- Email: [email protected]
Improved Criterion for Assessing Lung Function Reversibility
[
Original Research Asthma
]
Improved Criterion for Assessing Lung Function Reversibility Helen Ward, MBChB; Brendan G. Cooper, PhD; and Martin R. Miller, MD
Consensus on how best to express bronchodilator reversibility (BDR) is lacking. We tested different BDR criteria against the null hypotheses that BDR should show no sex or size bias. To determine the best criterion for defining BDR, we hypothesized that clinically important BDR should be associated with better survival in respiratory patients compared with that of patients without BDR.
BACKGROUND:
We used the first BDR test of 4,231 patients who had known subsequent survival status (50.8% male sex; mean age, 60.9 years; mean survival, 5.2 years [range, 0.1-16.5 years]). BDR for FEV1 was expressed as absolute change, % baseline change, and change as % predicted FEV1.
METHODS:
Having BDR defined from absolute change was biased toward men (male to female ratio, 2.70) and toward those with larger baseline FEV1. BDR defined by % change from baseline was biased toward those with lower baseline values. BDR defined by % predicted had no sex or size bias. Multivariate Cox regression found those with FEV1 BDR . 8% predicted (33% of the subjects) had an optimal survival advantage (hazard ratio, 0.56; 95% CI, 0.45-0.69) compared with those with FEV1 BDR ⱕ 8% predicted. The survival of those with FEV1 BDR . 8% predicted was not significantly different from that of those with FEV1 BDR . 14% predicted but was significantly better than that of those with FEV1 BDR , 0. RESULTS:
We have shown that expressing FEV1 BDR as % predicted avoids sex and size bias. FEV1 BDR . 8% predicted showed optimal survival advantage and may be the most appropriate criterion to define clinically important reversibility. CHEST 2015; 148(4):877-886
CONCLUSIONS:
Manuscript received September 30, 2014; revision accepted March 26, 2015; originally published Online First April 16, 2015. ABREVIATIONS: BD 5 bronchodilator; BDR 5 bronchodilator reversibility; HR 5 hazard ratio; ICS 5 inhaled corticosteroid; 1ve 5 positive; SR 5 standardized residual AFFILIATIONS: From the Department of Respiratory Medicine (Drs Ward and Cooper), Queen Elizabeth Hospital; and the Institute of Occupational and Environmental Medicine (Prof Miller), University of Birmingham, Birmingham, England.
journal.publications.chestnet.org
FUNDING/SUPPORT: The authors have reported to CHEST that no funding was received for this study. CORRESPONDENCE TO: Martin R. Miller, MD, Institute of Occupational and Environmental Medicine, University of Birmingham, Birmingham, B15 2TT, England; e-mail: [email protected] © 2015 AMERICAN COLLEGE OF CHEST PHYSICIANS. Reproduction of this article is prohibited without written permission from the American College of Chest Physicians. See online for more details. DOI: 10.1378/chest.14-2413
877
Reversibility testing of lung function using short-acting bronchodilators (BDs) is commonly undertaken in lung function laboratories, usually by recording any change in the FEV1. There are numerous guidelines as to how to determine whether the observed change in FEV1 is clinically important1-9 (these are shown in Table 1) and “the lack of a standardized procedure for assessing bronchodilator reversibility has led to significant confusion surrounding this concept.”10 The change may be examined as absolute values, as percentage change from baseline, as change as a percentage of the subject’s predicted value, or as combinations of these. Using change as a percentage of start value potentially biases the results to be positive in those with the lowest start value. Adding an absolute threshold change, based on observations about short-term variability in FEV1,11 was thought to overcome some of this criticism.
To determine how lung function reversibility test results should best be expressed, we tested the methods for expressing lung function reversibility against the null hypotheses that a suitable method should find no systematic difference in reversibility between the sexes and should not be size biased. A clinically important degree of reversibility is likely to be associated with the clinical condition of asthma and should mean that lung function may be improved by relevant drug therapy and that subsequent outcome may be improved. Lung conditions that usually do not have an important degree of reversibility (eg, COPD, pulmonary fibrosis, and bronchiectasis, to name a few) should have a worse prognosis than asthma. Therefore, we also tested the null hypothesis that a clinically important degree of BD reversibility (BDR) has no survival advantage.
Materials and Methods
All tests were conducted according to Association of Respiratory Technicians and Physiologists guidelines12 and the ratio FEV1/FVC was taken from the best FEV1 divided by the best FVC (which could be from separate blow efforts). Subjects referred with FEV1/FVC greater than the lower limit of normal were not routinely tested for BDR. Using the Global Lung Initiative (GLI) 2012 equations, baseline values were converted to standardized residuals (SRs), which are the same as Z scores.13 The SR values were termed FEV1 SR, FVC SR, and FEV1/FVC SR and were derived from (recorded value 2 predicted value)/RSD, where RSD is the residual SD for the scatter of values in the normal reference population.3
All the records of patient visits to our hospital lung function laboratory were extracted from the database (N 5 41,411) on January 11, 2012. The database was set up in 1996, and the use of anonymized data from it for research purposes was approved by the hospital’s Caldicott Guardian, who was the Medical Director overseeing all regulatory issues. From these data, results were obtained for the first reversibility test for white subjects where salbutamol 4 3 100 mg was administered via a spacer device. The data for patients of Asian and Caribbean extraction were not included because of the continued debate around adjustments for ethnicity regarding lung function. There were valid data before and 40 min after BD for FEV1 and FVC for 4,227 separate subjects aged . 20 years with FEV1 ⱖ 0.2 L and FEV1/FVC ⱖ 0.2 (2,147 men [50.8%]) whose National Health Service number was available to determine their survival up to June 13, 2012. There were data for an additional 1,124 subjects referred by family doctors who did not have an National Health Service number and so could not be analyzed further. These 1,124 included significantly more women (55.5% vs 49.2%, P , .001 [x2 test]) than did the study group and they were slightly older (mean ⫾ SD, 62.9 ⫾ 14.5 years vs 60.9 ⫾ 13.6 years; P , .001 [KruskalWallis test]).
Results Table 2 shows the median and interquartile range for pre-BD results for the 4,227 subjects, separated by sex and by survival status. By June 13, 2012, 444 subjects (10.5%) had died, with a significantly higher mortality in the men than in the women (11.6% vs 9.3%, P , .02 [x2 test]). The mean survival was 5.2 years (range, 0.1-16.5 years; median, 4.7 years; interquartile range, 2.7-7.5 years). For both sexes, those who died had, on average, worse lung function than did those alive at the end of the study, but the time at risk was no different in the women who died and was slightly shorter in the men who died. When considering only those who had died, the men had a significantly greater pack-year smoking 878 Original Research
The difference between the largest pre-BD FEV1 and the largest postBD FEV1 was called DFEV1. This was standardized in two ways: (1) as a percentage of the pre-BD (ie, start) value and (2) as a percentage of the subject’s predicted value. Statistical analysis was undertaken using Stata/SE 11.0 (StataCorp LP). Because the distribution of the change in FEV1 after BD was skewed, comparisons were made using rank sum tests. Survival analysis was undertaken using Cox proportional hazards regression, with survival measured from the date of the BD test.
exposure than did the women (P , .05 [rank sum test]), but for age, time at risk, and SR values for lung function, there were no differences between the sexes. In those alive at the end of the study, the men had significantly higher pack-year smoking exposure, shorter time at risk, better FVC SR, and worse FEV1/FVC SR than did the women (P , .05 [rank sum test]). On the request forms, the putative diagnosis or reason for requesting the test had asthma mentioned in 25% and COPD in 30% of subjects. These possible diagnoses could not be verified independently. Most of the remaining indications were symptoms (28%), and 5% were preprocedures such as bronchoscopy or general surgery. Most subjects were already receiving inhaled
[
148#4 CHEST OCTOBER 2015
]
TABLE 1
] Various Guidelines for Assessing Reversibility Lung Function Tests
Study
Body
Guideline
Committee on Emphysema /1974
ACCP
DFEV1 . 15% baseline value
American Thoracic Society2/1991
ATS
DFEV1 or DFVC . 12% baseline value AND . 200 mL
1
Quanjer et al3/1993
ERS
DFEV1 . 9% predicted value
Siafakas et al4/1995
ERS
DFEV1 ⱖ 10% predicted value
British Thoracic Society5/2003
BTS
DFEV1 . 15% baseline value AND . 200 mL
National Institute for Clinical Excellence6/2004
NICE
DFEV1 . 400 mL
ATS/ERS
DFEV1 . 12% baseline value AND . 200 mL
GOLD
DFEV1 . 12% baseline value AND . 200 mL
BTS/SIGN
DFEV1 . 400 mL
Pellegrino et al /2005 7
Global Initiative for Chronic Obstructive Lung Disease8/2007 British Thoracic Society and Scottish Intercollegiate Guidelines Network9/2012
ACCP 5 American College of Chest Physicians; ATS 5 American Thoracic Society; BTS 5 British Thoracic Society; ERS 5 European Respiratory Society; GOLD 5 Global Initiative for Chronic Obstructive Lung Disease; NICE 5 National Institute for Clinical Excellence; SIGN 5 Scottish Intercollegiate Guidelines Network.
BDs, with 60% taking b-agonists, 25% taking short-term anticholinergics, and 8% taking long-acting agents. All these had been withheld prior to testing, but 1,862 (43%) were taking inhaled corticosteroids (ICSs) (47% being men), 190 (4%) were taking oral prednisolone (53% being men), and 279 (6%) were taking oral b-blockers (59% being men), and all these treatments were continued. Taking a b-blocker did reduce the probability of the subject being found reversible by the tested criteria (P 5 .03 [one-sided Fisher exact test]). Absolute change in FEV1 (DFEV1) after BD was right skewed, as shown in Figure 1 (as was DFVC), with negative values obtained for 346 subjects (8%) and 834 subjects (20%), respectively. Table 3 shows the percentage of subjects deemed reversible for various criteria for defining an important degree of FEV1 reversibility, together with the male to female ratio in those reversible and the median values for those reversible and not reversible. The isolated use of an absolute threshold value for change in FEV1 gave a significant bias toward finding men as being reversible and was also biased toward finding reversibility in those with larger absolute FEV1 values. Methods relating change to baseline values were significantly biased toward find reversibility in subjects with lower absolute values for FEV1. Only the % predicted method avoided sex and size biases. The same was found for post-BD change in FVC (data not shown). Results for univariate analysis for predicting mortality are shown in Table 4, with the best predictor being age, followed by FEV1/FVC SR and FEV1 SR. Taking an ICS at the time of testing significantly reduced the hazard ratio (HR) for subsequent mortality, whereas taking
journal.publications.chestnet.org
prednisolone or b-blockers had no significant effect on survival in these subjects. Multivariate analysis using Cox proportional hazards regression was undertaken to predict death, and the significant predictors were age, FEV1/FVC SR, reversibility status, whether the subject was taking ICSs, and pack-years of smoking exposure. To determine the optimum cut level in terms of % predicted for BDR, we derived Cox models using different cut points while stratifying for quartiles of age, FEV1/FVC SR, and pack-years smoking and for ICS use. Figure 2 shows for various cut points the number of subjects deemed reversible and the HR for mortality (with 95% CIs) for those reversible when compared with all those deemed not to be reversible. The survival advantage of being deemed reversible was greatest for a dichotomous cut level of . 8% predicted. Figure 3 shows the proportion of subjects dying and the proportion of subjects who were men within bins of unitary value of reversibility expressed as % predicted. The proportion of men was slightly lower in the bin . 7% to 8% predicted, but there was no obvious change in percentage mortality in this region of reversibility. Figure 4 shows the HR for mortality for all the subjects within bins of unitary value of reversibility expressed as % predicted with the 399 subjects with an increase of . 14% predicted used as the comparator group for survival. The HR was not significantly different from unity until the bin with 8% to , 9% predicted change. Those with less change in FEV1 than this had a higher HR (ie, increased mortality) compared with the people with the most reversibility. Doing this analysis with the comparator group for mortality being those 429 subjects with a reversibility of ⱕ 0% predicted (ie, a fall or no change 879
23.52 to 21.77 22.61 23.15 to 21.40 22.14 FEV1/FVC SR
Significance values are shown for comparing alive vs dead for each sex tested by rank sum tests. IQR5interquartile range; SR 5 standardized residual. P 5 .0001. bP , .005. cP . .05; not significant. dP , .05. a
24.02 to 21.78b 22.70 22.45
23.49 to 21.68
0.40 to 0.62a 0.53 0.62 FEV1/FVC
0.53 to 0.70
0.57
0.44 to 0.65a
a
0.58
0.47 to 0.66
23.46 to 21.55a 22.55 22.13 22.07 FEV1 SR
23.01 to 21.18
22.61
23.40 to 21.77
a
23.00 to 21.29
21.75 to 20.02b 21.01 20.62 21.93 to 20.35a 21.10 21.63 to 0.02 20.78 FVC SR
2.62 to 7.33 4.83 2.79 to 7.78 4.83 Time at risk, y
21.46 to 0.21
2.23 to 6.97d 4.13 2.65 to 7.31 4.70
17.2 to 62.3b 37.8 11.6 to 52.4a 30.5 20.2 Pack-y
0.0 to 41.8
61.5 to 77.2 69.7 60.8 Age, y
51.5 to 70.0
Median
Prebronchodilator Data
IQR
Median
IQR
a
c
30.6
6.2 to 54.0
61.3 to 76.9a 69.9 52.1 to 70.0 61.7
IQR Median IQR Median
Dead (n 5 250 [11.6%]) Men (n 5 2,147) Alive (n 5 1,897 [88.4%]) Dead (n 5 194 [9.3%]) Women (n 5 2,080) Alive (n 5 1,886 [90.7%])
] Median and IQR for Prebronchodilator Data for Each Sex and for Those Alive and Those Dead at the End of the Study TABLE 2
880 Original Research
Figure 1 – Histogram of the absolute change in FEV1 after BD for the 4,227 subjects. BD 5 bronchodilator.
in FEV1 after BD) gave a similar optimum cut point at 8% to , 9% predicted (Fig 5). We then tested the . 8% predicted criterion and five of the endorsed criteria from Table 1 in multivariate Cox regression, with the results shown in Table 5. Using DFEV1 . 8% predicted gave the best model, and only for the two criteria using % predicted thresholds was the prediction contribution by reversibility status greater than that for ICS. The HRs for age, for pack-years, and for taking ICSs were stable across all models. Adding an absolute threshold of 200 mL and then 250 mL to the . 8% predicted criterion did not improve the prediction of mortality, and the survival benefit from BD status worsened from 0.56 to 0.59 and 0.65, respectively. Figure 6 shows the Kaplan-Meier survival curves for the 1,414 subjects meeting the . 8% predicted criterion and the 1,229 subjects meeting the . 12% start value plus . 0.2 L criterion. Those meeting the former criterion had better survival throughout almost all the at risk period. Regression models for the . 8% predicted criterion and the . 12% start value plus . 0.2 L criterion were then run for selected groups of subjects to see how stable the findings were with respect to suggested diagnoses on the request form, sex, age and whether subjects were taking ICSs. Table 6 shows these results. The . 8% predicted criterion returned more positive (1 ve) for BDR (BDR 1ve) with a better survival HR compared with the other criterion in all groups. The . 12% start value plus . 0.2 L criterion returned significantly more men as BDR 1ve in all groups except for where COPD was the suggested diagnosis, where the . 8% predicted criterion found significantly more women BDR 1ve than expected by chance. The greatest survival advantage (HR, 0.25; 95% CL, 0.12-0.49) was for the . 8% predicted criterion in those suspected of having COPD.
[
148#4 CHEST OCTOBER 2015
]
TABLE 3
] Subjects Identified for Defining Reversibility From Change in FEV1 After Bronchodilator, With Male to Female Ratio in Those Deemed Positive, and Median Baseline FEV1 Values
Study/Year
Reversibility Definition
% 1ve
Male to Female Ratio
Median FEV1 1ve
2ve
…
…
1.03
ⱖ 15% baseline
28.4
1.06a
1.35
1.97b
. 12% baseline 1 . 0.2 L
29.1
c
1.36
1.66
1.89b
. 12% baseline
37.5
1.04a
1.42
2.02b
. 0.2 L
38.6
1.60
c
1.92
1.71b
Quanjer et al3/1993
. 9% predicted
27.7
1.09a
1.74
1.83a
Siafakas et al /1995
. 10% predicted
23.0
1.08
a
1.74
1.82a
. 0.4 L
10.2
2.70c
2.20
1.76b
Committee on Emphysema1/1974 Pellegrino et al /2005 7
4
British Thoracic Society and Scottish Intercollegiate Guidelines Network9/2012
1.80
For x2 test on sex ratio and rank sum test for FEV1 comparisons. –ve 5 negative; 1ve 5 positive. aNot significant. bP , .0001. cP , .001.
Discussion We found that expressing BDR by the change in FEV1 as % predicted avoided any sex and size bias and gave the best survival prediction. The current American Thoracic Society/European Respiratory Society recommendation for expressing BDR7 was found to be biased toward giving reversibility status to men and to those with lower FEV1 at baseline, the latter effect having been demonstrated before.14,15 The best survival advantage from having reversibility as a dichotomous variable was seen with DFEV1 . 8% predicted. The first significant survival advantage compared with those who had BDR , 0% predicted (ie, those with TABLE 4
a reduction in FEV1 after BD) was for DFEV1 . 8% predicted, which gave reversibility status to more people (33%). Our finding is close to that of Brand et al,16 who found the best separation (highest likelihood ratio) of asthma from COPD was at . 9% predicted, which has also been used as the criterion by the European Respiratory Society.3 Dales et al14 also showed that . 9% predicted was stable among sex, height, and age subgroups. In subjects younger than ours, a % predicted criterion has been found to best relate to asthma diagnosis.17 Our finding of 8% predicted as the threshold for defining BDR is slightly more inclusive than these two previous studies and, therefore, increases the potential benefit of treatment options to patients. Our finding also held true
] Results of Univariate Cox Regression With Indexes in Order of Their Significance in Predicting Mortality
Predictors
HR
95% CI
X2
Age
1.07
1.06-1.07
226.2
FEV1/FVC SR
0.81
0.76-0.87
36.6
Pack-y
1.01
1.00-1.01
20.8
FEV1 SR
0.87
0.81-0.93
15.7
ICS
0.69
0.57-0.83
15.2
Male sex
1.37
1.13-1.65
10.6
b-Blocker
1.34
0.93-1.93
2.3
FVC SR
0.96
0.90-1.04
1.0
Prednisolone
1.14
0.76-1.70
0.4
Pack-y is pack-y of smoking exposure. ICS, b-blocker, and prednisolone indicate if the subject was taking an ICS or b-blocker or oral prednisolone, respectively, at the time of testing. HR 5 hazard ratio; ICS 5 inhaled corticosteroid. See Table 2 for expansion of other abbreviation.
journal.publications.chestnet.org
Figure 2 – Percentage of the population deemed reversible () on the right-hand ordinate scale against various cut points in DFEV1 expressed as % predicted, together with the hazard ratio for death () for those deemed reversible plotted on the left-hand ordinate scale using Cox regression stratified for quartiles of age, pack-y of smoking, FEV1/FVC standardized residual, and whether on inhaled corticosteroids.
881
Figure 3 – Percentage of patients dying and percentage who were male sex within bins of level of change in postbronchodilator FEV1 expressed as % predicted. The number of subjects in each bin is shown above the percentage male plot.
when using the European Community for Steel and Coal prediction equations3 instead of GLI-2012 for the subjects (data not shown), but this may not be the case for other lung function equations. Factors that may affect studies on BDR include the class and dose of BD used and the timing of the spirometry. The American Thoracic Society/European Respiratory Society statement regarding BDR testing7 gave an example of using 400 mg salbutamol, but the method of administration was not defined. We ensured our data were all consistent by including only those subjects who received a single BD as 400 mg salbutamol via a spacer device. Most previous studies defining BDR have been based mainly on patients with known airways obstruction,11,16-20 whereas Tan et al21 undertook a population sample of . 10,000 subjects aged . 40 years across four continents and showed that the upper 95% CI for FEV1 change was 10.0% predicted (9.5%-10.5%). This was consistent across the sites and was also independent of sex.
Figure 4 – Hazard ratio, with upper and lower 95% CI, for subjects within bins of level of change in postbronchodilator FEV1 expressed as % predicted. The 399 subjects with change . 14% predicted were the comparator group for survival. The number of subjects in each bin is shown.
882 Original Research
Figure 5 – Hazard ratio, with upper and lower 95% CI, for subjects within bins of level of change in postbronchodilator FEV1 expressed as % predicted. The 429 subjects with change ⱕ 0% predicted were the comparator group for survival. The number of subjects in each bin is shown.
Testing for lack of sex bias in BDR testing assumes that the proportion of men vs women who have significant reversibility (ie, those with a higher probability of asthma) should be similar to the proportion in all the subjects undergoing the test. The incidence of asthma is known to be higher in boys than in girls but over adult life the incidence in women is higher than in men.22,23 These patients were aged 20 years to 97 years and although the incidence of asthma may vary across this age range, we believe the prevalence of asthma in this group of patients should not be especially biased toward one sex or the other. In an adult population sample, a higher prevalence of asthma in women has been found,24 and a survey of . 100,000 subjects up to age 52 years (male to female ratio, 0.93) found the prevalence of asthma had a male to female ratio of 0.84.25 We have shown that including an absolute change in FEV1 as a criterion for diagnosing reversibility biased the findings in favor of men in our population of patients, which is opposite to the trend expected from these population surveys. A possible limitation of our study is that we did not have any final clinical diagnoses to verify who should have significant BDR. Using the proposed diagnosis from the referral to analyze subgroups showed the . 8% criterion had less sex and size bias and a better survival advantage than did the . 12% start plus . 0.2 L criterion. However, the strength of our study is that instead of using a clinical diagnosis, which has inherent difficulties of verification, we used the premise that those who truly have a clinically important degree of reversibility will have a better survival because it is related to the condition of asthma, which has a better prognosis than do those conditions without an important degree of reversibility of airflow limitation, such as COPD and bronchiectasis, and subjects with other lung diseases
[
148#4 CHEST OCTOBER 2015
]
For pack-y, the HR is per 10 pack-y of smoking exposure. The Cox models are ordered with the best model to the left. BL 5 baseline FEV1 measurement; Pred 5 Predicted. See Tables 1, 2, and 4 for expansion of other abbreviations. aX 2 for the model.
2.5
24.3 0.54-0.79
1.01-1.06 2.5
0.65 24.2 0.54-0.80
1.01-1.06 2.6
0.66 0.55-0.80
1.01-1.06 1.03
0.66
1.00-1.05
24.1 0.55-0.81
2.4 1.03 1.01-1.05 1.03 2.6
24.2
1.03 Pack-y
0.55-0.80 0.66 Taking ICS
1.01-1.06
0.66
0.55-0.80 24.2 0.67
2.3
24.2
1.03
1.03
26.1
21.9 0.45-1.00
0.74-0.86
23.6 0.52-0.82
0.67
26.7 0.71-0.83
23.8
0.65
0.72-0.84
0.52-0.81 0.65
0.77
0.51-0.79
27.0 0.70-0.81
0.48-0.75 24.5 0.64 0.60 25.3
26.6
Reversible
0.73-0.84 0.78
0.56
FEV1/FVC SR
0.46-0.70
0.78
0.73-0.85 26.4 0.75
24.1
26.7
0.77
0.79
Z
13.3 1.06-1.08
95% CI HR
1.07
Z
13.7
95% CI
1.06-1.08
HR
1.07
Z
13.6 1.06-1.08
95% CI HR
1.07 13.9
Z 95% CI
1.06-1.08
HR
Z
13.8 1.07 1.06-1.08
95% CI HR
1.07
Z
13.8 1.06-1.08
95% CI HR
1.07
Criterion
Age
DFEV1 . 0.4 L BTS/SIGN 20129 433 Reversible (10%) 294a DFEV1 . 15% BL and . 0.2 L BTS 20035 1,005 Reversible (24%) 303a DFEV1 . 12% BL and . 0.2 L ATS/ERS 20057 1,232 Reversible (29%) 305a DFEV1 . 15% BL ACCP 19741 1,200 Reversible (28%) 306a DFEV1 . 9% Pred ERS 19933 1,172 Reversible (28%) 311a DFEV1 . 8% Pred 1,414 Reversible (33%) 320a
Indicating Their Individual Significance for the Predictors
] Number of Subjects Reversible by Six Criteria for Defining Reversibility With HRs for All-Cause Mortality, Their 95% CIs, and Z Values TABLE 5
journal.publications.chestnet.org
Figure 6 – Kaplan-Meier plots for survival in the 1,414 subjects with reversibility . 8% predicted () and the 1,229 subjects with reversibility . 12% start value plus . 0.2 L (line only).
referred for testing. Results from normal subjects could dilute or obscure this effect, but they were unlikely to receive reversibility testing in our laboratory. The fact that we used patient survival to determine BD reversibility status does not presuppose or suggest that reversibility testing is or should ever be used clinically for survival prediction. If physicians gave inhalers only to those subjects they perceived to be reversible, then the improved mortality from BD status may be directly related to clinical practice. However, our physicians were generous in their use of inhalers, with 60% receiving b-agonists and 43% receiving ICSs compared with the 32% we found with BD reversibility . 8% predicted; therefore, we believe our analysis was not defeated by the prevailing clinical practice. A further possible limitation is that we did not have details of the causes of death. Many of the subjects’ deaths will have nonrespiratory causes, but lung function predicts all-cause mortality in the general population26,27 and the signal of improved survival from being reversible was detectable irrespective of the cause of death because the large sample of our data mitigated any uneven distribution of comorbidities or particular diagnoses. If cause of death had been available, this may have helped sharpen the signal in this form of analysis, but not knowing the cause of death does not negate or diminish our findings. All our post-BD results were based on measurements taken on one occasion, so we are unaware of the reproducibility of the result in this population. Calverley et al28 found that . 38% of the patients with COPD included in a study who had previously been shown to be irreversible at their first visit were then reversible at a subsequent visit. However, their criteria for reversibility were based on change as a % of baseline, so small variations in baseline severely affected the repeatability of the result; this is not true for 883
884 Original Research
[
148#4 CHEST OCTOBER 2015
]
2,392
2,147
Age ⱖ 60 y
Male sex
2,080
1,835
Age , 60 y
Female sex
2,261
Other diagnosis
…
…
52.1
49.2
52.1
54.4
48.2
53.5
47.3
% Male
33.1 25.0
. 8% predicted
33.0
. 12% start and . 0.2 L
33.8
. 8% predicted
26.4
. 12% start and . 0.2 L
. 12% start and . 0.2 L
31.4
32.5
. 8% predicted
36.1
. 8% predicted
26.0
. 12% start and . 0.2 L
30.7
. 8% predicted
34.6
. 12% start and . 0.2 L
39.1
. 8% predicted
32.1
. 12% start and . 0.2 L
36.2
. 8% predicted
26.6
. 12% start and . 0.2 L
31.7
. 8% predicted . 12% start and . 0.2 L
35.7 32.1
. 8% predicted
% 1ve
. 12% start and . 0.2 L
Criterion
1.74
1.79c b
1.83
1.87b a
1.90
1.94b c
2.22
2.34b
1.70 1.64 1.51 1.43 1.80 1.72 2.11 1.94
54.3b c
57.3a a
60.7b a
…
…
…
…
62.0b
53.4a
53.0c
48.9
54.8
47.0
48.0
0.53
0.80
0.59
0.81
0.70
0.34
0.25
0.52
0.46
0.42-0.67
0.52-1.25
0.38-0.93
0.61-1.07
0.53-0.92
0.18-0.67
0.12-0.49
0.35-0.78
0.31-0.68
0.61-1.06
0.54-0.92
1.48
1.47
1.82
2.04
0.44-0.77 0.50-0.87 0.36-0.70 0.40-0.84
0.58 0.66 0.51 0.58
c
1.58c
1.59
2.12
2.27b
0.45-0.74
a
a
0.80
0.70
a
1.98b
1.76
0.58
1.93
1.83
59.9b
53.5
0.34-0.69
0.28-0.57
1.55b
a
a
0.48
0.40
1.40
1.72b
1.55
95% CI
1.55b
1.67
1.61
55.3b
HR 1ve
1.39
a
a
48.7
2ve
1ve
% Male
Median FEV1
Significant sex difference or difference in median FEV1 between BDR 1ve and BDR 2ve. BDR 5 bronchodilator reversibility. See Tables 3 and 4 for expansion of other abbreviations. Not significant. bP , .005. cP , .05.
a
1,628
Asthma
338
2,369
Not taking ICS
COPD
1,858
No.
1ve
] Analysis of Subgroups of Subjects by Diagnosis and by Drug Treatment of Two Criteria for Defining BDR
Taking ICS
TABLE 6
BDR defined from change as % predicted. Because these subjects were those routinely referred for reversibility testing as part of their clinical care, our findings should pertain to similar clinical populations. Because we had no data for children, our observations may not necessarily pertain to a pediatric population; however, there is no obvious flaw to the principle that relating change to a predicted value is also true for children.
Conclusions We conclude that there is a lack of consensus across international societies as to how to determine a clini-
Acknowledgments Author contributions: M. R. M. had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis, including and especially any adverse effects. H. W. and M. R. M. contributed to the formulation of the research question and analysis of the data; and H. W., B. G. C., and M. R. M. contributed to the acquisition of the data and the writing of the paper. Conflict of interest: H. W. has participated in speaking activities funded by pharmaceutical industries including Novartis, Boehringer-Ingelheim GmbH, Almirall, GlaxoSmithKline, AstraZeneca, and NAPP. She has also received industry funding for travel to and accommodation at international respiratory conferences. B. G. C. has received payment for consultancy work from pharmaceutical companies for advice on spirometry. M. R. M. has participated in a spoken activity funded by Boehringer-Ingelheim GmbH. Other contributions: The authors thank the lung function physiologists of the Lung Function and Sleep Department at Queen Elizabeth Hospital, Birmingham, for their dedication in obtaining the best results possible for the patients.
References 1. Report of the Committee on Emphysema American College of Chest Physicians. Criteria for the assessment of reversibility in airways obstruction. Chest. 1974;65(5): 552-553. 2. American Thoracic Society. Lung function testing: selection of reference values and interpretative strategies. Am Rev Respir Dis. 1991;144(5):1202-1218. 3. Quanjer PH, Tammeling GJ, Cotes JE, Pedersen OF, Peslin R, Yernault JC; Official Statement of the European Respiratory Society. Lung volumes and forced ventilatory flows. Report Working Party Standardization of Lung Function Tests, European Community for Steel and Coal. Eur Respir J Suppl. 1993;16:5-40.
journal.publications.chestnet.org
cally important degree of BDR. Our study has shown that change in FEV1 as % predicted avoids any sex and size bias and was better at predicting survival advantage, which would help distinguish asthma, which has a better prognosis than other chronic respiratory diseases not associated with clinically important reversibility. Our data also suggested that using change in FEV1 . 8% predicted may be the optimal cutoff because this was the threshold showing the best consistent survival advantage and gave a better yield of reversible subjects than did the previously endorsed3 and validated14,16 criterion of . 9% predicted.
4. Siafakas NM, Vermeire P, Pride NB, et al; The European Respiratory Society Task Force. Optimal assessment and management of chronic obstructive pulmonary disease (COPD). Eur Respir J. 1995;8(8): 1398-1420. 5. British Thoracic Society; Scottish Intercollegiate Guidelines Network. British guideline on the management of asthma. Thorax. 2003;58(suppl 1):i1-i94. 6. Chronic obstructive pulmonary disease: management of chronic obstructive pulmonary disease in adults in primary and secondary care (partial update). National Institute for Clinical Excellence (NICE) website. http://www.nice.org.uk/guidance/ cg101/chapter/1-recommendations. Accessed April 8, 2015. 7. Pellegrino R, Viegi G, Brusasco V, et al. Interpretative strategies for lung function tests. Eur Respir J. 2005;26(5):948-968. 8. Global strategy for the diagnosis, management and prevention of chronic obstructive pulmonary disease. GOLD website. http://www.goldcopd.com. Published 2010. Accessed November 20, 2012. 9. British Thoracic Society and Scottish Intercollegiate Guidelines Network. British guideline on the management of asthma: a national clinical guideline. British Thoracic Society website. http:// www.brit-thoracic.org.uk/Portals/0/ Guidelines/AsthmaGuidelines. Accessed November 2, 2012. 10. Hanania NA, Celli BR, Donohue JF, Martin UJ. Bronchodilator reversibility in COPD. Chest. 2011;140(4):1055-1063. 11. Tweeddale PM, Alexander F, McHardy GJR. Short term variability in FEV1 and bronchodilator responsiveness in patients with obstructive ventilatory defects. Thorax. 1987;42(7):487-490. 12. Guidelines for the measurement of respiratory function. Recommendations of the British Thoracic Society and the Association of Respiratory Technicians and Physiologists. Respir Med. 1994;88(3): 165-194. 13. Quanjer PH, Stanojevic S, Cole TJ, et al; ERS Global Lung Function Initiative. Multi-ethnic reference values for spirometry for the 3-95-yr age range: the global
lung function 2012 equations. Eur Respir J. 2012;40(6):1324-1343. 14. Dales RE, Spitzer WO, Tousignant P, Schechter M, Suissa S. Clinical interpretation of airway response to a bronchodilator. Epidemiologic considerations. Am Rev Respir Dis. 1988;138(2):317-320. 15. Lehmann S, Bakke PS, Eide GE, Humerfelt S, Gulsvik A. Bronchodilator reversibility testing in an adult general population; the importance of smoking and anthropometrical variables on the response to a beta2-agonist. Pulm Pharmacol Ther. 2006;19(4):272-280. 16. Brand PL, Quanjer PH, Postma DS, et al; The Dutch Chronic Non-Specific Lung Disease (CNSLD) Study Group. Interpretation of bronchodilator response in patients with obstructive airways disease. Thorax. 1992;47(6): 429-436. 17. Bussamra MH, Cukier A, Stelmach R, Rodrigues JC. Evaluation of the magnitude of the bronchodilator response in children and adolescents with asthma. Chest. 2005;127(2):530-535. 18. Eliasson O, Degraff AC Jr. The use of criteria for reversibility and obstruction to define patient groups for bronchodilator trials. Influence of clinical diagnosis, spirometric, and anthropometric variables. Am Rev Respir Dis. 1985;132(4): 858-864. 19. Sourk RL, Nugent KM. Bronchodilator testing: confidence intervals derived from placebo inhalations. Am Rev Respir Dis. 1983;128(1):153-157. 20. Tweeddale PM, Merchant S, Leslie M, Alexander F, McHardy GJR. Short term variability in FEV1: relation to pretest activity, level of FEV1, and smoking habits. Thorax. 1984;39(12):928-932. 21. Tan WC, Vollmer WM, Lamprecht B, et al; BOLD Collaborative Research Group. Worldwide patterns of bronchodilator responsiveness: results from the Burden of Obstructive Lung Disease study. Thorax. 2012;67(8):718-726. 22. Becklake MR, Kauffmann F. Gender differences in airway behaviour over the human life span. Thorax. 1999;54(12): 1119-1138.
885
23. Dodge RR, Burrows B. The prevalence and incidence of asthma and asthma-like symptoms in a general population sample. Am Rev Respir Dis. 1980;122(4):567-575. 24. Chittleborough CR, Taylor AW, Dal Grande E, et al; North West Adelaide Health Study Team. Gender differences in asthma prevalence: variations with socioeconomic disadvantage. Respirology. 2010;15(1):107-114.
886 Original Research
27. Chinn S, Gislason T, Aspelund T, Gudnason V. Optimum expression of adult lung function based on all-cause mortality: results from the Reykjavik study. Respir Med. 2007;101(3):601-609. 28. Calverley PM, Burge PS, Spencer S, Anderson JA, Jones PW. Bronchodilator reversibility testing in chronic obstructive pulmonary disease. Thorax. 2003;58(8): 659-664.
25. Leynaert B, Sunyer J, Garcia-Esteban R, et al. Gender differences in prevalence, diagnosis and incidence of allergic and non-allergic asthma: a population-based cohort. Thorax. 2012;67(7):625-631. 26. Miller MR, Pedersen OF, Lange P, Vestbo J. Improved survival prediction from lung function data in a large population sample. Respir Med. 2009;103(3): 442-448.
[
148#4 CHEST OCTOBER 2015
]
Original Research Asthma
]
Improved Criterion for Assessing Lung Function Reversibility Helen Ward, MBChB; Brendan G. Cooper, PhD; and Martin R. Miller, MD
Consensus on how best to express bronchodilator reversibility (BDR) is lacking. We tested different BDR criteria against the null hypotheses that BDR should show no sex or size bias. To determine the best criterion for defining BDR, we hypothesized that clinically important BDR should be associated with better survival in respiratory patients compared with that of patients without BDR.
BACKGROUND:
We used the first BDR test of 4,231 patients who had known subsequent survival status (50.8% male sex; mean age, 60.9 years; mean survival, 5.2 years [range, 0.1-16.5 years]). BDR for FEV1 was expressed as absolute change, % baseline change, and change as % predicted FEV1.
METHODS:
Having BDR defined from absolute change was biased toward men (male to female ratio, 2.70) and toward those with larger baseline FEV1. BDR defined by % change from baseline was biased toward those with lower baseline values. BDR defined by % predicted had no sex or size bias. Multivariate Cox regression found those with FEV1 BDR . 8% predicted (33% of the subjects) had an optimal survival advantage (hazard ratio, 0.56; 95% CI, 0.45-0.69) compared with those with FEV1 BDR ⱕ 8% predicted. The survival of those with FEV1 BDR . 8% predicted was not significantly different from that of those with FEV1 BDR . 14% predicted but was significantly better than that of those with FEV1 BDR , 0. RESULTS:
We have shown that expressing FEV1 BDR as % predicted avoids sex and size bias. FEV1 BDR . 8% predicted showed optimal survival advantage and may be the most appropriate criterion to define clinically important reversibility. CHEST 2015; 148(4):877-886
CONCLUSIONS:
Manuscript received September 30, 2014; revision accepted March 26, 2015; originally published Online First April 16, 2015. ABREVIATIONS: BD 5 bronchodilator; BDR 5 bronchodilator reversibility; HR 5 hazard ratio; ICS 5 inhaled corticosteroid; 1ve 5 positive; SR 5 standardized residual AFFILIATIONS: From the Department of Respiratory Medicine (Drs Ward and Cooper), Queen Elizabeth Hospital; and the Institute of Occupational and Environmental Medicine (Prof Miller), University of Birmingham, Birmingham, England.
journal.publications.chestnet.org
FUNDING/SUPPORT: The authors have reported to CHEST that no funding was received for this study. CORRESPONDENCE TO: Martin R. Miller, MD, Institute of Occupational and Environmental Medicine, University of Birmingham, Birmingham, B15 2TT, England; e-mail: [email protected] © 2015 AMERICAN COLLEGE OF CHEST PHYSICIANS. Reproduction of this article is prohibited without written permission from the American College of Chest Physicians. See online for more details. DOI: 10.1378/chest.14-2413
877
Reversibility testing of lung function using short-acting bronchodilators (BDs) is commonly undertaken in lung function laboratories, usually by recording any change in the FEV1. There are numerous guidelines as to how to determine whether the observed change in FEV1 is clinically important1-9 (these are shown in Table 1) and “the lack of a standardized procedure for assessing bronchodilator reversibility has led to significant confusion surrounding this concept.”10 The change may be examined as absolute values, as percentage change from baseline, as change as a percentage of the subject’s predicted value, or as combinations of these. Using change as a percentage of start value potentially biases the results to be positive in those with the lowest start value. Adding an absolute threshold change, based on observations about short-term variability in FEV1,11 was thought to overcome some of this criticism.
To determine how lung function reversibility test results should best be expressed, we tested the methods for expressing lung function reversibility against the null hypotheses that a suitable method should find no systematic difference in reversibility between the sexes and should not be size biased. A clinically important degree of reversibility is likely to be associated with the clinical condition of asthma and should mean that lung function may be improved by relevant drug therapy and that subsequent outcome may be improved. Lung conditions that usually do not have an important degree of reversibility (eg, COPD, pulmonary fibrosis, and bronchiectasis, to name a few) should have a worse prognosis than asthma. Therefore, we also tested the null hypothesis that a clinically important degree of BD reversibility (BDR) has no survival advantage.
Materials and Methods
All tests were conducted according to Association of Respiratory Technicians and Physiologists guidelines12 and the ratio FEV1/FVC was taken from the best FEV1 divided by the best FVC (which could be from separate blow efforts). Subjects referred with FEV1/FVC greater than the lower limit of normal were not routinely tested for BDR. Using the Global Lung Initiative (GLI) 2012 equations, baseline values were converted to standardized residuals (SRs), which are the same as Z scores.13 The SR values were termed FEV1 SR, FVC SR, and FEV1/FVC SR and were derived from (recorded value 2 predicted value)/RSD, where RSD is the residual SD for the scatter of values in the normal reference population.3
All the records of patient visits to our hospital lung function laboratory were extracted from the database (N 5 41,411) on January 11, 2012. The database was set up in 1996, and the use of anonymized data from it for research purposes was approved by the hospital’s Caldicott Guardian, who was the Medical Director overseeing all regulatory issues. From these data, results were obtained for the first reversibility test for white subjects where salbutamol 4 3 100 mg was administered via a spacer device. The data for patients of Asian and Caribbean extraction were not included because of the continued debate around adjustments for ethnicity regarding lung function. There were valid data before and 40 min after BD for FEV1 and FVC for 4,227 separate subjects aged . 20 years with FEV1 ⱖ 0.2 L and FEV1/FVC ⱖ 0.2 (2,147 men [50.8%]) whose National Health Service number was available to determine their survival up to June 13, 2012. There were data for an additional 1,124 subjects referred by family doctors who did not have an National Health Service number and so could not be analyzed further. These 1,124 included significantly more women (55.5% vs 49.2%, P , .001 [x2 test]) than did the study group and they were slightly older (mean ⫾ SD, 62.9 ⫾ 14.5 years vs 60.9 ⫾ 13.6 years; P , .001 [KruskalWallis test]).
Results Table 2 shows the median and interquartile range for pre-BD results for the 4,227 subjects, separated by sex and by survival status. By June 13, 2012, 444 subjects (10.5%) had died, with a significantly higher mortality in the men than in the women (11.6% vs 9.3%, P , .02 [x2 test]). The mean survival was 5.2 years (range, 0.1-16.5 years; median, 4.7 years; interquartile range, 2.7-7.5 years). For both sexes, those who died had, on average, worse lung function than did those alive at the end of the study, but the time at risk was no different in the women who died and was slightly shorter in the men who died. When considering only those who had died, the men had a significantly greater pack-year smoking 878 Original Research
The difference between the largest pre-BD FEV1 and the largest postBD FEV1 was called DFEV1. This was standardized in two ways: (1) as a percentage of the pre-BD (ie, start) value and (2) as a percentage of the subject’s predicted value. Statistical analysis was undertaken using Stata/SE 11.0 (StataCorp LP). Because the distribution of the change in FEV1 after BD was skewed, comparisons were made using rank sum tests. Survival analysis was undertaken using Cox proportional hazards regression, with survival measured from the date of the BD test.
exposure than did the women (P , .05 [rank sum test]), but for age, time at risk, and SR values for lung function, there were no differences between the sexes. In those alive at the end of the study, the men had significantly higher pack-year smoking exposure, shorter time at risk, better FVC SR, and worse FEV1/FVC SR than did the women (P , .05 [rank sum test]). On the request forms, the putative diagnosis or reason for requesting the test had asthma mentioned in 25% and COPD in 30% of subjects. These possible diagnoses could not be verified independently. Most of the remaining indications were symptoms (28%), and 5% were preprocedures such as bronchoscopy or general surgery. Most subjects were already receiving inhaled
[
148#4 CHEST OCTOBER 2015
]
TABLE 1
] Various Guidelines for Assessing Reversibility Lung Function Tests
Study
Body
Guideline
Committee on Emphysema /1974
ACCP
DFEV1 . 15% baseline value
American Thoracic Society2/1991
ATS
DFEV1 or DFVC . 12% baseline value AND . 200 mL
1
Quanjer et al3/1993
ERS
DFEV1 . 9% predicted value
Siafakas et al4/1995
ERS
DFEV1 ⱖ 10% predicted value
British Thoracic Society5/2003
BTS
DFEV1 . 15% baseline value AND . 200 mL
National Institute for Clinical Excellence6/2004
NICE
DFEV1 . 400 mL
ATS/ERS
DFEV1 . 12% baseline value AND . 200 mL
GOLD
DFEV1 . 12% baseline value AND . 200 mL
BTS/SIGN
DFEV1 . 400 mL
Pellegrino et al /2005 7
Global Initiative for Chronic Obstructive Lung Disease8/2007 British Thoracic Society and Scottish Intercollegiate Guidelines Network9/2012
ACCP 5 American College of Chest Physicians; ATS 5 American Thoracic Society; BTS 5 British Thoracic Society; ERS 5 European Respiratory Society; GOLD 5 Global Initiative for Chronic Obstructive Lung Disease; NICE 5 National Institute for Clinical Excellence; SIGN 5 Scottish Intercollegiate Guidelines Network.
BDs, with 60% taking b-agonists, 25% taking short-term anticholinergics, and 8% taking long-acting agents. All these had been withheld prior to testing, but 1,862 (43%) were taking inhaled corticosteroids (ICSs) (47% being men), 190 (4%) were taking oral prednisolone (53% being men), and 279 (6%) were taking oral b-blockers (59% being men), and all these treatments were continued. Taking a b-blocker did reduce the probability of the subject being found reversible by the tested criteria (P 5 .03 [one-sided Fisher exact test]). Absolute change in FEV1 (DFEV1) after BD was right skewed, as shown in Figure 1 (as was DFVC), with negative values obtained for 346 subjects (8%) and 834 subjects (20%), respectively. Table 3 shows the percentage of subjects deemed reversible for various criteria for defining an important degree of FEV1 reversibility, together with the male to female ratio in those reversible and the median values for those reversible and not reversible. The isolated use of an absolute threshold value for change in FEV1 gave a significant bias toward finding men as being reversible and was also biased toward finding reversibility in those with larger absolute FEV1 values. Methods relating change to baseline values were significantly biased toward find reversibility in subjects with lower absolute values for FEV1. Only the % predicted method avoided sex and size biases. The same was found for post-BD change in FVC (data not shown). Results for univariate analysis for predicting mortality are shown in Table 4, with the best predictor being age, followed by FEV1/FVC SR and FEV1 SR. Taking an ICS at the time of testing significantly reduced the hazard ratio (HR) for subsequent mortality, whereas taking
journal.publications.chestnet.org
prednisolone or b-blockers had no significant effect on survival in these subjects. Multivariate analysis using Cox proportional hazards regression was undertaken to predict death, and the significant predictors were age, FEV1/FVC SR, reversibility status, whether the subject was taking ICSs, and pack-years of smoking exposure. To determine the optimum cut level in terms of % predicted for BDR, we derived Cox models using different cut points while stratifying for quartiles of age, FEV1/FVC SR, and pack-years smoking and for ICS use. Figure 2 shows for various cut points the number of subjects deemed reversible and the HR for mortality (with 95% CIs) for those reversible when compared with all those deemed not to be reversible. The survival advantage of being deemed reversible was greatest for a dichotomous cut level of . 8% predicted. Figure 3 shows the proportion of subjects dying and the proportion of subjects who were men within bins of unitary value of reversibility expressed as % predicted. The proportion of men was slightly lower in the bin . 7% to 8% predicted, but there was no obvious change in percentage mortality in this region of reversibility. Figure 4 shows the HR for mortality for all the subjects within bins of unitary value of reversibility expressed as % predicted with the 399 subjects with an increase of . 14% predicted used as the comparator group for survival. The HR was not significantly different from unity until the bin with 8% to , 9% predicted change. Those with less change in FEV1 than this had a higher HR (ie, increased mortality) compared with the people with the most reversibility. Doing this analysis with the comparator group for mortality being those 429 subjects with a reversibility of ⱕ 0% predicted (ie, a fall or no change 879
23.52 to 21.77 22.61 23.15 to 21.40 22.14 FEV1/FVC SR
Significance values are shown for comparing alive vs dead for each sex tested by rank sum tests. IQR5interquartile range; SR 5 standardized residual. P 5 .0001. bP , .005. cP . .05; not significant. dP , .05. a
24.02 to 21.78b 22.70 22.45
23.49 to 21.68
0.40 to 0.62a 0.53 0.62 FEV1/FVC
0.53 to 0.70
0.57
0.44 to 0.65a
a
0.58
0.47 to 0.66
23.46 to 21.55a 22.55 22.13 22.07 FEV1 SR
23.01 to 21.18
22.61
23.40 to 21.77
a
23.00 to 21.29
21.75 to 20.02b 21.01 20.62 21.93 to 20.35a 21.10 21.63 to 0.02 20.78 FVC SR
2.62 to 7.33 4.83 2.79 to 7.78 4.83 Time at risk, y
21.46 to 0.21
2.23 to 6.97d 4.13 2.65 to 7.31 4.70
17.2 to 62.3b 37.8 11.6 to 52.4a 30.5 20.2 Pack-y
0.0 to 41.8
61.5 to 77.2 69.7 60.8 Age, y
51.5 to 70.0
Median
Prebronchodilator Data
IQR
Median
IQR
a
c
30.6
6.2 to 54.0
61.3 to 76.9a 69.9 52.1 to 70.0 61.7
IQR Median IQR Median
Dead (n 5 250 [11.6%]) Men (n 5 2,147) Alive (n 5 1,897 [88.4%]) Dead (n 5 194 [9.3%]) Women (n 5 2,080) Alive (n 5 1,886 [90.7%])
] Median and IQR for Prebronchodilator Data for Each Sex and for Those Alive and Those Dead at the End of the Study TABLE 2
880 Original Research
Figure 1 – Histogram of the absolute change in FEV1 after BD for the 4,227 subjects. BD 5 bronchodilator.
in FEV1 after BD) gave a similar optimum cut point at 8% to , 9% predicted (Fig 5). We then tested the . 8% predicted criterion and five of the endorsed criteria from Table 1 in multivariate Cox regression, with the results shown in Table 5. Using DFEV1 . 8% predicted gave the best model, and only for the two criteria using % predicted thresholds was the prediction contribution by reversibility status greater than that for ICS. The HRs for age, for pack-years, and for taking ICSs were stable across all models. Adding an absolute threshold of 200 mL and then 250 mL to the . 8% predicted criterion did not improve the prediction of mortality, and the survival benefit from BD status worsened from 0.56 to 0.59 and 0.65, respectively. Figure 6 shows the Kaplan-Meier survival curves for the 1,414 subjects meeting the . 8% predicted criterion and the 1,229 subjects meeting the . 12% start value plus . 0.2 L criterion. Those meeting the former criterion had better survival throughout almost all the at risk period. Regression models for the . 8% predicted criterion and the . 12% start value plus . 0.2 L criterion were then run for selected groups of subjects to see how stable the findings were with respect to suggested diagnoses on the request form, sex, age and whether subjects were taking ICSs. Table 6 shows these results. The . 8% predicted criterion returned more positive (1 ve) for BDR (BDR 1ve) with a better survival HR compared with the other criterion in all groups. The . 12% start value plus . 0.2 L criterion returned significantly more men as BDR 1ve in all groups except for where COPD was the suggested diagnosis, where the . 8% predicted criterion found significantly more women BDR 1ve than expected by chance. The greatest survival advantage (HR, 0.25; 95% CL, 0.12-0.49) was for the . 8% predicted criterion in those suspected of having COPD.
[
148#4 CHEST OCTOBER 2015
]
TABLE 3
] Subjects Identified for Defining Reversibility From Change in FEV1 After Bronchodilator, With Male to Female Ratio in Those Deemed Positive, and Median Baseline FEV1 Values
Study/Year
Reversibility Definition
% 1ve
Male to Female Ratio
Median FEV1 1ve
2ve
…
…
1.03
ⱖ 15% baseline
28.4
1.06a
1.35
1.97b
. 12% baseline 1 . 0.2 L
29.1
c
1.36
1.66
1.89b
. 12% baseline
37.5
1.04a
1.42
2.02b
. 0.2 L
38.6
1.60
c
1.92
1.71b
Quanjer et al3/1993
. 9% predicted
27.7
1.09a
1.74
1.83a
Siafakas et al /1995
. 10% predicted
23.0
1.08
a
1.74
1.82a
. 0.4 L
10.2
2.70c
2.20
1.76b
Committee on Emphysema1/1974 Pellegrino et al /2005 7
4
British Thoracic Society and Scottish Intercollegiate Guidelines Network9/2012
1.80
For x2 test on sex ratio and rank sum test for FEV1 comparisons. –ve 5 negative; 1ve 5 positive. aNot significant. bP , .0001. cP , .001.
Discussion We found that expressing BDR by the change in FEV1 as % predicted avoided any sex and size bias and gave the best survival prediction. The current American Thoracic Society/European Respiratory Society recommendation for expressing BDR7 was found to be biased toward giving reversibility status to men and to those with lower FEV1 at baseline, the latter effect having been demonstrated before.14,15 The best survival advantage from having reversibility as a dichotomous variable was seen with DFEV1 . 8% predicted. The first significant survival advantage compared with those who had BDR , 0% predicted (ie, those with TABLE 4
a reduction in FEV1 after BD) was for DFEV1 . 8% predicted, which gave reversibility status to more people (33%). Our finding is close to that of Brand et al,16 who found the best separation (highest likelihood ratio) of asthma from COPD was at . 9% predicted, which has also been used as the criterion by the European Respiratory Society.3 Dales et al14 also showed that . 9% predicted was stable among sex, height, and age subgroups. In subjects younger than ours, a % predicted criterion has been found to best relate to asthma diagnosis.17 Our finding of 8% predicted as the threshold for defining BDR is slightly more inclusive than these two previous studies and, therefore, increases the potential benefit of treatment options to patients. Our finding also held true
] Results of Univariate Cox Regression With Indexes in Order of Their Significance in Predicting Mortality
Predictors
HR
95% CI
X2
Age
1.07
1.06-1.07
226.2
FEV1/FVC SR
0.81
0.76-0.87
36.6
Pack-y
1.01
1.00-1.01
20.8
FEV1 SR
0.87
0.81-0.93
15.7
ICS
0.69
0.57-0.83
15.2
Male sex
1.37
1.13-1.65
10.6
b-Blocker
1.34
0.93-1.93
2.3
FVC SR
0.96
0.90-1.04
1.0
Prednisolone
1.14
0.76-1.70
0.4
Pack-y is pack-y of smoking exposure. ICS, b-blocker, and prednisolone indicate if the subject was taking an ICS or b-blocker or oral prednisolone, respectively, at the time of testing. HR 5 hazard ratio; ICS 5 inhaled corticosteroid. See Table 2 for expansion of other abbreviation.
journal.publications.chestnet.org
Figure 2 – Percentage of the population deemed reversible () on the right-hand ordinate scale against various cut points in DFEV1 expressed as % predicted, together with the hazard ratio for death () for those deemed reversible plotted on the left-hand ordinate scale using Cox regression stratified for quartiles of age, pack-y of smoking, FEV1/FVC standardized residual, and whether on inhaled corticosteroids.
881
Figure 3 – Percentage of patients dying and percentage who were male sex within bins of level of change in postbronchodilator FEV1 expressed as % predicted. The number of subjects in each bin is shown above the percentage male plot.
when using the European Community for Steel and Coal prediction equations3 instead of GLI-2012 for the subjects (data not shown), but this may not be the case for other lung function equations. Factors that may affect studies on BDR include the class and dose of BD used and the timing of the spirometry. The American Thoracic Society/European Respiratory Society statement regarding BDR testing7 gave an example of using 400 mg salbutamol, but the method of administration was not defined. We ensured our data were all consistent by including only those subjects who received a single BD as 400 mg salbutamol via a spacer device. Most previous studies defining BDR have been based mainly on patients with known airways obstruction,11,16-20 whereas Tan et al21 undertook a population sample of . 10,000 subjects aged . 40 years across four continents and showed that the upper 95% CI for FEV1 change was 10.0% predicted (9.5%-10.5%). This was consistent across the sites and was also independent of sex.
Figure 4 – Hazard ratio, with upper and lower 95% CI, for subjects within bins of level of change in postbronchodilator FEV1 expressed as % predicted. The 399 subjects with change . 14% predicted were the comparator group for survival. The number of subjects in each bin is shown.
882 Original Research
Figure 5 – Hazard ratio, with upper and lower 95% CI, for subjects within bins of level of change in postbronchodilator FEV1 expressed as % predicted. The 429 subjects with change ⱕ 0% predicted were the comparator group for survival. The number of subjects in each bin is shown.
Testing for lack of sex bias in BDR testing assumes that the proportion of men vs women who have significant reversibility (ie, those with a higher probability of asthma) should be similar to the proportion in all the subjects undergoing the test. The incidence of asthma is known to be higher in boys than in girls but over adult life the incidence in women is higher than in men.22,23 These patients were aged 20 years to 97 years and although the incidence of asthma may vary across this age range, we believe the prevalence of asthma in this group of patients should not be especially biased toward one sex or the other. In an adult population sample, a higher prevalence of asthma in women has been found,24 and a survey of . 100,000 subjects up to age 52 years (male to female ratio, 0.93) found the prevalence of asthma had a male to female ratio of 0.84.25 We have shown that including an absolute change in FEV1 as a criterion for diagnosing reversibility biased the findings in favor of men in our population of patients, which is opposite to the trend expected from these population surveys. A possible limitation of our study is that we did not have any final clinical diagnoses to verify who should have significant BDR. Using the proposed diagnosis from the referral to analyze subgroups showed the . 8% criterion had less sex and size bias and a better survival advantage than did the . 12% start plus . 0.2 L criterion. However, the strength of our study is that instead of using a clinical diagnosis, which has inherent difficulties of verification, we used the premise that those who truly have a clinically important degree of reversibility will have a better survival because it is related to the condition of asthma, which has a better prognosis than do those conditions without an important degree of reversibility of airflow limitation, such as COPD and bronchiectasis, and subjects with other lung diseases
[
148#4 CHEST OCTOBER 2015
]
For pack-y, the HR is per 10 pack-y of smoking exposure. The Cox models are ordered with the best model to the left. BL 5 baseline FEV1 measurement; Pred 5 Predicted. See Tables 1, 2, and 4 for expansion of other abbreviations. aX 2 for the model.
2.5
24.3 0.54-0.79
1.01-1.06 2.5
0.65 24.2 0.54-0.80
1.01-1.06 2.6
0.66 0.55-0.80
1.01-1.06 1.03
0.66
1.00-1.05
24.1 0.55-0.81
2.4 1.03 1.01-1.05 1.03 2.6
24.2
1.03 Pack-y
0.55-0.80 0.66 Taking ICS
1.01-1.06
0.66
0.55-0.80 24.2 0.67
2.3
24.2
1.03
1.03
26.1
21.9 0.45-1.00
0.74-0.86
23.6 0.52-0.82
0.67
26.7 0.71-0.83
23.8
0.65
0.72-0.84
0.52-0.81 0.65
0.77
0.51-0.79
27.0 0.70-0.81
0.48-0.75 24.5 0.64 0.60 25.3
26.6
Reversible
0.73-0.84 0.78
0.56
FEV1/FVC SR
0.46-0.70
0.78
0.73-0.85 26.4 0.75
24.1
26.7
0.77
0.79
Z
13.3 1.06-1.08
95% CI HR
1.07
Z
13.7
95% CI
1.06-1.08
HR
1.07
Z
13.6 1.06-1.08
95% CI HR
1.07 13.9
Z 95% CI
1.06-1.08
HR
Z
13.8 1.07 1.06-1.08
95% CI HR
1.07
Z
13.8 1.06-1.08
95% CI HR
1.07
Criterion
Age
DFEV1 . 0.4 L BTS/SIGN 20129 433 Reversible (10%) 294a DFEV1 . 15% BL and . 0.2 L BTS 20035 1,005 Reversible (24%) 303a DFEV1 . 12% BL and . 0.2 L ATS/ERS 20057 1,232 Reversible (29%) 305a DFEV1 . 15% BL ACCP 19741 1,200 Reversible (28%) 306a DFEV1 . 9% Pred ERS 19933 1,172 Reversible (28%) 311a DFEV1 . 8% Pred 1,414 Reversible (33%) 320a
Indicating Their Individual Significance for the Predictors
] Number of Subjects Reversible by Six Criteria for Defining Reversibility With HRs for All-Cause Mortality, Their 95% CIs, and Z Values TABLE 5
journal.publications.chestnet.org
Figure 6 – Kaplan-Meier plots for survival in the 1,414 subjects with reversibility . 8% predicted () and the 1,229 subjects with reversibility . 12% start value plus . 0.2 L (line only).
referred for testing. Results from normal subjects could dilute or obscure this effect, but they were unlikely to receive reversibility testing in our laboratory. The fact that we used patient survival to determine BD reversibility status does not presuppose or suggest that reversibility testing is or should ever be used clinically for survival prediction. If physicians gave inhalers only to those subjects they perceived to be reversible, then the improved mortality from BD status may be directly related to clinical practice. However, our physicians were generous in their use of inhalers, with 60% receiving b-agonists and 43% receiving ICSs compared with the 32% we found with BD reversibility . 8% predicted; therefore, we believe our analysis was not defeated by the prevailing clinical practice. A further possible limitation is that we did not have details of the causes of death. Many of the subjects’ deaths will have nonrespiratory causes, but lung function predicts all-cause mortality in the general population26,27 and the signal of improved survival from being reversible was detectable irrespective of the cause of death because the large sample of our data mitigated any uneven distribution of comorbidities or particular diagnoses. If cause of death had been available, this may have helped sharpen the signal in this form of analysis, but not knowing the cause of death does not negate or diminish our findings. All our post-BD results were based on measurements taken on one occasion, so we are unaware of the reproducibility of the result in this population. Calverley et al28 found that . 38% of the patients with COPD included in a study who had previously been shown to be irreversible at their first visit were then reversible at a subsequent visit. However, their criteria for reversibility were based on change as a % of baseline, so small variations in baseline severely affected the repeatability of the result; this is not true for 883
884 Original Research
[
148#4 CHEST OCTOBER 2015
]
2,392
2,147
Age ⱖ 60 y
Male sex
2,080
1,835
Age , 60 y
Female sex
2,261
Other diagnosis
…
…
52.1
49.2
52.1
54.4
48.2
53.5
47.3
% Male
33.1 25.0
. 8% predicted
33.0
. 12% start and . 0.2 L
33.8
. 8% predicted
26.4
. 12% start and . 0.2 L
. 12% start and . 0.2 L
31.4
32.5
. 8% predicted
36.1
. 8% predicted
26.0
. 12% start and . 0.2 L
30.7
. 8% predicted
34.6
. 12% start and . 0.2 L
39.1
. 8% predicted
32.1
. 12% start and . 0.2 L
36.2
. 8% predicted
26.6
. 12% start and . 0.2 L
31.7
. 8% predicted . 12% start and . 0.2 L
35.7 32.1
. 8% predicted
% 1ve
. 12% start and . 0.2 L
Criterion
1.74
1.79c b
1.83
1.87b a
1.90
1.94b c
2.22
2.34b
1.70 1.64 1.51 1.43 1.80 1.72 2.11 1.94
54.3b c
57.3a a
60.7b a
…
…
…
…
62.0b
53.4a
53.0c
48.9
54.8
47.0
48.0
0.53
0.80
0.59
0.81
0.70
0.34
0.25
0.52
0.46
0.42-0.67
0.52-1.25
0.38-0.93
0.61-1.07
0.53-0.92
0.18-0.67
0.12-0.49
0.35-0.78
0.31-0.68
0.61-1.06
0.54-0.92
1.48
1.47
1.82
2.04
0.44-0.77 0.50-0.87 0.36-0.70 0.40-0.84
0.58 0.66 0.51 0.58
c
1.58c
1.59
2.12
2.27b
0.45-0.74
a
a
0.80
0.70
a
1.98b
1.76
0.58
1.93
1.83
59.9b
53.5
0.34-0.69
0.28-0.57
1.55b
a
a
0.48
0.40
1.40
1.72b
1.55
95% CI
1.55b
1.67
1.61
55.3b
HR 1ve
1.39
a
a
48.7
2ve
1ve
% Male
Median FEV1
Significant sex difference or difference in median FEV1 between BDR 1ve and BDR 2ve. BDR 5 bronchodilator reversibility. See Tables 3 and 4 for expansion of other abbreviations. Not significant. bP , .005. cP , .05.
a
1,628
Asthma
338
2,369
Not taking ICS
COPD
1,858
No.
1ve
] Analysis of Subgroups of Subjects by Diagnosis and by Drug Treatment of Two Criteria for Defining BDR
Taking ICS
TABLE 6
BDR defined from change as % predicted. Because these subjects were those routinely referred for reversibility testing as part of their clinical care, our findings should pertain to similar clinical populations. Because we had no data for children, our observations may not necessarily pertain to a pediatric population; however, there is no obvious flaw to the principle that relating change to a predicted value is also true for children.
Conclusions We conclude that there is a lack of consensus across international societies as to how to determine a clini-
Acknowledgments Author contributions: M. R. M. had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis, including and especially any adverse effects. H. W. and M. R. M. contributed to the formulation of the research question and analysis of the data; and H. W., B. G. C., and M. R. M. contributed to the acquisition of the data and the writing of the paper. Conflict of interest: H. W. has participated in speaking activities funded by pharmaceutical industries including Novartis, Boehringer-Ingelheim GmbH, Almirall, GlaxoSmithKline, AstraZeneca, and NAPP. She has also received industry funding for travel to and accommodation at international respiratory conferences. B. G. C. has received payment for consultancy work from pharmaceutical companies for advice on spirometry. M. R. M. has participated in a spoken activity funded by Boehringer-Ingelheim GmbH. Other contributions: The authors thank the lung function physiologists of the Lung Function and Sleep Department at Queen Elizabeth Hospital, Birmingham, for their dedication in obtaining the best results possible for the patients.
References 1. Report of the Committee on Emphysema American College of Chest Physicians. Criteria for the assessment of reversibility in airways obstruction. Chest. 1974;65(5): 552-553. 2. American Thoracic Society. Lung function testing: selection of reference values and interpretative strategies. Am Rev Respir Dis. 1991;144(5):1202-1218. 3. Quanjer PH, Tammeling GJ, Cotes JE, Pedersen OF, Peslin R, Yernault JC; Official Statement of the European Respiratory Society. Lung volumes and forced ventilatory flows. Report Working Party Standardization of Lung Function Tests, European Community for Steel and Coal. Eur Respir J Suppl. 1993;16:5-40.
journal.publications.chestnet.org
cally important degree of BDR. Our study has shown that change in FEV1 as % predicted avoids any sex and size bias and was better at predicting survival advantage, which would help distinguish asthma, which has a better prognosis than other chronic respiratory diseases not associated with clinically important reversibility. Our data also suggested that using change in FEV1 . 8% predicted may be the optimal cutoff because this was the threshold showing the best consistent survival advantage and gave a better yield of reversible subjects than did the previously endorsed3 and validated14,16 criterion of . 9% predicted.
4. Siafakas NM, Vermeire P, Pride NB, et al; The European Respiratory Society Task Force. Optimal assessment and management of chronic obstructive pulmonary disease (COPD). Eur Respir J. 1995;8(8): 1398-1420. 5. British Thoracic Society; Scottish Intercollegiate Guidelines Network. British guideline on the management of asthma. Thorax. 2003;58(suppl 1):i1-i94. 6. Chronic obstructive pulmonary disease: management of chronic obstructive pulmonary disease in adults in primary and secondary care (partial update). National Institute for Clinical Excellence (NICE) website. http://www.nice.org.uk/guidance/ cg101/chapter/1-recommendations. Accessed April 8, 2015. 7. Pellegrino R, Viegi G, Brusasco V, et al. Interpretative strategies for lung function tests. Eur Respir J. 2005;26(5):948-968. 8. Global strategy for the diagnosis, management and prevention of chronic obstructive pulmonary disease. GOLD website. http://www.goldcopd.com. Published 2010. Accessed November 20, 2012. 9. British Thoracic Society and Scottish Intercollegiate Guidelines Network. British guideline on the management of asthma: a national clinical guideline. British Thoracic Society website. http:// www.brit-thoracic.org.uk/Portals/0/ Guidelines/AsthmaGuidelines. Accessed November 2, 2012. 10. Hanania NA, Celli BR, Donohue JF, Martin UJ. Bronchodilator reversibility in COPD. Chest. 2011;140(4):1055-1063. 11. Tweeddale PM, Alexander F, McHardy GJR. Short term variability in FEV1 and bronchodilator responsiveness in patients with obstructive ventilatory defects. Thorax. 1987;42(7):487-490. 12. Guidelines for the measurement of respiratory function. Recommendations of the British Thoracic Society and the Association of Respiratory Technicians and Physiologists. Respir Med. 1994;88(3): 165-194. 13. Quanjer PH, Stanojevic S, Cole TJ, et al; ERS Global Lung Function Initiative. Multi-ethnic reference values for spirometry for the 3-95-yr age range: the global
lung function 2012 equations. Eur Respir J. 2012;40(6):1324-1343. 14. Dales RE, Spitzer WO, Tousignant P, Schechter M, Suissa S. Clinical interpretation of airway response to a bronchodilator. Epidemiologic considerations. Am Rev Respir Dis. 1988;138(2):317-320. 15. Lehmann S, Bakke PS, Eide GE, Humerfelt S, Gulsvik A. Bronchodilator reversibility testing in an adult general population; the importance of smoking and anthropometrical variables on the response to a beta2-agonist. Pulm Pharmacol Ther. 2006;19(4):272-280. 16. Brand PL, Quanjer PH, Postma DS, et al; The Dutch Chronic Non-Specific Lung Disease (CNSLD) Study Group. Interpretation of bronchodilator response in patients with obstructive airways disease. Thorax. 1992;47(6): 429-436. 17. Bussamra MH, Cukier A, Stelmach R, Rodrigues JC. Evaluation of the magnitude of the bronchodilator response in children and adolescents with asthma. Chest. 2005;127(2):530-535. 18. Eliasson O, Degraff AC Jr. The use of criteria for reversibility and obstruction to define patient groups for bronchodilator trials. Influence of clinical diagnosis, spirometric, and anthropometric variables. Am Rev Respir Dis. 1985;132(4): 858-864. 19. Sourk RL, Nugent KM. Bronchodilator testing: confidence intervals derived from placebo inhalations. Am Rev Respir Dis. 1983;128(1):153-157. 20. Tweeddale PM, Merchant S, Leslie M, Alexander F, McHardy GJR. Short term variability in FEV1: relation to pretest activity, level of FEV1, and smoking habits. Thorax. 1984;39(12):928-932. 21. Tan WC, Vollmer WM, Lamprecht B, et al; BOLD Collaborative Research Group. Worldwide patterns of bronchodilator responsiveness: results from the Burden of Obstructive Lung Disease study. Thorax. 2012;67(8):718-726. 22. Becklake MR, Kauffmann F. Gender differences in airway behaviour over the human life span. Thorax. 1999;54(12): 1119-1138.
885
23. Dodge RR, Burrows B. The prevalence and incidence of asthma and asthma-like symptoms in a general population sample. Am Rev Respir Dis. 1980;122(4):567-575. 24. Chittleborough CR, Taylor AW, Dal Grande E, et al; North West Adelaide Health Study Team. Gender differences in asthma prevalence: variations with socioeconomic disadvantage. Respirology. 2010;15(1):107-114.
886 Original Research
27. Chinn S, Gislason T, Aspelund T, Gudnason V. Optimum expression of adult lung function based on all-cause mortality: results from the Reykjavik study. Respir Med. 2007;101(3):601-609. 28. Calverley PM, Burge PS, Spencer S, Anderson JA, Jones PW. Bronchodilator reversibility testing in chronic obstructive pulmonary disease. Thorax. 2003;58(8): 659-664.
25. Leynaert B, Sunyer J, Garcia-Esteban R, et al. Gender differences in prevalence, diagnosis and incidence of allergic and non-allergic asthma: a population-based cohort. Thorax. 2012;67(7):625-631. 26. Miller MR, Pedersen OF, Lange P, Vestbo J. Improved survival prediction from lung function data in a large population sample. Respir Med. 2009;103(3): 442-448.
[
148#4 CHEST OCTOBER 2015
]