Clinical Chemistry AACC Online Job Center
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
QUICK SEARCH:   [advanced]


     

Clinical Chemistry 45: 1224-1234, 1999;
This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF)
Right arrow Submit an electronic Letter to
the Editor about this paper
Right arrow Alert me when this article is cited
Right arrow Alert me when eLetters are posted
Right arrow Alert me if a correction is posted
Right arrow Citation Map
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in ISI Web of Science
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via ISI Web of Science (35)
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Fitzgerald, R. L.
Right arrow Articles by Herold, D. A.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Fitzgerald, R. L.
Right arrow Articles by Herold, D. A.
Related Collections
Right arrow Drug Monitoring and Toxicology
Right arrow Automation and Analytical Techniques
(Clinical Chemistry. 1999;45:1224-1234.)
© 1999 American Association for Clinical Chemistry, Inc.

Articles

Broad Spectrum Drug Identification Directly from Urine, Using Liquid Chromatography-Tandem Mass Spectrometry

Robert L. Fitzgerald1,2,a, Jeffrey D. Rivera1 and David A. Herold1,2

1 Veterans Administration Medical Center San Diego and
2 University of California San Diego, 3350 La Jolla Village Dr., San Diego, CA 92161.
a Address correspondence to this author at: VAMC-113, 3350 La Jolla Village Dr., San Diego, CA 92161. Fax 619-552-7479; e-mail rlfitzgerald{at}vapop.ucsd.edu


   Abstract
Top
 Abstract
Introduction
 Materials and Methods
 Results
 Discussion
 References
 
Background: Currently the rate-limiting step for mass spectrometric analysis of drugs in biological samples is sample preparation. Many gas chromatography/mass spectrometry (GC/MS) methods are specific for a certain class of compounds, requiring extraction and/or derivatization before analysis. The purpose of this study was to develop a broad spectrum liquid chromatography/mass spectrometry (LC/MS) procedure that allowed for direct analysis of urine specimens with potential for quantitative analysis.

Methods: We modified a commercially available column-switching instrument, the REMEDi HS from Bio-Rad Diagnostics, to make it compatible with atmospheric pressure ionization. The system we developed was based on electrospray ionization and used three LC columns to extract, purify, and separate drugs directly from urine specimens. Drugs and metabolites were tentatively identified on the basis of retention times and (M+H)+ ions. Tandem mass spectrometry (MS/MS) was used to confirm the qualitative identification of suspected drugs, using data-dependent acquisition. For quantitative analysis, the cocaine metabolite benzoylecgonine was analyzed using isotope dilution and selected reaction monitoring.

Results: Seventeen basic drugs from a variety of classes of compounds were identified directly from urine without the need for prior sample extraction, using LC and MS/MS. Quantitative analysis was demonstrated for benzoylecgonine. When benzoylecgonine-d3 was used as the internal standard, the method was linear from 30 to 10 000 µg/L (range tested). At these concentrations, the within-run accuracy was ± 10% of the target concentration, with CVs <10%. Analytical results by LC/MS/MS compared favorably with GC/MS values for 50 benzoylecgonine-containing specimens and for 25 negative specimens.

Conclusions: The ability to directly analyze urine for a wide variety of drug classes, combined with the sensitivity and specificity of LC/MS/MS makes this technique attractive for many clinical, forensic, and biotechnology applications.© 1999 American Association for Clinical Chemistry


   Introduction
Top
 Abstract
 Introduction
Materials and Methods
 Results
 Discussion
 References
 
Currently, the rate-limiting step for isolating, identifying, and quantifying the presence of drugs in biological samples is sample preparation. For most clinical and forensic applications, initial screening is done by an immunoassay, and presumptive positive samples are confirmed by a more specific method such as gas chromatography/mass spectrometry (GC/MS).1 The sensitivity and specificity, combined with its widespread availability, has established GC/MS as the technique of choice for many biomedical applications. Despite the many advantages of GC/MS, it does have limitations, primarily relating to the gas phase properties of the analyte as it passes through the GC column. A successful GC/MS procedure requires a volatile, thermally stable analyte. These requirements generally necessitate extraction from the biological matrix, followed in many cases by a derivatization reaction to obtain desirable chromatographic characteristics. These limitations led us to investigate other possibilities for analyzing biological specimens.

Automated solid phase extraction (SPE) is one potential solution to minimizing the time spent by technical staff preparing samples for GC/MS analysis. Robotic systems for the analysis of a variety of drugs are commercially available, and several publications have demonstrated their utility for the analysis of drugs of abuse (1)(2)(3)(4). These studies demonstrated that robotic systems can prepare samples in a consistent manner with accuracy and precision equal or superior to that of manual extractions. However, a major drawback with SPE is that many procedures are generally class specific. For example, an extraction/derivatization method designed for GC/MS analysis of amphetamines will not necessarily work for opiates. Additionally, once the robotic system has prepared the samples for analysis, many SPE methods include manual steps such as transferring the extract to an autosampler vial. These inherent sample preparation requirements make GC/MS analysis of many biologically important compounds a time-consuming process that cannot be bypassed with automation.

Liquid chromatography/mass spectrometry (LC/MS) is an alternative approach that can reduce the off-line sample preparation required with GC/MS because relatively nonvolatile compounds can be analyzed (5)(6)(7)(8)(9). LC/MS applications for forensic and clinical toxicology have been reviewed recently (10). With the proper choice of stationary and mobile phases, acceptable LC separations can be achieved for very polar analytes such as amines, alcohols, and carboxylic acids. This capability of analyzing a variety of compounds in a liquid phase makes LC attractive for the direct analysis of biological samples. Several studies have reported the analysis of compounds directly from biological matrices by LC (11)(12)(13)(14)(15)(16)(17)(18)(19)(20)(21), but only a few have reported direct analyses of biofluids by LC/MS (22)(23)(24). Generally, the approach for the direct LC analysis of biological specimens involves restricted access media (RAM) columns and/or column switching.

RAM columns are multifunctional stationary phases that are inert to large biomolecules such as proteins, allowing them to pass through the column while preferentially binding the smaller hydrophobic molecules of interest. RAM stationary phases have a porous hydrophilic exterior and a hydrophobic interior. The pore size of the exterior coating allows small molecules access to the reversed-phase interior coating, which binds the molecules until the mobile phase is changed to selectively elute them. This combination of size exclusion and reversed phase has been used for a variety of chromatographic separations, primarily involving serum (17)(18)(19)(20)(21).

A second approach to the direct analysis of compounds in a biological matrix uses multiple LC columns to isolate the compounds of interest on-line before analysis (5)(6)(7)(8)(9)(10)(11)(12)(25), a method that was first developed in the early 1970s (26)(27). The REMEDi HS from Bio-Rad Laboratories is a commercially available column-switching LC instrument designed to detect >700 different drugs directly from urine or serum. The REMEDi HS bases identification on retention times relative to an internal standard and on ultraviolet (UV) spectra that are searched against a library. Several reports have demonstrated the utility of the REMEDi HS in both clinical and forensic settings (28)(29)(30)(31)(32)(33)(34)(35). One limitation of the REMEDi HS is its reliance on the UV spectra for identification purposes. Compared with MS, the use of UV spectra for identification is a relatively insensitive and nonspecific technique that requires good chromatographic peak resolution when searching library spectra. Many compounds have similar UV spectra, and when they coelute, positive identification is difficult. Because UV spectra are not as definitive as mass spectra, we were interested in developing LC/MS for the direct analysis of biological specimens.

The purpose of the present work was to demonstrate that a column-switching LC/MS system could identify a broad range of analytes directly from a urine specimen on the basis of retention times, full-scan spectra, and product-ion spectra. Building on established chromatographic techniques and taking advantage of recent improvements in MS, we designed an online LC extraction, purification, and separation procedure compatible with atmospheric pressure ionization (API). This system was used to show, for the first time, that a wide variety of basic drugs could be qualitatively identified from a urine specimen in a single analysis. Using selected reaction monitoring (SRM) and isotope dilution, we validated the use of this system for the quantitative analysis of benzoylecgonine.


   Materials and Methods
Top
 Abstract
 Introduction
 Materials and Methods
Results
 Discussion
 References
 
apparatus
A Finnigan LCQ ion trap mass spectrometer was used in the electrospray ionization (ESI) mode. The spray voltage was ~2.4 kV with a current of about 20 µA. Both the sheath gas and the auxiliary gas were nitrogen set at 80 and 20 relative units, respectively. The capillary entrance to the ion trap was at an offset of 3.3 V and was maintained at 200 °C.

For qualitative identifications, the LCQ was operated in the data-dependent mode. The mass spectrometer was scanned from 50 to 500 atomic mass units (amu), with a full-scan target of 5 x 107 relative units and a tandem mass spectrometry (MS/MS) target of 2 x 107 relative units. The maximum ionization time was set to 400 ms, and three microscans were collected for each data point.

For SRM of benzoylecgonine, the same basic MS settings were used with the exception that the maximum ionization time was 200 ms. The precursor ions of benzoylecgonine and benzoylecgonine-d3 were 290.2 and 293.2 m/z with an isolation width of 2 amu. Product-ion scans were from 165.5 to 170.5 m/z and from 168.5 to 173.5 m/z for benzoylecgonine and benzoylecgonine-d3, respectively. Quantitative analysis was based on peak area ratios of the 168.1 ± 0.5 amu ion relative to the internal standard ion at 171.1 ± 0.5 amu. Each day of analysis, a two-point calibration curve at 150 and 1500 µg/L was run in duplicate. Benzoylecgonine controls were analyzed at 112 and 246 µg/L (n = 5) each day of analysis for 5 different days (total n = 25).

chemicals and reagents
The acetonitrile, isopropanol, and methanol were all HPLC grade or better. The "application buffer" was 1 g/L potassium borate, pH 8. The "exchange buffer" was 16 mmol/L ammonium acetate, pH 6.3, containing 20 g/L isopropyl alcohol. The "mobile phase" was 16 mmol/L ammonium acetate, pH 7.05, containing 330 mL/L acetonitrile. The "transfer buffer" contained 2 g/L pentanesulfonic acid, 200 mL/L ethanol, and 20 g/L acetic acid. The "wash solution" was HPLC-grade methanol.

lc columns
Column 1 was a PRP-1 [16 µm poly(styrene-divinylbenzene) copolymer; 3.2 mm (i.d.) x 20 mm], column 2 was an Aminex, an anion-exchange column [11 µm; 4.6 mm (i.d.) x 20 mm], and column 3 was unmodified silica [45 µm; 6 mm (i.d.) x 160 mm]. All LC columns were obtained from Bio-Rad Diagnostics.

sample preparation
Sample preparation consisted of adding 200 µL of the internal standard solution to 1 mL of urine. The internal standard was a 6 mol/L ammonium acetate buffer, pH 8.0, containing 20 mL/L isopropanol and 7-chloro-1,3,-dihydro-1-ethyl-5-phenyl-2H-1,4-benzodiazepine-2-one (N-ethylnordiazepam) as internal standard 1 and chlorpheniramine as internal standard 2. When added to 1 mL of urine, the concentrations of internal standards 1 and 2 were 2000 and 3000 µg/L, respectively. After centrifugation (2 min at 9500g), 1 mL of the sample solution was injected by an autosampler. For quantitative analysis of benzoylecgonine, the same procedure was used with the addition of 10 µL of 15.8 mg/L benzoylecgonine-d3.

Specimens containing >10 000 or 100 000 µg/L benzoylecgonine as determined by GC/MS were diluted 1:10 or 1:20, respectively, with drug-free urine before LC/MS/MS analysis. Comparisons of GC/MS and LC/MS/MS were made on specimens collected from routine drug screens that had been stored frozen (less than -20 °C) for ~1 year.

time event table
The time event table for column switching shown in Table 1 was developed by varying the settings and monitoring the recovery of drugs of interest as a function of interfering substances and stability of ion current. It begins with two wash steps immediately after the urine sample is loaded onto column 1, including a forward-flow wash and a reversed-flow wash using the application buffer. Drugs and metabolites with different polarities are absorbed onto the resin of column 1, whereas proteins, minerals, salts, and very hydrophilic components are washed off to the waste by the application buffer. Upon completion of the washing steps, a small amount of exchange buffer is applied to column 1 from the reverse direction to rinse off the application buffer and to prepare the cartridge for the release of drugs and metabolites. The mobile phase then continues the transfer from column 1 to column 2. Up to this point, all solvents are delivered by pump A. Immediately after step 5, pump B is used to complete the transfer of analytes to column 2. Basic drugs and metabolites pass through column 2 with minimal retention, whereas most neutral, acidic, and very polar components are retained by the anion-exchange resin. Both columns 1 and 2 are switched out of the loop immediately after the basic drugs are transferred to column 3 and are washed with the wash solution. From this point on, pump B continues to deliver mobile phase for the separation of basic drugs and metabolites to column 3 at a rate of ~1.7 mL/min. The total process takes ~12 min per sample: the analytical separation time is 9 min, and the sample loading and online purification take 3 min.


View this table:
[in this window]
[in a new window]
  		Table 1. Time event table for column and mobile phase switching.

REMEDi LC/MS INTERFACE
The effluent from the UV cell of the REMEDi HS was connected to an adjustable PEEK splitter (Upchurch) with ~6 feet of 0.01-inch (i.d.) PEEK tubing. Effluent from the splitter was sent to waste or to the LCQ through 0.005-inch (i.d.) PEEK tubing. The split ratio was adjusted so that 95% of the mobile phase was sent to waste with ~85 µL/min going to the mass spectrometer. The LCQ analysis list was started by a contact closure initiated from the REMEDi HS.


   Results
Top
 Abstract
 Introduction
 Materials and Methods
 Results
Discussion
 References
 
qualitative analysis
LC effluents can be introduced into mass spectrometers using ESI, which typically produces protonated molecular ions. Fig. 1 A shows a full-scan chromatogram of a urine specimen that had been supplemented with a variety of drugs from several different classes of compounds. The relatively clean chromatogram combined with the wide range of compounds identified in a single experiment demonstrates the analytical power of this system. Under full-scan conditions (50–500 m/z), some of the peaks (e.g., amphetamine) do not appear very intense. However, when selected ions are plotted, as shown for amphetamine (136 m/z) in Fig. 1B , the signal-to-noise ratio is >600:1.



View larger version (21K):
[in this window]
[in a new window]
  		Figure 1. Direct LC/MS analysis of a urine specimen containing N-ethylnordiazepam, diazepam, amphetamine, imipramine, morphine, chlorpheniramine, and hydrocodone at concentrations of 2000, 1500, 2200, 2000, 2000, 3000, and 2000 µg/L, respectively.

The y-axis has been normalized to the most intense peak. (A), total ion chromatogram (50–500 m/z). (B), selected ion plot (136 m/z) showing the signal-to-noise ratio for amphetamine. (C), data-dependent scan.

One disadvantage of using ESI to identify unknowns is that it often produces very little fragmentation. Compounds, such as morphine and hydromorphone, that have the same nominal molecular weight will have similar ESI spectra as shown in Fig. 2 , A and B. Although the chromatographic system used separates these two compounds by >2 min, it is necessary to perform collisionally induced dissociation (CID) MS/MS experiments to clearly distinguish these compounds. The LCQ enables MS/MS data to be collected in a data-dependent mode. Fig. 1CUp shows the data-dependent scanning chromatogram for the urine sample shown in Fig. 1AUp . The jagged nature of the peaks in Fig. 1CUp are the result of the instrument switching from a full-scan mode to an MS/MS mode. When an ion exceeds a preset threshold in the full-scan mode, the instrument switches to collect a CID spectrum of the most intense ion. Because the relative ion current drops when MS/MS is performed, the total signal drops, causing a sharp decline in the chromatographic peak. Following the MS/MS scan, the mass spectrometer reverts back to full-scan mode, collects a spectrum, and checks to see if the data-dependent threshold is exceeded. If the threshold is still exceeded, it then collects another MS/MS spectrum of the most intense peak. This process is repeated until the end of the chromatographic run. Data-dependent scanning is important because CID spectra of drugs generally contain more characteristic ions than full-scan single MS experiments. Although CID spectra can be obtained in several different modes (e.g., source CID and separate injections), the ability to perform data-dependent scanning can eliminate the need for a second injection. The CID spectra of morphine and hydromorphone are shown in Fig. 2 , C and D. Although the full-scan spectra of these drugs are similar, the CID spectra of morphine and hydromorphone are clearly distinct.



View larger version (22K):
[in this window]
[in a new window]
  		Figure 2. Electrospray and CID spectra of morphine and hydromorphone.

(A), electrospray spectrum of morphine; (B), electrospray spectrum of hydromorphone; (C), CID spectrum of morphine; (D), CID spectrum of hydromorphone.

It is possible to generate CID spectra using a single-stage MS instrument within the ion source by changing the electrospray voltages, a technique sometimes referred to as "source CID" or "up-front CID". However, a primary limitation of doing such experiments is that the lineage of the product ions is lost. This is an important point when analyzing biological extracts and for identifying coeluting compounds. For example, if source CID was done for benzoylecgonine, which coelutes with several unidentified endogenous compounds, the resulting spectrum would be a composite of both the benzoylecgonine CID spectrum and the endogenous peaks CID spectrum. By performing MS/MS in the ion trap, we can first isolate the benzoylecgonine protonated molecular ion (290 m/z) and then perform CID in the trap, with the knowledge that any resulting product ions originated from the 290 m/z ion.

We tested 17 drugs by adding them to urine specimens and analyzing the specimens, using the data-dependent mode to collect CID spectra. The retention times, observed protonated molecular ions, and relative intensities of the product ions in the CID spectra are shown in Table 2 . Weinmann and Svoboda (36) recently published several CID spectra for drugs of abuse obtained with a triple quadrupole mass spectrometer. Although there is some general agreement between our spectra and those obtained by Weinmann and Svoboda (36), there are some notable differences. For example, in our experiments, the CID spectrum of amphetamine shows almost exclusively a single product ion at 119 m/z, with 91 m/z having a relative abundance of <1%. The amphetamine CID spectrum obtained with the triple quadrupole is nearly the reverse of what we observed with the ion trap: 91 m/z was the most intense product ion, with 119 m/z having a relative abundance of a few percent (36). There are some obvious differences in the way in which the triple quadrupole CID data were collected compared with our ion trap data (e.g., collision energy, collision gas, and ion transmission) that could explain the observed differences. However, the important point is that for drugs of abuse, CID spectra collected on different types of instruments will not be identical. At present, there are no widely available libraries of CID spectra for drugs of abuse, but it appears that the best CID libraries will be instrument specific.


View this table:
[in this window]
[in a new window]
  		Table 2. Retention times, precursor ions, and the five most abundant product ions observed for commonly encountered drugs analyzed directly from urine specimens.

The commercially available REMEDi HS UV uses chlorpheniramine as an internal standard, making it difficult to identify the use of this drug when it is present at low concentrations. However, previous experience with REMEDi showed that the chlorpheniramine metabolite norchlorpheniramine is often detected in urine samples from patients taking this drug (32). We did not analyze any samples from patients taking chlorpheniramine to determine whether its metabolites were detected with the mass spectrometer, but we anticipate eventually substituting a deuterium-labeled internal standard for chlorpheniramine to help ensure that exposure to this antihistamine will be detected.

Many drugs are excreted as glucuronide metabolites, including the opiates and benzodiazepines, and it is important to know how the system we developed works for conjugated metabolites. Although we did not fully evaluate the system for the analysis of glucuronide metabolites, two patient samples that were immunoassay positive for opiates did not show any glucuronide metabolites when injected into the REMEDi/MS. GC/MS analysis of these samples, after glucuronidase hydrolysis, SPE, and derivatization, showed that sample 1 contained morphine at 611 µg/L and sample 2 contained both morphine and codeine at 435 and 827 µg/L, respectively. ß-Glucuronidase pretreatment before REMEDi/MS analysis was successful in detecting morphine in sample 1 and both morphine and codeine in sample 2 without a noticeable increase in sample background. These preliminary REMEDi/MS experiments showed that off-line hydrolysis was compatible with the system we developed, but further studies are required to validate this finding.

quantitative analysis
Benzoylecgonine was chosen as the target compound for initial quantitative studies because it is widely measured as an indicator of cocaine abuse. With the MS/MS settings used, benzoylecgonine predominately formed a product ion at 168 m/z; this ion probably is produced by cleavage of the benzoyloxy group as shown in Fig. 3 . This product ion is consistent with CID fragmentation of other cocaine metabolites and a previous report (37). Because the deuterium labels are on the methyl group attached to the nitrogen, the internal standard forms an analogous product ion at 171 m/z.



View larger version (16K):
[in this window]
[in a new window]
  		Figure 3. CID spectrum of benzoylecgonine.

To eliminate space charging and ion molecule reactions, it is necessary to regulate the number of ions in the trap (38). The LCQ uses a short prescan to determine the number of ions entering the trap and then adjusts the ionization time to ensure that the trap is storing ions in an optimal manner. Consequently, with the ion trap, the number of data points collected per unit of time depends on the intensity of the ion signal. This is in contrast to linear quadrupole mass spectrometers, which collect a certain number of data points per unit time interval. The three important settings that determine sampling rate with the ion trap are target setting, maximum ionization time, and number of microscans. We left the MS/MS target, which reflects the desired number of ions in the trap, at the default setting of 2 x 107 relative units and varied the maximum ionization time and the number of microscans to optimize precision of SRM for benzoylecgonine. Table 3 shows that the combination of three microscans and a 200-ms maximum ionization time provided the best precision for quantifying ion ratios of benzoylecgonine at 150 µg/L. The settings were optimized at this concentration because it reflects the current cutoff for a positive confirmation of a presumptive positive immunoassay result. At 150 µg/L, the within-run CV was 3.4% (n = 6).


View this table:
[in this window]
[in a new window]
  		Table 3. Effect of maximum ion time and number of microscans on imprecision of ion ratio measurements (168 relative to 171) for SRM of benzoylecgonine (n = 6 for each condition).

The REMEDi UV chromatograms of a blank urine specimen and a urine specimen containing 150 µg/L of benzoylecgonine are shown in Fig. 4 , A and B. These chromatograms are indistinguishable, clearly demonstrating the that REMEDi UV would not work for quantitative analysis of benzoylecgonine at this concentration. Representative SRM chromatograms of a blank urine specimen and a 150 µg/L urine calibrator are shown in Fig. 4 , C and D. As can be seen from these chromatograms, there are no peaks identifiable as benzoylecgonine in the blank, whereas the 150 µg/L calibrator clearly shows a peak corresponding to benzoylecgonine at 5.17 min. The benzoylecgonine retention time difference of ~0.35 min between the UV chromatogram and the SRM plots is produced by the interface between the UV detector and the mass spectrometer.



View larger version (17K):
[in this window]
[in a new window]
  		Figure 4. REMEDi and SRM chromatograms of a blank urine sample and a urine sample containing benzoylecgonine.

(A), REMEDi UV trace of a blank urine sample; (B), REMEDi UV trace of a urine sample containing 150 µg/L benzoylecgonine, which is not detectable. (C), SRM chromatogram of a blank urine specimen for benzoylecgonine [290 to 168 m/z; retention time (RT), 5.17 min] and benzoylecgonine-d3 (293 to 171 m/z; retention time, 5.19 min). The noise at 5.17 min is equivalent to <0.5 µg/L benzoylecgonine. (D), SRM chromatogram of a urine calibrator containing 150 µg/L benzoylecgonine and benzoylecgonine-d3. AU, absorbance units.

The accuracy and imprecision of the SRM method for measuring benzoylecgonine added to urine specimens is shown in Table 4 . The within-run CVs were 2.8–8.8%. The relative errors were 0.2–9.3% at concentrations between 75 and 10 000 µg/L. The lowest concentration tested, 30 µg/L, had a relative error of 19.6%, which meets the acceptability criteria used in most forensic urine drug testing laboratories. We did not evaluate the lowest concentration detectable because 30 to 10 000 µg/L met the needs of a drug testing laboratory. At 30 µg/L, the signal-to-noise ratio was >100:1. Although more experiments are required, better limits of detection could probably be achieved by increasing the chromatographic efficiency of the system. Narrow bore columns could be developed that would introduce more concentrated peaks to the mass spectrometer, and the current practice of sending 95% of the sample to waste before MS analysis could be minimized.


View this table:
[in this window]
[in a new window]
  		Table 4. Within-run accuracy and imprecision for quantifying benzoylecgonine, using SRM.1

The linearity of the method was evaluated by adding 30 to 10 000 µg/L benzoylecgonine to drug-free urine samples (n = 4 for each concentration). As shown in Fig. 5 , the method was linear in this concentration range [y = (0.97 ± 0.01)x - (6.1 ± 47); r2 = 0.999]. This dynamic range is better than most GC/MS procedures based on trimethylsilyl derivatization, primarily because the inherent limitation of the isotope cascade caused by the A+2 effect of 30Si is bypassed in the LC/MS procedure. The isotope cascade effect of silyl derivatives is particularly problematic when internal standards with three or less deuteriums are used, as is commonly done in many toxicology laboratories (e.g., benzoylecgonine-d3, codeine-d3, and morphine-d3).



View larger version (12K):
[in this window]
[in a new window]
  		Figure 5. Linearity of benzoylecgonine analyzed directly from urine specimens.

The equation for the line is: y = (0.97 ± 0.01)x - (6.1 ± 47). Each data point represents the mean ± SD (n = 4).

Carryover was evaluated by injecting a blank urine specimen containing the benzoylecgonine-d3 internal standard immediately after a sample that contained 10 000 µg/L benzoylecgonine. There was no identifiable benzoylecgonine peak in the blank sample that followed the 10 000 µg/L calibrator.

To establish that our system provided satisfactory results in a realistic setting, we evaluated 50 urine specimens collected from random drug screens that had previously confirmed positive for benzoylecgonine by GC/MS. The GC/MS procedure was based on isotope dilution, SPE, and trimethylsilyl derivatization. Fig. 6 shows the correlation between GC/MS and LC/MS/MS over a wide range of concentrations [y = (0.97 ± 0.02)x - (263 ± 839); r2 = 0.987]. All LC/MS/MS results agreed within ± 20% of the original concentrations assigned by GC/MS. The accuracy of the GC/MS method over this concentration range was ± 20%, and there were no significant differences between the LC/MS/MS and GC/MS quantitative values, using a paired t-test (P = 0.05). To ensure that no false-positive results would occur, 25 specimens that were negative for benzoylecgonine were analyzed by LC/MS/MS. As expected, none of the negative specimens had peaks corresponding to benzoylecgonine.



View larger version (11K):
[in this window]
[in a new window]
  		Figure 6. Comparison of LC/MS/MS values with GC/MS values for 50 samples that contained benzoylecgonine.

The equation for the line is: y = (0.97 ± 0.02)x - (263 ± 839). There were no differences between the two methods, based on a paired t-test (P = 0.05). This is a log/log plot to display the wide range of concentrations analyzed.

The precision of the SRM method for quantifying benzoylecgonine was evaluated over a period of several weeks by analyzing a low-concentration control (112 µg/L benzoylecgonine) and a high-concentration control (246 µg/L benzoylecgonine) with each batch of specimens. A typical batch size was 38 samples and consisted of donor specimens mixed with calibrators and controls. The concentration of controls was selected to demonstrate that we could reliably distinguish a positive sample from one that contained benzoylecgonine at a concentration below the cutoff of 150 µg/L. As shown in Table 5 , the imprecision was 13% for the low control and 8.7% for the high (above cutoff) control. All of the controls were properly classified as to being above or below the cutoff, demonstrating that the method has good discriminating power around the 150 µg/L cutoff. The precision of the method using the LC/MS/MS ion trap for analyzing controls was not as good as we usually get with our routine linear quadrupole GC/MS method, which has CVs of ~4%. This is consistent with our previous experiments that showed an ion trap similar to the one used in the present studies did not measure ion ratios as precisely as a linear quadrupole (39).


View this table:
[in this window]
[in a new window]
  		Table 5. Imprecision of analyzing benzoylecgonine urine controls with LC/MS/MS.1


   Discussion
Top
 Abstract
 Introduction
 Materials and Methods
 Results
 Discussion
References
 
The first publication that demonstrated the application of MS/MS for the analysis of drugs directly from serum samples did not use any chromatographic separation (40). These authors screened specimens using a direct insertion probe and SRM to look for a variety of drugs on the basis of known transitions from precursor ions. Once tentatively identified, the drug identity was confirmed by collecting the full product-ion spectrum. Results of this initial study demonstrated that the direct analysis of drugs in serum was possible, but that significant background existed. Using solvent extraction, these authors reduced the background and were able to identify a variety of drugs in the microgram per liter concentration range (40). Later experiments showed that these types of MS/MS experiments could be successfully used to detect a variety of drugs directly from urine specimens (41). Recently, Weinmann and Svoboda (36) demonstrated good sensitivity and linear calibration curves for a variety of drugs of abuse, using SPE and flow injection ion spray MS/MS, but they concluded that for forensic cases, results obtained with flow injection MS/MS analysis must be verified by a second specific method. Despite the success of using MS/MS for the rapid analysis of drugs from biological samples and extracts, these methods have not been put to widespread use.

A primary limitation of direct analysis of biofluids by MS/MS is that the resolving power of a chromatographic system is eliminated. The importance of a chromatographic separation should not be overlooked. Even with the moderate resolution of the LC system we used, compounds with the same nominal molecular weight, such as morphine and hydromorphone, are separated by >2 min. This has important implications for quantitative analysis when both these compound are present in the same sample because there is some overlap of virtually all of the ions in the CID spectra of these two isomers (Fig. 2Up ). Fig. 2DUp shows three ions in the hydromorphone spectrum that could potentially be used to distinguish it from morphine [m/z (relative abundance), 185 (100%), 227 (34%), and 243 (23%)]. However, a careful look at the morphine spectrum (Fig. 2CUp ) shows that these ions are also present [m/z 185 (10%), 227 (4%), and 243 (0.5%)]. When performing quantitative analysis of a sample containing a mixture of these two compounds, it would be difficult to get accurate results, especially if hydromorphone was present at low concentrations in a sample with high amounts of morphine. In addition to eliminating spectral overlap, the retention time is also characteristic of the molecule of interest and provides additional support for conclusive identifications.

The choice of mobile phases and columns will determine what type of compounds can be analyzed with LC/MS/MS. Various combinations of column switching, RAM, and immunoaffinity extraction have been used successfully for the direct analysis of compounds from biological specimens. Rule and Henion (42) demonstrated that an immunoaffinity column before a RAM column dramatically improved sample cleanup for the analysis of propranolol directly from urine specimens. These authors also showed that large volumes (20 mL) of dilute urine could be analyzed with immunoaffinity chromatography to detect low concentrations of lysergic acid diethylamide present in urine. Another particularly attractive feature of the procedure of Rule and Henion is the general applicability of the protein G column used for immunoaffinity purification. Because protein G has a high affinity for the Fc region of IgG, it can be potentially be used for any IgG antibody. The limitation of this procedure is that an antibody is required for each compound of interest and consequently would be less useful as a broad-spectrum screen.

Our choice of LC columns was based on previous publications demonstrating that various combinations of a PRP-1 column, a reversed-phase column, an anion-exchange column, and a bare silica column provided retention and separation of the majority of drugs of abuse directly from urine specimens. De Jong et al. (43) were the first to show that a PRP-1 column (column 1) in series with an Aminex (column 2) and a C18 column could be used for the direct analysis of barbiturates from urine specimens with UV detection. Slais et al. (44) used a similar configuration for the direct analysis of amphetamines from urine specimens. Binder et al. (11) improved the chromatographic separation by adding an unmodified silica column and showed that by using four columns, they could identify a wide variety of compounds when a scanning UV detector was used. In our preliminary experiments, we also used the four columns described by Binder et al., but because each column adds an additional expense, we were interested in developing a three-column system. We eliminated the C18 column and were able to resolve most of the drugs of interest with the exception of the benzodiazepines, which coelute at the beginning of the chromatographic run. Although we did not do conclusive studies, the major impact of eliminating the C18 column was on the separation of benzodiazepines, which are not retained on the bare silica column. However, because the clinically important benzodiazepines all have different molecular weights, they are easily differentiated by the mass spectrometer, obviating the need for chromatographic separation. Other than the effect on benzodiazepine separation, eliminating the C18 column appeared to have little effect on the overall performance of the system. Our final combination of a PRP-1 column, an anion-exchange column, and a bare silica column is the first description of this combination for broad spectrum drug isolation and separation.

Previous work demonstrated that sodium borate was optimal for binding drugs to the first column during the application step and that borate was useful as the exchange buffer when initiating the transfer of drugs from column 1 to column 2 (11). Sodium borate is not volatile and, consequently, not compatible with API. As the electrospray droplets evaporate, any sodium borate present will crystallize on the ESI needle and capillary inlet, preventing ions from reaching the mass analyzer. We retained the sodium borate in the application step but eliminated it from the exchange buffer. As expected, by using the previously described application buffer, we were able to bind drugs of interest to the first column. Developing a new exchange buffer by substituting ammonium acetate for sodium borate allowed us to initiate drug transfer from column 1 to column 2 and prevented sodium borate from reaching the electrospray interface. Our new exchange buffer retains the desired properties of the sodium borate buffer used previously yet is amenable to API. The final mobile phase used in previous studies also contained the nonvolatile buffer potassium phosphate as well as tetramethylammonium hydroxide and dimethyloctylamine, which clearly are not compatible with our ESI interface (11). We designed a new mobile phase that uses ammonium acetate and acetonitrile. Various concentrations of ammonium acetate (from 4 to 30 mmol/L) in 670 mL/L water–330 mL/L acetonitrile were evaluated as the final mobile phase. The best compromise between peak shape and resolution was 16 mmol/L ammonium acetate; consequently, this was used for all of our studies.

The ability to bypass the need for manual extractions while maintaining the capabilities of a broad-spectrum analyzer makes this LC/MS method attractive in a variety of settings. Clinically, it is critical that any screening technique be sufficiently sensitive to detect a range of potential intoxicants, whereas specificity is required to help identify treatment strategies. The sensitivity of electrospray combined with the specificity of MS/MS enhances the diagnostic value of a urine drug screen and should aid in the treatment of suspected overdosed patients. In a postmortem forensic setting, this LC/MS/MS system could provide vital information as to cause of death, especially for compounds not detected by other routine techniques. Finally, from a more global perspective, this methodology could allow analytical chemists, through the proper selection of columns, mobile phases, and timed events, to design online sample preparations for virtually any compound amenable to API.

In conclusion, there have been several reports that used LC/MS to analyze single components or single classes of drugs directly from biological specimens. For the first time we have demonstrated that a wide range of basic drugs, from a variety of classes of compounds, can be identified directly from urine specimens using LC/MS/MS. We were successful in coupling recent advances in API and ion trap MS with evolving online chromatographic sample preparation to identify drugs based on retention times and full-scan and product-ion spectra. Using SRM and isotope dilution, we were able to validate the method for the quantitative analysis of the cocaine metabolite benzoylecgonine at concentrations encountered in routine forensic urine drug testing. The sensitivity can be improved by reducing the inner diameter of the LC columns to eliminate the need for effluent splitting just before MS analysis. Future experiments will be devoted to optimizing the LC system for more efficient separations, applications for other biofluids, and to software development for automated drug identification.


   Acknowledgments
 
We thank Jason Lai and Steve Binder for expert assistance and acknowledge Bio-Rad Diagnostics, Hercules, CA, for providing reagents, LC columns, and instrument support.


   Footnotes
 
1 Nonstandard abbreviations: GC, gas chromatography; MS, mass spectrometry; SPE, solid phase extraction; LC, liquid chromatography; ESI, electrospray ionization; amu, atomic mass unit; MS/MS, tandem mass spectrometry; RAM, restricted access media; UV, ultraviolet; API, atmospheric pressure ionization; SRM, selected reaction monitoring; and CID, collisionally induced dissociation.


   References
Top
 Abstract
 Introduction
 Materials and Methods
 Results
 Discussion
 References
 

  1. McCambly K, Kelly RC, Johnson T, Johnson JE, Brown WC. Robotic solid-phase extraction of amphetamines from urine for analysis by gas chromatography-mass spectrometry. J Anal Toxicol 1997;21:438-444. [Medline] [Order article via Infotrieve]
  2. Repique EV, Sacks HJ, Farver SJ. Isolation of drugs from blood serum by the DuPont Prep I automated sample processor and a C18 bonded phase adsorption column. Clin Biochem 1981;14:196-200. [Medline] [Order article via Infotrieve]
  3. Balkon J, Donnelly B, Prendes DA. A rapid isolation technique for drugs from tissues and fluids: use of the DuPont Prep 1 system. J Forensic Sci 1982;27:23-31. [Medline] [Order article via Infotrieve]
  4. Gere JA, Platoff GE. Solid phase extraction of abused drugs in urine. Lui RH Goldberger BA eds. Handbook of workplace drug testing 1995:23-44 AACC Press Washington, DC. .
  5. Henion JD, Thompson BA, Dawson PH. Determination of sulfa drugs in biological fluids by liquid chromatography/mass spectrometry/mass spectrometry. Anal Chem 1982;54:451-456. [Medline] [Order article via Infotrieve]
  6. Covey TR, Lee ED, Henion JD. High-speed liquid chromatography/tandem mass spectrometry for the determination of drugs in biological samples. Anal Chem 1986;58:2453-2460. [Medline] [Order article via Infotrieve]
  7. Tatsuno M, Nishikawa M, Katagi M, Tsuchihashi HJ. Simultaneous determination of illicit drugs in human urine by liquid chromatography-mass spectrometry. J Anal Toxicol 1996;20:281-286. [Medline] [Order article via Infotrieve]
  8. Clauwaert KM, Van Bocxlaer JF, Lambert WE, Van den Eeckhout EG, Lemiere F, Esmans EL, De Leenheer AP. Narrow-bore HPLC in combination with fluorescence and electrospray mass spectrometric detection for the analysis of cocaine and metabolites in human hair. Anal Chem 1998;70:2336-2344. [Medline] [Order article via Infotrieve]
  9. Nishikawa M, Nakajima K, Tatsuno M, Kasuya F, Igarashi F, Fukui M. The analysis of cocaine and its metabolites by liquid chromatography/atmospheric pressure chemical ionization-mass spectrometry (LC/APCI-MS). Forensic Sci Int 1994;66:149-158. [Medline] [Order article via Infotrieve]
  10. Maurer HH. Liquid chromatography-mass spectrometry in forensic and clinical toxicology. J Chromatogr B 1998;713:3-25.
  11. Binder SR, Regalia M, Baiggi-McEachern M, Mazhar J. Automated liquid chromatographic analysis of drugs in urine by on-line sample cleanup and isocratic multicolumn separation. J Chromatogr 1989;473:325-341. [ISI][Medline] [Order article via Infotrieve]
  12. Lauber R, Mosimann M, Buhrer M, Zbinden AM. Automated determination of midazolam in human plasma by high-performance liquid chromatography using column switching. J Chromatogr B 1994;654:69-75. [Medline] [Order article via Infotrieve]
  13. Huang JL, Morgan DJ. Simple direct injection high-performance liquid chromatographic method to determine quinidine in plasma. J Chromatogr 1993;620:278-280. [Medline] [Order article via Infotrieve]
  14. Polettinin A, Montagna M, Hogeddoorn EA, Dijkman E, van Zoonen P, van Ginkel LA. Applicability of coupled-column liquid chromatography to the analysis of beta-agonists in urine by direct sample injection. I. Development of a single-residue reversed-phase liquid chromatography-UV method for clenbuterol and selection of chromatographic conditions suitable for multi-residue analysis. J Chromatogr A 1995;695:19-31.
  15. McMahon GP, Kelly MT. Determination of aspirin and salicylic acid in human plasma by column-switching liquid chromatography using on-line solid phase extraction. Anal Chem 1998;70:409-414. [Medline] [Order article via Infotrieve]
  16. Yu Z, Westerlund D. Direct injection of large volumes of plasma in a column-switching system for the analysis of local anaesthetics. II. Determination of bupivacaine in human plasma with an alkyl-diol silica precolumn. J Chromatogr A 1996;725:149-155. [Medline] [Order article via Infotrieve]
  17. Cook SE, Pinkerton TC. Characterization of internal surface reversed-phase silica supports for liquid chromatography. J Chromatogr 1986;368:233-248. [Medline] [Order article via Infotrieve]
  18. Desilets CP, Rounds MA, Regnier FE. Semipermeable-surface reversed-phase media for high-performance liquid chromatography. J Chromatogr 1991;544:25-39. [Medline] [Order article via Infotrieve]
  19. Vielhauer S, Rudolphi A, Boos K, Seidel D. Evaluation and routine application of the novel restricted access precolumn packing material alkyl-diol silica: coupled-column high-performance liquid chromatographic analysis of the photoreactive drug 8-methoxypsoralen in plasma. J Chromatogr 1995;666:315-322.
  20. Brewster JD, Lightfield AR, Barford RA. Evaluation of restricted access media for high-performance liquid chromatographic analysis of sulfonamide antibiotic residues in bovine serum. J Chromatogr 1992;598:23-31. [Medline] [Order article via Infotrieve]
  21. Ohta T, Niida S, Nakamura H. Selective extraction of beta-blockers from biological fluids by column-switching high-performance liquid chromatography using an internal-surface phenylboronic acid precolumn. J Chromatogr 1996;675:168-173.
  22. Hogendoorn EA, van Zoonen P, Polettini A, Bouland GM, Montagna M. The potential of restricted access media columns as applied in coupled-column LC/LC-TSP/MS/MS for the high-speed determination of target compounds in serum. Application to the direct and trace analysis of salbutamol and clenbuterol. Anal Chem 1998;70:1362-1368. [Medline] [Order article via Infotrieve]
  23. Vidal C, Kirchner GI, Wunsch G, Sewing K. Automated simultaneous quantification of the immunosuppressants 40-O-(2-hydroxyethyl) rapamycin and cyclosporine in blood with electrospray-mass spectrometric detection. Clin Chem 1998;44:1275-1282. [Abstract/Free Full Text]
  24. McLoughlin DA, Olah TV, Gilbert JD. A direct technique for the simultaneous determination of 10 drug candidates in plasma by liquid chromatography-atmospheric pressure chemical ionization mass spectrometry interfaced to a Prospekt solid-phase extraction system. J Pharm Biomed Anal 1997;15:1893-1901. [Medline] [Order article via Infotrieve]
  25. Ramsteiner KA. Systematic approach to column switching. J Chromatogr 1988;456:3-20.
  26. Huber JJK, Van Der Linden R, Ecker E, Oreans M. Column switching in high-pressure liquid chromatography. J Chromatogr 1973;83:267-277.
  27. Scott CD, Chilcote DD, Katz S, Pitt WW. Advances in the application of high resolution liquid chromatography to the separation of complex biological mixtures. J Chromatogr Sci 1973;11:96-100. [Medline] [Order article via Infotrieve]
  28. Sadeg N, Francois G, Petit B, Dutertre-Catella H, Dumontet M. Automated liquid-chromatographic analyzer used for toxicology screening in a general hospital: 12 months experience. Clin Chem 1997;43:498-504. [Abstract/Free Full Text]
  29. Valentine JL, Middleton R, Sparks C. Identification of urinary benzodiazepines and their metabolites: comparison of automated HPLC and GC-MS after immunoassay screening of clinical specimens. J Anal Toxicol 1996;20:416-424. [Medline] [Order article via Infotrieve]
  30. Ohtsuji M, Lai JS, Binder SR, Kondo T, Takayasu T, Ohshima T. Use of REMEDi HS in emergency toxicology for a rapid estimate of drug concentrations in urine, serum, and gastric samples. J Forensic Sci 1996;41:881-886. [Medline] [Order article via Infotrieve]
  31. Kalasinsky KS, Schaefer T, Binder SR. Forensic application of an automated drug-profiling system. J Anal Toxicol 1995;19:412-418. [ISI][Medline] [Order article via Infotrieve]
  32. Chen JS, Chang KJ, Charng RC, Lai SJ, Binder SR, Essien H. The development of a broad-spectrum toxicology screening program in Taiwan. J Toxicol Clin Toxicol 1995;33:581-589. [Medline] [Order article via Infotrieve]
  33. Ferrara SD, Tedeschi L, Frison G, Brusini G, Castagna F, Bernardelli B, Soregaroli D. Drugs-of-abuse testing in urine: statistical approach and experimental comparison of immunochemical and chromatographic techniques. J Anal Toxicol 1994;18:278-291. [ISI][Medline] [Order article via Infotrieve]
  34. Demedts P, Wauters A, Franck F, Neels H. Evaluation of the REMEDi drug profiling system. Eur J Clin Chem Clin Biochem 1994;32:409-417. [ISI][Medline] [Order article via Infotrieve]
  35. Guitton J, Durand D, Bouillod N, Manchon M. Development of a solid/liquid extraction method for analysis of toxic drugs in serum using an automated liquid chromatograph: the REMEDi. Ann Biol Clin (Paris) 1994;52:111-115. [Medline] [Order article via Infotrieve]
  36. Weinmann W, Svoboda M. Fast screening for drugs of abuse by solid-phase extraction combined with flow-injection ionspray tandem mass spectrometry. J Anal Toxicol 1998;22:319-328. [Medline] [Order article via Infotrieve]
  37. Wang P, Bartlett MG. Collision-induced dissociation mass spectra of cocaine, and its metabolites and pyrolysis products. J Mass Spectrom 1998;33:961-967.
  38. Todd JFJ. Introduction to practical aspects of ion trap mass spectrometry. March RE Todd JFJ eds. Practical aspects of ion trap mass spectrometry 1995;Vol. 1:3-24 CRC Press New York. .
  39. Fitzgerald RL, O'Neal CL, Hart BJ, Poklis A, Herold DA. Comparison of an ion-trap and a quadrupole mass spectrometer using diazepam as a model compound. J Anal Toxicol 1997;21:445-450. [Medline] [Order article via Infotrieve]
  40. Brotherton HO, Yost RA. Determination of drugs in blood serum by mass spectrometry/mass spectrometry. Anal Chem 1983;55:549-553. [Medline] [Order article via Infotrieve]
  41. Yo MH. Analytical methods in forensic chemistry. Chichester: Ellis Horwood, 1990:440pp..
  42. Rule GS, Henion JD. Determination of drugs from urine by on-line immunoaffinity chromatography-high performance liquid chromatography-mass spectrometry. J Chromatogr 1992;582:103-112. [ISI][Medline] [Order article via Infotrieve]
  43. De Jong J, Nielen MWF, Frei RW, Brinkman UAT. Selective on-line sample handling for the determination of barbiturates in urine by liquid chromatography with precolumn technology. J Chromatogr 1986;381:431-437. [Medline] [Order article via Infotrieve]
  44. Slais K, Nielen MWF, Brinkman UAT. Screening of amphetamines by gradient microbore liquid chromatography and pre-column technology. J Chromatogr 1987;393:57-68. [Medline] [Order article via Infotrieve]



The following articles in journals at HighWire Press have cited this article:


Home page
 Clin. Chem.Home page
F.-L. Sauvage, F. Saint-marcoux, B. Duretz, D. Deporte, G. Lachatre, and P. Marquet
Screening of Drugs and Toxic Compounds with Liquid Chromatography-Linear Ion Trap Tandem Mass Spectrometry
Clin. Chem., September 1, 2006; 52(9): 1735 - 1742.
[Abstract] [Full Text] [PDF]

Home page
 J Exp BotHome page
J. M. Halket, D. Waterman, A. M. Przyborowska, R. K. P. Patel, P. D. Fraser, and P. M. Bramley
Chemical derivatization and mass spectral libraries in metabolic profiling by GC/MS and LC/MS/MS
J. Exp. Bot., January 1, 2005; 56(410): 219 - 243.
[Abstract] [Full Text] [PDF]

Home page
 Appl. Environ. Microbiol.Home page
B. Combourieu, P. Besse, M. Sancelme, J.-P. Godin, A. Monteil, H. Veschambre, and A.-M. Delort
Common Degradative Pathways of Morpholine, Thiomorpholine, and Piperidine by Mycobacterium aurum MO1: Evidence from 1H-Nuclear Magnetic Resonance and Ionspray Mass Spectrometry Performed Directly on the Incubation Medium
Appl. Envir. Microbiol., August 1, 2000; 66(8): 3187 - 3193.
[Abstract] [Full Text]

This Article
Right arrow Abstract Freely available
Right arrow Full Text (PDF)
Right arrow Submit an electronic Letter to
the Editor about this paper
Right arrow Alert me when this article is cited
Right arrow Alert me when eLetters are posted
Right arrow Alert me if a correction is posted
Right arrow Citation Map
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in ISI Web of Science
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via ISI Web of Science (35)
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Fitzgerald, R. L.
Right arrow Articles by Herold, D. A.
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Fitzgerald, R. L.
Right arrow Articles by Herold, D. A.