Genotoxic Impurities

Pharmaceutical process chemists want to make the compound, the whole compound, and nothing but the compound. They know, however, that chemistry isn’t that simple and that a multistep synthesis often involves undesirable materials. When that happens, the burden is on them to eliminate, or at least mitigate, the risk of harmful impurities in the final product.

Harmonized international guidelines long addressed impurities in drugs but skirted the issue of genotoxic impurities (GTIs)—ones that damage DNA and potentially cause cancer. To fill this gap, regulatory agencies and the drug industry have been trying to balance what is technically feasible and scientifically necessary against the risks of potential GTIs.

The agencies have more or less passed judgment. The European Medicines Agency (EMA) issued draft guidance in 2004 and final GTI guidelines in January 2007. In 2008, the U.S. Food & Drug Administration came out with nearly identical draft guidelines. EMA officials tell C&EN that pharmaceutical makers already routinely apply the guidelines when submitting drug approval filings.

Although drug companies are complying, many in the industry find the guidelines to be overly strict. Executives are loathe to be openly critical of the regulatory agencies, but former industry scientists now working as consultants are challenging the basic principles. And most companies have been raising questions at least to understand what is required and learn whether their proposed GTI mitigation strategies will satisfy regulatory dictates.

According to EMA, regular meetings with the European Federation of Pharmaceutical Industry Associations (EFPIA) have been constructive. One positive outcome is an evolving Q&A document that provides clarification on applying the new guidelines. EFPIA, for its part, has indicated that it would like to see the process expanded to resolve uncertainties and concerns, and even potentially modify the guidelines.

When GTIs are unavoidable, regulators have set a default threshold of toxicological concern (TTC), a level under which the risk is considered negligible. The guidelines stipulate that exposure to a GTI must be below 1.5 μg/day, which represents a 1 in 10⁵ incremental lifetime cancer risk. This level doesn’t apply to the most potent carcinogens or to GTIs for which safe limits already exist based on evidence of a threshold mechanism for genotoxicity.

The TTC comes from an analysis of 730 compounds and related carcinogenicity data in the Carcinogenic Potency Database housed at the University of California, Berkeley. Originally proposed by regulators as a 0.15-μg limit on carcinogenic impurities in food, the value was adjusted upward for pharmaceuticals because of the benefit gained from taking medications. Even so, the 1.5-μg limit is about 1,000 times lower than typical thresholds on impurities.

Some scientists question the derivation and validity of the TTC, especially as it applies to pharmaceuticals. The underlying analysis is nontransparent, and it’s not clear which compounds were included or whether they are actually carcinogens, says David J. Snodin of Xiphora Biopharma Consulting in England. A Ph.D. organic chemist, Snodin spent 12 years as a toxicology reviewer in the U.K. before moving into consulting in 2002. He says the TTC also relies on misleading linear extrapolations of carcinogenic effects from high doses to low ones.

“The data set used is horridly skewed in that it includes a lot of compounds that will hardly, if ever, be found as impurities in pharmaceuticals,” Snodin says. They include such highly potent carcinogens as aflatoxin-like, N-nitroso, and azoxy compounds that are of more concern in food. The analysis also is “incredibly worst case,” he says, because the lowest statistically significant values for carcinogenic potency, rather than ones derived from a weight-of-evidence approach, were selected.

Reagents and intermediates that are reactive and synthetically useful, which likely makes them interact with DNA, are often unavoidable, pointed out Edward J. Delaney in a 2007 paper (Regul. Toxicol. Pharmacol. 2007, 49, 107). Then at Bristol-Myers Squibb and now president of New Jersey-based Reaction Science Consulting, Delaney reanalyzed the TTC data set after discounting the high-potency, food-associated compounds. He found that the remaining chemicals of concern that tend to appear in the course of pharmaceutical process chemistry were at the weaker end of the potency spectrum.

As a result, Delaney, Snodin, and others consider the TTC much too conservative. According to Snodin, “The TTC value is at least one order of magnitude, and probably several, too low, but I think regulators are reluctant to move away from it.” Meeting such restrictive limits, some industry scientists suggest, requires significant time and effort that diverts resources and impedes pharmaceutical R&D, without providing an appreciable return in patient safety.

FDA’s decision to follow EMA in adopting the 1.5-μg level indicates “broad consensus among regulatory and industry scientists in using this TTC limit,” EMA tells C&EN in written responses to questions. “The TTC concept is a risk management tool that allows defining acceptable human exposure levels in the absence of adequate toxicological data,” the agency explains. “This can only be done on the basis of worst-case assumptions resulting in quite conservative limits.”

Alternatives to the TTC exist, regulators point out. If there are adequate safety data for a known GTI—which isn’t common—it can be used to set limits that may differ from the TTC. Lacking data, drug developers have the option to conduct toxicological studies with a GTI to enable an estimation of a compound-specific limit, or to default to the TTC.

With impurities suspected to be GTIs, drug developers may not bother to classify them, depending on the expected risk, needed analytical efforts, and considerations for controlling it. If no testing is done, potential GTIs must be controlled at or below the TTC. If testing finds that a potential GTI is genotoxic, either the TTC or a level based on safety data must be used. If it’s not genotoxic, the impurity is handled like any other.

Easier said than done, Snodin points out. “How one proceeds with further testing isn’t elaborated in either guidance,” he tells C&EN, “and there could be difficulties in obtaining scientific advice on appropriate additional testing, because many regulators favor elimination rather than qualification.” To help others leverage what information is available, he has compiled data on a list of compounds in several chemical classes tagged as potential genotoxins (Org. Process Res. Dev. 2010, 14, 960).

Although they don’t like it, Snodin and others in industry concede that the TTC is probably here to stay. “The problem is that at this moment nobody has an alternative to the TTC principle that everyone can universally agree on,” explains Andrew Teasdale, a senior quality assurance executive at AstraZeneca and chair of the company’s GTI advisory group. From this perspective, he says, it’s pragmatic to realize that the TTC is the “least worst option” and to view it instead as the principle around which future discussions can be framed.

In fact, Teasdale says, history shows that the drug industry actually asked that the TTC concept be applied because an alternative was carcinogenicity testing for every GTI. Industry helped tweak the TTC concept more to its liking when scientists from 12 major drug firms responded to the emerging EMA guidelines by proposing a “staged” version (Regul. Toxicol. Pharmacol. 2006, 44, 198).

In simple terms, staging means drugs can contain higher levels of impurities if they are given for shorter periods of time, with the level staged to the duration of exposure. At this point, regulators have accepted the staged TTC idea for compounds in clinical trials, although they halved most of the levels industry proposed.

For pharmaceutical companies, the staged TTC greatly eased the burden on characterizing and controlling impurities during the drug development process, says Larry Wigman, former associate director of analytical sciences at Sanofi-Aventis and now principal consultant at Regulitics in Pennsylvania. In early-stage work, impurity information is limited and analysis methods are undeveloped. As candidates advance and synthetic processes are optimized, impurities are routinely assessed and plans are made for avoiding or controlling them.

A risk assessment of the synthetic process—including starting materials, intermediates, solvents, by-products, and impurities—identifies GTIs that are or might be present. Certain chemical functional groups or structures associated with DNA reactivity are considered “alerts” for genotoxicity. If an impurity has a structural alert, a bacterial mutagenesis screen such as the Ames test can be run to confirm its genotoxicity. A negative Ames test result will overrule a structural alert, and the impurity can be considered nongenotoxic.

Structural considerations can easily turn complicated. Going by the assumption that structurally similar impurities have the same genotoxic mode of action and molecular target, regulatory authorities want them grouped together and subject to a single 1.5-μg TTC level. This move can necessitate reducing two or more similar impurities to sub-part-per-million levels, a task that poses big process development and analytical challenges.

Some scientists ask whether such restrictions are necessary when research has suggested that synergistic effects are unlikely at very low doses. The TTC is already conservative, and they question adding an extra level of protection when the supporting science would argue that some impurities can be treated as individual entities.

In a recent paper, David P. Elder, director of preclinical development at GlaxoSmithKline, and a coworker discuss several arguments against cumulative control, among them the human body’s mechanisms for coping with low-level impurities in the environment and diet (Org. Process Res. Dev. 2010, 14, 1037).

EMA, however, counters that the scientific understanding around exposure to mixtures of toxic compounds does suggest that additive effects are more likely for compounds with similar structural alerts. “We do not agree that this can be considered as an ‘increased level of caution’ for pharmaceuticals,” the agency tells C&EN.

Structural similarity determinations aren’t straightforward. Attributes may make impurities look similar when in fact they are not—either from a toxicological perspective or in how they behave chemically, Elder says. For example, methyl chloride and ethyl chloride are both carcinogenic alkyl halides but induce different tumor types and shouldn’t be grouped together, he argues.

At other times, similar compounds may behave alike, and industry scientists will draw connections between a potential GTI and another that has tested negative for genotoxicity or has a known threshold limit. Likewise, although regulators are proponents of cumulative control in theory, they have been reluctant to accept extrapolations between similar compounds for setting specific impurity limits, such as for the alkyl halides, Teasdale says. Regulators and industry can seem to be flip-flopping on the issue because, ultimately, data on how the compounds behave are more important than surface similarities.

Ironically, despite the debate over structural similarity, in silico structure-activity relationship (SAR) methods, which have become popular for characterizing GTIs, rely on such similarities. Two widely used software tools are Lhasa’s DEREK and MultiCASE’s MCASE. Regulators have left open the possibility for quantitative SAR determinations of cancer potency, and researchers are trying to develop in silico methods to predict actual carcinogenic potency values for use in risk assessments (Regul. Toxicol. Pharmacol. 2010, 57, 300).

In silico methods can be used instead of lab tests, according to the regulatory agencies. “Being able to use in silico tools to discount genotoxic concerns is probably more beneficial than trying to argue against grouping impurities,” Teasdale says about the apparent contradiction in industry’s embrace of structural-similarity-based software. Regulators will accept the absence of an in silico hit as sufficient to conclude that the impurity isn’t genotoxic even if it has an alerting structure. A positive SAR result is usually followed by an Ames test, which, if negative, overrides the SAR determination.

Although seemingly clear-cut, GTI evaluations have become a significant issue, Teasdale says. “There is no standard, agreed-upon process as to how you do an SAR evaluation.” Most companies use the commercial databases, but regulators have approached GTI evaluation in disparate ways, he explains. Some European national agencies don’t use any in silico method, because the systems are too costly to run and maintain. These agencies rely instead on simple structural alerts.

On the other hand, FDA has multiple systems, some of which are proprietary and inaccessible to outsiders, Teasdale says. Several times, company researchers have gotten negative SAR predictions, yet agency reviewers will flag impurities as genotoxic. Challenging the regulator’s decision is nearly impossible, he says. “You simply don’t know in enough detail the basis on which they have made their prediction to generate a response.

“Imagine the consternation that generates,” he says about the uncertainty companies face. “You could have gone all the way through development, following an appropriate risk assessment based on the commercial in silico systems. You get to a point where you make a regulatory submission, and all of a sudden you get concerns expressed over impurities that you had been treating as normal impurities.”

Even minor changes in the GTI assessment can have major implications on process chemistry development and the analytical workload. “The potential impact is massive,” Teasdale says. “The cost becomes almost incalculable if you are talking about something that could lead to having to reprocess material, delay a Phase III clinical study, or in some instances even lead to the death of a compound.”

Although the new guidelines have not had an immediate big impact, “it’s been a pretty rapid step-change in the way that industry has had to deal with impurities,” Snodin says. “The big pharma companies have tried really hard to get their act together and have put a lot of resources in making sure they have these issues covered. Generic drug companies, however, were often taken completely unaware and were totally unprepared for how to deal with the tidal wave of deficiency questions coming from the regulators.”

More broadly, companies have been complaining about differences in how European national agency assessors and individual FDA reviewers interpret and apply the GTI guidelines. EFPIA recently submitted several case studies to an EMA advisory group, which agreed with industry’s view that virtually all of the questions shouldn’t have been raised because they were outside the scope of the guidelines or about compounds for which there were no structural alerts.

EMA acknowledges to C&EN that “cases of inconsistent and inappropriate interpretation of the guidelines” have occurred. This issue has been discussed among the relevant parties to achieve a more harmonized approach, and the need for assessor training on this topic is under discussion, the agency adds.

After identifying GTIs, drug manufacturers face two challenges: figuring out how to prevent or limit GTIs in the final active pharmaceutical ingredient (API) and convincing regulators that their efforts have been successful. Approaches vary by company. Some opt to alter synthetic routes to avoid using or generating GTIs altogether. Others consider this impractical, especially when multiple GTIs must be handled and making changes might only lead to new ones.

Although GTIs are a major process-design consideration, they must be dealt with in the context of other factors such as operator safety, environmental impact, and cost-efficient operations. Increasingly, companies are saying they want to demonstrate GTI control through risk assessment techniques rather than brute force testing. Based on some company experiences, however, their perception is that regulators are not accepting these chemistry-based arguments on their own. Although they understand the arguments, overly cautious reviewers are looking instead for detailed data showing that GTIs are absent, companies report.

The last thing drug manufacturers want to do is test the final product. Doing so likely will require sophisticated and expensive techniques to measure trace levels of impurities (see page 27). In a 500-mg drug dose, for example, the TTC is just 3 ppm.

The frustration, industry scientists say, comes in trying to prove a negative and show something isn’t present. “What we have been trying to do in industry is a bit more pragmatic,” but in line with international standards for risk assessment, Elder says.

“You do a genotoxic risk assessment and identify the bad actors within the synthesis,” he explains. “Then you look to the downstream chemistries to assess the likelihood of that reagent, intermediate, or impurity being carried forward into the API.” Because many GTIs are inherently reactive, the argument is that they are unlikely to survive further reaction steps.

Industry chemists would like to rely on controlling GTIs through quality by design and their process understanding, both of which are principles that regulators encourage companies to follow. But because chemical and mechanistic arguments alone don’t always convince regulators that GTIs won’t be present at a level of concern, “we substantiate the theoretical arguments with an analytical assessment,” Elder says.

A common method is “impurity fate mapping,” or “spike and purge testing,” to monitor the purging capability of a synthetic process. This testing involves spiking the impurity—for example, to a level of several thousand parts per million—where it occurs and then tracking it through the synthesis. Considerable time and effort are put into developing analytical methods to track impurities.

Showing that the GTI is effectively purged by the end of the process allows a specification limit to be set at the point where the impurity appears. During manufacturing, simple analytical tools can be used to check this limit more easily upstream than in the final product. When there’s a high level of confidence even without fate mapping that the GTI won’t persist, “it seems overly bureaucratic to do the analysis on something that occurs 10 steps back,” Elder suggests.

Teasdale and coworkers have proposed a semiquantitative means to assess the risk of a GTI carrying over into the API (Org. Process Res. Dev. 2010, 14, 943). They created “purge factors,” or numerical values assigned to physicochemical parameters, to score the risk of any carryover without having to do exhaustive analytical testing. A comparison with experimental results found that the calculated purge factors were conservative and even indicated where a process has a limited capacity to remove an impurity and might call for other controls.

There is no set point at which to start tracking an impurity, and the decision depends on the process and the nature of the impurity itself. When an impurity arises late in a synthesis, which many consider to be fewer than three or four steps from the end, alternative means may have to be explored to show it is removed or below a level of concern. For example, process chemists may be able to alter physical process parameters or purify a product further to reduce impurity levels.

Targeting starting materials and intermediates, GSK scientists have published an example of how they assessed the fate of five GTIs and controlled them upstream in the process for making the recently approved cancer drug pazopanib (J. Pharm. Biomed. Anal. 2009, 50, 144). The control strategy was part of the firm’s regulatory filing. “We got full endorsement from FDA,” Elder says.

Success is more likely when regulators see “that we have thought of everything that enters the process, what’s formed, and what’s rejected,” says Keith M. DeVries, senior director for process design and development at Eli Lilly & Co. “It’s critical for us to actively articulate our control strategy and the scientific rationale.” In late 2008, he and coworkers at Lilly published a general process and analytical control strategy they routinely use today (Org. Process Res. Dev. 2009, 13, 285).

Control plans usually are laid out for regulators in meetings that come after Phase II clinical trials to test the regulatory waters and prepare for later drug approval submissions, DeVries says. Like other major drug firms, Lilly maintains a cross-disciplinary scientific group that reviews projects and offers guidance during drug development. The group ensures that efforts are in keeping with the company’s overall GTI policy.

In 2005, Lilly began to more systematically evaluate GTIs across its portfolio and for all phases of development. Although addressing the government guidelines has required more resources, DeVries considers the task to be manageable. “We believe we have a scientific and risk-based approach that ensures patient safety,” he says. “It also is achievable and sustainable from both technology and resource perspectives.”

The new regulatory guidelines aren’t applied retrospectively to approved products. That said, any modifications in commercial manufacturing processes or transfers to new production sites can require that spike-and-purge studies be revalidated. And when changes are made in starting materials, suppliers, or the synthetic route, testing is needed to determine whether new GTIs are generated.

Concerns don’t stop with manufacturing the API. Stephen P. Raillard, a scientist in chemical development at XenoPort, in Santa Clara, Calif., collaborated with Lilly scientists and a consultant to look at the possibility that GTIs arise during drug degradation. Although the government guidelines are silent on the subject, industry scientists are starting to look at whether GTIs form during the months-long stability testing process, a development that would have implications for product storage and handling.

The team used available software to evaluate Pharma D3, a pharmaceutical drug degradation database from software maker CambridgeSoft, for structural alerts (Org. Process Res. Dev. 2010, 14, 1015). Among 320 parent drug molecules analyzed, 220 alerting structures were found, Raillard says. At the same time, 336 alerts were found among the more than 1,020 degradants of those parent molecules. “Of those, 155 are unique, meaning they are not similar to the parent or shared with the parent,” he explains.

“This is important because it’s likely that the parent drug has been classified as nongenotoxic, but alerting structures start appearing upon degradation,” he says. “You may have to go the extra mile and show the degradants are not genotoxic.”

Impurities also may arise during formulation and necessitate low-level impurity detection in the final product. For example, sulfonate esters have attracted regulatory attention because sulfonic acid salt forms of APIs are commonly used for their good solubility. When these salts are crystallized in the presence of lower alcohols, potentially genotoxic sulfonate esters are possible by-products.

To understand how these impurities can be controlled during formulation, a group of industry researchers supported by the nonprofit Product Quality Research Institute (PQRI) conducted studies to determine the mechanism and conditions for sulfonate ester formation. Despite the high level of regulatory concern, Teasdale, who led the group, says the researchers found that the esters did not occur to any significant extent under normal API processing conditions. Furthermore, simply adding water and controlling the pH, temperature, sulfonic acid level, and time exposed to the alcohols were adjustable limiting factors.

At the same time the PQRI work was under way, Roche had to recall its HIV drug Viracept (nelfinavir mesylate) because of contamination with ethyl methanesulfonate (EMS) that arose from a manufacturing problem. Roche ended up spending millions of dollars and about a year on extensive toxicology work, which was highlighted in a special 12-article issue of Toxicology Letters in November 2009.

Sulfonate esters show genotoxic effects in bacterial and mammalian cell assays, but in animal studies, Roche researchers found a threshold level below which EMS does not have a harmful effect on DNA. EMA has accepted this threshold, which is five orders of magnitude higher than the recommended TTC, along with other risk mitigation measures for controlling the impurity in Viracept.

Other GTIs no doubt have thresholds, and knowing them would be valuable, but the amount of work done by Roche for just one compound was “huge,” Teasdale points out. Industry needs to think about how to extend such information within chemical classes, he suggests, but he’s not convinced that the impetus to do so will come from industry. “Is anybody prepared to bite the bullet, probably through some kind of industry collaboration, and find the several million dollars that’s required and do the three or four years of work that needs to be done?”

EMA has been rumored to have a list of chemicals with agreed-upon acceptable limits that it might offer as an appendix to its guidance. Although EMA won’t confirm its existence, such a list “would be very helpful in order to avoid duplication of work and to ensure the same limits are applied for the same GTI,” the agency tells C&EN. “As internationally harmonized limits would be most appropriate for this purpose, we would propose generating such a list in the framework of ICH,” the International Conference on Harmonization of Technical Requirements for Registration of Pharmaceuticals for Human Use.

Genotoxic impurities will move to that bigger stage this fall and become an ICH discussion topic. Regulatory watchers say a concept paper could appear soon, followed by a first official draft toward the end of 2011.

Because FDA and EMA policies are so similar, harmonization seems to be a fait accompli. But the pharmaceutical industry hopes the discussion will allow it to “get its point across and the right issues on the table,” Teasdale says. Industry participants hope the work delves into the details and provides clarity on how the guidelines are applied to the evaluation, qualification, and control of GTIs in medicines.