Estimand Framework Enhances Analysis of Patient-Reported Outcomes in Lung Cancer Trials

The estimand framework offers a structured approach to define the treatment effect of interest in clinical trials, especially when analyzing patient-reported outcomes (PROs). A recent study published in BMC Medical Research Methodology demonstrates the application of this framework in a single-arm trial (SAT) of an anticancer treatment for patients with locally advanced or metastatic anaplastic lymphoma kinase (ALK)-positive non-small cell lung cancer (搜索)搜索该疾病.

The study re-analyzed data from a phase 2 trial where the co-primary outcomes were objective tumor response and adverse events, with overall quality of life (QoL) measured by the EORTC QLQ-C30 (搜索)搜索该靶点 global QoL scale as a secondary endpoint. The original trial, conducted before the widespread adoption of the estimand framework, lacked an explicit strategy for handling intercurrent events or missing PRO data.

Applying the Estimand Framework

The estimand framework, as defined in ICH E9-R1, comprises five attributes: treatment, population, variable of interest, population-level summary, and strategy for handling intercurrent events. In this case study, the treatment was the trial medication, and the population was defined by the original trial's inclusion/exclusion criteria. The researchers focused on the absolute numerical value of the PRO, specifically the mean QoL value at each cycle, as the population summary.

Strategies for Intercurrent Events

The study explored various strategies for handling intercurrent events such as death, treatment discontinuation (TD), and disease progression (PD). These strategies significantly impact the interpretation of PRO data.

Handling Death

• While Alive Strategy: Estimates mean QoL only for patients alive at each cycle, acknowledging that the group's characteristics change over time due to mortality and censoring. Kaplan-Meier estimates of survival were provided alongside QoL estimates.
• Composite Strategy: Sets all global QoL values after death to 0, combining QoL and survival into a single composite outcome. The mean of this composite outcome was then estimated.
• Hypothetical Strategy: Estimates mean QoL under the hypothetical scenario where all patients remain alive until at least cycle 40, using linear mixed models to extrapolate QoL trajectories.

Handling Treatment Discontinuation

• While on Treatment Strategy: Estimates mean QoL only for patients still on treatment at each cycle, removing data after treatment discontinuation.
• Hypothetical Strategies: Uses linear mixed models to predict QoL under hypothetical scenarios, such as assuming no treatment discontinuation or death.
• Treatment Policy Strategy: Includes data with imputed measurements after TD, estimating mean QoL regardless of TD.

Handling Disease Progression

Strategies for handling disease progression (PD) were defined analogously to those for TD, using a treatment policy strategy for TD.

Impact on Interpretation

The study highlights that different strategies for handling intercurrent events lead to varying results and interpretations of PRO data. The choice of estimand and the corresponding analytical approach should be carefully considered based on the clinical question of interest and the potential impact of intercurrent events on patient outcomes.

By applying the estimand framework, researchers can gain a more comprehensive and nuanced understanding of the treatment's effect on patient well-being, accounting for the complexities introduced by intercurrent events in clinical trials.