Comparative effectiveness of Umeclidinium+Vilanterol versus Tiotropium Bromide+Olodaterol among patients with Chronic Obstructive Pulmonary Disease in a real-world primary care setting in the UK

Date of Approval
Application Number
Technical Summary

Aim: To compare real-world effectiveness of Umeclidinium/Vilanterol (UMEC/VI) versus Tiotropium Bromide/ Olodaterol (TIO/OLO) on rescue medication use and medication adherence among a General Practice (GP) cohort of patients with chronic obstructive pulmonary disease (COPD) in England. Greater understanding of the effectiveness of these maintenance therapy options will better-inform early treatment choices for patients with COPD in the UK.

Objectives: To compare i) mean number of rescue medication prescriptions and ii) medication adherence in patients newly-initiating UMEC/VI versus TIO/OLO. The exploratory objective is to describe the proportion of patients receiving triple therapy and time until triple therapy initiation among UMEC/VI and TIO/OLO patients.

Primary exposures: Single inhaler UMEC/VI or TIO/OLO initiation.

Primary outcome: Rescue medication use (inhaled or nebulized short-acting beta agonist [SABA]- or muscarinic antagonist [SAMA]- containing medication) 12 months following treatment initiation.

Methods: A new-user, active comparator, retrospective cohort study using inverse probability of treatment weighting (IPTW) to adjust for measured confounders will be employed to assess UMEC/VI efficacy using linked CPRD and Hospital Episode Statistics (HES) data. The first/earliest date of dual therapy (UMEC/VI or TIO/OLO) initiation between July 2015 and September 2019 will determine the index date. No minimum follow-up is required for study inclusion, a 24 months maximum follow-up will be used to assess effectiveness outcomes.

Data analysis: A propensity score (PS) based methodology will be implemented to minimise potential confounding and evaluate average treatment effects in the population. Logistic regression will generate the PS, which will be applied via IPTW. Rescue medication use will be compared between unweighted and weighted treatment cohorts, and 95% CIs and p-values generated from a generalised linear least squares regression model. Adherence will be classified using proportion of days covered (PDC): PDC<80% considered non-adherent, PDC≥80% adherent. Time-to-triple therapy will be assessed using Kaplan-Meier survival analysis and Cox proportional hazards models.

Health Outcomes to be Measured

Mean number of rescue medication prescriptions; Medication adherence (PDC ≥80%); Proportion of patients receiving triple therapy; Time-to-triple therapy

NOTE TO REVIEWER: Your comment raised concern that the primary outcome variable “rescue medication use” will be observed over a variable follow-up period, and that the study design does not account for this. We would ask you to consider that this outcome will be observed in distinct time periods after index (6, 12, 18 and 24 months) to allow us to observe any change in comparative rescue medication use over time since index. We also chose to observe mean number of rescue medication prescriptions because this was a more common approach in similar studies in the literature, and to align with a related GSK US study. For a given time period being assessed, only patients with follow-up extending at least as long as that time period will be included in that analysis. Therefore, while patients with varying lengths of follow-up may be included in the study cohort, only those with sufficient follow-up will be included in specific analyses. This is described in the third paragraph of the Study design section. We are unable to make amendments to the content of the form to make this clearer because both the Study Design and Data/Statistical Analysis sections are locked to editing, and the Outcomes to be measured section is specified to only be a list of variables with no further explanation. We hope this approach to responding to your feedback is sufficient.
Regarding your second point, we have added a statement to the Limitations section to explain that rescue medication may be prescribed for other reasons than worsened symptoms, but that these other reasons are not expected to be differential between groups, and that if prescriptions are occurring for other reasons, it would lead to an underestimation of treatment effect for this outcome.


Gema Requena - Chief Investigator - GlaxoSmithKline - UK
Victoria Banks - Corresponding Applicant - Adelphi Real World
Afisi Ismaila - Collaborator - GSK
Alexandrosz Czira - Collaborator - GSK
Catherine Castillo - Collaborator - Adelphi Real World
Eunmi Ha - Collaborator - Adelphi Real World
Jie Yeap - Collaborator - Adelphi Real World
Robert Wood - Collaborator - Adelphi Real World
Rosie Wild - Collaborator - Adelphi Real World
Theo Tritton - Collaborator - Adelphi Real World


HES Accident and Emergency;HES Admitted Patient Care;Patient Level Index of Multiple Deprivation