Cavity-Enhanced Near-Infrared Laser Absorption Spectrometer for the

Jun 9, 2015 - Elevated concentrations of acetonitrile have been found in the exhaled breath of patients with cystic fibrosis1 and may indicate the sev...
0 downloads 0 Views 2MB Size
Download PDF
Article pubs.acs.org/ac

Cavity-Enhanced Near-Infrared Laser Absorption Spectrometer for the Measurement of Acetonitrile in Breath Michele Gianella and Grant A. D. Ritchie* Department of Chemistry, Physical and Theoretical Chemistry Laboratory, University of Oxford, South Parks Road, Oxford, OX1 3QZ, United Kingdom S Supporting Information *

ABSTRACT: Elevated concentrations of acetonitrile have been found in the exhaled breath of patients with cystic fibrosis1 and may indicate the severity of their condition or the presence of an accompanying bacterial infection of the airways. There is therefore interest in detecting acetonitrile in exhaled breath. For this purpose, a cavity-enhanced laser absorption spectrometer (λ = 1.65 μm) with a preconcentration stage was built and is described here. The spectrometer has a limit of detection of 72 ppbv and 114 ppbv of acetonitrile in nitrogen and breath, respectively, with a measurement duration of just under 5 min. The preconcentration stage, which employs a carbon molecular sieve and an adsorption/thermal desorption cycle, can increase the acetonitrile concentration by up to a factor 93, thus, lowering the overall limit of detection to approximately 1 ppbv. The suitability of the system for acetonitrile measurements in breath is demonstrated with breath samples taken from the authors, which yielded acetonitrile concentrations of 23 ± 3 ppbv and 29 ± 3 ppbv, respectively.

M

unequivocal indicator of P. aeruginosa,15,16 several studies have found elevated HCN concentrations in the breath of CF patients.17−20 Another potential biomarker for CF or bacterial infections of the airways is acetonitrile (methyl cyanide, CH3CN, abbreviated as ACN).21 In an earlier study conducted in our group, among the 97 measured molecules (HCN was not among them), ACN was found to be the single most effective indicator of CF: acetonitrile concentrations in the breath of CF patients were on average 2.5× larger than in healthy controls.1,22 Typical ACN concentrations in the breath of healthy nonsmokers have been reported by Jordan et al. (below 15 ppbv);23 somewhat larger values (11.71 ± 11.65 ppbv) have been published by Lirk et al.24 Due to the complex chemical composition of breath and the low concentrations of most biomarkers, their detection and quantification requires a sensor that is both sensitive and selective. Infrared laser-based cavity enhanced absorption (CEA) spectroscopy satisfies these criteria. In its simplest form, a CEA spectrometer consists of a laser source, an optical cavity formed by two high-reflectivity dielectric mirrors between which the gaseous sample is held, and a photodetector. While most of the laser light directed at the first cavity mirror is lost (back-reflection), the small amount that is injected into the cavity can bounce between the mirrors thousands or even tens

edical diagnosis based on breath smell has been practiced for centuries, although only for a handful of ailments such as uncontrolled diabetes, lung abscesses, and liver failure.2 It is only after 1971, when Pauling et al.3 discovered approximately 250 substances in the breath of people that were put on a special diet, that the diagnostic potential of breath analysis truly unfolded.4 Molecules in exhaled breath that correlate with a specific disease are called biomarkers. A few examples of breath tests include those for nitric oxide (NO), which is an indicator of airway inflammation such as asthma;5−9 hydrogen (H2), which is routinely employed to diagnose sugar intolerance or bacterial overgrowth in the gastrointestinal tract;10,11 and carbon dioxide isotope ratio (13CO2/12CO2), which allows diagnosis of Helicobacter pylori infections (urea breath test).12 Nitrogen (N2) and oxygen (O2) make up about 89% of each exhaled breath (by volume), while the remaining 11% consists of water vapor (H2O, about 5%), carbon dioxide (CO2, about 5%), argon (Ar, about 1%), and several hundred compounds present in trace amounts. The exhaled concentrations of the latter vary greatly, both from person to person and for the same person throughout the day depending on metabolism, but are generally of the order of ppmv or less. Biomarkers end up in breath either after being transported to the lungs in the bloodstream and diffusing through the alveolar membrane, or by being produced within the airways. An example of the latter is the production of hydrogen cyanide (HCN) by Pseudomonas aeruginosa,13,14 an opportunistic pathogen that often plagues patients affected by cystic fibrosis (CF). Although there is evidence that HCN may not be an © 2015 American Chemical Society

Received: April 9, 2015 Accepted: June 9, 2015 Published: June 9, 2015 6881

DOI: 10.1021/acs.analchem.5b01341 Anal. Chem. 2015, 87, 6881−6889

Article

Analytical Chemistry of thousands of times, resulting in path lengths of the order of several kilometres.25 These enormous path lengths, which can be achieved with cavities whose physical length is less than 1 m, are at the core of the technique’s sensitivity. Its selectivity stems, on the one hand, from the uniqueness of the quantum mechanical rovibrational energy levels associated with each molecule, which give rise to a unique absorption spectrum, and, on the other hand, from the narrow line width (16 ppmv). The average PCF and standard deviation are 7.8 ± 0.4. By increasing the adsorption time (and sample volume) the preconcentration factor can be increased significantly: with 15 min adsorption time at 59 scm3/min the PCF increases to about 30 ± 2 (LODeff = 2.4 ppbv) and the required gas volume is about 1 L (at atmospheric pressure and room temperature); with 1 h adsorption time at 59 scm3/min the PCF reaches 93 ± 5 (LODeff = 0.8 ppbv) and about 4 L of gas are required (Table 2). The flow was kept constant at 59 scm3/min for practical reasons, but it is conceivable that it could be readily increased in order to shorten the adsorption time. Moreover, it is possible that with even larger sample volumes the PCF could be increased further, but this was not tested. It should also be pointed out that the PCF might be dependent on the initial concentration and that, therefore, significantly different values could be obtained with lower initial ACN concentrations. The maximum theoretical preconcentration factor PCFmax,th is achieved when all the ACN present in the initial sample volume is transferred into the optical cell, and is simply the ratio N0/N1, where N0 is the number of molecules

measurements. The baseline noise is the standard deviation of the absorption spectrum of a vacuum measurement. Generally, the LOD is the more interesting quantity, but the baseline noise figure characterizes the performance of the spectrometer independently of the algorithm used to extract the concentrations from the spectra. Both quantities depend on the acquisition time: short measurements are relatively immune to drifts but suffer from excessive noise, whereas long measurements are subject to various drifts (baseline, amplifier offsets, wavelength). Furthermore, the sample under investigation may not be stable over long times. The optical cavity was evacuated and 175 vacuum spectra were recorded. Each of the 175 measurements consisted of 4 × 104 sweeps at a tuning rate of 14 GHz/ms averaged together. The duration for each measurement was 4 min 51 s. We grouped consecutive measurements in groups of size n, averaged the spectra in each group, and computed the baseline noise (standard deviation) for 1 ≤ n ≤ 175, corresponding to acquisition times between 4 min 51 s and 849 min (14 h 9 min). The baseline noise as a function of acquisition T time is shown in Figure 6: since it does not exhibit the T−1/2-

Figure 6. Baseline noise as a function of acquisition time.

dependence typical for measurements dominated by white noise, we cannot provide meaningful bandwidth-normalized or acquisition-time-normalized values. The baseline noise of a single measurement (acquisition time: 4 min 51 s) is 3.5 × 10−10 cm−1, which decreases to 1.5 × 10−10 cm−1 when 12 measurements are averaged together (acquisition time: 58 min), and to 1.1 × 10−10 cm−1 when all 175 measurements are averaged together (acquisition time: 14 h 9 min). Sufficiently large samples could be measured over long times while they are kept flowing through the optical cell. Small samples, however, need to be kept in the closed cell. Mixtures containing ACN are not stable over long times, as will be shown later, and thus, acquisition times should be kept as short as possible (4 min 51 s). The ACN concentration of each of the 175 measurements was then computed with 4. Figure 7a shows the first measurement of the series (dashed black) and the average of all 175 (solid red). Concentrations for each measurement in the series are shown in Figure 7b. The expected value is zero since these are vacuum spectra, but the measured mean was μc = −12 ppbv and the standard deviation σc = 24 ppbv. With the usual 3σ convention, the LOD is 72 ppbv. The negative value of the mean is due to the small dip visible at the location of the ACN absorption feature (around 0 GHz in Figure 7a). A correction has been applied to the data shown in Figure 7b to cancel this error. 6885

DOI: 10.1021/acs.analchem.5b01341 Anal. Chem. 2015, 87, 6881−6889

Article

Analytical Chemistry

Figure 7. (a) Absorption spectra of the evacuated optical cavity: single measurement (dashed black, duration: 4 min 51 s) and average of all 175 measurements of the series (solid red, duration: 14 h 9 min). (b) Concentrations computed for each of the 175 measurements of the series. (c) Absorption spectra of 50 Torr of 112 ppbv of ACN in nitrogen: single measurement (dashed black, duration: 4 min 51 s), average of all 50 measurements of the series (solid red, duration: 4 h 3 min), and reference spectrum scaled to the average concentration of the series (dotted green). (d) Concentrations computed for each of the 50 measurements of the series.

Table 1. Key Properties of the CEA Spectrometer effective path length sample pressure optical cell volume tuning range tuning rate average sweep rate averaged sweeps measurement time baseline noise LOD

Table 2. Preconcentration Factor (PCF) and Effective Limit of Detection (LODeff) of the CEA Spectrometer with Sample Preconcentration, Depending on Available Sample Volumea

7.05−7.43 km 50 Torr, 66.7 mbar 350 cm3 60 GHz, 2 cm−1 14 GHz/ms, 0.47 cm−1/ms 138 sweeps/s 4 × 104 4 min 51 s 3.5 × 10−10 cm−1 72 ppbv ACN

PCFmax,th

LODeff (ppbv)

7.8 30 93

13 43 172

9.2 2.4 0.8

The maximum theoretical preconcentration factor PCFmax,th is given in eq 5.

in the optical cell. It is assumed that the initial sample volume V0 does not exceed the breakthrough volume. Adsorption Losses of Acetonitrile. Losses in Optical Cell. Due to the high polarity of the ACN molecule (dipole moment: 3.91 D),31 there is some concern that when working at low concentrations there might be significant losses due to adsorption effects. To quantify the extent of these losses, the optical cell was filled with a mixture containing 11.1 ppmv ACN in nitrogen at a total pressure of 50 Torr and then sealed off. A total of 100 spectra were measured consecutively over 4 h 34 min starting immediately after the filling. In Figure 9 the concentrations for all 100 measurements are shown as a function of time since filling. The concentration decreases nearly exponentially with a time constant of 135 min. The inset shows that the relative change in concentration from its initial value in the first 15 min is approximately −9%. A long acquisition time will yield a time-averaged concentration value which is smaller than the initial value, and the discrepancy increases with time. It is therefore important to either measure while the sample is kept flowing (only possible with sufficiently large samples), or to start the measurement immediately after the filling and keep the measurement time below 5 min to prevent errors larger than 5%.

in the initial sample, and N1 is the number of molecules in the preconcentrated sample. With the ideal gas law, we obtain p V0 N0 V = 0 ≈ 43 0 N1 p1 V1 1L

PCF

0.3 1 4 a

Figure 8. Preconcentration factor and initial ACN concentration for 11 adsorption/thermal desorption runs.

PCFmax,th ≡

sample volume V0 (L)

(5)

where V0 is the initial sample volume, V1 = 0.35 L is the volume of the optical cell, p0 = 760 Torr is the initial sample pressure, and p1 = 50 Torr is the pressure of the preconcentrated sample 6886

DOI: 10.1021/acs.analchem.5b01341 Anal. Chem. 2015, 87, 6881−6889

Article

Analytical Chemistry

determined with pure nitrogen earlier. Although this is an increase of more than 50%, the resulting LOD is still satisfactory. Real Breath Samples. Finally, we present two spectra of breath samples taken from the two healthy, nonsmoking authors. Both subjects exhaled into a 1.5 L breath sample bag (Fischer Analysen Instrumente F201-VP-05c) through a disposable mouthpiece. The exhalation maneuver was not standardized. The contents of the breath sample bags were allowed to flow through the TDT for 15 min at a flow of 59 scm3/min and subsequently desorbed and measured. The absorption spectra are shown in Figure 11a,b. Carbon dioxide and water vapor absorption lines are easily identified through comparison with Figure 5. In Figure 11c,d, the ordinate axes have been expanded to magnify the ACN absorption feature. The reference spectra scaled to the respective computed ACN concentrations are overlaid. The ACN concentrations in breath are obtained by dividing through the PCF, assumed to be 30 (see Table 2), and are 23 ± 3 and 29 ± 3 ppbv, respectively. The uncertainties taken into account are the repeatability of the PCF and the uncertainty of the concentration of the reference sample. However, as mentioned earlier, the PCF could be significantly larger due to the very small concentration of ACN in breath compared with the concentrations at which the PCF was measured (1.5−3 ppmv). With the maximum theoretical PCF (PCFmax,th = 43, see Table 2) the ACN concentrations would be 16 ± 2 and 20 ± 2 ppbv, respectively, and in line with what has previously been reported.23,24 These proof of principle experiments clearly demonstrate the ability to measure ACN in the breath of healthy subjects and will find ready application.

Figure 9. Concentration and relative change from the initial value of the 100 measurements taken over 4 h 34 min.

Losses in Breath Sample Bags. The preconcentration stage requires breath samples to be temporarily stored in some type of airtight container. A popular choice are bags lined with a chemically inert material, and with low permeability and adsorption with respect to the molecules of interest.32−35 Nevertheless, losses through adsorption and permeation through the walls cannot be ignored if the sample is stored in the bag over longer periods. In Figure 10 we show ACN

■

CONCLUSIONS We have presented a near-infrared laser-based cavity-enhanced absorption spectrometer for the measurement of acetonitrile (ACN) in breath. To obtain concentrations from measured absorption spectra, a simple linear model requiring the spectrum of a reference sample at a known concentration was employed. The limit of detection (LOD) for ACN in nitrogen was found to be 72 ppbv (4 min 51 s measurement time), while in breath the value increased to 114 ppbv due to interference with carbon dioxide. With a preconcentration stage based on a carbon molecular sieve, the concentration of ACN could be increased by up to a factor 93 (with a sample volume of 4 L), thus, lowering the effective LOD to approximately 1 ppbv. We have also shown that losses occur when a mixture containing ACN is kept in the closed optical cell, but that such losses are negligible when the mixture is stored in a breath sample bag. Finally, we have presented spectra of real breath samples taken from the two authors, in which 23 ± 3 and 29 ± 3 ppbv of ACN were detected; these are likely to be upper limits. A number of improvements can be suggested to enhance the performance of the system. The thermal desorption procedure should be fully automated to stabilize the PCF of the preconcentration stage. This would include controlling the heating and the gas flows with a computer program, so that every desorption cycle would be carried out exactly the same. Different sorbents with lower affinity for carbon dioxide could be tested in order to reduce the amount of CO2 present in the measured samples. Alternatively, carbon dioxide scrubbing prior to preconcentration could be employed, although that may also affect ACN. Because of adsorption of ACN to the walls of the optical cell and tubing, long evacuation times are required to prevent cross-contamination of consecutively measured sam-

Figure 10. Preconcentrated samples of 102 ppbv of ACN in nitrogen: drawn directly from the gas bottle (black) or drawn from a breath sample bag (light gray) after a variable amount of time.

concentrations measured after preconcentrating samples of 102 ppbv of ACN in nitrogen taken directly from the bottle or from a breath sample bag (Fischer Analysen Instrumente F201-VP05c) in which the mixture had been kept for some time. The times in brackets represent the time elapsed between the filling of the bag and the drawing of the sample. There are no appreciable ACN losses if the samples are not stored longer than about 3 h and possibly even longer. The fluctuation of the ACN concentration can be attributed to the uncertainty of the preconcentration factor. Interference with Carbon Dioxide. Due to the partial overlap of the ACN absorption feature with neighboring carbon dioxide lines (see Figure 5), some reduction of the LOD and/ or bias in the computed ACN concentration should be expected. We measured the absorption spectrum of a mixture of carbon dioxide and synthetic air without any ACN 20 times. The CO2 concentration was in the several percent range, but was not measured exactly. The ACN concentration was computed with 4 by only considering wavelengths which fall into one of the four overlap-free regions (Figure 5). The measured mean was 36 ppbv and the standard deviation 38 ppbv. This leads to a LOD of 114 ppbv versus 72 ppbv 6887

DOI: 10.1021/acs.analchem.5b01341 Anal. Chem. 2015, 87, 6881−6889

Article

Analytical Chemistry

Figure 11. (a, b) Absorption spectra of the preconcentrated breath samples (PCF = 30) of the two authors. (c, d) Expanded view of the ACN absorption feature with reference spectrum scaled to the computed ACN concentrations. The concentrations in parentheses correspond to the values in breath (prior to preconcentration).

■

ples. Evacuation could be hastened by heating all parts that come in contact with ACN with resistive wire or similar. Water vapor and carbon dioxide will always be a concern, no matter at what wavelength the spectrometer operates. Water vapor is probably the lesser of two evils as it can be removed by freezing. By shifting the laser wavelength to 4400 (2.3 μm) or to 3000 cm−1 (3.3 μm), one could get rid of carbon dioxide interference in return for increased water vapor interference, while also taking advantage of the significantly larger absorption cross sections of ACN. Since optics and especially photodetectors are of inferior quality at these wavelengths compared to the near-infrared, the benefit of the increased absorption cross-section would have to be verified experimentally. Without an extensive clinical trial it is difficult to say whether ACN alone can serve as a biomarker for CF or for those bacterial infections of the airways that often accompany the disease. The same doubts apply to hydrogen cyanide,16,17 and perhaps the answer lies in the combined measurement of ACN and HCN. To be carried out with a single laser source and a single set of cavity mirrors, such measurements would necessarily have to take place in a spectral region of concurrent ACN and HCN absorption. Stamyr et al.36 have shown that it is possible to measure HCN in breath near 1.57 μm; such a wavelength could easily be multiplexed with our wavelength of 1.65 μm, yielding a spectrometer that could measure HCN and ACN simultaneously in breath.

■

AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS The authors thank the European Commission for funding a Marie Curie Intra-European Fellowship (MG).

■

REFERENCES

(1) Bennett, L.; Ciaffoni, L.; Denzer, W.; Hancock, G.; Lunn, A. D.; Peverall, R.; Praun, S.; Ritchie, G. A. D. J. Breath Res. 2009, 3, 046002. (2) Phillips, M. Sci. Am. 1992, July, 74−79. (3) Pauling, L.; Robinson, A. B.; Teranishi, R.; Cary, P. Proc. Natl. Acad. Sci. U.S.A. 1971, 68, 2374−2376. (4) Manolis, A. Clin. Chem. 1983, 29, 5−15. (5) Kosterev, A. A.; Malinovsky, A. L.; Tittel, F. K.; Gmachl, C.; Capasso, F.; Sivco, D. L.; Baillargeon, J. N.; Hutchinson, A. L.; Cho, A. Y. Appl. Opt. 2001, 40, 5522−5529. (6) Wojtas, J.; Bielecki, Z.; Stacewicz, T.; Mikolajczyk, J.; Nowakowski, M. Opto-Electronics Rev. 2011, 20, 26−39. (7) Crader, K. M.; Repine, J. J. D.; Repine, J. E. J. Pulm. Respir. Med. 2012, 2, 1000111. (8) Risby, T. H. J. Breath Res. 2008, 2, 30302. (9) Kharitonov, S.; Alving, K.; Barnes, P. J. Eur. Respir. J. 1997, 10, 1683−1693. (10) Levitt, M. D. N. Engl. J. Med. 1969, 281, 122−127. (11) Simrén, M.; Stotzer, P.-O. Gut 2006, 55, 297−303. (12) Peng, N. J.; Lai, K. H.; Liu, R. S.; Lee, S. C.; Tsay, D. G.; Lo, C. C.; Tseng, H. H.; Huang, W. K.; Lo, G. H.; Hsu, P. I. Dig. Dis. Sci. 2001, 46, 1772−1778. (13) Goldfarb, W. B.; Margraf, H. Ann. Surg. 1967, 165, 104−110. (14) Neerincx, A. H.; Mandon, J.; van Ingen, J.; Arslanov, D. D.; Mouton, J. W.; Harren, F. J. M.; Merkus, P. J. F. M.; Cristescu, S. M. J. Breath Res. 2015, 9, 027102. (15) Stutz, M. D.; Gangell, C. L.; Berry, L. J.; Garratt, L. W.; Sheil, B.; Sly, P. D. Eur. Respir. J. 2011, 37, 553−558. (16) Dummer, J.; Storer, M.; Sturney, S.; Scott-Thomas, A.; Chambers, S.; Swanney, M.; Epton, M. J. Breath Res. 2013, 7, 017105.

ASSOCIATED CONTENT

S Supporting Information *

Protocols and gas flow paths employed for adsorption, thermal desorption, and conditioning. Derivation of 4. Obtaining peak absorption coefficient values from line strengths of water and carbon dioxide absorption lines. Acetonitrile absorption spectra in the regions around 6000, 4400, and 2500 cm−1. Verification of effective path length with carbon dioxide absorption line. The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.analchem.5b01341. 6888

DOI: 10.1021/acs.analchem.5b01341 Anal. Chem. 2015, 87, 6881−6889

Article

Analytical Chemistry (17) Enderby, B.; Smith, D.; Carroll, W.; Lenney, W. Pediatr. Pulmonol. 2009, 44, 142−147. (18) Lenney, W.; Gilchrist, F. J. Eur. Respir. J. 2011, 37, 482−483. (19) Gilchrist, F. J.; Bright-Thomas, R. J.; Jones, A. M.; Smith, D.; Spanel, P.; Webb, A. K.; Lenney, W. J. Breath Res. 2013, 7, 026010. (20) Sohrabi, M.; Zhang, L.; Zhang, K.; Ahmetagic, A.; Wei, M. Q. Clin. Microbiol. 2014, 3, 1000151. (21) Carroll, W.; Lenney, W.; Wang, T.; Spanel, P.; Alcock, A.; Smith, D. Pediatr. Pulmonol. 2005, 39, 452−456. (22) Ciaffoni, L. Laser Spectroscopy for the Detection of Volatile Sulfur-Containing Compounds in Breath. Ph.D. Thesis, University of Oxford, Department of Chemistry, 2010. (23) Jordan, A.; Hansel, A.; Holzinger, R.; Lindinger, W. Int. J. Mass Spectrom. Ion Process. 1995, 148, L1−L3. (24) Lirk, P.; Bodrogi, F.; Deibl, M.; Kähler, C. M.; Colvin, J.; Moser, B.; Pinggera, G.; Raifer, H.; Rieder, J.; Schobersberger, W. Wien. Klin. Wochenschr. 2004, 116, 21−25. (25) Mazurenka, M.; Orr-Ewing, A. J.; Peverall, R.; Ritchie, G. A. D. Annu. Rep. Prog. Chem. 2005, 101, 100−142. (26) Woolfenden, E. A., McClenny, W. A. Compendium Method TO17. Determination of Volatile Organic Compounds in Ambient Air Using Active Sampling Onto Sorbent Tubes, U.S. EPA Report EPA/625/R-96/ 010b, 2nd ed.; EPA: Cincinnati, OH, U.S.A., 1999; p 53. (27) Paul, J. B.; Lapson, L.; Anderson, J. G. Appl. Opt. 2001, 40, 4904−4910. (28) Rothman, L.; et al. J. Quant. Spectrosc. Radiat. Transfer 2013, 130, 4−50. (29) Sharpe, S. W.; Johnson, T. J.; Sams, R. L.; Chu, P. M.; Rhoderick, G. C.; Johnson, P. A. Appl. Spectrosc. 2004, 58, 1452−1461. (30) Parameswaran, K. R.; Rosen, D. I.; Allen, M. G.; Ganz, A. M.; Risby, T. H. Appl. Opt. 2009, 48, B73−B79. (31) Alston Steiner, P.; Gordy, W. J. Mol. Spectrosc. 1966, 21, 291− 301. (32) Steeghs, M. M. L.; Cristescu, S. M.; Harren, F. J. M. Physiol. Meas. 2007, 28, 73−84. (33) Steeghs, M. M. L.; Cristescu, S. M.; Munnik, P.; Zanen, P.; Harren, F. J. M. Physiol. Meas. 2007, 28, 503−514. (34) Hakim, M.; Billan, S.; Tisch, U.; Peng, G.; Dvrokind, I.; Marom, O.; Abdah-Bortnyak, R.; Kuten, A.; Haick, H. Br. J. Cancer 2011, 104, 1649−1655. (35) Ma, W.; Gao, P.; Fan, J.; Hashi, Y.; Chen, Z. Biomed. Chromatogr. 2015, 29, 961−965. (36) Stamyr, K.; Vaittinen, O.; Jaakola, J.; Guss, J.; Metsala, M.; Johanson, G.; Halonen, L. Biomarkers 2009, 14, 285−291.

6889

DOI: 10.1021/acs.analchem.5b01341 Anal. Chem. 2015, 87, 6881−6889