Value of lung diffusing capacity for nitric oxide in systemic sclerosis

Abstract A decreased lung diffusing capacity for carbon monoxide (DLCO) in systemic sclerosis (SSc) is considered to reflect losses of alveolar membrane diffusive conductance for CO (DMCO), due to interstitial lung disease, and/or pulmonary capillary blood volume (VC), due to vasculopathy. However, standard DLCO does not allow separate DMCO from VC. Lung diffusing capacity for nitric oxide (DLNO) is considered to be more sensitive to decrement of alveolar membrane diffusive conductance than DLCO. Standard DLCO and DLNO were compared in 96 SSc subjects with or without lung restriction. Data showed that DLNO was reduced in 22% of subjects with normal lung volumes and DLCO, whereas DLCO was normal in 30% of those with decreased DLNO. In 30 subjects with available computed tomography of the chest, both DLCO and DLNO were negatively correlated with the extent of pulmonary fibrosis. However, DLNO but not DLCO was always reduced in subjects with ≥ 5% fibrosis, and also decreased in some subjects with < 5% fibrosis. DMCO and VC partitioning and Doppler ultrasound‐determined systolic pulmonary artery pressure could not explain individual differences in DLCO and DLNO. DLNO may be of clinical value in SSc because it is more sensitive to DMCO loss than standard DLCO, even in nonrestricted subjects without fibrosis, whereas DLCO partitioning into its subcomponents does not provide information on whether diffusion limitation is primarily due to vascular or interstitial lung disease in individual subjects. Moreover, decreased DLCO in the absence of lung restriction does not allow to suspect pulmonary arterial hypertension without fibrosis.


Introduction
Systemic sclerosis (SSc) is a generalized fibrotic disease classified among the autoimmune connective-tissue disorders (van den Hoogen et al. 2013). Widespread microangiopathy (Asano and Sato 2015) and interstitial lung disease (ILD) (Baldwin Ede et al. 1949;Schoenfeld and Castelino 2015) are commonly found, whereas pulmonary arterial hypertension (PAH) affects only 5-12% of patients . The impairment of lung function is generally inferred from measurements of forced vital capacity (FVC) and lung diffusing capacity for carbon monoxide (DL CO ) (Schneider et al. 1982). However, FVC is not an accurate measure of lung restriction, that is, decreased total lung capacity (TLC) (Aaron et al. 1999;Pellegrino et al. 2005), and decrement of DL CO in SSc may reflect loss of alveolar surface or thickening of blood-gas barrier (Schoenfeld and Castelino 2015), but also changes in pulmonary microvasculature (Asano and Sato 2015).
Pulmonary CO uptake is mostly limited by its reaction rate with blood hemoglobin (Hb), thus DL CO is considered to be more sensitive to changes of pulmonary capillary blood volume (V C ) than alveolar membrane diffusive conductance for CO (DM CO ) (Gu enard et al. 1987;Borland and Higenbottam 1989). Accordingly, the reduced DL CO in isolation or disproportionately to FVC seen in a minority of SSc subjects has been interpreted as a sign of PAH with limited or no ILD (Wilson et al. 1964;Steen et al. 1992), though the correlation between DL CO and mean pulmonary artery pressure is weak (Mukerjee et al. 2004).
Nitric oxide (NO) has a much greater affinity and faster reaction with free Hb than CO (Gibson and Roughton 1957), which makes the lung diffusing capacity for nitric oxide (DL NO ) presumably more sensitive to alveolar membrane diffusive conductance than DL CO (Gu enard et al. 1987;Borland and Higenbottam 1989). Thus, the analysis of combined DL NO and DL CO measurement has been proposed as a method to partition DM CO and V C subcomponents (Gu enard et al. 1987) in various parenchymal (Phansalkar et al. 2004; Barisione et al. 2014Barisione et al. , 2016 and vascular (Farha et al. 2013) disorders. By using this method in SSc, one study (Sivova et al. 2013) suggested that partitioning of DL CO might be of interest to detect PAH either with or without ILD, whereas two other studies (Guarnieri et al. 2015;Degano et al. 2017) concluded that partitioning of DL CO is of little use in distinguish the patients with only ILD from those with ILD complicated by PAH. Two studies (Overbeek et al. 2008;Pernot et al. 2012) using the classical multistep alveolar O 2 partial pressure (PAO 2 ) method (Roughton and Forster 1957) to calculate DM CO and V C also gave contrasting results. Although DL NO has been demonstrated to be more sensitive than DL CO in idiopathic pulmonary fibrosis and this was due to the prevailing DM CO impairment (Barisione et al. 2016), this information in SSc is lacking. Therefore, the added value of DL NO and the partitioning of diffusion subcomponents in the diagnostic workup of SSc is still unclear.
The aim of the present study was to investigate whether DL NO and standard DL CO can provide different information on the lung involvement in individuals with SSc that can be attributed to interstitial or vascular abnormalities by DM CO and V C partitioning.

Study subjects
We retrospectively collected data from 96 Caucasian consecutive subjects fulfilling the current diagnostic criteria for SSc with a total score ≥ 9 (van den Hoogen et al. 2013). They had pulmonary function and other diagnostic tests completed between March 2014 and October 2017, when they were in stable clinical conditions. Sixteen of them received long-term intravenous iloprost, with various combinations of oral prednisone (n = 11), bosentan (n = 9), mofetil mycophenolate (n = 6), methotrexate (n = 4), and nifedipine (n = 3). Thirty-nine healthy subjects, matched for anthropometric and life-style data, were selected among health-care professionals to serve as a control group (Table 1). The study was approved by the Regional Ethics Committee (Registry no.: P.R.170REG/ 2015) and written informed consent was obtained from each subject to use his/her personal data.

Standard lung function measurements
Spirometry (Miller et al. 2005) and lung volumes (Wanger et al. 2005) were measured with the subject sitting in a whole-body plethysmograph (V62J, SensorMedics-Viasys, CareFusion; H€ ochberg, Germany). FVC, forced expiratory volume in one second (FEV 1 ), FEV 1 /FVC ratio, and TLC were determined and compared with available predicted values (Quanjer et al. 1993(Quanjer et al. , 2012. Standard DL CO was measured (MasterScreen PFT System, Jaeger-Viasys, CareFusion, H€ ochberg, Germany) by single-breath technique with 9-11 breath-hold time free of Valsalva or M€ uller maneuvers (Macintyre et al. 2005). CO back pressure was measured in expired gas prior to the inspiration of test mixture and compensated for analytically (Graham et al. 2002). Measured values were compared with reference values from Stanojevic et al. (2017) and adjusted for effective [Hb] in venous blood (Cotes et al. 1972).

DL NO measurement
At least 5-10 min after standard DL CO , single-breath DL NO -DL CO were simultaneously measured in sitting position (MasterScreen PFT System, Jaeger-Viasys, Care-Fusion, H€ ochberg, Germany) from the exponential disappearance rate of NO and CO with respect to helium (He) in exhaled air (Gu enard et al. 1987). A gas mixture of 0.28% CO, 9.0% He, and 20.9% O 2 balanced with N 2 was blended with 450 ppm NO in N 2 and inhaled from a plastic bag containing a final concentration of NO of 50.1 AE 4.8 ppm obtained ≤ 2 min before its use. The linearity of the electrochemical cell was checked by factory and the apparatus was calibrated for gas fractions using automated procedures. The subject was in a relaxed sitting position, wearing a nose-clip and breathing quietly through a mouthpiece with filter connected to a screentype pneumotachograph. After breathing 4-5 stable tidal volumes, he/she was requested to perform a full expiration to RV. Then a valve was opened allowing the subject to forcefully inhale the gas mixture achieving an inspired volume ≥ 90% of inspiratory vital capacity in < 2.5 sec. A breath-hold time of 4-6 sec duration, at near atmospheric intrapulmonary pressure and free of Valsalva and M€ uller maneuvers, was then requested followed by a rapid and smooth expiration to RV (Zavorsky et al. 2017). The total breath-hold time was calculated from the beginning of inspiration, minus 30% of inspiratory time, to the middle of expiratory gas sampling (Jones and Meade 1961). The first 750 mL of expired gas was discarded and the following 750 mL was sampled from the bag to be analyzed for NO, CO, and He fractional concentration. When the subject's vital capacity was < 2.0 L, sample and washout volumes were reduced to 500 mL (Macintyre et al. 2005). Between successive tests, an interval of ≥ 4-5 min was allowed to ensure complete elimination of prior test gases from the lungs. DL NO -DL CO values were accepted if two successive measurements were within 17.0 and 3.2 mLÁmin À1 ÁmmHg À1 , respectively, and the mean of two properly performed maneuvers was retained for analysis (Zavorsky and Murias 2006).

Derivation of diffusion subcomponents
DM CO and V C were derived by assuming a fixed value of blood conductance for NO (h NO ) of 4.5 mL (NO)Ámin À1 ÁmmHg À1 ÁmL (blood) À1 (Carlsen and Comroe 1958) which is largely independent of pulmonary blood [Hb] (van der Lee et al. 2005) and mean (expired) PAO 2 PAO 2 (Borland and Cox 1991). By contrast, the value of blood conductance for CO (h CO ) is variable and strictly dependent on [Hb] (Cotes et al. 1972) and PAO 2 , Data are absolute numbers or mean AE SD. SSc-N, subjects with all standard lung function data > 5 th percentile (LLN 5 ); SSc-D, subjects with DL CO < LLN 5 but TLC > LLN 5 ; SSc-R, subjects with TLC < LLN 5 with (n = 21) or without (n = 5) DL CO < LLN 5 ; c/f/n, current/former/never; FVC, forced vital capacity; FEV 1 , forced expiratory volume in one second; TLC, total lung capacity; DL CO , standard single-breath (9-11 s breath-hold time) lung diffusing capacity for carbon monoxide; K CO , DL CO /alveolar volume; DL NO, lung diffusing capacity for nitric oxide; K NO , DL NO /V A . *Significantly different from control. †Significantly different from SSc-N. ‡Significantly different from SSc-D. due to the competitive binding of CO and O 2 for Hb-accessible sites (Crapo et al. 1988). This makes the inspired O 2 fractional concentration (F I O 2 ) critical to DL CO measurement (Crapo et al. 1988) and, in turn, the DM CO and V C partitioning (Roughton and Forster 1957;Forster 1987). Thus, h CO values were calculated according to Gu enard et al.(2016) in vivo equation: (Bohr 1909), with _ VO 2 =DL o 2 being the ratio of O 2 uptake to lung diffusing capacity for O 2 which was assumed as %DL CO Á1.61 (Hsia et al. 2008). Moreover, [Hb] standard values Hb stand ½ ð Þ were set at 14.6 and 13.4 g•dL -1 for males and females, respectively (Cotes et al. 1972). For the simultaneous DL NO -DL CO measurement, due to an automated flushing procedure with 100% O 2 preceding each test and specific to this apparatus (Munkholm et al. 2018), the actual mean F I O 2 ( F I O 2 ) in the inspiratory bag was 0.228 AE 0.01, thus resulting in a PbO 2 of 125.4 AE 8.5 mmHg instead of the expected 99.1 AE 3.3 mmHg. To account for this potential bias, the corresponding values of h CO were calculated from Equation 1 to be 0.50 AE 0.04 and 0.55 AE 0.04 mL (CO)•min À1 •mmHg À1 •mL (blood) À1 , with PbO 2 of 125.4 AE 8.5 and 99.1 AE 3.3 mmHg, respectively. These values were used to derive the diffusion subcomponents in mild hyperoxia and normoxia from simultaneous DL NO -DL CO and separate standard DL CO and DL NO measurements, respectively, as follows (Zavorsky et al. 2017): where DM CO = DM NO /1.97 according to NO/CO physical diffusivity ratio (Wilhelm et al. 1977). Predicted values for DL NO and diffusion subcomponents were from Zavorsky et al.(2017).

Chest CT scanning
In 30 subjects, a thin-section CT scan obtained within 3 months before or after pulmonary function measurements was available. Scans of the entire chest had been routinely obtained in a supine position, during breathhold at full inspiration by a multidetector row-spiral scanner (SOMATOM Emotion 6, Siemens AG Medical, Forchheim, Germany). Images were acquired by 110 kVp tube voltage at 1.25 mm slice thickness and reconstructed at 1-mm increments by using smooth (B41s) and sharp (B70s) convolution kernels. All CT scans were independently analyzed quantitatively as well as semi-quantitatively by two experienced readers (G.B. and A.G.), with the latter being blind to clinical diagnosis and lung function data. For both analyses, the following six axial levels were considered: (1) aortic arch, (2) tracheal carina, (3) right pulmonary venous confluence, (4) midpoint between level three and level five, (5) 1 cm above the dome of the right hemi-diaphragm, and (6) 2 cm below the dome of the right hemi-diaphragm (Colombi et al. 2015). Mean lung density (gÁmL À1 ) was calculated by adding CT numbers (HU) and parenchymal volume (Gattinoni et al. 2005) measured in 2D at the six axial levels. In 13 subjects, mean lung density could also be obtained by automatic quantitative 3D analysis of the entire lungs (VIDA's Lung Volume Analysis, Coralville, IA, US). At each axial level, ground glass opacities and fibrotic changes, that is, subpleural honeycombing, reticular opacities, and traction bronchiectasis, were visually bounded and their respective volumes computed by a three-dimensional active contour segmentation software (ITK-Snap 3.6.0, Philadelphia, PA, USA) (Yushkevich et al. 2006). In each patient, the total percentages of lung volume with ground glass attenuation and/or fibrotic changes were calculated in relation to the total parenchymal volume of the six axial levels (Barisione et al. 2016).

Doppler echocardiography
In all subjects, a transthoracic Doppler echocardiography (Vivid E90, GE Healthcare, Little Chalfont, UK) was routinely obtained within 3 months before or after pulmonary function testing. In subjects with measurable tricuspid regurgitation velocity (TRV), systolic pulmonary artery pressure (sPap) was calculated as sPap = (4ÁTRV 2 + RAP), where RAP is right atrial pressure, estimated from the diameter and breath-induced variability in the inferior vena cava (Hatle et al. 1981). Values of sPap ≥ 36 mmHg were considered as suggestive of likely PAH (Greiner et al. 2014).

Statistical analysis
Lung function data were expressed as z-score, which indicates how many standard deviations a given measure differs from predicted, with a value À1.645, corresponding to the 5 th percentile of the reference population, assumed as the lower limit of normal (LLN 5 ). Categorical variables were compared by z-test with Yates correction and continuous variables by ANOVA with Holm-Sidak post hoc test for multiple comparisons. Associations between variables were determined by Spearman's rank correlation. Data are presented as mean AE SD or median with 25-75% interquartile range (IQR 25-75% ). In all analyses, the acceptable type I error was set at P < 0.05.

Results
Based on the standard lung function data, SSc subjects as the whole (n = 96) were divided into three groups, one with both TLC and DL CO above the LLN 5 (SSc-N) (n = 51), one with DL CO below the LLN 5 but TLC above the LLN 5 (SSc-D) (n = 19), and one with TLC below the LLN 5 with or without DL CO below the LLN 5 (SSc-R) (n = 26). Two SSc-D subjects showed mild-to-moderate airflow obstruction, but no CT evidence of emphysema. Eighteen out of the 39 subjects with measurable TRV had CT scan analysis and nine of them showed sPap values > 36 mmHg. Although, by definition all subjects of SSc-N group had standard lung function parameters above the LLN 5 , the group mean values were significantly (P < 0.001) lower than in the control group. Considering individual data, DL NO was below the LLN 5 in all restricted subjects, all but two SSc-D, eight out of nine with sPap > 36 mmHg, and also 11 out of 51 SSc-N subjects (Fig. 1). DL CO was below the LLN 5 in 21 out of 26 SSc-R subjects and six out of nine with sPap > 36 mmHg. DM CO was below the LLN 5 in the majority but not all SSc-R subjects and V C was below the LLN 5 in the majority but not all SSc-D subjects (Fig. 2). Overall, using LLN 5 as cutoff value, neither DL NO nor DL CO provided falsely positive results (100% specificity) but their sensitivities to the presence of SSc were 56% and 42%, respectively.
In 30 subjects with analyzable CT scans, the median extent of fibrosis was 15% (IQR 25-75% , 2 to 26%), while ground glass attenuation was 1% (IQR 25-75% , 0 to 9%). As the between-observer agreement of CT scans was very good for both fibrosis (weighted K, 0.92; 95% confidence interval, 0.84 to 0.99) and ground glass (weighted K, 0.82; 95% confidence interval, 0.70 to 0.94), the mean of two readings was retained for analysis. Both FVC and TLC zscores correlated significantly with mean lung density and fibrosis extent (Fig. 3A and B) but were above the LLN 5 not only in subjects with < 5% fibrosis, but also in a consistent number (11 and 8, respectively) of those with ≥ 5% fibrosis. DL CO and DL NO also correlated significantly with mean lung density and fibrosis extent ( Fig. 4A and B). However, DL NO but not DL CO was below the LLN 5 in all subjects with ≥ 5% fibrosis. Moreover, DL NO was below the LLN 5 in four out of nine subjects with ≤ 5% fibrosis. There were no significant correlations between ground glass extent and any measure of lung function.

Discussion
The main findings of the present study can be summarized as follows: (1) DL NO was reduced in 56% of subjects with SSc and in 22% of those with normal lung volumes and standard DL CO , (2) standard DL CO was normal in 30% of subjects with reduced DL NO , (3) DM CO and V C partitioning did not provide consistent explanations for individual differences in DL NO and DL CO , and (4) both DL NO and DL CO were significantly correlated with CT measurements of ILD but only the former was consistently reduced in all subjects with fibrosis extent ≥ 5%.

Comments on methodology
To account for the possible bias due to an increased F I O 2 (Munkholm et al. 2018) by an automatic flushing procedure with 100% O 2 before each test in the commercially available system used in the present study, we determined DM CO and V C using DL CO measured either in combination with DL NO or separately by standard technique (Figs. 5 and 6). These required different breath-hold times, that is, 4-6 and 9-11 sec, respectively, but this seems to have a negligible effect on final DL CO values (Graham et al. 1985).
In contrast to previous studies (Overbeek et al. 2008;Pernot et al. 2012;Sivova et al. 2013;Guarnieri et al. 2015), lung function data were expressed as z-scores, instead of absolute or percentage values, to avoid age-, sex-, and height-biases. Although the 2.5 th percentile (zscore of À1.96) has been suggested as a lower limit of normal for DL NO measurement (Zavorsky et al. 2017), we Figure 1. Relationships between the z-scores of lung diffusing capacity for nitric oxide (DL NO ) (y-axis) and standard (9-11 sec breath-hold time) lung diffusing capacity for carbon monoxide (DL CO ) (x-axis) in healthy controls (asterisks) and SSc subjects. Symbols indicate subjects with spirometry and total lung capacity (TLC) > LLN 5 (white), DL CO < LLN 5 (light grey), and TLC < LLN 5 (dark grey). Triangles indicate subjects with systolic pulmonary artery pressure (sPap) > 36 mmHg. The dashed lines indicate the LLN 5 (z-score < À1.645). The absolute values indicate the number of the total 96 test results for DL NO that fall into each quadrant (Q1-Q4). used the 5 th percentile (z-score À1.645) in order to make the results comparable with standard DL CO and avoid false negatives in subjects with established clinical diagnosis (Quanjer et al. 1993). On the other hand, no subject of the control group had DL NO , DM CO , or V C z-scores below À1.645, thus making the possibility of false positives unlikely.

Study limitations
First, the whole group of our control subjects showed a DL NO mean z-score of -0.69 AE 0.53. However, we used the reference equation (Zavorsky et al. 2017) that provided the lowest SD (0.58) of z-scores of our local dataset of healthy subjects. Moreover, all individual z-scores were

VC (z-score)
Hyperoxia Normoxia DM CO (z-score) Figure 2. Relationships between the z-scores of alveolar membrane diffusive conductance for CO (DM CO ) (y-axis) and pulmonary capillary blood volume (V C ) (x-axis) in mild hyperoxia (left panel) and normoxia (right panel) in healthy controls (asterisks) and SSc subjects. Symbols indicate subjects with spirometry and total lung capacity (TLC) > LLN 5 (white), DL CO < LLN 5 (light grey), and TLC < LLN 5 (dark grey). Triangles indicate subjects with systolic pulmonary artery pressure (sPap) > 36 mmHg. The dashed lines indicate the LLN 5 (z-score < À1.645). The absolute values indicate the number of the total 96 test results for DM CO that fall into each quadrant (Q1-Q4). definitely above the quoted LLN 5 (À1.645) for DL NO . Second, only 30 CT scans taken within 3 months from lung function studies were considered to reduce temporal variability and were analyzed at six axial levels. The CT images of 13 subjects were also suited for advanced quantitative analysis using all CT slices, and mean lung density was strongly correlated with that of the six axial levels (r = 0.99; P < 0.001). Thus, it seems reasonable to assume that the latter was representative of the whole lungs. Third, sPap was derived from TRV, which may not be detectable by transthoracic Doppler echocardiograpy in about half of subjects with invasively determined PAH (O'Leary et al. 2018). Although an estimated sPap > 50 mmHg has been arbitrarily recommended by previous guidelines to make PAH likely (Gali e et al. 2009), in a large recent study values ≥ 36 mmHg have been shown to provide a > 90% positive predictive value (Greiner et al. 2014). Nevertheless, PAH cannot be ruled out in subjects with undetected TRV. Fourth, while lung function tests were obtained in a sitting position, CT scans were acquired in supine posture, which might have altered V C (Cotton et al. 1990) and possibly mean lung density. It seems, however, unlikely that changes in body position might have altered the extent of fibrotic changes. Finally, 40 ppm of NO in the inspired gas could decrease hypoxic pulmonary vasoconstriction (Asadi et al. 2015), but this was found for PAO 2 values < 60 mmHg (Glenny and Robertson 2011), thus well below the values of 125.4 AE 8.5 and 99.1 AE 3.3 mmHg calculated for the hyperoxic and normoxic conditions of this study, respectively.

Comments on results
The value of DL CO partitioning into DM CO and V C subcomponents was investigated in five previous studies, two  . Bland-Altman plot of the difference between absolute values of lung diffusing capacity for CO measured in mild hyperoxia (DL CO,dual ) and normoxia (DL CO,stand ) (y-axis) vs. mean DL CO value (xaxis) in healthy controls (white circles) and SSc subjects (grey circles). The standard deviation (SD) of mean difference is bounded by the shaded area included between the horizontal dashed lines indicating 95% confidence interval (CI 95% ). It is noteworthy the scattered fluctuations of data around the mean value and the limits of agreement exceeding CI 95% in four cases. using the multistep PAO 2 method (Overbeek et al. 2008;Pernot et al. 2012) and three using the DL NO -DL CO method (Sivova et al. 2013;Guarnieri et al. 2015;Degano et al. 2017). Altogether, these studies showed that DM CO could be reduced in subjects with ILD, irrespective of the presence of PAH (Overbeek et al. 2008;Pernot et al. 2012;Sivova et al. 2013;Guarnieri et al. 2015;Degano et al. 2017), but also in subjects with isolated PAH without ILD (Guarnieri et al. 2015). By contrast, V C was reduced in subjects with PAH without ILD (Guarnieri et al. 2015;Degano et al. 2017) but highly variable in subjects with ILD, whether associated with PAH (Overbeek et al. 2008;Pernot et al. 2012;Sivova et al. 2013;Guarnieri et al. 2015;Degano et al. 2017), or not (Overbeek et al. 2008Pernot et al. 2012;Guarnieri et al. 2015;Degano et al. 2017). Although it has been suggested that DL NO may be more sensitive than standard DL CO in subjects without ILD or PAH (Guarnieri et al. 2015), none of the above studies reported direct comparisons between these two measures in individual subjects. Consistent with previous investigations (Abramson et al. 1991;Steen et al. 1992;Jacobsen et al. 1997;Guarnieri et al. 2015;Degano et al. 2017), a substantial number of SSc subjects of the present study had standard lung function measurements, including DL CO , within their respective normal ranges, that is, the 90% confidence interval. New findings of this study are that DL NO was reduced not only in all subjects with restriction, but also in 95% of those with reduced DL CO but no restriction, and about 22% of those with otherwise normal lung function. Consistent with these findings is the reduction in DL NO also observed in subjects with minimal or no fibrosis on CT scan. By converse, DL CO was normal in 30% of subjects with reduced DL NO , including individuals with lung restriction or likely PAH. In keeping with this finding is the normal DL CO also observed in some subjects with 10-40% fibrosis on CT scan. As 70-80% of the total resistance to CO uptake is deemed to be in the blood red cells (Gu enard et al. 1987;Borland and Higenbottam 1989), a reduction in DL CO without lung restriction has been considered as a sign of PAH (Wilson et al. 1964;Steen et al. 1992). Moreover, as the red cell resistance to NO uptake is only 40% of total (Borland et al. 2014), DL CO should be expected to be reduced in subjects without restriction or ILD more than DL NO and this difference be associated with V C rather than DM CO decrement. Neither of these predictions is supported by the present findings.
A possible reason for DL NO being more sensitive than DL CO even in the absence of restriction or fibrosis may be that DM CO is not simply related to membrane thickness and alveolar surface but to total gas exchange area, including vessel and red cell surfaces, whereas V C is related to pulmonary blood volume only (Kang and Sapoval 2016). Disproportionate reductions in gas exchange area may be the consequence of geometric heterogeneity of vascular bed, due to capillary remodeling or obliteration with blood volume being redistributed to unaffected lung regions, thus reducing DM CO but not total V C (Pande et al. 1975;Oppenheimer et al. 2006). DM CO impairment might be also the consequence of uneven blood red cell distribution within the alveolar capillaries with enhanced erythrocyte clustering and deformation (Hsia et al. 1997) due to increased plasma viscosity (Spengler et al. 2004), thus reducing their surface but not volume (Oppenheimer et al. 2006). These mechanisms are in keeping with previous data from Farha et al. (2013) showing that decreased gas transfer in idiopathic PAH may be due to loss of either DM CO or V C and may explain the variability in results of the present study in subjects with sPap > 36 mmHg. Another reason for decreased DM CO without V C changes could be a thickening of alveolar-to-capillary membrane by the presence of interstitial edema. However, there was a strict Figure 6. Bland-Altman plots of the difference between absolute values of alveolar membrane diffusive conductance for CO (DM CO ) (left panel) and pulmonary capillary blood volume (V C ) (right panel) both derived in mild hyperoxia and normoxia versus their respective mean value (x-axis) in healthy controls (white circles) and SSc subjects (grey circles). The SD of mean differences is bounded by the shaded areas included between the horizontal dashed lines indicating CI 95% . It is noteworthy the scattered fluctuations of data around the mean value of the relevant parameter and the limits of agreement exceeding CI 95% in six and seven cases for DM CO  correlation between mean lung density and % fibrosis extent (r = 0.80; P < 0.001, data not shown) and there was no significant ground glass attenuation in most of subjects, which makes this mechanism unlikely. On the other hand, owing to the strict dependence of DL CO on specific blood h CO (Roughton and Forster 1957), a preserved V C may attenuate the effects of alveolar-to-capillary membrane thickening by expanding the surface area available for gas exchange, which may explain the preserved DL CO in subjects with lung restriction or fibrosis.

Conclusions
Collectively, the results of the present study suggest that the measurement of DL NO may be of clinical value in the diagnostic workup of SSc, because it is more sensitive than DL CO , either in the presence or absence of lung restriction or fibrosis. However, DM CO and V C partitioning does not seem to be useful to tell whether different results of DL NO and DL CO are primarily due to vascular or interstitial lung disease in individual subjects. Finally, decreased DL CO in the absence of lung restriction does not allow to suspect PAH without ILD.

Conflict of Interest
G.B., A.G., M.B. and M.P. have no financial/nonfinancial interests to disclose; M.O. received personal fees from Imbio LLC for consultancies and a grant for Ph.D. course from Menarini Foundation; V.B. received personal fees and nonfinancial support for consultancy, given lecture, and travel reimbursement from ndd Medizintechnik.