Respiration 37: 301-308 (1979)
Relative Sensitivities and Specificities of Tests for Small Airways Obstruction Ronald D. Fairshter and Archie F. Wilson Department of Medicine, University of California, Irvine, Calif.
Key Words. Small airway obstruction • Pulmonary function tests • Cigarette smoking Abstract. This study compared several physiologic maneuvers for detecting small airways obstruction (SAO). Volume of isoflow best separated young smokers (average age - 27.6; average cigarette consumption - 9.3 pack years) from nonsmokers; this was true for com parisons of both population means and incidence of individual abnormality. However, assessment of individual abnormality by any test was greatly influenced by the criteria used to define abnormality. The best combination of test sensitivity and specificity was achieved by using the 95% confidence limits derived from a control population from the same laboratory. Even so, test results did not correlate significantly with pack years of cigarette smoking. Then, although tests for SAO are very sensitive, their prognostic value in young smokers is uncertain since abnormality presumably should correlate with extent of smoking.
Recent studies have demonstrated that the earliest lesions of obstructive lung dis ease appear to be located within airways less than 2 mm in internal diameter (small airways) [Macklem, 1972], Since these air ways contribute less than 20% to total air way resistance, clinical manifestations of small airways obstruction (SAO) are usually mild or nonapparent [Macklem and Mead, 1967]; similarly, standard physiologic pro cedures such as measurement of forced ex piratory volume in 1 sec (FEV,), forced vi tal capacity (FVC), and airways resistance (Raw) are also usually normal. Hence, de tection of obstructive lung disease is ordi
narily delayed until more advanced obstruc tion is present [Macklem, 1972]. Early detection of SAO has recently been made possible by the development of noninvasive pulmonary function tests which are far more sensitive than routine spirome try [Hutcheon el al., 1974; Morris et al., 1975; McCarthy et al., 1972; McFadden et al., 1974; McFadden and Linden, 1972].1 A 1 Abnormality of the new pulmonary function tests used in this study is generally considered evi dence of SAO. However, it is recognized that the physiologic determinants of most of the tests in clude lung elastic recoil. In this study, the term SAO is used to refer to early obstructive lung dis ease in cigarette smokers. No attempt is made to differentiate subjects with loss of lung elastic re coil from those with loss of airways caliber.
Downloaded by: Nagoya University 133.6.82.173 - 1/11/2019 7:25:31 PM
Introduction
variety of methods with differing technical and equipmental requirements are available for evaluation of SAO. The abilities of these techniques to detect SAO are not identical; indeed, the sensitivity of any one test rela tive to the others may vary considerably de pending upon the circumstances of evalua tion [Sobol, 1976]. Recent studies have also indicated that the results of some tests are highly variable [Black et al., 1974; McCar thy et al., 1975; McFadden et al., 1975]. Variability might be expected to be associat ed with increased nonspecificity; yet, there is little available data about the specificity of tests of small airways function. Further more, the sensitivity and specificity of any method must be influenced by the criteria used to define abnormality [Sobol, 1976]; in many studies, abnormality has been defined by arbitrary criteria [Sobol, 1976]. In order to compare several different noninvasive, physiologic methods of assess ing SAO, we performed extensive pulmon ary function tests in a group of asymptomat ic cigarette smokers. We then examined the sensitivity and specificity of the test meth ods by varying the criteria for abnormality. Methods and Materials Subjects The nonsmoking group consisted of 50 subjects (33 male, 17 female) from the medical, laboratory, and nursing staffs of the University of California, Irvine Medical Center (UCIMC). None of the subjects had ever smoked cigarettes, pipes, or ci gars. The smoking group consisted of 30 adults (20 male, 10 female) who were also from the med ical, laboratory, or nursing staffs of UCIMC. Cig arette smoking averaged 9.3 + 5.1 (1 SD) pack years. No history of chronic cough, sputum pro duction, asthma, previous cardiopulmonary dis ease, or recent (within 3 months) symptomatic up per or lower respiratory tract infection was elicit ed from subjects in either group. Both groups
were selected to exclude subjects who had smoked marijuana recently or subjects who had smoked marijuana on more than a few occasions. All sub jects answered a questionnaire detailing respirato ry symptoms (childhood and adult), occupational exposures, geographic and family histories, aller gies, and degree of exposure to cohort cigarette smokers. Forced expiratory volume in 1 sec (FEV,), forced mid-expiratory flow (FEF25_75), forced ex piratory flow at 75-85% of forced vital capacity (FEF jj _ 85), vital capacity (VC), expiratory reserve volume (ERV), and the ratio of FEV, to forced vital capacity (FEV,/FVC%) were measured using a 13.5-liter water-sealed bell spirometer (Warren E. Collins, Inc.). Specific conductance (SGaw) was calculated by dividing the reciprocal of Raw by Vtg [Briscoe and DuBois, 1958], Residual volume (RV) and total lung capacity (TLC) were calculat ed by standard techniques (RV = Vtg - ERV and TLC = RV + VC). All lung volumes and spirographic flow rates were expressed as a percentage of predicted [Boren et at., 1966; Goldman and Becklake, 1959; Kory et al., 1961; Morris el al., 1971, 1975], Single breath diffusing capacity for carbon monoxide (DLCO) was measured using the breathholding method of Ogilvie et al. [1957], Al veolar volume (VA) was determined from dilution of neon. Results were expressed as DLCO/VA. Closing volume (CV) was performed using a mo dified single breath nitrogen dilution technique [Buist and Ross, 1973a]. Results were expressed as the ratio of closing volume to vital capacity (CVAfC). The slope of the alveolar plateau (SBN2) was also calculated from the record [Buist and Ross, 1973b], Expiratory flow rates were con trolled at less than 0.5 liters/sec. Maximal expiratory flow volume (MEFV) curves were performed using a relatively friction less spirometer (Cardiopulmonary Instruments Corporation Model 220 dry-rolling seal). Results were recorded on an oscilloscope (Electronics for Medicine, Inc., DR-8). Volume history was stand ardized by 3 slow vital capacity inhalations prior to performance of all MEFV maneuvers. MEFV maneuvers were repeated until 3 curves with vis ually similar slopes and VC within 2.5% of each other were obtained. Maximal expiratory flow rates at 20% (Vmax20) and 50% (Vmax50) of VC
Downloaded by: Nagoya University 133.6.82.173 - 1/11/2019 7:25:31 PM
Fairshter/Wilson
302
Relative Sensitivities and Specificities of Tests for Small Airways Obstruction
2 Less than 5% probability of being normal (one-tailed test of significance). Since abnormality on some tests is unidirectional, a two-tailed test of significance (2 X SEE) is not required. Hence, for tests in which abnormal values are always higher than predicted, such as CV/VC%, a result greater than predicted + (1.64 X SEE) is abnormal. For tests in which abnormal results are always less than predicted, such as FEFJ5_85, a value less than predicted - (1.64 X SEE) is abnormal.
ed as a percentage of the mean results in the con trol poulation (range of normal was defined as + 25% of the mean). Separation of the smoking population from the nonsmokers was evaluated on the basis of inci dence of individual abnormality using two by two contingency table analysis (chi square). The Yates modification was used where appropriate. Test sensitivity and specificity were defined as follows [Vecchio, 1966]: Sensitivity (a) = diseased persons with a posi tive test/all diseased subjects tested. Specificity (b) = nondiseased persons negative to the test/all nondiseased subjects tested. For this analysis all cigarette smokers were considered to be ‘diseased subjects’. This assump tion was based upon the results of a recent patho logic study [Niewoehner et al., 1974] which showed that each of 20 young cigarette smokers evaluated had evidence of respiratory bronchioli tis. Although this same study showed that 25% of nonsmokers also had respiratory bronchiolitis, nonsmokers were considered to be ‘nondiseased subjects’ since: (1) respiratory bronchiolitis was considerably milder in nonsmokers [Niewoehner et al., 1974], and (2) nonatopic, nonsmokers rarely develop severe chronic obstruction to airflow.
Results Results for smokers and nonsmokers are shown in table I. Most measurements of ex piratory flow rates were significantly higher in nonsmokers than in smokers. These re sults were true for spirographic (FEV,/ FVC, FEF23.„ , FEF75_8s) as well as flowvolume techniques (Vmax20, Vmax50). Distribution of ventilation, CV/VC%>, dV Emax50 and Viso V also differed signifi cantly between the groups. Lung volumes, FEV,, SGaw, and DLCO/VA did not differ significantly between the groups. Statisti cally, the best separation of mean results of smokers from nonsmokers was achieved by determining dV Emax50, VisoV, and flow rates at low lung volumes (table I).
Downloaded by: Nagoya University 133.6.82.173 - 1/11/2019 7:25:31 PM
were calculated. Volume of isoflow (VisoV) was determined by superimposing MEFV curves ob tained after 3 VC inhalations of a helium-oxygen gas mixture (He-02) upon maneuvers obtained af ter breathing room air. All curves were matched precisely at RV and VisoV was expressed as a percentage of VC. The increment in airflow be tween VKmaXj,, (air) and VEmax50 (H c-02) was also measured. This determination (/1Vf;max50) is thought to be specific for changes in airway cal iber and independent of changes in lung elastic re coil [Dosman et al., 1975]. Mean results and standard deviations for both groups were determined by standard techniques; differences between means were evaluated by Stu dent’s ‘t’ test. A p value of 0.05 or less was consid ered statistically significant. Since various and often arbitrary definitions of individual abnormality have been utilized by dif ferent investigators [Sobol, 1976], we evaluated individual abnormality on tests of small airways function by several different criteria: (1) The results in our subjects were compared to predicted regression equations determined in other laboratories [Buist and Ross, 1973a, b; Gelb el al., 1975; Goldman and Becklake, 1959; Morris et al., 1975]. The range of normal was defined by using both percentage of predicted criteria (nor mal = + 25%> of predicted) and by using the 95°/o confidence limits of the predicted regression (2 X the standard error of the estimate (SEE); 1.64 X SEE2). (2) Individual results were also compared to the normal population tested in our laboratory. The range of normal was defined by the 95% con fidence limits of the control population using both one-tailed (1.64 X the standard deviation (SD)) and two-tailed (2 X SD) tests of significance. For completeness, individual results were also evaluat
303
Fairshter/Wilson
304
Tabic 1. Age, height, and pulmonary function results in smokers and nonsmokers (mean ± 1 SD) Param eter1
Nonsmokers
27.1 ±3.1 69.3 ± 3.2 3.95 ± 0.69 79.7 ± 7.0 4.56 ± 0.93 1.62 ± 0.57 0.28 ±0.12 5.67 ± 0.84 4.96 ± 1.01 1.55 ± 0.47 6.51 ± 1.15 110.1 ±45.6 330.0 ± 104.7 49.4 ± 12.9 9.6 ± 4.2 0.87 ± 0.37 8.6 ±4.8
Age, years Height, inches FEVi, liters FEV i/FV C, % FEF-25-75, 1/sec F E F 7 5 -85 ,1/sec SGaw, 1/sec/cm/l D lCO/V a, ml/min/mm Hg/1 VC, liters RV, liters TLC , liters VEmax 2o (air), 1/min VEmaxso (air), 1/min dVr;niax5o, % VisoV, % VC SBNo, % N 2 /I CV/VC, %
Smokers
p value
27.6 ± 3.7 68.5 ± 2.7 3.68 ± 0.60 75.4 ± 5.6 3.89 ± 1.01 1.17 ± 0.45 0.22 ±0.16 5.48 ± 0.91 4.88 ± 0.80 1.66 ± 0.47 6.54 ± 0.95 74.7 ± 18.5 263.0 ± 62.4 37.2 ± 10.2 16.7 ± 6.6 1.22 ± 0.58 11.7 ± 5.4
NS NS NS
Relative Sensitivities and Specificities of Tests for Small Airways Obstruction Ronald D. Fairshter and Archie F. Wilson Department of Medicine, University of California, Irvine, Calif.
Key Words. Small airway obstruction • Pulmonary function tests • Cigarette smoking Abstract. This study compared several physiologic maneuvers for detecting small airways obstruction (SAO). Volume of isoflow best separated young smokers (average age - 27.6; average cigarette consumption - 9.3 pack years) from nonsmokers; this was true for com parisons of both population means and incidence of individual abnormality. However, assessment of individual abnormality by any test was greatly influenced by the criteria used to define abnormality. The best combination of test sensitivity and specificity was achieved by using the 95% confidence limits derived from a control population from the same laboratory. Even so, test results did not correlate significantly with pack years of cigarette smoking. Then, although tests for SAO are very sensitive, their prognostic value in young smokers is uncertain since abnormality presumably should correlate with extent of smoking.
Recent studies have demonstrated that the earliest lesions of obstructive lung dis ease appear to be located within airways less than 2 mm in internal diameter (small airways) [Macklem, 1972], Since these air ways contribute less than 20% to total air way resistance, clinical manifestations of small airways obstruction (SAO) are usually mild or nonapparent [Macklem and Mead, 1967]; similarly, standard physiologic pro cedures such as measurement of forced ex piratory volume in 1 sec (FEV,), forced vi tal capacity (FVC), and airways resistance (Raw) are also usually normal. Hence, de tection of obstructive lung disease is ordi
narily delayed until more advanced obstruc tion is present [Macklem, 1972]. Early detection of SAO has recently been made possible by the development of noninvasive pulmonary function tests which are far more sensitive than routine spirome try [Hutcheon el al., 1974; Morris et al., 1975; McCarthy et al., 1972; McFadden et al., 1974; McFadden and Linden, 1972].1 A 1 Abnormality of the new pulmonary function tests used in this study is generally considered evi dence of SAO. However, it is recognized that the physiologic determinants of most of the tests in clude lung elastic recoil. In this study, the term SAO is used to refer to early obstructive lung dis ease in cigarette smokers. No attempt is made to differentiate subjects with loss of lung elastic re coil from those with loss of airways caliber.
Downloaded by: Nagoya University 133.6.82.173 - 1/11/2019 7:25:31 PM
Introduction
variety of methods with differing technical and equipmental requirements are available for evaluation of SAO. The abilities of these techniques to detect SAO are not identical; indeed, the sensitivity of any one test rela tive to the others may vary considerably de pending upon the circumstances of evalua tion [Sobol, 1976]. Recent studies have also indicated that the results of some tests are highly variable [Black et al., 1974; McCar thy et al., 1975; McFadden et al., 1975]. Variability might be expected to be associat ed with increased nonspecificity; yet, there is little available data about the specificity of tests of small airways function. Further more, the sensitivity and specificity of any method must be influenced by the criteria used to define abnormality [Sobol, 1976]; in many studies, abnormality has been defined by arbitrary criteria [Sobol, 1976]. In order to compare several different noninvasive, physiologic methods of assess ing SAO, we performed extensive pulmon ary function tests in a group of asymptomat ic cigarette smokers. We then examined the sensitivity and specificity of the test meth ods by varying the criteria for abnormality. Methods and Materials Subjects The nonsmoking group consisted of 50 subjects (33 male, 17 female) from the medical, laboratory, and nursing staffs of the University of California, Irvine Medical Center (UCIMC). None of the subjects had ever smoked cigarettes, pipes, or ci gars. The smoking group consisted of 30 adults (20 male, 10 female) who were also from the med ical, laboratory, or nursing staffs of UCIMC. Cig arette smoking averaged 9.3 + 5.1 (1 SD) pack years. No history of chronic cough, sputum pro duction, asthma, previous cardiopulmonary dis ease, or recent (within 3 months) symptomatic up per or lower respiratory tract infection was elicit ed from subjects in either group. Both groups
were selected to exclude subjects who had smoked marijuana recently or subjects who had smoked marijuana on more than a few occasions. All sub jects answered a questionnaire detailing respirato ry symptoms (childhood and adult), occupational exposures, geographic and family histories, aller gies, and degree of exposure to cohort cigarette smokers. Forced expiratory volume in 1 sec (FEV,), forced mid-expiratory flow (FEF25_75), forced ex piratory flow at 75-85% of forced vital capacity (FEF jj _ 85), vital capacity (VC), expiratory reserve volume (ERV), and the ratio of FEV, to forced vital capacity (FEV,/FVC%) were measured using a 13.5-liter water-sealed bell spirometer (Warren E. Collins, Inc.). Specific conductance (SGaw) was calculated by dividing the reciprocal of Raw by Vtg [Briscoe and DuBois, 1958], Residual volume (RV) and total lung capacity (TLC) were calculat ed by standard techniques (RV = Vtg - ERV and TLC = RV + VC). All lung volumes and spirographic flow rates were expressed as a percentage of predicted [Boren et at., 1966; Goldman and Becklake, 1959; Kory et al., 1961; Morris el al., 1971, 1975], Single breath diffusing capacity for carbon monoxide (DLCO) was measured using the breathholding method of Ogilvie et al. [1957], Al veolar volume (VA) was determined from dilution of neon. Results were expressed as DLCO/VA. Closing volume (CV) was performed using a mo dified single breath nitrogen dilution technique [Buist and Ross, 1973a]. Results were expressed as the ratio of closing volume to vital capacity (CVAfC). The slope of the alveolar plateau (SBN2) was also calculated from the record [Buist and Ross, 1973b], Expiratory flow rates were con trolled at less than 0.5 liters/sec. Maximal expiratory flow volume (MEFV) curves were performed using a relatively friction less spirometer (Cardiopulmonary Instruments Corporation Model 220 dry-rolling seal). Results were recorded on an oscilloscope (Electronics for Medicine, Inc., DR-8). Volume history was stand ardized by 3 slow vital capacity inhalations prior to performance of all MEFV maneuvers. MEFV maneuvers were repeated until 3 curves with vis ually similar slopes and VC within 2.5% of each other were obtained. Maximal expiratory flow rates at 20% (Vmax20) and 50% (Vmax50) of VC
Downloaded by: Nagoya University 133.6.82.173 - 1/11/2019 7:25:31 PM
Fairshter/Wilson
302
Relative Sensitivities and Specificities of Tests for Small Airways Obstruction
2 Less than 5% probability of being normal (one-tailed test of significance). Since abnormality on some tests is unidirectional, a two-tailed test of significance (2 X SEE) is not required. Hence, for tests in which abnormal values are always higher than predicted, such as CV/VC%, a result greater than predicted + (1.64 X SEE) is abnormal. For tests in which abnormal results are always less than predicted, such as FEFJ5_85, a value less than predicted - (1.64 X SEE) is abnormal.
ed as a percentage of the mean results in the con trol poulation (range of normal was defined as + 25% of the mean). Separation of the smoking population from the nonsmokers was evaluated on the basis of inci dence of individual abnormality using two by two contingency table analysis (chi square). The Yates modification was used where appropriate. Test sensitivity and specificity were defined as follows [Vecchio, 1966]: Sensitivity (a) = diseased persons with a posi tive test/all diseased subjects tested. Specificity (b) = nondiseased persons negative to the test/all nondiseased subjects tested. For this analysis all cigarette smokers were considered to be ‘diseased subjects’. This assump tion was based upon the results of a recent patho logic study [Niewoehner et al., 1974] which showed that each of 20 young cigarette smokers evaluated had evidence of respiratory bronchioli tis. Although this same study showed that 25% of nonsmokers also had respiratory bronchiolitis, nonsmokers were considered to be ‘nondiseased subjects’ since: (1) respiratory bronchiolitis was considerably milder in nonsmokers [Niewoehner et al., 1974], and (2) nonatopic, nonsmokers rarely develop severe chronic obstruction to airflow.
Results Results for smokers and nonsmokers are shown in table I. Most measurements of ex piratory flow rates were significantly higher in nonsmokers than in smokers. These re sults were true for spirographic (FEV,/ FVC, FEF23.„ , FEF75_8s) as well as flowvolume techniques (Vmax20, Vmax50). Distribution of ventilation, CV/VC%>, dV Emax50 and Viso V also differed signifi cantly between the groups. Lung volumes, FEV,, SGaw, and DLCO/VA did not differ significantly between the groups. Statisti cally, the best separation of mean results of smokers from nonsmokers was achieved by determining dV Emax50, VisoV, and flow rates at low lung volumes (table I).
Downloaded by: Nagoya University 133.6.82.173 - 1/11/2019 7:25:31 PM
were calculated. Volume of isoflow (VisoV) was determined by superimposing MEFV curves ob tained after 3 VC inhalations of a helium-oxygen gas mixture (He-02) upon maneuvers obtained af ter breathing room air. All curves were matched precisely at RV and VisoV was expressed as a percentage of VC. The increment in airflow be tween VKmaXj,, (air) and VEmax50 (H c-02) was also measured. This determination (/1Vf;max50) is thought to be specific for changes in airway cal iber and independent of changes in lung elastic re coil [Dosman et al., 1975]. Mean results and standard deviations for both groups were determined by standard techniques; differences between means were evaluated by Stu dent’s ‘t’ test. A p value of 0.05 or less was consid ered statistically significant. Since various and often arbitrary definitions of individual abnormality have been utilized by dif ferent investigators [Sobol, 1976], we evaluated individual abnormality on tests of small airways function by several different criteria: (1) The results in our subjects were compared to predicted regression equations determined in other laboratories [Buist and Ross, 1973a, b; Gelb el al., 1975; Goldman and Becklake, 1959; Morris et al., 1975]. The range of normal was defined by using both percentage of predicted criteria (nor mal = + 25%> of predicted) and by using the 95°/o confidence limits of the predicted regression (2 X the standard error of the estimate (SEE); 1.64 X SEE2). (2) Individual results were also compared to the normal population tested in our laboratory. The range of normal was defined by the 95% con fidence limits of the control population using both one-tailed (1.64 X the standard deviation (SD)) and two-tailed (2 X SD) tests of significance. For completeness, individual results were also evaluat
303
Fairshter/Wilson
304
Tabic 1. Age, height, and pulmonary function results in smokers and nonsmokers (mean ± 1 SD) Param eter1
Nonsmokers
27.1 ±3.1 69.3 ± 3.2 3.95 ± 0.69 79.7 ± 7.0 4.56 ± 0.93 1.62 ± 0.57 0.28 ±0.12 5.67 ± 0.84 4.96 ± 1.01 1.55 ± 0.47 6.51 ± 1.15 110.1 ±45.6 330.0 ± 104.7 49.4 ± 12.9 9.6 ± 4.2 0.87 ± 0.37 8.6 ±4.8
Age, years Height, inches FEVi, liters FEV i/FV C, % FEF-25-75, 1/sec F E F 7 5 -85 ,1/sec SGaw, 1/sec/cm/l D lCO/V a, ml/min/mm Hg/1 VC, liters RV, liters TLC , liters VEmax 2o (air), 1/min VEmaxso (air), 1/min dVr;niax5o, % VisoV, % VC SBNo, % N 2 /I CV/VC, %
Smokers
p value
27.6 ± 3.7 68.5 ± 2.7 3.68 ± 0.60 75.4 ± 5.6 3.89 ± 1.01 1.17 ± 0.45 0.22 ±0.16 5.48 ± 0.91 4.88 ± 0.80 1.66 ± 0.47 6.54 ± 0.95 74.7 ± 18.5 263.0 ± 62.4 37.2 ± 10.2 16.7 ± 6.6 1.22 ± 0.58 11.7 ± 5.4
NS NS NS
