INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Numerous studies have confirmed the predictive value of lung function measurements in the assessment of prognosis in obstructive lung disease as well as in subjects without apparent pulmonary symptoms. FEV₁ has traditionally been considered the "gold standard," but a few epidemiologic surveys have used peak expiratory flows (PEF) in assessing ventilatory function (1). The preferential use of FEV₁ in assessing ventilatory function is based on several factors. Firstly, PEF is highly effort-dependent and is thus characterized by a greater intrasubject variability than is FEV₁ (4). PEF is also claimed to be less sensitive than FEV₁ in detecting mild airway obstruction, as no information is obtained on the caliber of the small airways (4, 5). Furthermore, the accuracy and linearity of devices used to measure PEF are generally poorer than that of spirometers; although, with an acceptable within-device repeatability (6, 7).

When assessing prognosis in asthma and chronic obstructive pulmonary disease (COPD) it has become clear that it is the irreversible airway obstruction that is important; thus, FEV₁ measurements should be done after maximal bronchodilation, probably including the use of a course of oral corticosteroids to optimize lung function (8, 9). In respect to PEF, very little is known about the possible benefits of repeated measurements after optimizing lung function with bronchodilators and corticosteroids. No studies have compared best FEV₁ and best PEF in a multivariate model of survival in asthma and COPD. We examined if best PEF provides additional prognostic information in subjects with asthma or COPD, even after controlling for best FEV₁, age, smoking, sex, and body mass index (BMI).

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Study Population

The study population comprised 1,586 subjects who performed combined bronchodilator and corticosteroid reversibility testing in a standardized manner during a 6-yr period (1983 to 1988) at the lung clinic in Copenhagen City. The subjects were selected from a data base with 2,095 patients, from which we excluded those younger than 18 yr of age (36 subjects), not having asthma or COPD (283 subjects), or with missing values essential to our analyses (190 subjects). Asthma was defined by a consultant in pulmonary medicine based on a clinical judgment that included evaluation of variability in airflow obstruction, first appearance of symptoms, history of allergy/atopy, smoking history, and the results of corticosteroid reversibility testing. COPD was defined by a reduced FEV₁/FVC (< 89% of predicted) in the absence of asthma. Vital status of the study population as on 25 September 1997 was obtained from the Danish Civil Registration System. Details of the selection procedure and demographic data have previously been reported (9, 10).

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. PEF was measured using a Wright peak flow meter. The best of three readings was recorded. On the first day, FEV₁, FVC, and PEF were recorded before and 30 min after inhalation of salbutamol 0.3 mg and ipratropium bromide 0.06 mg. Patients were subsequently treated with oral prednisone 30 mg daily for 7 d. On the eighth day spirometry was repeated before and after inhalation of salbutamol and ipratropium bromide. No systematic quality control was applied to the spirometers or PEF meters. However, spirometers were controlled approximately every 3 mo as one of the authors (A.K-J.) performed a test maneuver. Leaks were controlled for by routine examination of the FVC plateau for a possible downward slope.

Calculations and Data Management

For each individual subject, predicted values of FEV₁, FVC, FEV₁/ FVC, and PEF were calculated using published standard reference equations (11). The actual values were expressed relative to predicted values (% pred). For each lung function variable the individuals best value was defined as the highest of the four readings (two on Day 1 and two on Day 8). BMI was calculated based on self-reported height and weight at the first visit to the lung clinic. In case of doubt or obvious misjudgment, height and weight were measured with indoor clothing, including flat shoes.

Data on smoking were limited to current smoking status and current amount smoked. Subsequently, we transformed this information into a categorical variable with three categories: (1) never- and ex-smokers, (2) light smokers (< 15 cigarettes/d), and (3) heavy smokers (> 15 cigarettes/d). The cut point of 15 cigarettes/d was chosen as it approximated the median value of cigarettes per day among current smokers.

Statistical Analyses

Statistical analyses were performed using the SPSS statistical program for Windows, version 8.0 (SPSS Inc., Chicago, IL, USA). Demographic distributions were described with mean and SD. Lung function values were tested for significance with a two-sided t test. Cox proportional hazard analysis (12) was used to test for relationship between time until death (all causes) and age, sex, smoking category, BMI, and pulmonary function. These covariates were forced into the regression models, whereas interaction terms were tested and considered significant if p values were less than 0.05. Age and pulmonary function variables were entered as continuous variables, whereas sex, smoking category, and BMI 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 fulfilling the inclusion criteria, 491 were classified as having asthma and 1,095 as having COPD. Demographic data on the two populations are shown in Table 1. Both asthmatics and patients with COPD had a substantial component of irreversible airway obstruction; COPD subjects were moderately to severely obstructed, with a mean FEV₁ after corticosteroid and bronchodilator of 1.27 L (49.2%p), whereas the asthmatics after similar medication reached a mean FEV₁ of 2.13 L (73.0%p). The corticosteroid reversibility was pronounced among the asthmatics (17.8% of predicted FEV₁) and insignificant among the subjects with COPD (2.6% of predicted FEV₁). The total response to treatment, measured as best value versus baseline value, was modestly greater in terms of PEF than by FEV₁; thus asthmatics increased FEV₁ by 27.7%p and PEF by 29.9%p (p < 0.001), whereas patients with COPD increased FEV₁ by 10.8%p and PEF by 11.8%p (p < 0.001).

PEF values were generally greater than FEV₁ values, when compared with what was expected from reference equations, and this difference was more pronounced in asthmatics than in subjects with COPD (10.2 versus 6.1 percentage points using best values, p < 0.001). However, the ratio between best PEF and best FEV₁ (PEF%p/FEV₁%p) was not significantly different between asthmatics and patients with COPD (115 versus 114, NS). Pearson's correlation coefficient (r) between best PEF%p and best FEV₁%p was similar among asthmatics and subjects with COPD (r = 0.79 in both populations).

Lung function parameters and BMI were not significantly correlated in the asthmatic population, whereas in COPD, BMI was correlated to best FEV₁ (r = 0.18, p < 0.001) and best PEF (r = 0.25, p < 0.001). PEF%p/FEV₁%p was also significantly correlated to BMI in the COPD population (r = 0.13, p < 0.001) but not in asthmatics.

The results of the Cox regression analyses are shown in Table 2. The effect of age, sex, and smoking status on overall mortality in the two populations has previously been reported. Interaction terms between sex and all other covariates were tested one by one in both asthma and COPD groups. All interaction terms proved non-significant, allowing us to analyze both sexes together.

Subjects with COPD displayed a U-shaped relationship between BMI and mortality; thus, a low BMI (< 20 kg/m²) as well as a high BMI (> 30 kg/m²) was associated with an increase in mortality (RR of 1.3 and 1.4, respectively). For subjects with asthma, only a low BMI (< 20 kg/m²) was associated with a significantly increased risk (RR of 2.0).

After controlling for age, sex, smoking, and BMI, combinations of PEF and FEV₁ were added to the model. The relative risk ratios associated with a 25% decrease in best FEV₁ or best PEF, added either singly or simultaneously to the Cox model, are shown in Table 2. In asthma, best PEF significantly contributed to survival prediction if best FEV₁ was not controlled for (RR of 1.9, p < 0.0001), but not after adjusting for best FEV₁ (RR of 1.3, NS). In COPD, best PEF was a significant contributor to survival prediction irrespective of adjustment for best FEV₁, however, with a smaller value of RR if FEV₁ was controlled for (RR of 1.9, p < 0.0001 versus RR of 1.5, p < 0.0001).

Another approach to evaluate the power of survival prediction by PEF and FEV₁ is shown in Figure 1, where improvements in the Cox model ² (decrease in 2 log likelihood) are plotted for baseline and best values of FEV₁ and PEF (see legend for detailed explanation). In asthma, the predictive power improved significantly by using best values of PEF and/or FEV₁ instead of baseline values (² values of 27, 22, and 28 versus 8, 6, and 8). However, PEF was inferior to FEV₁ in predictive power and could not improve on the survival model if FEV₁ was controlled for, irrespective of baseline or best values being used. In COPD, best values of PEF and/or FEV₁ were also clearly superior to baseline values in predictive power (² values of 104, 108, and 120 versus 80, 64, and 84). In contrast to what was seen in asthma, the use of best values instead of baseline values improved relatively more on the predictive power of PEF than on the predictive power of FEV₁. Thus, at baseline, FEV₁ was superior to PEF (improvement in model ² of 80 versus 64), but at best, PEF was marginally superior to FEV₁ (improvement in model ² of 108 versus 104). Furthermore, if both covariates were added simultaneously into the model, ² improved to 120, which was highly significant.

View larger version (33K): [in this window] [in a new window]   Figure 1.   Improvements in model 2 (decrease in 2 log likelihood) in the Cox proportional hazard analyses by adding either baseline values of FEV1 and/or PEF or the corresponding best values. The improvements are stated relative to the 2 improvements for a model that includes only age, sex, smoking category, and body mass index. The 2 improvements cannot be compared between asthma and COPD because of different survival rates and number of subjects and are thus depicted on different scales.

To test whether the independent effect of PEF was present irrespective of the level of FEV₁, we performed similar survival analyses on two COPD subgroups divided according to the median value of best FEV₁ for the whole COPD population (49.2%p). The two groups had mean values of best FEV₁ of 38%p and 61%p, respectively, and in both, PEF was an independent predictor of survival after controlling for best FEV₁. We also tested a formal interaction between stratified values of PEF and FEV₁ in the COPD population in toto. This interaction term proved nonsignificant.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

This study has demonstrated that in a population of moderately to severely obstructed patients with COPD, PEF is at least as important for prognosis as FEV₁. Furthermore, PEF is partially independent of FEV₁ as predictor of outcome in terms of overall mortality. The precondition is that PEF is measured subsequent to lung function being optimized with bronchodilators and presumably also with corticosteroids, although the latter is quantitatively less important in patients with COPD. If PEF is measured at a single time point without prior medication, it is not equal to FEV₁ in predictive power and it does not add substantially to the predictive model. In asthma the situation is apparently different, as PEF generally seems inferior to FEV₁ in predictive power and adds no further information to a predictive model if best FEV₁ is controlled for. These findings must be cautiously interpreted, as especially our asthmatic population may not be representative of asthmatics in the general population, as later discussed.

There are no obvious explanations for the observed differences between asthma and COPD with respect to the importance of PEF. We tested the relationship between PEF and FEV₁ in the two populations and found no major differences; however, PEF values were generally higher than FEV₁ values in both COPD and asthma. Among subjects with COPD, PEF was significantly closer associated with BMI than among asthmatics, and the association to BMI tended to be better for PEF than for FEV₁, indicating that PEF, apart from airway obstruction, also may reflect an extrapulmonary component, presumably muscle strength. In contrast to asthma, COPD is likely to affect muscle strength. Studies have shown loss of fat-free mass in advanced COPD (13) and low BMI is in itself a predictor of mortality, both in this study and in others (14). Although we have adjusted for BMI in our model, inclusion of BMI as an indicator variable does not preclude that the effect of PEF is due to loss of muscle mass. Other extrapulmonary components than muscle strength may also be reflected to a different extent by measuring FEV₁ and PEF. It is well known that inability to perform acceptable spirometry in itself is a predictor of mortality (18), probably acting as a surrogate marker of poor lung function as well as comorbidity and general weakening. We have no data to support that PEF rather than FEV₁ correlates to extrapulmonary disabilities such as cognitive dysfunction and mental disorders, but it is likely, as effort and thereby "vitality and vigour" is a prerequisite for an optimal PEF measurement, whereas FEV₁ is less dependent of this.

It is well known that some patients with emphysema can have a marked reduction in FEV₁ because of dynamic airway collapse, with a disproportionate smaller reduction in PEF. In such patients a reduction in PEF may reflect progression from solitary emphysema to the combination of emphysema and proximal airways disease; i.e., chronic bronchitis. It has been established that chronic mucus hypersecretion, the key feature of chronic bronchitis, is an independent predictor of outcome irrespective of FEV₁ (19, 20). Our finding of PEF as an independent predictor may well be the result of the same pathophysiologic mechanisms. We would consequently expect to find lower FEV₁ values relative to PEF values in patients with emphysema than in patients with chronic bronchitis or asthma. Unfortunately, we were not able to distinguish pathoanatomical subgroups within the COPD population, but we found no disproportionate reduction in FEV₁ for the COPD population in toto compared with asthmatics, which does not preclude the presence of such a reduction in a smaller subgroup of patients with COPD. Others, however, have not been able to discriminate asthma from emphysema on basis of the PEF/ FEV₁ ratio (21), which is somewhat contradictory to the above hypothesis.

A limitation of this study originates from the selection procedures applied at the lung clinic in connection with the pulmonary function tests. Subjects who normalized their lung function on bronchodilator therapy alone did not proceed with a course of oral corticosteroids and were thus not included in our study. This is reflected in the irreversible component of lung function reduction seen in both asthmatics and subjects with COPD (Table 1). This has little importance for the COPD population, which by definition must have an irreversible component, but for asthma we cannot infer any conclusions for subjects without chronic airflow limitation as they are very sparsely represented in the study population. Furthermore, it must be acknowledged that in the present study the number of subjects and especially the number of end points were smaller for asthma than for COPD, limiting our power to detect a small independent effect of PEF in the asthmatic population.

Our smoking history was very crude and it is possible that an adjustment for full smoking history; i.e., pack-years, might have changed the results. However, from a theoretical point of view, the past smoking history is somewhat accounted for by adjusting for FEV₁, and what further needs to be accounted for is the "prospective smoking history," which is only possible by adjusting for current smoking level, as we have done. This view is substantiated by the fact that we found no significant difference in model parameters for age, sex, and best FEV₁ in a population of truly ex-smokers (subjects with COPD not currently smoking) and a population of mixed never- and ex-smokers (asthmatic subjects not currently smoking).

We have chosen not to emphasize the magnitude of the relative risks associated with a reduction in FEV₁ and PEF, respectively, as the close correlation between the two parameters makes it impossible to determine their exact individual contribution. However, if either parameter was entered singly into the model, the results were uniform with relative risk ratios between 1.9 and 2.2 per 25% decrease in lung function relative to predicted values. This applied for FEV₁ as well as PEF and for both asthma and COPD. When both parameters were entered simultaneously, the risk reduction was smaller for each parameter, as seen in Table 2.

In conclusion, we find that PEF is at least as important to prognosis in moderate to severe COPD as FEV₁, conditional on PEF being measured after optimizing lung function with bronchodilators and corticosteroids. Also, PEF adds further information in terms of prognosis, even after controlling for FEV₁. This finding cannot be reproduced in a population of asthmatics with an irreversible component of airway obstruction, in which FEV₁ generally seems to be superior to PEF, although some independent effect of PEF cannot be excluded. These observations have not been reported previously. However, the implications are important, as the possibility of substituting FEV₁ by PEF as a measure of severity in COPD can be very attractive, especially in settings with limited resources. In this context it is important to emphasize that by definition measurement of FEV₁ and FVC is still a sine qua non to establishing the diagnosis of COPD.