INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The determinants of the outcome of chronic obstructive pulmonary disease (COPD) and asthma have been the focus of several studies. Airways obstruction has been established as an important predictor of survival in population-based studies (1, 2) as well as in patient-based studies on COPD (3) and asthma (6).

The importance of the reversible part of airways obstruction is less clear. In a study of 985 patients with COPD the response to a bronchodilator was found to be a positive prognostic factor along with FEV₁ at baseline (7). However, if baseline FEV₁ was substituted with postbronchodilator FEV₁, the bronchodilator reversibility became nonsignificant. In a summary of three Dutch studies a higher reversibility was found to be an important predictor of survival and a slower decline in FEV₁ (8). In one of these studies the survival prediction was improved by expressing reversibility as a percentage of predicted minus baseline FEV₁ (9). In contrast to traditional measures of reversibility, this parameter was significantly and positively correlated with survival even after controlling for postbronchodilator FEV₁. Others have found that airway responsiveness, measured as the response to a bronchodilator, was negatively correlated with prognosis in terms of annual decline in FEV₁ independently of baseline FEV₁ (10).

In asthma the importance of reversibility is sparsely documented. One large study found that a high degree of reversibility was negatively associated with survival, i.e., subjects with more than 50% reversibility had a 7.0 times higher risk of death from asthma compared with subjects with less than 25% reversibility (11). The postbronchodilator FEV₁ was not used in the analyses. The investigators speculated that a high degree of reversibility was a marker of poor asthma control which in turn was associated with increased mortality.

The interpretation of the studies has been complicated by the use of different markers or surrogate markers of prognosis and the use of different measures of reversibility. Clearly, reversibility will be biased toward higher values in obstructive patients by expressing the value in percentage of baseline FEV₁ and toward higher values in normal subjects by expressing the value relative to the difference between predicted and baseline FEV₁. Thus, reversibility may become a marker not only of airways reactivity but of airways obstruction as well.

All studies of the importance of reversibility in prognosis have focused on short-term reversibility with bronchodilators but none has examined the importance of reversibility with corticosteroids. Given the fact that corticosteroid and bronchodilator reversibility are very poorly correlated (12) and probably reflect different aspects of airways obstruction and reactivity, the impact of these responses on prognosis is not necessarily the same.

We examined the importance of airways obstruction and reversibility to corticosteroids and bronchodilators expressed in different terms as predictors of survival in a population of patients with predominantly severe asthma or COPD attending a chest clinic.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Study Population

During a 6-yr period (1983-1988) all patients referred to the chest clinic at Copenhagen Municipal Hospital, Bispebjerg, who performed combined bronchodilator and corticosteroid reversibility testing were registered. At the end of 1988, the data base contained demographic data, smoking history, clinical diagnosis, and pulmonary function tests of 2,095 patients. The bronchodilator and corticosteroid reversibility of a subset of these patients have previously been reported (12). The vital status of the 2,095 patients as of September 25, 1997 was obtained from the Danish Civil Registration System. In case of death, time of death was obtained. The study was approved by the local ethical committee and the Danish Data Protection Agency.

At the time of entry into the database, subjects were classified by a consultant as having asthma or not, based on a clinical judgment which included evaluation of variability in airflow obstruction, first appearance of symptoms, history of allergy/atopy, smoking history, and the results of corticosteroid reversibility testing. This initial diagnosis was retained in our analyses. Subjects with a baseline FEV₁/FVC of less than 89% of predicted value, who did not have asthma, were classified as having COPD in our analyses.

Patients were excluded from subsequent analyses if (numbers excluded are given in parentheses): age was below 18 at testing time (36), or the subject did not have asthma or COPD (283), or if any of the following information was missing: age, height, and gender (10); FVC at baseline (4); vital status and possible time of death (32); information about current smoking at testing time (140). Finally, four subjects were excluded because of one or more extreme FEV₁ values. The remaining 1,586 subjects comprised the study population.

Pulmonary Function Testing

FEV₁ and FVC were measured using a dry wedge spirometer (Vitalograph Ltd., Buckingham, UK), and the highest value from at least two technically satisfactory maneuvers differing by less than 5% was recorded. Measurements were corrected to BTPS. Baseline values of FVC and FEV₁ were recorded at Day 1. Bronchodilator reversibility was measured half an hour after inhalation of salbutamol 0.3 mg and ipratropium bromide 0.06 mg. Patients received treatment with oral prednisone 30 mg daily for 7 d and performed spirometry on Day 8. Finally, bronchodilator reversibility with salbutamol and ipratropium bromide was repeated on Day 8.

Calculations and Data Management

Using published standard reference equations for ventilatory flows (13), predicted FEV₁ and FEV₁/FVC were calculated for each individual. The actual FEV₁ values were expressed in relation to predicted values (%p): (1) baseline FEV₁ %p (Day 1); (2) postbronchodilator FEV₁ %p (Day 1); (3) postcorticosteroid FEV₁ %p (Day 8); (4) postcorticosteroid and postbronchodilator FEV₁ %p (Day 8). The highest of the four values was used as a measure of individual best FEV₁ %p.

The measured reversibilities were calculated in four ways: as absolute values (ml), absolute values relative to baseline FEV₁ (%), absolute values relative to predicted FEV₁ (%), and absolute values relative to predicted minus baseline FEV₁ (%).

Owing to sparse information about smoking habitsonly reflecting current status and amountsmoking information was transformed into a categorical variable with three categories: (1) never-smokers and ex-smokers; (2) current smokers smoking less than 15 cigarettes/day; (3) current smokers smoking at least 15 cigarettes/day. The cut point of 15 cigarettes/day was arbitrarily chosen, but approximated the median value of cigarettes per day among the current smoking subjects.

Statistical Analyses

Statistical analyses were performed using the SPSS statistical program for Windows, version 7.5. Demographic distributions are presented as mean ± SD. Cox proportional hazards analysis was used to test for relationship between time until death (all causes) and age, gender, smoking habits, pulmonary function, and reversibility. The above explanatory covariates were forced into the Cox regression models, whereas other covariates such as interaction terms were tested and considered significant if p values were less than 0.05. Age, FEV₁, and reversibility were entered as continuous variables whereas gender and smoking habits were coded as indicator variables. Separate analyses were performed on asthma and COPD. The assumption of proportional hazards was tested by plotting the logarithm to the cumulated hazard against time for different categories or strata of the explanatory variables. The predictive power of the regression models was evaluated by the decrease in 2 log likelihood (2 LL).

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Of the 1,586 subjects included, 491 subjects were classified as having asthma and 1,095 as having COPD. Demographic data are shown in Table 1.

Mean follow-up time was 11.2 yr (range, 8.5 to 14.8 yr) for subjects surviving the study period. The degree of airways obstruction in general was moderate to severe. The patients with COPD had a mean baseline FEV₁ of 1.00 L (38.5 %p) which increased to a mean of 1.27 L (49.2 %p) after corticosteroid and bronchodilator. The asthmatics had a mean baseline FEV₁ of 1.34 L (45.3 %p) which increased after corticosteroid and bronchodilator to a mean of 2.13 L (73.0 %p).

The excluded 509 subjects were a very heterogeneous group, consisting of subjects with other diseases, children/adolescents, and subjects with missing information on essential variables. We compared a subgroup of those excluded, consisting of 164 subjects with missing values on vital status or smoking history, with our study population of 1,586 subjects, and found no significant difference on age, baseline FEV₁, baseline FVC, or gender distribution.

The results of the Cox analyses are shown in Table 2 (asthma) and Table 3 (COPD). Two models were tested with different FEV₁ measures as covariates. Model 1 included baseline FEV₁ %p and Model 2 included best FEV₁ %p. Furthermore, both models included age, gender, smoking category, and bronchodilator/corticosteroid reversibilities as covariates. Interaction terms between gender and the other covariates were tested and found nonsignificant, which allowed us to analyze the two sexes together. The bronchodilator responses before as well as after corticosteroid were tried in all analyses. However, the initial bronchodilator response was consistently nonsignificant after adjusting for the postcorticosteroid bronchodilator response, for which reason the latter was used in subsequent analyses.

In asthma, Model 1 decreased 2 LL from an initial value of 1,486 to 1,352 resulting in a model ² of 134 with 7 degrees of freedom (df) whereas Model 2 decreased 2 LL to 1,353 resulting in a model ² of 133 with 7 df. This differerence was not significant. However, the reversibilities could be eliminated as covariates from Model 2 without loss of prognostic information, resulting in a model ² of 133 with 5 df, for which reason Model 2 was considered the best model owing to its simplicity and comparable predictive power. Similarily in COPD, Model 1 had a ² of 314 with 7 df, compared with a Model 2 ² of 313 with 7 df, which was not significantly different. Eliminating the reversibilities from Model 2 resulted in a model ² of 312 with 5 df. Thus, Model 2 could also in COPD be considered the best model because it had fewer covariates and the same predictive power as Model 1.

In Model 2, which we considered the final model, age came out as a strong and significant predictor (p < 0.0005) with a relative risk (RR) ratio of 2.2 per 10 yr of age for asthmatic subjects and a RR of 1.8 for subjects with COPD. Best FEV₁ %p was also a highly significant predictor (p < 0.0005) with a RR of 0.63 per 15% for asthmatic subjects and a RR of 0.62 per 15% for subjects with COPD. Heavy smokers had an elevated risk compared with never- and ex-smokers, irrespective of disease. However, moderate smoking only contributed to a significantly elevated risk among COPD subjects but not among asthmatics. Gender did not significantly contribute to survival prediction, despite a tendency toward higher risk for male subjects than for females.

In Model 1 bronchodilator and corticosteroid reversibility came out as significantly positive predictors (p < 0.0005). In Model 2 where baseline FEV₁ %p was substituted for best FEV₁ %pwhich typically was the FEV₁ measured after corticosteroid and bronchodilatornone of the reversibilities was significant and the RR estimates approximated one.

Reversibility expressed in absolute values, relative to baseline FEV₁ and relative to predicted minus baseline FEV₁ were tested in both models but did not improve the predictive power, and all became nonsignificant when best FEV₁ %p was introduced as predictor.

The importance of best FEV₁ to survival is visualized in Figure 1, where cumulative survival rates are plotted for strata of best FEV₁ %p. The differences in survival between all four strata were highly significant (log-rank test, p < 0.0001).

View larger version (16K): [in this window] [in a new window]   Figure 1.   Survival in strata of best FEV1. Analyses include asthmatics as well as subjects with COPD (n = 1,586).

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

This study lends strong support to the hypothesis that maximal attainable lung function is the best spirometric index in survival prediction irrespective of the need for medication to attain it. This applies to asthma as well as COPD; furthermore, it seems unimportant whether the highest FEV₁ is attained with corticosteroids and/or bronchodilators. When baseline FEV₁ was used in the survival prediction, both corticosteroid and bronchodilator reversibility were prognostic factors with high reversibility indicating better prognosis. The RR associated with corticosteroid and bronchodilator reversibility were of equal size irrespective of diagnosis. Despite different pharmacological and pathophysiological mechanisms, a 15% increase in FEV₁ obtained by corticosteroids in an asthmatic subject constituted the same survival benefit as a 15% increase in FEV₁ obtained by bronchodilators in a patient with COPD. However, when the best FEV₁ was used in the model, both reversibilities became nonsignificant. This means that whether best FEV₁ is composed of a low baseline FEV₁ and a high degree of reversibility, or a high baseline FEV₁ and a low degree of reversibility, is of no importance to prognosis.

The number of subjects in the present study and the follow-up period was adequate to establish a significant association between all-cause mortality and best FEV₁ and the unimportance of reversibility per se as a prognostic factor. The end-point (death) occurred in 850 of 1,586 subjects and even though only 127 deaths occurred in the asthma group, the associations are significant in this group as well. Furthermore, few subjects were lost to follow-up, and only 1.4% of otherwise eligible subjects were excluded owing to missing vital data.

A possible limitation of this study is that the study population was derived from a clinical setting, and the findings might not apply to asthmatics and patients with COPD in general. The subjects were referred to a chest clinic where they subsequently underwent bronchodilator and corticosteroid reversibility tests, and they were more likely to suffer from a more severe disease than asthmatics and patients with COPD in a general population. Selection was further biased toward severity, as patients who had normal lung function after bronchodilators did not always perform corticosteroid reversibility testing. We are not able to estimate the magnitude of this selection bias, but it is reflected in the fact that only 49 of 1,586 subjects in our study population had a postbronchodilator FEV₁ equal to or above 80% of predicted. Thus, our conclusions cannot be generalized to apply for asthmatics fully reversible on bronchodilators or to COPD patients with very mild disease, reflected in a reduced FEV₁/FVC ratio but almost normal FEV₁.

We used the "doctor's diagnosis" of asthma, whereas COPD was diagnosed on the basis of a reduced baseline FEV₁/FVC. Undoubtedly some patients have been misclassified and some may have had both diagnoses. However, demographic data (Table 1) show that the two populations differed in almost all parameters most pronounced in corticosteroid reversibility, smoking habits, age, and survival rates. Despite these differences, the analyses provided uniform results in the two diagnostic groups. Considering the very similar estimates for the importance of baseline FEV₁, best FEV₁, and reversibility in the two diseases, it is unlikely that other methods of differentiating between asthma and COPD would have changed the basis for our overall conclusions.

Our data on smoking history were unfortunately very crude, and it can be argued that the pooling of never-smokers and ex-smokers may have caused underestimation of the effect of smoking and may have influenced the other estimates, too. In order to look into this we made separate analyses of the never- or ex-smoking COPD subjects (who presumably are almost exclusively ex-smokers) and the never- or ex-smoking asthmatics (probably a real mixture of never- and ex-smokers). Those subgroup analyses gave identical estimates of the effect of baseline FEV₁, best FEV₁, and reversibilities compared with the overall models, which assured us that our conclusions are valid, despite the crude smoking data.

The diversity of the results from various studies of prognosis in relation to reversibility in COPD has led to speculations that regular treatment with bronchodilators was a prerequisite for a positive prognostic impact of reversibility. This has been proposed by Postma and Sluiter (8) and also was part of the hypothesis behind the Lung Health Study (14), which, however, showed no beneficial effect of bronchodilator treatment on the decline in FEV₁ over time (15). The nature of our study did not allow for evaluating the effect of treatment, but the convincing results of the Lung Health Study indicate that prognosis would probably not be different if regular bronchodilator treatment had been used systematically.

Apparently, the results of our study of asthmatic subjects are quite different from the results of Ulrik and Frederiksen who found reversibility to be a highly negative prognostic sign, yet without taking postbronchodilator FEV₁ into account (11). Furthermore, this study examined the risk of asthma-related death whereas we looked at all-cause mortality. The asthmatics in the study of Ulrik and Frederiksen were considerably younger than the ones in our study (mean age: 38.2 yr versus 53.4 yr) and they had better lung function (mean baseline FEV₁: 86.7 %p versus 45.3 %p). Probably, asthma-related death in a young population is related to poor asthma control, which in turn is associated with a high degree of reversibility and a low baseline FEV₁, whereas mortality in an older and more irreversibly obstructed group of asthmatics is related to a broader range of diseases, where FEV₁ is a common predictor in accordance with the results of large population studies (16). If this is the case, asthma-related deaths might be negatively associated with reversibility, whereas deaths from all causes might be associated with best FEV₁ as is shown in our study.

In conclusion, we find that prognosis in terms of all-cause mortality is strongly correlated with age, smoking, and the best attainable FEV₁. This applies to asthmatics as well as patients with COPD, provided an element of chronic airflow obstruction. Gender is not a significant predictor of mortality after controlling for the above covariates, despite a tendency toward poorer survival among male subjects. By using best FEV₁ as a predictor, the predictive power of reversibility is eliminated regardless of type or measure. The combined use of corticosteroid and bronchodilator reversibility in survival analyses is a novel approach, and we have demonstrated that both contribute to survival prediction to the extent that best FEV₁ of the subject is modified, but neither reversibilities independently influence survival.