|Home | About | Journals | Submit | Contact Us | Français|
In forensic bioanalytical methods, there is a general agreement that calibrators should be prepared by fortifying analytes in matrix-based blank samples (matrix-based). However, in the case of vitreous humor (VH), the collection of blank samples for the validation and for routine analysis would require the availability of many cadavers. Besides the difficulty of obtaining enough blank VH, this procedure could also represent an ethical issue. Here, a study of matrix effect was performed taking into consideration human and bovine vitreous and saline solution (SS) (NaCl 0.9%). Tricyclic antidepressants [amitriptyline (AMI), nortriptyline (NTR), imipramine (IMI) and desipramine (DES)] were used as model analytes and were extracted from samples by means of liquid-phase microextraction and detected by gas chromatography–mass spectrometry. Samples of human and bovine VH and SS were prepared in six different concentrations of antidepressants (5, 40, 80, 120, 160 and 200 ng/mL) and were analyzed. Relative matrix effect was evaluated by applying a two-tailed homoscedastic Student's t-test, comparing the results obtained with the set of data obtained with human VH and bovine VH and SS. No significant matrix effect was found for AMI and NTR in the three evaluated matrices. However, a great variability was observed for IMI and DES for all matrices. Once compatibilities among the matrices were demonstrated, the method was fully validated for AMI and NTR in SS. The method was applied to six VH samples deriving from real cases whose femoral whole blood (FWB) was analyzed by a previously published method. An average ratio (VH/FWB) of ~0.1 was found for both compounds.
Vitreous humor (VH) often plays an important role in the field of forensic toxicology in cases where it is not possible to obtain blood or urine samples. Its privileged anatomic location, protected in the ocular environment, provides protection against contaminations by microorganisms. It is also less susceptible to postmortem redistribution phenomenon, and it is easily obtained by inserting a needle through the sclera. VH comprises 99% water and the remaining 1% is made up of salts, sugars, phagocytes and a network of collagen (1).
There is a general agreement that all forensic bioanalytical validation should be matrix-based (2–4). However, the lack of availability and the complexity of some biological samples used in postmortem toxicology make the analytical process a difficult task. In some circumstances, in which samples have unique features such as decomposed or embalmed specimens, the method of standard addition has been used for accurate quantitation (5, 6). In other cases, such as the analysis in VH, one of the difficulties is to obtain enough volume of blank specimen to perform validation and to prepare calibrators and quality control samples for routine analysis. The analytical strategy of some authors is to perform quantification in different specimens with a calibration curve constructed in blood after a study of matrix effect (1, 5). In fact, matrix-matched calibrators are recommended by the Society of Forensic Toxicologists and the American Academy of Forensic Sciences when analyses are performed on unusual specimens (decomposed tissue, vitreous fluid, etc.) (7).
In this work, a matrix effect study was carried out to verify the analytical compatibility among human VH, bovine VH and saline solution (0.9% NaCl). Tricyclic antidepressants [amitriptyline (AMI), nortriptyline (NTR), imipramine (IMI) and desipramine (DES)] were used as a model. Antidepressants are used worldwide for the treatment of different types of depression and other psychiatric disorders. As a result, these drugs are frequently found in emergency toxicology cases and forensic medical examinations (8, 9). Based on our previous experience in the analysis of whole blood (WB), urine and liver samples (10–13), liquid-phase microextraction (LPME) was also used here for the analysis of VH. Afterwards, the validated method was applied to human VH collected from six deceased persons previously exposed to antidepressants.
Standard solutions (1.0 mg/mL) of AMI, NTR, IMI, DES in methanol, and the respective deuterated internal standard (IS) solutions (100 µg/mL) of AMI-d3, NTR-d3, IMI-d3 and DES-d3 were purchased from Cerilliant Analytical Reference Standards (Round Rock, TX, USA). Sodium hydroxide, sodium chloride and formic acid were purchased from Merck (Darmstadt, Germany). Dodecane was purchased from Sigma-Aldrich (St. Louis, MO, USA).
Working solutions of AMI, NTR, IMI and DES, at concentrations of 100 µg/mL, and working solutions of ISs, NTR-d3, IMI-d3 and DES-d3, at concentrations of 10 µg/mL, were prepared with methanol in volumetric glassware. AMI-d3 was prepared in the same manner at a concentration of 1.0 µg/mL. The stock solutions were refrigerated at 2–8°C when not in use.
Hollow fiber Q3/2 Accurel KM polypropylene (600 µm internal diameter, 200 µm wall thickness, 0.2 µm pore size) was purchased from Membrana (Wuppertal, Germany). A multistation magnetic stirrer with a temperature control, model RT-10, was purchased from IKA® (Staufen, Germany). Gas chromatography–mass spectrometry (GC–MS) analyses for antidepressants were performed using a gas chromatograph, model GC-2010, coupled with a mass-selective detector (MSD) (GCMS-QP2010; Shimadzu, Kyoto, Japan). Chromatographic separation was achieved with a HP-5MS fused-silica capillary column (30 m × 0.25 mm i.d., 0.25 µm film thickness) using helium as carrier gas at 0.8 mL/min in a constant ﬂow rate mode. Injections were made in the splitless mode. The MSD was operated using electron ionization at 50 eV by selected ion monitoring (SIM). The temperatures of the injection port and the interface were 220 and 280°C, respectively. The oven temperature was maintained at 125°C for 1 min, increased at 50°C/min to 190°C, increased at 5°C/min to 225°C, and held for 3 min. Then the temperature was increased at 50°C/min to 230°C and held for 1 min. The following ions were chosen for specific quantification AMI-d3 at m/z 61, AMI at m/z 277, NTR-d3 at m/z 266, NTR at m/z 263, IMI-d3 at m/z 283, IMI at m/z 280, DES-d3 at m/z 269, DES at m/z 266.
Human drug-free VH samples and real cases were obtained from the Death Verification Service, University of São Paulo. Bovine VH was acquired in the slaughterhouse. Human drug-free VH samples were combined for the blank vitreous matrix. Postmortem specimens were collected in tubes containing ethylene-diamine-tetra-acetic acid disodium salt. This study was approved by the Research Ethics Committee of the Clinics Hospital at the Faculty of Medicine of the University of São Paulo, CAPPesq HC-FMUSP (ethics protocol approval no. 0352/09).
An aliquot of 0.50 mL of VH sample (or saline solution—0.9% NaCl) was transferred into a 5-mL glass tube containing a 10-mm magnetic stirrer, followed by 3.5 mL of 0.1 M NaOH solution. Deuterated ISs were added to this sample solution: 10 ng of AMI-d3; 100 ng of DES-d3, IMI-d3 and 200 ng of NTR-d3. A 8-cm hollow fiber, whose pores were filled with dodecane (organic phase), was used for each extraction. The lumen of the hollow fiber was filled with 30 µL of 0.1 M formic acid (acceptor phase) using a micropipette, and the filled hollow fiber was placed into the sample solution in a U-shaped configuration. During extraction for 10 min, the solution was stirred at 1,000 rpm at 55°C. After extraction, the acceptor phase was withdrawn from the fiber using a thin tip and dried under a nitrogen stream at 40°C. The residue was suspended in 30 µL of methanol. No derivatization step was required. Subsequently, 2.0 µL of the solution was injected into the GC–MS system. Each fiber was used only one time and after the extraction it was discarded.
The interfering compounds present in each matrix, either identified or unidentified, can affect the analyte measurement, leading to different results from the analysis of samples with the same analyte concentration, but in different matrices (14). This test was performed by preparing three sets of samples (n = 6), comprising of human VH, bovine VH and saline solution. The samples had the same concentration of analytes (5, 40, 80, 120, 160 and 200 ng/mL). Data were analyzed by the application of the two-tailed homoscedastic Student's t-test. If the calculated t-value is below the value that is expected in the t-test tables, it is possible that the analytes do not suffer a different interference from the matrices and vice versa. Calibration curves were also constructed with all the matrices for comparison. Microsoft Excel® was used for all statistical analyses and preparation of x–y scatter plots.
In the study of matrix effect, a great variability was observed for IMI and DES for all matrices (human and bovine VH and saline solution). No significant matrix effect was found for AMI and NTR in the three evaluated matrices. Once compatibilities among the matrices were demonstrated, the method was optimized and validated for AMI and NTR in saline solution. The parameters were evaluated independently, and only one parameter was varied at a time. In this case, all of the parameters were fixed, with the exception of the one that was varied, until the best result was obtained. The following parameters were taken into consideration: the velocity of magnetic stirring (250, 500, 800, 1,000 or 1,200 rpm), the extraction time (1, 5, 10, 20, 40 or 60 min) and the extraction temperature (25, 35, 45, 55, 65 or 75°C). Fortified saline solution samples at a concentration of 100 ng/mL of AMI and NTR were subjected to the previously described procedure. The efficiency of the extraction for each compound was evaluated by measuring the absolute average chromatographic peak areas produced by the analytes in triplicate.
The method was validated taking into consideration the limits of detection (LODs), limits of quantification (LOQs), the calibration equations, the intraday and interday precision and accuracy and dilution integrity using international guidelines and recommendations in the field of forensic toxicology (2–4).
The limit of detection (LOD) was defined as the concentration giving a signal-to-noise ratio of >3. The LOQ was defined as the lowest concentration of a sample that can still be quantified with acceptable precision and accuracy; the acceptance criteria for these two parameters were 15% for precision (relative standard deviation: RSD) and 15% for accuracy bias.
The linearity study was performed by analyzing aliquots of saline solution containing AMI and NTR in six replicates at the following concentrations for each analyte: 5, 40, 80, 120, 160 and 200 ng/mL.
The precision and accuracy were evaluated by the analysis of saline solution containing concentrations of 15, 100 and 180 ng/mL for all analytes on three different consecutive days. The analyses were performed on six replicates for each day. Precision, defined as RSD, was determined by intraday and interday repetitions. Accuracy was expressed as a percentage of the known concentration, that is, the mean measured concentration/nominal concentration × 100, or percent bias.
The extraction efficiency (recovery) was determined as the percentage of the remaining amount of the antidepressant in the sample at the end of the extraction, compared with the set of samples not extracted by the LPME procedure. The unprocessed represented 100% recovery. Three concentrations of antidepressants denominated low, medium and high in the respective concentrations at 15, 100 and 180 ng/mL, with six replicates at each concentration, were assayed.
Dilution integrity is a parameter that allows evaluating samples with concentrations of analytes above the calibration curve. Three high concentrations of antidepressants (600, 1,500 and 2,100 ng/mL) were prepared in saline solution and diluted at the following ratios: 1:5; 1:15 and 1:42, respectively. Analyses were carried out in triplicate for each diluted sample. Accuracy and precision were required to be within the set criteria (within ±15%) (4). Stability of the analytes in prepared samples at a concentration of 150 ng/mL was evaluated in the autosampler (22°C/22 h).
To prove the applicability of this analytical method, real cases that involved antidepressant consumption were analyzed. The six cases corresponded to patients who had used antidepressants therapeutically for psychiatric disorders and died in the hospital. Rapid immunoassay screening for drugs in the urine (TriageTM, Biosite) was performed during their autopsies, and all the subjects tested positive for tricyclic antidepressants. No other details concerning the history of the cases were accessed. The quantification was based on the ratios of the ion peak areas of the compounds to the IS ion peak areas. The calibration curves performed in saline solutions were used to determine the antidepressants (AMI and NTR) concentrations in human VH specimens.
Matrix-matched calibrators are recommended by forensic toxicologists when dealing with the analysis of biological specimens such as VH (2–4). Nevertheless, due to the difficulty of obtaining enough human VH for routine analysis, the question then arose is to whether there is a significant matrix effect when quantification of vitreous fluid is performed using calibrators prepared in animal VH or even in saline solution (NaCl 0.9%). Although matrix effect studies are more common in LC–MS methods, this effect can also be observed in other techniques, such as GC–MS (1, 14). Therefore, a comparison was made among the measurements of antidepressants performed in different matrices (human and bovine VH and saline solution). Certainly, animal vitreous and especially saline solution are more available materials than human VH.
With the application of Student's t-test in a wide range of analyte concentration, a great variability of measurements was observed when data obtained with the analysis of IMI and DES fortified in human VH samples were compared with those obtained with bovine VH and saline solution. Calibration curves were built with the matrices and are shown in Figure 1. As it can see in this figure, calibration curves of IMI and DES constructed in bovine VH and saline solution are quite similar but differs from the curve built in human vitreous. It is possible that constituents present in human VH interfere with the analysis and decrease the response of the IMI and DES, since the slope of this curve is lower than the other curves.
In contrast, no significant matrix effect was found for the analytes AMI and NTR in the three evaluated matrices by the application of Student's t-test (P < 0.05). In fact, calibration curves constructed in these matrices were quite similar, as shown in Figure 1. Therefore, measuring human vitreous specimens through calibration performed in bovine VH and saline solution is acceptable. Figure 1 presents the results and it is possible to observe the matrix effect (ME) in the analysis in which IMI and DES were fortified in the human, bovine VH and saline solution. Coefficients of variation (CV) of slopes greater than 4% mean the presence of ME (15). On other hand, there was no influence of this phenomenon in the replicates containing AMI and NTR. Once compatibilities among the matrices were demonstrated, the LPME/GC–MS method was fully validated for AMI and NTR in saline solution. The present LPME/GC–MS method was based on our previously published article, in which antidepressants were determined in WB samples (10). However, the new method was afresh optimized since the matrices are different. In fact, compared with the conditions set for WB, the results obtained with VH demonstrated milder conditions, except for the temperature, which was 55°C. The velocity of magnetic stirring was reduced by ~17% (1,000 rpm). The more significant change was the reduction of time for extraction: 10 min, which corresponded to 67% faster. The result may be related to the difference in viscosity between the two matrices. Contrary to the data obtained with WB, a higher extraction time resulted in a reduction in the yield of extraction of antidepressants from the bovine biological matrix. In addition, as reported by other authors, here we also observed that higher stirring speeds caused the appearance of bubbles on the outer wall of polypropylene membrane of the hollow fiber and affected negatively the extraction (16, 17).
The validation parameters of the method (LOD, LOQ, intraday and interday precision, accuracy, recovery and dilution integrity) obtained for the determination of the antidepressants are shown in Table I. After the optimization of the extraction procedure, the recovery rates for AMI and NTR ranged from 62.0 to 97.6%. The stability study showed that extracted analytes were stable in the autosampler at room temperature during 24 h.
Based on the articles that performed the measurements of antidepressants in WB and HV in real cases, the vitreous humor/femoral whole blood (VH/FWB) ratio is ~0.1 (18). Therefore, a method to detect antidepressants in VH should be more sensitive than a method to detect these substances in WB. With the aim to increase sensitivity, without the need for derivatization reactions, the voltage of the ion source of the MS was changed to 50 eV. Lower ionization energy favors the produce of molecular and higher ions, less fragmentation process and consequently increases the sensitivity, especially when the spectrometric method monitors and quantifies ions with such characteristics (19). To test this hypothesis, the analytes (50 ng) were injected ten (10) times in both voltages (50–70 eV). In fact, the decrease of the ion source voltage to 50 eV represented an increase in response for both analytes (37% for AMI and 87% for NTR), calculated by absolute area. However, this change of the voltage also represented a decrease in the working range of the calibration curve. Concentrations exceeding 300 ng/mL were not able to be determined in the GC–MS due to saturation of the detector. For this reason, the dilution integrity test was also performed, taking into consideration samples with high concentration of antidepressants (600, 1,500 and 2,100 ng/mL). After appropriate dilution of these samples, the accuracy values were satisfactory (93.3–110%).
The developed method to detect AMI and NTR in VH was applied in six real cases. FWB also were collected from these cases and were measured as the analytical method published previously (10). The results are shown in Table II, and the chromatograms obtained by analysis with LPME and GC–MS for the new method and an actual postmortem are showed in Figure 2. The average ratios (VH/FWB) of antidepressants AMI and NTR found in this study were, respectively, 0.1 and 0.09. In general, substances cross the ocular membrane through passive diffusion from blood. Therefore, only parent drugs or their metabolites that are not bound to plasma proteins are able to penetrate this barrier (20, 21). Consequently, the concentrations of drugs with high affinity for plasma proteins will be diminished in VH. The tricyclic antidepressants are highly bound to plasma proteins. Particularly, the percentage of plasma protein binding of AMI and NTR are 91–97% and 90–95%, respectively (22). This phenomenon can explain the low HV/FWB ratios found in this study.
The matrix effect should be considered in forensic toxicology, even in methods based on GC–MS. Even analytes with similar chemical structures may exhibit different behaviors. Each practitioner laboratory should verify matrix effect based upon his own specific method. The HF-LPME method developed for the determination of AMI and NTR in VH proved to be appropriate for the analysis of real postmortem cases. An average ratio (VH/FWB) of ~0.1 was found for both compounds.
There are no financial or other relations that could lead to a conﬂict of interest.
This study was supported by the Faculty of Medicine, University of São Paulo (LIM 40/HC-FMUSP). Financial support provided by the Fundação de Amparo à Pesquisa do Estado de São Paulo (FAPESP Grant No. 2009/08314-9) and the Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq Grant No. 470643/2009-9) is also gratefully acknowledged. Marcelo Filonzi dos Santos is a fellow of the FAPESP (Grant No. 2010/06530-3). Adrian Yamada is a fellow of the Programa Institucional de Bolsas de Iniciação Científica-CNPq.