2.1.

Gas Sensors

The QCL-based NO sensor uses wavelength modulation spectroscopy technique (WMS) and consists of a continuous-wave distributed feedback quantum cascade laser (CW-DFB QCL) operating in the wavelength region of 5.2 µm ($1891to1908{\mathrm{cm}}^{-1}$), a multi-pass cell, and a room temperature detector.²⁴ The laser beam is sent through an astigmatic Herriott absorption cell of 400 ml (AMAC-76, Aerodyne Research, USA), offering a total optical path length of 76 m and which consists of two concave mirrors where light is reflected multiple times, enhancing the effective path length through the breath sample. The absorbed amount of light at the output of the cell is proportional to the NO concentration present in the cell. The detection is performed by a photovoltaic detector (PV-6, Vigo Systems) working at room-temperature, thus eliminating the need for a highly sensitive liquid nitrogen-cooled IR detector, simplifying daily use of the system and allowing long-term automated operation. In order to improve sensitivity toward the NO target limit of $\le 1\mathrm{pbbv}$ required for breath analysis, WMS is implemented by modulating the injection current of the laser at a frequency of 100 kHz. The absorption signal is transposed into a frequency domain where noise sources are weaker, and absorption spectra is obtained by demodulating the detected signal at the second harmonic (2f) using a lock-in amplifier (model SR844, Stanford Research Systems) with the time constant set to 100 μs. This approach allows a sensitivity of 0.5 ppbv within 1 s averaging. To reach a fast response of the system, a pressure of $70\mathrm{cm}{\mathrm{H}}_2\mathrm{O}$ is maintained in the multi-pass cell and a sample flow rate of $900\mathrm{ml}/\min$ is generated. Therefore, the cell is refreshed in less than 2 s, making the sensor suitable for online monitoring of exhaled breath. A schematic arrangement of the experimental setup is shown in Fig. 1. The QCL-based sensor was calibrated with a reference mixture of $100\pm 3\mathrm{ppbv}$ of NO in ${\mathrm{N}}_2$ (prepared by VSL-National Dutch Metrology Institute) before analysis of a breath sample. A ${\mathrm{N}}_2$ gas bottle was used as a NO-free gas reference.

Fig. 1

Schematic of the QCL-sensor. The QCL beam at 5.2 µm is sent into a 30 -cm multi-pass cell, thus enhancing the absorption path length to 76 m. Wavelength modulation is performed by modulating the laser current at 100 kHz, and a 2f lock-in amplifier is used to demodulate the signal.

In this study, the performances of the QCL-sensor were investigated. To do that, 11 gas mixtures were prepared from our reference gas mixture (100 ppbv NO in ${\mathrm{N}}_2$) diluted in pure ${\mathrm{N}}_2$ to cover the range 10 to 100 ppbv, and 14 gas mixtures were made from a $20\mathrm{ppm}\pm 2\%$ bottle of NO for concentrations up to 4 ppm. The different concentrations of NO were produced by using two mass flow controllers (Brooks Instrument, max flow: $25\mathrm{l}/\mathrm{h}$ and $5\mathrm{l}/\mathrm{h}$ with an accuracy of $\pm 1\%$).

Contrary to the QCL sensor which directly measures the NO concentration, the chemiluminescence system is based on the reaction between NO and ${\mathrm{O}}_3$, which generates ${\mathrm{NO}}_2$ in the exited form.³⁴ When ${\mathrm{NO}}_2$ returns to a stable state, light is emitted. The amount of light, which is proportional to the amount of NO, is measured by a photomultiplier. Calibration of the Sievers NOA 280 was performed each day prior to use with the same mixture of NO as with the QCL-based sensor, i.e. with ${\mathrm{N}}_2$ gas bottle and $100\pm 3\mathrm{ppbv}$ of NO in ${\mathrm{N}}_2$. The sample flow into the Sievers is $200\mathrm{ml}/\min$. The chemiluminescence analyzer is connected to a computer with the NO analysis software, which provides a graphical display of ${\mathrm{F}}_{\mathrm{E}}\mathrm{NO}$ concentration during the analysis process.

Despite the chemiluminescence NO sensor being the gold standard reference for NO measurements in breath, its use for routine clinical practice is limited by its size and expense. Portable hand-held, relatively inexpensive NO analysers based on electrochemical analysis have been introduced on the market a few years ago. The NIOX MINO (Aerocrine AB, Sweden) Asthma Inflammation Monitor is one of the available sensors. This sensor is ideally suited for use in primary care, where the majority of asthma patients are managed. In this study, the accuracy of NIOX MINO measurements was assessed. The manufacturer stateed an accuracy of $\pm 5\mathrm{ppbv}$ of measured value below 50 ppbv and $\pm 10\%$ at or above 50 ppbv.

2.2.

BreathCollection

The QCL-based sensor was used for ${\mathrm{F}}_{\mathrm{E}}\mathrm{NO}$ monitoring with both online and offline breath-collection.

To perform online measurements, a commercially available breath sampler (Loccioni, Italy) was used to monitor the pressure of the exhaled breath within an acceptable range and to measure the breath ${\mathrm{CO}}_2$ concentration level. The breath sampler meets the American Thoracic Guidelines (ATS) for collecting breath by providing a back pressure of $10\mathrm{cm}{\mathrm{H}}_2\mathrm{O}$, to ensure soft palate closure and prevent nasal contamination, and allowing the patient to maintain constant exhalation flow. The exhalation flow ranged from 15 to $250\mathrm{ml}/\mathrm{s}$. The target flow rate was maintained within 5%.

During the sampling, the collected breath entered the Loccioni breath sampler. The CO2 concentration profile andairway pressure were simultaneously displayed in graphical forms on the sampler display. When the ${\mathrm{CO}}_2$ concentration reached 3%, a part of the breath was sent to the NO sensor. The 3% value has been chosen since it is the start of the ${\mathrm{CO}}_2$ plateau for most of people. The ${\mathrm{F}}_{\mathrm{E}}\mathrm{NO}$ level in ppbv was acquired and plotted in real-time on a computer screen. In addition, the breath sampling pipe was heated to $\ge 38°\mathrm{C}$ to prevent water condensation. A typical recorded signal at flow $50\mathrm{ml}/\mathrm{s}$ is shown in Fig. 2. ${\mathrm{F}}_{\mathrm{E}}\mathrm{NO}$ level was sampled after the estimated start of the ${\mathrm{CO}}_2$ plateau region. The end-tidal of the ${\mathrm{CO}}_2$ exhalation trend determines the point where the NO plateau is measured, i.e. 5 s before the end-tidal ${\mathrm{CO}}_2$ point. One breath sample was collected at each flow rate (15, 50, 100 and $250\mathrm{ml}/\mathrm{s}$) from each patient.

Fig. 2

Recorded data during online sampling of a single exhalation at $50\mathrm{ml}/\mathrm{s}$. During the exhalation process the mouth pressure is monitored to keep a constant flow rate (a). The exhaled breath is sent to the QCL-based sensor when the ${\mathrm{CO}}_2$ concentration reaches 3% (b). The average of the last 5 seconds of the exhaled breath was used to calculate the ${\mathrm{F}}_{\mathrm{E}}\mathrm{NO}$ value (c).

Offline collections were also performed as samples can be collected at a distant collection site from the NO sensor. A custom-built breath-collection device was used to collect single breaths into bags with the subject exhaling at specific constant flow rates.³⁵ The breath-collection device was based on the guidelines of the ATS for the sampling of exhaled NO; it is very simple, inexpensive, and has been tested in previous studies. The exhaled breath line consists of a mouth-piece connected to a discard bag (400 ml) and an NO-impermeable aluminum-foil air bag of 500 ml capacity (Mylar balloon, ABC ballonnen, Zeist, The Netherlands).³⁶ To maintain a constant exhalation flow, the mouth pressure was monitored by the patient during the sampling process. Breath was collected at various constant flow rates from each subject by changing the resistance of the breath line to maintain a mouth pressure of $10\mathrm{cm}{\mathrm{H}}_2\mathrm{O}$.

To assess the agreement of the two sampling procedures, offline ${\mathrm{F}}_{\mathrm{E}}\mathrm{NO}$ measurements in comparison to online measurements, 23 healthy individuals performed online and offline single breath maneuvers at a flow of $50\mathrm{ml}/\mathrm{s}$. To prevent systematic errors, 11 persons did first online followed by offline breath maneuvers, and 12 in the reverse order.

2.3.

Extended NO Analysis

Different exhalation flow rates can be performed with both online and offline exhaled breath measurements by changing the resistance of the sampling line and keeping a back pressure of $10\mathrm{cm}{\mathrm{H}}_2\mathrm{O}$ during the exhalation. Measuring the NO plateau for multiple flow rates allows an estimation of the NO extended parameters, which may provide additional useful clinical information. The simple two-compartment model¹² describes exhaled NO arising from two compartments: the airways and the alveolar region. It is based on three flow-independent exchange parameters, one describing the steady-state NO alveolar concentration (${\mathrm{C}}_{\mathrm{A}}\mathrm{NO}$, ppbv), and two describing the airway region (the airway NO diffusing capacity (${\mathrm{D}}_{\mathrm{aw}}\mathrm{NO}$, $\mathrm{pl}·{\mathrm{s}}^{-1}·{\mathrm{ppbv}}^{-1}$) and the airway wall NO concentration (${\mathrm{C}}_{\mathrm{aw}}\mathrm{NO}$, ppbv).¹¹ The potential of these parameters lies in their ability to split exhaled NO into two important anatomic subdivisions of the lungs and also to provide both structural and metabolic information relevant to the NO pathways. To determine the airway and alveolar contribution to exhaled NO, multiple exhalation flow rates must be accomplished. Different approaches are described in the literature¹¹ and recently used in combination with a QCL-based sensor to determine extended NO parameters in chronic obstructive pulmonary disease.³⁷ We selected the method described by Högman et al.^13,38 as it requires only three exhalation flows. This method is based on a non-linear model with a quality control to notify erroneous NO values. NO values from different flow rates are included into a software program with a second order algorithm, which will render values of ${\mathrm{C}}_{\mathrm{A}}\mathrm{NO}$, ${\mathrm{C}}_{\mathrm{aw}}\mathrm{NO}$ and ${\mathrm{D}}_{\mathrm{aw}}\mathrm{NO}$.

We measured ${\mathrm{F}}_{\mathrm{E}}\mathrm{NO}$ at different flow rates controlled by the Loccioni sampler: a low exhalation flow ($15\mathrm{ml}/\mathrm{s}$), a medium flow ($100\mathrm{ml}/\mathrm{s}$), and a high flow rate ($250\mathrm{ml}/\mathrm{s}$). To validate the sampling system, the exhalation flow of $50\mathrm{ml}/\mathrm{s}$ is also measured (${\mathrm{F}}_{\mathrm{E}}{\mathrm{NO}}_{0.05}$) and compared to calculated value by the model. A group of 20 subjects performed online single breath maneuvers at four expiratory flows as mentioned previously. After each maneuver, the subject was given a 1-minute pause before performing the next exhalation at a different flow.

3.1.

QCL-Based Sensor Performance

Prior to the study on patients, the performance of our QCL-sensor was evaluated. With the prepared NO concentrations, the QCL-sensor was tested in terms of linearity, precision, accuracy, sensitivity, and selectivity. To minimize random error, the concentrations were measured six times. One measurement consists of flushing the absorption cell for 5 min with the gas and by taking the last value displayed by our sensor. Table 1 summarizes the main characteristics of the sensor. The linearity response of the system, Fig. 3, is given by the coefficient of determination ($R^2=0.998$). Precision of the sensor was evaluated in conditions of their repeatability, and accuracy was determined by comparing the mixtures and the measured concentration. The calculated accuracy was between 99 and 104% and is limited by the accuracy of the mass flow controllers used for the dilution. The precision of the system expressed as RSD (relative standard deviation) was below 5%.

Table 1

Main characteristics of the QCL-sensor.

Fig. 3

Linearity of the QCL sensor. Measured NO concentrations versus prepared dilutions from a standard calibration mixture of $100\pm 3\mathrm{ppbv}$ of NO in ${\mathrm{N}}_2$ are plotted ($R^2=0.9998$).

The best minimum detectable NO concentration achieved with QCL-based sensor with a 1 s averaging time is 0.5 ppbv. A better sensitivity can be reached by increasing the integration time. An Allan variance analysis of the data, which presents the variance of the data as a function of integration time of the sampling, is shown in Fig. 4. It displays the reduction in noise level as the integration time is increased.

Fig. 4

Allan variance showing the QCL-based sensor detection limit. The sensitivity is 0.5 ppbv of NO with an acquisition time of 1 second. A minimum detection limit of 0.24 ppbv is reached with 26 s integration time.

As exhaled breath carries a number of compounds at high concentration such as water and ${\mathrm{CO}}_2$, the selectivity of the sensor is important. The emission spectrum of the QCL allows accessing of the NO absorption line within the fundamental vibration band without interference of water and ${\mathrm{CO}}_2$, as shown in the HITRAN simulation.³⁹ The response of the QCL sensor to humid air and 5% ${\mathrm{CO}}_2$ has been experimentally investigated and has provided good agreement with the HITRAN calculations.

3.2.

Offline/Online Comparison

The QCL-based sensor can perform ${\mathrm{F}}_{\mathrm{E}}\mathrm{NO}$ measurements online or offline. If the advantages of online analysis are obvious, the benefits of offline analysis are apparent when analyzing large sample numbers, for example during clinical studies. To examine the diagnostic capability of our sampling procedure, online and offline measurements in the same subjects are compared, using the Bland-Altman method.⁴⁰ The Bland-Altman plot, Fig. 5, consists of the difference between a pair of measurements (in this case offline–online) versus their average. On average, the difference in NO concentration between the two sampling methods (offline-online) was $-0.1\mathrm{ppbv}$ and the limit of agreements was $\pm 1.0\mathrm{ppbv}$. By measuring the NO concentration in bags within 24 h of collection, no significant differences or combined effects of measurement technique or flow rate were noted. The sample is reportedly stable for up to 24 h, but it was observed that NO concentrations in bags decreased over time up to 7% after 48 h (data not shown).

Fig. 5

Bland-Altman plot comparing 23 offline and online sampling procedures. Solid line represents the mean difference between values obtained using the two sampling methods ($-0.1\mathrm{ppbv}$). Dashed lines represent the limits of agreement $-2.1\mathrm{ppbv}$ and 1.9 ppbv; SD is standard deviation. The plot indicates the sampling procedures are in good agreement.

3.3.

Validation of QCL-Based Sensor for Multiple Flows Analysis

For each individual, the three independent parameters are calculated. Figure 6 shows the data as medians with lower and upper quartiles (25 to 75%). The level of agreement between the measured and calculated ${\mathrm{F}}_{\mathrm{E}}{\mathrm{NO}}_{0.05}$ is determined by the coefficient of determination $R^2$ (average between calculated and measured: $-0.4\mathrm{ppbv}$, $\mathrm{SD}=0.72$, $R^2=0.9997$). Data are also compared to previous results, and these results are in agreement with values presented by Högman et al.⁴¹ Regardless of the lack of agreement about what flow rates to use to calculate extended NO parameters, those results demonstrate the validity of our sampling procedure and confirm the efficiency of the theoretical model to give values which are consistent.

Fig. 6

Calculated NO extended parameters characterizing the alveolar region (${\mathrm{C}}_{\mathrm{A}}\mathrm{NO}$), airway wall (${\mathrm{C}}_{\mathrm{aw}}\mathrm{NO}$), diffusing capacity (${\mathrm{D}}_{\mathrm{aw}}\mathrm{NO}$), and flux (${\mathrm{J}}_{\mathrm{aw}}\mathrm{NO}$) NO exchanges using the model proposed by. Ref. 13 Minimum, median, maximum and, interquartile ranges (25%, 75%) are displayed. The data from the QCL-based sensor are included in the ranges suggested by Högmann and et al.⁴¹ for healthy subjects (hatched boxes).

3.4.

Comparison of the QCL-Based Sensor with Chemiluminescence and Electrochemical Sensors

${\mathrm{F}}_{\mathrm{E}}\mathrm{NO}$ from 22 asthmatic children (age range: 6 to 16 years) was measured with three sensors. They provided offline samples in aluminum bags for the QCL-based sensor and the chemiluminescence, respectively, at flow $50\mathrm{ml}/\mathrm{s}$, followed by online ${\mathrm{F}}_{\mathrm{E}}\mathrm{NO}$ sample measured by the NIOX MINO two minutes later. Ambient NO varied between 1 and 4 ppbv over the study. The bags were measured simultaneously with the QCL-based and the chemiluminescent sensors on the same day as the sampling. Measurements were achieved with a 1 s integration time for both devices. The Bland-Altman plot, Fig. 7(a), shows a high degree of agreement between the QCL-based sensor and the chemiluminescence device with an average difference of $-0.1\mathrm{ppbv}$ and a standard deviation of 1.1 ppbv.

Fig. 7

Bland-Altman plot comparing the QCL-based sensor with chemiluminescence device (a) and with NIOX MINO (b), respectively. (a) Solid line represents the mean difference between values obtained using the QCL-based sensor and chemiluminescence ($-0.1\mathrm{ppbv}$). Dashed lines represent the limits of agreement, $-2.3\mathrm{ppbv}$ and 2.1 ppbv. The plot suggests the two sensors are in good agreement. (b) Solid line represents the mean difference between values obtained using the QCL-based sensor and NIOX MINO (1.6 ppbv). Dashed lines represent the limits of agreement, $-1.8\mathrm{ppbv}$ and 5 ppbv; SD is standard deviation. The plot suggests the QCL-sensor and NIOX MINO are quite similar and reveals one outlier.

The comparison between the QCL-based sensor and the NIOX MINO, Fig. 7(b), shows an average value of 1.6 ppbv with a SD of 1.7 ppbv.