6G-01 - The ECHO Study: Evaluation of Cloning-Censor-Weighting to assess the effect of time to treatment initiation of single inhaler dual ICS/LABA therapy after a first observed COPD diagnosis in the United States

Background: Proactive and timely delivery of healthcare is often an important aspect of healthcare quality and of the optimization of the real-world effectiveness of treatments. Evaluations of time-dynamic treatment strategies are increasingly popular, but using retrospective data to classify patients by observed treatment patterns can introduce immortal time bias or issues of reverse causality. Novel methods that avoid these biases exist but need to be assessed in practical applications.

Objectives: To explore the use of novel statistical methods to evaluate the impact of delayed treatment initiation following a first chronic obstructive pulmonary disease exacerbation (ECOPD) on time to first subsequent exacerbation, and exacerbation rate, within the first year of follow-up.

Methods: Data Source(s): administrative claims data from the USA sourced from Optum Clinformatics DataMart.

Population: 31276 adults aged ≥40 years with a first observed COPD diagnosis between January 1, 2018, and December 31, 2018, and an ECOPD occurring ≥30-days following the incident COPD diagnosis. The index date was set to the date of the ECOPD.

The study seeks to assess the degree to which classifying patients according to their observed treatment patterns (prompt [within 90 days of the index ECOPD] or delayed [within 90 to 180 days] treatment start) is subject to bias, by contrasting it to Marginal Structural Models (MSM) and Cloning-Censor-Weighting (CCW) approaches. Demographics, pre-index HCRU, respiratory medications, and concomitant diseases were considered to be confounders. The time to the first post-index ECOPD was analyzed using a weighted Cox regression and the annualized exacerbation rate was analyzed using a weighted negative binomial model where the weights were used and set up for the MSM and CCW separately.

Results: The number of patients with prompt and delayed treatment initiation was 23409, and 4214, respectively. The hazard ratios comparing delayed to prompt treatment initiation (with 95% CI) for time to first event were 1.16 [1.11, 1.20], 1.13 [1.09, 1.18] and 1.10 [1.06, 1.14] for the simple approach, MSM, and CCW, respectively. The corresponding ratio of event rates (with 95% CI) were 1.11 [1.08, 1.15], 1.10 [1.07, 1.13] and 1.08 [1.05, 1.10] for the simple approach, MSM and CCW respectively.

Conclusions: Analyses of both time to first-post index ECOPD and annualized exacerbation rates show similar results across all methods. However, findings may not be generalisable to other settings, populations, or outcomes, and researchers should consider novel and appropriate methods when developing studies to prevent self-inflicted biases, as estimated treatment effects may be impacted by immortal time bias or reverse causality.