INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Chronic obstructive pulmonary disease (COPD) is characterized by periodic exacerbations of cough, sputum, and dyspnea, frequently due to infections of the lower respiratory tract. Studies of the effect of such episodes on the progression of COPD have been inconclusive. Fletcher and coworkers studied 792 London working men, smokers and nonsmokers, with and without airway disease and concluded that episodes of lower respiratory illness had no effect on the annual rate of change of the FEV₁ (1). Similarly, Howard (2) followed British workers with and without respiratory problems for 11 yr and detected no effect of acute respiratory illness on the change in FEV₁. He noted that FEV₁ declines occurred prior to any documented respiratory illness. A 12-yr study of male smoking Canadian war veterans with "chronic bronchitis" also failed to demonstrate a role for respiratory infections in the progression of airways obstruction (3). On the other hand, in the United States, patients with COPD were followed for an average of 4 yr with weekly surveillance, and a more rapid FEV₁ decline was associated with more frequent episodes of lower respiratory infection (4). None of these studies evaluated interactions between lower respiratory illnesses and smoking status on the annual change of lung function.

The Lung Health Study (LHS) was a 5-yr, prospective, randomized, multicenter clinical trial that followed a large number of smokers with mild COPD (5). This report presents the LHS experience regarding the frequency of respiratory illnesses and the effect of respiratory illnesses on changes in lung function, as influenced by smoking status.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Study Design

The LHS was designed to determine whether smoking intervention and regular use of a bronchodilator could slow the decline in FEV₁ in persons at high risk for significant COPD (6). The LHS recruited 5,887 smokers, aged 35 to 60 yr, who were free of significant morbidity, had a baseline FEV₁/FVC less than 0.70, and an FEV₁ between 55 and 90% of predicted (7). Participants were randomized into three groups: (1) smoking intervention and inhaled ipratropium bromide (Atrovent) (SI-A); (2) smoking intervention and inhaled placebo (SI-P); and (3) usual care (UC), only advice to stop smoking.

Details of the study have been published previously (5). Participants underwent follow-up spirometry annually for 5 yr (6). Respiratory symptoms were recorded at baseline and annually with the ATS-DLD respiratory symptom questionnaire with minor modifications (8). Annually participants were asked about episodes of bronchitis, pneumonia, influenza, or chest colds resulting in physician visits during the previous year. These self-reported illnesses were grouped together as LRI. Chronic bronchitis was defined as cough and sputum on most days for 3 mo of the preceding year.

Participants were categorized as sustained quitters (stopped smoking for the entire 5-yr study period), intermittent smokers (smoking at one or more but not all annual visits), or continuous smokers (those that never stopped smoking).

Statistical Methods

Baseline differences between SI versus UC were evaluated using t tests for quantitative variables such as FEV₁ and chi-square tests for categorical variables such as sex. LRI rates were computed using the number of self-reported visits at each annual follow-up. Adjusted rates were estimated using analysis of covariance, controlling for sex, age, educational level, baseline cigarettes /d, race (African-American versus other), baseline rates of physician visits for LRI, baseline FEV₁, and O'Connor slope as a measure of airway reactivity (9).

The relationship between changes in FEV₁ during follow-up and changes in illness status was examined using a mixed-effects repeated-measures analysis of covariance (10). The dependent variable was change in FEV₁ expressed either as a percentage of predicted or in milliliters, from baseline to each annual visit. LRI status was entered as a time-dependent categorical variable.

Analyses looking at short-term changes in FEV₁ as a function of LRI during the preceding year were carried out comparing treatment groups and smoking status categories and were adjusted for the baseline covariates noted above.

The effects of LRI frequency on the 5-yr rate of decline of FEV₁ were analyzed using analyses of covariance. Separate analyses were performed for SI versus UC and by final smoking status. Analyses were controlled for age, sex, cigarettes/d, baseline FEV₁ percentage predicted, percentage bronchodilator response, and airways reactivity (O'Connor slope). Long-term LRI effects on FEV₁ decline were expressed as the change in annual rate of decline associated with one LRI/yr. Analyses were repeated excluding those participants reporting chronic bronchitis at baseline.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Baseline characteristics of participants in the two treatment groups are presented in Table 1; the two SI groups were combined, as there was no significant difference in the annual LRI frequency between SI-A and SI-P (p = 0.149). There were more women in the SI-A group (39.2%) than in the SI-P group (36.0%, p = 0.037), but otherwise they were similar. The baseline characteristics of the SI (SI-A plus SI-P) and the UC groups were similar, without significant differences (Table 1). The SI group averaged more LRI visits to a physician in the 12 mo prior to the time of randomization, but this difference did not reach significance (p = 0.098). Follow-up rates have been previously reported and were very high being  93% each year for interviews and  88% for spirometry with the highest follow-up rates occurring in the final annual visit (5). Cross-sectional and 5-yr validated sustained smoking cessation rates for the SI and UC groups are presented in Table 2. There were more sustained quitters in the SI group than in the UC group at each annual visit, and in both groups the number of sustained quitters decreased with time. In addition, there were more participants that were not smoking at each annual visit in the SI group than in the UC group. These data have been published previously (5).

Over the 5-yr study period, participants averaged 0.12-0.15 physician visits per year for "acute bronchitis," about 0.05 physician visits per year for both "pneumonia" and "influenza," and 0.07 visits per year for "chest colds." Participants averaged 0.24 physician visits per year for the combination of all these diagnoses (LRI). Baseline characteristics that were significantly associated with LRI during the study were female sex, LRI in the previous year, a relatively low FEV₁, and the presence of respiratory symptoms, including dyspnea, wheeze, and chronic cough and sputum. The associations were not particularly strong, explaining about 10% of the variability of LRI. Previous LRI and sex accounted for 8.4% of the total variation.

Significant differences were few for each specific LRI disease category between the SI and UC groups. Where significance was reached, the SI group had fewer physician visits. The mean number of LRI in each treatment group for the entire 5-yr study period is shown in Figure 1A. Only a minority of the participants attending the annual visits reported having one or more LRI, and this proportion tended to increase as the study progressed. The UC group had significantly more LRI than the SI group (p = 0.0008). Analysis by final smoking status (Figure 1B) demonstrated similar differences. Overall, continuous smokers had significantly more LRI than sustained quitters (p = 0.0003). This was because LRI increased with time in continuous smokers but not in sustained quitters. Intermittent smokers showed intermediate LRI rates.

View larger version (18K): [in this window] [in a new window]   View larger version (23K): [in this window] [in a new window]   Figure 1.   (A) Mean number (± 2 SE) of physician visits per year for lower respiratory illness by year of follow-up and treatment group. Smoking intervention versus usual care. (B) Mean number (± 2 SE) of physician visits per year for lower respiratory illness by year of follow-up and final smoking status. Sustained quitters versus intermittent quitters versus continuous smokers.

The presence or absence of chronic bronchitis ascertained at annual visits had a strong influence on LRI frequency reported at the same visits. Figure 2 shows LRI frequency in continuous smokers with and without chronic bronchitis at the five annual visits. At each visit, continuous smokers with chronic bronchitis had LRI rates that were 158-189% of those without chronic bronchitis (p < 0.001). LRI rates increased with time in continuous smokers whether or not they reported chronic bronchitis (p = 0.008). LRI rates in intermittent smokers were similar to those in continuous smokers in that they were strongly influenced by the coexistence of chronic bronchitis. In sustained quitters LRI rates were not significantly higher at annual visits in those with chronic bronchitis than in those without, but the small numbers of sustained quitters with symptoms of chronic bronchitis weakened the power of these comparisons.

View larger version (16K): [in this window] [in a new window]   Figure 2.   Mean number (± 2 SE) of physician visits per year for lower respiratory illness by year of follow-up in the continuous smoking group according to whether they had chronic bronchitis.

Table 3 presents the relationship between LRI and the short-term adjusted mean change in FEV₁ in sustained quitters and continuous smokers. FEV₁ changes are expressed as percentages of the predicted normal value and refer to data collected during the immediate previous and current annual visit, classified according to the presence or absence of LRI. Thus NI refers to the change when there were no LRI (N) in the year prior to the year of an annual visit in which an LRI occurred (I). In continuous smokers, LRI had highly significant adverse associations with FEV₁, which differed significantly according to LRI status. FEV₁ changes were less in smokers with two consecutive years without LRI (NN) than in those with LRI during the year of study (NI, II). There also appeared to be some recovery from the acute loss associated with LRI in that the IN mean decline in FEV₁ was less than that of the NI and II categories. Similar analyses of short-term changes in lung function in the SI and UC groups produced similar results. In both groups NN changes were smaller than II changes, and IN changes approached NN values. Differences between the SI and UC groups were less striking than those between sustained quitters and continuous smokers, but LRI related lung function loss was somewhat less in the SI group.

The influence of chronic bronchitis on the short-term changes in lung function shown in Table 3 was evaluated in two ways. The analysis was repeated excluding participants with chronic bronchitis at baseline, and the presence of chronic bronchitis at annual visits was entered as a covariate in the analyses of annual changes of lung function. Neither had a significant influence on the results.

The effect of LRI on the annual rate of change in FEV₁ averaged over 5 yr is shown in Figure 3, which presents the mean annual rate of change of FEV₁ in the three smoking categories as a function of the number of LRI. In sustained quitters the frequency of LRI had no effect on change of FEV₁, whereas in intermittent and continuing smokers the frequency of LRI had adverse effects on the 5-yr averaged annual rate of decline of FEV₁ (p value between sustained quitters and continuing smokers = 0.086). In both the SI and UC groups, FEV₁ decline increased with more frequent LRI, and there was no significant difference between the slopes of the relationship between LRI frequency and lung function loss (data not shown). As previously reported (5), the overall rate of decline was significantly less in the SI group than in the UC group.

View larger version (22K): [in this window] [in a new window]   Figure 3.   Mean annual change (± 2 SE) in FEV1 (ml/yr) versus mean number of physician visits for lower respiratory illnesses by final smoking status (sustained quitters, intermittent quitters, and continuous smokers).

The magnitude of the effect of LRI on the 5-yr averaged annual rate of change in FEV₁ is presented in Table 4. Sustained quitters had relatively little decline and this was unaffected by LRI. In both continuous and intermittent smoking categories, the estimated effect of one LRI/yr added to the 5-yr averaged annual decline in FEV₁ was slightly more than 7 ml /yr, or 0.23-0.24%/yr of the predicted normal value. Neither eliminating participants with chronic bronchitis at baseline nor entering annually ascertained chronic bronchitis as a covariate in the analysis altered the data of Table 4.

Cumulative effects of LRI on the 5-yr averaged annual decline in FEV₁ are shown in Table 5. Again no significant associations between LRI and annual FEV₁ decline were noted in sustained quitters. In the two smoking categories LRI frequency was positively associated with loss of FEV₁ and as LRI frequency increased it accounted for an increasingly larger fraction of the total annual loss of lung function. Smokers who had 1.5 or more LRI/yr had substantially more loss of lung function than those who had none.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

In LHS participants with mild COPD, self-reported LRI were associated with long-term adverse effects on lung function in those who continued to smoke cigarettes. Smokers with one LRI/ yr experienced an increased rate of loss of FEV₁ that averaged 7 ml/ yr (0.24% of the predicted normal) over 5 yr. Stopping smoking not only reduced the frequency of LRI, but also prevented the accelerated loss of lung function associated with these events. The finding that LRI affected FEV₁ decline only in smokers may explain the contradictory findings reported in other studies that analyzed current, former, and never smokers with and without airway obstruction as a single group (1).

The data on LRI were collected by questionnaire annually and relied on the participant reporting the information to us. We combined positive responses to questions regarding episodes of acute bronchitis, chest cold, flu, or pneumonia that necessitated a visit to a physician during the previous year into LRI as they are often used interchangeably, and any could represent an exacerbation of COPD. Analysis of only self- reported physician visits for bronchitis and chest colds gave similar, but less significant, results. Hospitalizations were too few to permit a separate analysis, although they probably included most episodes of true pneumonia. This approach has drawbacks, although it has been used in virtually all other studies assessing the frequency and significance of COPD exacerbations. We relied on participant recall, which can be erroneous. Underreporting probably occurred, but it is unlikely that quitters were more likely to underreport as they were more apt to be health conscious. However, it is also possible that quitters might deny illnesses, reinforcing their feelings that stopping smoking had improved their health. In addition to self-reported physician visits, we obtained self-reported information on the number of days in bed due to respiratory infections, which may have been recalled more accurately than physician visits. Analyses of days in bed gave similar results that were less significant due to fewer episodes. It was likely that participants had differing degrees of access to physicians, and restricted access would mean fewer visits irrespective of health status. Again, because of concern regarding personal health, a bias toward underreporting was probably more likely in continuing smokers than in quitters. Participants with symptoms and relatively low values of FEV₁ might be more likely to see a physician and thus report LRI. However, sex and LRI during the year previous to the onset of the observation period were the best predictors of LRI and the latter was more common in SI, a bias against the finding of fewer LRI in the SI group. When possible, we adjusted our analyses for these predictors.

LRI frequency in the Lung Health Study averaged 0.24/ yr. This was lower than that noted in most other studies of COPD. Several studies report exacerbation frequencies of 1- 3 /yr (13), but evaluated patients with worse baseline lung function than LHS participants. It seems reasonable that exacerbation frequency would increase with COPD severity, and we noted that LRI were more common in participants with poorer baseline lung function. On the other hand, a Danish study of largely healthy smokers recently reported an exacerbation frequency of 0.4-0.5/yr (18). Part of the discrepancy may be due to recall, as we questioned our participants annually whereas the Danish study did this every 6 mo (18). Also, our rates may have been lower because we required an associated physician visit, whereas the Danish investigators simply asked about symptoms. A recent report of respiratory viral infections in COPD where participants were contacted every 2 wk noted that in mild COPD there were 1.8 acute respiratory illnesses/yr. In those with moderate and severe COPD this number rose to 3.0/yr. Again, however, these illnesses include those with and without a visit to a physician (19). The frequency of LRI in the present study is similar to the frequency of "chest episodes" noted by Fletcher and coworkers (1). We therefore believe that differences in LRI frequency between this and other studies are largely explicable, and not due to error.

We replicated the finding of Fletcher and coworkers (1) that people with chronic cough and sputum had considerably more "chest episodes" than people who did not. In our study, differences in LRI frequency were confined to smokers, partly because sustained quitters who continued to have cough and sputum were uncommon. Other studies have shown that smokers were more likely to have respiratory infections, including pneumonia, than nonsmokers and that quitting smoking reduced this tendency (20, 21).

Remarkably, though the presence of chronic cough and sputum increased the frequency of LRI, it apparently did not influence the effect of LRI on lung function. Excluding participants with chronic bronchitis at baseline had no effect on LRI-associated decreases in FEV₁ either in the short or long term. Entering annually ascertained cough and sputum as a covariate in the analyses of the effects of LRI on lung function also did not alter the results. Thus, independent of chronic bronchitis, the data are consistent with smoking being the major determinant of the effect of LRI on lung function.

We attempted to examine the short-term effects of LRI on lung function by comparing the 1-yr change in FEV₁ according to whether an LRI occurred since the previous visit. This has the sampling problem that lung function measurements were carried out at fixed annual intervals, and not immediately before and after LRI. In spite of this, continuous smokers who reported LRI during the year preceding testing showed a larger fall in FEV₁ than those who did not. Continuous smokers in the IN category had changes that approached those of smokers in the NN category (Table 3). These data were consistent with LRI being associated with a decrease in lung function that was at least partially reversible. This interpretation is in agreement with studies of patients with established COPD (13, 17, 22) showing a transient decrease in lung function with exacerbations. The unique aspect of our data was that in sustained quitters LRI had no significant effect on short-term changes in lung function (Table 3). Thus, in people with mild COPD, the effects of LRI on relatively short-term lung function changes were dependent on smoking status; there was an additive effect of smoking and LRI on the 1-yr change in FEV₁. We are unaware of other studies addressing this issue.

Over the 5-yr study period the data are consistent with an effect of LRI on decline in FEV₁ in continuing and intermittent smokers that was not seen in sustained quitters. LRI were associated with similar degrees of 5-yr lung function decline in the SI and UC groups. However, continuing and intermittent smokers were distinct majorities in both groups and dominated intergroup analyses. When analyses compared participants by final smoking status, substantial and significant differences were observed (Figure 3). Although this kind of post hoc analysis carries risks, the simplest interpretation is that smoking and LRI had additive effects in accelerating the rate of decline of lung function in people with mild COPD. Our finding that LRI accelerated lung function decline differs from the results of most previous studies (1). This may be because we studied a large relatively homogeneous population, because of the high quality of our lung function measurements (23) and high rates of follow-up, and /or because we analyzed smokers and nonsmokers separately. Figure 3 indicates that smoking cessation favorably influenced the effect of LRI on lung function. To our knowledge this has not been previously reported.

This investigation demonstrates that LRI were associated with an accelerated loss of lung function in smokers; this may not be a cause and effect relationship. A plausible hypothesis as to how LRI might compromise lung function is that CD8⁺ T lymphocytes are found in increased numbers in the airways of smokers with COPD and in animals infected with respiratory syncytial virus (RSV) (24). A recent study noted normal CD8⁺ cell counts in ex-smoking patients with COPD (29). Blood T lymphocyte subsets in current smokers also demonstrated increased levels of CD8⁺ cells, but when the smokers had quit for 6 wk these levels reverted to those noted in never smokers (30). The role of the CD8⁺ T lymphocyte is complex (31) but in one study in RSV-infected mice the infusion into the lung of specific CD8⁺ T lymphocytes accelerated RSV clearance but resulted in additional lung damage (32). Thus, the combination of current smoking and a viral infection, a likely cause of LRI in our population (19), may have increased the numbers of lung CD8⁺ T lymphocytes. Nicotine also may act on lymphocyte receptors and affect the function of this inflammatory cell. This could result in cytokine release with recruitment of neutrophils and other cells causing further airway and/or parenchymal damage. In former smokers the background level of CD8⁺ T lymphocytes may have decreased and this decrease may have minimized airway damage. Also, the possible absence of nicotine affecting lymphocyte function could be important.

In smokers with airways obstruction, the effect of LRI upon lung function may not be trivial, especially in those with chronic cough and sputum. Although a single LRI/yr was associated with an increase in rate of FEV₁ decline of only 7 ml/yr, multiple LRI were associated with larger losses, and the presence of chronic bronchitis increased LRI frequency. Indeed, our results may well explain the finding that smokers with chronic cough and sputum have faster declines of lung function and more morbidity than those without these symptoms (33). We also observed an increase in frequency of LRI with time in smokers. This was compatible with LRI frequency increasing as lung function is compromised, a reasonable postulate, in view of studies of more severe COPD (13). The interactive effects of smoking and LRI on lung function are therefore likely accentuated as COPD progresses. As shown in Table 5, smokers with more than one LRI/yr had substantially more rapid declines of FEV₁ than those with no LRI. There are many data indicating that 1.5 or more exacerbations per year are common for patients with COPD (13).

In summary, the LHS findings demonstrate that in study participants with mild COPD stopping smoking reduced the number of reported lower respiratory illnesses that cause a visit to a physician. In addition, continuing smokers suffered both short- and long-term adverse lung function changes associated with these illnesses. In active smokers one LRI/yr resulted in a 5-yr annualized averaged acceleration in the decline of FEV₁ of about 7 ml/yr (0.24% of the FEV₁ as a percentage of the predicted normal value). Although the presence of chronic bronchitis did not affect the influence of LRI on FEV₁, it increased the frequency of LRI and presumably their cumulative effect on lung function. Stopping smoking protected these people with mild COPD from this additional loss of lung function.