Abstract
The present authors aimed to determine if the criteria for the diagnosis of chronic obstructive pulmonary disease (COPD) and its classification by severity as recommended by the Global Initiative for Chronic Obstructive Lung Disease are supported by measurements of respiratory health-related quality of life.
A community-based sample of adults aged 25–75 yrs had pre- and post-bronchodilator spirometry and completed the St George's Respiratory Questionnaire (SGRQ). Loess scatter plot smoothers of the SGRQ versus post-bronchodilator forced expiratory volume in one second (FEV1)/forced vital capacity (FVC) ratio and post-bronchodilator FEV1 % predicted together with receiver operating characteristic (ROC) curve analysis were used to determine the relationship between spirometric variables and clinically important differences in the SGRQ score.
The scatter plot smoother and ROC curve analyses supported the value of 0.7 for post-bronchodilator FEV1/FVC ratio, which was ∼4 units higher than the nadir of the SGRQ. To represent a distance of 8 units on the SGRQ, the cut-off points for post-bronchodilator FEV1 that delimit COPD severity stages were 80, 60 and 40% pred for mild, moderate and severe COPD, respectively.
To diagnose chronic obstructive pulmonary disease the use of post-bronchodilator forced expiratory volume in one second/forced vital capacity ratio of 0.7 is supported by health-related quality of life measurements. There may be advantages in using forced expiratory volume in one second cut-off points of 80, 60 and 40% predicted for the classification of mild, moderate and severe chronic obstructive pulmonary disease, respectively, similar to the approach recommended for asthma.
The Global Initiative for Chronic Obstructive Lung Disease (GOLD) defines chronic obstructive pulmonary disease (COPD) as a post-bronchodilator forced expiratory volume in one second (FEV1)/forced vital capacity (FVC) ratio of <0.7, and not explained by another lung disease 1. However, use of this fixed ratio has been criticised for not reflecting physiological changes of this variable with aging, potentially leading to the overdiagnosis of COPD, particularly in the elderly 2–5.
The utility of spirometry-based definitions of COPD is supported if other measurements of the impact of lung disease show a definite change in relation to spirometry. This constitutes a form of external validity for whatever measurement of lung function is chosen, be it FEV1, FEV1 % predicted or absolute definitions based on the FEV1/FVC ratio. The St George's Respiratory Questionnaire (SGRQ), which is a measurement of respiratory quality of life, has been validated in COPD 6–8. However, it was developed prior to the GOLD guidelines with their rigorous spirometric definition of COPD. Furthermore, these guidelines include severity cut-off criteria based on FEV1 % pred, which are acknowledged to be based on simplicity rather than clinical validation 1. Large longitudinal epidemiological studies would be required to validate the choice of such severity cut-off criteria; however, corroboration of these with meaningful differences in health-related quality of life could potentially serve as a useful surrogate for this. For the purposes of randomised controlled trials, a change in the total score for the SGRQ that represents a clinically important difference is 4 units 6.
The present study aims to examine if the value for the FEV1/FVC ratio of 0.7 can be supported by the measurement of respiratory quality of life in the SGRQ. Secondly, the current authors explore whether the boundaries for severity of COPD based on the GOLD criteria, using cut-off points of FEV1 % pred, are supported by clinically important quality of life changes reflected in the SGRQ.
METHODS
Subjects
Participants (n = 3,500) in the Wellington Respiratory Survey (WRS), were randomly selected from the electoral register, equally distributed by sex across the five-decade age groups from 25 to 75 yrs 9. Subjects were sent a simple postal questionnaire seeking demographic, respiratory and smoking history data. All subjects who completed and returned questionnaires were invited to undertake more detailed questionnaires, pulmonary function tests and computed tomography scanning, as previously described in detail 9. Subjects who completed all investigative modules formed the study group for the present analysis. The reference range population comprised 212 subjects who were never-smokers with no diagnosis of respiratory disease, no recent respiratory symptoms and no use of inhaled medication. The survey was approved by the Wellington Ethics Committee (Wellington, New Zealand) and written informed consent was obtained from each subject.
Pulmonary function testing
Pulmonary function tests in the WRS have been described in detail elsewhere 9, 10. In brief, spirometry measurements were carried out according to American Thoracic Society guidelines 11, a minimum of three acceptable manoeuvres were carried out and the best FEV1 and FVC results were selected for analysis. Spirometry was repeated 45 min after the administration of 400 µg of salbutamol (Ventolin; GlaxoSmithKline NZ Ltd, Auckland, New Zealand) via a spacer (Space Chamber; Medical Developments International Ltd, Springvale, Australia). FEV1 and FVC values were expressed as % pred of normal values 10.
Quality of life questionnaire
The SGRQ is a disease specific quality of life assessment tool validated in both COPD and asthma 12–14. The questionnaire consists of 76 items divided into three parts measuring symptoms, activity limitation and social and emotional impact of disease. Each item is accorded a weight determined by the degree of distress accorded to each symptom or state described. Overall scores range from 0 (no effect on quality of life) to a maximum score of 100 (maximum perceived distress); thus, a higher score means a poorer quality of life and the questionnaire is suitable for administration in healthy people 8. The English for New Zealand version was self completed by the subjects during the 45 min bronchodilation time. Results were entered into a spreadsheet, supplied by the questionnaire producers and pre-programmed with formulae to calculate a total score and scores for each of the individual components of the SGRQ (symptoms, activity and impacts). A change in the total score for the SGRQ that represents a clinically important difference is 4 units 6.
Statistical methods
Quantile regression 15 was used to find a formula to predict the upper 90% and 95% quantiles for the SGRQ based on age. The dataset used for this was the 178 out of 212 subjects from a reference range dataset, who had values for the total score on the SGRQ (data for the SGRQ was missing in 32 subjects from the reference population sourced from outside of the WRS dataset 10 and from two of the WRS cohort).
Loess lines were fitted to the overall WRS dataset for all 713 subjects who had values for the total score on the SGRQ, plotting the SGRQ versus the post-bronchodilator FEV1/FVC ratio and post-bronchodilator FEV1 % pred. Loess plots are a form of robust scatter plot smoothing suitable for noisy data 16, providing information on the form of a relationship between variables where it may not be easily described by conventional regression. The loess technique uses a weighted average of a set of data points to fit curves rather than the individual points as in conventional regression. The technique is relatively resistant to outlying values and does not require a prespecified global linear function to be chosen. The degree of smoothing is determined by the width of the window determining the weighted average. Optimal smoothing is obtained by balancing the residual sum of squares for each point against the degree of smoothness.
It was evident that both curves had a point of inflection, where the value of the SGRQ increased from a minimum level, and the value of the SGRQ 4 and 8 units above this point of inflection were examined. Receiver operating characteristic (ROC) curves for two cut-off points for post-bronchodilator FEV1/FVC ratio of 0.7 and 0.65 and the total score on the SGRQ, were fitted to determine discriminatory values, sensitivity and specificity for the SGRQ.
The relationship of the post-bronchodilator FEV1 % pred, based on equations from the WRS 10, and the loess lines were used to examine the cut-off points for FEV1 % pred in relation to the GOLD severity cut-off points of 80, 50 and 30% pred.
RESULTS
Subjects
There were 2,319 responders to the postal questionnaire, of whom 795 agreed to attend the present authors’ research facility for further testing and 713 were able to complete all WRS investigative modules and form the current study group. Subjects in the study group were broadly similar to the 1,606 postal questionnaire responders not in the study group. Subjects in the study group were more likely to be of male sex (54% versus 44%; p≤0.001), smoke tobacco (13% versus 10%; p = 0.007), have a history of chronic cough (15% versus 11%; p = 0.006) or breathing trouble (28% versus 24%; p = 0.04) and had a greater mean age (53.9 versus 51.6 yrs; p<0.001) than postal questionnaire responders not in the current study group.
Quantile regression prediction equations for SGRQ
The quantile regression equations for the 95% and 90% quantiles for the reference range subjects are 1.53+(age×0.24) and 5.76+(age×0.096), respectively. Using these equations, the 95% or 90% quantile is found by inserting the age in years into the equation, for example the 90% quantile for a 50 yr old is 5.76+(50×0.096), which equals 10.56. The predicted 95% and 90% quantiles ranged from 7.9 to 19.0 and 8.3 to 12.8, respectively, across the age range of 27–74 yrs. Sex was not an important predictor for these two quantiles. Different coefficients were observed for different quantiles as is allowed by the method of quantile regression.
SGRQ versus post-bronchodilator FEV1/FVC ratio
Figure 1⇓ shows a scatter plot of the SGRQ total score versus post-bronchodilator FEV1/FVC ratio with the superimposed loess smoothed line and 90% confidence interval (CI). The point of inflection, the minimum level for the SGRQ, in the loess plot was at a post-bronchodilator FEV1/FVC ratio of ∼0.8. The SGRQ predicted at this value is 8.1 units. A score 4 units higher than this value (SGRQ score 12.1 units), corresponds to a post-bronchodilator FEV1/FVC ratio of 0.71. A score 8 units higher (SGRQ score 16.1 units), corresponds to a value of the post-bronchodilator FEV1/FVC ratio of 0.65.
The St George’s Respiratory Questionnaire (SGRQ) total score versus post-bronchodilator forced expiratory volume in one second (FEV1)/forced vital capacity (FVC) ratio with superimposed loess scatter plot smoother (––––) and 90% confidence interval (– – – –). The loess smoother is a form of regression line using a weighted average of a set of data points at each part of the curve and is robust to outlying values.
ROC curve analysis
In total, 119 (16.7%) out of 713 subjects had a post-bronchodilator FEV1/FVC ratio of <0.7. The area under the ROC curve for the SGRQ for this value of post-bronchodilator FEV1/FVC ratio was 0.692, suggesting moderate discrimination of the SGRQ for this cut-off point. In total, 72 (10.1%) out of 713 subjects had a post-bronchodilator FEV1/FVC ratio of <0.65. The area under the ROC curve for the SGRQ for this value of post-bronchodilator FEV1/FVC ratio was 0.812, suggesting moderate to good discrimination.
Table 1⇓ shows the sensitivity and specificity values and the corresponding SGRQ total scores for the two post-bronchodilator FEV1/FVC ratio cut-off points of 0.7 and 0.65. Sensitivity refers to the number of subjects correctly allocated below the cut-off value by the corresponding value of the SGRQ total score divided by the total number of subjects who, in fact, had a ratio below the cut-off point. Specificity refers to the number of subjects correctly allocated above the cut-off point divided by the total number who, in fact, had a ratio above the cut-off point. For example, a SGRQ total score of 12.0 has 66.7% sensitivity and 70% specificity for a post-bronchodilator FEV1/FVC ratio of 0.65.
Comparison of the sensitivity and specificity of the SGRQ to identify a low post-bronchodilator FEV1/FVC ratio using two different post-bronchodilator FEV1/FVC ratio cut-off points
Figure 2⇓ shows the full ROC curves for the SGRQ and two values of the post-bronchodilator FEV1/FVC ratio.
Receiver operating characteristic plot showing the ability of the St George’s Respiratory Questionnaire to identify a low post-bronchodilator forced expiratory volume in one second (FEV1)/forced vital capacity (FVC) ratio for post-bronchodilator FEV1/FVC ratios of 0.65 (– – – –) and 0.7 (–––––).
SGRQ versus post-bronchodilator FEV1
Figure 3⇓ shows a scatter plot of the SGRQ total score versus post-bronchodilator FEV1 % pred with superimposed loess smoothed line and 90% CI. The point of inflection in the loess plot is a post-bronchodilator FEV1 ∼100% pred. The SGRQ predicted at this value is 7.9 units. A score 4 units higher than this value (SGRQ score 11.9 units) corresponds to a value of post-bronchodilator FEV1 88% pred. A score 8 units higher (SGRQ score 15.9 units) corresponds to a value of post-bronchodilator FEV1 80% pred.
St George’s Respiratory Questionnaire (SGRQ) versus post-bronchodilator forced expiratory volume in one second (FEV1) % predicted with superimposed loess scatter plot smoother (––––) and 90% confidence interval (– – – –). The loess smoother is a form of regression line using a weighted average of a set of data points at each part of the curve and is robust to outlying values.
In figure 3⇑, it can be seen that the relationship between the SGRQ and post-bronchodilator FEV1 % pred is approximately linear from a FEV1 value of 90% with a 10% change in FEV1 associated with a 4 unit change in the SGRQ. The differences in the SGRQ score associated with the GOLD boundaries for severity are shown in table 2⇓. These data show that there are major differences in the magnitude of change in the SGRQ score between the GOLD severity boundaries of 80, 50 and 30% pred, which define mild, moderate and severe COPD, respectively. In contrast, the magnitude of change in the SGRQ score between the severity boundaries of 80, 60 and 40% pred (as recommended in asthma guidelines to define mild, moderate and severe disease 17) were similar, corresponding to approximately twice the clinically important difference for SGRQ (8 units).
Difference in the SGRQ between GOLD defined COPD severity boundary points
DISCUSSION
The present authors have shown that the suggested boundary of normality for a post-bronchodilator FEV1/FVC ratio of 0.7 for COPD is supported by analysis of the current dataset. This FEV1/FVC value corresponded to a value of the SGRQ score close to the 90% quantile in the reference range sample representing a clinically important difference in the SGRQ from the nadir of the relationship. The present findings also suggest that the current GOLD cut-off values of 80, 50 and 30% pred to define mild, moderate and severe COPD, respectively, do not result in a similar magnitude of change in the SGRQ. If the boundaries of severity are to correspond to approximately equal distances on the scale of the SGRQ, boundaries of FEV1 80, 60, and 40% pred would each correspond to approximately twice the clinically important difference for the SGRQ.
The use of a fixed FEV1/FVC ratio of 0.7 to diagnose COPD has been widely criticised for failing to take account of physiological changes in the value of the ratio with ageing, which can result in overdiagnosis of COPD 3. The current authors’ study group has shown a considerable difference in the prevalence of COPD using a fixed ratio compared with age-determined predicted values for the FEV1/FVC ratio 9. In its defence, use of a fixed ratio results in guideline simplicity 1, 18, has been supported by some epidemiological data 19 and values below this have been shown to be predictive of poorer outcomes despite apparent good health 20. Since, in practical terms, use of a fixed ratio is likely to continue, it is useful to show that this ratio corresponds to clinically significant changes in health status on a validated respiratory questionnaire with reasonable specificity.
Normal reference range values for the SGRQ were derived based on its distribution in healthy nonsmokers, identifying that the 90% quantile for the SGRQ across the age range in the present reference range group was 12.8 units. This is not dissimilar to values reported in a previous study, although different methods of reporting make direct comparison difficult 8. The present results also reflect the previously acknowledged effect of age, irrespective of respiratory disease, on health-related quality of life 21. Using smoothed plots of SGRQ versus post-bronchodilator FEV1/FVC ratio the nadir was 8.1 units for the SGRQ and 0.8 for the post-bronchodilator FEV1/FVC ratio. An increase in the SGRQ score of 4 units, an amount that has been proposed to represent a clinically important difference 6, 8, resulted in a value of 12.1 units, which corresponded to a post-bronchodilator FEV1/FVC ratio of 0.7. An increase in the SGRQ score by a further 4 units corresponded to an FEV1/FVC ratio of 0.65. Construction of ROC curves identified that a SGRQ value between 12 and 13 units, corresponding to the 90% quantile in the present reference range sample, had moderate to good discriminative power for a cut-off value for post-bronchodilator FEV1/FVC ratio of 0.7. Not surprisingly, use of a lower fixed ratio of 0.65 improves the specificity associated with the same SGRQ score, at the cost of reduced sensitivity. Thus, a post-bronchodilator FEV1/FVC ratio of 0.7 was represented in the present dataset by a clinically important difference in the SGRQ compared with the nadir of the relationship and a value of the SGRQ close to the 90% quantile in a reference range sample.
The current data was also used to examine the validity of the severity criteria cut-off points for FEV1 espoused by GOLD. Whilst it is known that lower FEV1 values are associated with poorer health status the strength of this correlation is not always that large 21. Such poor correlations indeed strengthen the case for holistic patient assessment using tools such as quality of life questionnaires and the more recently developed BODE (body mass index, airflow obstruction, dyspnoea and exercise capacity) index 22. Nonetheless, a FEV1 cut-off point is simple to apply and has been used in guidelines to recommend preferred therapy and in some settings to control access to specific treatments 23. Whilst the arbitrary nature of cut-off points is acknowledged, it is reassuring that the GOLD severity criteria broadly correlate with the SGRQ, a tool which takes into account the mutlifaceted impact of COPD. However, the present findings suggest that the current cut-off values of FEV1 80, 50 and 30% pred to define mild, moderate and severe COPD, respectively, do not result in a similar magnitude of changes in SGRQ as the relationship is essentially linear at values <80%. If the boundaries of severity are to correspond to approximately equal distances on the scale of the SGRQ, boundaries of FEV1 80, 60 and 40% pred would each correspond to about twice the clinically important difference for the SGRQ. The use of the FEV1 80, 60 and 40% pred cut-off points would also have the advantage of consistency with the grading of severity for asthma 17. In support of this approach, many older patients with asthma will meet the spirometrically-defined criteria for COPD 24.
There are methodological issues to consider in the current study. There were minor differences in the characteristics of the current study group compared with postal questionnaire responders who did not complete all investigative modules, raising the possibility of selection bias. However, it is unlikely that these differences would systematically affect the relationship between the SGRQ and spirometry. There is a high prevalence of asthma in New Zealand and some of the subjects identified as having COPD from post-bronchodilator spirometry may actually have asthma with incomplete bronchodilator reversibility rather than COPD. In a separate analysis of the WRS data, ∼25% of subjects identified as having COPD had features of asthma and no apparent chronic bronchitis or emphysema 24. As the proportion of these individuals that have COPD with an asthma-like phenotype and the proportion that have “pure” asthma cannot be determined, individuals with an asthma-like phenotype of airway disease were not excluded from the analysis. There were few subjects in the current study group with severe COPD and, thus, the present analysis primarily applies to those with mild or moderate COPD. The CIs for the relationship between FEV1 and the SGRQ are consequently wider at a low FEV1 and could support a greater, lesser or even nonlinear relationship between FEV1 and the SGRQ in advanced disease. The SGRQ measurements were “noisy” and although the loess scatter plot smoother is useful for examining the underlying pattern of the dataset it may give a spurious impression of precision. Reducing the degree of smoothing in the algorithm fitting the loess scatter plot smoother may better reflect the noisiness of the data but would reduce the ability to see the overall pattern.
A considerable strength of the present study was that it was a community-based random sample and, therefore, may not be subject to biases created when using diseased populations, such as those attending hospital clinics. The SGRQ is an accepted method of measuring respiratory health quality of life and the definition of a clinically important difference with this instrument means that the analysis reflects associations of clinical importance.
The current authors conclude that, in the diagnosis of chronic obstructive pulmonary disease, the use of a post-bronchodilator forced expiratory volume in one second/forced vital capacity ratio of 0.7 is supported by health-related quality of life measurements. There may be advantages in using post-bronchodilator forced expiratory volume in one second cut-off points of 80, 60 and 40% predicted for the classification of mild, moderate and severe COPD, respectively, similar to the approach recommended for asthma.
Statement of interest
None declared.
Footnotes
-
For editorial comments see page 953.
- Received July 30, 2008.
- Accepted November 28, 2008.
- © ERS Journals Ltd
References
