Clinical outcome of corticosteroids in the treatment of COPD exacerbations in China - CONTAIN

This is a Post-hoc analysis based on the database of “A non-interventional, retrospective study on AECOPD treatment status in China”. A large nation-wide retrospective non-interventional study has carried out from January to September 2014 in China. The study aimed to observe clinical practice including glucocorticoids treatment in AECOPD in China. The data for 5067 cases were collected. These cases met the following inclusion criteria: more than 40 years old, diagnosed by GOLD 2013 (GOLD: The Global Initiative for Chronic Obstructive Lung Disease) as COPD at least 3 months before AECOPD based on treating physician’s judgment; the patients received hospitalization due to AECOPD since Sep 2013. Demographic information, administration of corticosteroids, medical and surgical history, comorbidities, clinical outcome, laboratory tests and lung function tests were recorded in database. There were 43 sites in the main study. These sites located in 22 provinces in China. A majority of sites (40 sites) were tier 3 hospitals in major cities. The data in study was provided by each site via medical records. In 5091 screening cases, 5067 cases were recruited and included in full analysis set (FAS). All cases in FAS will be included into this post-hoc analysis. Through data mining and analysis, it is to explore the relationship between corticosteroids based treatment regimen and clinical outcome and the optimal treatment regimen for corticosteroids used in inpatients with COPD exacerbations based on our database.

In general, the statistical analyses based on raw data and the data using propensity score matching (PSM) will be primarily descriptive in nature. The statistics are as in the following: • Continuous variables: N, mean, standard deviation, lower quartile Q1, median, upper quartile Q3, minimum, maximum; • Categorical variables: frequencies and percentages of patients at each category. Before the statistical analyses, the medical history and co-morbidities need to be coded using ICD10. At the same time, the data collected should be clearly understood and provide the following information but not limited to the number of subjects enrolled in each site, the list of variables, the number of observations of each variable and the missing proportion of the variable. The factors involved into the logistic model for propensity score matching include but not limited to demographic, baseline characteristics, severity of COPD, etc. A factor research will be performed before the matching is performed aiming to include appropriate factors into the model. Matching ratio could be exact or approximate. Programming software is SAS 9.3 (SAS: Statistical Analysis System) or higher in Windows system. R statistical software could be another tool for the exploration.