Study
PLCO
(Learn more about this study)
Project ID
PLCO-1821
Initial CDAS Request Approval
Feb 12, 2025
Title
Leveraging causal inference approaches to bridge knowledge gaps in colorectal cancer screening
Summary
Colorectal cancer (CRC) is the third most common cancer in US adults and the second leading cause of cancer related deaths. Despite recent advances in CRC screening (CRCS) and treatment, approximately 152,810 adults were diagnosed with CRC in 2024, with 53,010 CRC related deaths. Efforts to inform comprehensive CRCS policy are complicated by the limitations of ‘gold-standard’ estimates of effectiveness from randomized screening trials, intention-to-screen (ITS) effects, which target the effect of randomization to a screening strategy (e.g. sigmoidoscopy) in a particular trial (e.g. PLCO trial). ITS effects vary widely across trials, driven primarily by differential non-adherence, which essentially dilutes the ‘true’ effect of screening. While IV methods may reconcile variations in ITS effects by estimating local average screening effects (LASE), such efforts fail to target the primary insight of interest in patient/provider decision making - the causal benefit of adhering to a given CRCS strategy. Further, existing estimates are challenging to transport as trial samples are often highly selected, and IV estimates among ‘compliers’ (those that adhere to the randomized strategy) constitute an ultimately unidentifiable group. Equitable patient-oriented CRCS policy requires understanding the benefit of adhering to a CRCS strategy (as opposed to targets of ITS and LASE effects), and the transportability of such effects.
Aims
Aim 1: Estimate per-protocol effects of flexible sigmoidoscopy in the Prostate, Lung, Colorectal & and Ovarian (PLCO) Cancer Screening trial using inverse probability weighted marginal structural models.
• Leverage inverse probability of adherence weighted marginal structural models to estimate causal effects of adherence to the flexible sigmoidoscopy screening strategy in the PLCO trial. We will compare our estimated per-protocol effects to the ITS effects from the original trial, with the larger goal of improving the translation of trial findings to clinical practice and patient decision-making.
Aim 2: Transport per-protocol effects of sigmoidoscopy in the PLCO trial to the CRCS-eligible US target population using inverse odds re-weighting under the counterfactual framework.
• We will combine PLCO trial data with population-level data (NHIS, 2010) via inverse-odds re-weighting to estimate per-protocol effects of CRCS (specifically flexible sigmoidoscopy) in the corresponding screening-eligible target population, to generate externally valid estimates of CRCS effectiveness.
Collaborators
Katarzyna Wyka (CUNY Graduate School of Public Health and Health Policy)
Zachary Shahn (CUNY Graduate School of Public Health and Health Policy)
