At the time of the Duff Report, the MABEL approach to dose selection had not been described in the literature, although it is essentially similar to the PAD approach described by the FDA (2005a). Examples of how to calculate the MABEL dose, as submitted by the ABPI/BIA, were included in the final report. It was also noted that a conservative approach should be taken (i.e., the lowest of the doses calculated by the NOAEL and MABEL methods should be used). It is worth noting that the focus is on the starting dose and that very little emphasis has been placed on understanding the effects of the subsequent doses. While not recommended in the Duff Report, it would be prudent to project the likely effects of all planned doses in a Phase 1 trial.
Use of Pharmacology Data for Selection of The FIH Dose for TGN1412
If the use of pharmacology-based approaches for selecting FIH doses is broadly adopted, one effect will be to place more of the responsibility for selecting the starting dose for the FIH study in the hands of the pharmacologist rather than the toxicologist. This can be demonstrated by using available pharmacology data to calculate an FIH dose for TGN1412. When TeGenero calculated the FIH dose, it was based on the NOAEL in cynomolgus monkeys and the procedure described in the draft FDA guidance for Exploratory INDs (FDA 2005b) but did not include calculation of a PAD for comparison. Thus, the toxicologic NOAEL, which was considered to be 50 mg/kg in monkeys, was corrected for body surface area with an allometric correction factor of 3.1, which resulted in a putative NOAEL HED of 16 mg/kg. TeGenero then selected a starting dose of 0.1 mg/kg, which appeared to provide a cumulative safety factor of 160. It is important to note that the TeGenero documents make no mention of the projected plasma TGN1412 concentration and CD28 receptor occupancy necessary for the desired pharmacologic effect (Treg cell stimulation), nor of the projected pharmacologic effects of the FIH dose in immunocompetent human volunteers. Thus, this FIH dose, derived from a toxicology-based NOAEL algorithm, was not placed in any pharmacologic context.
In keeping with the MABEL approach of using all available information and emphasizing pharmacologic effects, there are several different ways to calculate a starting dose for TGN1412, all of which result in markedly lower FIH doses. The first method would be to use data from mAb 5.11A1, the murine precursor to TGN1412, for which the minimally active concentration was 0.1 μg/mL in an in vitro assay of human T cell proliferation (Lühder et al. 2003). It should be noted that the concentration-response curve in this assay was markedly bell-shaped, with maximal activity at ~0.2 to 1 μg/mL. Assuming IV administration to a 70-kg human with 2.5 L plasma volume, an immediate postinjection concentration of 0.1 μg/mL would be expected to be provided by a total dose of 0.25 mg, or approximately 0.003 mg/kg (Sims 2007). This FIH dose (3 μg/kg) would be expected to be associated with minimal, not optimal or maximum, pharmacologic activity in vivo. This FIH dose is 33–fold lower than the dose used in the clinical trial. By the same logic, the actual FIH dose of 0.1 mg/kg might be expected to have resulted in postdose concentrations of ~3 μg/mL, which are similar to those producing maximal effects in vitro.
A second method would be to use data from a surrogate molecule (JJ316, a rat CD28-specific superagonist antibody). For this mAb, the NOEL in studies conducted in healthy and arthritic rats was < 0.3 mg/kg and optimal pharmacological responses were achieved between 1 and 5 mg/kg. Therefore, the MABEL dose could be considered to be between 0.3 and 1 mg/kg. If one assumes that the MABEL dose was 0.5 mg/kg and uses the 1/100th safety factor proposed for microdosing (EMEA 2004), the safe starting dose would have been calculated to be 0.005 mg/kg or 5 μg/kg. This FIH dose is 20-fold lower than the dose used in the clinical trial.
Implicit in the use of MABEL to calculate the FIH dose is the assumption that the mechanism of action (MOA) is known prior to the initiation of the clinical trial and that both toxicity and efficacy are related to the MOA. This is generally true for most biologics but not always true for small molecules, where the toxicity may be off-target toxicity rather than on-target toxicity or exaggerated pharmacology.
An alternative method of determining the FIH dose relies on estimation of the pharmacodynamic (PD) effects, such as receptor occupancy, likely to occur at a given dose. Calculations of receptor occupancy were included in the Duff Report. These calculations were based on the dissociation constant for TGN1412 and an assumption of the volume of distribution for TGN1412. It was calculated that a dose of 0.1 mg/kg would have resulted in 86.2% receptor occupancy. The Duff Report also includes a slightly more detailed calculation for the predicted receptor occupancy that was performed by the ABPI/ BIA taskforce in which the percentage of CD28 receptors occupied by TGN1412 was 90.6%. Lowe et al. (2007) have provided an even more comprehensive set of calculations and also concluded that a dose of 0.1 mg/kg TGN1412 would have resulted in > 90% receptor occupancy. Using the same method of calculation, a dose of 0.005 mg/kg would have resulted in ~33% receptor occupancy. This is interesting because Lowe at al. state that within Novartis, for biotherapeutics with potentially antagonistic modes of action on key body systems, no more than 10% receptor occupancy is proposed for the first dose in humans. To achieve that low level of occupancy for TGN1412, a dose of 0.001 mg/kg should have been recommended. Such an FIH dose would be 1/100th of the actual dose and is 3– to 5–fold lower than the FIH doses calculated above by the MABEL approach.
There are of course limitations to be expected with any of these pharmacology-based methods. The most important of these is the extent to which the target biology and product pharmacologic and toxicologic activities within the preclinical animal species are representative of those of humans. It is only with hindsight that the relevance of the preclinical work can be truly known. However, assuming that the chosen species are highly relevant, several limitations remain. For instance, while calculations can show that a dose of 0.01 mg/kg would result in ~90% of CD28 receptors on T cells being occupied by TGN1412, they cannot address the functional consequence of the binding of TGN1412. It would remain unknown whether 90% receptor occupancy is required for cytokine release or whether a lesser extent would elicit the same response, whether response is dictated by the extent of saturation of any one cell or the average saturation of all cells, how many T cells must be stimulated to release cytokines for the adverse effects to be seen, or whether functional activity requires a secondary event (e.g., recognition of the Fc region by Fc receptors on effector cells). To understand these relationships, it is extremely important that in vitro studies be performed to correlate concentrations with effect before dose-selection decisions are made. Even then, the in vitro pharmacologic response may be highly dependent on variables that may not be immediately apparent, such as a requirement for “fixation” of TGN1412 (Stebbings et al. 2007). Finally, these theoretical calculations of receptor occupancy need to be placed in the context of the PD assays that will be used in the clinic. Often, the PD assays are receptor occupancy assays based on flow cytometry methods that determine the percentage of a given cell type that either stain positive for the presence of the exogenously administered antibody or have “free sites.” Such assays do not provide us with information regarding the extent of receptor occupancy on any given cell.
Simply understanding the PD properties of a given product in terms of a concentration-effect relationship may not be sufficient to project a FIH dose. To fully understand the pharmacological effect of a product, it may be necessary to perform PK/PD modeling (e.g., Emax modeling) as was emphasized by both the Duff Report and the EMEA guideline. However, creating such models is not a trivial task, and they almost invariably require refinement based on actual human data. Once created, a PK/PD model should be continually refined as real data are obtained from the FIH clinical trial, and thus, the utility of PK/PD models in helping establish the doses for the FIH study may be limited. In addition, with the use of PK/PD models, there is the potential for a misunderstanding regarding the accuracy of the predictions. For example, the pharmacokineticist may feel that the prediction of the PK was successful if the observed and the predicted results were within twofold (i.e., the observed clearance is within 50%–200% of the predicted clearance; Rowland 2007). It has been reported that even with such a wide acceptance range, the success rate is < 53% (Nagilla and Ward 2004, 2005a, 2005b; Ward and Smith 2004a, 2004b; Mahmood 2005; Huang et al. 2008). Yet, for a product with potentially serious PD-mediated effects, this level of variability in PK/PD parameters may be undesirable. Most often, allometric scaling based on the animal PK data is used to predict the PK of a small molecule drug in humans. However, for mAbs, it may not be feasible to apply the principles of allometric scaling due to species differences in affinity for the FcRn receptor. It may be more reliable to assume that a “humanized” mAb will follow the typical kinetics of an immunoglobulin, unless data indicate otherwise. This greatly simplifies the predictions of the PD response in humans. In addition, it should be recognized that predictions of the PK of a product in humans are most often based on mean animal data and the predictions do not take into account any potential variability that may be observed in humans. The lack of accounting for variability could lead to surprises in the observed response in the clinic, although the potential for interindividual variability of PK and/or PD in humans is often mitigated by the use of a safety factor (e.g., 10). Thus, our ability to accurately predict the PD response to FIH doses is limited and needs to be enhanced by the generation of physiologically relevant PK/ PD models and better understanding of the dose and concentration-effect relationships and potential sources of variability. Even so, the accuracy of our predictions can be known only with hindsight.
Conclusions
Administration of TGN1412 to healthy volunteers was associated with rapid T cell activation, cytokine release, depletion of circulating T cells, and the development of clinical symptoms and signs consistent with a “cytokine storm,” followed by modest T cell proliferation two weeks later. Similar T cell activation, depletion, and/or rebound proliferation have been documented in rats that received a surrogate antirat CD28 mAb (JJ316), mice expressing human CD28 and human Fc receptors that received the murine precursor to TGN1412 (mAb 5.11A1), and monkeys that received TGN1412. These results confirm that these animal species were to some degree pharmacologically relevant for testing the safety of TGN1412 but do not indicate that they will be predictive of human pharmacology. The in vitro and in vivo studies in animals demonstrated that TGN1412-induced cytokine release and/or observation of clinical signs consistent with cytokine release were not prominent features. This is not a new observation; both HuM291 and OKT3, directed against CD3 on T cells, are known for inducing marked T cell activation, depletion, and cytokine release in humans yet did not result in a cytokine release syndrome in chimpanzees. This knowledge should have influenced the design of the nonclinical studies toward assessing all aspects of T cell activation and invoked a strong sense of caution in interpreting their results. It would further dictate that a concerted effort to predict the pharmacodynamic effects in humans at the proposed FIH doses was essential. The observation of differences in the timing and magnitude of the response to the surrogate JJ316 and TGN1412 suggests differences in either their superagonist qualities or the biology of CD28 in these species and should have prompted additional caution about interpretation of the results.
The TeGenero incident has demonstrated once again that pharmacologically relevant animal species are reasonable predictors for human T cell activation, depletion, and rebound proliferation but not for cytokine release or the accompanying symptoms. It has also demonstrated that inappropriate study designs, incomplete assessment of available data, and a failure to reconcile inconsistencies in the nonclinical safety data may lead to inappropriate conclusions about human safety. Thus, in deference to the Duff Report conclusions, the observed clinical adverse events in the TeGenero incident were in fact predictable based on accumulated historical experience with anti–T cell mAbs, specific information that was already available or could have been generated with TGN1412 and appropriate consideration of the pharmacodynamic properties at the chosen doses. The severity of the TeGenero incident was, however, attributable as much to the inappropriate design and conduct of the Phase 1 study as to the activity of TGN1412. The vast experience with other anti–T cell mAbs has demonstrated that FIH trials can be conducted safely if appropriate pharmacologic projections are made and the attendant clinical precautions are taken. Rather than demonstrating the nonpredicitiveness of safety testing in animals, the TeGenero incident has served to highlight the invaluable contributions of appropriate, well-designed animal safety studies to the informed clinical development of novel therapeutics.