A Factor Analysis of Dyspnea Ratings, Respiratory Muscle Strength, and Lung Function in Patients with Chronic Obstructive Pulmonary Disease 1- 4
DONALD A. MAHLER and ANDREW HARVER Introduction
Various outcomes are used to evaluate the severity of disease and to assess response to therapy in chronic respiratory disorders. Traditionally, lung function, especially spirometry, has been the primary metric for evaluating and assessing patients with chronic obstructive pulmonary disease (COPO). Recent studies have indicated that the symptom of breathlessness is an important outcome measure in the management of patients with COPO (1-4). Although a number of different investigators have demonstrated significant correlations between clinical ratings of dyspnea and lung function in patients with chronic respiratory disease, the magnitude of these relationships is modest (5-10). Possibly, ratings of dyspnea may account for variability in patients with COPO in addition to that provided by standard spirometric measures. The purpose of this study was to demonstrate that clinical ratings of breathlessness and physiologic function comprise separate or distinct components, or dimensions, underlying the pathophysiology of patients with chronic respiratory disease. The demonstration of this is important for at least two reasons. First, it would indicate that dyspnea and physiologic function are independent and nonoverlapping quantities in the patient with COPO. Although it may be intuitively obvious that dyspnea and lung function are quite different, previous studies have not readily established this. Second, such statistical evidence would further support the utility for the routine measurement of the severity of dyspnea in the care of symptomatic patients. In this investigation, we used factor analysis to analyze data collected prospectively in 86 symptomatic patients with COPO. Briefly, the aim of factor analysis is to condense as much of the variability in the data to as few a number of factors as possible (11).
SUMMARY The purpose of this study was to demonstrate that clinical ratings of dyspnea and physiologic function are separate dimensions underlying the pathophysiology of chronic obstructive pulmonary disease (COPO). We used principal-components factoranelysls to confirm these dimensions using data collected prospectively In 86 symptomatic patients with COPD. Three dlffarent Instruments were used to rate dyspnea: a modified Medical Research Council (MRC) scale, the oxygen cost diagram (OCO),and the baseline dyspnea Index (BOI). Measures of physiologic function Included standard spirometric measures (forced vital capacity [FVCI and forced expiratory volume In one second [FEV,)) and maximal Inspiratory (P1max) and expiratory (PEmax) mouth pressures. Age of the 65 male and 21 female subJects was 62.9 ± 1.2 yr (mean ± SEM). All three clinical scales were significantly correlated with physiologic function (range of r velues, 0.32 to 0.45; p < 0.05), except for the relationship between the MRCscale and PEmax (r 0.14; P NS). The factor analysis yielded three factors that accounted for 71.9% of the total ,verlance of the data: clinical ratings of dyspnea (MRCscale, OCO, and BOI) loaded on the first factor; maximal respiratory pressures and gender loaded on the second factor; and lung function and age loadttd on the third factor. Additional posf hoc factor analysis provided similar results when the sampte ~s divided Into two subgroups by randomization, by severity of dyspnea ratings, or by severity Qf airflow obstruction. We conclude that dyspnea ratings, maximal respiratory pressures, end lung function are separate factors or quantities that Independently characterize the condition of patients, with COPO. These results suggest that the clinical eveluatlon of patients with COPOshould Include the routine measurement of dyspnea In addition to spirometry and respiratory muscle strength.
=-
=
AM REV RESPIR DIS 1992: 145:467-470
Methods Patients The subjects were 86 patients (65 men and 21 women) with COPD who were recruited in the outpatient clinic at DartmouthHitchcock Medical Center. Their age was 62.9 ± 1.2 yr (mean ± SEM). Inclusion criteria were a diagnosis of either chronic bronchitis or emphysema (COP D); a ratio of forced expiratory volume in one second (FEV,) to forced vital capacity (FVC) < 70070; and a complaint of dyspnea on exertion. None of the patients had any medical illnessthat limited functional activities more than COPD. Each patient gave signed written consent, and the study protocol was approved by the institutional review board. Apparatus and Procedures Dyspnea ratings, spirometry, and respiratory mouth pressures were measured in this sequence in each patient during a single clinic visit. Each patient was clinically stable at the time of testing. Three different scales wereused to rate dyspnea: a modification of the scale devised by the MedicalResearchCouncil (MRC)of Great Britain (12), the oxygen cost diagram (OCD) (13), and the baseline dyspnea index (BDI)
(6). The MRC, OCD, and BDI were administered in random order to patients by one of two pulmonary technicians before pulmonary function testing. The modified MRC scale consists of five grades based on various physicalactivities that provokedyspnea. Each patient was instructed to read the descriptive statements and then select the number that best fit his or her breathlessness. The OCD
(Receivedin originalform February 25, 1991 and in revised form August 16, 1991) 1 From the Department of Medicine, Dartmouth Medical School, Hanover, New Hampshire, and the Department of Psychology, State Universityof New York at Stony Brook, New York. 2 Supported by grants from the American Lung Association of New Hampshire and by No. MH45731 from the National Institute of Mental Health. 3 Presented in part at the Annual Meeting of the American Thoracic Society in Cincinnati, Ohio, May 15, 1989. 4 Correspondence and requests for reprints should be addressed to Donald A. Mahler, M.D., Pulmonary Section, Department of Medicine, HB7500,Dartmouth Medical School, Hanover, NH 03756.
467
468
MAHLER AND HARVER
is a visual analog scale consisting of a line 100mm long with descriptive phrases at various points along the line that correspond to oxygen requirements of different activities (13). The top of the vertical line represents "no breathlessness"; the bottom of the line reflects "the greatest breathlessness."The patient wasinstructedto "mark the lineat a point above which you would become breathless." Usually, further explanation was necessary for the patient to understand the relationship between the vertical line and the listed activities. The distancefrom the bottom of the scale to the patient's mark was measured in millimeters and provided a quantification of the subject's dyspnea. It took approximately 1 min for the patient to complete the OCD. The BDI is a multidimensional instrument that consists of five specific grades for each of the three categories: functional impairment, magnitude of task, and magnitude of effort (6). To grade dyspnea with the BDI, an observerinterviewed the patient and asked open-endedquestionsconcerningthe patient's symptoms. The observerthen focusedon specific criteria for the severityof breathlessness in each categoryas outlined in the BDl. Based on the patient's responses, the observer was able to grade the degreeofimpairment (range o to 4) related to dyspnea for all three components. A baseline focal score was obtained by adding the three ratings for functional impairment, magnitude of task, and magnitude of effort. The range for the focal score was o to 12. The interview required 2 to 3 min. Spirometry (pvC and FEV,) was measured in the seated position using the Gould Ml00B testing system.The highest values wereselected from the best of at least three efforts for FVC and FEV, (14). Predicted normal values were taken from Morris and colleagues (15). Maximal inspiratory (PImax) and expiratory (PEmax) mouth pressure was measured usingthe system describedby Black and Hyatt (16). Both Plmax and PEmax were measured at least five times at functional residual capacity. Predicted normal values were selected from Black and Hyatt (16). Statistical Analysis Correlations among dyspnea ratings and
physiologic parameters wereevaluated using Pearson's correlation coefficient (r). Principal-components factor analysis was used to determine the dimensions underlying the pattern of interrelationships. Factor analysis is a set of statistical tools used to reduce or rearrange a large set of variables to smaller sets of clusters or factors of related variables(11). Theseclusters or factors are taken to represent virtual "source variables" because they account for, or emerge from, fundamental patterns of (statistical) relationships in the data set. Variablesare grouped together in this process based on the calculated estimates of shared variance among variables, with the restriction that the factors reflect independent sources of variation. Accordingly,it is important to recognize that factor analysis does not merely regroup variables that are highly correlated with one another. In the final solution, correlations are obtained among all variablesentered in the analysisand the virtual factors. Each of the original variablesemployedin the analysis are said to "load" on the factors, to greater or lesser extent, based on the magnitude of the obtained correlations betweeneach variable and factor. In general, each variable loads (i.e., correlated most highly with) a single factor, and groups of variables exhibit similar patterns of correlations across the various factors. Factorsare labeled,conveniently, in terms of the pattern of these loadings. Thus, the purposes of using factor analytic techniques in the present study were (1) to determine whether measures of dyspnea, respiratory muscle strength, and lung function obtained in symptomatic COPD patients would reduce to similar or different factors; and (2) to define or label these factors based on the composition or pattern of factor loadings. Once the factors are extracted (at the end of the process), factors are frequently "rotated" in multidimensional space to obtain the simplest and most interpretable factors and to preservethe independence among factors. Several rotation options are available. We chose to rotate the initial solution to simplify the factors themselves(Varimax rotation option). Factor loadings are reported for the rotated factor matrix.
CLINICAL DYSPNEA RATINGS AND PHYSIOLOGIC FUNCTION IN 86 PATIENTS WITH COPO Mean ± SEM
Correlations Correlation coefficients computed among clinical ratings of dyspnea, spirometry, and maximal respiratory pressures are shown in table 2. Dyspnea values were significantly related to physiologic parameters except for the relationship between the modified MRC scale and PEmax (r = -0.14; p = 0.19). Similar correlations were obtained for men and women in separate analyses.
Factor Analyses In the primary analysis, we conducted principal-components factor analysis in the 86 patients incorporating the following variables: age, gender, FVC, FEV" PI max, PEmax, MRC, OCD, and BDI. The solution yielded three factors that accounted for 71.90/0 of the total variance in the data. The Varimax rotation yielded three interpretable factors, and the factor loadings, that is, the correlations with original variables, obtained for each rotated factor are displayed in table 3. As shown, the three clinical ratings of dyspnea loaded on, that is, correlated most highly with, the first factor; maximal respiratory pressures and gender loaded on the second factor; and lung function and age loaded on the third factor. The results of a factor analysis conducted without the gender variable yielded nearly identical solutions; accordingly, the gender variable was excluded in subsequent post hoc analyses.
0-4 1-90 1-12
+ + + + + + + +
1.12-6.03 41-125 0.42-3.59 14-109 30-150 29-167 45-235 21-128
0.11 2.1 0.08 2.3 3.1 3.1 4.9 2.3
Subjects were divided into two groups, TABLi~ 2
Range
1.5 + 0.1 51.9 + 2.1 5.8 + 0.3 3.17 79.5 1.59 56.7 86.0 87.8 133.9 71.5
Descriptive Statistics Values for clinical dyspnea ratings and physiologic function for the 86 patients with COPD are listed in table 1. The range of scores for each variable is also presented and shows the variability in the severity of breathlessness, as well as physiologic impairment, in the sample.
Post Hoc Analyses
TABLE 1
Clinical dyspnea ratings MRC scale OCD BDI Physiological function FVC, L FVC, % of predicted FEV" L FEV" % of predicted Plmax, cm H2O Plmax, % of predicted PEmax, cm H2O PEmax, % of predicted
Results
CORRELATIONS AMONG' CLINICAL DYSPNEA RATINGS, SPIROMETRY, AND MAXIMAL RESPIRATORY PRESSURE IN 86 PATIENTS WITH COPD'
FVC FEV, Plmax PEmax
BDI
OCD
MRC
0.41 0.35 0.41 0.36
0.34 0.32 0.45 0.37
-0.39 -0.37 -0.32 -O.14t
• Values are Pearson's correlation coefficients. Unless otherwise indicated, values are significant at p < 0.05. t Notsignificant.
DYSPNEA RATINGS, RESPIRATORY MUSCLE STRENGTH, AND WNG FUNCTION IN COPD
TABLE 3 VARIMAX ROTATED FACTOR MATRIX FROM THE PRIMARY PRINCIPAL COMPONENTS ANALYSIS OF DYSPNEA, RESPIRATORY MUSCLE STRENGTH, AND LUNG FUNCTION CONDUCTED IN 86 PATIENTS WITH COPD' Factor 1
Factor 2
Factor 3
0.86 ] 0.85 -0.83
0.16 0.25 -0.02
0.15 0.05 -0.17
PEmax Gender Pl max
0.23 0.07 0.38
0.86 -0.77 0.70
FEV, FVC Age
0.29 0.26 0.03
0.22 0.44 0.09
BDI OCD MRC
]
-0.05 -0.20 0.14 0.79 ] 0.74 -0.72
• Values arecorrelation coefficients between factorsandthe originalsets of variables. Brackets are added to highlight the similar patterns of correlations among the variables and the factors.
randomly, and the principal-component analysis (previously described) was repeated for each subgroup of patients. A preliminary series of unpaired t tests conducted between groups for each variable indicated that the groups were comparably matched for age, lung function, maximal respiratory pressures, and dyspnea. The rotated solutions yielded only the same three factors already described, with nearly identical factor loadings, accounting for 750/0 of the variance in one subgroup and 79% in the other. We conducted two additional analyses. In one analysis, patients were divided into two groups on the basis of clinical ratings of dyspnea obtained on the BDI: those with mild to moderate levels of breathlessness (BDI ~ 7, n = 40) comprised one group, and those with more severe levels of breathlessness (BDI < 7, n = 46) comprised the other group. A series of preliminary t tests conducted between groups for each variable entered into these analyses indicated that the more breathless group exhibited significantly lower levels of lung function and respiratory muscle strength compared with the moderately breathless group. Only the age of patients was comparable. Nonetheless, the rotated solutions yielded only the same three factors already described, with nearly identical factor loadings for each group. These three factors accounted for 67% of the variance in the moderately breathless group and 76% in the more severely breathless group. Finally, the total sample of patients was divided on the basis of the severity of airflow obstruction: patients with mild
to moderate levelsof airflow obstruction (FEV1 percent 0 f predicted ~ 50%, n = 52) comprised one group, and those with severeairflow obstruction (FEV 1 percent of predicted < 50%, n = 34) comprised the other. Results of t tests conducted between groups indicated that only age and PEmax were similar between groups. The factor analysis conducted in the less obstructed group yielded only two factors accounting for 65 % of the variance. Dyspnea ratings and maximal respiratory pressures comprised one factor, and lung function and age comprised the other. The analysis conducted in the more obstructed group yielded a slightly different outcome than any other analysis. Three factors that accounted for 77% of the variance were extracted. Both lung function and maximal respiratory pressures loaded on the first factor; clinical ratings of dyspnea loaded on the second factor; and age alone loaded on the third factor. Discussion
Examination of our data by standard correlation analyses revealed that dyspnea ratings were significantly correlated with lung function and respiratory pressures (table 2). The levelsof relationship that weobserved are similar to those reported in other studies in patients with asthma, COPD, and interstitial lung disease (5-10). Wepreviously suggested that these intermediate levels of correlation indicated that dyspnea and physiologic function measure complementary but not overlapping characteristics of COPD (6, 8, 10). In the present investigation we used the method of principal-components factor analysis to examine the independence of dyspnea and physiologic function in the assessment of patients with COPD. In factor analysis, a set of variables is reduced or rearranged into smaller sets of separate dimensions or factors that account for significant portions of the variance in the interrelations among the variables. The solutions are characterized first by the number of independent factors extracted and, second, by the pattern of correlations between the original variables and the factors. In our application of the technique, we sought to determine whether multiple estimates of dyspnea, lung function, and respiratory muscle strength would reduce to similar or different factors in symptomatic patients with COPD. The results of our analyses indicate that dyspnea ratings, maximal respiratory pressures, and lung
469
function are independent factors that characterize or describe the status or condition of the COPD patient. Dyspnea was the first factor extracted in the analysis. This was not unexpected because difficulty in breathing is the most common symptom reported by patients with chronic respiratory disease, including COPD, and can severely limit functional activities. More importantly, our findings using the statistical method of factor analysis indicate that the clinical measurement of dyspnea represents a separate dimension in the assessment of the patient with COPD not provided by physiologic data. The severityofdyspnea can be graded using various clinical instruments (17). Multidimensional tools, including the BDI, appear sensitive to reflect subjective improvement as a result of specific therapy (1, 3, 4, 18). We included measures of respiratory muscle strength in our analyses because of the importance of the respiratory musclesin the pathophysiology of COPD (19, 20). Our results demonstrate that maximal respiratory muscle pressures provide another unique dimension relating to the status of patients with COPD that was independent of both dyspnea ratings and spirometry. Although this information is not surprising, these results support the clinical utility of routinely measuring PImax and PEmax in patients with COPD. Such measurements may be helpful in understanding the severityof symptoms, especially dyspnea, and in establishing a baseline status before any therapeutic intervention. The third factor extracted from the data set was comprised of lung function and age. Certainly, FVC and FEV 1 are primary metrics for establishing the severity of disease in the COPD population. The inclusion of age with spirometry is appropriate given that age is an independent predictor variable of lung function. In our analysis, three factors - dyspnea, maximal respiratory pressures, and lung function - account for a significant proportion (about 72%) of the variance in the interrelationships among a relatively large set of variables designed to assess the status of 86 patients with COPD. Exploratory analyses of patient subgroups (by randomization, by severity of dyspnea, and by severity of airflow obstruction) generally confirmed the results of the primary analysis. Together, the results support the hypothesis that dyspnea ratings, respiratory muscle strength, and lung function are nonoverlapping quantities in the patient with
470
MAHLER AND HARVER
COPD. Accordingly, the clinical evaluation of patients with COPD should include routine measurement of dyspnea and respiratory muscle strength in addition to spirometry. References 1. Mahler DA, Matthay RA, Snyder PE, Wells CK, Loke J. Sustained-releasetheophylline reducesdyspnea in nonreversible obstructive airway disease. Am Rev Respir Dis 1985; 131:22-5. 2. Dullinger 0, Kronenberg R, Niewoehner DE. Efficacy in inhaled metaproterenol and orallyadministered theophylline in patients with chronic airflow obstruction. Chest 1986; 89:171-3. 3. Guyatt GH, Townsend M, Pugsley SO, et al. Bronchodilators in chronic air-flow limitation. Am Rev Respir Dis 1987; 135:1069-74. 4. Harver A, Mahler DA, Daubenspeck JA. Targeted inspiratory muscle training improves respiratory muscle function and reduces dyspnea in patients with chronic obstructive pulmonary disease. Ann Intern Med 1989; 11l:1l7-24. 5. Thrner-Warwick M, Burrows B, Johnson A. Cryptogenic fibrosing a1veolitis: clinical features
and their influence on survivaL Thorax 1980; 35:171-80. 6. Mahler DA, WeinbergDH, WellsCK, Feinstein AR. The measurement of dyspnea: contents, interobserver agreement, and physiologic correlates of two new clinical indexes. Chest 1984; 85:751-8. 7. Stoller JK, Ferranti R, Feinstein AR. Further specifications of a new clinical index for dyspnea. Am Rev Respir Dis 1986; 134:1l29-34. 8. Mahler DA, Wells CK. Evaluation of clinical methods for rating dyspnea. Chest 1988;93:580-6. 9. Gift AG. Validation of a vertical visual analogue scale as a measure of clinical dyspnea. Rehab Nurs 1989; 14:323-5. 10. Mahler DA, Harver A, Rosiello RA, Daubenspeck lA. Measurement of respiratory sensation in interstitial lung disease: evaluation of clinical dyspnea ratings and magnitude scaling. Chest 1989; 96:767-71. 11. Linderman RH, Merenda PF, Gold RZ. Introduction to bivariate and multivariate analysis. Glenview, IL: Scott, Foresman, 1980. 12. Surveillance for respiratory hazards in the occupational setting. Official statement of the American Thoracic Society. Am Rev Respir Dis 1982; 126:952-6. 13. McGavin CR, Artvinli M, Naoe H, McHardy
GJR. Dyspnea, disability, and distance walked: comparison of estimates of exercise performance in respiratory disease. Br Med J 1978; 2:241-3. 14. Gardner RM. Snowbird workshop on standardization of spirometry. Am RevRespirDis 1979; 119:831-8. IS. Morris JF, Koski A, Johnson Le. Spirometric standards for healthy nonsmoking adults. Am Rev Respir Dis 1971; 103:57-67. 16. BlackFL, Hyatt RE. Maximal respiratory pressures; normal values and relationships to age and sex. Am Rev Respir Dis 1969; 99:696-702. 17. Mahler DA, Harver A. Clinical measurement of dyspnea. In: Mahler DA, ed. Dyspnea. Mount Kisco, NY: Futura Publishing, 1990; 75-126. 18. Guyatt GH, Berman LB, Townsend M, Pugsley SO, Chambers LW.A measure ofquality of life for clinical trials in chronic lung disease. Thorax 1987; 42:773-8. 19. O'Connell JM, Campbell AH. Respiratory mechanics in airways obstruction associated with inspiratory dyspnoea. Thorax 1976; 31:669-77. 20. Sharp JT. The chest walland respiratory muscles in airflow limitation. In: Roussos C, Macklem PT, eds. The thorax, Part B. New York: Dekker, 1985; 1155-202.
DONALD A. MAHLER and ANDREW HARVER Introduction
Various outcomes are used to evaluate the severity of disease and to assess response to therapy in chronic respiratory disorders. Traditionally, lung function, especially spirometry, has been the primary metric for evaluating and assessing patients with chronic obstructive pulmonary disease (COPO). Recent studies have indicated that the symptom of breathlessness is an important outcome measure in the management of patients with COPO (1-4). Although a number of different investigators have demonstrated significant correlations between clinical ratings of dyspnea and lung function in patients with chronic respiratory disease, the magnitude of these relationships is modest (5-10). Possibly, ratings of dyspnea may account for variability in patients with COPO in addition to that provided by standard spirometric measures. The purpose of this study was to demonstrate that clinical ratings of breathlessness and physiologic function comprise separate or distinct components, or dimensions, underlying the pathophysiology of patients with chronic respiratory disease. The demonstration of this is important for at least two reasons. First, it would indicate that dyspnea and physiologic function are independent and nonoverlapping quantities in the patient with COPO. Although it may be intuitively obvious that dyspnea and lung function are quite different, previous studies have not readily established this. Second, such statistical evidence would further support the utility for the routine measurement of the severity of dyspnea in the care of symptomatic patients. In this investigation, we used factor analysis to analyze data collected prospectively in 86 symptomatic patients with COPO. Briefly, the aim of factor analysis is to condense as much of the variability in the data to as few a number of factors as possible (11).
SUMMARY The purpose of this study was to demonstrate that clinical ratings of dyspnea and physiologic function are separate dimensions underlying the pathophysiology of chronic obstructive pulmonary disease (COPO). We used principal-components factoranelysls to confirm these dimensions using data collected prospectively In 86 symptomatic patients with COPD. Three dlffarent Instruments were used to rate dyspnea: a modified Medical Research Council (MRC) scale, the oxygen cost diagram (OCO),and the baseline dyspnea Index (BOI). Measures of physiologic function Included standard spirometric measures (forced vital capacity [FVCI and forced expiratory volume In one second [FEV,)) and maximal Inspiratory (P1max) and expiratory (PEmax) mouth pressures. Age of the 65 male and 21 female subJects was 62.9 ± 1.2 yr (mean ± SEM). All three clinical scales were significantly correlated with physiologic function (range of r velues, 0.32 to 0.45; p < 0.05), except for the relationship between the MRCscale and PEmax (r 0.14; P NS). The factor analysis yielded three factors that accounted for 71.9% of the total ,verlance of the data: clinical ratings of dyspnea (MRCscale, OCO, and BOI) loaded on the first factor; maximal respiratory pressures and gender loaded on the second factor; and lung function and age loadttd on the third factor. Additional posf hoc factor analysis provided similar results when the sampte ~s divided Into two subgroups by randomization, by severity of dyspnea ratings, or by severity Qf airflow obstruction. We conclude that dyspnea ratings, maximal respiratory pressures, end lung function are separate factors or quantities that Independently characterize the condition of patients, with COPO. These results suggest that the clinical eveluatlon of patients with COPOshould Include the routine measurement of dyspnea In addition to spirometry and respiratory muscle strength.
=-
=
AM REV RESPIR DIS 1992: 145:467-470
Methods Patients The subjects were 86 patients (65 men and 21 women) with COPD who were recruited in the outpatient clinic at DartmouthHitchcock Medical Center. Their age was 62.9 ± 1.2 yr (mean ± SEM). Inclusion criteria were a diagnosis of either chronic bronchitis or emphysema (COP D); a ratio of forced expiratory volume in one second (FEV,) to forced vital capacity (FVC) < 70070; and a complaint of dyspnea on exertion. None of the patients had any medical illnessthat limited functional activities more than COPD. Each patient gave signed written consent, and the study protocol was approved by the institutional review board. Apparatus and Procedures Dyspnea ratings, spirometry, and respiratory mouth pressures were measured in this sequence in each patient during a single clinic visit. Each patient was clinically stable at the time of testing. Three different scales wereused to rate dyspnea: a modification of the scale devised by the MedicalResearchCouncil (MRC)of Great Britain (12), the oxygen cost diagram (OCD) (13), and the baseline dyspnea index (BDI)
(6). The MRC, OCD, and BDI were administered in random order to patients by one of two pulmonary technicians before pulmonary function testing. The modified MRC scale consists of five grades based on various physicalactivities that provokedyspnea. Each patient was instructed to read the descriptive statements and then select the number that best fit his or her breathlessness. The OCD
(Receivedin originalform February 25, 1991 and in revised form August 16, 1991) 1 From the Department of Medicine, Dartmouth Medical School, Hanover, New Hampshire, and the Department of Psychology, State Universityof New York at Stony Brook, New York. 2 Supported by grants from the American Lung Association of New Hampshire and by No. MH45731 from the National Institute of Mental Health. 3 Presented in part at the Annual Meeting of the American Thoracic Society in Cincinnati, Ohio, May 15, 1989. 4 Correspondence and requests for reprints should be addressed to Donald A. Mahler, M.D., Pulmonary Section, Department of Medicine, HB7500,Dartmouth Medical School, Hanover, NH 03756.
467
468
MAHLER AND HARVER
is a visual analog scale consisting of a line 100mm long with descriptive phrases at various points along the line that correspond to oxygen requirements of different activities (13). The top of the vertical line represents "no breathlessness"; the bottom of the line reflects "the greatest breathlessness."The patient wasinstructedto "mark the lineat a point above which you would become breathless." Usually, further explanation was necessary for the patient to understand the relationship between the vertical line and the listed activities. The distancefrom the bottom of the scale to the patient's mark was measured in millimeters and provided a quantification of the subject's dyspnea. It took approximately 1 min for the patient to complete the OCD. The BDI is a multidimensional instrument that consists of five specific grades for each of the three categories: functional impairment, magnitude of task, and magnitude of effort (6). To grade dyspnea with the BDI, an observerinterviewed the patient and asked open-endedquestionsconcerningthe patient's symptoms. The observerthen focusedon specific criteria for the severityof breathlessness in each categoryas outlined in the BDl. Based on the patient's responses, the observer was able to grade the degreeofimpairment (range o to 4) related to dyspnea for all three components. A baseline focal score was obtained by adding the three ratings for functional impairment, magnitude of task, and magnitude of effort. The range for the focal score was o to 12. The interview required 2 to 3 min. Spirometry (pvC and FEV,) was measured in the seated position using the Gould Ml00B testing system.The highest values wereselected from the best of at least three efforts for FVC and FEV, (14). Predicted normal values were taken from Morris and colleagues (15). Maximal inspiratory (PImax) and expiratory (PEmax) mouth pressure was measured usingthe system describedby Black and Hyatt (16). Both Plmax and PEmax were measured at least five times at functional residual capacity. Predicted normal values were selected from Black and Hyatt (16). Statistical Analysis Correlations among dyspnea ratings and
physiologic parameters wereevaluated using Pearson's correlation coefficient (r). Principal-components factor analysis was used to determine the dimensions underlying the pattern of interrelationships. Factor analysis is a set of statistical tools used to reduce or rearrange a large set of variables to smaller sets of clusters or factors of related variables(11). Theseclusters or factors are taken to represent virtual "source variables" because they account for, or emerge from, fundamental patterns of (statistical) relationships in the data set. Variablesare grouped together in this process based on the calculated estimates of shared variance among variables, with the restriction that the factors reflect independent sources of variation. Accordingly,it is important to recognize that factor analysis does not merely regroup variables that are highly correlated with one another. In the final solution, correlations are obtained among all variablesentered in the analysisand the virtual factors. Each of the original variablesemployedin the analysis are said to "load" on the factors, to greater or lesser extent, based on the magnitude of the obtained correlations betweeneach variable and factor. In general, each variable loads (i.e., correlated most highly with) a single factor, and groups of variables exhibit similar patterns of correlations across the various factors. Factorsare labeled,conveniently, in terms of the pattern of these loadings. Thus, the purposes of using factor analytic techniques in the present study were (1) to determine whether measures of dyspnea, respiratory muscle strength, and lung function obtained in symptomatic COPD patients would reduce to similar or different factors; and (2) to define or label these factors based on the composition or pattern of factor loadings. Once the factors are extracted (at the end of the process), factors are frequently "rotated" in multidimensional space to obtain the simplest and most interpretable factors and to preservethe independence among factors. Several rotation options are available. We chose to rotate the initial solution to simplify the factors themselves(Varimax rotation option). Factor loadings are reported for the rotated factor matrix.
CLINICAL DYSPNEA RATINGS AND PHYSIOLOGIC FUNCTION IN 86 PATIENTS WITH COPO Mean ± SEM
Correlations Correlation coefficients computed among clinical ratings of dyspnea, spirometry, and maximal respiratory pressures are shown in table 2. Dyspnea values were significantly related to physiologic parameters except for the relationship between the modified MRC scale and PEmax (r = -0.14; p = 0.19). Similar correlations were obtained for men and women in separate analyses.
Factor Analyses In the primary analysis, we conducted principal-components factor analysis in the 86 patients incorporating the following variables: age, gender, FVC, FEV" PI max, PEmax, MRC, OCD, and BDI. The solution yielded three factors that accounted for 71.90/0 of the total variance in the data. The Varimax rotation yielded three interpretable factors, and the factor loadings, that is, the correlations with original variables, obtained for each rotated factor are displayed in table 3. As shown, the three clinical ratings of dyspnea loaded on, that is, correlated most highly with, the first factor; maximal respiratory pressures and gender loaded on the second factor; and lung function and age loaded on the third factor. The results of a factor analysis conducted without the gender variable yielded nearly identical solutions; accordingly, the gender variable was excluded in subsequent post hoc analyses.
0-4 1-90 1-12
+ + + + + + + +
1.12-6.03 41-125 0.42-3.59 14-109 30-150 29-167 45-235 21-128
0.11 2.1 0.08 2.3 3.1 3.1 4.9 2.3
Subjects were divided into two groups, TABLi~ 2
Range
1.5 + 0.1 51.9 + 2.1 5.8 + 0.3 3.17 79.5 1.59 56.7 86.0 87.8 133.9 71.5
Descriptive Statistics Values for clinical dyspnea ratings and physiologic function for the 86 patients with COPD are listed in table 1. The range of scores for each variable is also presented and shows the variability in the severity of breathlessness, as well as physiologic impairment, in the sample.
Post Hoc Analyses
TABLE 1
Clinical dyspnea ratings MRC scale OCD BDI Physiological function FVC, L FVC, % of predicted FEV" L FEV" % of predicted Plmax, cm H2O Plmax, % of predicted PEmax, cm H2O PEmax, % of predicted
Results
CORRELATIONS AMONG' CLINICAL DYSPNEA RATINGS, SPIROMETRY, AND MAXIMAL RESPIRATORY PRESSURE IN 86 PATIENTS WITH COPD'
FVC FEV, Plmax PEmax
BDI
OCD
MRC
0.41 0.35 0.41 0.36
0.34 0.32 0.45 0.37
-0.39 -0.37 -0.32 -O.14t
• Values are Pearson's correlation coefficients. Unless otherwise indicated, values are significant at p < 0.05. t Notsignificant.
DYSPNEA RATINGS, RESPIRATORY MUSCLE STRENGTH, AND WNG FUNCTION IN COPD
TABLE 3 VARIMAX ROTATED FACTOR MATRIX FROM THE PRIMARY PRINCIPAL COMPONENTS ANALYSIS OF DYSPNEA, RESPIRATORY MUSCLE STRENGTH, AND LUNG FUNCTION CONDUCTED IN 86 PATIENTS WITH COPD' Factor 1
Factor 2
Factor 3
0.86 ] 0.85 -0.83
0.16 0.25 -0.02
0.15 0.05 -0.17
PEmax Gender Pl max
0.23 0.07 0.38
0.86 -0.77 0.70
FEV, FVC Age
0.29 0.26 0.03
0.22 0.44 0.09
BDI OCD MRC
]
-0.05 -0.20 0.14 0.79 ] 0.74 -0.72
• Values arecorrelation coefficients between factorsandthe originalsets of variables. Brackets are added to highlight the similar patterns of correlations among the variables and the factors.
randomly, and the principal-component analysis (previously described) was repeated for each subgroup of patients. A preliminary series of unpaired t tests conducted between groups for each variable indicated that the groups were comparably matched for age, lung function, maximal respiratory pressures, and dyspnea. The rotated solutions yielded only the same three factors already described, with nearly identical factor loadings, accounting for 750/0 of the variance in one subgroup and 79% in the other. We conducted two additional analyses. In one analysis, patients were divided into two groups on the basis of clinical ratings of dyspnea obtained on the BDI: those with mild to moderate levels of breathlessness (BDI ~ 7, n = 40) comprised one group, and those with more severe levels of breathlessness (BDI < 7, n = 46) comprised the other group. A series of preliminary t tests conducted between groups for each variable entered into these analyses indicated that the more breathless group exhibited significantly lower levels of lung function and respiratory muscle strength compared with the moderately breathless group. Only the age of patients was comparable. Nonetheless, the rotated solutions yielded only the same three factors already described, with nearly identical factor loadings for each group. These three factors accounted for 67% of the variance in the moderately breathless group and 76% in the more severely breathless group. Finally, the total sample of patients was divided on the basis of the severity of airflow obstruction: patients with mild
to moderate levelsof airflow obstruction (FEV1 percent 0 f predicted ~ 50%, n = 52) comprised one group, and those with severeairflow obstruction (FEV 1 percent of predicted < 50%, n = 34) comprised the other. Results of t tests conducted between groups indicated that only age and PEmax were similar between groups. The factor analysis conducted in the less obstructed group yielded only two factors accounting for 65 % of the variance. Dyspnea ratings and maximal respiratory pressures comprised one factor, and lung function and age comprised the other. The analysis conducted in the more obstructed group yielded a slightly different outcome than any other analysis. Three factors that accounted for 77% of the variance were extracted. Both lung function and maximal respiratory pressures loaded on the first factor; clinical ratings of dyspnea loaded on the second factor; and age alone loaded on the third factor. Discussion
Examination of our data by standard correlation analyses revealed that dyspnea ratings were significantly correlated with lung function and respiratory pressures (table 2). The levelsof relationship that weobserved are similar to those reported in other studies in patients with asthma, COPD, and interstitial lung disease (5-10). Wepreviously suggested that these intermediate levels of correlation indicated that dyspnea and physiologic function measure complementary but not overlapping characteristics of COPD (6, 8, 10). In the present investigation we used the method of principal-components factor analysis to examine the independence of dyspnea and physiologic function in the assessment of patients with COPD. In factor analysis, a set of variables is reduced or rearranged into smaller sets of separate dimensions or factors that account for significant portions of the variance in the interrelations among the variables. The solutions are characterized first by the number of independent factors extracted and, second, by the pattern of correlations between the original variables and the factors. In our application of the technique, we sought to determine whether multiple estimates of dyspnea, lung function, and respiratory muscle strength would reduce to similar or different factors in symptomatic patients with COPD. The results of our analyses indicate that dyspnea ratings, maximal respiratory pressures, and lung
469
function are independent factors that characterize or describe the status or condition of the COPD patient. Dyspnea was the first factor extracted in the analysis. This was not unexpected because difficulty in breathing is the most common symptom reported by patients with chronic respiratory disease, including COPD, and can severely limit functional activities. More importantly, our findings using the statistical method of factor analysis indicate that the clinical measurement of dyspnea represents a separate dimension in the assessment of the patient with COPD not provided by physiologic data. The severityofdyspnea can be graded using various clinical instruments (17). Multidimensional tools, including the BDI, appear sensitive to reflect subjective improvement as a result of specific therapy (1, 3, 4, 18). We included measures of respiratory muscle strength in our analyses because of the importance of the respiratory musclesin the pathophysiology of COPD (19, 20). Our results demonstrate that maximal respiratory muscle pressures provide another unique dimension relating to the status of patients with COPD that was independent of both dyspnea ratings and spirometry. Although this information is not surprising, these results support the clinical utility of routinely measuring PImax and PEmax in patients with COPD. Such measurements may be helpful in understanding the severityof symptoms, especially dyspnea, and in establishing a baseline status before any therapeutic intervention. The third factor extracted from the data set was comprised of lung function and age. Certainly, FVC and FEV 1 are primary metrics for establishing the severity of disease in the COPD population. The inclusion of age with spirometry is appropriate given that age is an independent predictor variable of lung function. In our analysis, three factors - dyspnea, maximal respiratory pressures, and lung function - account for a significant proportion (about 72%) of the variance in the interrelationships among a relatively large set of variables designed to assess the status of 86 patients with COPD. Exploratory analyses of patient subgroups (by randomization, by severity of dyspnea, and by severity of airflow obstruction) generally confirmed the results of the primary analysis. Together, the results support the hypothesis that dyspnea ratings, respiratory muscle strength, and lung function are nonoverlapping quantities in the patient with
470
MAHLER AND HARVER
COPD. Accordingly, the clinical evaluation of patients with COPD should include routine measurement of dyspnea and respiratory muscle strength in addition to spirometry. References 1. Mahler DA, Matthay RA, Snyder PE, Wells CK, Loke J. Sustained-releasetheophylline reducesdyspnea in nonreversible obstructive airway disease. Am Rev Respir Dis 1985; 131:22-5. 2. Dullinger 0, Kronenberg R, Niewoehner DE. Efficacy in inhaled metaproterenol and orallyadministered theophylline in patients with chronic airflow obstruction. Chest 1986; 89:171-3. 3. Guyatt GH, Townsend M, Pugsley SO, et al. Bronchodilators in chronic air-flow limitation. Am Rev Respir Dis 1987; 135:1069-74. 4. Harver A, Mahler DA, Daubenspeck JA. Targeted inspiratory muscle training improves respiratory muscle function and reduces dyspnea in patients with chronic obstructive pulmonary disease. Ann Intern Med 1989; 11l:1l7-24. 5. Thrner-Warwick M, Burrows B, Johnson A. Cryptogenic fibrosing a1veolitis: clinical features
and their influence on survivaL Thorax 1980; 35:171-80. 6. Mahler DA, WeinbergDH, WellsCK, Feinstein AR. The measurement of dyspnea: contents, interobserver agreement, and physiologic correlates of two new clinical indexes. Chest 1984; 85:751-8. 7. Stoller JK, Ferranti R, Feinstein AR. Further specifications of a new clinical index for dyspnea. Am Rev Respir Dis 1986; 134:1l29-34. 8. Mahler DA, Wells CK. Evaluation of clinical methods for rating dyspnea. Chest 1988;93:580-6. 9. Gift AG. Validation of a vertical visual analogue scale as a measure of clinical dyspnea. Rehab Nurs 1989; 14:323-5. 10. Mahler DA, Harver A, Rosiello RA, Daubenspeck lA. Measurement of respiratory sensation in interstitial lung disease: evaluation of clinical dyspnea ratings and magnitude scaling. Chest 1989; 96:767-71. 11. Linderman RH, Merenda PF, Gold RZ. Introduction to bivariate and multivariate analysis. Glenview, IL: Scott, Foresman, 1980. 12. Surveillance for respiratory hazards in the occupational setting. Official statement of the American Thoracic Society. Am Rev Respir Dis 1982; 126:952-6. 13. McGavin CR, Artvinli M, Naoe H, McHardy
GJR. Dyspnea, disability, and distance walked: comparison of estimates of exercise performance in respiratory disease. Br Med J 1978; 2:241-3. 14. Gardner RM. Snowbird workshop on standardization of spirometry. Am RevRespirDis 1979; 119:831-8. IS. Morris JF, Koski A, Johnson Le. Spirometric standards for healthy nonsmoking adults. Am Rev Respir Dis 1971; 103:57-67. 16. BlackFL, Hyatt RE. Maximal respiratory pressures; normal values and relationships to age and sex. Am Rev Respir Dis 1969; 99:696-702. 17. Mahler DA, Harver A. Clinical measurement of dyspnea. In: Mahler DA, ed. Dyspnea. Mount Kisco, NY: Futura Publishing, 1990; 75-126. 18. Guyatt GH, Berman LB, Townsend M, Pugsley SO, Chambers LW.A measure ofquality of life for clinical trials in chronic lung disease. Thorax 1987; 42:773-8. 19. O'Connell JM, Campbell AH. Respiratory mechanics in airways obstruction associated with inspiratory dyspnoea. Thorax 1976; 31:669-77. 20. Sharp JT. The chest walland respiratory muscles in airflow limitation. In: Roussos C, Macklem PT, eds. The thorax, Part B. New York: Dekker, 1985; 1155-202.
