Survival Among Copd Patients Using Fluticasone/Salmeterol in Combination Versus Other Inhaled Steroids and Bronchodilators Alone

Abstract

Recent retrospective studies have suggested that use of inhaled corticosteroids (ICS) may improve survival in chronic obstructive pulmonary disease (COPD), particularly when combined with a long-acting beta-agonist (LABA). However, the study methodologies have been questioned, and no study has examined the survival effect of the newer combination ICS/LABA inhalers. The goal of this project was to further examine the relationship between ICS treatment, with or without LABA, and survival in COPD. COPD patients were identified from the administrative databases of four different integrated health care delivery systems. All patients who were diagnosed with COPD between September 1, 2000 and August 31, 2001 and who had at least 3 months treatment with either a combined fluticasone/salmeterol inhaler (FSI, N = 866), any ICS used with a LABA (ICS/LABA, N = 525), ICS alone (N = 742), LABA alone (N = 531), or a short-acting bronchodilator alone (SABD, N = 1832), were included. Analyses were conducted using three different analysis approaches that adjust for various biases that may affect the results. In the basic Cox proportional hazards models, use of FSI, ICS/LABA, ICS alone, and LABA alone had significant survival benefits as compared to SABD, after adjustment for differences in age, gender, comorbidities, asthma status, and disease severity (HRs 0.61 [0.45–0.83], 0.59 [0.46–0.77], 0.76 [0.61–0.95], 0.75 [0.57–0.98], respectively). Propensity score matching to reduce the clinical differences between the treatment groups versus the SABD reference group found very similar results. Nested case-control analyses, which are based on survival status instead of treatment, continued to show a significant survival benefit for FSI, ICS/LABA, and ICS alone. Treatment with FSI or another ICS with or without LABA is associated with improved survival in COPD. The treatment benefit is reproducible and is robust to application of a number of different analysis techniques designed to adjust for differences in confounding variables and for bias by indication.