The Viracept (nelfinavir)—Ethyl methanesulfonate case

Abstract

In May 2007, the F. Hoffmann-La Roche Company became aware of a contamination of Viracept (nelfinavir) tablets by the mutagenic DNA-ethylating agent ethyl methanesulfonate (EMS) as a result of a production incident. HIV-patients could have been exposed for 3 months to daily doses of up to 2.75 mg EMS, i.e., about 50 μg/kg per day. In this special issue, 12 manuscripts have been assembled to provide comprehensive insight in what happened and how the incident was managed by Roche and handled by the regulatory agencies. In the first four papers, the course of events and the toxicological information available at the outset are summarized and a traditional cancer risk assessment on the basis of a linear default dose–response is made. Three articles then report on the experiments performed for an improved risk assessment. A standard 4-week toxicity study with EMS in the rat indicated an NOAEL of 20 mg/kg per day. Extensive studies on the genotoxicity showed threshold-like dose responses for both chromosome damage (bone marrow micronucleus test) and gene mutation (lacZ transgenic MutaMouse test) in various organs of mice treated for up to 4 weeks, whereas ethylation of hemoglobin at the N-terminal valine increased linearly with dose. The difference between adduct formation in DNA and protein was interpreted by repair of DNA adducts that becomes saturated above a threshold concentration of EMS, regarded as the metrics for the rate of DNA ethylation. Elaborate toxicokinetic investigations in various animal species, coupled to appropriate modeling, were performed in order to extrapolate the animal data to humans. Using a threshold risk assessment based on estimated c_max of EMS, a safety factor of 370 was derived for maximum doses ingested by Viracept patients. A number of critical points are addressed in this editorial, concerning (i) definitions and types of “thresholds”, (ii) estimation of a confidence limit for a slope below the threshold dose, interpreted as an increment within background variation, (iii) implementation for other mutagens and for human risk assessment.

Introduction

Following reports by patients of a strange smell of Viracept^® 250 mg tablets, analysis revealed that Roche's protease inhibitor nelfinavir mesylate produced between March 2007 and May 2007 was contaminated with up to 2300 ppm ethyl methanesulfonate (EMS), as a result of a production accident (Gerber and Toelle, 2009). This led to a global recall of the drug on June 6th 2007 and initiated an extensive testing strategy in close coordination with regulatory agencies (Müller and Singer, 2009) in order to establish a comprehensive risk assessment for an estimated daily EMS dose of 0.055 mg/kg/d in patients taking 10 tablets daily over 3 months (Müller et al., 2009).

EMS is a genotoxic carcinogen, inducing DNA damage and mutation by DNA ethylation. Although EMS has widely been used as model mutagen, the database regarding carcinogenic potency is poor and information on genotoxicity limited to high dose levels (Gocke et al., 2009b). An initial cancer risk assessment had therefore to be based on very limited data and the default procedure for extrapolation to low dose, i.e., a linear dose–response relationship. A maximum cancer risk based on lifetime exposure was estimated to be on the order of 10⁻³, a 3-month exposure was considered to be associated with less than one additional tumor in 10,000 exposed patients (Gocke et al., 2009c).

In view of the absence of reliable toxicity data, a large testing program was set up. This included a standard 4-week toxicity study in Wistar rats that showed a NOAEL of 20 mg/kg per day (Pfister and Eichinger-Chapelon, 2009). A major effort was made for in vivo genotoxicity studies, including the mouse bone marrow micronucleus test (MNT) and the MutaMouse assay for lacZ mutant induction. A wide dose range was covered and periods of exposure were up to 28 days (Gocke et al., 2009a). The rationale for these efforts was a report on threshold-like dose responses for a number of endpoints of genotoxicity by EMS in cultured cells (Doak et al., 2007). Ethylation of the N-terminal valine of mouse globin was determined in parallel (Gocke et al., 2009a). The data showed pronounced sublinearity in the dose–response (slope increasing with dose) for EMS-induced genotoxicity, while globin ethylation increased approximately linearly with dose. Doses up to 25 mg/kg per day (for lacZ mutant frequency) and up to 80 mg/kg per day (for micronucleus induction) did not produce any deviation from background measures in mouse bone marrow, while the slope above this putative threshold dose was prominent. Statistical assessment of the data (Gocke and Wall, 2009) indicated that the null hypothesis of linearity could be rejected with high significance against a hockey stick model (Lutz and Lutz, 2009). The distinct threshold-like appearance of the dose–response curve can be explained by saturation of the alkylguanine DNA alkyltransferase repair system above a certain threshold dose. This occurs whenever the rate of adduct formation exceeds the rate of instantaneous repair.

Under these conditions, the concentration of EMS at the DNA rather than the AUC becomes the critical dose metrics. In order to estimate c_max and AUC in humans for the estimated dose of about 0.05 mg/kg per day, a comprehensive characterization of the kinetics of EMS in vitro and in vivo was necessary (Lavé et al., 2009a), including translation of the data to exposed patients by appropriate modeling (Lavé et al., 2009b). This all done, a comprehensive human risk assessment was performed, based on the assumption of a threshold and using margins of exposure (Müller et al., 2009).

Contamination of drug mesylates by EMS is a general pharmaceutical problem, and a daily dose of 1.5 μg EMS per patient is considered the threshold of toxicological concern (TTC). In view of the threshold-type risk assessment described here, a permitted daily exposure (PDE) for EMS of 100 μg per person is proposed (Müller and Gocke, 2009). This exposure would still include a product safety factor of 12,000.

The last contribution to this special issue discusses impact and provides an outlook by external experts in the field (Walker et al., 2009). The authors recommend biologically based models in lieu of default models also for risk assessments for DNA-reactive compounds.