| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
1 Division of Clinical Epidemiology, Royal Victoria Hospital, McGill University Health Centre, and the Departments of Epidemiology and Biostatistics and of Medicine, McGill University, Montreal, Canada; 2 Division of Pulmonary Sciences and Critical Care Medicine, Denver Health and Hospital Authority, University of Colorado School of Medicine, Denver, Colorado; 3 Veterans Affairs Medical Center, University of Minnesota, Minneapolis, Minnesota; and 4 Division of Pulmonary Sciences and Critical Care Medicine, National Jewish Medical and Research Center, University of Colorado School of Medicine, Denver, Colorado
Correspondence and requests for reprints should be addressed to Barry Make, M.D., National Jewish Medical and Research Center, 1400 Jackson Street, J211, Denver, CO 80206. E-mail: makeb{at}njc.org
ABSTRACT
The effectiveness of inhaled corticosteroids (ICS) in patients with chronic obstructive pulmonary disease (COPD) remains controversial. Randomized controlled trials, meta-analyses, medication withdrawal studies, and observational reports have examined this question, with mixed results. Observational studies have been subject to criticism because of study design involving immortal time bias. Some randomized controlled trials suggest small benefits in lung function and health status, and a reduction in the rate of acute exacerbations of COPD and mortality, but their incomplete follow-up and statistical methods have been criticized. The greatest benefits of ICS in COPD have been reported with use of ICS and long-acting ß-agonist combination therapy, although no benefit was found for the primary outcome studied under the most rigorous methodology by the recent TORCH and Optimal randomized trials. Thus, although future randomized trials will need to be conducted with the most rigorous methodology for all outcomes, much uncertainty remains regarding the potential benefits of ICS in COPD.
Key Words: chronic obstructive pulmonary disease • inhaled corticosteroids • beta-agonist bronchodilators • survival • quality of life
The use of inhaled corticosteroids (ICS), particularly in combination with a long-acting ß-agonist (LABA), has gained widespread acceptance among clinicians in the management of patients with stable COPD. ICS for COPD were widely used before rigorous evidence was gathered about their clinical benefit. This phenomenon can probably be attributed to several factors. The clinically obvious value of ICS for asthma may have encouraged the use of these drugs for COPD as well. The efficacy of systemic corticosteroids for acute exacerbations of COPD provided a rationale for prescribing ICS in stable disease. ICS became widely available in combination with a LABA for inhalation. Finally, physicians may have perceived ICS as being safer and easier to administer than certain other respiratory drugs, such as theophylline.
Regulatory approval of ICS in patients with COPD was based largely on their use in combination with a LABA and the demonstrated effect of combination therapy on lung function. Current clinical practice guidelines emphasize the use of ICS to reduce COPD exacerbations, although the approach to statistical analysis of data from the randomized trials of COPD exacerbations that form the basis for this recommendation has recently been criticized. The effects of ICS on mortality are also controversial, with the results of a large randomized controlled trial suggesting lack of statistically significant benefit in this important outcome. Although the mechanism of action of ICS has been debated, there is a growing body of evidence that ICS may affect pulmonary and systemic inflammation. This article reviews the evidence, including that from three recent randomized trials, for the effectiveness of the chronic use of ICS in patients with stable COPD.
The randomized controlled trial is the primary source of evidence of the efficacy of a drug. Once these trials consistently demonstrate a drug's effects, observational studies become an important part of the knowledge base about the drug's effectiveness due to the ability of this study design to assess infrequent and latent outcomes, and benefit on major disease outcomes. Concordance of the findings between randomized controlled trials and observational studies is an essential element in assessing the overall effect of the drug. Such coherence regarding the benefit of ICS has been observed in asthma (1). In the treatment of COPD, however, the data on the effectiveness of ICS are at times contradictory among and between randomized trials and observational studies. The purpose of this article is to review the controversy over the interpretation of the published evidence on the use of ICS as chronic therapy in patients with COPD.
MECHANISM OF ACTION
COPD is characterized by both pulmonary and systemic inflammation. Because of their antiinflammatory properties, corticosteroids have been used for decades in the treatment of COPD exacerbations. However the side effects of systemic corticosteroids make them unsuitable for long-term treatment of stable COPD. Inhaled delivery of corticosteroids provides antiinflammatory effects while minimizing systemic absorption and side effects. ICS alone or in combination with an LABA reduce airway inflammation detected by endobronchial biopsy (2–4). Although no study has compared the antiinflammatory effects of ICS alone with an ICS/LABA combination, the observed antiinflammatory effects were greater in the study using ICS/LABA than in the previous smaller studies looking at ICS alone. In addition to their influence on airway inflammation, ICS, like systemic corticosteroids, cause significant reduction in systemic inflammation as measured by C-reactive protein (5).
Corticosteroids act primarily through binding to intracellular glucocorticoid receptors, which regulate gene expression through glucocorticoid response elements (GREs). Recent work has demonstrated that much of the antiinflammatory effects of corticosteroids are not directly regulated by GREs but are instead mediated by downstream influences on the state of histone acetylation (6). Corticosteroids promote chromatin compaction and silencing of gene expression by inhibiting histone acetyltransferase activity and by promoting the activity of a histone deacteylase (HDAC2) (7). Impaired HDAC2 function has been observed in patients with COPD (8), and may explain the observation that the inflammatory state of COPD is relatively resistant to the effects of corticosteroids when compared with more steroid-sensitive inflammatory diseases such as asthma (9).
Molecular mechanisms provide a rationale for administering ICS in combination with bronchodilators. Corticosteroids up-regulate ß2-adrenoreceptors (10) and can prevent or reverse down-regulation (tachyphylaxis) of ß2-adrenoreceptors in response to agonists (11, 12). ß2-Agonists, in turn, increase translocation of glucocorticoid receptors from the cytoplasm to their site of action in the nucleus (13–16). Laboratory studies suggest complementary biochemical effects of ICS and LABAs in airway cells that might have clinical utility (17). Corticosteroids may also be effective in combination with theophylline, because theophylline activates HDAC2 and may accentuate corticosteroid action through this mechanism (6, 9).
EFFECTS ON LUNG FUNCTION
Because corticosteroids are potent antiinflammatory agents, it was posited that long-term administration of ICS might favorably affect the natural course of COPD, as demonstrated by the decline in FEV1 over time. Numerous trials tested this hypothesis, and the largest of these trials used a nearly identical study design (18–21). High doses of ICS or placebo were administered to patients with mild to moderate COPD for a period of several years, with the slope of the FEV1 over time being the primary outcome. In none of these trials did ICS have any statistically significant beneficial effect on the decline in FEV1, although trends were noted in some. Two meta-analyses of the results from these four studies together with several other smaller trials reported contradictory conclusions (22, 23). One group of investigators concluded that ICS reduce FEV1 decline by a small but statistically significant mean rate of 7.7 ml/year (95% confidence interval [CI], 1.3–14.2 ml/yr), whereas the second group concluded that the reduction in the mean rate was only 5.0 ml/year (95% CI, –1.2 to 11.2 ml/yr) and was not statistically significant.
Although there is uncertainty as to whether ICS have any effect the long-term decline of the FEV1 in patients with COPD, most randomized trials did demonstrate a small, sustained improvement in the FEV1 from ICS alone. Average improvements were usually in the range of 50 to 75 ml. Bronchodilation is apparent shortly after initiating ICS treatment, and it persists for the duration of therapy, albeit declining at the same rate as placebo in these individual studies (19, 20, 24, 25). Results from the recently published TORCH (Toward a Revolution in COPD Health) and the recent Canadian trials are consistent with these prior findings (25, 26). A third meta-analysis of the influence of ICS on lung function was recently published, showing that this initial increase in FEV1 was more pronounced in former smokers and in women; again, there was no significant change in the rate of decline in FEV1 over time (27). Investigators from the Inhaled Steroids in Obstructive Lung Disease in Europe (ISOLDE) study demonstrated that FEV1 changes after a 2-week course of prednisolone were not predictive of ICS-induced FEV1 responses, or of any other efficacy outcome at any time during the 3-year follow-up period (28).
EFFECTS ON SYMPTOMS AND QUALITY OF LIFE
Most large randomized ICS trials assessed symptoms and quality of life as secondary outcomes. Several trials variably reported small improvements in cough and dyspnea (29–31). A number of the larger trials evaluated quality of life and health status with the well-validated and disease-specific St. George's Respiratory Questionnaire (SGRQ) (19, 24, 28, 32, 33). Compared with placebo, patients receiving ICS experienced mean improvements in the SGRQ of between approximately 1 and 3 units in each of the individual trials. Results from the TORCH trial are consistent with these findings (25); over a 3-year period, the adjusted mean SGRQ improved by –2.0 units (95% CI, –1.0 to –2.9) when the fluticasone treatment arm was compared with the placebo arm. Hence, ICS appear to confer an average improvement in the SGRQ that falls short of the 4-unit change in the SGRQ that is judged to be clinically important (34). Responses in individual patients vary, so it is likely that a small proportion of patients treated with ICS may achieve a clinically noticeable improvement of 4 units or better.
EFFECTS ON EXACERBATIONS
Exacerbations of COPD adversely impact health status and exact a huge economic toll. COPD exacerbations requiring hospitalization are particularly important not only because they are highly morbid events but they also account for one-half or more of total medical costs for treating this disease (35, 36). Thus, preventing exacerbations represents an important management priority. Numerous randomized trials, most conducted within the past 10 years, suggest that newer inhalation therapies, such as ICS and long-acting inhaled bronchodilators, have a favorable impact on COPD exacerbation rates (37). The pathogenesis and etiology of COPD exacerbations are not well delineated and the definition of COPD exacerbations used in different trials has not been uniform. Earlier definitions based on symptoms alone have recently been modified to include the need for changes in therapy (38, 39). Although hospital admissions for COPD-related illnesses is a more well-defined outcome, less severe exacerbations are not always captured and may be subject to recall bias when the interval between assessments is long. The variation in exacerbation definitions confounds interpretation between different trials and influences study interpretation. Moreover, the study design, statistical methodology, and interpretation of the results from many of the randomized trials have been a source of controversy.
A recent systematic review identified 10 randomized studies in patients with COPD that reported exacerbation data, comparing ICS with placebo (37). These 10 trials met prespecified quality criteria, had a follow-up period of least 6 months, and enrolled a total of 3,724 patients. Compared with placebo, there was a 22% relative reduction (relative risk [RR], 0.78; 95% CI, 0.70–0.88) in the percentage of ICS-treated subjects who experienced one or more exacerbations. The pooled absolute risk reduction was 5% (95% CI, –8 to –3). There is no evidence that any single drug used in the various trials (fluticasone, budesonide, or beclomethasone) was more effective than any other. Medical care utilization related to exacerbations, such as hospitalization, was infrequently reported and described differences were small. In the TORCH trial, fluticasone, compared with placebo, reduced moderate to severe exacerbations by 18% (RR, 0.82; 95% CI, 0.76–0.89) but did not statistically significantly reduce hospitalizations due to COPD (RR, 0.88; 0.74–1.03) (25).
Sin and associates pointed out that larger ICS-related reductions in exacerbations tend to be seen in trials that enrolled patients with lower mean baseline values of FEV1 (40). Limited data indicate that a similar phenomenon may be discernible within individual trials. In the ICS withdrawal trial of van der Valk and associates, the hazard ratio (HR) was 2.1 (95% CI, 1.1–3.6) for time to first exacerbation among the patients with an FEV1 less than 50% predicted, but the HR was only 1.2 (95% CI, 0.8–2.0) in patients with a higher FEV1 (33). A secondary analysis of results from the ISOLDE trial indicated that fluticasone, compared with placebo, caused a larger absolute decrease in exacerbations among patients with worse lung function than in those patients with better function (41). Because relatively few exacerbations occurred in patients with better lung function, these results should be interpreted with caution.
One meta-analysis of randomized trials evaluating exacerbations as secondary outcomes reported an impressive 30% reduction (95% CI, 16–42%) in the rate of COPD exacerbations associated with ICS (42). Such an effect would clearly be important not only to patients but also to the health care system because exacerbations of COPD are associated with reduction in health status and those leading to hospitalization are an important factor leading to increased health care utilization and medical costs. This meta-analysis, however, presented findings that were at odds with those of the original studies. The rate ratio of exacerbation for ICS relative to placebo for some studies was presented as statistically significant when, in reality, the individual studies had reported these differences as nonsignificant or much less significant. For example, the Lung Health Study reported a nonsignificant p value of 0.07 for exacerbation rates, whereas the meta-analysis calculated a 95% CI that corresponds to a significant p value of 0.005. The ISOLDE trial reported a median rate ratio of 0.75 with a 95% CI of 0.6 to 1.0, whereas the meta-analysis computed a rate ratio of 0.67 with a 95% CI of 0.63 to 0.71. These contradictions in estimation from the same data were discussed by Suissa in whose report the original papers were reviewed and compared with the meta-analysis and methodological aspects presented (43).
The correct statistical methodology stresses that the length of follow-up should be considered when analyzing exacerbation rates in clinical trials. Most randomized trials, however, had estimated the mean rate of exacerbation for the group by using an average that was unweighted by the length of follow-up time. Suissa showed that such unweighted mean rates can be biased downward, so that ICS could be believed to be more effective than they actually are (43). Instead, the mean rates have to be weighted by the length of follow-up time to provide unbiased estimates of the true exacerbation rate of the group. As a result, the estimates from each of the randomized trials used in the meta-analysis, generated from unweighted averages, are necessarily biased to various extents.
The meta-analysis has also been criticized for the approach used to estimate the CI and p value of the rate ratio (43). The method used only variation in exacerbations within a patient over time but not the variation on exacerbation rates between patients. Thus, the Poisson distribution assumes that all patients in a group are homogeneous with respect to their rate of exacerbation. This assumption does not recognize that patients with COPD are very heterogeneous, with some patients having frequent exacerbations and many who may have no exacerbations. As a result, the p value and CI surrounding the rate ratio of exacerbation computed in the meta-analysis underestimated the true p value and CI for each of the studies and consequently for the overall estimate. These biases introduced in the meta-analysis of these studies thus resulted in an exaggeration of the benefits of ICS.
The TORCH trial reported a significant reduction in the overall 3-year rate of exacerbations with fluticasone relative to placebo (RR, 0.82; 95% CI, 0.76–0.89), but not for severe exacerbations requiring hospitalization (RR, 0.88; 95% CI, 0.74–1.03) (25). This study used the negative binomial model to fit the exacerbation rate. This model also uses the same Poisson distribution for the number of exacerbations within patients, and thus correctly weights for follow-up time, but uses a 
distribution to fit the variability in the exacerbation rate between patients. This assumes that the between-patient rates are well fit by the model, which may be incorrect, whereas the pure Poisson approach simply estimates the variance directly from the data without imposing a model. Thus, the rate reductions are correct but the confidence limits may be too narrow in view of the very wide between-patient variability in COPD exacerbation rates.
ICS Withdrawal Studies
In addition to trials which compare treatment with ICS to placebo, four studies examined the withdrawal of ICS therapy from patients with COPD. In a report of the run-in phase from the ISOLDE trial, patients who had ICS withdrawn experienced a much higher rate of exacerbation than those who were not on ICS therapy (38 vs. 6%) (44). In a small crossover study, O'Brien and colleagues report that withdrawal of inhaled budesonide was associated with decreases in FEV1 and increased exercise-induced dyspnea (45). Two larger randomized controlled trials examined the effect of ICS withdrawal after a run-in period of ICS treatment. Wouters and coworkers found that those randomized to ICS withdrawal experienced decreases in FEV1 and FEV1/FVC ratio, increases in the rate of mild exacerbations, increased dyspnea, and more nights of sleep disturbance from respiratory symptoms (46). van der Valk and associates also found increased exacerbations and decreased health status in those randomized to ICS withdrawal compared with ICS maintenance therapy (33). The deleterious effects of ICS withdrawal also suggest pharmacologic efficacy of these medications.
EFFECTS ON MORTALITY
Meta-analysis of Older Randomized Controlled Trials
Until recently, randomized controlled trials of ICS were powered on clinical endpoints other than mortality. Individually, these trials were unable to detect differences in all-cause mortality between the patients assigned to ICS and those assigned to placebo. However, several meta-analyses of these earlier ICS trials have been performed in the attempt to overcome the limitations of their smaller sample sizes.
The Inhaled Steroid Effects Evaluation in COPD (ISEEC) review obtained individual patient-level data from all randomized controlled trials comparing ICS with placebo that were 12 months or longer in duration (47). This study included seven trials and 5,085 patients. The authors demonstrated a statistically significant 27% relative reduction in all-cause mortality (HR, 0.73; 95% CI, 0.55–0.96). However, because the overall mortality was only 4%, there was only a 1% absolute risk reduction (with corresponding number needed to treat to prevent one death = 50). The strength of this study stems from having obtained individual patient-level data from each trial, which allowed for the application of survival analysis techniques (i.e., Kaplan-Meier curves). The main weakness of this pooled analysis is that ascertainment of mortality was incomplete for subjects who did not complete the study, which was more likely for the placebo patients and thus can result in an apparent, rather than a real, difference in mortality (48). Because dropouts were more common in the patients receiving placebo, this may have introduced a bias favoring treatment. The complete ascertainment of outcomes during follow-up, as was done in the TORCH trial for mortality, is essential to provide an unbiased intent-to-treat analysis. Such bias is suggested by the observation that the reduction in all-cause mortality in the pooled analysis essentially stemmed from deaths due to cancer and to other causes, but not to a reduction in respiratory or cardiac deaths. This bias is confirmed by the TORCH trial, which found no reduction in the 3-year mortality rate with fluticasone therapy (RR, 1.06; 95% CI, 0.89–1.27) (25). The biologic plausibility of the role of ICS in reducing deaths has not been demonstrated.
Two additional meta-analyses also included trials with shorter durations of follow-up. Wilt and others, on behalf of the Agency for Healthcare Research and Quality, found trends toward reduced mortality with ICS alone (RR, 0.81; 95% CI, 0.60–1.10) or in combination with an LABA (RR, 0.66; 95% CI, 0.32–1.38), although these trends did not achieve statistical significance (37). Gartlehner and others examined trials of ICS alone and found a similar nonsignificant trend toward reduced mortality in the patients assigned to treatment (RR, 0.81; 95% CI, 0.60–1.08) (49). These were also based on studies that did not observe patients until the end of the prestated follow-up period.
Observational Studies
Many observational studies using a variety of administrative databases have examined the relationship between use of inhaled steroids and mortality in patients with COPD (49–63). Most of these studies have reported an association between ICS use and decreased all-cause mortality and additional benefit when ICS were used in conjunction with an LABA (58–60). The magnitude of benefit reported in these studies ranges from 16 to 54% relative risk reduction with ICS alone, and from 35 to 66% relative risk reduction with the combination of ICS and LABA. Although many trials report estimates consistent with those seen in the meta-analyses described above, others report much different estimates of risk. Although it is possible that inhaled steroids may have more benefit in the "real world" patients included in these cohort studies than in the carefully selected populations enrolled in the clinical trials, it is more likely that these findings may be due to methodological flaws in individual studies. The observational studies cited above were all shown to be affected by immortal time bias, where exposure to ICS is defined in such a way to artificially give the ICS-exposed patients a survival advantage, thus explaining the large effects on mortality reported using these methods. This bias has cast doubt on the robustness of these reports' findings (52, 63–65).
Immortal time refers to a period of follow-up or observation time during which the outcome under study cannot occur (66, 67). For example, the Sin and Tu study used a cohort of over 22,000 patients followed for 1 year after discharge for COPD hospitalization (cohort entry) until death (57). Patients were then classified as exposed if they were dispensed a prescription for an ICS during the first 90-day period after cohort entry and unexposed otherwise. Data analysis was based on comparing the time from cohort entry until death between the two groups, resulting in a 29% reduction in the rate of all-cause mortality (HR, 0.71; 95% CI, 0.65–0.78). However, the time between cohort entry and the time of the first ICS prescription, for the exposed subjects, is immortal and unexposed because the subjects have yet to receive their first ICS (as depicted in Figure 1). Thus, in comparing the survival times between the two groups, the ICS users will appear to have a longer survival. This difference, however, is simply the result of having artificially created a survival advantage by this immortal time in the ICS users, but not in the patients who did not receive ICS (52). This bias appears to be confirmed by the TORCH trial, which did not show a survival advantage in the fluticasone group (RR, 1.06; 95% CI, 0.89–1.27).
|
Recently, Kiri and colleagues presented another observational study of the effect of ICS on mortality claiming it to be "free of immortal time bias" (55). This claim was shown to be false (64). The authors excluded from the analysis 2,769 patients who were not prescribed ICS on the day of discharge but received an ICS later in the year of follow-up. In doing so, the authors excluded an important amount of unexposed and immortal time, thus introducing a significant degree of immortal time bias in the results. A reanalysis of these data shows that the reported HR of mortality with ICS of 0.70 was biased and should have been 1.48 (64).
The solution to studies that introduced immortal time bias is to use a time-dependent definition of exposure that properly classifies this immortal person-time as unexposed until the start of drug use and exposed thereafter (52, 63, 64). Until time-dependent analyses are performed, observational studies suggesting that ICS have such large effects in reducing mortality should be suspected of being affected by immortal time bias (68, 69).
The TORCH Trial
The TORCH study was designed to examine compare the twice-daily inhalation of a combined inhaled steroid/LABA (fluticasone, 500 µg, and salmeterol, 50 ug) to placebo on all-cause mortality (23). This study followed 6,112 subjects with COPD for 3 years; subjects were randomized to receive ICS, an LABA, an ICS/LABA combination, or placebo. All subjects were followed for the entire 3-year period for mortality, regardless of the time the assigned drug was discontinued. Compared with placebo, there was a 17.5% reduction in all-cause mortality with the ICS/LABA combination (HR, 0.825; 95% CI, 0.681–1.002), although this reduction was not statistically significant after adjusting for interim analyses (p = 0.052). There was no effect of ICS alone on mortality (HR, 1.06; 95% CI, 0.89–1.27) compared with placebo.
The TORCH trial apparently did not find a significant interaction between treatment and disease severity as measured by FEV1. This is in contrast with a meta-analysis of smaller trials that reported that ICS are more effective in patients with more severe disease (40), although here again, this meta-analysis may have been subject to the methodological limitations discussed above. If the relative risk reduction in mortality is equal across all stages of COPD severity as measured by the FEV1, the potential absolute benefit of treatment still depends on disease severity, as illustrated in Table 1. This could have important implications for the risk/benefit profile and cost-effectiveness of these medications.
|
ICS and LABAs are now widely prescribed for COPD as combined formulations (e.g., salmeterol/fluticasone, formoterol/budesonide). A number of large randomized controlled studies, including the recent TORCH trial, evaluated LABAs and ICS, individually and in combination, against placebo (24, 25, 29–32). In addition, Kardos and colleagues recently reported a large two-arm randomized trial in which combined therapy with fluticasone plus salmeterol for 1 year was compared with salmeterol alone in patients with severe, exacerbation-prone COPD (70). Aaron and colleagues reported this year on a Canadian three-arm randomized trial of patients with moderate or severe COPD, all given a base therapy of tiotropium and allocated randomly to additionally receive placebo, salmeterol, or fluticasone/salmeterol for 1 year (26).
Effects on Lung Function
When ICS are given together with an LABA, an additive bronchodilator effect from ICS is apparent, with most studies showing a mean FEV1 improvement in the range of 50 to 75 ml (19, 25, 29–32, 70).
Effects on Quality of Life
Most trials also show some further improvement in the SGRQ score when ICS are added to an LABA, but in none of the trials did the average increment achieve the 4-unit change that is judged to be clinically important (19, 25, 29–32, 70).
Effects on Exacerbations
Most trials show some reduction in exacerbation frequency when an ICS is combined with an LABA, but the effect tends to be small, and a recent systematic review of published trials was unable to show that combination therapy was statistically superior to LABA monotherapy (RR, 0.82; 95% CI, 0.65–1.04) (37). However, the recently published study by Kardos and colleagues reports a 35% reduction in exacerbation rate with fluticasone plus salmeterol relative to salmeterol alone (RR, 0.65; 95% CI, 0.57–0.76) (71). Although this study used a Poisson analysis weighted for follow-up time, it did not account for the wide between-subject variation in exacerbation rates, explaining the unusually and erroneously tight confidence limits and low p value (p < 0.0001). A recent Canadian study, on the other hand, using the proper data analysis, found no difference in the proportion of subjects who experienced an exacerbation during the year or in the exacerbation rate with fluticasone plus salmeterol versus placebo (RR, 0.85; 95% CI, 0.65–1.11), but found a reduction for exacerbations requiring hospitalization (RR, 0.53; 95% CI, 0.33–0.86) (26). The results from TORCH are generally consistent with previous trials (25). Adding fluticasone to salmeterol resulted in a statistically significant improvement in the SGRQ (–2.2 units; 95% CI, –3.1 to –1.2 units) and reduction in the annualized COPD exacerbation rate (RR, 0.88; 95% CI, 0.81–0.95).
Until recently, a common flaw of these randomized trials has been the interruption of follow-up as soon as a patient stopped using the study drug. The TORCH trial was the first to actually follow patients for mortality until the end of the 3-year study period, regardless of drug discontinuation (25). Unfortunately, the secondary outcomes such as exacerbations and quality-of-life measures, which are reported as being significantly improved with the combination therapy, were not assessed until the end of follow-up. The recent Canadian study, which is the only other trial to have followed patients beyond drug discontinuation, did so for all outcomes including exacerbations (26). The authors presented the findings both ways: using the complete follow-up, the rate ratio of exacerbation with fluticasone plus salmeterol versus placebo was 0.85 (95% CI, 0.65–1.11) compared with 0.79 (95% CI, 0.54–1.14) when follow-up was stopped at the time of drug discontinuation.
ADVERSE EFFECTS OF ICS
ICS in patients with COPD are associated with a variety of adverse effects. The most common of these are local effects in the upper airway, presumably secondary to the high concentrations of deposited drug. Sin and coworkers identified an increased incidence of oral thrush in pooled data from six randomized trials (RR, 2.98; 95% CI, 2.09–4.26) as well as an increased incidence of dysphonia in a summary estimate from four of these same studies (RR, 2.02; 95% CI, 1.43–2.83) (40). These complications can be largely mitigated through use of spacers and rinsing the mouth after inhalation of drug, and only infrequently cause discontinuation of ICS therapy.
All trials of ICS in patients with COPD have used moderate to high doses of drug, and higher doses are known to be associated with some systemic absorption (29). Systemic effects to some degree may be specific to the ICS formulation that is administered (72). Compared with placebo, patients with COPD taking high doses of ICS exhibit a higher incidence of skin bruising together with small reductions in mean morning cortisol levels (19, 29, 32, 40). Judging from the degree of adrenal suppression, systemic effects from ICS appear to be small and none of the study patients exhibited clinical signs or symptoms of adrenal insufficiency. However, there is still some concern that small systemic effects over long periods of treatment might exert deleterious effects, particularly on bone and eye health.
Extended therapy with oral corticosteroids frequently causes severe osteoporosis (73). Bone mineral density (BMD) has been evaluated in three randomized placebo-controlled trials of ICS therapy in patients with COPD. The Lung Health Study II reported that inhaled triamcinolone acetonide, 1,200 µg daily, reduced femoral neck BMD by 2% over a 4-year period, whereas the (European Respiratory Society Study on COPD) EUROSCOP trial found no effect of budesonide, 800 µg daily, on BMD over a 3-year period (21, 74). The TORCH study evaluated bone density at the hip and lumbar spine in a safety substudy of 658 subjects as well as the development of new fractures with high-dose inhaled steroid (fluticasone, 500 ug) (25). The reasons for these discrepant results are not apparent. Three recent systematic reviews pooled results from prospective studies that evaluated ICS in patients with COPD or adult asthma. Authors of each of these reviews concluded that ICS had no effect on BMD or fracture rates and probably had no effect on bone markers except at very high doses of ICS (75–77). Observational studies of the relation between ICS use in patients with COPD and fracture risk have yielded mixed results, but some have reported that high-dose ICS may increase nonvertebral fracture rate (78, 79), whereas others suggest that only long-term use and very high doses increase this risk (80). There were no differences compared with patients receiving placebo.
Posterior subscapular cataracts are a well-known complication of extended oral corticosteroid exposure (73). Several large observational studies reported an association between ICS in patients with COPD and posterior subscapular and nuclear cataracts (48, 81–83). Although these risks were generally more pronounced in patients who had used higher doses of ICS for longer periods, one study suggests that this risk could be apparent even at relatively low doses (48). On the other hand, several placebo-controlled trials found no increased incidence of cataracts or other adverse eye effects from high-dose ICS during follow-up periods of 3 to 4 years (19–21). The TORCH study performed slit lamp examinations in a safety substudy of 658 subjects with high-dose inhaled steroid (fluticasone, 500 µg, twice daily) (23). Only 29% of subjects did not have evidence of cataracts at the baseline examination. There were no differences in the development of new cataracts compared with patients receiving placebo. Thus, the relation of ICS to cataract formation is not totally clear. Randomized trials are generally regarded as the more reliable source of evidence, but they typically can provide only a few years of follow-up. Observational studies offer the advantages of larger sample sizes and more extended follow-up, but they are subject to the effects of unidentified confounders.
Kardos and associates as well as the TORCH investigators recently reported a previously unrecognized complication of fluticasone, 500 µg, given twice daily (25, 71). Patients in all fluticasone treatment arms in both studies experienced an excess in the rate of nonfatal pneumonia of about 3 per 100 per year compared with no fluticasone. This complication may be related to the known immunosuppressive effects of corticosteroids and the locally high concentrations achieved in the lung. The clinical significance of this finding is unclear and awaits further information.
CONCLUSIONS
ICS are commonly used to treat patients with COPD in an attempt to improve outcomes by decreasing airway inflammation. Although some randomized controlled trials suggest that treatment with ICS results in small improvements in lung function and health status in patients with COPD, others do not, rendering the clinical significance of these findings uncertain. The scientific evidence to support ICS use derives largely from secondary analyses of randomized controlled trials, meta-analyses of these trials, and observational studies of major outcomes such as exacerbations and mortality.
A critical review of the statistical treatment of exacerbation rates revealed major discrepancies between the data reported in the randomized trials and those used in the meta-analysis. In addition, the statistical techniques used in trials and in the meta-analysis were shown to be biased from their use of unweighted estimates of the rates, while also exaggerating the statistical significance in the meta-analysis. Consequently, the validity of the oft-quoted 25% reduction in the rate of COPD exacerbations associated with ICS from the meta-analysis of these randomized trials can be questioned. The data on mortality that come mostly from observational studies based on a variety of administrative databases have been shown to be affected by immortal time bias. This bias tends to exaggerate the benefit of ICS by giving an artificial advantage to the users of ICS arising from the definition of ICS use. The pooled data analysis of all randomized controlled trials comparing ICS with placebo, which suggests a significant reduction in all-cause mortality with ICS, is biased by the incomplete follow-up of subjects in all these trials. This bias was subsequently further suggested by the TORCH randomized trial, which found no reduction in mortality with ICS after complete follow-up of patients.
Recently, ICS have been used clinically in conjunction with LABAs, and there appear to be some benefits of such combination therapy. The evidence for ICS/LABA use also comes from randomized controlled trials, meta-analyses of randomized controlled trials, and an observational study. Whereas the observational study has been shown to be subject to immortal time bias, it is uncertain whether the randomized trials and meta-analyses are not also limited by the same methodological problems that have plagued the studies of ICS alone. Although the results of the TORCH trial did not demonstrate benefit from ICS alone or in combination with an LABA on mortality, these reports suggest that combined ICS plus LABA therapy may have a beneficial role in the management of patients with COPD due to improvements in lung function, health status, and rates of exacerbation and hospitalization. Unfortunately, however, although the trial was impeccably conducted with complete follow-up for mortality, it had incomplete follow-up to assess these very same secondary outcomes of lung function, health status, and exacerbations that suggest benefit.
Future randomized trials will need to be conducted with the most rigorous methodology for all outcomes. Moreover, future work should attempt to quantify the cost–benefit and risk–benefit profiles of these medications, and focus on identifying whether certain patients with COPD derive particular benefit from these medications.
FOOTNOTES
This article was presented verbally at a symposium, funded by Sepracor, Inc., held on October 13, 2006, in Boston, Massachusetts.
Conflict of Interest Statement: S.S. has been reimbursed for attending advisory board meetings and conferences and participating as a speaker in scientific meetings financed by various pharmaceutical companies (AstraZeneca, Boehringer Ingelheim, GlaxoSmithKline, Pfizer, and Sepracor). He received funding for research grants from AstraZeneca and GlaxoSmithKline. R.M. does not have a financial relationship with a commercial entity that has an interest in the subject of this manuscript. D.N. received fees from Boehringer Ingelheim, Pfizer, Adams Respiratory Therapeutics, AstraZeneca, Shering Plough, and Sanofi Aventis for consultancies, or serving on advisory boards. He also received grants from Boehringer Ingelheim and GlaxoSmithKline to perform clinical studies. B.M. received $3,000 from Sepracor in 2006 for participating in a conference and presenting some of the information in this manuscript. He received $17,500 from GlaxoSmithKline in 2006, $23,500 in 2005, and $22,000 in 2004 for attending advisory boards and providing talks on its products; and $12,500 from Schering in 2006 for conducting and attending advisory boards. His institution has received funds from GlaxoSmithKline over the last 3 years for conducting clinical trials in which he has been the principal investigator. His medical center has received unrestricted educational grants and research funds and has purchased products for its pharmacy from pharmaceutical companies that manufacture inhaled corticosteroids and long-acting ß-agonists over the last 3 years.
(Received in original form January 22, 2007; accepted in final form April 14, 2007)
REFERENCES
This article has been cited by other articles:
|
|
 |
|
  J. F. Dummer, M. J. Epton, J. O. Cowan, J. M. Cook, R. Condliffe, C. E. Landhuis, A. D. Smith, and D. R. Taylor Predicting Corticosteroid Response in Chronic Obstructive Pulmonary Disease Using Exhaled Nitric Oxide Am. J. Respir. Crit. Care Med., November 1, 2009; 180(9): 846 - 852. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 G. J. Rodrigo, J. A. Castro-Rodriguez, and V. Plaza Safety and Efficacy of Combined Long-Acting {beta}-Agonists and Inhaled Corticosteroids vs Long-Acting {beta}-Agonists Monotherapy for Stable COPD: A Systematic Review Chest, October 1, 2009; 136(4): 1029 - 1038. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 S. Singh, A. V. Amin, and Y. K. Loke Long-term Use of Inhaled Corticosteroids and the Risk of Pneumonia in Chronic Obstructive Pulmonary Disease: A Meta-analysis Arch Intern Med, February 9, 2009; 169(3): 219 - 229. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
