Inhaled corticosteroids and the risk of a first exacerbation in COPD patients

Inhaled corticosteroids (ICS) are currently the mainstay of adult asthma management and, although their benefit in chronic obstructive pulmonary disease (COPD) is still disputed 1, 2, they are also widely used in patients with COPD 3. Certain similarities between asthma and COPD, e.g. the presence of both inflammatory and bronchospastic components, provide a rationale for the use of ICS in COPD patients. During the second part of the last decade several studies designed to investigate the efficacy of ICS in the treatment of COPD have been published, with most reporting negative results 4–14. However, a few studies have suggested a beneficial effect of ICS in reducing the number and severity of exacerbations 4, 8, 12, 13. More recently, a systematic review of randomised clinical trials of ICS in COPD showed a 30% reduction in the rate of exacerbations with the use of ICS 15. However, none of the studies in this review had acute exacerbations as their primary end-point and the definition of an acute exacerbation varied or was not provided.

An acute exacerbation or “flare” of COPD is a frequent occurrence in COPD patients, although the definition of an exacerbation has varied between studies. It is characterised by an increase in symptoms often requiring a change in regular medication 16. It is believed that bacterial infection is the underlying cause in as many as 70–75% of cases, with the remainder being attributed to viruses 17 and other causes 18. Exacerbations have a considerable effect on morbidity and quality of life of COPD patients 19, and constitute an important cause of hospital admission and re-admission, with the consequent high burden on the healthcare system 19, 20. In 1995, in the USA, total direct costs of inpatient and physician services for the treatment of acute exacerbations of COPD accounted for ∼1.6 billion dollars 21 and in 1998, 662,000 hospitalisations (1.9% of all hospitalisations) were attributed to COPD 22.

In a cohort of COPD patients followed for 2.5 yrs it was observed that 35 days after an exacerbation ∼25% of cases had not reached their baseline lung function and that by day 91 complete recovery of lung function had still not occurred in 7% of cases 12. These results suggest a persistent deterioration in overall well-being with recurring episodes during the course of the disease. Hence, one important goal of treatment is to delay the time to a first exacerbation. Such a delay to first exacerbation has been shown with ICS in asthma 23. This study assessed the effect of ICS in preventing a first exacerbation of COPD in a population-based cohort of COPD patients.

Methods

Results

The inception cohort of COPD patients was composed of 4,455 individuals, after 1,192 subjects were excluded because they had an exacerbation prior to cohort entry, that is, prior to the initiation of more than occasional bronchodilator use. All patients with a first COPD exacerbation after cohort entry were identified and matched with all the available controls. Between January 1st 1990, and December 31st 1999, there were 995 patients identified with a first exacerbation, of which 381 patients were identified with a moderate exacerbation and 614 with a severe exacerbation. The majority of the patients were males and mean age was 74 yrs.

Table 1⇓ describes the distribution of the study subjects by selected characteristics, medications dispensed and possible confounding variables. The two groups were well matched, with the same mean age and similar follow-up time. Case patients consisted of a higher proportion of males and were sicker than controls, as attested to by the higher mean number of prescriptions for respiratory medications and higher mean number of prescriptions for antibiotics dispensed during the 12-month period that preceded a first exacerbation (mean number of antibiotic prescriptions for cases 1.86 versus 1.23 for controls, p<0.05). Current use of β‐blockers and/or benzodiazepines was similar in cases and controls. In addition, case patients had increased cardiac disease (p<0.025).

Table 2⇓ summarises the results of the regression analyses for different time windows of ICS exposure. The adjusted RR for any use of ICS in the previous year, as compared with no use, was 1.27 (95% CI: 1.08–1.48). Current exposure, excluding the 15-day period prior to the index date, as compared to no exposure, yielded an adjusted RR of 1.51 (95% CI: 1.22–1.87). When the prodromic period was included, the adjusted RR did not change appreciably (adjusted RR=1.54, 95% CI: 1.28–1.86). Moreover, past use of ICS and exposure in both the past and current periods were also associated with the risk of a first exacerbation (adjusted RR for past use 1.40 (95% CI: 1.17–1.67), and adjusted RR for both current and past periods 1.52 (95% CI: 1.21–1.92)). Furthermore, the risk of current exposure to ICS for a first exacerbation occurring in the first 2 yrs of follow-up was slightly greater than the risk of a first exacerbation occurring after 2 yrs from cohort entry (table 2⇓).

Crude and adjusted RR for the effectiveness of various daily doses of ICS are displayed in table 3⇓. The majority of cases and controls that took ICS used a daily dose of ≤500 µg of beclomethasone or an equivalent dose of another ICS. The risk of a first exacerbation increased with increasing daily doses of ICS, varying from 1.28 for a daily dose of ≤500 µg to 2.94 for a daily dose >1,500 µg. This corresponds to an increase of 83% for every additional 1,000 µg of beclomethasone or an equivalent dose of another ICS.

Discussion

The present study found that the use of ICS was not associated with a reduction in the risk of a first exacerbation of COPD. In fact, the risk appears to be slightly increased among users of ICS.

The current results are in accordance with the study of Renkema et al. 13, which showed that frequency and duration of exacerbations were not significantly different between the two active treatments (budesonide alone, budesonide and prednisolone) and placebo. Paggiaro et al. 8 found similar results with fluticasone, administered for 6 months, with a similar number of exacerbations in the active and placebo groups, but less severe exacerbations in the fluticasone group. In contrast, results from the Lung Health Study and from the Inhaled Steroids in Obstructive Lung Disease (ISOLDE) trial suggested fewer exacerbations 4, 7 and fewer visits to general practitioners 4 among those receiving ICS. However, these were not the primary endpoints of these studies and the patients involved were younger.

The current authors results suggest no benefit of ICS, which is in opposition to the results of a recent systematic review that showed a 30% protective effect of ICS in preventing acute exacerbations of COPD 15. However, this systematic review used information from studies where acute exacerbations were either not defined or where definitions varied substantially between studies. Furthermore, these studies combined were heterogeneous in other relevant characteristics, such as dose of ICS used and duration of follow-up. In particular, the authors did not study the first exacerbation, unlike the present work, to reduce the heterogeneity from varying disease severity and duration.

Confounding by indication is a major concern in observational studies of drug benefit, such as the current work, because the reason for prescribing a certain medication is often associated with the outcome of interest 26, introducing a bias. In theory, it would be possible to control for this, if one could measure with sufficient accuracy the effect of this confounder 26. In studies using information from large drug databases, why a particular drug was prescribed at a particular time is not known, which makes it difficult if not impossible, to control completely for this bias. The current study used the frequency of prescriptions for respiratory disease dispensed during the year that preceded the exacerbation as a measure of severity. This assumes that sicker patients are more likely to receive more medications to treat their symptoms. Moreover, a matched design was used to control for effects of age and duration of disease, above and beyond the effect of severity of COPD. The present authors noticed that after controlling for sex and severity of the disease, the association of an increase in the risk of an acute exacerbation with the use of ICS was reduced significantly. However, residual confounding was probably present and remains the most likely explanation for the increased risk of a first exacerbation observed with increasing daily doses of ICS. With increasing severity, patients are more likely to be prescribed higher doses of ICS or are more likely to be prescribed these drugs more frequently.

Protopathic bias is also a concern in observational studies and is present when one attributes an outcome to an exposure that actually resulted from early signs and symptoms of the outcome under study 26. The present study tried to avoid this bias by classifying as nonexposed the individuals who had been dispensed an ICS only during the prodromic phase, that is, during the 15‐day period that preceded an exacerbation. However, in the present study, there was no difference in the risk of a first exacerbation in current users of ICS when the period of exposure was extended to allow exposure during the prodromic phase.

The authors chose to study a true first exacerbation because the period of time between exacerbations is believed to decrease with the progression of the disease, and full recovery after an acute exacerbation may not occur 12. Therefore, if there was a benefit in using ICS it might be better perceived in the early stage of the disease. Any other exacerbation would be correlated with the preceding ones and the possible benefit of ICS might be difficult to assess. The current authors have previously reported a lack of benefit of ICS in preventing hospitalisations for an exacerbation of COPD among subjects who had been previously hospitalised for this same indication and had, therefore, progressed further in the natural course of COPD 27. Moreover, by excluding subjects who had exacerbations before cohort entry, the current study can only generalise its results to patients who experience exacerbations after starting more than occasional treatment with bronchodilators.

The current study attempted to exclude patients who were likely to have asthma by excluding subjects who had received any respiratory drugs in the 5 yrs before cohort entry. This was done to isolate the effect of ICS in subjects with COPD since the authors have previously shown that ICS are effective at decreasing severe exacerbations of asthma 23. It is unlikely that the current study was completely successful in excluding patients with asthma since this disease may have its onset after the age of 55 yrs. Inclusion of patients with asthma will have underestimated the increased risk of exacerbations of COPD that was observed with the use of ICS. Furthermore, the limitations due to assuming the current study patients had COPD based on the onset of more than occasional bronchodilator treatment after the age of 55 yrs should be considered. This was made necessary by the lack of access to medical records with information on smoking and lung function, as is usually the case in database studies. While recurring therapy with bronchodilators may be prescribed for various reasons, it appears safe to assume that most of these subjects have a chronic respiratory disease, the most common at this age being COPD.

An individual was considered to be currently exposed if there was a dispensing of a prescription in the 60‐day period prior to an exacerbation. Selection of a 60‐day period was based on two factors: 1) ICS are often prescribed in a high dose (1,000 µg·day⁻¹) for COPD patients; 2) the most commonly prescribed high‐dose metered-dose-inhaler delivered 200 doses of 250 µg each. Hence, it was estimated that a compliant patient would need to refill a prescription every 50 days. A period of 60 days was used to allow for some noncompliance and to allow for patients who were prescribed <1,000 µg·day⁻¹.

This study used information from the Prescription Drug Database of the province of Saskatchewan, which is subject to several validation checks at the time a prescription is dispensed. In addition, submitting pharmacies routinely undergo audits 24. However, in this database, information available represents only dispensed prescriptions and thus may not correspond entirely to medication used. This would result in a nondifferential misclassification of exposure with a consequent underestimation of the RR.

The results of this study depend in part on the definition of a moderate exacerbation. The authors defined a moderate exacerbation as a same day dispensing of an oral corticosteroid and antibiotic. This definition is consistent with that recently proposed during a workshop 16. However, because there is still a dispute on whether the use of antibiotics in treating exacerbations of COPD is useful 28, 29, many physicians may not prescribe antibiotics if they are uncertain that the exacerbation has a bacterial aetiology. Seemungal et al. 12 followed a cohort of COPD patients for 2.5 yrs, and observed that treatment was started at a median of 3 days after onset of symptoms; prednisolone was administered in 27.3% of exacerbations; antibiotics in 85.6%; and both an antibiotic and a oral corticosteroid were administered in only 25% of patients with an exacerbation 12. Moreover, 12% of exacerbations were left untreated. Hence, it is likely that the cases in the current study represent an underestimate of all occurrences of exacerbations, in which case point estimates of RR may be too low. In addition, for some cases the outcomes may not be their first exacerbation, but only their first exacerbation that was treated with both an antibiotic and an oral corticosteroid on the same day. A strict definition was used in this study to avoid counting as exacerbations the dispensing of antibiotics for reasons other than COPD and, because of the difficulty, in some situations, of distinguishing between prescriptions of oral corticosteroid bursts for the treatment of exacerbations from dispensing of oral corticosteroids for daily regular therapy of COPD or other conditions. Moreover, this study did not try to assess mild exacerbations, which patients usually manage themselves without seeking the assistance of a physician, for instance, by increasing the frequency of administration of his bronchodilator 16. This would have been difficult to capture in a database study. The accuracy of the information on severe exacerbations leading to hospitalisation is dependent on how well the codes of discharge diagnosis are entered. Rawson et al. 30 have previously shown the diagnosis of COPD taken from records of hospital discharge in the Saskatchewan databases to be quite reliable.

The current results can only be generalised to patients ≥55 yrs, with the onset of more than occasional bronchodilator therapy for COPD who were not treated for COPD exacerbations as the initial presentation of their illness and who had not received respiratory medications for the previous 5 yrs. Therefore, patients with COPD who remain untreated or patients whose onset of regular therapy for COPD was before the age of 55 yrs were not included in the present study.

In conclusion, the current study suggests that inhaled corticosteroids are not efficacious in preventing a first exacerbation of chronic obstructive pulmonary disease. The small increase in risk is most likely due to residual confounding by indication, but it is unlikely that the remaining unmeasured confounding would mask a significant benefit of inhaled corticosteroids.