Use of Genotoxicity Data to Support Clinical Trials or Positive Genetox Findings on a Candidate Pharmaceutical or Impurity …. Now What?

INTRODUCTION AND HISTORY

Carcinogenicity can be a legitimate concern for human drugs. Pharmaceuticals are taken at doses designed to have pharmacological effects and often for protracted time periods. Because of the long latency period for human cancers, even the longest clinical trials are too short to determine if a drug is potentially carcinogenic. Postmarketing studies are only informative if the drug induces an unusual form of cancer as was seen in the daughters of women who took diethylstilbestrol during pregnancy (Waggoner et al. 1994) or if the drug is especially potent and specific tumors are monitored as with psoralen plus ultraviolet light (PUVA) therapy (Stern, Nichols, and Vakeva 1997). If a drug simply increases the risk for a common cancer, it is unlikely that epidemiological studies would be able to detect such an effect.

Typically, the results of carcinogenicity studies are not known until the time of submission of a New Drug Application (NDA). By the time the NDA is submitted, large scale human clinical trials have been completed, which can involve exposure of hundreds or thousands of subjects for time periods ranging from a few weeks to many months or even years. This long-standing paradigm for drug development necessitates that the Center for Drug Evaluation and Research (CDER) depend on surrogate markers for cancer risk. Typically, CDER review divisions rely on the results of genetic toxicology tests to signal possible cancer hazards. Prior to “first-in-man” clinical studies, the International Conference on Harmonization (ICH guidelines M3, S2A, and S2B) suggest sponsors perform a bacterial test for gene mutation (Ames assay) and an in vitro assay for clastogenicity (metaphase analysis for chromosome breakage or the mouse lymphoma gene mutation assay). Prior to beginning phase 2 clinical trials, the guidelines suggest sponsors complete the third test in the battery, an in vivo rodent assay for chromosomal effects, often a micronucleus test in bone marrow. Negative results in these three assays are generally sufficient to assure safety in clinical trials if no other signals of a cancer hazard are evident. Other risk factors include previous demonstration of carcinogenic potential in the product class considered relevant to humans; a structure-activity relationship suggesting carcinogenic risk; evidence of preneo-plastic lesions in repeated-dose toxicity studies; long-term tissue retention of parent compound or metabolite(s) resulting in local tissue reactions or other pathophysiologic responses. A positive result in one or more of the three genetox assays and/or a concern in regards to one of the other risk factors may signal a need for additional studies prior to initiation of repeated-dose clinical trials.

For certain disease indications, results in genetic toxicology tests will have little or no regulatory impact. For example, many drugs designed to treat serious types of cancer are known mutagenic carcinogens (Benedict et al. 1997). Positive results from genotoxicity tests on drugs designed to treat serious illnesses near the end of life such as Alzheimer’s disease can influence the choice of the patient populations selected for clinical trials but generally would not trigger requests for additional testing. Clearly, the number of exposed subjects and the duration of clinical exposure are important risk factors that must be considered. Drugs or diagnostics designed for only a single administration are thought to present low risk. Repeated-dose clinical trials proposed for healthy subjects or for patients with mild, non–life-threatening indications, must present minimal risks.

The ICH guidelines give minimal direction concerning how to proceed when positive results are seen in a genetic toxicology test. For example, S2A indicates “compounds giving positive results in the standard test battery may, depending on therapeutic use, need to be tested more extensively.” The guideline also states “a positive result in vitro is followed up by a second in vivo study—using tissue other than bone marrow or peripheral blood.” Frequently, the second in vivo assay is the rat liver unscheduled DNA synthesis (UDS) assay. Anecdotal information, primarily from contract testing labs, suggests results the in vivo UDS assay are, in almost all instances, negative.

A survey of marketed pharmaceuticals taken from the 1999 Physician’s Desk Reference and from the open literature suggests that of a total of 467 drugs examined, approximately 75% have at least one genetox result (Snyder and Green 2001). Drugs for which there were no data tended to be older drugs that were not generally given chronically, such as antibiotics, antifungals, antihistamines, and anesthetics. For those drugs for which data were available, 29% had at least one positive genetox result. Not surprisingly, positive results for the bacterial mutation assay and in vivo chromosomal aberration assay were rare. Most of the positive findings were for the in vitro chromosomal aberration assay and the mouse lymphoma gene mutation test. Although there were more than twice as many in vitro chromosomal aberration assay results as compared to the mouse lymphoma assay, the percentage of positives was the same for the two assays, 25%. Of the universe of drugs evaluated, 201 had both genetic toxicology and carcinogenicity data. Of those, 124 were negative for carcinogenicity and 77 were either positive or equivocal. Of the 124 negatives, 100 were also negative for genetox; of the 24 “false positives,” 19 were in vitro cytogenetic assays. Of the 77 compounds positive or equivocal for carcinogenicity, only one-third was also positive for genotoxicity. These results indicate that there are many nongenotoxic carcinogens. Thus, it is worth noting that at least for drugs, most positive carcinogenicity results are not related to mutagenicity. Exaggerated pharmacological effects, immune suppression and hormonal disequilibrium are common modes of action.

In the recent past a number of CDER review divisions have requested sponsors perform either a Syrian hamster embryo (SHE) cell transformation assay or a p53 mouse carcinogenicity study to help clarify positive genetox results. In some cases, the positive results were seen only in the in vitro chromosomal aberration or the mouse lymphoma gene mutation assay. Performing a transgenic mouse carcinogenicity study prior to initiating repeated-dose phase 1 clinical trials is problematic. The time required to perform the 28-day range-finding study combined with the 6-month in-life for the definitive assay and the postlife phases for the two studies would result in more than a year’s delay in the clinical program. The SHE cell transformation assay is shorter in duration but typically would still be expected to take several months. Furthermore, the pharmaceutical industry has not embraced the SHE cell assay. Validation studies, primarily from one or two laboratories, have demonstrated good concordance between results in the SHE cell assay and the outcome of 2-year rodent carcinogenicity studies (Isfort, Krckaert, and LeBoeuf 1996). In particular the “low pH” SHE cell assay is reported to have even better concordance (Isfort, Krckaert, and LeBoeuf 1996). These validation studies have been primarily on industrial and agricultural chemicals and pollutants and not on pharmaceuticals. A validation program organized by the International Life Sciences Institute examined the concordance among a series of transgenic mouse carcinogenicity models, the SHE cell transformation assay, and traditional 2-year rodent bioassays for a series of 19 human drugs (Mauthe et al. 2001). Again the data suggested good concordance between the SHE cell assay and the 2-year studies but the correlation with human carcinogens was poor, 37%. The SHE cell transformation assay also suffers from a number of technical challenges: cells have to be repeatedly isolated and qualified; frozen cells have limited shelf life; serum must be screened and qualified (shown to be capable of morphological transformation at acceptable frequencies); evaluation of transformed foci is highly subjective; feeder cells must be irradiated; the difference between a positive and negative study can sometimes hinge on one focus; high failure rate with many repeats; significant overlap in the historical values for the vehicle and positive control transformation frequencies (Myhr 2005). As indicated above, many drugs are known to induce tumors by a variety of mechanisms unrelated to mutagenesis. Modes of action such as exaggerated pharmacological effects, immune suppression, and hormonal imbalances are only expected to be operant in whole animal systems. It is unclear how a cell culture–based assay could respond to these mechanisms.

CDER has recently published a draft guideline entitled “Recommended Approaches to Integration of Genetic Toxicology Study Results” (CDER 2004). This guidance suggests possible alternatives to the SHE cell and p53 assays for clarifying positive genetic toxicology results. The guidance focuses on three approaches: weight-of-evidence (WOE), mode of action (MOA), and conduct of additional studies to add to the WOE.

WEIGHT-OF-EVIDENCE

In using a WOE evidence approach, the following questions are germane: Was the positive response observed in an in vitro or in vivo assay? Was the positive response reproducible? Was the response dose related? What was the magnitude of the response? Was the response seen only at a highly cytotoxic dose? Was a positive seen in more than one assay? Positive responses that are dose related, seen at doses that induce little or no toxicity, and seen in multiple test systems raise the level of concern. Positive in vitro responses that are not reproducible, are not dose related, and are seen only at highly cytotoxic doses are generally considered to represent little or no human risk. For the in vitro chromosomal aberration assay in particular, it is vital to look not just at the concurrent negative-control values but also at the historical control values. Occasionally, one can see statistically significant, dose-related increases in aberration frequencies compared to the concurrent vehicle control but all the values are within the historical control range. Such a “positive” result likely does not have biological significance. The guidance also points out that it is also useful to look for concordance between segments within a study and between studies. For example, both the in vitro chromosomal aberration assay and the mouse lymphoma gene mutation assay have short-term (∼4 h) and long-term (∼24 h) exposures in the absence of S9. Examination of concordance within studies at comparable levels of cytotoxicity can be informative as well as comparable exposures between studies. If the WOE suggests a lack of human risk, additional studies may not be required.

MODE-OF-ACTION

Examination of the MOA of a positive response can help elucidate whether the response represents a human risk. The best known and often cited examples are false-positive results in in vitro mammalian cell assays induced by extremes in osmolarity and pH (Galloway et al. 1987). Since the report of these potential artifacts, these parameters are routinely measured in testing laboratories. Another relevant parameter that should be considered is the possibility of a threshold. Unlike other types of toxicity, risk assessments on carcinogens and mutagens have historically not included consideration of thresholds. This practice was based on the conventional wisdom that a single molecule of an alkylating agent could theoretically cause a mutation and initiate a tumor. However, it is also likely that some agents that induce genotoxic end points may do so by mechanisms that have thresholds. For example, genetic lesions scored as micronuclei can originate through clastogenic events giving rise to acentric chromosome fragments, or through damage to the mitotic spindle that would give rise to whole chromosomes. Because DNA is not the target in the latter situation, this MOA is likely to be associated with a threshold. Other examples of MOA likely to be associated with a threshold include enzyme inhibition such as induction of chromosomal aberrations by topoisomerase inhibitors, inhibitors of DNA synthesis, and excessively high toxicities (Hilliard et al. 1998; Galloway et al. 1998; Galloway 2000).

When positive genetox responses are seen, they are often in one of the in vitro mammalian cell assays. Frequently, in vivo assays on the same study material are negative. The negative in vivo studies do not necessarily allay concern. There are multiple possible explanations for this commonly seen discrepancy: the drug may be metabolized differently with genotoxic metabolites generated in vitro but not in vivo; potentially genotoxic products may be metabolically inactivated in vivo; the genotoxic product may not reach the target cell in vivo. The most common reason is that blood levels of a genotoxic product cannot be achieved that are comparable to concentrations that gave positive results in vitro.

ADDITIONAL TESTING TO SUPPORT THE WOE

In instances where results for genetic toxicology studies or other indicators suggest potential risk for carcinogenicity, additional testing may be warranted to either allay concern or con-firm the potential risks. Data from a broad spectrum of assays can be used to add to the WOE assessment and help regulators and sponsors determine the existence and magnitude of a cancer risk. For example, concerns regarding a positive in vitro clastogenic response can be explored by examining cytogenetic endpoints from subacute or subchronic studies. The National Toxicology Program (NTP) routinely makes blood smears from mouse 90-day studies for the evaluation of micronuclei. Peripheral blood lymphocytes from rats and primates are easily cultured for metaphase analysis. Bone marrow metaphases from various species can be assessed with fluorescent in situ hybridization for the assessment of stable chromosomal rearrangements. Evaluation of micronuclei in bone marrow from animals on repeated-dose studies has limited value since the results are only informative on events in the last day or two of the study.

When a positive response is seen in one of the two in vitro mammalian cell assays, it is reasonable to complete the four-test battery and perform the other mammalian cell assay. Other end points that can influence the WOE include assessment of DNA adducts, evaluation of DNA damage using the “comet” assay, mutation in vivo using a transgenic mutagenesis model, in vitro cell transformation assays (including SHE), short-term carcinogenesis studies, and patient monitoring. Although this last option is technically straightforward, the associated ethical and legal issues are not. For example, whereas increased genetic damage can suggest a cohort may be at increased risk for cancer, nothing can be said regarding individual risk. What obligations would a sponsor have to a patient population found to have elevated levels of markers of genotoxicity?

LIMITS OF GENOTOXIC IMPURITIES

Impurities in a drug can result from residual starting materials, intermediates, side reactions, degradation products, and contaminants from packaging. Excipients can also be a source of impurities. ICH Q3A gives guidance on impurities in drug substances; Q3B(R) deals with impurities in drug products whereas Q3C gives guidance on residual solvents. Guidance from ICH for dealing with genotoxic impurities is limited. Q3A and Q3B(R) define levels of impurities that trigger reporting, identification, and qualification. Although the specifications for these triggers are linked to the total quantity of drug substance or product consumed, the requirement for qualification is elicited when the impurity reaches 0.15% or a total daily intake of approximately 1 mg. Qualification for genotoxicity involves performance of a bacterial mutation assay and an in vitro mammalian cell assay for chromosomal damage (metaphase test or mouse lymphoma assay). Furthermore, the guidelines indicate “Such studies can be conducted on the new drug substance containing the impurities to be controlled, although studies using isolated impurities can sometimes be appropriate.” This approach begs the question as to whether genetox assays have sufficient sensitivity to detect levels of genotoxic impurities that could potentially induce adverse health effects. Typical positive control chemicals used in an Ames assay include 2-aminoanthracene (1.0 μg/plate); sodium azide (1.0 μg/plate); 2-nitrofluorene (10 μg/plate); 9-aminoacridine (75 μg/plate), and methyl methane-sulfonate (1000 μg/plate). Assuming an impurity is present at 0.15%, (minimum for qualification) and the drug is nontoxic, the drug substance would be tested to 5 mg/plate or 7.5 μg/plate of the impurity. At this concentration, 2-aminoanthracene and sodium azide would be detected, 2-nitrofluorene would probably be detected, 9-aminoacridine and methyl methanesulfonate would not be detected. Clearly, powerful genotoxins could be missed using this strategy. Furthermore, any toxicity associated with the drug substance would further limit the sensitivity of this approach to qualification.

Unlike an active pharmaceutical ingredient, an impurity imparts no benefit, only risk. In this sense, these chemicals resemble environmental contaminants. In dealing with environmental contaminants, regulatory agencies such as the Environmental Protection Agency (EPA) generally perform quantitative risk assessments and determine acceptable daily intakes. For health effects that are thought not to have thresholds, such as cancers caused by DNA-reactive chemicals, acceptable exposures are based on calculation of “virtually safe doses.” A virtually safe dose is defined as a lifetime daily exposure that increases an individual’s risk of contracting cancer by one in one hundred thousand or one in one million. A prerequisite for performing these types of quantitative risk assessments is the availability of dose-response data for increased risk of the adverse event (e.g., cancer) and exposure to the chemical. Quantitative risk assessments on most drug impurities will not be possible because lifetime rodent bioassays are not typically performed on isolated impurities. The European Medicines Agency (EMEA) has recently proposed a method for dealing with this challenge (EMEA 2004). This draft guideline proposed the application of a “threshold of toxicological concern” (TTC). The TTC defines an exposure to an unstudied chemical that does not pose a significant risk for carcinogenicity or other health effect (Munro, Kennopohl, and Kroes 1999; Kroes and Kozianowski 2002). Based on a database of more than 700 carcinogens and defining a virtually safe dose as increasing cancer risk by no more than one in one million, the guideline proposes a TTC for genotoxic impurities of 1.5 μg/person/day for lifetime exposures. The guideline goes on to note that although this number is reasonable for a carcinogen of average potency, potency can vary by orders of magnitude and that 0.15 μg/person/day might be a more reasonable number for the more potent carcinogens. The guideline then circles backs to suggest that because pharmaceuticals are designed to provide benefit, a more reasonable definition of virtually safe dose may be one in one hundred thousand that brings us back to a TTC of 1.5 μg/person/day.

The Food and Drug Administration (FDA) is currently working on a guideline for genotoxic impurities and considers the reasoning outlined in the EMEA publication to be an excellent starting point for FDA’s own deliberations. Issues that are currently under discussion include the observations that (1) carcinogens can vary in potency by several orders of magnitude; (2) because drugs may contain many impurities, whether there should be an upper limit to the number of TTCs permitted; and (3) TTCs are based on lifetime exposures whereas early clinical trials may be only weeks or months in duration. In this latter situation, higher levels of impurities may be acceptable but reaching a consensus on what those levels should be will no doubt be a challenge.