Published ahead of print on May 29, 2008, doi:10.1164/rccm.200712-1869OC
© 2008 American Thoracic Society doi: 10.1164/rccm.200712-1869OC
Effect of Pharmacotherapy on Rate of Decline of Lung Function in Chronic Obstructive Pulmonary DiseaseResults from the TORCH Study1 Tufts University School of Medicine, Pulmonary and Critical Care Division, Caritas-St. Elizabeth's Medical Center, Boston, Massachusetts; 2 Respiratory Medicines Centre, GlaxoSmithKline, Greenford, Middlesex, United Kingdom; 3 Pulmonary Research, Institute of Southeast Michigan, Livonia, Michigan; 4 Woolcock Institute of Medical Research, Camperdown, New South Wales, Australia; 5 Division of Cardiac and Vascular Science, St George's University of London, London, United Kingdom; 6 Cardiology and Respiratory Medicine, Hvidovre Hospital, Hvidovre, Denmark; 7 North West Lung Centre, Wythenshawe Hospital, Manchester, United Kingdom; 8 Respiratory Medicines Centre, GlaxoSmithKline, Research Triangle Park, North Carolina; and 9 Department of Medicine, University Hospital Aintree, Liverpool, United Kingdom Correspondence and requests for reprints should be addressed to Bartolomé Celli, M.D., Professor of Medicine, Tufts University School of Medicine, Pulmonary and Critical Care Division, Caritas-St. Elizabeth's Medical Center, Boston, MA, 02135-2997. E-mail: bcelli{at}copdnet.org
Rationale: Chronic obstructive pulmonary disease (COPD) is characterized by an accelerated decline in lung function. No drug has been shown conclusively to reduce this decline. Objectives: In a post hoc analysis of the Toward a Revolution in COPD Health (TORCH) study, we investigated the effects of combined salmeterol 50 µg plus fluticasone propionate 500 µg, either component alone or placebo, on the rate of post-bronchodilator FEV1 decline in patients with moderate or severe COPD. Methods: A randomized, double-blind, placebo-controlled study was conducted from September 2000 to November 2005 in 42 countries. Of 6,112 patients from the efficacy population, 5,343 were included in this analysis. Measurements and Main Results: Spirometry was measured every 24 weeks for 3 years. There were 26,539 on-treatment observations. The adjusted rate of decline in FEV1 was 55 ml/year for placebo, 42 ml/year for salmeterol, 42 ml/year for fluticasone propionate, and 39 ml/year for salmeterol plus fluticasone propionate. Salmeterol plus fluticasone propionate reduced the rate of FEV1 decline by 16 ml/year compared with placebo (95% confidence interval [CI], 7–25; P < 0.001). The difference was smaller for fluticasone propionate and salmeterol compared with placebo (13 ml/year; 95% CI, 5–22; P = 0.003). Rates of decline were similar among the active treatment arms. FEV1 declined faster in current smokers and patients with a lower body mass index, and varied between world regions. Patients who exacerbated more frequently had a faster FEV1 decline. Conclusions: Pharmacotherapy with salmeterol plus fluticasone propionate, or the components, reduces the rate of decline of FEV1 in patients with moderate-to-severe COPD, thus slowing disease progression. Clinical trial (GSK Study Code SCO30003) registered with www.clinicaltrials.gov (NCT00268216).
Key Words: FEV1 • salmeterol • fluticasone propionate • disease progression
Chronic obstructive pulmonary disease (COPD), a major cause of morbidity worldwide (1), is characterized by airflow obstruction, as determined by the ratio of the forced expiratory volume in one second (FEV1) and forced vital capacity (FVC). Disease progression has been assessed using the rate of FEV1 decline, which is greater than normal in COPD (2, 3). To date, smoking cessation is the only intervention that has conclusively been shown to alter the rate of decline in FEV1 (4). While the pathogenesis of COPD is complex, studies suggest that airway inflammation plays an important role in disease progression (3, 5, 6). The intensity of inflammation relates to the degree of airflow obstruction (5), and may result from oxidant-induced damage. However, neither the antioxidant drug N-acetyl cysteine nor the nonspecific antiinflammatory effects of inhaled corticosteroids have been shown to modify the rate of decline in FEV1 (7–11). Meta-analyses of the inhaled corticosteroid (ICS) studies have yielded conflicting results (12–14). Salmeterol and other long-acting β-agonists are highly selective bronchodilators that have been shown to improve lung function, dyspnea, and health status in relatively short-term studies (15, 16). However, their possible long-term effect on rate of decline in FEV1 has never been evaluated. It has recently been shown that the administration of an ICS combined with a long-acting β-agonist modifies the expression of inflammation in mucosal biopsies and sputum of patients with COPD (6), raising the possibility that this pharmacologic combination could have an effect on the rate of decline of lung function. The Toward a Revolution in COPD Health (TORCH) study investigated the effect of salmeterol/fluticasone propionate (SFC) and either component alone compared with placebo on mortality, as well as the impact on the rate of exacerbations, health-related quality of life, and postbronchodilator FEV1. The primary efficacy analysis has already been published (17). The original report concentrated on the effect of therapy on mortality as the primary outcome, and presented the mean effect on lung function over 3 years as a supportive analysis, without addressing the change in the rate of FEV1 decline, a variable that has been accepted as a reasonable surrogate marker for disease progression. Before treatment unblinding, we decided to test the hypothesis that pharmacotherapy would modify the rate of decline of postbronchodilator FEV1, compared with placebo. We also conducted an exploratory analysis of the factors that could affect FEV1 rate of decline, since an association has been reported between frequency of exacerbation and an increased rate of decline of FEV1 (18, 19). Some of these results have been previously reported in the form of an abstract (20).
Design Overview Details of the TORCH study design have been published elsewhere (17, 21). TORCH was a multi-center, randomized, double-blind, parallel-group, placebo-controlled study. All corticosteroids and inhaled long-acting bronchodilators were stopped before the run-in period, but other COPD medications were allowed. After a 2-week run-in period, eligible patients were stratified by smoking status and randomized to receive either SFC 50/500 µg, salmeterol (SAL) 50 µg, fluticasone propionate (FP) 500 µg, or placebo twice daily for 3 years via a Diskus/Accuhaler inhaler (GlaxoSmithKline, Greenford, UK) (see online supplement). The primary efficacy endpoint of TORCH was all-cause mortality at 3 years. Other efficacy endpoints included rate of exacerbations (see online supplement), health status, and post-bronchodilator spirometry every 24 weeks.
Setting and Participants
Randomization and Interventions
Outcomes and Follow-up Spirometers were regularly calibrated according to manufacturer recommendations and a calibration log was kept. Lung function data were reviewed centrally during the study and queried if values differed significantly in consecutive visits (criteria used for the query are published in the online supplement). After completion of the study, the variability of the spirometric values was assessed by analyzing the variance of individual regression slopes and comparing them with those obtained in the ISOLDE trial (11), in which spirometric measurements were the primary endpoint and were closely monitored.
Statistical Analysis The effect of treatment on rate of decline of absolute FEV1, percentage change in a year and as percentage of predicted FEV1 was analyzed using a random coefficients model, including terms for treatment, time on treatment in years, treatment by time interaction, and covariates of smoking status, sex, age, baseline FEV1, region (see Table 2 footnote for countries included in each region), and body mass index (BMI). This methodology was the same as that used in other landmark studies, which assessed rate of decline in lung function in COPD (9–11). To derive the percentage change in a year, the logarithm of FEV1 was analyzed. To eliminate immediate improvements, the decline was evaluated from 24 weeks onward (the time at which the first on-treatment measurement was made). The effects of covariates on the rate of FEV1 decline were investigated using this model as exploratory analyses, including the covariate by time interaction individually for smoking status, sex, age, baseline percentage predicted FEV1, region, ethnic origin, BMI, previous exacerbation history, and baseline St George's Respiratory Questionnaire (SGRQ). We also tested whether the treatment effect on the rate of decline was consistent for subgroups by including a treatment-by-covariate-by-time interaction individually in this model. A further analysis was performed by calculating individual patient slopes from the regression analysis of each subject's FEV1 values and applying ANCOVA to these slopes. At least two on-treatment FEV1 measurements were required for this analysis. In addition, we report exploratory summary statistics of individual patient slopes categorized by number of exacerbations reported during the study, and by whether subjects survived or died during the 3 years of the study. All analyses were performed on an intention-to-treat basis using SAS software version 8.2 (SAS Institute, Inc., Cary, NC) on a Unix platform. For the principal analyses, a threshold for statistical significance was set at 0.05. For the effect of covariates on the slopes, which were exploratory analyses, the threshold was set at 0.10.
Role of the Funding Source
Patients A total of 6,112 patients composed the efficacy population of TORCH. Of these, 5,343 (87%) had at least one on-treatment FEV1 and were included in the decline analysis (Figure 1). The characteristics of these patients at baseline are shown in Table 1. The number of patients was smaller in the placebo compared with the active treatment arms because more patients withdrew within the first 24 weeks from the placebo arm (17% in placebo compared with 12% in the SAL and FP arms, and 9% in the combination arm). During the study, 187 (3%) patients took tiotropium while on study medication (44 [3%] placebo, 62 [4%] SAL, 40 [3%] FP, and 41 [3%] SFC).
Rate of Decline of FEV1 A total of 26,539 on-treatment observations were available for the analysis. The maximum number of on-treatment measurements a patient could contribute to the estimation of the rate of FEV1 decline was 6, and 64% of patients contributed this number. The average number of measurements was 5, with only 19% of patients having 3 or fewer, primarily due to early withdrawal or death. In the placebo arm, patients withdrawing before the end of the study had a faster rate of decline (76 ml/year) compared with those completing the trial (54 ml/year). The rate of absolute decline of FEV1 for each arm is summarized in Table 2, and that of % predicted FEV1 is shown in Table 3. Figure 2 shows the adjusted means and standard errors at each visit and fitted lines from the random coefficients model of FEV1. The rate of decline of FEV1 was slowest in patients on SFC and fastest in those randomized to the placebo arm. From Week 24 onwards, the adjusted rate of decline in FEV1 was 39 ml/year for SFC, 42 ml/year for both SAL and FP and 55 ml/year for placebo, a reduction of 16 ml/year with SFC compared with placebo (P < 0.001), and 13 ml versus placebo for both FP and SAL (P = 0.003) (Figure 2). These treatment differences remained when the values were expressed as % predicted FEV1 (Table 3) or as percentage of the baseline value (where the rate of decline was 3%/yr for SFC, 4%/yr for SAL and FP, and 5%/yr for placebo). In addition, the analysis of individual regression slopes produced similar findings. The standard deviations of individual regression slopes were similar in all treatment groups ranging from 160 to 180 ml/year. These values are similar to those observed in the ISOLDE trial (166 ml/yr for FP and 187 ml/yr for placebo).
Effect of Covariates on FEV1 Slopes of Rate of Decline A slower rate of decline in absolute ml/year was observed in former smokers, females, patients 65 years and older, and those with FEV1 less than 30% predicted. Patients with a BMI greater than or equal to 25 showed a slower decline in lung function (Table 4).The rate of decline in patients from the Asia Pacific and Eastern Europe regions was slower than that of patients from the other regional groups. These relationships were preserved when the rate of FEV1 decline was expressed as a percentage change in a year for all of these covariates except sex, where there was no difference, and in patients with FEV1 less than 30% predicted (see Table E1 in the online supplement). In this group of patients, the FEV1 declined by 28 ml/year, compared with 47ml/year for the other patients (Table 4), but when this change was expressed as a percentage of baseline (4%/yr) it was within the range of the other groups (4.4%/yr and 3.3%/yr for 30–49% and  50% predicted FEV1, respectively; Table E1).
The effect of treatment on FEV1 decline was similar irrespective of smoking status, sex, age, baseline FEV1, region of the world, ethnicity, BMI, previous exacerbations, and baseline SGRQ. The differences between placebo and the treatment arms were unaffected by whether the patients had taken ICS or LABA in the 12 months before the study (Tables E2 and E3). We observed no association between previous exacerbation history based on patient recall and FEV1 decline (Table 4). However, there appeared to be an association between number of exacerbations documented during the duration of the study and the rate of decline of FEV1 (see Table 5), with higher rates of decline being evident in patients experiencing more exacerbations.
COPD is characterized by airflow obstruction, which is usually progressive (2, 3), and hence, the measured decline in FEV1 has been accepted as a key marker for disease progression and a target for therapeutic trials. The longitudinal analysis of lung function from the TORCH data set presented here is the first to identify significant reductions in FEV1 decline in those patients receiving active treatment. The normal rate of FEV1 decline in healthy subjects is approximately 30 ml/year (24, 25). The modeled rate of decline in post-bronchodilator FEV1 in patients receiving placebo in TORCH was 55 ml/year; similar to that seen in the Lung Health Study 1 (–52 ml) (26), Lung Health Study 2 (–47 ml) (10), BRONCUS (–54 ml) (7), and ISOLDE studies (–59 ml) (11), and slightly lower than in EUROSCOP (–69 ml) (9), where the baseline FEV1 was higher and all randomized subjects were current smokers. We identified a significantly lower rate of decline in FEV1 (by 13–16 ml/yr) in those patients receiving active therapy. Rate of decline was similar among the three active treatment arms of the study. Although treatment did not abolish the accelerated decline in lung function, it did ameliorate it substantially, decreasing the excess FEV1 decline attributable to historically obtained values in patients with COPD (27). All three treatments showed improvements in post-bronchodilator FEV1 relative to placebo at each visit, but the mechanism responsible for the effect on rate of decline is not clear, as all treatments have potentially significant nonbronchodilator effects (6, 28, 29). Whether the maintenance of airway patency and reduction in hyperinflation, improvements in mucociliary clearance, or decreases in airway inflammation contribute singly or together to produce the observed functional change cannot be determined in TORCH, and further mechanistic studies are needed. The results of the forthcoming UPLIFT trial, where a long-acting bronchodilator drug tiotropium is compared with placebo with lung function decline as its primary outcome, may help clarify mechanisms, since tiotropium is a bronchodilator without primary antiinflammatory action (30, 31). In the TORCH trial, there were significant reductions in exacerbations in all treatment arms, with the greatest reductions observed with the SFC combination (17). This is consistent with our data, in which treatment decreased the rate of decline in FEV1 and this effect was greatest in patients receiving SFC. There was an association between exacerbation frequency documented during the study and FEV1 decline, supporting previous observations (18, 19) (Table 5). However, in patients who had no exacerbations during the study, the rate of decline was significantly faster in the placebo group compared with active treatments (56 ml/yr versus 27–31 ml/yr), which suggests that the effect of treatment on exacerbations was not the sole mechanism responsible for the reduced rate of decline with active treatment. Our results confirm those of previous studies, which have shown that smoking status, age, and baseline percent predicted FEV1 affect the rate of lung function decline (32). However, our data extend these observations in the Lung Health Study population to patients with more severe COPD. In addition, we have identified two novel factors associated with FEV1 decline, specifically BMI and region of origin, although these could also be due to differences in height. Together with already known variables such as baseline lung function, smoking status, and exacerbation frequency, they may help explain between-subject differences in FEV1 decline. Lung function declined least (35 ml/yr) in patients with a BMI of 29 or higher, was higher in patients with BMI between 25 and 29 (42 ml/yr), but was greatest in patients with a baseline BMI below 25 (51 ml/yr). This suggests an important association between systemic consequences of the disease and disease progression in the lungs (33, 34), but does not necessarily indicate causality. Interestingly, patients from the Asia Pacific and Eastern Europe regions, as well as patients of Asian and American Hispanic ethnic origins, had a slower rate of decline compared with Western Europeans and North Americans, even when expressed as percentage change in FEV1. This may be related to the fact that patients in Asia Pacific and Eastern Europe had lower mean FEV1 absolute and percent predicted at baseline (Table E4), thus providing less capacity for FEV1 to decline over time. Alternatively, other factors yet unexplored such as genetic, socioeconomic, or environmental differences may be important. Female patients lost FEV1 at a slower rate than that of male patients, a result similar to that reported in the long-term follow-up of the Lung Health Study (4). Women who quit smoking in that study lost an average of 22 ml/year, compared with men who lost an average of 30 ml/year. In the smokers, the values were 54 and 66 ml/year, respectively. In this TORCH dataset, women lost 39 ml/year, whereas men lost 47 ml/year irrespective of smoking status. This difference disappeared when the rate of decline was expressed as a percentage change in a year, with women losing 4.2% versus 3.9% per year for men. These results suggest that the sex difference was related to airway size rather than intrinsic biologic differences in the progression of COPD. There were some limitations to our study. As in other long-term COPD trials (35, 36), many patients failed to complete the study, with significantly more withdrawing from the placebo limb. Moreover, those withdrawing showed more rapid deterioration in lung function, a finding noted by others (37). This preferential dropout in the placebo arm of those patients whose function worsens more rapidly (evidenced by the greater decline in patients who withdrew early compared with those who completed) actually minimizes the differences observed in rate of FEV1 decline. In addition, the random coefficients model (11) gives most weight to patients who complete the trial, and hence, the differences in lung function decline we report may be conservative estimates of the true treatment effect. We are confident that our principal findings are reliable, since they were consistent whether expressed in ml/year or as percentage change per year. It has been suggested that changing the usual therapy of the patient with COPD can influence the results of interventional studies (38). This was not the case in TORCH, where prior therapy with ICS or LABA was unrelated to the beneficial effect of therapy on the rate of FEV1 decline. Another limitation of the study was that the FEV1 was not a primary outcome in this mortality trial. However, postbronchodilator lung function was extensively measured, and the more than 26,000 spirometric assessments obtained over the 3 years of the study provided a unique opportunity to evaluate how lung function evolved in patients randomized to different treatments. A theoretical limitation was the less rigorous monitoring of spirometry compared with other trials primarily evaluating lung function decline. However, the standard deviation of our FEV1 measurements was comparable with that in previous studies, in which spirometry was performed more frequently and using more rigorous quality control (10, 11). These data suggest that measuring postbronchodilator spirometry in a larger number of patients, as in TORCH, compensated for any inherent between-tests variability in FEV1. In summary, we have shown for the first time that pharmacologic therapy slows the decline in lung function in patients with COPD. Given the progressive nature of COPD, halving of the excess decline in FEV1 is likely to be clinically important in patients such as those who participated in TORCH.
The authors acknowledge technical support from K. Runcie, a professional medical writer with Gardiner-Caldwell Communications who was compensated by GlaxoSmithKline, and M. Sayers (GlaxoSmithKline) in the preparation of this manuscript. TORCH Steering Committee members: P.M.A. Calverley (Chairman), Liverpool, UK; J.A. Anderson, Greenford, UK; B. Celli, Boston, MA; G.T. Ferguson, Livonia, MI; C.R. Jenkins, Sydney, Australia; P.W. Jones, London, UK; K. Knobil, Research Triangle Park, USA; J. Vestbo, Manchester, UK; J.C. Yates, Research Triangle Park, NC.
The Toward a Revolution in COPD Health (TORCH) study was funded by GlaxoSmithKline. This article has an online supplement, which is accessible from this issue's table of contents at www.atsjournals.org. Originally Published in Press as DOI: 10.1164/rccm.200712-1869OC on May 29, 2008 Conflict of Interest Statement: B.R.C. has been reimbursed by GlaxoSmithKline (GSK), Boehringer Ingelheim (BI), AstraZeneca (AZ), and Almirall for participating in advisory boards and for lectures. N.E.T. is an employee of GSK and holds stock options with GSK. J.A.A. is an employee of GSK and holds stock options with GSK. G.T.F. in the last 3 years has participated as a speaker in scientific meetings or courses organized and financed by GSK, BI, and Pfizer; served on advisory boards for GSK, BI, Novartis, and Schering Plough; served as a consultant to GSK and Novartis; received an investigator initiated research grant from GSK; and has participated in multi-center clinical trials for GSK, BI, Novartis, Emphasys, Dey, Forest, and Altana. C.R.J. has been reimbursed by GSK, AZ, Novartis, and BI for participating on advisory boards and speaking at educational symposia. The studies she has managed have received research funding paid to the Woolcock Institute of Medical Research from GSK and AZ. P.W.J. has received consultancy fees averaging $10,000 over 3 years (2004–2007), $14,000 in lecture fees in 2006, and a grant of $190,000 in 2004, all from GSK. J.V. has been reimbursed by GSK for presenting of various meetings, receiving fees of $14,000 in 2004, $12,000 in 2005, and $12,000 in 2006. J.V. received $450,000 in 2006 and expects to receive $400,000 per year as research grants for participating in multi- and single-center clinical studies sponsored by GSK. J.V.'s wife is an employee of AZ Denmark; neither J.V. nor his wife owns shares in any pharmaceutical company. K.K. is an employee of the sponsor of the study (GSK) and owns shares of GSK. J.C.Y. is an employee of GSK and owns shares in GSK. P.M.A.C. has acted as a study leader for investigations sponsored by GSK, Altana, and Chiesi; has served on advisory boards supported by GSK, Pfizer, and AZ; and his department has received research funding from GSK to conduct non–product-related research. Received in original form December 21, 2007; accepted in final form May 29, 2008
Related articles in AJRCCM:
This article has been cited by other articles:
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
TOP
ABSTRACT
AT A GLANCE COMMENTARY
0.003 compared with placebo. Note: Vertical bars represent standard errors of the adjusted means at each visit. The number of patients at each clinic visit with FEV1 measurements is shown below the graph.
