Project Summary

Non-small cell lung cancer (NSCLC) is among the leading causes of cancer mortality in the United States and worldwide and is typically diagnosed at an advanced stage. Five-year survival among NSCLC patients diagnosed with advanced disease barely exceeds six percent. Individuals with advanced NSCLC (aNSCLC) have varying prognoses, with treatment occurring in stages. In recent years, there have been advances in treatment involving new therapies and combination of therapies. These advances have increased the treatment options. Given the high mortality, heterogeneous prognostic landscape, multi-stage nature of treatment and potential differential responses to treatment, it is essential to individualize patients’ treatment to yield the best outcomes. 

A dynamic treatment regime (DTR), also known as an adaptive treatment strategy, is a sequence of decision rules determining how to individualize treatments and their timing for patients based on their evolving health history. The overarching goal of the proposed research is to use electronic health record (EHR) data to develop DTRs for aNSCLC that optimize patient outcomes. 

The study team plans to utilize two large EHR databases: the Flatiron aNSCLC database and the CancerLinQ lung cancer database. These databases include extensive data on about 89,000 and about 54,000 unique patients with aNSCLC, respectively, including demographics, clinical measurements, treatments, laboratory values, cancer characteristics and outcomes. These databases include patients treated at academic, community and managed integrated care providers. These databases will be augmented with patient-reported outcome (PRO) data from two National-Cancer-Institute-designated Comprehensive Cancer Centers: Huntsman Cancer Institute and Moffitt Cancer Center. 

There are three key analytic challenges that need to be addressed to estimate optimal DTRs for aNSCLC patients using these databases. These challenges are driven by the messiness of EHR data and the limitations of existing DTR optimization methods. In addressing these challenges, the researchers will be guided by an advisory team comprised of lung oncologists, a patient advocate, a patient caregiver and a PRO specialist. The study team’s first challenge is to extract the components needed to define DTRs, including patient characteristics, aspects of their health history and cancer, possible treatments and outcomes. These data elements may not be available for some patients in their EHR. The second challenge is to develop a statistical method to fill in these missing data elements. This will require assumptions that the study team will elicit from their advisory team. Existing methods developed for determining optimal DTRs from observational data rely on strong modeling assumptions and may be unduly influenced by the weights given to specific individuals in the databases, leading to biased results. There has also been limited work on time-to-event outcomes, such as survival time. The third challenge is to develop a machine-learning-based method that is more robust than previous methods and can handle survival-type outcomes. 

The proposed project will produce evidence on optimal treatment strategies that are responsive to evolving clinical characteristics for patients with aNSCLC, as well as essential methodological contributions for the analysis of large EHR data, from problem refinement and feature extraction to missing data handling and optimal DTR identification. The pipeline that the study team will create, carefully document and disseminate will be useful for recomputing optimal DTRs for aNSCLC patients as new treatments are developed and EHR databases evolve. In addition, the pipeline will be applicable to optimal DTR estimation for other diseases requiring multi-stage treatment decision making (e.g., melanoma, prostate cancer, epilepsy and rheumatoid arthritis).

Project Information

Jincheng Shen, B.S., M.S., Ph.D.
Daniel Scharfstein, Ph.D.
University of Utah
$1,069,908 *

Key Dates

36 months *
April 2024

*All proposed projects, including requested budgets and project periods, are approved subject to a programmatic and budget review by PCORI staff and the negotiation of a formal award contract.


Project Status
Award Type
State State The state where the project originates, or where the primary institution or organization is located. View Glossary
Last updated: April 23, 2024