New England Journal of Medicine February 22, 2007

Salmeterol and Fluticasone Propionate and Survival in Chronic Obstructive Pulmonary Disease

Peter M.A. Calverley, Julie A. Anderson, Bartolome Celli, Gary T. Ferguson, Christine Jenkins, Paul W. Jones, Julie C. Yates, Jørgen Vestbo

Bottom Line

In patients with moderate-to-severe COPD, combination therapy with salmeterol and fluticasone propionate did not achieve a statistically significant reduction in all-cause mortality compared to placebo, though it significantly reduced exacerbation rates and increased the risk of reported pneumonia.

Key Findings

1. Over 3 years, all-cause mortality (the primary endpoint) was 12.6% in the combination-therapy group compared to 15.2% in the placebo group [1.1.1].
2. The hazard ratio for death in the combination therapy group versus placebo was 0.825 (95% CI, 0.681 to 1.002; P=0.052), which narrowly missed the pre-specified threshold for statistical significance.
3. Combination therapy significantly reduced the annual rate of COPD exacerbations to 0.85, compared to 1.13 in the placebo arm (a 25% risk reduction; P<0.001).
4. The probability of having pneumonia reported as an adverse event was significantly higher among patients receiving fluticasone propionate (19.6% in the combination group and 18.3% in the fluticasone-alone group) compared to the placebo group (12.3%, P<0.001).

Study Design

Design
Randomized Controlled Trial
Double-Blind
Sample
6,112
Patients
Duration
3 yr
Median
Setting
Multicenter, international
Population Patients aged 40 to 80 years with moderate-to-severe COPD (pre-bronchodilator FEV1 <60% of predicted) and a smoking history of at least 10 pack-years.
Intervention Combination of salmeterol 50 μg and fluticasone propionate 500 μg twice daily administered via a single Diskus inhaler.
Comparator Placebo, salmeterol 50 μg twice daily alone, or fluticasone propionate 500 μg twice daily alone.
Outcome Death from any cause (all-cause mortality) over a 3-year period (comparing combination therapy with placebo).

Study Limitations

There was a highly disproportionate study withdrawal rate, with 44% of patients in the placebo group withdrawing compared to 34% in the combination group [1.1.3]. This extensive dropout and subsequent potential crossover to active rescue therapies likely diluted the observable mortality benefit.
The mortality rate in the placebo arm was lower than statistically anticipated during power calculations (observed 15.2% vs. expected 17%), rendering the study potentially underpowered to detect the 2.6 percentage point absolute difference observed.
Pneumonia was recorded based on adverse event reporting rather than prospectively defined, objective radiographic criteria, leaving ambiguity around the exact nature and severity of the respiratory infections.

Clinical Significance

The TORCH trial is a landmark study that definitively established the efficacy of long-acting beta-agonists and inhaled corticosteroids (LABA/ICS) in reducing exacerbations, preserving lung function, and improving quality of life in moderate-to-severe COPD. However, it also unmasked a substantial, class-effect risk of pneumonia associated with inhaled fluticasone. Although the trial missed its primary survival endpoint by a razor-thin margin (P=0.052), its findings fundamentally changed COPD guidelines, prompting clinicians to weigh the clear benefits against the elevated pneumonia risk, ultimately steering modern guidelines to reserve ICS for frequent exacerbators or those with eosinophilic phenotypes.

Historical Context

Prior to 2007, the use of inhaled corticosteroids and LABAs in COPD was heavily extrapolated from asthma paradigms, but definitive prospective evidence of a survival benefit was elusive. The TORCH trial was the first mega-trial in COPD—unprecedented in its scale and 3-year duration—specifically designed to test if pharmacological intervention could alter the natural history and mortality of the disease. Its legacy heavily influenced subsequent major trials (like UPLIFT and FLAME) and cemented the necessity of phenotyping COPD patients rather than applying a 'one-size-fits-all' approach to ICS therapy.

Guided Discussion

High-yield insights from every perspective

Med Student
Medical Student

How does the mechanism of action of fluticasone and salmeterol theoretically work together synergistically to manage COPD, even if an overall mortality benefit was not definitively proven in the TORCH trial?

Key Response

Salmeterol, a Long-Acting Beta-2 Agonist (LABA), causes bronchodilation by increasing intracellular cAMP, which also enhances the nuclear translocation of glucocorticoid receptors. Fluticasone, an Inhaled Corticosteroid (ICS), reduces airway inflammation and simultaneously upregulates beta-2 receptor expression on airway smooth muscle, preventing LABA tachyphylaxis. This molecular synergy improves lung function and reduces exacerbation frequency more effectively than either drug alone.

Resident
Resident

Based on the TORCH trial findings, how do you clinically weigh the benefits of reduced exacerbations against the increased risk of pneumonia when deciding to prescribe an ICS/LABA combination for a patient with moderate-to-severe COPD?

Key Response

The trial demonstrated a significant reduction in COPD exacerbations and an improvement in health status, but it also revealed a higher incidence of reported pneumonia in the fluticasone-containing arms. Clinicians must actively phenotype their patients, reserving ICS/LABA for the 'frequent exacerbator' phenotype or those with asthma-COPD overlap, while avoiding ICS in patients with a history of recurrent pneumonias or those who are well-controlled on bronchodilators alone.

Fellow
Fellow

The TORCH trial's primary endpoint of all-cause mortality narrowly missed statistical significance with a p-value of 0.052. How should this near miss influence your interpretation of the data, and how does informative censoring complicate this finding?

Key Response

A p-value of 0.052 represents a 17% relative risk reduction in mortality, suggesting a strong clinical trend despite missing the arbitrary 0.05 threshold. Fellows must recognize that informative censoring likely occurred: a much higher percentage of patients withdrew from the placebo arm due to worsening COPD and were subsequently prescribed active rescue medications off-trial. This 'rescue' likely improved survival in the placebo cohort, biasing the intention-to-treat mortality difference toward the null.

Attending
Attending

Given the TORCH trial's finding of increased pneumonia risk without a definitive all-cause mortality benefit, how should this evidence drive long-term stewardship of inhaled corticosteroids in your practice?

Key Response

Historically, ICS was overprescribed across all stages of COPD. The TORCH data quantified the real-world harm (pneumonia) against a backdrop of symptomatic and exacerbation benefit rather than survival. This should prompt attendings to proactively evaluate patients for ICS de-escalation, shifting the foundation of maintenance therapy to LABA/LAMA combinations unless there is a clear, evidence-based indication for ICS, such as elevated blood eosinophils or frequent severe exacerbations.

Scholarly Review

Critical appraisal through the lens of expert reviewers and guideline development

PhD
PhD

The TORCH trial experienced differential dropout rates, with 44% withdrawing from the placebo arm versus 34% from the combination arm. From a methodological standpoint, how does this differential attrition threaten the validity of the intention-to-treat survival analysis, and what statistical models could adjust for this?

Key Response

Differential dropout creates a significant risk of attrition bias and informative censoring. If patients dropping out of the placebo arm are the sickest but receive standard-of-care rescue therapies outside the trial, their survival inflates the placebo group's outcomes. Researchers could employ Inverse Probability of Censoring Weighting (IPCW) or marginal structural models to re-weight the data and estimate the true treatment effect as if all patients had remained on their randomized assignments.

Journal Editor
Journal Editor

If reviewing the TORCH manuscript today, how would you critically evaluate the investigators' handling of vital status follow-up for the massive number of early withdrawals, and how does the fragility index apply here?

Key Response

A major threat to validity in mortality trials is loss to follow-up, especially with high withdrawal rates. The TORCH investigators aggressively sought vital status for all randomized patients, capturing 99% of mortality data. However, as an editor, I would flag that with a p-value of 0.052, the fragility index is essentially zero. Even one or two misclassified deaths or missing vital statuses could have swung the result across the threshold of significance, highlighting the fragility of drawing binary conclusions from this specific p-value.

Guideline Committee
Guideline Committee

How do the findings of the TORCH trial regarding exacerbation reduction versus pneumonia risk directly inform the current GOLD guideline recommendations for the placement of ICS-containing regimens?

Key Response

The TORCH trial's dual findings heavily influenced current GOLD guidelines, which now restrict ICS use primarily to Group E patients (those with frequent or severe exacerbations) who also have elevated blood eosinophils (typically >300 cells/microL). The guidelines explicitly reference the pneumonia risk highlighted by TORCH to recommend against ICS monotherapy in COPD and to position LABA/LAMA combinations as preferable initial therapy for most non-exacerbating patients, balancing efficacy with safety.

Clinical Landscape

Noteworthy Related Trials

2008

UPLIFT Trial

n = 5,993 · NEJM

Tested

Tiotropium 18 mcg daily

Population

Patients with moderate to very severe COPD

Comparator

Placebo

Endpoint

Rate of decline in FEV1

Key result: Tiotropium improved lung function and quality of life, and reduced exacerbations, but did not significantly reduce the rate of decline in FEV1.
2016

FLAME Trial

n = 3,362 · NEJM

Tested

Indacaterol/Glycopyrronium (LABA/LAMA) 110/50 mcg daily

Population

COPD patients with a history of exacerbations

Comparator

Salmeterol/Fluticasone (LABA/ICS) 50/500 mcg twice daily

Endpoint

Annual rate of all COPD exacerbations

Key result: LABA/LAMA was superior to LABA/ICS in preventing COPD exacerbations and was associated with a lower incidence of pneumonia.
2018

IMPACT Trial

n = 10,355 · NEJM

Tested

Fluticasone furoate/Umeclidinium/Vilanterol (Triple therapy)

Population

Symptomatic COPD patients with a history of exacerbations

Comparator

ICS/LABA or LAMA/LABA dual therapies

Endpoint

Annual rate of moderate or severe COPD exacerbations

Key result: Triple therapy resulted in a significantly lower rate of moderate or severe COPD exacerbations than dual therapy with either ICS/LABA or LAMA/LABA.

Tailored to your role

Want this tailored to you?

Add your specialty or training stage to get role-specific takeaways and more questions.

Personalize this analysis