Multiple Breath Nitrogen Washout: A Feasible Alternative to Mass Spectrometry

Background The lung clearance index (LCI), measured by multiple breath washout (MBW), reflects global ventilation inhomogeneity and is a sensitive marker of early cystic fibrosis (CF) lung disease. Current evidence is based on a customized mass spectrometry system that uses sulfur hexafluoride (SF6) as a tracer gas, which is not widely available. Nitrogen (N2) washout may be better suited for clinical use and multi-center trials. Objective To compare the results obtained from a N2 washout system to those generated by the SF6 based system in healthy children and children with CF. Methods Children with CF were recruited from outpatient clinics; healthy children were recruited from the Research4Kids online portal. Participants performed MBWSF6 (Amis 2000, Innovision, Denmark) and MBWN2 (ExhalyzerD, EcoMedics, Switzerland) in triplicate, in random order on the same day. Agreement between systems was assessed by Bland-Altman plot. Results Sixty-two healthy and 61 children with CF completed measurements on both systems. In health there was good agreement between systems (limits of agreement −0.7 to 1.9); on average N2 produced higher values of LCI (mean difference 0.58 (95% CI 0.42 to 0.74)). In CF the difference between systems was double that in health with a clear bias towards disproportionately higher LCIN2 compared to LCISF6 at higher mean values of LCI. Conclusion LCIN2 and LCISF6 have similar discriminative power and intra-session repeatability but are not interchangeable. MBWN2 offers a valid new tool to investigate early obstructive lung disease in CF, but requires independent normative values.


Introduction
Pathologic changes associated with cystic fibrosis (CF) lung disease occur in early childhood, but have historically gone undetected until the onset of clinical symptoms, at which point irreversible lung damage may have already occurred [1]. Consequently, over the last ten years the focus of clinical care in CF has shifted to early intervention and prevention of these structural changes. To facilitate early intervention there is a pressing need for surrogate markers of early obstructive lung disease that are also sensitive enough to detect treatment effects. [2] Spirometric measures, such as forced expired volume in one second (FEV 1 ), have traditionally been used in the assessment of CF lung disease due to their direct correlation with morbidity and mortality. [3] However, FEV 1 tends to remain within normal limits in a high percentage of children, despite radiographic evidence of airway damage. [4,5,6,7] This is likely due to the fact that these measures are primarily influenced by resistive changes in the large airways and thus not reflective of the patchy distribution of small airway pathology characteristic of early CF lung disease. [8] In addition to this inherent insensitivity, young children are also often not developmentally advanced enough to perform complicated respiratory maneuvers. The lung clearance index (LCI), as measured by multiple breath washout (MBW), reflects global ventilation inhomogeneity (VI) and as such is a highly sensitive marker of early obstructive lung disease. [9,10,11] Furthermore, LCI is more sensitive than other measures of lung function in detecting structural changes identified by high resolution computed tomography (HRCT) imaging [4,6,7]. MBW is performed during tidal breathing and requires only passive co-operation, it is therefore feasible during infancy and early childhood. Importantly, LCI tracks from preschool to school-age and has been found to precede subsequent abnormalities in spirometric indices [12].
To date most evidence for LCI has been collected using mass spectrometry based MBW systems. [9,10,11] The equipment is immobile, expensive and uses sulfur hexafluoride (SF 6 ) as its inert tracer gas. Therefore, the current customized system is neither suitable for multi-center clinical research nor clinical practice. Multiple breath nitrogen washout (MBW N2 ) offers a possible alternative to mass spectrometry based SF 6 washout (MBW SF6) . N 2 is a resident gas and permeates even poorly ventilated lung units, which may not be the case during MBW SF6 . Thus, the physiological attributes of the respective tracer gases may lead to differences in measurements obtained with the two systems. The aim of this study was to determine whether the results of MBW N2 and MBW SF6 can be used interchangeably in both healthy children and children with CF. In addition, we aimed to quantify the discriminatory power of LCI, as measured by MBW N2 and MBW SF6 , to differentiate health and disease throughout a range of pulmonary function abnormalities in CF.

Methods
This study was approved by the research ethics board (REB) at the Hospital for Sick Children (HSC), Toronto, Canada (REB# 1000019945). Informed written consent was obtained from the parents or guardians of healthy children and children with CF. Assent was obtained from subjects when appropriate.

Study Subjects
Families with eligible children between the ages of 3 and 18 years attending a routine visit to the CF outpatient clinic of the HSC were invited to participate in our study. Eligibility was defined as a diagnosis of CF by a positive newborn screening test or at least one clinical feature of CF in combination with either a documented sweat chloride $60 mEq/L by quantitative pilocarpine iontophoresis or a genotype with two CF-causing mutations. Children with acute respiratory symptoms, inter-current respiratory infections, or chronic lung disease not related to CF were excluded from participation; as were patients requiring supplemental oxygen.
Healthy controls were recruited from siblings of children attending our Respiratory Medicine outpatient clinics, children of staff members and through the Research4Kids online portal supported by the SickKids Research Institute. Health was defined as no history of chronic use of bronchodilator or controller medication for asthma symptoms, no chronic lung disease and no active or passive exposure to cigarette smoke. All subjects were free of acute respiratory tract symptoms for at least four weeks prior to testing. Children with any history of wheeze within the previous two years were excluded from the study.
Participants performed MBW SF6 and MBW N2 in triplicate, in random order on the same day. All children attempted to perform spirometry, while plethysmographic lung volume measurement was attempted by children age seven and older. Lung function testing was performed according to American Thoracic Society (ATS) standards using the Vmax system (VIASYS CareFusion San Diego, California, USA). [13,14] Children between the ages of 3 and 6 years performed spirometry to ATS ERS standards for preschool lung function testing [15] using the Easy-on-PC system (ndd, Zurich, Switzerland). Height, weight, BMI and spirometry outcomes were standardized for age, body size and sex. [16,17,18] MBW Testing MBW SF6 . A mass spectrometer (AMIS 2000; Innovision A/S, Odense, Denmark) based set up and technique was used to perform MBW testing with a SF 6 /He gas mixture as previously described. [9,10,11] Briefly, subjects breathed a gas mixture containing 4% SF 6 , 4% He, 21% O 2 , balance N 2 via an open circuit bias flow system through either a mask or mouthpiece and an attached heated pneumotachograph (3700 series Hans Rudolph, Shawnee, KS, USA) which measures flow by pressure differential, until equilibrium was reached. Once the inert tracer gas (SF 6 ) stabilized at 4%, the gas source was removed during the start of exhalation and the subject breathed room air until endtidal SF 6 concentration reached below 1/40 th of its starting concentration for at least three breaths. Depending on individual feasibility, either a mask (Silkomed, Rendell Baker Masks size 3, Rusch Canada Inc., Benson Medical Industries, Markham, Ontario) filled with therapeutic putty (Air Putty, Sammons Preston Canada Inc., Mississauga, Ontario) or mouthpiece (VacuMed model #1004, Ventura, CA, USA) with nose clips was used. All subjects used the same size pneumotachograph with a total post gas sampling point dead space of 15.4 ml; pre-gas sampling point dead space was considered to be zero for mouthpiece and 10 mls for mask and putty [19]. Calculation of signal delay and subsequent alignment of flow and gas concentration signals with appropriate BTPS correction was performed as previously described. [9,10,11] MBW N2 . MBW N2 was performed using an open circuit, bias flow system (Exhalyzer D H , EcoMedics AG, and Duernten, Switzerland) and associated software (Spiroware H 3.1 EcoMedics AG). This MBW N2 device uses an indirect technique to determine N 2 concentration. Oxygen (O 2 ) and carbon dioxide (CO 2) were measured during testing; N 2 was then calculated based on Dalton's law of partial pressures.[20] CO 2 was measured using a mainstream infrared CO 2 sensor (Capnostat H 5, Respironics Novametrix LLC, Wallingford CT, USA). Incorporated into the CO 2 sensor was a sampling port where O 2 was measured side stream at a rate of approximately 3 ml/s to an internal O 2 analyzer (Oxigraf Inc, Mountain View, CA, USA). Flow was measured by an ultrasonic flow head [21] inline along the breathing circuit, and volume was derived from the flow signal by integration. Due to differences in O 2 and CO 2 sensor response times a speeding algorithm was applied to the O 2 signal to reduce the response time to approximately 110 ms in order to align gas signals. Synchronized gas signals were time-shifted to align with flow as described by . [20] In contrast to MBW SF6 , a wash-in phase using a test gas was not required. The subject breathed 100% O 2 during wash out to reduce the concentration of N 2 in the lungs to below 1/40 th of the starting concentration. The switch from room air to 100% O 2 was automated, eliminating the need for manual disconnect as was done during MBW SF6 . As there was no parallel wash-in phase during MBW N2 subjects were allowed to re-equilibrate in room air between test trials. Time between trials was at minimum the time required to washout on the previous trial.

Offline Data Analysis
Synchronized data files from both systems were analyzed by trained observers using custom written analysis software (Test-Point, Capital Equipment Corp., Billerica, MA, USA). To assess inter-observer variability of offline MBW results, the N 2 data files from 40 subjects (20 HC and 20 CF) were independently overread by two observers. Quality control standards, as proposed by the ERS working group [19], were used as guidelines for technical acceptability during offline data analysis.

Indices calculated
Functional residual capacity (FRC) is calculated by dividing the net amount of inert tracer gas exhaled over the course of the washout by the difference in end-tidal marker gas concentration (Cet) from the beginning to the end of washout. [22] LCI represents the number of FRC turnovers required to reduce the end-tidal concentration of tracer gas to 1/40 th of the starting concentration and is calculated by dividing the sum of exhaled tidal breaths (cumulative exhaled volume (CEV)) by simultaneously measured FRC. [22] Statistical Analysis For each outcome, agreement between the SF 6 and N 2 systems was assessed using Bland-Altman plots. [23] A t-test was used to test whether MBW outcomes in healthy controls were different from children with CF. Additional analysis used simple linear regression to determine whether the differences between the two systems could be explained by body size and/or lung function. A p-value ,0.05 was regarded as statistically significant.
Results 144 children (68 healthy controls and 76 CF) were enrolled into this study ( Figure 1). Subjects who failed to meet MBW SF6 and or MBW N2 quality control criteria were excluded from analysis ( Figure 1). In most cases, subjects failed to meet quality control criteria due to inability to maintain stable breathing pattern, leak around interface, or incomplete washout. In total 62 HC (91%) and 61 CF (80%) had paired measurements on both systems available for analysis. Both groups were well matched for age and sex. As expected the healthy group were taller and heavier than CF subjects (Table 1). Spirometry (FEV 1 z-scores) was reduced in the CF group compared to healthy controls, whereas FRC measured by plethysmography (percent predicted) was elevated in CF compared to healthy controls (Table 1). Each subject completed at least two acceptable MBW trials. Overall the within test occasion variability (coefficient of variation (CV) of all trials) was similar for both systems, and similar in health and disease ( Table 2). There was no evidence that the CV was affected by increased ventilation inhomogeneity as CV was constant across the range of LCI.

LCI comparison between systems
In both systems LCI identified the same proportion (96%) and the same subjects as abnormal. On average, in healthy subjects   (Figure 2b). The same bias was not observed when LCI SF6 was compared to LCI measured using another low density gas, helium (LCI He ). While the variability in the difference between LCI SF6 and LCI He increased as the average LCI increased, the scatter was uniform on both sides of the mean difference (data not shown).

FRC comparison between systems
As a crude way to adjust for body size, FRC measurements from both systems were adjusted for height (FRC/height)*100 and expressed as relative FRC. In health MBW N2 produced higher values of FRC (mean difference (FRC N2 2RC SF6 ) = 0.21 (95% CI 0.16; 0.25)), with no bias observed between systems (limits of agreement 20.15; 0.56) (Figure 3a). In CF the difference between the two systems was greater than in health (mean difference = 0.33 (95%CI 0.27; 0.38)), and the difference was disproportionately greater with higher average adjusted FRC (Figure 3b).
Thirty CF and 44 HC had measurements of all three FRC outcomes (FRC pleth , FRC SF6 and FRC N2 ) ( Table 2); for comparison each FRC measure was corrected for body size in the same manner (FRC/height*100). FRC N2 more closely agreed with FRC pleth (Figure 4). As the difference between FRC pleth and The solid horizontal line represents the mean difference, and the dashed lines represent the limits of agreement (mean difference+/22SD). In health, there was good agreement between the systems, the mean difference (LCI N2 2LCI SF6 was 0.61 (95% CI 0.45 to 0.78), limits of agreement (20.7 to 1.9)); whereas in CF there was an obvious bias (mean difference = 1.41 (95% CI 0.92 to 1.90), limits of agreement (22.4 to 5.2)) such that LCI N2 increased disproportionately to LCI SF6 as mean LCI increased. doi:10.1371/journal.pone.0056868.g002 Comparison of MBWN2 to MBWSF6 PLOS ONE | www.plosone.org FRC SF6 may represent the volume of gas in extremely slowly ventilated lung units, we compared the difference in LCI between systems to trapped gas volume (FRC pleth 2FRC SF6 ). We observed that the volume of trapped gas increased as LCI N2 increased disproportionately to LCI SF6 suggesting that the N 2 system is measuring volume not captured using SF 6 ( Figure 5).

Additional comparisons between systems
As LCI is the cumulative expiratory volume (CEV) divided by FRC, we examined the agreement of CEV N2 and CEV SF6 , corrected for pre and post gas sampling point dead space, between systems and found good agreement in health with no bias observed (limits of agreement 20.001 to 0.041) ( Figure 6). In CF, there was a strong bias such that CEV N2 was disproportionately higher than CEV SF6 with increasing mean values of CEV (limits of agreement 20.041 to 0.150).
Since CEV is the product of tidal volume (Vt) and number of breaths required to complete washout, we compared the Vt/FRC ratio between systems. Both variables were corrected for pre and post gas sampling point dead space. While the variability of Vt/ FRC was greater in health than in CF, there was minimal difference and no bias observed when the two systems were compared (data not shown). Healthy subjects required an additional 5 breaths to complete washout during MBW N2 compared to MBW SF6 (mean (SD): 35 (14) vs. 30 (13), p,0.001). CF subjects required an additional 18 breaths to complete washout using the N 2 system (mean (SD): 56 (26) vs. 38 (14), p,0.001). This indicates that the bias observed in CEV between systems is related to number of breaths. When the difference in breath number was compared to volume of trapped gas we found that number of breaths required to complete washout using N 2 increases proportionally to volume of trapped gas (data not shown).
Respiratory rate was lower during MBW N2 compared to MBW SF6 in both health (17 breaths/minute vs. 19; p,0.001) and disease (18 breaths/minute vs. 21; p,0.001)), but was constant across the range of LCI; there was no bias observed in respiratory rate between the two systems (data not shown).

Comparison between systems and disease severity
To determine whether the difference in LCI between systems  Figure 5. Comparison of the mean difference in LCI between systems to volume of trapped gas (FRC pleth 2FRC SF6 ). The volume of trapped gas increased as LCI N2 increased disproportionately to LCI SF6 suggesting that the N 2 system was measuring volume not captured during MBW SF6 . doi:10.1371/journal.pone.0056868.g005 was related to lung function we compared the difference in LCI across a range of lung function abnormalities. The difference in LCI between the two systems was greater as lung function worsened (i.e. lower values of FEV 1 (Figure 7) and higher values of FRC pleth (data not shown)), such that on average LCI N2 was disproportionately higher than LCI SF6 in subjects with abnormal lung function compared to those with normal spirometric and plethysmographic findings (data not shown). The observed differences could not be explained by differences in age or body size (height, weight, BMI (data not shown)). Finally, to investigate the contribution of factors explaining the observed differences in LCI between systems, a linear regression was performed for each factor separately (Table 3). Greater breath number during MBW N2 compared to MBW SF6 explained most of the variability (24%) in the difference in LCI while trapped gas and zFEV1 explained 15% and 13% of the variability respectively.

Discussion
To the best of our knowledge, no other study has directly compared outcomes measured by MBW N2 to those measured by both MBW SF6 and traditional lung function tests in healthy children and children with CF. LCI N2 and LCI SF6 had similar discriminative power and intra-session repeatability but are not interchangeable as LCI N2 was on average higher than LCI SF6 . As such, interpretation of parameters measured by MBW N2 will require independent normative values to define an appropriate upper limit of normal.
The feasibility of using MBW N2 in a pediatric clinical setting has recently been described but this study did not include head to head comparison to other technologies [24]. Two studies have previously compared alternative MBW systems to mass spectrometry based MBW SF6 . [25], [26]. However, both used SF 6 as the tracer gas and neither performed between system comparisons in the same individual nor compared MBW based lung volume measurements to plethysmographic FRC measurements; therefore results are not directly comparable to our study.
Although the LCI and FRC were comparable between systems in health, albeit higher using N 2 , the bias observed in CF subjects clearly demonstrates that the two systems cannot be used interchangeably. These observed differences could potentially be explained by differing physiological properties of SF 6 and N 2 . SF 6 is a heavy gas and thus may behave differently in the periphery of the lung than a lighter gas (He or N 2 ); however comparison of LCI SF6 to LCI He in CF did not demonstrate the same bias observed between LCI SF6 and LCI N2 . The endogenous nature of N 2 results in the contribution of gas from very slowly ventilated lung units not captured by MBW SF6 as evidenced by the relationship between trapped gas, number of breaths and difference in LCI between systems. However, this will also increase washout time in subjects with uneven ventilation distribution as it will take longer to clear endogenous tracer gas from their lungs compared to SF 6 , which may not equilibrate in extremely slowly ventilated lung units.
FRC measured by MBW is subject to the same limitations as other gas dilution techniques in that only communicating lung units will contribute to measured volume, while FRC measured by body plethysmography includes all compressible gas volume. Thus, in subjects with significant peripheral airway obstruction we would expect differences between FRC pleth and FRC MBW , and   indeed FRC measured by both MBW techniques was lower than that measured by plethysmography. However, we observed that FRC N2 more closely agreed with FRC pleth . These results suggest that the difference between FRC pleth and FRC SF6 may reflect trapped gas volume and that the volume contribution of slowly ventilated lung regions, not captured during MBW SF6 , results in lower FRC SF6 values. Consequently, during MBW N2 subjects with CF required significantly more breaths to complete washout leading to the disproportionately higher CEV N2 compared to CEV SF6 . Our data demonstrate that these differences are progressively more pronounced with worsening obstructive lung disease. LCI N2 was shown to increase disproportionately more than LCI SF6 with greater disease severity (increased FRC pleth and lower FEV 1 ) and as such may be able to more accurately reflect the degree of VI than LCI SF6 . These interpretations are based on the assumption that the additional gas volume measured during MBW N2 can be attributed to measurement of gas in extremely slowly ventilated lung units. However, a further unquantifiable amount of tissue dissolved N 2 will diffuse from the blood into the alveoli during MBW N2 , particularly during long washouts seen in subjects with significant VI. Most evidence would suggest unless lung disease is severe the tissue N 2 contribution will be relatively low. [19] The close correspondence of FRC N2 and FRC pleth observed in this study would support this hypothesis.
While it would appear that MBW N2 is better able to reflect the degree of peripheral airway disease than MBW SF6 , washout times will be substantially longer in subjects with significant VI. Long washout times may limit the feasibility of MBW N2 in the clinical setting. This limitation could potentially be overcome by choosing higher cut-off concentrations earlier in the washout. Preliminary evidence [27] would suggest that this is possible without compromising the sensitivity of MBW N2 . Investigation into the minimal number of trials required to achieve reproducible results; another option to shorten the test duration, is ongoing.
In conclusion, MBW N2 offers a valid tool to investigate obstructive lung disease in CF. Furthermore, future studies in younger patients are required to better understand the sensitivity of multiple breath N 2 washout in this age group. In addition, interventional studies similar to those performed with MBW SF6 are needed to further clarify the role of MBW N2 as an outcome measure in clinical trials in CF patients.