Specificity and Sensitivity of Respiratory Impedance in Assessing Reversibility of Airway Obstruction in Children

Specificity and Sensitivity of Respiratory Impedance in Assessing Reversibility of Airway Obstruction in Children* Henry K. Mazurek;f Francois Marchal, MD; Jocelyne Derelle; Riad Hatahet; Denise Moneret- Vautrin; and Pierre M on in Flow in the upper airway wall induces significant error in estimating respiratory impedance by the standard forced oscillation technique in subjects with airway obstruction and may be minimized by oscillating pressure around the subject's head (head generator technique). The aim of this study was to determine whether the latter improves the power of forced oscillations in detecting airway response to bronchodilators in children. Seventy-five children with airway obstruction were studied (ages 5.5 to 15 years old). Fifty-three had asthma and 22, cystic fibrosis. A bronchodilator was administered, and the percent changes in respiratory resistance at 10 Hz (Rrs10), 20 Hz (Rrs20), respiratory compliance (Crs), and resonant frequency (fn) with standard and head generator were compared with the corresponding change in FEV 1• The response was positive in 38 (~% FEV 1 2': 15%) and negative in 37 patients. Data on RrslO, Crs, and fn could not be obtained in 7, 8, and 4 subjects, respectively, for technical reasons. The ~ %Rrs20 was not different between head and standard generator in nonresponders (mean± SEM: -19.0 ± 4.5, vs -11.8 ± 3.1 o/o), but significantly larger with head than standard generator in responders (-54.1 ± 3.0 vs -26.5 ± 2.4%; p<0.001). The optimal decision level determined by Receiver Operation Characteristic analysis showed that, compared with the standard method, the

head generator improved the specificity of Rrs20 (78 vs 65%) with no change in sensitivity (76% for both). Resonant frequency had larger sensitivity with standard than with head generator (91 vs 53%) but slightly lower specificity (70 vs 78%). Finally, ~%Crs was more specific (72 vs 67%) and more sensitive (68 vs 52%) with standard than with head generator. The overall incidence of false results was lower with the head generator than with the standard generator for resistance and lower with the standard generator than with the head generator for fn and compliance. Thus, the head generator improves the diagnostic power of the forced oscillation resistance in establishing the reversibility of airway obstruction, but parameters derived from the reactance may have better diagnostic value with the standard method. (Chest 1995; 107:996-1002)

The forced oscillation technique is popular in children because it is completely noninvasive and requires minimal cooperation 1 -6 A common method consists of oscillating pressure and measuring flow directly at the airway opening, ie, the conventional input impedance technique, that will be referred to as standard generator method (SG). From the measurement of airway pressure and flow, respiratory impedance modulus (!Zrsi) and phase(¢) are calculated, and may conveniently be expressed as respiratory resistance (Rrs=!Zrsl • cos¢) and react-

ance (Xrs=!Zrsl• sin¢), respectively, the in-phase and out-of-phase component of respiratory impedance. From the latter, respiratory compliance may be estimated. It has been recognized that as the impedance of the upper airway wall (Zuaw) decreases with increasing frequency, some flow is lost in upper airway wall motion and leads to an error on Zrs. 7-9 With large respiratory impedances, eg, adults with airway obstruction,10.1 1 infants, 12 and young children, 11 •13 the upper airway artefact is enhanced and may be minimized by oscillating pressure around the head rather than directly at the mouth,ll.l 4·15 the so-called head generator (HG) technique. A recent comparison of the change in Rrs induced by bronchomotor agents in asthmatic children, as measured with SGs and HGs, has shown that the latter provided significantly larger changes than the former, suggesting that the HG improved the sensitivity of the forced oscillation technique. On the other hand, larger changes in fn

*From Laboratoire d'Exploration Fonctionnelles Pediatriques, Service de Pediatrie I, Hopital d'Enfants Service de Medecine D, Centre Hospitalier Universitaire de Nancy, Vandoeuvre les Nancy, France. f Dr. Mazurek was on leave from the Institute of Tuberculosis and Lung Diseases, Rabka, Poland. Manuscript received April 6, 1994; revision accepted August 17. Reprint requests: Dr. Marchal, Laboratorie de Physiologie, Faculte de Medecine de Nancy, 9 Avenue de la Foret de Haye BP 184, Vandoevre-Les-Nancy 54505 France

996

Crs=respiratory compliance; fn=resonant frequency; HG=head generator; ROC=receiver operating characteristic; Rrs=respiratory resistance; SG=standard generator; Xrs=reactance; Zrs=respiratory impedance modulus; Zuaw=upper airway wall impedance

Key words: asthma; bronchodilator; compliance; frequency dependence of resistance; reactance; resistance; resonant frequency; respiratory mechanical impedance

Assessment of Reversibility of Airway Obstruction in Children (Mazurek eta/)

and compliance were found with SG compared with HG, suggesting that the Xrs may provide more useful information with SG than with HG.l 6 However , that study lacked independent criteria to evaluate the airway response to provocation or relaxation . The aim of the present study was to evaluate the practical benefit of using a HG in the routine assessment of the airway response to bronchodilators in children. For that purpose, the respective power of the parameters derived from ZrsSG and ZrsHG in identifying a response to bronchodilator was evaluated in reference to FEY 1 .

0.95 were discarded. The data from at least two measuring periods were averaged. SG and HG measurem ents were done in an unbiased order, and the tim e interval b etween the two techniques was 2 t o 3 min. The real part of impedance (Rrs) was analyzed in term s o f resistance at 10, 20 H z (Rrs10, Rrs20), and the m ean resistance was obtained b y averaging values obtained at all frequencies (Rrs). The c hange of resistance with frequency (S) also was estim ated b y linear regression over the entire frequ ency range. Similarl y, total Crs and inertance (Irs) were derived from Xrs data, assuming Xrs=Irs.w-1 /(Crs.w), where w is angular frequency (w=21r.frequency). Resonant fr equency was calculated as the frequency where Xrs=O. When the Xrs was still negati ve at 32 H z, fn was obtained b y extrapolation.

MATERIA LS AND M ETHODS

The bronchodilator was administered with a m etered-dose inhaler and spacer device. In 26 asthmatic subjects, it was gi ven after a positive airway response to c hallenge, i e, a reduction in FEY 1 of 20% or more by either m ethacholine ( n =12), allergenic ex tracts o f h ouse dust mite (n= 12), or molds (n=2), as described previously 16 In the other children (27 asthm atic subjects and 22 cystic fibrosis patients), the bronchodilator was administered after the baseline m easurements. The following medicati ons were used : salbutamol (100 J.Lg/ kg) in 65 subjects, ipratropium bromide (60 J.Lg ) in 5 subjects, and a mixture o f f enoterol(150 J.Lg) and ipratropium bromide (60 J.Lg) in 5 subjects. The F E Y1, ZrsSG, and ZrsHG were measured before and 5 min after salbutam ol in all children . A positive response was defined om prebronchodilator by an increase in F E Y 1 of 15% o r m ore fr value. Spirometry and respiratory impedance were m easured in a nonpredetermined order, but the two measurem ents of respiratory impedance were always performed in rapid succession.

Subjects Seventy-five children (40 boys and 35 girls) aged 5.5 t o15 years (9.2 ± 0.3 [SE M] years) were studied ; 53 had a history o f sathma. Based on attack r ate, the severity was d fein ed as moderate in 33, mild in 12, and severe in 8, with a baseline F E Y1 of 89.9 ± 2.6, 84.6 ± 5.4, and 67.6 ± 6.3% predi cted , respecti vely 17 The patients l 12 h b efore discontinued their 13-agonist medications for at east the study. The other 22 children h ad cystic fibrosis, and their baseline FE Yt was 62. 6 ±5.1% predicted . The stud y was approved b y the re gional committee on hum an subject experim entation.

Measurement of Forced E xpirator y Flows Forced expiratory volume in 1 s was measured with a num erical fl owmeter (Spiromatic, Gauthier, France). Forced expiratory maneuvers were repeated until three reproducible recordings were obtained . The data from th e flow-volum e curve with the highest sum of forced vital capacity and FEY 1 were used f or calculations.

Measurem ent of Respiratory Im pedance The apparatus for measuring ventilatory im pedance (Pulmosfor, SEF AM ) has been described in detail previously 16•18 and was in conformity with the recomm endations o f the European echanical Respiratory Impedance.19 Briefl y, Working G roup o n M respiratory input impedance was measured u sing 6 to 3 2 H z pseudo-rand om pressure oscillations applied with both standard (ZrsSG) and head generator tec hniques (ZrsHG ). Wi th the form er, pressure variati ons only were applied at the airway opening, and the c heeks were supported . With the latter, they were applied t o a 35-L chamber enclosing the subject's head and the pneum otachograph, and the c heeks were not supported . For both measuremen ts, the child was breathing through amouthpiece and was wearing a nose clip. Airway fl ow was measured with a Fleisch No. 1 pneum otachograph connected to a differential pressure transducer (Honey well ± 35 em H zO ), and the in put pressure was measured with an identical transducer, matched t othe first within 1% of amplitude and 2° of phase up to 32Hz. The comm on m ode rejection r atio o f the flow chan nel was 60 dB at 32 Hz. The pneum otachograph was calibrated b y the integral m ethod , and the pressure channel was calibrated with a p recision fluidmanometer. The accuracy of the apparatus was checked d aily with a ph ysical analog, and th e calibration procedure was repeated whenever the results were found to dev iate by m orethan 5% from the expected value. at 128 Hz for p eriods o f 1 6 s,and a s The s ignals were ampled Fourier analysis was perform ed on 7blocks of 4 s, with 50% overlapping. The coherence function (-y 2) , an index of signal-tonoise ratio, was calculated and impedance data with 'Y2 less than

Protocol

Data Analysis The percent change from prebronchodilator values were calculated f or the following parameters: Rrs10, Rrs20, Rrs, fn , and Crs. The diagnostic value o f Irs was not evaluated, since this parameter previously has been f ound to have little physical meaning with the SG and not t o change significantl y after bronchodilator with the HG. 16 The difference between post- and prebronchodilator values was used for the s lo pe of fr equency depend ence of Rrs. Mean values of these param eters were then computed in children with negative (~ % FEY 1 <15 %) and positive responses (~ % FE Y12:I5 %) . Analysis of variance and linear regression analysis were used as necessary. D ata a re xe pressed as mean ± SEM, unless otherwise indicated. The sensiti vity and specifity o f~ % Rrs10, ~ % Rrs20 , ~ % Rrs, ~S , ~ % fn , and ~ % C r s obtained with H G and SG were computed with reference to ~ % F E Y 1. The incidence of positi ve and negative responses to bronchodilation was calculated and compared b etween m ethods using receiver operating c haracteristic (ROC) analysis. An ROC curve is a graphic presentation of the statistical relationshi p between true-positive and false-positive rate over t a range of possible decision levels. The diagnostic value of a est depends o n its ability to separate patients with a particular condition fr om those without this condition. Each decision l evel has a corresponding pair of true-positive rate and false-positive rate, representing o ne point on the ROC curve. The whole ROC curve shows c hanges in the true-positive rate, ie, sensitivity-as a fun ction of the false-positive rate (expressed as 1 minus s pecificity), over the range of decision levels e xamined 20 This allows for comprehensive comparisons a mong altern ati ve analytical techniques in similar condition (for equal true- or f alse-positive rate) and over all possible decision l evels. 20 On a ROC curve, the optimal cutoff point which discrimin ates a partiea most effi ciently between p atients having o r n ot h ving CHEST / 107 / 4 / APRIL, 1995

997

Table !-Impedance Data Before Bronchodilator With SG and IIG*

SG HG

Rrs10, hPa.1- 1 s

Rrs20, hPa.1- 1 s

Rrs, hPa.1- 1.s

S, Pa.l- 1.s2

fn , Hz

Crs, mL.hPa- 1

8.2±0.4 ll.4±0.8

7.3±0.3 11.3 ±0.8

7.4±0.3 11.6±0.8

-8.2±0.9 7.2 ± 1.6

24.0± 1.8 10.4±0.4

5.6±0.6 6.6±0.4

*For all parameters, p
The overall mean change in FEY 1 induced by bronchodilator drugs was 22.5 ± 3.0%. Thirty-eight subjects responded positively to bronchodilator, and 37 had ~ %FEY 1 less than 15%. Rrs10 could not be determined in seven subjects, because of low )' 2 coherence function. In four subjects, the shape of the reactance frequency function did not allow the determination of resonant frequency . In eight subjects, particularly those with missing low frequency data , the estimated compliance was negative, and these values were discarded. It is also worth noting that such values of compliance were observed in seven children with asthma with FEY 1 response greater than 15%, and in 1 with FEY 1 response less than 15%. The values of the different impedance parameters measured with standard and HG before bronchodilation are listed on Table 1. It may be seen that all coefficients are significantly

different between methods, as previously reported. The general pattern of response to bronchodilator for both methods was a decrease in Rrs at all frequencies , an increase in Crs, and a decrease in fn. Generally also, the changes in Rrs were larger with HG than with SG , while those observed for the parameters derived from the reactance (fn and Crs) were larger for SG than HG. After bronchodilator use , S was less negative with the standard generator (SSG) and less positive with the head generator (SHG). The average percent change in resistance, fn and Crs, and change in S for the children with, respectively, negative and positive responses , is reported in Table 2. It is interesting that ~ % Rrsl0 and ~ % Rrs20 are not significantly different between SGs and HGs in the children not responding to bronchodilator use, while they are significantly different in those with positive response. In contrast, for the other parameters, the difference between SGs and HGs is significant in children with both positive and negative response. There was a significant correlation between ~ %FEY 1 and the corresponding changes in impedance parameters. As shown in Table 2, the correlation coefficients for resistance are higher with HG than SG. In contrast, the correlation with ~ % fn is better with SG than HG. The diagnostic value of ~ % Rrsl0 and ~ % Rrs20 is compared with that of ~ %FEY 1 in Figure 1. The ROC curve for Rrs is not presented, since it was quite similar to that of Rrs20. Optimal cutoff points for SGs and HGs were, respectively, about -30 and -35% for ~ % Rrs10SG and ~ % Rrs10HG, respectively, and

Table 2-Changes in Impedance Parameters in Children With Negative (FEV 1 <15%) and Positive (FEV1 ?=.15%)

Response to Bronchodilators and Their Correlation Coefficients to Changes in FEV1 FEV1 <15 %

Rrs10, % Rrs20, % Rrs, % S, Pa•l- 1es 2 fn ,% Crs, %

FEV1

Correlation to 6.%FEV 1 *

~15 %

SG

HG

SG

HG

SG

HG

-17.7±3.6 -11.8±3.1 -13.4±2.9 3.5±0.8 -20.4± 3.6 39.5±8.0

-21.6±3.9 -19.0±4.5 -19.8±3.9 -1.7±1.0 -10.7±2.5 10.7±3.7

- 36.7±2.4 -26.5±2.4 -28.2±2.3 9.2 ± l.l -45.8±2.9 ll2.0±21.4

-50.6±3.3 -54.1 ±3.0 -53.1 ±3.1 -8.4 ± 1.8 -19.1 ±4.8 45.0± .ll .1

-0.41 t -0.37 -0.39 034 -0.501 0.40

-0.59f -0.60 -0.61 -0.39 -0.311 0.41§

*All correlations significant at p
998

Assessment of Reversibility of Airway Obstruction in Children (Mazurek eta/)

FIGURE l. ROC curve (A) and distance of the ROC curve (B) from the ideal point for percent changes in RrslO (left panel) and Rrs20 (right panel) with standard (solid symbols) and head generation (HG) (open symbols). A , The shift of the HG ROC curve to the left of the standard generator curve for RrslO and Rrs20 indicates better sensitivity at a given incidence of false-negative results for the former. B, The distance from the ideal point is lower and corresponds to a higher d ecision level for HG than SG for both RrslO and Rrs20. Furthermore, the distance for HG curve remains minimal between -35 and -50% for RrslO and between -40 and -55% for Rrs20.

A

~

100

100

50

50

~ ·;;:

:~

.,c:: "'

0

0

25

50

75

0

100

100

!-specificity (%)

!-specificity (%)

B

.,

~

'f3

1,00

1,00

0,50

0,50

0,00

0

-25

-50

-75

-100

0,00

0

-25

-50

-100

-75

Rrs20 (%change)

Rrs10 (%change)

about -20 and -40% for Ll%Rrs20SG and Ll%Rrs20HG, respectively. Moreover, it can be seen that the shape of the curves are different: the distance increases rapidly above optimal cutoff with Ll %RrsSG10 or 20, but it plateaus with Ll%RrsHG. Hence, cutoff points of -50% for RrslOHG and -55% for Rrs20HG may reasonably be taken without altering the sensitivity of the method. The sensitivity of Ll%Rrsl0 was lower for SG than HG. The sensitivity of Ll%Rrs20SG at the cutoff point of -20% was similar to that of Ll%Rrs20HG at the cutoff point of -40% . For both Ll%Rrsl0 and Ll%Rrs20, the specificity was better with HG than SG . Consequently, the incidence of false results was smaller for the former than for the latter (Table 3). The Ll%fnSG had a surprisingly good sensitivity, with a specificity equivalent to that of Ll%RrsSG10, and low incidence of false results. The Ll%fnHG had comparatively lower sensitivity but larger specificity (Fig 2 and Table 3). Finally, changes in compliance with SG were both more sensitive and more specific than with HG (Fig 2 and Table 3).

rized as follows. First, the mean change in the resistance at 10 and 20 Hz was significantly larger with head than with standard generator (SG) in children with positive response to bronchodilator but not in those with negative response. Second, the best correlations to Ll%FEV 1 were observed for the changes in resistance with HG and in fn with SG. Third, the ROC analysis showed that compared with the standard method the changes in resistance with the HG were more specific and usually more sensitive, with a smaller overall incidence of false results. In contrast, changes in the parameters derived from the Xrs, particularly fn, were more sensitive with a lower incidence of false results, with SG than HG. The characteristics of the frequency response of Zrs after bronchodilation are in conformity with prior studies for both HG and SG.3•5•16 The interesting finding that the changes in resistance at 10 and 20Hz are significantly larger with HG than SG in the case of a positive response to bronchodilator indicates that the HG technique does improve the sensitivity of the forced oscillation resistance, without altering its specificity, since the corresponding changes are not different between SG and HG in nonresponsive children. This is supported by the fact that the cor-

DISCUSSION

The findings of the present study may be summa-

Table 3-Sensitivity, Specificity, and False Results of Impedance Parameters at Optimal Decision Level as DeterOmined by Receiver Operating Characteristic Analysis* HG

SG

A%Rrs10 A%Rrs20 A%Rrs AS A%fn A%Crs

DL

SE

SP

FR

DL

SE

SP

FR

::S-30% ::S-20% ::s-20%

74 76 76 66

70 65 63 73

28 29 30 31

::S-35% -40% ::S-40% ::S-3 Pa.1- 1 -s2

83 76 74 74

76 78 81 76

21 23 22 25

91 68

70 72

20 30

::S-20%

53 52

78 67

34 40

~6

Pa el- 1s2 ::S-30% ~40 %

~20 %

*All data are reported as percentages. Abbreviations are DL, decision level; SE, sensitivity; SP, specificity; FR, false results. CHEST / 107 / 4 / APRIL, 1995

999

FIGURE 2. ROC curve (A) and distance of the ROC curve (B) from the ideal point for percent changes in fn (left panel) and Crs (right panel) with standard (solid symbols) and HG (open symbols). A , ROC curves of head and ~tandard generator (SG) are close to each other for fn and Crs. B, For both parameters, the distance is lower for SG than HG and also corresponds to a higher decision level.

A 100

50

0

0

25

!-specificity (%) B

1,00

., ~

.

:a

0,50

0,00

0

-25

-50

fn (%change)

-75

-100

o.oo

0

25

relation with changes in FEV 1 is stronger for RrsHG than RrsSG. There appears to be a striking parallel between our findings and the study by Konig et aP where the only statistically significant correlation with change in FEV 1 after bronchodilator use was observed for the resistance at 6 Hz. Smaller changes in Rrs at high frequency compared to low frequency were also reported by Buhr et al 5 in response to bronchomotor agents. In these studies, transrespiratory pressure was varied with the conventional technique. In tracheostomized rabbits, methacholine induces a similar change in the resistance measured at high and low frequency. 23 It is, therefore, clear that minimizing the upper airway wall motion improves the relationship between the changes in resistance and FEV 1 resulting from bronchodilation (Table I). ROC analysis showed that the optimal cutoff point that defines a positive response to bronchodilator use is a change in RrsSGlO or RrsSG20 of, respectively, -30 and -20%. These figures are close to twice the coefficient of variation usually reported in normal subjects.1·2·4·6·24 A variation of this magnitude is often used as the threshold defining a positive bronchomotor response. 25 Other investigators, however, use a variation of three to four times the mean coefficient of variation. 3 Indeed, larger coefficients of variation may be observed in patients compared with healthy subjects. 26 In the present study, the HG allows a larger change in resistance to define the threshold response, without decreasing the sensitivity. Consequently, the incidence of false results estimated with ~ % RrsHGIO or ~%RrsHG20 is lowered, compared with the standard method. Whatever method is used for measuring resistance, a complete agreement with FEV 1 is not reached. Indeed, FEV1 and high frequency Rrs do not explore the same mechanical properties of the airways. An 1000

50

75

100

Crs (%change)

increase in resistance may be rP!ated to the changing caliber of any part of the airways, including the larynx,27 while the change in FEV 1 is related to expiratory flow limitation and hence intrathoracic airways mechanisms. One would, thus, suspect that the latter would generally be more specific and perhaps less sensitive than Rrs. Moreover, forced expiratory maneuvers per se may alter bronchomotor tone and thus further complicate the comparability of these methods. In our study, the effect of volume history was likely to be similar on both estimates of respiratory impedance, since SG and HG techniques were always performed in close succession. Frequency dependence of resistance was found to change in the opposite way for SG and HG and became less negative with SG and less positive with HG after bronchodilation. These findings are in agreement with previous reports in children and moderately obstructed patients, where the negative frequency dependence of Rrs with the standard method tended to disappear with the HG.ll.l 6 Negative frequency dependence of Rrs as well as increased fn with the standard method in the presence of airway obstruction is exaggerated by upper airway wall motion, 10 but may also reflect pulmonary inhomogeneity,28 intrathoracic airway shunting, or both.29 Both these parameters have been reported to separate between patients and healthy subjects.l 1·30 However, frequency dependence of resistance alone was not found suitable to distinguish healthy subjects from patients with bronchial hyperresponsiveness. 5·25 In our study, fn obtained with the standard method appears as a surprisingly good parameter to estimate the bronchodilator response. In the study of Hayes et al ,31 fn was significantly higher in asymptomatic smokers than in normal subjects. Consequently, it was described as being highly sensitive to incipient Assessment of Reversibility of Airway Obstruction in Children (Mazurek eta/)

airway obstruction. 3 ·30 The decrease in fn with bronchodilator use may be explained by the increase in the apparent compliance of the respiratory system. Respiratory compliance estimated at high frequency bears little information on the change in elastic properties of lung tissues, as could be expected from measurements of the pressure-volume relationship of the lung under static conditions. Rather, an increase in Crs at these frequencies after bronchodilator use is likely to reflect an improved distribution of ventilation, or a decreased flow shunting from the central intrathoracic airway wall. In some children, the estimation of compliance had no physical meaning because low frequency data, which are important to the correct estimation of compliance, were missing, as a result of poor signal-to-noise ratio. The clinical value of Crs may be questioned, as its estimation was not consistent. Also, this parameter was estimated according to a lumped second-order model, a far from accurate description of the respiratory system of obstructed patients. The change in fn appears to be a better index in assessing airway response, since it is obtained directly from the raw data, does not rely on fitting to a theoretic model, and appears to have a better power of decision. An interesting alternative to estimating Crs would be the raw Xrs data at the lowest frequencies available. Several practical conclusions may be drawn from this study. The most useful part of the frequency spectrum for resistance appears to be the l 0- to 20-Hz range. The low frequency limit is set by the signalto-noise ratio, which depends on the harmonic content of the subject's breathing. The high frequency limit depends on the dynamic response of the pressure transducers and the common mode rejection ratio of the flow channel and on the upper airway artifact. The effect of upper airway wall motion is, therefore, to minimize the change in Rrs and exaggerate that of Xrs. In the presence of airway obstruction, the latter will minimize the response at high frequency with SG and exaggerate it with HG. The RrsHG was in general more specific and sometimes more sensitive to detect bronchodilation, compared with RrsSG, and the overall incidence of false results was smaller. Changes in resonant frequency by SG had a better concordance with .::l%FEV 1 than by HG. Resonant frequency is relatively easy to measure and could thus be used as an additional criterion to evaluate the airway response to bronchodilators.

REFERENCES

2

3

4

5

6

7

8

9 10

ll

12

13

14

15

16

17

18

19 20

ACKNOWLEDGMENTS: The authors thank C. Chone, G. Colin, B. Delorme, F. Fortin, 0. Lacome, and S. Meline for helpful technical assistance; C. Creusat for the typing; and R. Peslin for advice and discussion.

21

Aronsson H, Solymar L, Dempsey J, et al. A modified forced oscillation technique for measurements of respiratory resistance. J Appl Physiol1977; 42:650-55 Konig P, Hordvik NL, Pimmel RL. Forced random noise resistance determination in childhood asthma. Chest 1984; 86:884-90 Duiverman EJ, Nejiens HJ, Vander Snee-Van Smalen M, et al. Comparison of forced oscillometry and forced expiration for measuring dose-related responses to inhaled methacholine in asthmatic children. Bull Eur Physiopath Respir 1986; 22:433-36 Lebecque P , SpierS, Lapierre JG, et al. Histamine challenge test in children using forced oscillation to measure total respiratory resistance. Chest 1987; 92:313-18 Buhr W, Jorres R, Berdel D, et al. Correspondence between forced oscillation and body plethysmography during bronchoprovocation with carbachol in children. Pediatr Pulmonol 1990; 8:280-88 Holmgren D, Engstrom I, Bjure J, et al. Respiratory resistance and transcutaneous P0 2 during histamine provocation in children with bronchial asthma. Pediatr Pulmonol1993; 15:168-74 Michaelson ED, Grassman ED, Peters WR. Pulmonary mechanics by spectral analysis of forced random noise. J Clin Invest 1975; 56:1210-30 Peslin R, Duvivier C, Jardin P. Upper airway walls impedance measured with head plethysmograph. J Appl Physiol 1984; 57:596-600 Cauberghs M, Van de Woestijne KP. Mechanical properties of the upper airway. J Appl Physiol1983; 55:335-42 Peslin R, Duvivier C, Gallina C, et al. Upper airway artifact in respiratory impedance measurements. Am Rev Respir Dis 1985; 132:712-14 Cauberghs M, Van de Woestijne KP. Effect of upper airway shunt and series properties on respiratory impedance measurements. J Appl Physiol1989; 66:2274-79 Marchal F, Peslin R, Duvivier C, et al. Mechanics of the ventilatory system in sedated infants: forced oscillations versus single breath method. Pediatr Pulmonol 1988; 5:9-26 Marchal F, Haouzi P, Peslin R, et al. Mechanical properties of the upper airway wall in children and their influence on respiratory impedance measurements. Pediatr Pulmonol 1992; 13:28-33 Peslin R, Duvivier C, Didelon J, et al. Respiratory impedance measured with head generator to minimize upper airway shunt. J Appl Physiol1985; 59:1790-95 Marchal F, Peslin R, Duvivier C, et al. Measurement of ventilatory mechanical impedance in infants using a head pressure generator. Pediatr Pulmonol1989; 7:209-16 Marchal F, Mazurek H, Habib M, et al. Input respiratory impedance to estimate airway hyperreactivity in children: standard method versus head generator. Eur Respir J 1994; 7:601-07 Zapletal A, Motoyama EK, Van de Woestijne K, et al. Maximum expiratory flow-volume curves and airway conductance in children and adolescents. J Appl Physiol1969; 26:308-16 Peslin R, Marchal F, Duvivier C, et al. Evaluation of a modified head generator for respiratory impedance measurements. Eur Respir Rev 1991; 1:140-45 Van de Woestijne K. Recommendations and future plans. Eur Respir Rev 1991; 1:236-37 Robertson EA, Zweig MH. Use of receiver operating characteristic curves to evaluate the clinical performance of analytical systems. Clin Chern 1981; 27:1569-74 Oilier S, Osman J, Hordle DA, et al. Skin prick test preparation of Dermatophagoides pteronyssinus for prediction of a positive response to provocation testing. Clin Exp Allergy 1989; 19: 445-62 CHEST /107/4/ APRIL, 1995

1001

22 Stafanger G, Kock Andersen J, et a!. Specific diagnosis of exogenous bronchial asthma in children. Allergy 1986; 41: 110-17 23 Belaguid A, Marchal F , Mazurek H , eta!. Low-frequency vs high-frequency respiratory mechanics after methacholine challenge in artificially ventilated rabbits. Pediatr Pulmonol1993; 16:297-302 24 Desager K, Van Bever H , Landser F, et al. Use of the forced oscillation technique in infants. Eur Respir J 1991; 1:248 25 ChinetT, Macquin-Mavier I, Lorino H, eta!. Comparison of the dose-response curves obtained by forced oscillation and plethysmography during carbachol inhalation. Eur Respir J 1988; 1:600-05 26 van den Elshout FJJ, van den Woestijne KP, Folgering HTM. Variations of respiratory impedance with lung volume in

1002

bronchial hyperreactivity. Chest 1990; 98:358-64 27 Collet PW, Brancastisano T, Engel LA. Changes in the glottic aperture during bronchial asthma. Am Rev Respir Dis 1983; 128:719-23 28 Otis AB, McKerrow CB, Bartlett RA, eta!. Mechanical factors in distribution of pulmonary distribution. J Appl Physiol1956; 8:427-43 29 Mead J. Contribution of compliance of airways to frequency dependent behavior of lungs. J Appl Physioll969; 26:670-73 30 Clement J, Landser FJ, Van de Woestijne KP. Total resistance and reactance in patients with respiratory complaints with and without airways obstruction. Chest 1983; 83:215-20 31 Hayes DA, Pimmel RL, Fullton JM, et al. Detection of respiratory mechanical dysfunction by forc ed random noise impedance parameters. Am Rev Respir Dis 1979; 120:1095-1100

Assessment of Reversibility of Airway Obstruction in Children (Mazurek eta/)