Determination of bromadiolone residues in fox faeces by LC/ESI-MS in relationship with toxicological data and clinical signs after repeated exposure

Environmental Research 110 (2010) 664–674 Contents lists available at ScienceDirect Environmental Research journal homepage: www.elsevier.com/locate/envres Determination of bromadiolone residues in fox faeces by LC/ESI-MS in relationship with toxicological data and clinical signs after repeated exposure$, $$ Mickaël Sage a,n, Isabelle Fourel b, Michaël Cœurdassier a, Jacques Barrat c, Philippe Berny b, Patrick Giraudoux a a b c University of Franche-Comte, Department of Chrono-Environment, UMR UFC/CNRS 6249 UsC INRA, 25030 Besanc- on Cedex, France VetAgro Sup. College of Veterinary Medicine, Toxicology Laboratory, UMR INRA 1233, BP 83, 69280 Marcy l’Etoile, France French Food Safety Agency—Unité épidémiosurveillance des maladies de la faune sauvage, Domaine Piéxécourt, BP 9, 54220 Malzeville, France a r t i c l e in f o a b s t r a c t Article history: Received 28 January 2009 Received in revised form 12 July 2010 Accepted 14 July 2010 Available online 7 August 2010 In many countries, the fox (Vulpes vulpes), predator of small mammals, is particularly affected by anticoagulant rodenticides such as bromadiolone due to secondary poisoning. Nevertheless, to date, no method of exposure monitoring is applicable in the field over large areas, and no toxicological data are available concerning sensitivity of foxes to bromadiolone. The aim of this work was to compare excretion kinetics of bromadiolone in fox faeces with clinical and haemostatic effects after repeated exposure to intoxicated voles. A sensitive method for the quantification of bromadiolone excretion in fox faeces and plasma was developed, using liquid chromatography combined with electrospray ionisation mass spectrometry (LC/ESI-MS). The LoD was 0.9 mg/kg and 0.15 mg/L, and the LoQ was 3.0 mg/kg and 0.5 mg/L, in faeces and in plasma, respectively. Four captive foxes were fed for 2 or 5 days with water voles (Arvicola terrestris Sherman) spiked with bromadiolone at concentrations close to those measured in the field. Faeces and blood were collected for bromadiolone titration, and blood-clotting tests were performed to monitor fox health daily during 10 days and then every 3–4 days until the end of the experiment (D28). Then, after euthanasia, a complete necropsy was performed, and levels of bromadiolone residues in the liver were determined. Bromadiolone residues were detected in faeces 15 h after the first exposure. They increased dramatically during the exposure period and then gradually decreased, but they remained detectable at the end of the experiment, i.e., 26 days after the last exposure. Bromadiolone residues in plasma showed a similar pattern but were no longer detectable 7–24 days after the last exposure. Two foxes presented very severe external haemorrhages, requiring the administration of the antidote vitamin-K1. Bromadiolone residues in faeces and their relationships with exposure and other direct-markers that were measured are discussed. Liver residues and the toxicity data of our study will help to interpret data from fox carcasses collected by wildlife disease surveillance networks. These findings provide a basis for programs aiming to monitor the exposure of wild fox populations to bromadiolone using non-invasive methods based on standard sampling and analysis of residues in faeces. & 2010 Elsevier Inc. All rights reserved. Keywords: Second generation anticoagulant rodenticide (SGAR) Secondary poisoning hazard Exposure monitoring Faeces Liver and blood plasma bromadiolone residues Blood-clotting time $ Information on funding sources: This study received support from the Conseils généraux of the Doubs and the Jura departments and from the Région of Franche-Comté. Studies on experimental animals: Experimentations were achieved at the experimental farm of AFSSA Nancy (Agence Franc-aise de Sécurité Sanitaire des Aliments—French Food Safety Agency), a laboratory animal facility registered by the French Ministry of Agriculture, in accordance to European and French legislation on Laboratory Animal Care and Use (French Decree 2001-464 and European Directive CEE86/609). All staff was qualified through mandatory trainings, and worked under veterinary supervision. Ethical issues were considered by the research team in the protocol design, as required by the French legislation: animal housing was respectful of the species’ needs and animal handling was limited. Based on previous studies on dogs (Mount, M.E., 1988. Diagnosis and therapy of anticoagulant rodenticide intoxicants. Vet. Clin. N. Am. Small Anim. Pract. 18, 115–130), experimental end-points were defined by the maximal level of prothrombin time (greater than 50 s) in the blood coagulation test, or every unexpected health deterioration such weakness or anaemia, or when haemorrhages were observed. In these cases, foxes were treated with the antidote Vitamin-K1 (Mount, 1988). n Corresponding author. Fax: + 33 381 665 797. E-mail address: mickael.sage@univ-fcomte.fr (M. Sage). $$ 0013-9351/$ - see front matter & 2010 Elsevier Inc. All rights reserved. doi:10.1016/j.envres.2010.07.009 M. Sage et al. / Environmental Research 110 (2010) 664–674 1. Introduction Second generation anticoagulant rodenticides (SGARs) are currently used for controlling small mammalian pests. Evidence of exposure on non-target fauna has been reported worldwide (e.g., Britain: McDonald et al., 1998; Shore et al., 2003a; France: FournierChambrillon et al., 2004; Lambert et al., 2007; USA: Stone et al., 2003; New Zealand: Eason et al., 2002). The red fox (Vulpes vulpes) is the most widespread fox species, found in both rural and urban habitats throughout the world (Mitchell-Jones et al., 1999). Depending on food availability, its diet can be mainly or even exclusively composed of live or dead rodents (Weber and Aubry, 1993). This makes it particularly susceptible to exposure and possibly secondary poisoning where and when rodenticides are used (e.g., Barnett et al., 2006; Berny, 2007; Berny et al., 1997; Shore et al., 2003b). In some countries, particularly in Europe, the SGAR bromadiolone is intensively used in the field. It is the only rodenticide authorised in France for controlling the population outbreaks of the water vole Arvicola terrestris Sherman (Ministe re de l’Agriculture et de la Pêche, 2002). Some 400 tonnes of bromadiolone baits are used by farmers annually in the country (Liphatech, personal communication). Vole control operations using bromadiolone are undertaken over large areas (hundreds of km2), and dozens to hundreds of secondary poisonings of red foxes have been reported each year (Berny et al., 1997; Kupper et al., 2006; SAGIR, 1990–2007). During fox population surveys carried out in eastern France, no foxes were counted locally along standard night road side counts the year following a treatment (Raoul et al., 2003). Ecotoxicological risk is a function of both exposure and toxicity. While non-invasive direct-markers (such as bloodclotting time or anticoagulant analysis in plasma) have been used, this has been mainly in domestic species (Boermans et al., 1991; Mount and Feldman, 1982) and, requires live trapping of animals which can be difficult when studying wild carnivores and raptors. To date, there are no available methods for measuring the exposure of fox populations to SGARs over large areas. Therefore, for practical and ethical reasons, tissue analysis on carcasses collected from wildlife surveillance networks is the only method that is currently widely used to determine the level of exposure to SGARs of foxes and other predators (e.g., Alterio, 1996; Alterio et al., 1997; Barnett et al., 2006; Berny et al., 1997; Brown et al., 1998; Eason et al., 1999; Eason and Spurr, 1995; FournierChambrillon et al., 2004; McDonald et al., 1998; Murphy et al., 1998; Shore et al., 2003a; Walker et al., 2008). However, searching for carcasses typically has a low success rate (Howald et al., 1999). Therefore, the impact of poisoning on predator populations may be underestimated and records of poisoning are only obtained after mortality has occurred (Brakes and Smith, 2005). In previous laboratories experiments, the analysis of SGARs (but not bromadiolone) in regurgitated barn owl pellets has been used to monitor exposure in owls (Eadsforth et al., 1991; Gray et al., 1994; Newton et al., 1994). We hypothesised that similarly, bromadiolone residue analysis in faeces could be used as a non-invasive indirect method of monitoring the exposure of foxes in the field. Indeed, faeces are the major excretion route for bromadiolone. For instance, 53.3% of a radiolabelled dose was excreted by rats within 2 days (Lipha, 1987). However, no information is available concerning anticoagulant excretion via faeces for foxes. The utilisation of radiolabelled molecules is inconceivable in the field over large areas, and current analytical methods using high performance liquid chromatography (HPLC) and gas chromatography (GC) (Erickson and Urban, 2002; Kemikalieinspektionen, 2006; USEPA, 1998) are neither selective nor sensitive enough to detect and quantify bromadiolone in faeces. HPLC coupled with a mass spectrometry detector (LC/MS) is a method known to allow quantification at trace levels in complex matrices because of its 665 high specificity and sensitivity (Jin et al., 2007). This method could therefore be used, but no analytical methods have been validated on anticoagulant in faeces. Although foxes may be at high risk of exposure, to our knowledge, no data are available concerning bromadiolone toxicokinetics and toxicodynamics for this species (AGRITOX INRA, 2004; Erickson and Urban, 2002; Giraudoux et al., 2006; Kupper et al., 2006; USEPA, 1998). Neither blood-clotting time nor bromadiolone plasma kinetics has been used to investigate the effects on foxes of feeding on bromadiolone-poisoned rodents. Analytical method for bromadiolone titration (Grobosch et al., 2006; Jin et al., 2007) have not been validated for fox blood, nor has a relationship between concentrations in tissues and effects been established (Erickson and Urban, 2002). Thus, the likelihood of fox mortalities following field operations to control water voles is poorly understood (Giraudoux et al., 2006). The first objective of this study was to develop a new LC/MS method for bromadiolone titration in fox faeces and plasma. Secondly, we determined the excretion kinetics of bromadiolone in faeces of captive foxes fed a repeated dose of bromadiolone. The dose used was of a similar magnitude to that likely to be experienced by foxes eating poisoned water voles in the field. The main purpose here was to determine how long bromadiolone residues remain detectable after exposure. At the same time, we investigated the toxicological, clinical and haemostatic effects on those foxes by monitoring bromadiolone kinetics in plasma, residues in liver and blood-clotting tests (Erickson and Urban, 2002; USEPA, 1998). 2. Materials and methods 2.1. Fox exposure Experiments were conducted at the experimental farm of AFSSA Nancy (Agence Franc- aise de Sécurité Sanitaire des Aliments—French Food Safety Agency; agreement nb. A54747). The animal used in this study was the silver fox, which belongs to the same species as the ‘wild’ red fox, (Vulpes vulpes). A mixed sex group of five foxes, weighing 6.8–7.4 kg (mean¼ 7.0) were individually caged. They were born at the experimental farm of our laboratory, were 6 years old, and had never been exposed to anticoagulants. To simulate potential field exposure to bromadiolone, multiple low-dose oral administrations were used. As captive foxes did not eat easily always consume all of the rodents provided, they were fasted for 1 day prior to the feeding trial. According to Artois (1989), foxes eat between 0.3 and 0.6 kg of food per day, which represents 4–8 water voles. So, on each exposure day, captive foxes were fed 5 water voles trapped in an area that had never been treated with bromadiolone before (no bromadiolone residues were detected in liver of 5 randomly sampled voles). Foxes that were to be dosed with bromadiolone were fed voles that had been spiked with a quantity of bromadiolone similar to that found in voles during 20 days after treatment in a field study (Sage et al., 2008). This consisted of 200 mg of bromadiolone/vole (batch 511101, Liphatech, France) dissolved in 1 mL of ethanol (Carlo Erba analytical grade) that was injected in five places (i.e., 1000 mg of bromadiolone per day per fox). As bromadiolone metabolites and their eventual toxicity/interaction with the parent molecule remain unknown (Kemikalieinspektionen, 2006), it was impossible to consider this issue in our study. Unfortunately, The LD50 for foxes is unknown (Erickson and Urban, 2002; USEPA, 1998), but the quantity administered in our study was very similar to the lowest lethal dose for dogs reported in literature, i.e., 150 mg/kg/day for 5 days (Kolf-Clauw et al., 1995). This translates to 1050 mg/day for a 7.0 kg fox. Five foxes were available for this experiment, and the exposure operation was conducted on two consecutive days (D0 and D1) for one male (called F2) and for five consecutive days (from D0 to D4) for two males and one female (called F5.1, F5.2 and F5.3, respectively). As a control, one female fox (called F0) was fed during 5 days with rodents that were not poisoned but were spiked only with 1 mL of pure ethanol. After the exposure period, all foxes were fed with non-contaminated water voles until D10; then they were fed on their usual diet (dry food) until D28 (end of the experiment). Water was provided ad libitum for every fox. 2.2. Sampling Faeces were collected and immediately stored at  20 1C in the dark (Morin, 1988; Wright, 2002) for subsequent bromadiolone titration. Collection took place immediately before exposure (D0), 15 and 24 h after the first ingestion (H15 and 666 M. Sage et al. / Environmental Research 110 (2010) 664–674 D1), then daily for the first 10 days of the experiment and every 3 or 5 days (spacing out gradually) over the next 28 days (D2 to D10 and D12, D15, D17, D22 and D28). Blood was collected from the jugular vein into 10 mL vacuum tubes containing 1 mL of trisodium citrate (0.109 M) on the same sampling days as faeces. All blood samples were centrifuged immediately for 15 min at 2500g, and plasma was separated. Blood-clotting tests were determined, and the rest of the plasma was stored at  20 1C for later bromadiolone titration. All foxes were sacrificed by a T61s intravenous injection (Hoechst Roussel Vet, Brussels, Belgium) after 28 days. A complete necropsy was performed, and the liver was collected for bromadiolone titration. 2.3. Bromadiolone titration in faeces and plasma 2.3.1. Solvents and materials All chemicals, reagents, and solvents used were of analytical grade. Methanol was obtained from Carlo Erba (code no. 412532), acetone from Riedel de Haën Pestanal (code 34480), dichloromethane from Sigma-Aldrich (34856), hexane (1.04367.1000) and acetonitrile (1.14291.2500) from Merck, and ammonium acetatate (MS grade) from Fluka. Ultra-pure water was supplied by a Milli-Q Plus water purification system from Millipore. Bromadiolone (batch 511101) and difenacoum (batch 0395167) as an internal standard (IS) (Lipha-Tech, France) were dissolved in methanol to make individual stock solutions of 100 mg/L. Those where stored at  20 1C until they were diluted, as needed, with methanol to prepare spiking solutions at concentrations of 1, 1.5, 5, 10, 25, 50, 100, 175, 250, 500 mg/L for bromadiolone and 500 mg/L for IS. 2.3.2. HPLC/MS system Analytical methods for development and validation were performed on a 1100 Series ESI/LC/MSD ion Trap VL, consisting of an on-line solvent degasser, binary pump, autosampler and column temperature module, with an electrospray ionisation (ESI) source and ion trap analyser, all controlled with the HP Chemstation software (Agilent Technologies, Palo Alto, CA, USA). Chromatographic separation was performed using a Zorbax Eclipse XDB-C18 (2.1 mm  100 mm, 3.5 mm) column and a Zorbax Eclipse XDB-C8 (2.1 mm  12.5 mm, 5 mm) guard column from Agilent Technologies with a mobile phase of A: 10 mM ammonium acetate and B: methanol. The mobile phase gradient elution was 30%A:70%B (v:v) at 0 min, increasing to 80%B from 0 to 5 min, and to 90%B at 6 min, holding at 90%B during 4 min and back at 70% B at 11 min, for a total run time of 18 min. The column temperature was 30 1C. The temperature of the autosampler tray was set to 5 1C. The system was operated at a flow rate of 0.250 mL/min. The injection volume was 1 mL. The mass spectrometric conditions were optimised by infusing solutions of pure standards diluted in mobile phase. The capillary voltage was 2828 V. The source temperature was maintained at 350 1C. The settings for nebuliser gas (N2) were 40 psi, and the source gas flow was 8 L/min. The high vacuum in the trap was 1.1  10  5 mbar. The mass spectrum of bromadiolone showed deprotonated precursor ions ([M–H]  ) at m/z 527 and 525 (diastereoisomers). The mass spectrum of the IS showed a precursor ion ([M–H]  ) at m/z 443. Identification of bromadiolone was confirmed by the LC retention time compared to a calibration standard and the double m/z 527 and 525. Although the Atmospheric Pressure Chemical Ionisation (APCI) source was also tested, the Electro-Spray Ionisation (ESI) source, operated in the negative ion mode (  ESI), gave higher sensitivity for the bromadiolone, and its IS. MS/MS quantification was tested on spiked samples but did not show better sensitivity compared to MS. This was due to a poor fragmentation rate for bromadiolone. methanol and vortex mixed for 30 s. These solutions were used for LC/MS analyses. If samples were too concentrated and were not within the concentration range of the calibration curve, clean-up by solid phase extraction was repeated on another 4 mL of supernatant, and eluate was reconstituted with 1 mL of methanol and vortex mixed for 30 s. 2.3.4. Plasma extraction procedure Samples were thawed at room temperature and shaken for 10 s with a Vortex. For each sample, 1 mL was extracted with liquid–liquid extraction tubes (Toxitube Bs, Varian Inc., Lake Forest, CA, USA), and 200 mL of IS solution (2.5 mg/mL) was added. Extraction was conducted according to the manufacturer procedure and Berny et al. (2006). The tubes were shaken by gentle inversion for 2 min and centrifuged at 1800 rcf for 10 min. The upper layer was removed and evaporated to dryness under a stream of nitrogen at 50 1C. The dry residues were reconstituted with 200 mL of methanol. These solutions were used for LC/MS analysis. If the samples were too concentrated and were not within the concentration range of the calibration curve, these solutions were diluted 5-fold in methanol. 2.3.5. Validation of the analytical procedures for faeces and plasma samples Specificity was determined through assessment of four blank faeces samples or four blank plasma samples obtained from foxes before exposure (D0). Solvents were also checked for contamination with substances that could potentially interfere with analyses. Background noise levels associated with the analyses were determined based on the findings from blank faeces or plasma samples; for each analysis, a limit of detection (LoD; three times the noise level) and a limit of quantification (LoQ; ten times the noise level) were determined theoretically. To check the validity of the analyses, spiked samples were prepared and analysed until peaks measured on the chromatogram corresponded to the determined limits. Linearity was determined with ten-point calibration curves on both standard solutions and spiked samples. Preliminary titrations of faeces and plasma, sampled the last day of the intoxication period (theoretically the most concentrated) and at the end of the experiment, the last day of the postintoxication period (theoretically the least concentrated), provided the linear range. A linear regression curve was determined from the ratio of the peak area of bromadiolone to the peak area of IS and compared to the ratio of response observed in a calibration standard. Percentages of recovery were determined for six samples spiked at the same concentration (middle of the linear range) to verify that bromadiolone and internal standards were extracted to a similar extent and that extraction conditions were acceptable compared with expected concentrations. Because bromadiolone concentrations were determined by calculation of the ratio of the bromadiolone peak area to the peak area of its IS, results were not corrected for % recovery based on date for bromadiolone spiked samples. The repeatability of the extraction procedure and LC/MS analysis was determined with six samples. Coefficients of variation (SD  100/mean) were determined for bromadiolone and IS. All diluted standard solutions were shown to be stable for 24 h at room temperature (20–25 1C), for 1 week at 5 1C and for at least 1 month at  20 1C. Sample solutions are stable for 24 h at 5 1C, and analyses were always conducted within 10 h of sample extraction; meanwhile, they were stored at 5 1C. For quality assurance purposes, faeces and plasma samples spiked with bromadiolone, as well as blank samples, were assessed each day of the analysis. 2.4. Bromadiolone titration in liver 2.3.3. Faeces extraction procedure Samples were crushed, still frozen, to obtain a homogeneous mixture and were dried during 48 h at 30 1C. Fresh mass (F.M.) and dry mass (D.M.) were measured. In order to compare bromadiolone excretion between each fox in this experiment, as well as between our results and field data, faeces bromadiolone concentrations were expressed in mg of bromadiolone per kg dry matter (D.M.). From those concentrations, the quantity of bromadiolone excreted in the faeces was calculated. An aliquot of 0.5 g (D.M.) was weighed and placed in a 100 mL tube containing IS (200 mL of a 2.5 mg/mL solution) and 10 mL of acetone. This was homogenised for 30 s using an Ultra Turrax tissue disperser. Ten millilitres of acetone was poured into another tube to carefully rinse the disperser for 30 s and was added to the first tube. The tissue disperser was washed with water and soap and rinsed with absolute ethanol between each sample. The homogenate was centrifuged at 2100 rcf for 10 min. Four millilitres of supernatant was taken and evaporated to dryness under a stream of nitrogen at 50 1C. The residue was reconstituted with 2 mL of dichloromethane/hexane (50:50), shaken for 20 s with a Vortex (MS2 Minishaker IKA-WERK, Staufen, Germany) and cleaned up by solid phase extraction as follows: extract was loaded onto a Sep-Paks Silica cartridge Vac 3cc (Part no. WAT020810 Lot no. 044034057C Waters, USA). Each cartridge was rinsed with an additional 1 mL of dichloromethane/hexane (50:50) to fix the two compounds and wash away impurities, and it was then dried, under vacuum for 10 s. The analytes were eluted twice with 1.5 mL of dichloromethane/ acetonitrile (50:50), and the cartridge was dried under vacuum for 10 s to obtain a quantitative elution. The eluate was then dried, reconstituted with 200 mL of Samples were thawed at room temperature. Liver processing and bromadiolone concentrations were determined by High Performance Liquid Chromatography (HPLC) according to the method derived from Hunter (1983a; 1983b) and described by Giraudoux et al. (2006) for rodent liver with difenacoum (as an internal standard) and fluorescence detection. 2.5. Blood-clotting test Activated partial thromboplastin time (APTT); kaolin-activated partial thromboplastin time (kAPTT/KCT) and prothrombin time (PT) are the most commonly used screening assays for anticoagulant intoxication (Mount, 1988; Woody et al., 1992). They were assessed immediately after collection of the blood samples on centrifuged plasma by use of an Option 2+ apparatus (Biomérieux, Marcy l’Etoile, France). Kit reagents used were STAs—PTT A; C.K. Prests and Neoplastines Cl, respectively (Diagnostica Stago, Asnie res, France). Clotting-time tests were stated in seconds according to the manufacturer’s instructions. As reference values for foxes were lacking (Beklova et al., 2007) and no data were available with those kit reagents, those tests were achieved on ten uncontaminated foxes before the experiment. Each value was determined as the mean of two measurements; if there was a difference higher than 10% between these two values, clotting-time determination was repeated. M. Sage et al. / Environmental Research 110 (2010) 664–674 2.6. Therapeutic protocol Based on previous studies on dogs (Mount, 1988), we decided to treat the foxes with vitamin-K1 (Roche, Fontenay-sous-Bois, France) using intravenous injections (5 mg/kg) followed by oral administration (2.5–5 mg/kg until stability of normal PT). Foxes were treated when PT was greater than 50 s, when foxes had clinical signs such as weakness or anaemia, or when haemorrhages were observed. Under such circumstances, and according to the mode of action of the antidote, blood-clotting time should quickly decrease, but the kinetic curve of bromadiolone concentration in faeces and plasma should not be affected (Woody et al., 1992). After 2 days of vitamin-K1 administration, therapy was suspended. If no increase in blood-clotting time was observed, therapy was stopped. 2.7. Statistics The normality of blood-clotting test data was tested using the Kolmogorov– Smirnov test. Means and median were estimated by bootstrapping (replicates¼ 1000). The mean and the standard error (SE) were given if the variable proved normally distributed; if not, the median and the 95% empirical confidence interval (CI) were computed from the bootstrap replicate distribution. Due to lack of normality and problematical transformations to Gaussian distribution, comparisons were generally based on nonparametric statistics (Siegel and Castellan, 1988; Sokal and Rohlf, 1997). The Wilcoxon and Mann–Whitney U test was used to compare two independent samples. Blood-clotting time values were considered to be increased if they were 25% greater than the mean preintoxication values (Johnstone, 1988; Woody et al., 1992). Determination of the absorption kinetics of bromadiolone was not possible in our study due to the dosing regimen, with repeated exposure and sampling not frequent enough during the first hours of the experiment. However, bromadiolone has previously been reported to exhibit a biphasic and exponential decay pattern in different tissues (Erickson and Urban, 2002), and its persistence in plasma and faeces was analysed individually for each fox using a one-compartment model with the following equation, derived from Widianarko and Van Straalen (1996): Ct ¼ Ce ekðtteÞ ð1Þ Table 1 Results of the validation procedures for assays of bromadiolone in faeces and plasma samples. a Repeatability for bromadiolone (%) Repeatability for internal standard (%)a Linearity (R2) Linear range (mg/L) (standard solutions) Extraction efficiency of bromadiolone (%) Extraction efficiency of internal standard (%) Limit of detection (LOD) Limit of quantification (LOQ) Mean retention time for bromadiolone (min) Mean retention time for internal standard (min) Plasma 8.9 8.8 4.3 3.4 40.986 2–500 63.5 7 5.6 65.9 7 5.8 0.9 mg/kg D.M. 3.0 mg/kg D.M. 3.6 4.8 0.5–100 91.0 7 3.9 89.0 7 3.0 0.15 mg/L 0.5 mg/L Coefficient of variation. where Ct is the bromadiolone concentration in the tissue (mg/kg or mg/L for faeces and plasma, respectively), Ce the maximal concentration observed the day after the last exposure, k the concentration decrease rate (day  1), t the time since the beginning of the experiment (days) and te the day after the last exposure (day  1). Disappearance half-lives (DT50) were calculated according to the following (Toutain and Bousquet-Mélou, 2004): DT50 ¼ lnð2Þ k ð2Þ Biologicaly, DT50 correponds to the time for the plasma to eliminate or degradate half of the bromadiolone present the day after the last intoxication. In regards to faeces, it corresponds to the time for each fox to excrete half of the bromadiolone that it excreted on the first day post-exposure. The model was fitted using non-linear regression. Different variance functions were used to model the variance structure of the within-group errors using covariates (Pineiro and Bates, 2000). Models were compared using the information theoric approach, as outlined by Burnham and Anderson (2004) and Sakamoto et al. (1996). The differences in the parameter estimates between foxes were juged from the overlap of the 95% confidence intervals. Statistics were performed using R2.4.1 (R Development Core Team, 2004). 3. Results and discussion 3.1. Validation of the bromadiolone method titration in faeces and plasma The validation procedure of bromadiolone titration with the HPLC/ESI-MS assay in faeces and plasma extracts gave satisfactory results (Table 1). Although extraction rate was smaller in faeces than in plasma (Wilcoxon, Mann and Whitney; P¼ 0.02 for bromadiolone and IS), extraction was good for the experimental conditions, since bromadiolone was quantifiable from the first to the last day of our experiment in all faeces samples. There was a good repeatability of the extraction procedure (coefficients of variation o10%). A typical LC/ESI/MS chromatogram monitoring the titration of faeces containing 0.9 mg of bromadiolone/kg is shown in Fig. 1. The validation procedure gave satisfactory results, and this technique offers much-improved sensitivity and selectivity over other existing HPLC methods. For example, in fox plasma, our LoD and LoQ are 30–40-fold lower than values observed for bromadiolone in sheep as measured by HPLC (Berny et al., 2006). Sensitivity in faeces increased approximately 10–30fold in comparison to previous studies (Eadsforth et al., 1991; Gray et al., 1994; Newton et al., 1994) with other SGARs in barn owl pellets. In comparison with recently published HPLC/ESI-MS– MS methods, our LoQ in plasma was the same as results reported by Jin et al. (2007) in human whole blood, while it was 10 times more sensitive than results reported by Grobosch et al. (2006) in human blood serum. In faeces, our LoQ was similar to those observed in animal feed, ground beef and drink mix by Marek and Koskinen (2007). b Rel. Resp. x104 a 1500 1000 525.1 500 527.0 1 0.5 x104 . Rel. Resp. Intensity a Faeces 667 443.0 6 3 0 0 1 2 3 6 4 5 Time (min) 7 8 9 300 400 500 m/z 600 Fig. 1. Chromatogram and mass spectrum for a blank faeces sample spiked at the LoD 0.9 mg/kg: (a) representing the bromadiolone peak and (b) the IS. 668 M. Sage et al. / Environmental Research 110 (2010) 664–674 3.2. Fox feeding phase In the present experiment, foxes were fed with spiked water voles. Theoretically, this type of exposure allows controlling the precise dose of bromadiolone ingested by each fox but it did not consider the toxicity of eventual metabolites, binding or the distribution of the toxicant in intoxicated rodents. Fox F5.3 did not eat all the spiked voles, fox F2 ate little, and parts of rodents discarded by these foxes were eaten by insects. As we did not know the precise quantity of bromadiolone ingested by these two foxes, we expressed exposure semi-quantitatively; fox F2—low dose, F5.3—moderate dose, F5.1 and F5.2—total dose. 3.3. Kinetics of bromadiolone residues in faeces There were no detectable bromadiolone residues in the faeces of the control (F0) fox during the experiment. In contrast, bromadiolone was rapidly excreted by the intoxicated foxes and was detectable in all faeces from the first sampling, 15 h after dosing (Fig. 2a). Concentrations increased for all foxes during the dosing. Maximal residues were observed in the middle of the exposure, on D3 for F5.1 with 6226.0 mg/kg, the next day of the exposure, on D5, for F5.2 and F5.3 with 5196.8 and 6320.8 mg/ kg, respectively, and 2 days after the last exposure, on D3, for F2, with 480.3 mg/kg. Despite the fact that all foxes exposed during 5 days did not ingest the same quantity of vole, the voles they did eat were contaminated with the same concentration of bromadiolone. Thus, the maximal concentrations in faeces of F5.1, F5.2 and F5.3 were very close (mean and standard error of the mean: 5917.5  623.4 mg/kg) and were, respectively, 12.9, 10.8 and 13.2 times greater than that of F2. The very small consumption of contaminated voles by F2 induced an increase in bromadiolone excretion that was slower and a maximal concentration that was much lower than those of the other foxes. After this increase, concentration levels gradually decreased once dosing ceased. The half-life of bromadiolone in faeces ranged between 23.3 and 43.9 h and was shortest for F2 but similar for the three F5 foxes (Table 2). A small amount of bromadiolone was still detectable in faeces of all intoxicated foxes until the end of the experiment and concentrations measured in faeces on D28 were 4.4, 53.0, 28.1 and 15.0 mg/kg for F2, F5.1, F5.2, and F5.3, respectively. These were from 7.8 to 96 fold above the LoD. Since bromadiolone residues were significantly above our LoD at the end of the experiment, we may suppose that the molecule is excreted and detectable longer than 26 days after exposure. The stall conditions of F5.1 and F5.2 means that most (475%) but not all faeces could be collected daily. Measured concentrations might therefore be generally considered as representative of the total faeces excreted, but the quantity excreted per day for these two foxes might be underestimated. Despite that, the moderate consumption of voles by F5.3 was associated with lower faecal excretion of bromadiolone than observed in foxes F5.1 and F5.2 (Fig. 3a). The quantity of excreted bromadiolone reflected the vole consumption for each fox, and maximal values were observed either in the middle of the exposure, the next day or 2 days after exposure, with 6.2, 69.9, 73.3 and 26.6 mg for F2, F5.1, F5.2 and F5.3, respectively. The highest concentrations were observed on D3 (F2 and F5.1), and D5 (F5.2 and F5.3) (Fig. 2a). However, in comparison to concentrations, the quantity of bromadiolone excreted on some days may be low, due to a smaller mass of faeces on those particular days (i.e. D4 for F5.1 and F5.2, D10 for F5.3 and D15 for F5.1) (Fig. 3a). The greatest portion of bromadiolone ingested on D0 by F5.1 and F5.2 seems to have been excreted and sampled from 15 h after exposure. As no new exposure was made between H15 and D1, this may explained why the quantities measured on D1 for these two foxes were low. The calculated excretion half-life corresponds to the time for a fox to excrete faeces contaminated at concentrations divided by two, compared with concentrations observed the day after the last exposure. The presence of bromadiolone in faeces might be due to two processes: (1) the non-absorption by the gastrointestinal tract of the total bromadiolone ingested and its direct excretion via faeces and (2) a biliary excretion via faeces of the parent molecule after elimination from the liver (Lipha, 1987). To our knowledge, no data are available concerning the intestinal transit duration for foxes fed with rodents, but rodent hairs were present in faeces until 5–7 days after returning to dry food. So, we might suppose that bromadiolone in faeces could initially be due to both processes and then increasingly be due only to biliary excretion which would account for the detection of bromadiolone in faeces after exposure had ceased. Unfortunately, as the quantity of bromadiolone ingested by F2 and F5.3 was unknown, and the quantity excreted by F5.1 and F5.2 might be underestimated, our results did not allow us to establish an exact relationship between the quantity that was ingested and the amount that was excreted. Bromadiolone concentration in faeces seemed to be closely related to the dietary contamination but not the ingested dose. Our study demonstrates that both concentration and quantity of bromadiolone in faeces are complementary and essential for understanding the ingested bromadiolone dose. 3.4. Toxicological data and fox health conditions 3.4.1. Kinetics of bromadiolone in plasma The control fox (F0) did not show detectable residues in plasma during the experiment. In dosed foxes, bromadiolone was absorbed from the gastrointestinal tract at a relatively rapid rate and was detectable in the plasma of all intoxicated foxes at the time the first samples were taken (15 h after intoxication; Fig. 2b). During the exposure period, assimilation processes were higher than those of biotransformation and elimination, since bromadiolone plasma concentrations continually increased to reach a peak on the day following the end of the exposure period (D2 for the fox F2 with 41.6 mg/L and D5 for the foxes F5.1, F5.2 and F5.3, with 255.3, 195.3 and 107.2 mg/L, respectively). Maximal concentrations measured in foxes F5.1, F5.2 and F5.3 were 6.1, 4.7 and 2.6 times greater than in F2, respectively. All foxes were offered the same bromadiolone quantity during the two first days of the experiment and should have exhibited similar plasma concentrations. However, the very small consumption of contaminated voles by F2, and the moderate consumption by F5.3 modified the real exposure and might explain why the F2 plasma concentrations increased more slowly than did those of the others, and why those of F5.3 both decreased and increased during the exposure period. According to the literature, bromadiolone residues were detectable in rat plasma 1 h after oral administration in solution by stomach tube, and maximum levels were attained 6–9 h after ingestion (Nahas, 1986). The rapid assimilation of bromadiolone may explain why, in our study, it was detectable from the first sampling and why maximum residues were observed the day following the last exposure. After this day, biotransformation and elimination were higher than accumulation processes. As was previously reported in rats (Nahas, 1986), we showed that bromadiolone elimination exhibited a biphasic-exponential pattern and was no longer detectable in plasma from 7 to 24 days after the last dosing for the fox F2 and the three foxes F5, respectively. Our calculated half-lives ranged between 12.7 and M. Sage et al. / Environmental Research 110 (2010) 664–674 669 300 6000 200 F2 F2 4000 100 2000 0 0 0 5 10 15 20 25 30 0 5 10 15 20 25 30 300 6000 F5.1 200 F5.1 4000 100 2000 0 0 0 5 10 15 20 25 0 30 5 10 15 20 25 30 300 6000 200 F5.2 F5.2 4000 100 2000 0 0 0 5 10 15 20 25 0 30 5 10 15 20 25 30 300 6000 200 F5.3 F5.3 4000 100 2000 0 0 0 5 10 15 20 25 0 30 5 10 15 20 25 30 300 6000 F0 F0 200 4000 100 2000 0 0 0 5 10 15 20 25 30 0 5 10 15 20 25 30 Time (days) Fig. 2. Graphic representing measures on foxes feeding on five water voles spiked with 200 mg of bromadiolone/vole during two (F2) or 5 days (F5.1, F5.2, F5.3) and control (F0). Each point is the measure for each fox at one time for bromadiolone concentration in faeces (a) and bromadiolone concentration in plasma (b). Dotted lines are the expected curves as fitted by the computer with the one-compartment model derived from Widianarko and Van Straalen (1996). 72.2 h (Table 2); however, nothing was done to correct for possible hypovolaemia as a result of blood sampling and loss with haemorrhages. Our results could be improved with a larger sampling frequency (e.g., more than three blood samplings within the half-life duration (Toutain and Bousquet-Mélou, 2004)). Our study provides baseline values for a more accurate estimation of the slope of the concentration decrease curve. Our study did not confirm results of Lipha (1987), who demonstrated an increase of half-life with the ingested dose for rats. While F2 and F5.3 ingested a smaller bromadiolone quantity than F5.1 and F5.2, the 670 M. Sage et al. / Environmental Research 110 (2010) 664–674 Table 2 Estimates of kinetic parameters for bromadiolone elimination kinetics in plasma and faeces for each fox. Variance function k (d  1) min/max p-Value Half-life (h) r2 Plasma F2 F5.1 F5.2 F5.3 / varExp varExp varExp 1.30a 0.86b 0.91b 0.23c 1.236/1.375 0.766/0.964 0.701/1.232 0.167/0.307 o 0.001 o 0.001 o 0.001 o 0.001 12.7 19.4 18.2 72.2 0.99 0.99 0.97 0.97 Faeces F2 F5.1 F5.2 F5.3 / / / / 0.72a 0.38b 0.38b 0.38b 0.450/1.294 0.248/0.608 0.275/0.548 0.249/0.601 o 0.001 o 0.001 o 0.001 o 0.001 23.3 43.6 43.4 43.9 0.85 0.84 0.90 0.89 Fox Minimum and maximum values are likelihood-based 95% confidence intervals. For one parameter, within the same k0, values that share similar letters are not significantly different. fox F2 exhibited a first elimination phase more quickly than did the other three. On the other hand, the fox F5.3 exhibited a first bromadiolone elimination phase slower than F5.1 and F5.2 (Table 2). Nevertheless, during the second elimination phase, after reaching pseudo-equilibrium, bromadiolone residues in plasma of the most severely intoxicated foxes persisted longer than in the others (Fig. 2b). In accordance with Toutain and Bousquet-Mélou (2004), we confirmed that persistence of a drug in a body is influenced by pharmacokinetic parameters specific to molecule and species but is also increased with the ingested dose and is especially relevant to multiple dosing regimens. Our calculated bromadiolone plasma half-lives were within in the reference range for other species (Erickson and Urban, 2002), but our data confirmed their variability across mammals (e.g., from 26 to 58 h in rat plasma (Nahas, 1986) and around 20 h in cattle (Puyt et al., 2000)). Unfortunately, other literature data described the most common anticoagulant kinetics in plasma for poisoning diagnosis. Because the date of the intoxication remained unknown in these cases, it was impossible to compare our results with those data on different species for bromadiolone (Binev et al., 2005; Jin et al., 2007; Robben et al., 1998) or other anticoagulants (Hollinger and Pastoor, 1993). 3.4.2. Reference values of blood-clotting test The APTT and the kAPTT/KCT of the 10 control foxes have shown very high variability between individuals, ranging from 53.1 to 205.6 s (mean¼138.1 756.8) and from 12.8 to 210.4 s (mean¼126.4 776.4), respectively. With PT, these two tests were commonly used to confirm anticoagulant poisoning of mammals, for instance of rats (Batten and Bratt, 1990) or dogs (Mount, 1988). However, in our study, the high variability of these two tests prevented exploitable results. This high variability had never been reported in other species (e.g., dogs, Mount, 1988). Kit reagents might not be adapted for foxes and should be tested in a further experiment. On the other hand, we showed a notable stability of the PT (mean¼8.4 71.9) for the ten control foxes. Although they had not been exposed to anticoagulants before, one of them had a PT of 13.7 s (measures repeated during 3 days for this fox), a value 63% higher than the mean of the reference range. This outlier indicates a need for caution in interpreting isolated values measured in the field without reference values for each individual. Numerous other coagulation disorders, such as disseminated intravascular coagulopathy, hepatic disease, and hereditary coagulopathies, may influence the PT assay (Mount et al., 2003). Despite this fact, PT is the first blood-clotting time to be prolonged in anticoagulant rodenticide toxicosis and is considered the most sensitive among the 3 tested here (Woody et al., 1992). Therefore, we only measured PT on foxes of the experiment. Before dosing, the F0, F2, F5.1, F5.2 and F5.3 exhibited a PT of 7.8, 7.0, 6.8, 6.9 and 8.3 s, respectively. Beklova et al. (2007) reported a PT value of 7.7–9.2 s from one 6-month-old healthy fox. Our measured PT on controls and initial baseline values of intoxicated foxes allows us to propose reference data of 8.0 s on average (median¼7.7; 95% CI 7.4–8.9) for several individuals (n¼ 15). Those values include the data of Beklova et al. (2007) and fall within the reference range from seven to nine seconds reported for dogs (The Merck Veterinary Manual, 2005). 3.4.3. Relationships between PT kinetics and intoxication clinical signs The results clearly show an increase in clotting time in foxes that had received contaminated voles compared with pretreatment baseline values and values of the 10 uncontaminated foxes. The control fox (F0) PT remained unchanged (ranged from 6.7 to 7.4 s) (Fig. 3b). Bromadiolone effects were evident (PT425% greater than the pre-exposure value) on D3 for F2 and F5.3 and appeared more quickly (within 24 h of feeding contaminated voles) for F5.1 and F5.2 (PT+ 26 and 37%, respectively, on D1). After reaching a maximum value of 11.7 s on D3 and 20.8 s on D6 for F2 and F5.3, respectively, the PT of these two foxes decreased gradually within the reference range on D9–10 and D13, respectively. Concerning F5.1 and F5.2, the gradual prolongation of PT during the three first days was multiplied by (on average) 8 and reached 50.0 and 57.2 s on D6, respectively. This pattern is in accordance with the vitamin-K-dependent coagulation factor decrease described by Mount and Kass (1989). As was also described by Lorgue et al. (1985) and Kemikalieinspektionen (2006), severe signs related to bleeding were observed 5–6 days after the first bromadiolone administration in the two most intoxicated foxes, F5.1 and F5.2. These included depression, lethargy, weakness, mucous membrane pallor, melena and haematoma developed over the venipuncture site. Clinical signs became more intense on D6, especially for F5.2, with acute respiratory distress, profuse bleeding from the penis sheath and incessant bleeding of a venipuncture site. Therefore, F5.1 and F5.2 were immediately treated with vitamin-K1 on D6. Then, an obvious decrease in the PT and, as reported by Mount and Feldman (1982), a return to the normal reference range were observed within 24 h. Twelve hours after antidote administration, foxes were normal in both clinical signs and behaviour. As described for other species (Kemikalieinspektionen, 2006; Lorgue et al., 1985), we can assume that F5.1 and F5.2 would probably have died 1 or 2 days later without antidote injection. SGARs such as bromadiolone have a ‘‘long-acting’’ anticoagulation effect in animals (Erickson and Urban, 2002), and bromadiolone was still present in the plasma and liver of the poisoned foxes during at least 23 days after the last intoxication (see discussion below). However, contrary to what is usually reported (Mount, 1988), no PT increase was observed after the interruption of treatment on D7. Prothrombin time remained within the reference range for the duration of the study without additional antidote administration. In accordance with Robben et al. (1998), the consumed dose was low, and concentrations were probably lower than toxic values at that time. Following the autopsy, the foxes F2 and F5.3, which did not exhibit external signs of poisoning, had evidence of a moderate bleeding diathesis consistent with the pharmacological action of the active substance (Mount 1988; i.e., internal haemorrhages in abdominal fat and ecchymosed into the thoracic cavity). The foxes F5.1 and F5.2 exhibited more severe signs of hemopericardium, M. Sage et al. / Environmental Research 110 (2010) 664–674 80 671 60 60 F2 F2 40 40 20 20 0 0 5 10 15 20 25 30 0 0 5 10 15 20 25 30 60 80 60 F5.1 F5.1 40 40 20 20 0 0 0 5 10 15 20 25 0 30 5 10 15 20 25 30 60 80 60 F5.2 F5.2 40 40 20 20 0 0 0 5 10 15 20 25 30 0 5 10 15 20 25 30 60 80 60 40 F5.3 F5.3 40 20 20 0 0 0 5 10 15 20 25 0 30 5 10 15 20 25 30 60 80 60 F0 F0 40 40 20 20 0 0 5 10 15 20 25 0 0 30 Time (days) 5 10 15 20 25 30 Fig. 3. Graphic representing measures on foxes feeding on five water voles spiked with 200 mg of bromadiolone/vole during two (F2) or 5 days (F5.1, F5.2, F5.3) and control (F0). Each point is the measure for each fox at one time for bromadiolone quantity in faeces (a), and values for each fox represent the mean of two measurements of prothrombin time (b). Lines connect points only to facilitate reading. Vitamin-K1 administration for F5.1 an F5.2 is represented by the two arrows (b). hemomediastinum, mesenteric myocardial and pulmonary haemorrhages, as well as haemorrhages into the pleural cavity and subcutaneous tissues of the abdominal neck. 3.4.4. Bromadiolone concentrations in liver Bromadiolone analyses of the liver demonstrated that large amounts of the molecule were retained in this tissue at least 672 M. Sage et al. / Environmental Research 110 (2010) 664–674 24–26 days following exposure; no bromadiolone residues were observed in the control fox F0. While residues were no longer detectable in plasma, the liver exhibited residues in proportion with the acuteness of clinical signs observed. However, concentrations were all within a small range (i.e., 2.00, 2.54, 2.11 and 2.04 mg/kg in F2, F5.1, F5.2 and F5.3, respectively). Liver residues of our study may be compared to those reported in studies based on carcass collections from wildlife disease surveillance networks. Berny et al. (1997) confirmed poisoning of field foxes by a liver threshold of 0.2 mg/kg (corresponding to their routine LoD) and/or lesions consistent with anticoagulant poisoning. These parameters were also accepted for other carnivorous species by other authors (Barnett et al., 2006; Erickson and Urban, 2002; Fournier-Chambrillon et al., 2004). However, although a threshold value is a useful guide, to determine a ‘‘significant toxicological’’ liver concentration is a challenging problem that is affected by many variables, and the different studies demonstrate the high variability of the data. Berny et al. (1997) described liver concentrations ranging from 0.8 to 6.9 (median¼1.5 mg/kg) in confirmed bromadiolonepoisoned foxes. In the study of Beklova et al. (2007), a fox fed for 5 days with 5 pheasants previously exposed to bromadiolone baits exhibited a liver concentration of 0.198 mg/kg after death. However, no information was provided on the duration between exposure and death. This situation is complicated by the long time persistence of anticoagulants in the liver. For instance, the halflife of the second decline phase in rats is 318 days (Lipha, Unpublished; USEPA 1998). Also, many animals may have detectable residues of anticoagulants in their liver, but another cause of death (e.g., trauma, infectious disease) has been attributed in these cases (Fournier-Chambrillon et al., 2004; Shore et al., 2003a). The relationship between the dose of an anticoagulant ingested and the level retained in the liver remains complex and is poorly understood, exhibiting a large interindividual variability (Hegdal and Colvin, 1988). Our study brings original information concerning bromadiolone residue in fox liver that can be related to animal death. For instance, the two captive foxes with liver concentrations of 2.00 and 2.04 mg/kg 24 and 26 days after exposure, respectively, exhibited signs of only moderate internal haemorrhages. Others foxes with liver residues of 1.06 and 1.27 times higher than the value of the foxes mentioned above probably would have died without an antidote injection. Thus, measured concentration in livers in the field is difficult to interpret. In addition to an appropriate LoD for residue analysis, an important piece of evidence in determining poisoning is postmortem information with the degree of haemorrhages. 3.4.5. Implication of toxicity data on risk assessment To our knowledge, this is the first time that the PT kinetics after repeated ingestion of an anticoagulant has been investigated in a wild carnivore. The susceptibility to bromadiolone varies considerably among mammal species (USEPA, 1998). The present data showed that foxes are apparently more sensitive to bromadiolone than dogs. Prothrombin times twice as long (122 s) as the maximal values observed when our foxes exhibited the most severe symptoms were already observed in a dog before antidote treatment and recovery (Binev et al., 2005). Moreover, the dose inducing lethal symptoms on all our foxes that ingested the total administrated dose is listed as the smallest lethal dose for dogs in the literature and is 3.3 times less than 500 mg/kg/day, the LD50 after 5 days of exposure (Petterino and Paolo, 2001). Moreover, our study demonstrated that, after bromadiolone administration, the drug was not entirely eliminated at the time of the second administration, and plasma concentrations increased progressively with the repetition of dosing at the same dose rate, according to the superposition principle (Toutain and Bousquet-Mélou, 2004). These residual tissue concentrations in liver and in plasma to may have consequences for predators and lead accumulation following repeated sublethal exposure. To date, laboratory experiments (i.e., Grolleau et al., 1989) have been in contradiction with massive secondary poisoning hazards recorded in the field (Berny, 2007; Berny et al., 1997; SAGIR, 1990–2007). Our study realistically demonstrated that a captive fox feeding on contaminated voles for 5 days (a likely consumption under natural conditions) may die. Moreover, wild foxes are probably more sensitive than our captive foxes which were housed individually, thereby reducing movements and development of haemorrhages (Petterino and Paolo, 2001; Woody et al., 1992). 3.5. Relationship between bromadiolone residues in faeces, toxicological data and clinical signs 3.5.1. Detection of a recent and a previous exposure Analysis of bromadiolone residues in plasma and liver are probably the best way to detect a recent and a previous exposure, but it requires the catching or the death of animals (Erickson and Urban, 2002). Concerning non-invasive markers, the response time of the coagulation system and PT assay presented a snapshot of the exposure before 24–72 h. On the other hand, bromadiolone analyses in plasma and faeces was sensitive enough to detect a recent exposure within 15 h, whatever the quantity ingested by the fox. The assessment of fox exposure to moderate intoxication (e.g., F2) by bromadiolone analyses in faeces was possible during a period at least 1.6 and 2.6 times longer than by plasma residues or PT measurement, respectively. For foxes exposed over 5 days, bromadiolone analysis in faeces was also more sensitive, since residues were well above the LoD on the last day of exposure, while no residues were detectable in plasma. Therefore, unlike with other non-invasive direct-markers tested, analyses of faeces might determine if a fox has been exposed to bromadiolone from 15 h until at least 26 days after feeding on contaminated voles. 3.5.2. Measure sensitivity in relation to fox exposure Liver bromadiolone concentrations at the end of the experiment did not reflect obvious differences in exposure levels between foxes. Prothrombin time is considered to be a sensitive non-invasive marker of effect for vitamin-K antagonism (Erickson and Urban, 2002; Mount et al., 2003), and it closely reflected fox health conditions in relation to their exposure. However, high values of one control make us cautious concerning isolated PT values. Plasma concentration provide a picture of the actual intensity of exposure but we have shown that it is not sensitive enough to demonstrate the occurrence of exposures in the past. Both bromadiolone concentration and quantity measured in faeces are complementary, are related to the dose ingested 1 or 2 days before, and may relieves the effect of intestinal transit disturbance and natural inter-individual variations or the possible effects of parent molecule metabolisation (Markussen et al., 2008; Pelz et al., 2005). According to our feeding protocol, bromadiolone residues detected in faeces were higher than the LoD of the analytical method, and concentration data did not reach saturation point at the maximal exposure period. This indicates that lower and higher exposures than those tested here should be easily detectable. As anticoagulants are more toxic when ingested for several days (USEPA, 1998), ideally, to determine the poisoning risk for a fox, several samples taken at intervals of a few days would be better than isolated sampling. Furthermore, some interrogations remain such as the possible effects of UV radiation, M. Sage et al. / Environmental Research 110 (2010) 664–674 heat and other factors on bromadiolone persistence in faeces in the field. Until specific studies on this issue are achieved, sampling design should privilege situations which can be reasonably compared (e.g. faeces having been weathered similarly). Thus, it will be possible to detect an increasing, constant or decreasing exposure. To our knowledge, this is the first time that the detection of bromadiolone in faeces of a rodent predator has been investigated. This non-invasive technique provides an approach for monitoring both recent and previous exposure of foxes to bromadiolone that can be applied not only at the individual level in experiments but also at the population level in the field. Such monitoring could be used to compare how the severity of exposure varies over time, with location, and between baiting strategies, and so can be used to reduce the risk to non-target species form rodent control operations. Acknowledgments This study received support from the Conseils généraux of the Doubs and the Jura departments and from the Région of FrancheComté. We are grateful to F. Cliquet and M. Munier of the AFSSA and to E. Benoit of the ENVL for their hospitality, participation and technical support. References AGRITOX INRA, 2004. Fiche d’information sur les suptances actives phytopharmaceutiques—Bromadiolone/Lipha /http://www.inra.fr/aritoxS. Alterio, N., 1996. Secondary poisoning of stoats (Mustela erminea), feral ferrets (Mustela furo), and feral house cats (Felis catus) by the anticoagulant poison, brodifacoum. N. Z. J. Zool. 23, 331–338. Alterio, N., Brown, K.P., Moller, H., 1997. Secondary poisoning of mustelids in a New Zealand Northfagus forest. J. Zool. (London) 243, 863–869. Artois, M., 1989. Le Renard roux (Vulpes vulpes Linnaeus, 1758). In: Artois, M., Delattre, P. (Eds.), Encyclopédie des carnivores de France. SFEPM, Nort/Erdre, pp. 90. Barnett, E.A., Fletcher, M., Hunter, K., Taylor, M.J., Sharp, A.E., 2006. Pesticide poisoning of animals in 2006—investigations of suspected incidents in the United Kingdom. In: Department for Environment Food and Rural Affairs (Ed.), A Report of the Environmental Panel of the Advisory Committee on Pesticides 2007, Defra, London. Batten, P., Bratt, H., 1990. Brodifacoum: elimination from the tissues of rats following administration of single oral doses: Lab Project Number: UR0172: UR0211: CTL/P/1559. Unpublished study prepared by ICI Central Toxicology Laboratory, 65 pp. Beklova, M., Krizkova, S., Supalkova, V., Mikelova, R., Adam, V., Pikula, J., Kizek, R., 2007. Determination of bromadiolone in pheasants and foxes by differential pulse voltametry. Int. J. Environ. Anal. Chem. 87, 459–469. Berny, P., 2007. Pesticides and the intoxication of wild animals. J. Vet. Pharmacol. Ther. 30, 93–100. Berny, P., de-Oliveira, L., Videmann, B., Rossi, S., 2006. Assessment of ruminal degradation, oral bioavailability, and toxic effects of anticoagulant rodenticides in sheep. Am. J. Vet. Res. 67, 363–371. Berny, P.J., Buronfosse, T., Buronfosse, F., Lamarque, F., Lorgue, G., 1997. Field evidence of secondary poisoning of foxes (Vulpes vulpes) and buzzards (Buteo buteo) by bromadiolone, a 4-year survey. Chemosphere 35, 1817–1829. Binev, R., Petkov, P., Rusenov, A., 2005. Intoxication with anticoagulant rodenticide bromadiolone in a dog—a case report. Vet. Arhive 75, 273–282. Boermans, H.J., Johnstone, I., Black, W.D., Murphy, M., 1991. Clinical signs, laboratory changes and toxicokinetics of brodifacoum in the horse. Can. J. Vet. Res. 55, 11–27. Brakes, C., Smith, R., 2005. Exposure of non-target small mammals to rodenticides: short-term effects, recovery and implications for secondary poisoning. J. Appl. Ecol. 42, 118–128. Brown, K.P., Alterio, N., Moller, H., 1998. Secondary poisoning of stoats (Mustela erminea) at low mouse (Mus musculus) abundance in a New Zealand Nothofagus forest. Wildl. Res. 25, 419–426. Burnham, K.P., Anderson, D.R., 2004. Multimodel inference: understanding AIC and BIC in model selection. Sociol. Meth. Res. 33, 261–304. Eadsforth, C.V., Dutton, A.J., Harrison, E.G., 1991. A barn owl feeding study with [14C] flocoumafen-dosed mice: validation of a non-invasive methode of monitoring exposure of barn owls to anticoagulant rodenticides in their prey. Pestic. Sci. 32, 105–119. 673 Eason, C.T., Milne, L., Potts, M., Morriss, G., Wright, G.R.G., Sutherland, O.R.W., 1999. Secondary and tertiary poisoning risks associated with brodifacoum. N. Z. J. Ecol. 23, 219–224. Eason, C.T., Murphy, E.C., Wright, G.R.G., Spurr, E.B., 2002. Assessment of risks of brodifacum to non-target birds and mammals in New Zealand. Ecotoxicology 11, 35–48. Eason, C.T., Spurr, E.B., 1995. Review of the toxicity and impacts of brodifacoum on non-target wildlife in New Zealand. N. Z. J. Zool. 22, 371–379. Erickson, W., Urban, D., 2002. Potential Risks of Nine Rodenticides to Birds and Nontarget Mammals: a Comparative Approach, vol. 1. United States Environmental Protection Agency, Washington, DC 198 pp. Fournier-Chambrillon, C., Berny, P., Coiffier, O., Barbedienne, P., Dassé, B., Delas, G., Galineau, H., et al., 2004. Evidence of secondary poisoning of free-ranging riparian mustelids by anticoagulant rodenticides in France: implication for conservation of European mink (Mustela lutreola). J. Wildl. Dis. 40, 688–695. Giraudoux, P., Tremolie res, C., Barbier, B., Defaut, R., Rieffel, D., Bernard, N., Lucot, E., et al., 2006. Persistence of bromadiolone anticoagulant rodenticide in Arvicola terrestris populations after field control. Environ. Res. 102, 291–298. Gray, A., Eadsforth, C.V., Dutton, A.J., 1994. Non-invasive methode for monitoring the exposure of barn owls to second-generation rodenticides. Pestic. Sci. 41, 339–343. Grobosch, T., Angelow, B., Schonberg, L., Lampe, D., 2006. Acute bromadiolone intoxication. J. Anal. Toxicol. 30, 281–286. Grolleau, G., Lorgue, G., Nahas, K., 1989. Toxicité secondaire, en laboratoire, d’un rodenticide anticoagulant (bromadiolone) pour des prédateurs de rongeurs champe tres: Buse variable (Buteo buteo) et Hermine (Mustela erminea). Bull. OEPP/EPPO 19, 633–648. Hegdal, P.L., Colvin, B.A., 1988. Potential hazard to eastern screech-owls and other raptors of brodifacoum bait used for vole control in orchards. Environ. Toxicol. Chem. 7, 245–260. Hollinger, B.R., Pastoor, T.P., 1993. Case management and plasma half-life in a case of brodifacoum poisoning. Arch. Intern. Med. 153, 1925–1928. Howald, G.R., Mineau, P., Elliott, J.E., Cheng, K.M., 1999. Brodifacoum poisoning of avian scavengers during rat control on a seabird colony. Ecotoxicology 8, 431–447. Hunter, K., 1983a. Determination of coumarin anticoagulant rodenticide residues in animal tissue by high-performance liquid chromatography. I. Fluorescence detection using post-column techniques. J. Chromatogr. 270, 267–276. Hunter, K., 1983b. Determination of coumarin anticoagulant rodenticide residues in animal tissue by hight-performance liquid chromatography. II. Fluorescence detection using ion-pair chromatography. J. Chromatogr. 270, 277–283. Jin, M.C., Ren, Y.P., Xu, X.M., Chen, X.H., 2007. Determination of bromadiolone in whole blood by high-performance liquid chromatography coupled with electrospray ionization tandem mass spectrometry. Forensic Sci. Int. 171, 52–56. Johnstone, I.B., 1988. Clinical and laboratory diagnosis of bleeding disorders. In: Feldman, BF. (Ed.), The Veterinary Clinics of North America: Small Animal Practice. WB Saunders, Philadelphia, pp. 21–33. Kemikalieinspektionen, S.C.I., 2006. Competent Authority Report According to Directive 98/8/EC Bromadiolone Cas 28772-56-7; Sweden. Kolf-Clauw, M., Alvarez, E., Matray, O., 1995. Anticoagulant rodenticides: étiologie, diagnostic et traitement des intoxications. Recueil de Medecine Vétérinaire. Spécial Toxicologie des Carnivores Domestiques, 127–134. Kupper, J., Grobosch, T., Kistler, R., Sydler, T., Naegeli, H., 2006. Bromadiolone poisoning in foxes. Schweizer Archiv Fur Tierheilkunde 148, 405–408. Lambert, O., Pouliquen, H., Larhantec, M., Thorin, C., L’Hostis, M., 2007. Exposure of raptors and waterbirds to anticoagulant rodenticides (Difenacoum, bromadiolone, coumatetralyl, coumafen, brodifacoum): epidemiological survey in loire atlantique (France). Bull. Environ. Contam. Toxicol. 79, 91–94. Lipha., 1987. Kinetics of bromadiolone, anticoagulant rodenticide, in the Norway rat (Rattus norvegicus). ENVL–INRA, France. The Italian Pharmacological Society. Pharmacol. Res. Commun. 19 (11), 767–775. Lipha., 1991. Determination of the residues and the half-life of the rodenticides brodifacoum, bromadiolone and flocoumafen in the livers of rats during 200 days after a single oral dose of each at a dose level of 0.2 mg/kg. Non-public. Laboratory Report No. LPA 158/891590 GLP/Unpublished. Lorgue, G., Mally, C., Nahas, K., 1985. Traitement de l’intoxication par la bromadiolone, rodonticide anticoagulant, chez le chien. Recueil de Medecine Vétérinaire 161, 329–334. Marek, L.J., Koskinen, W.C., 2007. Multiresidue analysis of seven anticoagulant rodenticides by high-performance liquid chromatography/electrospray/mass spectrometry. J. Agric. Food Chem. 55, 571–576. Markussen, M.DK., Heiberg, A.C., Fredholm, M., Kristensen, M., 2008. Differential expression of cytochrome P450 genes between bromadiolone-resistant and anticoagulant-susceptible Norway rats: a possible role for pharmacokinetics in bromadiolone resistance. Pest. Manage. Sci. 269, 239–248. McDonald, R.A., Harris, S., Turnbull, G., Brown, P., Fletcher, M., 1998. Anticoagulant rodenticides in stoats (Mustela erminea) and weasels (Mustela nivalis) in England. Environ. Pollut. 103, 17–23. Morin, M.F., Merlet, N., Dore, M., Lechevin, J.C., 1989. Analyse de résidus de bromadiolone (rodonticide anticoagulant) dans l’eau et les tissus animaux. technique d’extraction et de purification. Analusis 17 (9), 526–531. 674 M. Sage et al. / Environmental Research 110 (2010) 664–674 Mitchell-Jones, A.J.J., Amori, G., Bogdanowicz, W., Krystufek, B., Reijnders, P., Spitzenberger, P., Stubbe, M., Thissen, J.B.M., Vohralik, V., Zima, J., 1999pp. The Atlas of European Mammals. Poyser, T. & A.D. Limited for the Societas Europaea Mammalogica, ISBN: 9780856611308 484 pp. Ministe re de l’Agriculture et de la Pêche, 2002. Arrêté ministériel du 17 décembre 2001 relatif aux conditions d’emploi de la bromadiolone pour la lutte contre le Campagnol terrestre. J.O. no. 25 du 30 janvier 2002, p. 2015. Mount, M.E., 1988. Diagnosis and therapy of anticoagulant rodenticide intoxicants. Vet. Clin. N. Am. Small Anim. Pract. 18, 115–130. Mount, M.E., Feldman, B.F., 1982. Vitamin K and its therapeutic importance. J. Am. Vet. Med. Assoc. 11, 1354–1356. Mount, M.E., Kass, P.H., 1989. Diagnostic importance of vitamin K1 and its epoxide measured in serum of dogs exposed to an anticoagulant rodenticide. Am. J. Vet. Res. 50, 1704–1709. Mount, M.E., Kim, B.U., Kass, P.H., 2003. Use of a test for proteins induced by vitamin K absence or antagonism in diagnosis of anticoagulant poisoning in dogs: 325 cases (1987–1997). J. Am. Vet. Med. Assoc. 222, 194–198. Murphy, E.C., Clapperton, B.K., Bradfield, P.M.F., Speed, H.J., 1998. Brodifacoum residues in target and non-target animals following large-scale poison operations in New Zealand podocarp–hardwood forests. N. Z. J. Zool. 25, 307–314. Nahas, K., 1986. Dosage de la bromadiolone (rodonticide anticoagulant) dans le plasma, le foie et le rein du rat. J. Chromatogr. 369, 445–448. Newton, I., Wyllie, I., Gray, A., Eadsforth, C.V., 1994. The toxicity of the rodenticide flocoumafen to barn owls and its elimination via pellets. Pestic. Sci. 41, 187–193. Pelz, H.J., Rost, S., Hunerberg, M., Fregin, A., Heiberg, A.C., Baert, K., MacNicoll, A.D., Prescott, C.V., Walker, A.S., Oldenburg, J., Muller, C.R.., 2005. The genetic basis of resistance to anticoagulants in rodents. Genetics 170, 1839–1847. Petterino, C., Paolo, B., 2001. Scientific reviews—toxicology of various anticoagulant rodenticides in animals. Vet. Hum. Toxicol. 46, 353–360. Pineiro, J.C., Bates, D.M., 2000. Mixed-Effects Models in S and S-Plus. New York. Puyt, J.D., Sachot, E., Larhantec-verdier, M., et al., 2000. Oral bioavailability of bromadiolone in cows. In: Proceedings of the Ninth Meeting of the European Association for Veterinary Pharmacology and Toxicology. R Development Core Team, 2004. R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria, ISBN: 3900051-00-3. Raoul, F., Michelat, D., Ordinaire, M., Decote, Y., Aubert, M., Delattre, P., Deplazes, P., et al., 2003. Echinococcus multilocularis: secondary poisoning of fox population during a vole outbreak reduces environmental contamination in a high endemicity area. Int. J. Parasitol. 33, 945–954. Robben, J.H., Kuijpers, E.A.P., Mout, H.C.A., 1998. Plasma superwarfarin levels and vitamin K1 treatment in dogs with anticoagulant rodenticide poisoning. Vet. Q. 20, 24–27. Sage, M., Coeurdassier, M., Defaut, R., Gimbert, F., Berny, P., Giraudoux, P., 2008. Kinetics of bromadiolone in rodent populations and implications for predators after field control of the water vole, Arvicola terrestris. Sci. Total Environ. 407, 211–222. SAGIR, 1990–2007. Traitement de campagnols aux anticoagulants. Evaluation de la mortalité de la faune non cible dans le Doubs. Office National de la Chasse et de la Faune Sauvage. Doubs, France. Sakamoto, Y., Ishiguro, M., Kitagawa, G., 1996. Akaike Information Criterion Statistics. D. Reidel Publishing Company. Shore, R.F., Birks, J.D.S., Afsar, A., Wienburg, C.L., Kitchener, A.C., 2003a. Spatial and temporal analysis of second-generation anticoagulant rodenticide residues in polecats (Mustela putorius) from throughout their range in Britain, 1992–1999. Environ. Pollut. 122, 183–193. Shore, R.F., Fletcher, M.R., Walker, L.A., 2003b. Agricultural pesticides and mammals in Britain. In: Tattersall, F., Manley, W. (Eds.), Conservation and Conflict: Mammals and Farming in Britain. Linnean Society Occasional Publication 4, The Linnean Society, London, pp. 37–50. Siegel, S., Castellan, N.J., 1988. Nonparametric Statistics for the Behavioral Sciences. McGraw-Hill International, New York. Sokal, P.R., Rohlf, F.J., 1997. Biometry: The Principles and Practice of Statistics in Biological Research. W.H. Freeman and Company, New York, USA. Stone, W.B., Okoniewski, J.C., Stedelin, J.R., 2003. Anticoagulant rodenticides and raptors: recent findings from New York. Bull. Environ. Contam. Toxicol. 70, 34–40. The Merck Veterinary Manual, 2005. In: Cynthia, M., Kahn, Line, Scott (Eds.), ninth ed. (February 8, 2005). Toutain, P.L., Bousquet-Mélou, A., 2004. Plasma terminal half-life. J. Vet. Pharmacol. Ther. 27, 427–439. USEPA, 1998. Registration Eligibility Decision (RED) Rodenticide Cluster. EPA 738R-98-007. Environmental Protection Agency, United States. Walker, L.A., Turk, A., Long, S.M., Wienburg, C.L., Best, J., Shore, R.F., 2008. Second generation anticoagulant rodenticides in tawny owls (Strix aluco) from Great Britain. Sci. Total Environ. 392, 93–98. Weber, J.M., Aubry, S., 1993. Predation by foxes, Vulpes vulpes, on the fossorial form of the water vole, Arvicola terrestris scherman, in western Switzerland. J. Zool. London 229, 553–559. Widianarko, B., Van Straalen, N., 1996. Toxicokinetics-based survival analysis in bioassays using nonpersistent chemicals. Environ. Toxicol. Chem. 15, 402–406. Woody, B.J., Murphy, M.J., Ray, A.C., Green, R.A., 1992. Coagulopathic effects and therapy of brodifacoum toxicosis in dogs. J. Vet. Intern. Med. 6, 23–28. Wright, G.R.G., 2002. Protocol for Tissue Sampling Testing for Vertebrate Pesticides in Animals. Landcare Research, P.O. Box 69, Lincoln, New Zealand o http:// www.landcareresearch.co.nz 4.