The bronchodilator test in chronic obstructive pulmonary disease: Interpretation methods

The bronchodilator test in chronic obstructive pulmonary disease: Interpretation methods

ARTICLE IN PRESS Respiratory Medicine (2007) 101, 34–42 The bronchodilator test in chronic obstructive pulmonary disease: Interpretation methods M. R...

148KB Sizes 2 Downloads 25 Views

ARTICLE IN PRESS Respiratory Medicine (2007) 101, 34–42

The bronchodilator test in chronic obstructive pulmonary disease: Interpretation methods M. Rodrı´guez-Carballeiraa,, J.L. Herediab, M. Rue ´c, S. Quintanad, P. Almagroa a

Department of Internal Medicine, Hospital Mu ´tua de Terrassa, Plaza Dr. Robert, 5, 08221 Terrassa, Barcelona, Spain b Department of Pneumology, Hospital Mu ´tua de Terrassa, Plaza Dr. Robert, 5, 08221 Terrassa, Barcelona, Spain c Universitat de Lleida-Facultat de Medicina, Sabadell, Barcelona, Spain d Department of Intensive Care, Hospital Mu ´tua de Terrassa, Plaza Dr. Robert, 5, 08221 Terrassa, Barcelona, Spain Received 12 July 2004; accepted 22 April 2006

KEYWORDS Chronic obstructive pulmonary disease; Interpretation of bronchodilator test

Summary The objective of the study was to evaluate the best method for interpreting the bronchodilator test (BDT). Five formulas for expressing the BDT results were analyzed and compared: changes experienced by maximum expiratory volume in 1 s (FEV1) and forced vital capacity (FVC) measured in milliliters, in percentage with respect to the baseline, in percentage with respect to the predicted, in percentage with respect to the possible, and in standardized residuals. Ninety-eight chronic obstructive pulmonary disease (COPD) patients were submitted to a respiratory function test on two different days. On each occasion three spirometries were conducted: basal, post-placebo and post bronchodilator. As a gold standard, a normality interval was defined using the variability experienced with the placebo between the two days of the study. The best formulas according to their sensitivity, specivity and area under receiver operating characteristic (ROC) curve were the ‘‘standardized residuals’’, with a cut point of .3, and the ‘‘percentage with respect to the predicted’’ with a cut point of 6%. & 2006 Elsevier Ltd. All rights reserved.

Introduction Corresponding author. Tel.: +34 93 736 50 50;

fax: +34 93 736 59 50. E-mail address: [email protected] (M. Rodrı´guez-Carballeira).

Spirometry in chronic obstructive pulmonary disease (COPD) patients, conducted prior to and after the administration of a bronchodilator (bronchodilator test, BDT), not only is useful for diagnosis but

0954-6111/$ - see front matter & 2006 Elsevier Ltd. All rights reserved. doi:10.1016/j.rmed.2006.04.018

ARTICLE IN PRESS Interpretation of the bronchodilator test in COPD also has therapeutic and prognostic implications.1–7 The possible association between bronchodilator response and prognosis has not been elucidated: some reviews emphasize that subjects who are hyperresponsive have great diminution of lung function,8,9 but recently the inhaled steroids in obstructive lung disease in Europe (ISOLDE) study3 and Anthonisen et al.4 found that baseline bronchodilator response did not relate to the subsequent decline in lung function. Therefore, the BDT is normally performed in respiratory function laboratories for patients with COPD. However, this test procedure and its interpretation have never been standardized; there is no general agreement about how to report results, or definitions of significant positive response thresholds.10 To determine the best reporting method,the following are required: I. The most appropriate parameter: The maximum expiratory volume in 1 s (FEV1) is the most utilized, since it is considered to have greater sensitivity to detect bronchoreversibility, smaller variability and better reproducibility.11,12 In a previous study, we analyzed FEV1, forced vital capacity (FVC) and peak flow (PEF): FEV1 and FVC showed the minor unexplained variability and better reproducibility, clearly better than PEF.13 So we decided to continue the study considering the variation in both for the interpretation of the BDT. II. The best way to express the change after medication: Many formulas have been proposed for expressing the bronchodilator response (Table 1),1,14–17 but few publications have compared these different methods, and those comparisons are not comprehensive and are done in heterogeneous groups of patients.15,16,18 Our group has analyzed the reproducibility—by the calculation of the intraclass correlation coefficient—and the dependency on the baseline—by the calculation of the Pearson correlation coefficient—of the formulas shown in Table 1. The best results were obtained for the change expressed in milliliters (ML), percentage with respect to the predicted (PCP), percentage with respect to the possible (%POSSIBLE), and standardized residuals (SDR).13 III. The placebo effect: The best way to evaluate the BDT is to take spontaneous variability and the placebo effect into account, with the goal of delimiting a pharmacological effect. IV. The best threshold for determining the existence of significant bronchoreversibility (positive BDT): It is difficult to establish a threshold because bronchodilation is a continuous, not

35 dichotomous, variable.19 There have been multiple proposals, but almost none are based on controlled studies.2,14,20,21 The aim of this study was to analyze the sensitivity, specificity and area under the receiver operating characteristic (ROC) curve and to determine the best cut point of differents formulas for discriminating between significant and non-significant bronchodilator responses in the BDT. The formulas studied are the more reproducible and less dependent on the baseline (ML, PCP, %POSSIBLE and SDR), Furthermore, we included the PC because it is most widely used in the clinical practice.

Methods The study included patients with COPD, defined by clinical and spirometric criteria (regular coughing and expectoration lasting more than 3 months/year during the last 2 years, FEV1o70% with respect to the predicted, and FEV1/FVCo.70) who came to our laboratory for a spirometry. The inclusion criteria were that their bronchopathy was stable (no decompensation in the previous 2 months), that they gave their informed consent to participate in the study and that the variation in their basal FEV1 between the two days of the study was less than 15% (as a spirometric criteria of stability). One hundred and four COPD patients were studied on two occasions, 7–14 days apart. On each day of the study, a basal spirometry was conducted, another 30 min after receiving two inhalations of placebo, and a third 30 min after receiving 1500 mg of terbutaline. Bronchodilator treatment and caffeine intake was suspended 12 h before the study. The spirometries were performed using the European Respiratory Society protocol.14 The medication was administrated using a pressurized cartridge and a 750 mL spacer device (NebuhalerR). A bell spirometry was used (SensorMedicsR PEFT Horizon). The ‘back extrapolation’ method was considered for determining the start of test. The values developed by Roca et al.22 for the Mediterranean population were taken as the reference values.

Statistical analysis The formulas that showed better behavior with regard to variability, reproducibility and dependence on the basal value in the previous study mentioned above13 were analyzed comparing their sensitivity, specificity and area under the ROC curve.















Percentage with respect to the achievable

Percentage with respect to the maximal

Percentage with respect to the possible

Standardized residualsy

Percentage with respect to the baseline modified

Percentage with respect to the predicted

Percentage with respect to the baseline

Change in milliliters

Measure of change

ðFEV1 post BD  FEV1 pre BDÞ  100 , FEV1 maximum  FEV1 pre BD where the maximum FEV1 is obtained based on multiple spirometries carried out in each patient

ðFEV1 post BD  FEV1 pre BDÞ  100 , Maximum_increase FEV1 where the maximum increase in FEV1 is calculated based on repeated BDTs for each subject

ðFEV1 post BD  FEV1 pre BDÞ  100 FEV1 predicted  FEV1 preBD

    FEV1 post BD  FEV1 predicted FEV1 pre BD  FEV1 predicted  sd of FEV1 predicted sd of FEV1 predicted

ðFEV1 post BD  FEV1 pre BDÞ  100  2 FEV1 pre BD þ FEV1 post BD

ðFEV1 post BD  FEV1 pre BDÞ  100 FEV1 predicted

ðFEV1 post BD  FEV1 pre BDÞ  100 FEV1 pre BD

Post BD FEV1pre BD FEV1



BD: bronchodilator.  All this formulas were also applied to the changes of FVC. y The SDR represents the number of SD that a given spirometric index is away from the predicted value for this index. Its a dimensionless measure without age, sex and height biases.18,24



Different formulas for interpreting the bronchodilator test.


Table 1

ARTICLE IN PRESS M. Rodrı´guez-Carballeira et al.

ARTICLE IN PRESS Interpretation of the bronchodilator test in COPD


We considered the ‘natural variability’ as the standard deviation (SD) of the variation seen in the spirometric indices after placebo administration during both days of the study. Subsequently, the Pearson’s correlation coefficient between this SD and the mean basal FEV1 was determined to analyze their relation with the basal value. The result shows association between the ‘natural variability’ of FEV1 and the grade of basal obstruction. So that to analyze the formulas with FEV1, the patients were subdivided into three groups on the basis of function of their basal FEV1: one group with FEV1o 1000 mL; the second with FEV1 between 1000 and 2000 mL and the third with FEV142000 ml. In order to define a reference standard for the interpretation of the BDT, we calculated a normality interval (NI) for each group of patients, according to their natural variability.

Table 2

This interval was calculated using the formula proposed by Chinn.23 This formula uses the intraindividual standard deviation obtained from a repeated-measure model to create a reference interval for changes. Thus, a threshold can be defined, beyond which changes in FEV1 and FVC may be attributed to a pharmacological effect. The upper limit of the NI is the point from which the changes experienced after the BD is due to a real therapeutic effect (positive BDT). The sensitivity and specificity of the different methods for interpreting the BDT were calculated taking the NI as a gold standard and their areas under the ROC curve were compared. The area under the curve was calculated using the Mann–Whitney U-test and its 95% confidence interval from the standard error (SE) using the formula proposed by Hanley and McNeil.24 The comparison

Spirometric basal characteristics of patients.


1st day

Age mean yr (SD) Male (%) Smokers (continous and quitters) FEV1(L) FVC(L) FEV1 (% predicted) FVC (% predicted) FEV1/FVC

2nd day 65 (10) 91 (93%) 96%

1.38 2.64 49 68 .52

(7.48) (7.68) (15) (14) (7.11)

1.42 2.70 50 70 .52

(7.48) (7.69) (15) (14) (7.12)

Results expressed as mean (7SD).

Table 3

Definition of the normality interval.


Increase with placebo


Upper limit of the NI

Total patients with COPD (N ¼ 98) FEV1 FVC

19 29

85 130

185 284

COPD patients with FEV1o1000 mL (N ¼ 23) FEV1 FVC

28 42

46 129

118 295

COPD patients with FEV1 1000–2000 mL (N ¼ 65) FEV1 FVC

10 21

90 133

186 280

COPD patients with FEV142000 mL (N ¼ 10) FEV1 FVC

46 45

116 124

273 287

Values are expressed in mL. SDi: intrasubject standard deviation.


Of the 104 patients included, four were excluded because they did not complete the study and two more because they did not meet the stability criteria. The description of the baseline and spirometric characteristics of the 98 patients who were included in the study appears in Table 2. Upon analyzing the relation between the natural variability and the basal value, no relationship was detected when analyzing FVC, but a significant association was detected in the case of FEV1 (r ¼ :37, P ¼ :0001). Thus, we calculated an NI after applying the formula suggested by Chinn both to the group of patients with COPD and to the three subgroups in the sample, defined by their basal FEV1, in order to get more homogeneous groups. Table 3 shows the variation in FEV1 and FVC experienced with placebo, and the resulting NI in each case. The upper limit of the NI defined for FVC varies only slightly in relation to basal FEV1. In the case of FEV1, it increases with basal FEV1, according to the analysis of natural variability described in the previous section. For this reason, we decided to utilize the threshold given by the NI of each subgroup of patients as a reference standard. For FVC we used the same threshold for all COPD patients. Considering this standard of reference, of the 196 BDT performed, 65 (33%) were positive according to the FVC criteria of significant bronchoreversibility, 84 (43%) were positive according to the FEV1 criteria of significant bronchoreversibility and 49 (25%) were positive according to both criteria. Considering one or both criteria 98 (50%) of BDT were positive. We compared the three best methods for expressing the BDT based on their reproducibility and dependency on the baseline: %POSSIBLE, PCP and SDR. Changes measured in milliliters (ML) were excluded from the comparison, since in the Methods section it was shown that when FEV1 is used, the threshold should depend on the basal FEV1. On the other hand, since this system of

Changes of FVC with the BDT 1.00 RSD %POSSIBLE

0.75 PCP



measure is used to define the reference standard, the comparison could be biased. When measuring both FEV1 and FVC, the ROC curves defined for each of the interpretation methods showed a greater area and a narrower confidence interval for the SDR, followed by the PCP, both clearly superior to the %POSSIBLE (Fig. 1). When comparing these two curves, SDR was significantly superior (Po:05). Upon analyzing the results in the three defined COPD subgroups, differences were not seen with regard to those systems that had the best interpretations (Table 4). Only the %POSSIBLE when measuring FVC deviates from this rule, since the area diminishes significantly in the FEV142 L group.



0.00 0.00


0.50 Specificity



Changes of FEV1 with BDT 1.00





among the best curves was made applying a Z-test according to the method proposed by Hanley and McNeil.25 In order to determine the best cut point for discriminating between positive and negative responses to the bronchodilator multiple correlative points were tested, with sensitivity and specificity calculated for each. The data were analyzed using the SPSS-PC statistical package (version 8.0).

M. Rodrı´guez-Carballeira et al.



0.00 0.00


0.50 Specificity



Figure 1 ROC curves of the different formules to express the changes experienced by FEV1 and FVC with the bronchodilator test.

ARTICLE IN PRESS Interpretation of the bronchodilator test in COPD Table 4


Sensitivity and specificity of different methods for interpreting the bronchoditlator test.

Group parameter




Area under ROC curve (CI 95%)


6% 10% .3SDR 12 6% 15% .3SDR 5

82.7 82.7 89.8 81.6 75.0 83.7 83.7 83.7

89.0 75.5 80.6 84.7 86.0 79.6 76.5 67.3

.92 .87 .94 .79 .88 .86 .90 .89

(.89–.95) (.82–.92) (.92–.96) (.70–.88) (.80–.92) (.80–.92) (.86–.94) (.85–.93)


4% 6% .3SDR 13 4% 10% .3SDR 5

96.4 96.4 92.9 96.4 92.9 92.9 89.3 96.4

93.3 93.3 100 86.7 93.3 86.7 80.0 60.0

.96 .96 .96 .95 .96 .97 .94 .94

(.94–.98) (.94–.98) (.94–.98) (.93–.97) (.94–.98) (.95–.99) (.91–.97) (.91–.97)


6% 10% .3SDR 12 4% 15% .3SDR 5

80.0 85.0 86.2 80.0 81.5 84.6 83.1 83.1

92.1 79.0 81.0 92.1 74.6 76.2 77.8 71.4

.92 .88 .92 .90 .89 .98 .91 .87

(.89–.95) (.83–.93) (.89–.95) (.86–.94) (.85–.93) (.97–.99) (.88–.94) (.82–.92)


6% 18% .3SDR 12 5% 15% .3SDR 5

100 100 100 87.5 75.5 75.0 75.5 75.0

85.7 85.7 71.4 100 93.0 92.9 93.0 92.9

.98 .96 .99 .99 .90 .76 .90 .88

(.95–1) (.92–1) (.97–1) (.97–1) (.84–.96) (.64–.88) (.84–.96) (.81–.95)

FEV1: maximum expiratory flow in 1 s; FVC: forced vital capacity; FEVPCP: percentage with respect to the predicted value of the FEV1; FVCPCP: percentage with respect to the predicted value of the FVC; FEVPOS: percentage with respect to the possible value of the FEV1; FVCPOS: percentage with respect to the possible value of the FVC; FEVSDR: standardized residuals of FEV1; FVCSDR: standardized residuals of FVC; FEVPC: percentage with respect to the baseline of the FEV1; FVCPC: percentage with respect to the baseline FVC.

In clinical practice, the most used formula to express the result of the BDT is the percentage of the baseline (PC), so we decided to analyze its ROC curve. Considering all patients, when measuring the changes of FEV1 the area of the ROC curve was .79, the cut off 12% with a sensitivity of 81.6 and specificity of 84.7. When measuring the changes of FVC, the area of the ROC curve was .89, the cut off 5% with a sensitibity of 84 and specificity of 67.

We observe that the most adequate significance thresholds for interpreting the response to the BDT, taking the NI of each subgroup as a reference, are very similar for both the FEV1 and the FVC and that they remain practically constant (Table 4). The only exception is in the case of %POSSIBLE: it is observed that the best threshold obtained increases with basal FEV1, on applying this formula to the change of FVC and although more so in the case of FEV1, in each patient subgroup (Fig. 2).


M. Rodrı´guez-Carballeira et al.

Change of FEV1 (%POSSIBLE)





20 Mean

0 -2SD

-20 0.0



1.5 2.0 Basal FEV1 (L)



Change of FVC (%POSSIBLE)


200 2SD

100 Mean




-200 0.0


1.0 1.5 2.0 Basal FEV1 (L)



Figure 2 Change experienced by FEV1 and the FVC after the bronchodilator expressed in percentage with respect to possible (%POSSIBLE) for basal FEV1 (L).

Discussion The lack of unanimity in the interpretation of the BDT was the principal reason for the present study—to analyze the different interpretation formulas proposed and attempt to determine which are the most adequate. We have compared their capacity to discriminate between positive and negative responses to the bronchodilator and we also have analyzed the best cut points. According to their reproducibility, variability and area under ROC curve, the best are PCP and SRS with a cut point of 6% and .3, respectively, for both FEV1 and FVC. The main problem for interpreting the BDT derives from the large spontaneous variability presented by the bronchodilator response. Kerstjens et al.18 observed that only 42% of the patients showed a consistent classification in their acute response to the bronchodilator during several years of follow-up. In this same vein, Anthonisen

and Wright6 upon following 985 patients with COPD for several years observed that up to 70% of the patients had variable bronchodilator responses. The reference standard for assessing whether or not the BDT response is significant should take this variability into account and also the possible placebo effect. Some authors attempted to survey the natural variability of spirometric indices using confidence intervals based on the variation of these parameters.26–28 In our study, repeated measures have enabled us to analyze the intraindividual variability, while adding the placebo has enabled us to measure its influence as well. In order to measure both the intraindividual variability and the placebo effect we apply the formula proposed by Chinn23 to the placebo response. Due to a direct association detected between the degree of basal obstruction and the natural variability experienced by FEV1, we define a specific threshold for each subgroup of patients according to their basal obstruction. It must be pointed out that these results contrast with those of Tweeddale et al.26 who did not find a relationship between basal FEV1 and changes in its absolute value. They proposed a single significance threshold, applicable to all patients, regardless of their basal FEV1. The area under the ROC curve makes it possible to compare the different systems, and interpret the changes that the bronchodilator causes in FEV1 and FVC. The best ways of expressing the BDT are the SDR and the PCP, both of which have a larger area under the ROC curve and narrower confidence intervals. %POSSIBLE shows a worse result in this analysis by having a smaller area under the ROC curve, and less specificity and accuracy. Upon analyzing the behavior of %POSSIBLE, it is important to point out that the best threshold for identifying a positive BDT varies with regard to the degree of basal obstruction, in the case of both FEV1 and FVC, something that does not happen with the other two formulas, which present constant significance thresholds (Table 4). This variation of %POSSIBLE may be because it exaggerates the responses of patients with smaller degrees of obstruction (Fig. 1). This behavior allows us to prefer PCP and SDR. These results partially contradict the conclusions of Dompeling et al.16 who compare six different formulas for expressing the BDT based on the increase in FEV1, and settle on %POSSIBLE and the %ACHIEVABLE. However, it is necessary to point out that this decision is exclusively based on an analysis of reproducibility and dependency on the baseline value. The area under ROC curve for PC was minor than that of SDR and PCP, moreover it has less reproducibility and more dependency on the

ARTICLE IN PRESS Interpretation of the bronchodilator test in COPD baseline value, age, sex and height, so we prefer PCP or RSD. SDR is the other clearly useful formula according to our results. Other works have shown evidence that they take into account the range of reference values, that they have the same scale for different parameters and, in addition, that they do not depend on age, sex or height.17–19 There is no unanimity of opinion or evidence based on contrasted studies for the adequacy of significant thresholds used to interpret the different formulas of expressing the bronchodilator response.5,11,18,29,30 Few attempts have been carried out in order to unify criteria on the basis of scientific studies. Tweeddale et al.26 used confidence intervals—which attempt to delimit natural variability—to extract significance limits for FEV1 and FVC. However, he did not consider either the placebo effect or intraindividual variability. In conclusion, to interpret the BDT in patients with COPD, changes in FEV1 and FVC should be considered. If the absolute value is used, the threshold with FVC should be 300 mL, regardless of the degree of basal obstruction. For FEV1, the significant threshold depends on the basal FEV1: 118 mL for basal FEV1o1000 mL, 186 mL for basal FEV1 between 1000 and 2000 mL, and 273 mL for basal FEV142000 mL. The most precise formulas for interpreting changes in FEV1 and FVC are SDR and PCP. An increase greater than .3 should be considered significant in the case of SDR and greater than 6% in the case of PCP.

References 1. Siafakas NM, Vermeire P, Pride NB, Paoletti P, Gibson J, Howard P, et al. On behalf of the Task Force. ERS Consensus Statement. Eur Respir J 1995;8:1398–420. 2. ATS Statement. Standards for the diagnosis and care of patients with chronic obstructive pulmonary disease. Am J Respir Cirt Care Med 1995;152(Suppl.):S77–S120. 3. Calverly PMA, Burge PS, Spencer S, Anderson JA, Jones PW for de ISOLDE Study Investigators. Bronchodilator reversibility testing in chronic obstructive pulmonary disease. Thorax 2003;58:659–64. 4. Anthonisen NR, Lindgren PG, Tashkin DP, Kanner RE, Scanlon PD, Connet JE for the Lung Health Study Research Group. Bronchodilator response in the Lung Health Study over 11 years. Eur Respir J 2005;26:45–51. 5. Nisar M, Walshaw M, Earis JE, Pearson MG, Calverley PM. Assessment of reversibility of airway obstruction in patients with chronic obstructive airways disease. Thorax 1990;45:190–4. 6. Anthonisen NR, Wright EC, IPPB Trial Groups. Response to inhaled bronchodilators in COPD. Chest 1987;91 (5 suppl): 36S–9S. 7. Postma DS, Vries KD, Koeter GH, Sluiter HJ. Independent influence of reversibility of airflow obstruction and non-




10. 11.






17. 18.





23. 24.



specific hyperreactivity on the long-term course of lung function in chronic airflow obstruction. Am Rev Respir Dis 1986;134:276–80. Anto ´ JM, Vermeire P, Vestbo J, Sunyer J. Epidemiology of chronic obstructive pulmonary disease. Eur Respir J 2001; 17:773–90. Xu X, Rijcken B, Schouten JP, Weiss ST. Airways responsiveness and development and remission of chronic respiratory symptoms in adults. Lancet 1997;350:1431–4. Harf A. How to express the reversibility of bronchial obstruciton? Eur Respir J 1992;5:919–20. Berger R, Smith D. Acute postbronchodilator changes in pulmonary function parameters in patients with chronic airways obstruction. Chest 1988;93(3):541–6. Pennock BE, Rogers RM, McCaffree DR. Changes in measured spirometric indices. What is significant? Chest 1981; 80(1):97–9. Rodrı´guez-Carballeira M, Heredia JL. Prueba bronco-dilatadora en pacientes con EPOC. Arch Bronconeumol Clin 2000;36(6):334–43. Quanjer PhH, Tammeling GS, Cotes JE, Pedersen OF, Peslin R, Yennault JC. Lung Volumes and forced ventilatory flows. Report working party standardization of lung function tests. European Community for steel and coal. Eur Respir J 1993; 6(Suppl. 16):5–40. Weir DC, Burge PS. Measures of reversibility in response to bronchodilators in chronic airflow obstruciton: relation to airway calibre. Thorax 1991;46:35–43. Dompeling E, van Schayck CP, Molema J, Akkermans R, Folgering H, van Grunsven PM, et al. A comparison of six different ways of expressing the bronchodilating response in asthma and COPD; reproductibility and dependence of prebronchodilator FEV1. Eur Respir J 1992;5: 975–81. Editorial. Predicted values: how should we use them? Thorax 1988;43:265–67. Kerstjens HAM, Brand PLP, Quanjer PH, van der BruggenBogaarts BAHA, Koe ¨ter GH, Postma DS on behalf of the Dutch CNSLD Study Group. Variability of bronchodilator response and effects of inhaled corticosteroid treatment in obstructive airways disease. Thorax 1993;48:722–9. Brand PL, Quanjer PH, Postma DS, Kerstjens HAM, Koe ¨ter GH, Dekhuijzen PNR, et al. and the Dutch chronic nonspecific lung disease (CNSLD) study group. Interpretation of bronchodilator response in patients with obstructive airways disease. Thorax 1992;47:429–36. Committe on emphysema of the American College of Chest Physicians. Criteria for the assessment of reversibility in airways obstruction. Chest 1974;65(5):552–3. American Thoracic Society. ATS Statement. Snowbird workshop on standardization of spirometry. Am Rev Respir Dis 1979;119:831–8. Roca J, Sanchis J, Agustı´-Vidal A, Segarra F, Navajas D, Rodrı´guez-Roisı´n R, et al. Spirometric reference values for a mediterranean population. Bull Eur Physiopathol Respir 1986;22:217–24. Chinn S. Repeatability and method comparison. Thorax 1991;46:454–6. Hanley JA, McNeil BJ. The meaning and use of the area under a receiver operating characteristic (ROC) curve. Radiology 1982;143:29–36. Hanley JA, McNeil BJ. A method of comparing the areas under Receiver Operating Characteristic curves derived from the same cases. Radiology 1983;148:839–43. Tweeddale PM, Alexander F, McHardy GJR. Short term variability in FEV1 and bronchodilator responsiveness in

ARTICLE IN PRESS 42 patients with obstructive ventilatory defects. Thorax 1987;42:487–90. 27. Sourk RL, Nugent KM. Bronchodilator testing: confidence intervals derived from placebo inhalations. Am Rev Respir Dis 1983;128:153–7. 28. Nickerson BG, Lemen RJ, Gerdes CB, Wegmann MJ, Robertson G. Within-subject variability and percent change for significance of spirometry in normal subjects and in

M. Rodrı´guez-Carballeira et al. patients with cystic fibrosis. Am Rev Respir Dis 1980; 122:859–66. 29. Barros MJ, Rees PJ. Bronchodilator responses to salbutamol followed by ipratropium bromide in partially reversible airflow obstruction. Respir Med 1990;84:311–5. 30. Anthonisen NR, Wright EC and the IPPB trial group. Bronchodilator response in chronic obstructive pulmonary disease. Am Rev Respir Dis 1986;133:814–9.