1.

Introduction

Exhaled carbon monoxide $\left(\mathrm{eCO}\right)$ is controversially discussed as a volatile marker of oxidative stress and inflammation that could be measured noninvasively. $\mathrm{CO}$ is generated endogenously during heme degradation and catalyzed by the heme oxygenase enzymes.¹ Recent studies showing an activation of heme oxygenase $\left(\mathrm{HO}\right)\text{-}1$ by agents that cause oxidative stress have generated interest in the study of the $\mathrm{CO}$ level as a marker of oxidation. Furthermore, accumulating evidence from animal models suggests that elevated $\mathrm{eCO}$ levels may occur in the case of respiratory inflammations (like asthma, etc.), and also with nonpulmonary disorders such as diabetes. However, conflicting studies prevent a firm conclusion on the value of this marker as a diagnostic tool.²

It was previously reported by Paredi that $\mathrm{eCO}$ was elevated in diabetic patients and that the level of $\mathrm{eCO}$ correlated with glucose concentration in the blood.³ The authors found that $\mathrm{eCO}$ concentration was significantly increased after an oral glucose tolerance test (OGTT) by acute elevations of the blood glucose level. The authors speculated that high $\mathrm{eCO}$ levels during OGTT may have been a reflection of $\mathrm{HO}$ activation in response to the induction of the lipid peroxidation cascade. In that study, as in many other studies on $\mathrm{eCO}$ , an electrochemical sensor (Bedfont EM50 Micro Smokerlyzer) was used. According to the manufacturer, this type of $\mathrm{CO}$ sensor is not free of cross sensitivities to other compounds present in exhaled breath, e.g., hydrogen. Alternatively, laser absorption spectrometry-based $\mathrm{CO}$ sensors can be used. The application of this technique to biogenic $\mathrm{CO}$ production has been demonstrated above vascular cells^{4, 5} and to breath $\mathrm{CO}$ analysis.⁶

The aim of our present study was to investigate whether the reported increase of $\mathrm{eCO}$ levels after OGTT could be reproduced with a novel type of $\mathrm{CO}$ analyzer that was recently developed. We used a high-precision mid-infrared laser-spectroscopic methodology that was previously evaluated for breath analysis.^{7, 8} For comparison, we employed an electrochemical $\mathrm{CO}$ sensor (Bedfont Smokerlyzer Micro 4).

2.

Materials and Methods

2.2.

General Measurement Procedure

At the beginning of the measurement, all subjects had been fasting for at least 10 hours. During the whole measurement, all subjects were calm and seated. As a baseline, three sets of data were recorded from each subject. The measurement procedure is illustrated in Fig. 1 . A set of data consists of a glycemia measurement, one $\mathrm{eCO}$ measurement with an electrochemical analyzer, one $\mathrm{eCO}$ measurement with our laser spectrometer, and one measurement of breath hydrogen. Glycemia was determined by a commercially available analyzer (Accu-Chek Aviva, Roche Pharma AG). The recording of one data set took $10\min$ . After recording the baseline, the subjects drank a 75-g glucose solution (Accu-Chek Dextro OGT, Roche Pharma AG) within $2\min$ . One set of data was recorded afterwards every $10\min$ for two hours. For all breath measurements, the subjects inhaled to maximum and exhaled afterwards within $20\mathrm{s}$ . The last $30\mathrm{ml}$ of breath was used for analysis.

Fig. 1

Measurement procedure. After measuring glycemia, the $\mathrm{eCO}$ concentration with the electrochemical (EC) device, the CALOS technique, and the breath hydrogen monitor for three times as a baseline, the subjects took $75\mathrm{g}$ of glucose solution and repeated the measurement cycle for 12 times. One measurement cycle took approximately $10\min$ .

2.5.

Laser-Spectroscopic $\mathrm{eCO}$ Measurements

Cavity leak-out spectroscopy (CALOS) is an extremely sensitive laser absorption spectroscopy technique that uses a high-finesse optical cavity to achieve effective absorption path lengths of several kilometers. Figure 2 shows a schematic of the entire gas system. The gas sample was dehumidified by a Nafion tube (PermaPure, length $2\mathrm{m}$ ). The Nafion tube removed the water but did not affect the $\mathrm{CO}$ concentration, which was checked with a certified gas mixture.

Fig. 2

Schematic of the gas sampling and analyzer setup for laser-spectroscopic $\mathrm{eCO}$ analysis. The breath sample was dried and cleaned with a Nafion tube and a cooling trap before entering the absorption cell. The pressure inside the absorption cell was stabilized via a pressure control loop.

In the mid-infrared spectral region near $4.969\mathrm{\mu }\mathrm{m}$ $\left(2012{\mathrm{cm}}^{-1}\right)$ , $\mathrm{CO}$ shows a characteristic “fingerprint” absorption spectrum [Fig. 3a ], which leads to the outstanding specificity of absorption spectroscopy techniques. We recently reported the technical details of this spectroscopic setup.^{8, 9} The noise-equivalent $\mathrm{CO}$ concentration was $7\mathrm{ppb}$ with a subsecond time resolution. For calibration, a certified gas mixture of $3.45\mathrm{ppm}$ $\mathrm{CO}$ in nitrogen was used. The corresponding calibration plot is shown in Fig. 3b. The accuracy of the spectrometer derived from this calibration series was approximately 1%.

Fig. 3

(a) Calculated absorption spectra of water vapor, carbon dioxide, and carbon monoxide near the laser frequency (HITRAN2000 data). The concentrations displayed are average values of a typical exhalation after drying with a Nafion tube. (b) Calibration plot obtained by a standard addition method using a certified gas mixture of $\mathrm{CO}$ in nitrogen. (c) Raw data of a single exhalation. The exhalation starts at $t=0$ and $V=0$ , respectively. To copy the sampling procedure used with the Smokerlyzer, only the last $30\mathrm{ml}$ of the expirogram was used for the laser-spectroscopic $\mathrm{eCO}$ analysis.

The $\mathrm{CO}$ level of the expired air was recorded for two exhalations. Simultaneously, the breath flow rate, ${\mathrm{CO}}_2$ , and ${\mathrm{O}}_2$ concentrations were measured by a capnograph (Capnomac Ultima, Datex Ohmeda). Since the gas sample traveled about $6\mathrm{m}$ from the mouthpiece to the absorption cell through the NAFION tube, the cooling trap, and the flow controller, the $\mathrm{CO}$ measurement was delayed by a few seconds, which was corrected via data acquisition software (homemade, LabView 7.0 programming language). From the raw data, plots of the $\mathrm{eCO}$ concentration over the exhaled volume (expirograms) were extracted [see Fig. 3c]. The expirograms exhibited three phases. The exhalation started with phase I, where the $\mathrm{eCO}$ concentration equaled the ambient $\mathrm{CO}$ concentration. During phase II the $\mathrm{CO}$ concentration rose rapidly up to phase III. The described breathing procedure resulted in a nearly constant $\mathrm{CO}$ level during phase III. To copy the breath sampling procedure used with the Smokerlyzer, we used only the last few data points corresponding to the last $30\mathrm{ml}$ of breath for $\mathrm{CO}$ analysis.

3.

Results

The laser spectrometer we used is capable of measuring $\mathrm{eCO}$ level changes down to $7\mathrm{ppb}$ at a time resolution of $1\mathrm{s}$ . This sensitivity is two orders of magnitude better than the electrochemical device, which has a resolution of $1\mathrm{ppm}$ .

After intake of glucose, the glycemia level increased within $30\min$ by 75% and decreased to about 30% above the initial value during the following $40\min$ . The mean initial glycemia was $83\mathrm{mg}∕\mathrm{dl}$ , and the standard deviation (SD) was $9.1\mathrm{mg}∕\mathrm{dl}$ . Initial measurements spread from 74 to $96\mathrm{mg}∕\mathrm{dl}$ .

The results of the laser-spectroscopic $\mathrm{eCO}$ measurements are shown in Fig. 4a . Initial $\mathrm{eCO}$ fractions varied from $1.3\mathrm{ppm}$ to $3.8\mathrm{ppm}$ ( $\mathrm{mean}=2.4\mathrm{ppm}$ , $\mathrm{SD}=0.72\mathrm{ppm}$ ). The $\mathrm{eCO}$ level significantly decreased by $4.2\pm 1.4\%$ during the maximum increase of glycemia in the time between 20 and $60\min$ after glucose administration $(p<0.05)$ .

Fig. 4

Glycemia level and simultaneous measurements of $\mathrm{eCO}$ with (a) the CALOS analyzer, and (b) the Smokerlyzer Micro 4. The graphs show the average of six healthy volunteers. The CALOS analyzer shows a significant decrease $(p<0.05)$ of the $\mathrm{eCO}$ level during the phase of maximum glycemia $(20\min <t<50\min )$ , while the change of $\mathrm{eCO}$ measured by the smokerlyzer is not significant $(p=0.08)$ .

The results of the $\mathrm{eCO}$ measurements with the Smokerlyzer Micro device are shown in Fig. 4b. In contrast to the results obtained with the CALOS analyzer, the change in $\mathrm{eCO}$ levels measured by the electrochemical sensor after glucose administration was not significant $(p=0.08)$ . The initial $\mathrm{eCO}$ concentrations ranged from 0 to $4\mathrm{ppm}$ ; the peak concentrations did not exceed $4\mathrm{ppm}$ . The mean initial concentration was $1.6\mathrm{ppm}$ $(\mathrm{SD}=1.3\mathrm{ppm})$ .

We found that the Smokerlyzer Micro 4 exhibited a slight cross sensitivity to hydrogen. Figure 5a shows the ${\mathrm{H}}_2$ dependence of the response of both the Smokerlyzer and the laser spectrometer for different ${\mathrm{H}}_2∕\mathrm{CO}∕{\mathrm{N}}_2$ mixtures, normalized to a pure $\mathrm{CO}∕{\mathrm{N}}_2$ mixture. According to the results displayed in Fig. 5a, the response of the Smokerlyzer to hydrogen (in the range up to $500\mathrm{ppm}$ ) was nearly linear with a slope of 0.014, whereas the CALOS analyzer was inherently insensitive to hydrogen fractions in the gas sample.

Fig. 5

(a) Analysis of $\mathrm{CO}$ in various ${\mathrm{H}}_2∕\mathrm{CO}∕{\mathrm{N}}_2$ mixtures with increasing ${\mathrm{H}}_2$ fraction from 0 to approximately $500\mathrm{ppm}$ . The deviation from the theoretical $\mathrm{CO}$ value, which is given by the analysis result of the zero-hydrogen mixture, is plotted over the hydrogen concentration. (b) Change in exhaled hydrogen and glycemia during the OGTT. The graphs show the average of six healthy volunteers. The maximum increase occurs $10\min$ after glucose ingestion.

The measurements of exhaled hydrogen during the OGTT are shown in Fig. 5b. Initial ${\mathrm{H}}_2$ concentrations ranged from $3\mathrm{ppm}$ to $97\mathrm{ppm}$ $(\mathrm{mean}=23\mathrm{ppm})$ . The breath hydrogen level increased by 40% within the first $10\min$ after glucose administration $(p<0.05)$ .

4.

Discussion and Conclusion

In comparison with the electrochemical $\mathrm{eCO}$ analysis, the laser-spectroscopic $\mathrm{eCO}$ measurement is an extremely sensitive and precise method for analyzing $\mathrm{eCO}$ in human breath. The Smokerlyzer has a resolution of $1\mathrm{ppm}$ , so the systematic measurement error is $0.5\mathrm{ppm}$ . For typical $\mathrm{eCO}$ levels of about $2\mathrm{ppm}$ , this error leads to a relative uncertainty of 25%. The uncertainty of the laser-based analyzer is around 1%. Also, the laser spectrometer is highly specific to $\mathrm{CO}$ due to the use of its “fingerprint” absorption spectrum in the mid-infrared spectral region around $5\mathrm{\mu }\mathrm{m}$ . Homonuclear compounds like nitrogen, oxygen, and hydrogen cannot affect this method due to the absence of infrared absorption of such molecules.

Using an electrochemical Smokerlyzer Micro 4 for $\mathrm{eCO}$ analysis, we did not find any significant change in $\mathrm{eCO}$ after glucose ingestion. This is in opposition to the observed strong elevation of $\mathrm{eCO}$ (i.e., 50% change) after glucose ingestion that was previously reported by Paredi ³ They used a Smokerlyzer EM50 for $\mathrm{eCO}$ analysis, which is an earlier version of the device that was used in our study.

Using our laser-spectroscopic technique, we confirmed that $\mathrm{eCO}$ levels are not elevated after gluose ingestion. In contrast, we found that $\mathrm{eCO}$ levels decreased a few percent after glucose intake. Due to the lower sensitivity and precision of the electrochemical sensor, this slight decrease could not be observed with the Smokerlyzer.

What are the possible reasons for the found discrepancy? We propose that the $\mathrm{eCO}$ measurements with the Smokerlyzer EM50 device reported by Paredi may have been considerably affected by this electrochemical device’s well-known cross sensitivity to hydrogen. Hydrogen is generated by bacteria in the colon from carbohydrates that escaped digestion in the small intestine,¹⁰ but also in the small intestine itself. For example, hydrogen breath tests are used to diagnose small intestine bacterial overgrowth.¹¹ Our speculation is strengthened by our measurements of breath hydrogen after glucose ingestion. The maximum increase of exhaled hydrogen was observed during the first $10\min$ after glucose ingestion. This characteristic course is almost identical with the course of the $\mathrm{eCO}$ measurement reported by Paredi

Generally, electrochemical sensors are sensitive to hydrogen, but the Smokerlyzer Micro 4 device used in our study has been considerably improved in this regard (private communication with manufacturer). This explains why we did not reproduce the findings of Paredi A measurement series with a $\mathrm{CO}∕{\mathrm{H}}_2∕{\mathrm{N}}_2$ mixture still showed a slight cross sensitivity to hydrogen. However, this resulted in a measurement error of only 1.4% of the hydrogen concentration. For breath hydrogen concentrations of up to $100\mathrm{ppm}$ , this results in only 1 to $2\mathrm{ppm}$ offset to the $\mathrm{CO}$ measurement. According to the manufacturer, aging of the sensor might increase this cross sensitivity to hydrogen. If we assume that the crosstalk reaches $1\mathrm{ppm}$ $\mathrm{CO}$ per $10\mathrm{ppm}$ hydrogen (i.e., 10%) in an older or aged version, a hydrogen increase from 25 to $35\mathrm{ppm}$ observed after glucose administration would appear as a considerable $\mathrm{eCO}$ increase by $1\mathrm{ppm}$ , corresponding to a 50% increase of $\mathrm{eCO}$ for a baseline $\mathrm{eCO}$ value of $1\mathrm{ppm}$ .

In conclusion, if an electrochemical sensor is used for $\mathrm{eCO}$ analysis, it is essential to make sure that no other constituents of exhaled breath, especially hydrogen, interfere with the measurement. Laser-based absorption spectroscopy techniques like CALOS are excellent methods to detect $\mathrm{eCO}$ with extremely high specificity, sensitivity, and speed. Ongoing projects in our laboratory seek to develop a more rugged and compact CALOS analyzer that eventually could be used in the doctor’s office or at the bedside.