Treating COPD in the Real World

Abstract

Evidence-basedmedicine is central tomodernmedical practice and relies on the availability of data from appropriately conducted randomized clinical trials (RCTs). These studies establish whether treatment is effective and, when an active comparator group is included in the trial,whether the new therapy is better than currently used treatment. In some conditions such as chronic obstructive pulmonary disease (COPD) for which no single surrogate end point predicts response to treatment, multiple trials of varying duration are neededtoconvincephysiciansandregulators thatdrugtherapy is beneficial. Themainstays of COPDmanagement, including inhaled long-acting antimuscarinic agents or long-acting β-agonists (LABAs) alone or combined with inhaled corticosteroids (ICSs),havebeenshown inRCTs to improve lung function and quality of life and reduce exacerbation frequency.1-3 Patients included in these trials are selectedonthebasisofhaving stable COPD without serious diseases that would lead to prematuredeath, and these studies are then incorporated into treatment guidelines that direct clinical practice.4 AlthoughRCTsprovideevidence thatadrugworks inaspecificpatientpopulation,whathappenswhen treatment isused in the general patient population seen in daily practice remains uncertain. Evidence from RCTs may be biased due to premature study withdrawal, which limits the ability of investigators to answer important supplementary questions about the effects of treatment on the natural history of the disease.5 Large clinical trials can identify unanticipated new adverse effects such as pneumonia with ICS use,6 although whether the newevents have the same clinical importance as non–steroid-associated pneumonia is less clear.7,8 More importantly, theprofile ofpatients included inRCTsdiffers in age and comorbidity from that of many patients with COPD who are seenbyphysicians inmost settings.Hence, it is important to establishwhether outcomes seen in clinical treatment trials apply in the so-called real world of clinical practice. Large administrative health care databases, statistical methods needed to interrogate these complex data sets, and, in somecountries likeCanada, comprehensivehealthcareprovision that reduces confounding due to socioeconomic treatment choices all provide the opportunity to study the effect ofclinicians’prescribingchoices. In this issueofJAMA,Gershon and colleagues9 report the results of an investigation of the comparative efficacy of using a LABA alone or together with an ICS in elderly patients with COPD living in Ontario. The authors identified patient records for 38 266 individuals who met their database-validated diagnosis of COPD and had received a newprescription for a LABAor a LABA-ICS combination.Approximately 1 in 10patients receivedaLABAasamonotherapy, possibly reflecting the influence of guidelines and pharmaceuticalpromotionontreatmentchoices.Patientswere matched for confounding variables using a propensitymatching protocol, and the resulting 3160 LABA and 8712 LABA-ICS users were followed up for as long as 5 years, with theprincipal outcomebeing riskofdeathorhospitalization for COPD. The matched patients were older (mean age, 77 years) and had more comorbid diseases (median, 7) than is usual in RCT populations. Approximately 30% had been hospitalized in the previous 6 months and 28% had a diagnosis of asthma recordedat some timepreviously.Approximately 25%hadnot undergone spirometry to confirm theCOPDdiagnosis, but the demographic characteristics of this subgroup and its subsequent outcomes were similar to the group as a whole. Median life expectancy in these older patients was approximately 2.5 years, leading to an annual mortality rate almost3 timeshigher thanwasseen in theTORCHstudy, theonly RCT to studymortality with ICS-LABA therapy as its primary endpoint.1 In the studybyGershonet al,9 amongpatients prescribed ICS-LABAdrugs, the relative risk of death or hospitalization was approximately 8% less than among those treated withLABAsalone.This is similar to the6.8%reduction in relative riskofdeathreportedwith thesedrugs in theTORCHstudy, although that difference in the latter study was not statistically significant. This resultmay reflect the greater number of deaths reported in the study by Gershon et al compared with the TORCH study (4353 vs 398, respectively) and the use of a combined end point with COPD hospitalizations. As Gershon et al report, the benefit associated with ICS-LABA use was greatest among patients with a prior diagnosisofasthmaandamongpatientswithout thisdiagnosiswho did not receive a long-acting antimuscarinic agent. However, there was no difference between treatments in the incidence of pneumonia. This is in contrast bothwith previousRCTs2,6,7 and with a large database study from Quebec10 that reported adoublingof the incidenceofpneumonia inCOPDpatients receiving ICSs. However, the population in the study by Gershonet alwasdefineddifferently, had adifferent baseline risk profile forpneumonia, andwas followedup for longer than the 60 days reported in the Quebec study. Database studies always raise concerns about why a physician prescribed a particular drug for a specific patient, often leading to important differences between study groups. The size and scope of the report byGershon et al provide some reassurance that this is not themain factor in explaining thedifferences they report. Their diagnostic criterion identified approximately 80%of patients correctly as havingCOPD, and in a series ofmodeled scenarioswith increasing degrees of diagRelated article page 1114 Opinion