Assessing analytical specificity in quantitative analysis using tandem mass spectrometry

Assessing analytical specificity in quantitative analysis using tandem mass spectrometry

Clinical Biochemistry 38 (2005) 319 – 327 Review Assessing analytical specificity in quantitative analysis using tandem mass spectrometry Mark M. Ku...

241KB Sizes 4 Downloads 46 Views

Clinical Biochemistry 38 (2005) 319 – 327

Review

Assessing analytical specificity in quantitative analysis using tandem mass spectrometry Mark M. Kushnir a,*, Alan L. Rockwooda, Gordon J. Nelsona, Bingfang Yuea, Francis M. Urry a,b b

a ARUP Institute for Clinical and Experimental Pathology, Salt Lake City, UT 84108, USA Department of Pathology, University of Utah School of Medicine, Salt Lake City, UT 84132, USA

Received 17 September 2004; received in revised form 7 December 2004; accepted 13 December 2004 Available online 22 January 2005

Abstract Objectives: The necessity of confirmation of compound identity in quantitative analysis is well recognized for methods utilizing single mass spectrometry detection but is not commonly addressed for applications utilizing multiple-stage mass spectrometry (MSn). For MSn detection, no commonly accepted rules for assessment of analytical specificity in quantitative analyses have been established to date. Methods: To assure compound identity, we evaluated approaches based on monitoring multiple mass transitions of a target compound followed by comparison of the branching ratios of the mass transitions. Results: Monitoring multiple mass transitions along with evaluation of the ratio of their relative intensities allows the analyst to distinguish the target analyte from interferences in quantitative analysis. The strategy and the acceptance criteria are compound and method specific and should be established during the method development and validation. Conclusions: The certainty of analyte identity is very important in quantitative analysis using MSn detection; methods to verify analyte identity should be used in all critical applications. D 2004 The Canadian Society of Clinical Chemists. All rights reserved. Keywords: Tandem mass spectrometry; MS/MS; Quantitative analysis; Analytical specificity; Deconvolution; Testosterone; Sirolimus; Cortisol; Cortisone; Methylmalonic acid

Contents Introduction . . . . . . . . . . . . . . . . . . . . . . . Materials and methods . . . . . . . . . . . . . . . . . . Calculations . . . . . . . . . . . . . . . . . . . . . . Apparatus and chromatographic conditions . . . . . . Evaluation of the analytical specificity . . . . . . . . . . Isotopic ions . . . . . . . . . . . . . . . . . . . . . The same parent/product ion transitions acquired with Other strategies for assessing analytical specificity . . Uncertainty of identity . . . . . . . . . . . . . . . . Quantitative deconvolution . . . . . . . . . . . . . . Conclusions . . . . . . . . . . . . . . . . . . . . . . . Acknowledgment. . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . different . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . collision energies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . .

. . . . . . . . . . . . .

* Corresponding author. Fax: +1 801 584 5207. E-mail address: [email protected] (M.M. Kushnir). 0009-9120/$ - see front matter D 2004 The Canadian Society of Clinical Chemists. All rights reserved. doi:10.1016/j.clinbiochem.2004.12.003

. . . . . . . . . . . . .

. . . . . . . . . . . . .

. . . . . . . . . . . . .

. . . . . . . . . . . . .

. . . . . . . . . . . . .

. . . . . . . . . . . . .

. . . . . . . . . . . . .

. . . . . . . . . . . . .

. . . . . . . . . . . . .

. . . . . . . . . . . . .

. . . . . . . . . . . . .

. . . . . . . . . . . . .

. . . . . . . . . . . . .

. . . . . . . . . . . . .

. . . . . . . . . . . . .

. . . . . . . . . . . . .

. . . . . . . . . . . . .

. . . . . . . . . . . . .

. . . . . . . . . . . . .

. . . . . . . . . . . . .

320 320 320 321 321 323 323 324 324 326 327 327 327

320

M.M. Kushnir et al. / Clinical Biochemistry 38 (2005) 319–327

Introduction Mass spectrometry (MS) detection is capable of providing a high degree of specificity for reliable identification and quantification of compounds of interest. The combination of chromatographic separation with MS and MSn detection yields a particularly specific detection, which is a major reason for the popularity of the technique. MS detection allows a reduction in sample cleanup and chromatographic separation compared to methods utilizing less selective detection; however, on some occasions mass selective detection may produce compromised results due to interference. The most common type of interference with MS detection is when in addition to identical fragment ions the compounds have similar chromatographic retention properties. Another commonly recognized type of interference characteristic of soft ionization techniques (e.g., atmospheric pressure chemical ionization, electrospray) is ion suppression [1]. Compounds that potentially may interfere with MSn detection are those which coelute with the target compound and have the same characteristic parent and product ions. Even though it seems unlikely that parent and product ions from an interfering substance are identical to those of a target analyte, the phenomenon does occur in biochemical applications. The degree of interference is determined by many factors, such as the efficiency of ionization, the relative intensity of the parent and product ions, and the relative concentration of the analytes. Compounds potentially interfering with the analysis may be isomers or isobars of the target analyte, isotopic analogs, or adducts of the impurities present in the sample, which are isobaric to the compound of interest. The molar response of the instrument to the interference may be small compared to the target compound, but when the interfering compound is present in high concentration the interference may be significant. The necessity of assessing analytical specificity is well recognized for methods utilizing single MS detection [2,3] but is just beginning to be recognized for applications utilizing MSn detection [4–7]. To date relatively few published biomedical applications utilizing LC-MS/MS for quantitative analysis include procedures for assessing specificity of the analysis. Several terms are used with respect to the qualitative identity of an analyte, e.g., analytical specificity, confirmation, identification, etc. Various organizations have developed definitions relating to this concept. We have chosen to use the term danalytical specificityT defined by the National Committee on Clinical Laboratory Standards [8]. In the manuscript the term danalytical specificityT is used interchangeably with the term dspecificityT. In the case of MSn detection no commonly accepted rules for assessing analytical specificity in quantitative analysis have been established to date. One of the approaches for evaluation of the presence of interference in MSn is through monitoring multiple product ions of a parent ion [5–7,9–14].

This approach is not always possible because collisioninduced dissociation (CID) fragmentation of a parent ion often produce only a single dominant product ion. In the manuscript we describe several strategies that may be used for assessing analytical specificity in methods utilizing MSn detection.

Materials and methods Testosterone, methylmalonic acid (MMA), succinic acid (SA), cortisol, cortisone, prednisolone, sirolimus, formic acid, and ammonium formate were purchased from Sigma (St. Louis, MO). The internal standards were deuterated analogs of the compounds d3-testosterone, d3-MMA, and d4-cortisol purchased from Cambridge Isotope Laboratories (Andover, MA); and desmethoxy-sirolimus (used as internal standard for sirolimus) was a gift from Wyeth Pharmaceuticals (Madison, NJ). Methanol, acetonitrile, methyl-tert-butyl ether (MTBE), and phosphoric acid were all HPLC grade from Fisher Scientific. Hydrochloric acid (3 mol/L) in n-butanol was purchased from Regis Technologies, Inc. All other chemicals were of the highest purity commercially available. Sample preparation for the methods is outlined in Table 1. Samples included in the evaluation of the branching ratios distribution had analyte concentrations within the linear range of the corresponding method and no apparent interference in the chromatogram. All studies with samples from human subjects were approved by the Institutional Review Board of the University of Utah. Calculations Different strategies for assessing analytical specificity were evaluated in the analyses of MMA, the steroids cortisone and testosterone, and the immunosuppressant drug sirolimus. In practice, for evaluation of the analytical specificity of the analysis the branching ratio is calculated as a ratio of peak intensities of multiple mass transitions. For evaluation purposes in the manuscript, the branching ratios were transformed to a logarithmic scale and were calculated according to the formula 8 9 > > > > > > > > IAx < = IBx ð1Þ Rx ¼ ln ( P IAical ) > > IBical > > > > > > : nn ; where I Ax and I Bx are the peak area of the two transitions in an unknown sample, I Aical and I Bical are the peak area of two transitions in the calibration standards, and n is the number of calibration standards. The mean, median, standard deviation, standard error, kurtosis, and skewness of R x were calculated for the distribution of R x and evaluated as a predictor of the utility and the tolerance for the branching

M.M. Kushnir et al. / Clinical Biochemistry 38 (2005) 319–327

321

Table 1 Outline of the sample preparation and instrumental analysis Analyte

Internal standard

Testosterone (T)

d3-T

Methylmalonic acid (MMA)

d3-MMA

Cortisol (CL), Cortisone (CN)

d4-CL

Sirolimus (SR)

DesmethoxySR

Sample preparation

0.2 mL of serum extracted with MTBE, followed by derivatization with hydroxylamine and solid phase extraction 1 mL of serum extracted with MTBE followed by derivatization with hydrochloric acid (3 mol/L) in n-butanol 0.5 mL of urine extracted with MTBE

0.3 mL of whole blood, proteins precipitated with acetonitrile

LC column

LC conditions

Mass transitions, m/z (collision energy) Quantitative

Qualitative

Reference

Luna C18, 50  2 mm, 5 Am particles (Phenomenex). Column temperature 45 8C.

Mobile phase: 70% methanol, 30% formic acid, 5 mM, flow rate 250 AL/min

304 to 124 (40 V)

304 to 112 (40 V)

[6]

Luna C18, 30  2 mm, 5 Am particles (Phenomenex). Column temperature 40 8C.

Mobile phase: 85% methanol; 15% ammonium formate, 5 mM; flow rate 700 AL/min

231 to 119 (15 V)

231 to 175 (11 V)

[8]

Luna phenyl–hexyl, 30  2 mm, 5 um particles (Phenomenex). Column temperature 45 8C. Guard cartridge C18 (Phenomenex). Column temperature ambient.

Mobile phase: 50% methanol; 50% water, 5 mM; flow rate 300 AL/min

CL 363 to 121 (35 V), CN: 361 to 163 (35 V)

CL 363 to 97 (45 V), CN: 361 to 163 (25 V)

([13], unpublished work)

Gradient 100% water to 100% methanol.

931 to 864 (24 V)

932 to 865 (24 V)

[11]

ratios. The expected value of R x in the absence of interference is zero. Apparatus and chromatographic conditions The HPLC system consisted of an LC pump, series 1100 (Agilent); a vacuum degasser and an autosampler PE series 200 (Perkin Elmer Analytical Instruments). An API 3000 (Applied Biosystems/MDS SCIEX) tandem mass spectrometer was used in the positive ion mode with TurboIonSprayk for testosterone, MMA, and sirolimus analysis, and atmospheric pressure chemical ionization (APCI) was used for cortisone analysis. The LC column, the mobile phase composition, LC separation conditions, and mass transitions along with corresponding collision energy are outlined in Table 1. Quantitative data analysis was performed using Analystk software (Applied Biosystems/MDS SCIEX).

Evaluation of the analytical specificity One of the common applications of MSn detection is quantitative analysis of target compounds. Most mass analyzers used for quantitative analysis are scanning instruments utilizing serial detection of the mass spectra. It is common practice in target analysis to increase the effective sensitivity of detection by monitoring only a few ions. This approach increases sensitivity at the expense of specificity. In quantitative analysis with electron ionization (EI) GC-MS

selected ion monitoring mode (SIM) operation, the most commonly utilized criterion for assuring analyte identity is the monitoring of at least three fragment ions of a compound followed by evaluation of the relative intensity of the ions [2]. This approach is widely recognized as a means to minimize the risk of false-positive identification. Soft ionization techniques in combination with MSn detection generally provide more selective detection compared to an EI-MS. It is commonly assumed that the connectivity relationship between the precursor and its product ion provides greater specificity compared with SIM EI-MS detection and utilization of a single mass transition is sufficient for compound identification and quantitation. This is reflected in the fact that relatively few published methods utilizing MSn detection include assessment of the specificity as part of the analysis. However, it is not always true that MSn detection is interference-free. Quantitative performance of a method utilizing MSn detection may be affected by the ion suppression [1] or presence of isobars of target compound having the same product ions. Because of the complex and variable composition of the biological sample matrix there is often a need to demonstrate that there is no interference present in a sample and that the quantitative results of the analysis are accurate. This is particularly true for biochemical and clinical applications because interference in an analytical result may be a contributing factor to misdiagnosis of a disease [5]. The above-mentioned b3-ionQ rule [2] may be adapted to MSn detection through monitoring multiple MS/MS transi-

322

M.M. Kushnir et al. / Clinical Biochemistry 38 (2005) 319–327

tions and evaluation of the ratio of their relative intensities for the confirmation of compound identity [5,7–14]. We utilized this strategy for the analysis of testosterone in blood [9]. Because of the need to accurately measure low concentrations (10–50 pg/mL in samples from women and children), there is a potential for interference from the background noise or endogenous constituents that may be present in some samples. Monitoring two mass transitions of testosterone m/z 304 to 124 and m/z 304 to 112 and evaluating branching ratio of their intensities allows detection of samples with interference. Method evaluation with patient samples showed branching ratio outside of the F2 SD limits in ~2% of the female patient samples. Fig. 1 shows a chromatogram of two transitions of testosterone and a histogram of the distribution of the ratio of the peak areas of the transitions. Descriptive statistics for the distribution of R x are included in Table 2. A variation of the above-described approach is monitoring two transitions at different fragmentation conditions. An example of this approach is the analysis of MMA in human blood [11]. Methylmalonic acid is a marker of two severe clinical disorders: vitamin B12 deficiency and methylmalonic acidemia, an inborn error of metabolism SA, are the final product of the same metabolic pathway as MMA. SA is present in samples at 10- to 500-fold greater concentrations than MMA. In patients with vitamin B12 deficiency, the

Fig. 1. Chromatograms of the primary (a) and secondary (b) mass transitions of testosterone (human serum sample, 0.8 ng/mL) and histogram (c) of the branching ratio (R x , Eq. (1)) distribution in a group of 215 patient samples.

Table 2 Descriptive statistics for the branching ratios distribution (calculated with Eq. (1))

N Mean Median Standard error Skewness Kurtosis Standard deviation 2 SDb

Testosterone (2 transitions, the same CEa)

Methylmalonic acid (2 transitions, transitionspecific CE)

Cortisone (the same transition, CE stepping)

Sirolimus (2 transitions, the same CE, isotopic ions)

216 0.024 0.021 0.009

269 0.07 0.07 0.01

125 0.041 0.033 0.0096

64 0.021 0.018 0.012

1.253 9.56 0.129

0.9 1.78 0.11

0.577 2.92 0.108

0.741 3.78 0.095

F29%

F25%

F24%

F21%

a

Collision energy. b Acceptance limits for the branching ratio at a confidence level of 95% (the values are converted from the logarithmic to the linear scale).

metabolic pathway from MMA to SA is blocked and MMA concentration is elevated, while SA concentration decreases. To be practically useful, a test for MMA must accurately quantitate MMA in the presence of a large excess of SA. The difficulty of analyzing MMA in the presence of SA is the absence of unique mass transitions between the compounds. Mass transitions m/z 231 to 119 (primary transition, collision energy (CE) 15 V) and m/z 231 to 175 (secondary transition, CE 11 V) are characteristic for both isomers, but approximately 100 times and 30 times more abundant for MMA than for SA, respectively. The relative abundance of the transitions is significantly different between the compounds. In a high throughput analysis without chromatographic separation of MMA and SA, the difference in the relative intensity between the mass transitions serves as an indication of interference in the sample. The branching ratio m/z 175/119 for MMA was 0.35 F 0.1, whereas the ratio for the same fragments of SA was 2.0 F 0.2. When both MMA and SA are present in a sample, the branching ratio of the two product ions enables the detection of SA interference with MMA quantitation. Evaluation of the results of the analysis of 380 patient samples showed that 89.2% of the analyzed samples produced acceptable results, and 10.8% of the samples needed to be reanalyzed by a method, which chromatographically resolves MMA and SA because of unacceptable branching ratio [11]. Statistical parameters describing the distribution of the relative intensities of the transitions during the analysis of MMA in a method chromatographically resolving MMA and SA are presented in Table 2. Urinary-free cortisol excretion is commonly used in the diagnosis of Cushing’s syndrome. We identified three compounds, which may potentially interfere with the analysis, which have primary mass transition m/z 363 to 121 identical to cortisol [13]. Two of the compounds, tetrahydroprednisone and dihydroprednisolone, are metabolites of the synthetic steroid prednisone. Another com-

M.M. Kushnir et al. / Clinical Biochemistry 38 (2005) 319–327

pound with a major mass transition identical to cortisol is the triglyceride-lowering drug fenofibrate (TricorR), which is structurally unrelated to cortisol [5]. If inadequately separated, the peaks may coelute with cortisol and produce a severe falsely elevated cortisol concentration. Utilization of additional mass transition (m/z 363 to 97) characteristic to cortisol allowed identifying the samples with interference and preventing misdiagnosis of the patients. While the expected branching ratio of the transitions (m/z 363 to 97– m/z 363 to 121) for cortisol was 0.45, the observed ratios were 0.01, 0.01, and 0.02 for tetrahydroprednisone, dihydroprednisolone, and fenofibrate, respectively. All the samples from patients taking prednisone and fenofibrate were consistently identified by this approach. The methods utilizing multiple mass transitions rather than a single transition resulted in accurate quantitation in the above cases. On some occasions, CID fragmentation results in a single dominant transition and the abovementioned specificity assessment strategies are not feasible. We propose alternative approaches that may provide assessment of compound identity. Isotopic ions Most chemical elements consist of more than one naturally occurring isotope. The relative intensity of the isotopic ions compared to the monoisotopic ion depends on the composition of a molecule and its molecular weight. The relative intensities of the isotopic ions represent a property that may be used for assessing the analytical specificity of the analysis. This approach may be utilized for compounds having an intense isotopic ion, which is common for high molecular weight compounds, or molecules containing atoms with intense isotopic ions (e.g., chlorine, bromine, sulfur, etc.). To be practically useful as a confirmation transition, the isotopic molecular ion should be of comparable intensity relative to the utilized major isotope molecular ion [16]. Product ions of the monoisotopic molecular ion always contain monoisotopic atoms, whereas the product ions of the isotopic molecular ion may be either monoisotopic (same as the product ion of the monoisotopic molecular ion), or isotopic ions. Depending on the molecular weight and the structure of the fragment lost during the CID, the relative intensity of the monoisotopic and isotopic product ions originating from the parent isotopic molecular ion would be different. If the fragment lost is less than 50% of the molecular weight of the parent ion, it is more likely that the isotopic product ion would be of greater intensity compared to the monoisotopic product ion, and vice versa [16]. In cases where the fragment lost during the CID is close to 50% of the molecular weight of the parent ion, the relative intensity of the isotopic and nonisotopic product ions would typically be comparable. We evaluated the use of a transition from an isotopic parent ion for assessing specificity in the analysis of the drug sirolimus in whole blood. The ammonium adduct of

323

sirolimus has m/z 931, and the major molecular isotopic ion (M + NH4 + 1)+ has an intensity of ~35%, compared to the (M + NH4)+. The major unique product ion originating from the molecular ion is m/z 864, and the major unique product ion originating from the ion isotopic to the molecular ion m/z 932 is m/z 865. The method does not have known interferences, but because of the need to accurately measure low concentrations (above 1 ng/mL) there is a potential for interference from an unknown compound or from the background noise. Monitoring two mass transitions of sirolimus and evaluating branching ratio of their intensities allows one to identify samples with a suspected interference [15]. Descriptive statistics of the distribution of the relative intensities of the transitions are included in Table 2. The same parent/product ion transitions acquired with different collision energies An alternative to monitoring multiple unique transitions resulting from the loss of different portions of the molecule is to monitor the same transition at different fragmentation conditions. Even though other molecules may produce the same parent/product transition, it is unlikely that the ratio of intensities of the transitions obtained at different fragmentation conditions would be the same among the compounds. One strategy to find suitable fragmentation conditions for the confirmation transition is to obtain breakdown curves of the target compound and the potential impurity (Fig. 2). Fragmentation conditions producing the maximum intensity may be used as the primary transition for the target compound, and the fragmentation conditions resulting in intermediate intensity as the confirmation transition. The above approach was utilized for analysis of the endogenous steroid cortisone in urine samples (unpublished work). Interconversion between cortisol and cortisone is controlled by 11h-hydroxysteroid dehydrogenase (11hHSD) isoenzymes [14]. Accurate measurement of the cortisol and cortisone concentrations and their ratio provides useful information about the enzymes deficiency. A potential problem with cortisone analysis is interference

Fig. 2. Breakdown curves for mass transition m/z 361 to 163 of isomers prednisolone (P) and cortisone (C).

324

M.M. Kushnir et al. / Clinical Biochemistry 38 (2005) 319–327

from its isomer, the synthetic steroid prednisolone. Up to 5% of clinical samples submitted to our laboratory for analysis of cortisone come from patients treated with synthetic steroid prednisone. Because cortisone and prednisolone are positional isomers, their CID fragmentations produce the same most intense mass transition (m/z 361 to 163); and all other mass transitions present in the mass spectrum of cortisone are not specific and not suitable to use for assessing specificity. The optimal conditions for the confirmation transitions were determined from the breakdown curves of cortisone and prednisolone. Fig. 2 displays an overlay of the breakdown curves of cortisone and prednisolone. The optimal value of CE was utilized for the quantitative mass transition, and the same mass transition at the CE corresponding to half of the signal intensity on the leading edge of the breakdown curve was used for the confirmation of its identity (Table 2). If prednisone is administered to a patient, concentrations of endogenous cortisone are significantly reduced, but because the peaks are not chromatographically separated, prednisolone may be misidentified as cortisone and falsely elevate the cortisone concentration. The branching ratio for the transitions of cortisone is 0.45 F 0.15 and the branching ratio for the same transitions of prednisolone is 1.35 F 0.25. The proposed approach of monitoring the same mass transition at variable collision energies allows identification of the samples from patients taking prednisolone from the samples containing endogenous cortisone and prevents the misdiagnosis. Other strategies for assessing analytical specificity Soft ionization techniques may generate protonated or deprotonated molecular ions, and adducts of the molecular ions, e.g., Na+, K+, NH4+, etc. Formation, stability, and relative intensity of the adduct ion relative to the other molecular ion are compound specific. The existence of an adduct depends on the chemical properties of a compound, the mobile phase composition, pH, and the ionization conditions. An isobaric interference will not likely produce both the protonated/deprotonated molecular ion and the adduct in the same proportions as the target compound. This phenomenon may be used for assessment of the specificity of the analysis. With this approach specificity may be assessed utilizing a unique transition from a protonated/ deprotonated molecular ion as the parent ion for one of the transitions and the adduct as the parent ion for the other transition. The majority of compounds under soft ionization conditions produce positive ions, and fewer compounds produce negative ions. If a compound contains functional groups, which may produce both positive and negative ions, alternation of acquisition between negative and positive ion modes may be used for assessing the specificity. Other techniques potentially useful for confirmation purposes are (a) alternation of the scan type between

multiple reaction monitoring (MRM) and full scan acquisition; (b) utilization of MSn on ion trap analyzers, and (c) MS3 on a hybrid quadrupole linear ion trap (LIT) analyzer. One of the operation modes on the LIT analyzer is to generate product ions in the collision cell followed by excitation and fragmentation in the LIT [17,18]. Exact mass measurement is another approach, which may be utilized to assure the identity of a target compound. At masses below m/z 400 Da for an analyzer capable of mass accuracy b5 ppm, the difference may be sufficient to confirm a unique elemental composition of a compound. Accurate mass measurement might resolve isobars, but it would not distinguish between isomers. As the molecular weight increases, the chance of misidentifying a compound significantly increases. This method of confirmation is widely used in qualitative analysis, and there is no fundamental reason why it could not be used for assessment of analytical specificity in the quantitative analysis as well. Modern single stage or hybrid time of flight instruments are useful for improvement of the specificity via accurate mass measurement, though rarely used in quantitative analysis. Uncertainty of identity As suggested above, the assessment of specificity may be based on any property that is unique to a compound. The required degree of certainty for analytical specificity depends on the application. Thompson et al. proposed a bfitness for a purposeQ approach that relates a magnitude of uncertainty associated with the analysis with the needs of the analysis [3]. An important concern that should be addressed for the analytical method is how much selectivity or specificity can be sacrificed without unduly compromising the quality of the result. Because of the diversity of analytical applications, it is not feasible to use a single set of rules for the assessment of specificity that would fit all purposes. Two examples of applications requiring different degrees of certainty for compound identification are analysis of a reaction product in a process control of a wellcharacterized reaction system and diagnostic testing of biological samples for endogenous compounds present at low concentrations. In the first case it is already known what compounds are present in a sample; in the second case the sample matrix composition may be complex and highly variable and a positive test result may mean a severe health consequence for a person. Clearly the identification criteria that are adequate for the process control are unlikely to be adequate for analysis of clinical samples. The best strategy for assessment of analytical specificity is to take into consideration the degree of specificity required in the application. This means that in some cases monitoring a single unique transition would be sufficient for the purpose, while in other cases monitoring more than two mass transitions may be necessary. In order to enhance specificity, it is necessary to avoid mass transitions resulting from commonly lost fragments (e.g., H2O, NH3, CO).

M.M. Kushnir et al. / Clinical Biochemistry 38 (2005) 319–327

Preferred approaches for identity confirmation are utilization of multiple MS/MS transitions resulted from different fragmentation pathways at either the same or transition-dependent collision energy. The choice of the approach depends on the application and one can expect this to be analyte specific. If a compound has a single unique mass transition, the strategy may be based on a unique property of a compound, e.g., alternation between positive/ negative ion mode acquisition; mass transitions resulting from the isotopic ions; the same parent/product ion transitions at a different collision energy; utilization of an adduct of the molecular ion as the parent ion for the secondary mass transition; etc. If it is known what compound may interfere with the method, it is possible to determine instrumental conditions, which could provide a sufficient degree of distinction between the compounds based on their fragmentation pattern. In cases of interference from unknown compounds, there is a low probability that the interference could produce multiple mass transitions in the same proportions as the target compound. When monitoring multiple MS/MS transitions, evaluation of the analytical specificity should include a comparison of retention time and peak shapes between individual transitions and evaluation of the branching ratios of the peak intensities. If one of the above conditions is not satisfied, the sample may be reanalyzed by a method utilizing more extensive chromatographic separation. If a result is still unacceptable, a new sample may be requested for the analysis, and if this proves unsuccessful the sample may be reported as bunable to quantitate due to interferenceQ. CID fragmentation is not consistent between instruments and may even vary from day to day on the same instrument. Therefore, one should analyze an authentic standard with every set of samples and use the branching ratio of the MS/ MS transitions as a baseline for the evaluation of the unknown samples; or perform the quantification based on the alternative mass transitions of the target compound and the internal standard and compare the observed concentrations. One can expect the acceptance limits for the branching ratio to be method, analyte, and instrument specific, and in order to provide reliable confirmation of identity the acceptance limits for the ratio should be established for the method during development and validation. Some of the factors affecting the acceptance limits are uniqueness of the mass transitions, consistency of instrument performance, strategy used for the identity confirmation, and the concentration range. One approach for establishing the acceptance limits is to calculate the branching ratio and standard deviation of its distribution in samples during the method validation, and set the range as 2 standard deviations of the distribution, or if the distribution is not gaussian as a central 95th percentile of the range. This approach would assure that up to 95% of all samples without interference would have the branching ratio within acceptance limits. For less critical applications, the range may be established as three standard deviations of the

325

distribution, and the false rejection rate would decrease to less than 1%. On the other hand it is also possible that the standard deviation may be too large compared to the requirements of the application. In this case the analytical procedure must be either modified for acceptable performance or replaced with a more suitable method. To verify that the acceptance limits are adequate and the method performance is acceptable, statistically significant number of samples should be analyzed and the distribution of the branching ratios should be compared to the established acceptance limits. Table 2 summarizes statistics for the branching ratio distribution in biological samples for four methods utilizing the above-described approaches for the assessment of the specificity. For these methods the acceptance limits for the branching ratios were established as F2 standard deviations for the distribution observed in the results of analysis of the samples. The determined acceptance limits for the methods ranged between F21% and F29% (corresponding to F0.19 to F0.26 on the natural logarithm scale). The distance of the mean value from zero, standard deviation, skewness, and kurtosis may serve as representation of potential trends for the interference in individual mass transitions. If more than two mass transitions are used for the confirmation, the evaluation of the relative intensities of the transitions becomes more complicated because multiple ratios would need to be evaluated. In order to simplify the evaluation it is possible to use similarity indices, SI [19–23]. Wan et al. proposed a comparison of product ion spectra based on spectral contrast angles (CA), where each spectrum is represented as a vector in an N-dimensional space [23] and the similarity is expressed as an angle between the vectors. The CA is calculated as dot product between two normalized vectors, according to the formula n P ai bi 1 ffi cosh ¼ rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð2Þ n n P P 2 2 ai d bi 1

1

where h is the contrast angle, and a i and b i are the intensities of the product ions. Spectra that resemble each other have vectors that point in the same direction in space and have angles between vectors of 08. A 908 angle indicates the greatest spectral difference. Utilization of SI eliminates the restriction for the number of mass transitions employed in the analysis and simplifies the evaluation of the results by comparing single value of SI instead of multiple ratios. The precision of the acquisition of signal intensity in scanning mass analyzers depends on the dwell time used for acquiring each mass transition. As the dwell time increases, the quantitative performance of the method improves. The necessity of monitoring multiple mass transitions leads to a decrease in the time spent on acquisition of each individual transition and consequently would reduce the precision of the method. This means that there is a compromise between the degree of confidence in compound identity and the

326

M.M. Kushnir et al. / Clinical Biochemistry 38 (2005) 319–327

quantitative performance of a method. As more mass transitions are monitored, the greater the confidence in the identity of the target analyte, but the quantitative accuracy and precision could be reduced. Quantitative deconvolution Typical applications of MSn often involve analysis of samples with unique product ions where each mass transition derives from a single compound. When isobars of a target compound with identical CID fragmentation are present in a mixture it is common practice to chromatographically separate them from the target analyte. A different approach is possible for the quantitative analysis of coeluting isobars if evidence exists that only known interferences may coelute with the target analyte. Product ion mass spectra of isomers commonly have identical characteristic product ions, but if conditions can be found at which different branching ratios of mass transitions could be obtained, monitoring multiple mass transitions can be used for the quantitative analysis of unresolved chromatographic peaks [24]. When similarly fragmenting compounds are present in a mixture, the mass spectrum of the mixture is a combination of the mass spectra of pure components added in propor-

tions corresponding to their relative concentrations. By monitoring multiple characteristic mass transitions of a compound, and knowing the branching ratio of the transitions for individual components, it is possible to determine the composition of the mixture by finding the proportion of the components that would produce the observed mass spectrum. If in addition to the above data a quantitative calibration is utilized, the absolute concentration of the compounds can be calculated [24]. This approach could be used if the following requirements are satisfied: (a) no peaks of the same mass transitions other than those originating from the isobars are present under the target peak; (b) the total acquired signal is a linear combination of signals from the coeluting isomers; and (c) the branching ratio of the monitored mass transitions is significantly different among the isomers. The algorithm relies on the accuracy of measurements of the acquired signal and an assumption that the observed signal results only from the target analyte and known interference. This approach was used for the analysis of the isomers MMA and SA [11,24]. Fig. 3 displays experimental and reconstructed chromatograms representing performance of the algorithm for MMA analysis in the presence of SA in human plasma for the methods with chromatographic separation (a and c) and without chromatographic separation

Fig. 3. MRM chromatograms of a patient sample containing 0.21 Amol/L of methylmalonic acid (MMA) and 4.0 Amol/L of succinic acid (SA) (a and b are transitions m/z 231 to 119, c and d are transitions m/z 231 to 175). Chromatograms a and c are for the method utilizing chromatographic separation of MMA and SA; chromatograms b and d are for the method utilizing the deconvolution and no chromatographic separation. Solid lines (b and d) correspond to the acquired data; dotted lines are deconvoluted peaks of MMA and SA [24]. (Reproduced from [24] with permission form John Wiley and Sons, Ltd.).

M.M. Kushnir et al. / Clinical Biochemistry 38 (2005) 319–327

(b and d) of the isomers. The solid line (Figs. 3b and d) represents the total intensity of the transitions of the unresolved peaks that were acquired by the instrument. The intensity of each individual transition (dotted lines) was calculated utilizing the branching ratios determined by analyzing pure standards of MMA and SA and the formulas derived in [24]. This approach eliminates the need for chromatographic separation and may be applied for high throughput analysis of isomers and isobars.

Conclusions In summary, evaluation of analytical specificity is essential in quantitative analysis in many applications, especially in clinical diagnostic field. We discussed various strategies that may be used for the assessment of a target analyte identity in quantitative analysis in the methods utilizing MSn detection. A basic approach for the identity confirmation is through monitoring of multiple mass transitions and evaluation of their relative intensities. Some alternative approaches that we proposed are utilization of mass transition-dependent fragmentation conditions, monitoring the same mass transition at different fragmentation conditions, use of isotopic ions, alternating polarity or type of the acquisition, and exact mass measurement. The strategy used may be analyte, instrument, and application specific. Acceptance limits for the relative intensity between different transitions should be established during the method validation depending on the confirmation strategy used with the acceptance limits determined for each application based on the purpose of the analysis and the uniqueness of the mass transitions. Stringent criteria may not be necessary for assays that measure noncritical analytes, or concentration, which are outside of the decision level. In cases when sufficient history of the method utilization supports the absence of interference, the ultimate specificity may be used for quantitative analysis of similarly fragmenting isobars without chromatographic separation through the concept of deconvolution.

Acknowledgment This work was supported by ARUP Institute for Clinical and Experimental Pathology.

References [1] Annesley TM. Ion suppression in mass spectrometry. Clin Chem 2003;49:1041 – 4. [2] Sphone JA. J Assos Off Anal Chem 1987;61:1247 – 52. [3] Thompson M, Ramsey MH. Quality concepts and practices applied to sampling—an exploratory study. Analyst 1995;120(2):261 – 70. [4] Laboratory Accreditation Program. Chemistry and toxicology checklist. http://www.cap.org/apps/docs/laboratory_accreditation/checklists/ chemistry_and_toxicology_march2004.doc; 2004.

327

[5] Meikle AW, Findling J, Kushnir MM, Rockwood AL, Nelson GJ, Terry AH. Pseudocushing’s syndrome cause by interference of drug fenofibrate with HPLC assay for cortisol. J Clin Endocrinol Metab 2003;88:3521 – 4. [6] Li LYT, Campbell DA, Bennett PK, Henion J. Acceptance criteria for ultra trace HPLC-tandem mass spectrometry: quantitative and qualitative determination of sulfonylurea herbicides in soil. Anal Chem 1996;68:3397 – 404. [7] Borts DJ, Bowers LD. Direct measurement of urinary testosterone and epitestosterone conjugates using high-performance liquid chromatography/tandem mass spectrometry. J Mass Spectrom 2000;35:50 – 61. [8] Terminology and Definitions for Use in NCCLS Documents; Approved Standard. NRSCL8-A, 1998 (November);18(16). [9] Kushnir MM, Rockwood AL, Yue B, Roberts WL, Meikle AW. Testosterone analysis in serum of women by LC-MS/MS. Clin Chem 2004;50(6):A91– 92. [10] Lehotay SJ, Gates RA. Determination of qualitative and quantitative accuracy in GC-MS and GC-MS/MS for the analysis of 16 pesticides in complex extracts. Proceedings 49th ASMS conference on mass spectrometry and allied topics, Chicago, Paper WPK 200; 2001. [11] Kushnir MM, Komaromy-Hiller G, Shushan B, Urry FM, Roberts WL. Analysis of dicarboxylic acids by tandem mass spectrometry. High throughput quantitative measurement of methylmalonic acid in serum, plasma and urine. Clin Chem 2001;47(11):1993 – 2002. [12] Kushnir MM, Urry FM, Frank EL, Roberts WL, Shushan B. Analysis of catecholamines in urine by positive ion electrospray tandem mass spectrometry. Clin Chem 2002;48(2):323 – 31. [13] Kushnir MM, Rockwood AL, Nelson GJ, Terry AH, Meikle AW. Analysis of cortisol in urine by LC-MS/MS. Clin Chem 2002;49(6): 965 – 7. [14] Kushnir MM, Neilson R, Roberts WL, Rockwood AL. Cortisol and cortisone analysis in serum and plasma by atmospheric pressure photoionization tandem mass spectrometry. Clin Biochem 2004;37(5): 357 – 62. [15] Nelson GJ, Rockwood AL, Kushnir MM. Using M++1 isotopic peaks to establish ion ratios for the analysis of sirolimus in human whole blood by tandem mass spectrometry. Proceedings 50th ASMS conference on mass spectrometry and allied topics, Orlando, Paper WPK 279; 2002. [16] Rockwood AL, Kushnir MM, Nelson GJ. Dissociation of individual isotopic peaks: predicting isotopic distributions of product ions in MSn. J Am Soc Mass Spectrom 2003;14(4):311 – 22. [17] King RC, Gundersdorf R, Ferna´ndez-Metzler CL. Collection of selected reaction monitoring and full scan data on a time scale suitable for target compound quantitative analysis by liquid chromatography/tandem mass spectrometry. Rapid Commun Mass Spectrom 2003;17:2413 – 22. [18] Gambel T, Sasaki TA, Impey G, Jones EB. Multiple techniques for qualitative and quantitative data acquisition using a hybrid quadrupole-linear ion trap mass spectrometer. Proceedings 51st ASMS conference on mass spectrometry and allied topics, Nashville, Paper ThPF 107; 2004. [19] Lay JO, Gross ML, Zwinselman JJ. Nibbering NM. Field ionization and collisionally activated dissociation/charge stripping study of some [C9H10] Ions. Org Mass Spectrom 1983;1:16 – 21. [20] Drahos L, Vekey K. Quantification of isomeric differences in mass spectra. Rapid Commun Mass Spectrom 1996;10:1309 – 15. [21] Stein SE. Chemical substructure identification by mass spectral library searching. J Am Soc Mass Spectrom 1995;6:644 – 55. [22] Ramanathan R, Cao K, Cavalieri E, Gross ML. Mass spectrometric methods for distinguishing structural isomers of glutathione conjugates of estrone and estradiol. J Am Soc Mass Spectrom 1998;9:612 – 9. [23] Wan KX, Vidavsky I, Gross ML. Comparing similar spectra: from similarity index to spectral contrast angle. J Am Soc Mass Spectrom 2002;13:85 – 9. [24] Kushnir MM, Rockwood AL, Nelson GJ. Simultaneous quantitative analysis of isobars by tandem mass spectrometry from unresolved chromatographic peaks. J Mass Spectrom 2004;39:532 – 40.