We investigated ventilation inhomogeneity in patients with chronic obstructive pulmonary disease (COPD) through use of the multiple breath N_{2} washout test (MBW). From an alveolar slope analysis throughout the MBW, we derived two indices, S_{cond} and S_{acin }, as a measure of ventilation inhomogeneity in conductive and acinar zones of the lungs, respectively ( *J. Appl. Physiol.* 1997;83:1807–1816). We evaluated the relationship of S_{cond} and S_{acin} to standard lung-function indices by means of a principal-components factor analysis, which linked correlated indices to independent factors accounting for 81% of the total variance within the COPD group. S_{acin} was linked to the so-called acinar lung-zone factor, which also comprises diffusion capacity measurements. S_{cond} was linked to the so-called conductive lung-zone factor, which also comprises specific airway conductance (SGaw) and forced expiratory flows. FEV_{1} divided by FVC (FEV_{1}/FVC) was the only variable linked to both the conductive and the acinar lung-zone factors. The fact that S_{cond} and S_{acin} were linked to independent factors provides statistical confirmation of the hypothesis that S_{cond} and S_{acin} reflect independent lung alterations, whereas FEV_{1}/FVC behavior indicates a combined conductive and acinar contribution to airways obstruction.

The evaluation of inhomogeneity in central versus peripheral ventilation with standard lung-function indices is rendered difficult because the dividing line between lung regions represented by one or another index is sometimes subject to interpretation. The same difficulty exists for the alveolar slope of the VC single-breath washout (SBW). Although often regarded as a measure of peripheral-airways inhomogeneity, increases in the alveolar slope can also result from unequal narrowing of the conductive airways, without alteration at the acinar level, as has been shown to be the case during bronchoprovocation (1). In particular, the same provocation study in which this was shown (1) also showed how the alveolar-slope analysis of a multiple-breath N_{2} washout (MBW) test can yield two indices that relate directly to ventilation inhomogeneity occurring either proximal or peripheral to the diffusion front. On the basis of theoretical modelling of convective and diffusive mixing in the lungs (2, 3), it is indeed possible to consider the diffusion front as a diving line. During normal breathing, the proximal part of the diffusion front is located at the entrance of the acinar lung zone, and spreads out over several generations of airways beyond this. That is why we refer to our MBW indices as reflecting ventilation inhomogeneity in the conductive airway region (S_{cond}) and acinar airway region (S_{acin}).

When ventilation distribution is impaired, S_{cond} and S_{acin} increase, since they are derived from alveolar slopes. In our provocation study (1), we showed a significantly increased baseline S_{acin} value in a group of subjects hyperresponsive to histamine as compared with a group of nonhyperresponsive subjects. Baseline S_{cond} was similar in both groups, and large increases in S_{cond} appeared only during bronchoprovocation. In the present study, we evaluated S_{cond} and S_{acin} in a group of patients with chronic obstructive pulmonary disease (COPD), in whom considerable conductive and acinar impairment of ventilation are expected to occur.

When introducing a new index for the characterization of pulmonary pathophysiology, a proper way to validate it is with a sensitivity–specificity analysis. This, however, requires a “gold standard.” In the absence of such a standard, the new index is usually correlated to indices that have been previously validated. In the present study of S_{cond} and S_{acin} as indices that characterize the contribution of conductive versus acinar airways to ventilation inhomogeneity, the main problem is the choice of lung-function indices with which to compare S_{cond} and S_{acin}. For instance, FEV_{1} divided by FVC (i.e., FEV_{1}/ FVC), which can be used to quantitate airways obstruction in COPD, primarily reflects malfunction of conductive airways, although a contribution from acinar airways cannot be ruled out. Conversely, S_{acin} potentially reflects acinar structural change, including changes in peripheral lung asymmetry that do not lead to loss of carbon monoxide diffusing capacity (Dl _{CO}) so that only part of S_{acin} is also reflected in the measurement of CO diffusing capacity.

A way to reduce the complexity of all the potential correlations between MBW and lung-function indices is to conduct a principal-component-factor analysis (4). In analogy with the study by Mahler and Harver (5), in which three indices of dyspnea were evaluated with respect to indices of respiratory muscle strength and standard lung function (also in COPD patients), we performed a factor analysis to evaluate how S_{cond} and S_{acin} relate to each other and to all other indices obtained in our COPD subjects. In particular, factor analysis is a statistical technique for condensing several correlated variables measured in a group of subjects into a few underlying, independent, ordered factors explaining most of the total variance of the group. Factor analysis can also reveal which measured variables are most closely associated with each of the retained factors. Therefore, if a variable is highly correlated with one of the factors, it can be considered as pertinent for the characterization of the group, and if two variables are linked to different factors, they can be considered independent (i.e., participating in different physiopathologic mechanisms). Factor analysis would therefore allow us to verify: (*1*) whether S_{cond} and S_{acin} contribute to the characterization of patients within a COPD group; (*2*) whether S_{cond} and S_{acin} belong to different (i.e., independent) factors; and if so, (*3*) which, if any, of the conventional lung-function indices under study belong to the factors that also comprise S_{cond} and S_{acin}.

All lung-function indices except those related to the MBW test were obtained with standard lung-function laboratory equipment (SensorMedics, Bilthoven, The Netherlands). The MBW tests were performed with a dedicated breathing assembly that has been extensively described elsewhere (1). This allowed the patient to breathe in a regular pattern with a tidal volume (Vt) of approximately 1 L, starting from FRC and using pure oxygen for inspiration. Dead-space volume (i.e., the volume that did not contain pure O_{2} prior to the onset of the MBW test) was 50 ml. Each patient was to perform three MBW tests, with an interval of ∼ 2.5 min between each test. Lung-function measurements included single-breath Dl _{CO}, Dl _{CO} divided by alveolar volume (Kco), FEV_{1}, FVC, forced expiratory flow after exhaling 75% of FVC (FEF_{75}), specific airway conductance (SGaw), and body plethysmographic FRC (FRC_{pl}).

We selected 20 stable COPD patients fulfilling the following criteria: ⩾ 45 yr of age, smoking history (⩾ 10 pack-yr), nonatopic (skin allergy tests negative), FEV_{1}/FVC ⩽ 70%, and with an increase in FEV_{1} after bronchodilation of less than 9% predicted. None of the patients took any medication in the 12-h period before the time of study, which was usually done between 9:00 and 10:00 a.m.

The analysis of MBW test data began with the computation of FRC (FRC_{mbw}) from the amount of cumulatively expired N_{2} throughout the MBW test. Each MBW expiration was then treated as a single-breath washout (SBW), computing the alveolar slope by linear regression of the N_{2} concentration between 0.65 L expired volume and the end of expiration (nominally 1 L), and dividing the result by the mean expired N_{2} concentration of that breath to provide a normalized alveolar slope (6). As a last step, the normalized alveolar slope (Sn) of each expiration was plotted as a function of lung turnover (TO) (i.e., cumulative expired volume of each breath divided by the subject's FRC_{mbw}). The use of lung turnover instead of breath number on the abscissa allows better comparison of subjects with different lung volumes and N_{2} dilutions (7). Figure 1 shows the Sn curves obtained from two COPD patients “copd1” and “copd2” with comparable lung volumes. The dotted line in Figure 1 illustrates the increase in Sn (smoothed curve) typical of an asymptomatic nonsmoker designated as “normal.” The line was reconstructed from the average data for the asymptomatic nonhyperresponsive group in our previous study (1), in which we have described in great detail how S_{cond} and S_{acin} were derived from Sn curves (also depicted in Figure 1). Here, we provide only some background information and reiterate the mathematical description of these MBW indices.

_{cond}and S

_{acin}

Model simulations predict that the increase in Sn as a function of TO is due to: (*1*) inhomogeneity in conductive-airways ventilation resulting from unequal conductive-airways narrowing and/or inhomogeneous expansion of any two lung units subtended from these conductive airways, resulting in convective flow sequencing between them; this leads to a steady increase in Sn throughout the MBW test, with Sn = 0 for TO = 0 (8); and (*2*) inhomogeneity in acinar-airways ventilation deriving from any change in structural asymmetry of the lung periphery through a change in airway caliber or air-space enlargement, even in the absence of convective flow sequencing; this produces an initial nonzero Sn that hardly increases beyond the first few breaths, to the extent that beyond TO = 1.5, any contribution to an increase in Sn from the acinar airways may be considered negligible (3, 6). Therefore, S_{cond} is defined as the difference in Sn per unit TO between TO = 1.5 and TO = 6, within which the acinar contribution is negligible, and S_{acin} is determined by subtracting from the Sn of the first breath the part that can be attributed to the conductive airways (i.e., S_{cond} multiplied by the TO value of the first breath). The ‘*normal*' subject (Figure 1, *dotted line*) was reconstructed as a curve corresponding to S_{cond} = 0.033 L^{−1} and S_{acin} = 0.075 L^{−1} (i.e., the average S_{cond} and S_{acin} values obtained in a group of normal, nonhyperresponsive subjects) (1).

In summary, a larger S_{cond} value, reflecting a steeper Sn curve, points to impaired ventilation originating in the conductive airways zone and therefore between large lung units. A larger S_{acin}, representing a larger offset of the entire Sn curve, indicates alteration at the level of the acinar airways structure. In the example shown in Figure 1, the fact that the “normal” and “copd1” curves have a different offset but run parallel between TO = 1.5 and TO = 6 is translated in similar S_{cond} values (0.032 L^{−1} versus 0.033 L^{−1}) and different S_{acin} values (0.225 L^{−1} versus 0.075 L^{−1}), suggesting only impairment of acinar ventilation. By contrast, “copd2” shows a three times steeper curve between TO = 1.5 and TO = 6 than does “normal,” and “normal” and “copd2” have a different offset, which is translated in different S_{cond} (0.104 L^{−1} versus 0.033 L^{−1}) and different S_{acin} (0.667 L^{−1} versus 0.075 L^{−1}) values, indicating both conductive and acinar ventilatory impairment.

All statistical analyses were done with the SPSS software package (SPSS Inc., Chicago, IL). The principal-components factor analysis was performed to extract from the set of nine variables the smallest number of independent factors explaining the maximum of the common variance among the variables. The number of factors extracted was based on the latent root criterion (i.e., factors with an eigenvalue > 1 were extracted) (4). As was done in the study by Mahler and Harver (5), we used the Varimax orthogonal rotation option to simplify factors, so that each variable tends to load highly on only one factor. In our particular case, an orthogonal rotation (i.e., a rotation that forces all factors to be independent) was preferred to an oblique rotation, to permit verifying whether S_{cond} and S_{acin} were indeed independent. A factor loading (i.e., the correlation of a variable with a factor) of greater than 0.56 was retained to define a variable contributing to a given factor (p < 0.01 for a sample size n = 20). This means that the factor accounts for at least 31% (0.56^{2} × 100) of the variance of the variable (4).

All patients participating in the study were able to perform the MBW tests adequately (i.e., producing sawtooth breathing patterns with a Vt of about 1 L). Despite the fact that the patients were not trained or locked into a 1-L breathing pattern by end-inspiratory valve switching, the average Vt obtained from the MBW tests on all patients was 1,021 ± 50 ml (mean ± SD), with a breathing frequency of 9 ± 3 breaths/min. The coefficient of variation (CV) of Vt within each MBW test averaged 10 ± 2%. The FRC_{mbw} value was not significantly different from the predicted value for the subjects in the study (paired *t* test; p > 0.1). By contrast, FRC_{pl} averaged 124 ± 30% predicted, pointing to a considerable degree of gas trapping in these subjects even if one takes into account that for methodologic reasons, FRC_{pl} is probably overestimated and FRC_{mbw} underestimated in these patients (9).

Table 1 lists the average values and standard deviations of all indices that were to be included in the factor analysis, as obtained in the COPD group in the study. All lung-function indices in Table 1 were abnormal and showed important variability around their mean values, with CVs between 20% and 60%. The MBW indices S_{cond} and S_{acin} obtained in this group were significantly different (unpaired *t* test; p < 0.001 for both indices) from “normal” values reported for a group of nonhyperresponsive asymptomatic subjects (1). In particular, S_{cond} ranged from one to five times the average normal value, whereas S_{acin} ranged between three to 10 times the average normal value.

Mean | SD | |||||
---|---|---|---|---|---|---|

Py | — | 44 | 22 | |||

SG_{aw} | (cm H_{2}O)^{−1} · sec^{−1} | 0.07 | 0.04 | |||

Kco | % predicted | 77 | 25 | |||

Dl _{CO} | % predicted | 69 | 24 | |||

FEV_{1} | % predicted | 60 | 18 | |||

FEV_{1}/FVC | % | 52 | 11 | |||

FEF_{75} | % predicted | 18 | 9 | |||

S_{acin} | L^{−1} | 0.43 | 0.18 | |||

S_{cond} | L^{−1} | 0.085 | 0.038 |

Table 2 summarizes the result of the principal-components factor analysis, using no more than the nine variables in Table 1 in order to comply with the recommendation that the number of variables equals half the sample size (i.e., number of patients) or less (4). Three factors were extracted, the first of which accounted for 51% of the variability. The explained variability cumulatively increased by up to 69% and 81%, respectively, with the second and third factors. Within each factor, the degree of correlation of each variable with that factor is expressed by a factor loading, which is represented in Table 2 by a *boldface* number when its absolute value was ⩾ 0.56. On the basis of the criterion that the absolute value of a variable's factor loading should be more than 0.56 in order to contribute to a particular factor, Dl _{CO}, Kco, and S_{acin} contribute only to the first factor. Strictly speaking, the opposite sign of the respective factor loadings means only that correlations with the common factor are of opposite sign, but by extrapolation we can roughly state that, for instance, S_{acin} increases as Dl _{CO} and Kco decrease. The second independent factor essentially contains all the conductive-airway indices in relation to an overall increase in airway resistance (FEV_{1}, FEF_{75}, SGaw) and sequential emptying of lung units through inhomogeneity in resistance of the conductive airways or inhomogeneity of compliance of the subtended units (S_{cond}). FEV_{1}/FVC is the only variable that is linked to more than one factor (Factors 1 and 2). Smoking history is not linked to any of the other variables, and makes up for an independent factor (Factor 3).

Factor 1 | Factor 2 | Factor 3 | ||||
---|---|---|---|---|---|---|

S_{acin} | −0.82 | −0.24 | −0.31 | |||

Kco | 0.91 | −0.01 | −0.25 | |||

Dl _{CO} | 0.86 | 0.30 | −0.25 | |||

FEV_{1}/FVC | 0.68 | 0.59 | −0.04 | |||

S_{cond} | 0.17 | −0.77 | 0.14 | |||

FEV_{1} | 0.39 | 0.82 | −0.02 | |||

FEF_{75} | 0.29 | 0.82 | −0.04 | |||

SG_{aw} | 0.29 | 0.78 | 0.01 | |||

Py | −0.13 | −0.10 | 0.95 |

The COPD patients in the present study showed considerable heterogeneity in terms of traditional lung-function indices as well as MBW-derived indices (Table 1). This heterogeneity made the data particularly suitable for a principal-components factor analysis, because in theory such analysis can yield two or more independent factors only if there is variability in at least two of the indices under study. The two most important findings of this factor analysis were that: (*1*) S_{cond} and S_{acin} contributed significantly to the characterization of COPD patients (i.e., their average values were not only grossly abnormal as compared with those of normal subjects, but the degree of abnormality was further translated into very different values within the COPD group); and (*2*) it assigned S_{cond} and S_{acin} to independent factors, confirming the theoretical concept underlying the MBW normalized slope analysis, that alterations at the level of conductive airways and acinar airways can be characterized through two independent indices.

These two characteristics of S_{cond} and S_{acin} make them complementary to standard lung function indices, because they both derive from a single test involving a breathing maneuver and underlying theory that differ completely from the standard lung-function techniques. Also, S_{cond} and S_{acin} can be linked to functional lung units. In particular, S_{acin} can be useful for characterization of the silent zone of the lungs, where resistance-related indices are less efficient in revealing airways alteration. This analysis shows that FEF_{75}, which is generally thought to reflect the so-called small airways, is linked to the same factor as S_{cond}, strongly suggesting that these small airways are nevertheless proximal to the diffusion front. Also, the significant factor loadings of FEV_{1}/FVC on both Factor 1 (containing S_{acin}) and Factor 2 (containing S_{cond}) shows that at least in COPD patients, impairment of FEV_{1}/FVC results from a combination of conductive and acinar airways alterations.

Factors extracted from a factor analysis are often named in order to give them a more concrete meaning (4). On the basis of the normalized slope theory, S_{cond} and S_{acin} should always end up in independent factors, even in a lung disease other than COPD. Therefore, it is tempting and straightforward to name Factor 1 (containing S_{acin}) the “acinar lung zone” factor, and Factor 2 (comprising S_{cond}) the “conductive lung zone” factor. Because forced expiration indices are potentially a combination of central and peripheral effects, factor analysis in a group of patients suffering from a different lung disease will not necessarily yield results that align with the same factors (e.g., in the case of an obstruction confined to the central airways, FEV_{1}/FVC would not be expected to align with the acinar lung zone factor at all). By contrast, if Dl _{CO} is affected by lung disease, it would always be expected to belong to the acinar lung zone factor.

It is important to consider all correlations between different types of indices, preferably derived from independent techniques, and see if they still fit in the overall pathophysiologic picture. An alternative approach would be to dismiss a new index whenever it shows good correlation with an established one. An obvious pitfall to such an approach is that in another type of disease, in which, for instance, inflammation changes the lumina of intraacinar airways without affecting Dl _{CO}, one could miss an intraacinar alteration that could have become apparent through S_{acin}. Although FEV_{1}/FVC would probably pick it up, this would not reveal whether the obstruction is peripherally located or not.

The interest in having S_{cond} and S_{acin} in addition to traditional lung-function indices is that through the MBW normalized slope analysis from which these indices are derived, they are a direct reflection of differences in gas concentrations between lung units that comprise anything from several clusters of acini up to whole regions (S_{cond}), or between much smaller lung units within the acini (S_{acin}). This contrasts with the VC SBW phase III slope, which has been used as an “early” indicator of small airways disease (10), and as a predictor of subsequent decline in FEV_{1}, with conflicting results (11-13). Although it was generally accepted that part of the phase III slope was due to gravity, recent investigations in sustained microgravity have shown that the contribution of gravity to the VC SBW N_{2} slope amounts to only 22% (14). It was also demonstrated that the VC SBW slope itself is mainly conditioned by events occurring at the extreme volumes (i.e., near closing volume and near TLC) (15). MBW experiments performed during a recent Spacelab Life Science mission (16) showed that the normalized slope curve, such as the one depicted in Figure 1, is unaffected by gravity. This, combined with the fact that the MBW test involves a near-tidal breathing pattern away from the extreme lung volumes, makes it particularly suitable for assessing structural alterations in the lung.

The more tedious analysis of an MBW test needed to obtain indices S_{cond} and S_{acin}, as compared with the analysis of an SBW test or a forced expiration, may restrict its use to targeted research. Three areas are particularly promising. First, research directed at relating structural changes to lung function. With this area limited to former studies in which the SBW phase III slope was used, S_{cond} and S_{acin} could be useful for clarifying the potential correlations between ventilation inhomogeneity and pathology scores provided in the pioneering work of Cosio and colleagues (17) but absent in a later study by Matsuba and coworkers (18). A second area of research could relate S_{cond} and S_{acin} to a new generation of indices that appear to be early indicators of small airways alteration (19, 20). It has, for instance, been shown that the dispersion of an exhaled aerosol bolus, after inhalation of a monodisperse aerosol bolus from just above FRC, could be a sensitive index for detecting early lung alteration in asymptomatic smokers, despite the fact that the exact origin of bolus dispersion is not yet fully understood (20). This type of research also points to the third area in which S_{cond} and S_{acin} could be useful: aerosol drug delivery. In combination with radioaerosol measurements to monitor whether acinar deposition has been achieved, the physiologic effect of such drug delivery on conductive and acinar airways could be estimated through this type of MBW analysis.

In conclusion, we have applied the MBW normalized slope analysis to estimate the contribution of conductive and acinar lung zones to the overall ventilation inhomogeneity in COPD through two indices, S_{cond} and S_{acin}. We used a factor analysis to simplify the complexity of correlations between S_{cond} and S_{acin} and the different lung-function indices, and to reduce them to three underlying and independent factors. It was shown that S_{cond} and S_{acin} indeed reflected independent lung alterations, and the way in which the combination of these MBW indices with the traditional lung-function indices can lead to a more complete pathophysiologic picture was also illustrated. A number of applications in which the computation of these MBW indices could be of interest have also been suggested.

The authors thank Johan Goris, of the Biotechnology Department of AZ-VUB, for technical support.

Supported by the Fund for Scientific Research, Flanders-Belgium, and the Federal Office for Scientific Affairs of Belgium.

1. | Verbanck S., Schuermans D., Van Muylem A., Noppen M., Paiva M., Vincken W.Ventilation distribution during histamine provocation. J. Appl. Physiol.83199718071816 |

2. | Paiva, M., and L. A. Engel. 1989. Gas mixing in the lung periphery. In H. K. Chang and M. Paiva, editors. Respiratory Physiology: An Analytical Approach. Marcel Dekker, New York. 245–276. |

3. | Verbanck S., Paiva M.Model simulations of gas mixing and ventilation distribution in the human lung. J. Appl. Physiol.69199022692279 |

4. | Hair, J. F., R. E. Anderson, R. L. Tatham, and W. C. Black. 1992. Factor analysis. In Multivariate Data Analysis, 3rd ed. Maxwell McMillan International, New York. 223–253. |

5. | Mahler D. A., Harver A.A factor analysis of dyspnea ratings, respiratory muscle strength, and lung function in patients with chronic obstructive pulmonary disease. Am. Rev. Respir. Dis.1451992467470 |

6. | Crawford A. B. H., Makowska M., Paiva M., Engel L. E.Convection-and diffusion-dependent ventilation maldistribution in normal subjects. J. Appl. Physiol.591985838846 |

7. | Crawford A. B. H., Cotton D. J., Paiva M., Engel L. A.Effect of lung volume on ventilation distribution. J. Appl. Physiol.66198925052510 |

8. | Paiva M.Two pulmonary functional indexes suggested by a simple mathematical model. Respiration321975389403 |

9. | Quanjer, H., G. J. Tammeling, J. E. Cotes, O. F. Pedersen, R. Peslin, and J.-C. Yernault. 1993. Lung volumes and forced expiratory flows. Eur. Respir. J. 6(Suppl. 16):5–40. |

10. | Dosman J. A., Cotton D. J., Graham B. L., Hall D. L., Li R., Froh F., Barnett G. D.Sensitivity and specificity of early diagnostic tests of lung function in smokers. Chest791981611 |

11. | Stanescu D. C., Rodenstein D. O., Hoeven C., Robert A.“Sensitive tests” are poor predictors of the decline in forced expiratory volume in one second in middle-aged smokers. Am. Rev. Respir. Dis.1351987585590 |

12. | Olofsson J., Bake B., Svarsudd K., Skoogh B.-E.The single-breath N_{2} test predicts the rate of decline in FEV_{1}. Eur. J. Respir. Dis.6919864656 |

13. | Buist A. S., Vollmer W. M., Johnson L. R., McCamant L. E.Does the single breath N_{2} test identify the smoker who will develop chronic airflow limitation? Am. Rev. Respir. Dis.1371988293301 |

14. | Guy H. J. B., Prisk G. K., Elliott A. R., Deutschman R., Jr, West J. B.Inhomogeneity of pulmonary ventilation during sustained microgravity as determined by single breath washouts. J. Appl. Physiol.76199417191729 |

15. | Dutrieue B., Lauzon A.-M., Verbanck S., Elliott A. R., Prisk G. K., West J. B., Paiva M.Reconstruction of single breath washins from multiple bolus washins under different gravity conditions (abstract). Am. J. Respir. Crit. Care Med.1551997A906 |

16. | Prisk G. K., Guy H. J. B., Elliott A. R., Paiva M., West J. B.Ventilatory inhomogeneity determined from multiple-breath washouts during sustained microgravity on Spacelab SLS-1. J. Appl Physiol.781995597607 |

17. | Cosio M., Ghezzo H., Hogg J. C., Corbin R., Loveland M., Dosman J., Macklem P. T.The relations between structural changes in small airways and pulmonary function tests. N. Engl. J. Med.298197712771281 |

18. | Matsuba K., Shirakusa T., Kuwano K., Hayashi S., Shigematsu N.Small airways disease in patients without chronic airflow limitation. Am. Rev. Respir. Dis.136198711061111 |

19. | Brand P., Tuch T., Manuwald O., Bischof W., Heinrich J., Wichmann H. E., Beinert T., Heyder J.Detection of early lung impairment with aerosol bolus dispersion. Eur. Respir. J.7199418301838 |

20. | Anderson P. J., Hardy K. G., Gann L. P., Cole R., Hiller F. C.Detection of small airway dysfunction in asymptomatic smokers using aerosol bolus behavior. Am. J. Respir. Crit. Care Med.15019949951001 |