Evaluation of the effect of inhaled corticosteroids (ICS) on blood eosinophil count (EOS) in patients with COPD

1. Impact of inhaled corticosteroids (ICS) on blood eosinophil count (EOS). To test whether the administration of ICS suppresses the levels of blood EOS count, we will compare EOS counts measured while participants were receiving ICS versus time-points while participants were not receiving ICS. More specifically, we will perform the following comparisons: (i) EOS count before versus after the run in period of ISOLDE trial, among participants who were previously prescribed on ICS, (ii) EOS count after the run in period versus after the first year of study treatment, among participants who were randomised to receive ICS in ISOLDE trial, (iii) EOS count after the run in period versus after the first year of study treatment, among participants who were randomised to receive ICS in ISOLDE trial and were steroid naive previously, (iv) EOS count at baseline versus EOS count at six and at twelve weeks, among patients who were randomised to receive ICS in SCO104925 trial. In addition, we will perform the following comparisons as control (we would not expect to find significant differences, as EOS administration status does not change between the two timepoints): (v) EOS count before versus after the run in period of ISOLDE trial, among participants who were not previously prescribed on ICS, (vi) EOS count after the run in period versus after the first year of study treatment, among participants who were randomised not to receive ICS in ISOLDE trial. For all these comparisons, we will use paired t-test if normal distribution of the data is confirmed or Wilcoxon signed-rank test if not.Furthermore, in both ISOLDE and SCO104925 trials, we will evaluate the impact of the administration versus no administration of ICS on EOS count, using mixed methods repeated measures methodology (MMRM). We will correct for potential confounding factors such as age, sex, COPD severity and smoking history.2. Could EOS measured while patients (i) are or (ii) are not receiving ICS predict responsiveness to ICS, with regards to the frequency of exacerbations, spirometric outcomes and quality of life?a. EOS cut points: We will repeatedly perform the following analyses using the following EOS cut points: 200 cells per ml, 400 cells per cubic mm, 600 cells per cubic mm, 2% or 4% of total white cells. Finally, we will repeat the analyses using the EOS count as a continuous variable.b. Study population: Each of the following models will be performed in the following study populations. For ISOLDE trial: (i) Firstly we will use EOS levels after the run in period (off steroids); in this model we will include all study participants. (ii) Then, we will use EOS levels before the run-in period and we will only include those participants who were receiving ICS prior to recruitment. (iii) Finally, we will use the difference in EOS levels before and after the run-in period, to assess whether patients whose EOS levels are suppressed by the administration of ICS respond better to ICS treatment. We will only include those participants who were receiving ICS prior to recruitment. We will use the following cutpoints: 200, 400 cells per cubic mm, 2% or 4% of total white cells.For SCO104925 trial, all participants will be included in all analyses. We will use EOS levels (i) at baseline, (ii) at week 6, (iii) at week 12 and the difference of EOS levels between baseline and week 12.c. Models to be used: - Time to first exacerbation: We will use a Cox proportionals hazard model. Covariates will include age, sex, FEV1% predicted at baseline, treatment group, eosinophil subgroup and treatment by eosinophil interaction. In the ISOLDE trial, the three-way interaction between treatment, eosinophil level and ICS status prior to recruitment will be considered.- Annual rate of moderate and severe exacerbations: We will use a generalised linear model assuming a negative binomial distribution. The response variable will be the number of on-treatment moderate and severe exacerbations per patients per year, with covariates including: age, sex, FEV1% predicted at baseline, treatment group, eosinophil subgroup and treatment by eosinophil interaction. In the ISOLDE trial, the three-way interaction between treatment, eosinophil level and ICS status prior to recruitment will be considered.- FEV1 decline rate: We will use mixed models repeated measures methodology (MMRM). Point estimates and 95% confidence intervals for the treatment differences in FEV1 will be generated for each visit according to each eosinophil subgroup. In the ISOLDE trial, the four-way interaction between treatment, eosinophil level, ICS status prior to recruitment and time will be considered. Various subgroup analyses will be considered to aid in the interpretation of the four-way interaction and any three-way interactions.- SGRQ total score analysis: We will use MMRM methodology and the rate of decline in health status will be assessed from 6 months, using a mixed effects model.d. Selection of the best model: In each of the previous analyses, we will compare model fit (using measures such as Akaike Information Criterion) to evaluate the relative quality of the different statistical models (based on EOS measured before or after the run-in period and using different EOS cut-points) and to identify the most clinically relevant ESO cut point and time point (measured while patients were receiving/not receiving ICS).