Effect of Fluticasone/Salmeterol on Chronic Bronchitis Symptoms and Associated Patient Reported Outcomes: a post-hoc analysis of the TORCH trial

Subjects: We will determine if the subjects at study entry had CB based on a novel definition that is comparable to the classic definition from the cough and phlegm questions in the SGRQ (15). The subjects will be classified as CB+ if they answer “almost every day” or “most days a week” to the 2 following questions: “Over the last 4 weeks I have coughed” and “Over the last 4 weeks I have brought up phlegm (sputum).” The groups (SGRQ CB+ vs. SGRQ CB-) in each arm will be compared using descriptive statistics; the groups will be tested for normality and if normally distributed will be compared with t test, and if non-normal will be compared with Mann Whitney U test or the data will be log transformed to make the distribution normal. Categorical variables will be compared with a chi squared test. Primary Endpoint: The primary endpoints are exacerbation frequency and time-to-first exacerbation. Our primary goal is to determine if SGRQ CB+ subjects receiving fluticasone with or without salmeterol have a greater reduction in exacerbation frequency in comparison to those receiving salmeterol alone or placebo. We also wish to determine if the reduction in exacerbations is greater in SGRQ CB+ subjects compared to SGRQ CB- subjects in each arm. Therefore, we will compare the exacerbation frequency in SGRQ CB+ subjects in the four arms of the study (fluticasone/salmeterol, fluticasone, salmeterol, or placebo). We will also compare the exacerbation frequency in SGRQ CB- subjects in the four arms. Finally, we will compare the exacerbation frequency in SGRQ CB+ subjects and SGRQ CB- subjects in each arm (i.e. SGRQ CB+ exacerbation frequency vs. SGRQ CB- exacerbation frequency in the fluticasone/salmeterol arm, in the fluticasone arm, in the salmeterol arm, and in the placebo arm). The frequency of exacerbations will be analyzed with a generalized linear model (assuming a negative binomial distribution, which accounts for variability among patients in the number and frequency of exacerbations), with the number of exacerbations as the outcome of interest and the logarithm of time during which treatment was received as an offset variable. Time to first exacerbation will also be a primary endpoint. Kaplan-Meier analysis will be performed using log-rank test to compare exacerbation-free survival curves between groups. A cox proportional hazards model will be used to develop hazard ratios for time to first exacerbation, using several known risk factors for exacerbations as covariates. Rate ratios will be calculated comparing the various groups (i.e. the four arms in SGRQ CB+ patients, the four arms in SGRQ CB- patients, and SGRQ CB+ vs. SGRQ CB- patients in each arm).Secondary Endpoints: The secondary endpoints are SGRQ scores, lung function and mortality. Total scores, as well as subscores (symptoms, activity, and impact) of the SGRQ and post-bronchodilator FEV1 will be analyzed as changes from baseline values with the use of repeated-measures analysis of covariance (ANCOVA). Estimated differences between treatments at each visit will be averaged with equal weights to determine the overall treatment effect during the 3-year study period. All efficacy analyses will be performed according to the intention-to-treat principle. Comparisons other than those between the combination regimen and placebo and between the combination regimen and salmeterol alone will be exploratory. The difference in times to death from any cause in each arm between the CB+ group and CB- group will be analyzed with the log-rank test (with stratification according to smoking status) and Cox regression. We will use CB status and smoking status as time varying covariates in our Cox regression models. We will also test for interactions between CB status and treatment arm in our Cox regression analyses.We understand now that CB may be a time varying covariate; recent evidence has shown that CB can wax and wane, particularly in relationship to smoking status (16, 17). To account for this, we will perform an analysis similar to the Evaluation of COPD Longitudinally to Identify Predictive Surrogate Endpoints (ECLIPSE) trial (18) (see publication for more detail). We will assess the stability of the SGRQ CB+ phenotype over time in each arm. We will determine how many subjects in each arm that were SGRQ CB+ at study initiation who were still SGRQ CB+ at years 1, 2, and 3 using a repeated measures analysis. We hypothesize that in those in the fluticasone arms will be more likely to have converted to SGRQ CB- in follow-up years compared to the salmeterol alone or placebo arms.Power analysis: According to our recent publication, 45.6% of the 2148 subjects in COPDGene with an FEV1=60% predicted (similar to the population in the TORCH trial) were SGRQ CB+ (15). Using the SGRQ CB definition at study initiation, we found that the SGRQ CB+ group had a greater exacerbation frequency compared to the SGRQ CB- group in longitudinal follow-up (1.11±1.62 vs. 0.77±1.17 exacerbations/patient/year, p<0.0001; mean±SD 4.44±1.75 years followed).SGRQ CB+ was statistically significant in multivariable linear regression using other covariates associated with exacerbations (ß coeff 0.266, SE 0.066, p<0.0001). The calculated effect size is 0.24 from these data. There were approximately 1,500 subjects in each arm of the TORCH trial. Using G-power to estimate the sample size, we will need 734 subjects in each arm to have 90% power with a of 0.05 to detect an effect size of 0.24. Therefore, there should be enough subjects in each arm to adequately detect at least a similar difference between SGRQ CB+ and SGRQ CB- groups.