WEUSKOP6416: Evaluating Pneumonia in Chronic Obstructive Pulmonary Disease (COPD) Subjects

Pneumonia remains an important cause of morbidity and mortality in older adults with obstructive lung disease. Risk factors for pneumonia, including episodes associated with a hospital admission, have been extensively characterized in clinical trials and observational studies of patients with COPD, and include older age, lower predicted FEV1 (<50%), prior COPD exacerbations, dyspnea , normal to low body mass index (<25), current smoking and certain co-morbid conditions (e.g. dementia). The use of inhaled corticosteroids (ICS) has also been identified, as associated with an increased risk of pneumonia in patients with COPD.

The primary objective of this study is to estimate the magnitude of known risk factors and the outcomes of pneumonia requiring hospitalization and the potential effect modification of these risk factors by ICS use. The primary endpoints will be severe pneumonia, defined as community-acquired pneumonia (CAP) resulting in hospitalization and/or death and hospital-acquired pneumonia (HAP) diagnosed after two days in the hospital. As a secondary endpoint, CAP that did not result in hospitalization or death will be examined. As a secondary objective, we will describe characteristics for those patients who develop pneumonia requiring hospitalization compared to those with pneumonia not requiring admission.

This study will use the General Practice Online Database (GOLD), formerly referred to as the General Practice research Database (GPRD), a primary care electronic medical record database.

A new user cohort will be defined among patients with COPD who are 45 years and older in the United Kingdom. Patients will be considered a new user of ICS-containing medications if they had not received a prescription for an ICS-containing medication in the prior year. The comparator treatment group will be new users of long-acting bronchodilators (LABD), including long-acting beta-agonists (LABA) or long-acting antimuscarinics (LAMA). In the one year washout period, all new users could not have either ICS-containing medications or LABD.

Prior to conducting the analysis, feasibility analyses will be conducted to evaluate of the number of pneumonia events and the number of new users separately to examine the available precision based on the study design.

Patients will be followed from the date of their first eligible prescription (Cohort Entry Date) until the earliest of the following: date of study end point (first pneumonia event of interest), date of treatment end (up to 60-day gap allowed for each inhaler), date of transfer to a new practice, date of ICS initiation (among LABD new users), death or study end (end of available data). As part of the primary analysis, patients will be examined for their first severe pneumonia (severe CAP, HAP). As a secondary analysis, time to non-severe CAP will be examined. Incidence rates of the pneumonia outcomes will be calculated as the number of patients experiencing an event divided by the person-years at risk.

Multivariable analysis will be performed using Cox proportional hazard model with adjustment for confounders and medication exposure. To adjust for differences confounding by severity due to differences in prescribing between ICS-containing medications and LABD, propensity scores (PS) will be utilized using inverse probability of treatment weighting (IPTW). The propensity score will be estimated to model the probability of a patient receiving ICS-containing medication prescription versus receiving a LABD prescription given a patient's observed set of baseline covariates.

Effect modification (statistical interaction) will be evaluated based on available theory and include ICS medication use by known risk factors for pneumonia (BMI<21, BMI 21-24.9, BMI ≥25, age, GOLD stage III/IV, MRC dyspnea score ≥4, history of pneumonia diagnosis, current smoking status, social deprivation quartiles). Additional interactions may be evaluated.

To test proportionality of the hazard functions, model diagnostics will be performed.

To compare severe pneumonia with non-severe pneumonia in patients with COPD, characteristics of patients experiencing non-severe CAP vs. severe CAP or HAP will be tabulated. To assess differences between treatments, clinical and patient characteristics will be compared using the chi-square tests or Wilcoxon tests for categorical or continuous data, respectively. Severe CAP and HAP may be combined. Modeling of clinical and patient characteristics may be considered using logistic regression using CAP vs. severe CAP and then with severe CAP vs. HAP.

Additional analysis or adjustments to the analytic or modeling strategy will be performed if the data warrants. A more detailed modeling strategy, including generation of the propensity scores and Cox modeling, will be created in a separate analysis plan. Adjustments to the a priori plan will be described in the final study report.