Early cancer screening can play a transformative role in healthcare, potentially saving lives by identifying cancer before it progresses to an advanced, less treatable stage. However, designing effective screening programs isn’t easy. We need solid evidence to show that screening tests improve health outcomes, which often means running long and expensive studies. In a recent study in Biometrics, Kehao Zhu, Ying-Qi Zhao, and Yingye Zheng offer a new approach to cancer screening trial design, focusing on an alternative endpoint: reducing the incidence of late-stage cancer rather than focusing solely on cancer mortality. This shift in focus could make screening trials more feasible and produce results that can be applied more rapidly in public health. Reduction in late-stage cancer incidence has been considered as an important surrogate endpoint that can inform mortality benefit. Additionally, focusing on late-stage cancer incidence can avoid the complex ethical considerations surrounding “overdiagnosis,” where screening detects cancers that may never become symptomatic or life-threatening. Designing a trial based on late-stage cancer reduction could make screening trials more feasible with a shorter follow-up period and produce results that can be applied more rapidly in public health.
Dr. Zheng emphasizes the urgency of adapting screening trial designs to new technologies: “We are at a pivotal moment in cancer early detection, with innovative liquid biopsy technologies enabling the possibility of screening for cancers through simple blood tests. It is urgent to establish evidence-based public policies that address both the benefits and potential harms of these tests as they become widely available.” Screening trials need to keep pace with these technological advances, but the current framework for designing these trials is both time-intensive and costly. Standard cancer treatment trials are based on measuring outcomes like survival in patients with a known diagnosis. Screening trials need to follow asymptomatic, healthy individuals over time to observe if early detection can lead to fewer cases of late-stage cancer, and by extension, potentially lower mortality. As Dr. Zhao notes, “Randomized screening trials are essential for determining the clinical utility of these tests—specifically, whether they lead to meaningful improvements in patient outcomes, such as reduced mortality.”
To address these challenges, the study introduces a five-state disease progression model designed to characterize cancer’s development and measure the impact of screening interventions on reducing late-stage cancer incidence. Depending on the disease progression and screening frequency, the impact of screening is often time-varying. The model tracks the progression of disease through various stages, from normal tissue to early-stage preclinical cancer, to late-stage preclinical cancer, and finally to early- and late-stage clinical disease. This framework addresses the reality that early detection can prevent cases from progressing to advanced stages, even if the direct impact on mortality requires longer follow-up to confirm. As Dr. Zhu explains, “To navigate these challenges, our disease model may provide a powerful tool to improve the rigor and efficiency of screening trials aimed at reducing late-stage cancer incidence.”
Their approach builds on an extensive background of prior biomarker studies that identify potential early detection tests for cancer. The trial design begins with a rigorous process of biomarker discovery and validation, ensuring the selected test has sufficient sensitivity (ability to detect cancer) and specificity (ability to avoid false positives). In a trial using this design, participants are divided into two groups: an experimental group, which receives the new screening test, and a control group, which follows the current standard of care without additional screening. Both groups are monitored for the incidence of late-stage cancer, and the model calculates whether the new test leads to a meaningful reduction in such cases compared to standard care.
As illustrated in the study’s summary figure, the study outlines key aspects of the trial: Q1a, the model investigates whether the screening test significantly changes the incidence of late-stage cancer compared to the control group. Moving to Q1b, the model explores whether reducing late-stage cancer incidence ultimately affects cancer-specific or all-cause mortality. Another vital question, Q2, addresses whether the clinical performance of the test justifies launching a larger utility trial. This step is crucial because it helps decide if the test's screening benefits outweigh the potential harms associated with screening, Q3 further examines how to measure these possible harms, ensuring that any new screening program is balanced between benefits and risks.
Through numerical examples based on data from the National Lung Screening Trial (NLST), the researchers demonstrated that the model could identify an optimal design that achieves high statistical power without the need for exceptionally large samples or extended follow-up times. The team’s analysis showed that high statistical power could be achieved faster than in traditional trials, allowing results to be obtained with fewer participants and shorter follow-up periods. Dr. Zhu highlights this practical impact: “Our model can also be used to establish performance criteria for new biomarker tests, which are critical for launching successful large-scale screening initiatives.”
The team is already extending their incidence model to incorporating mortality as an additional endpoint, alongside late-stage cancer incidence. “Our generic disease model can be adapted to design and interpret results for multi-cancer early detection tests, opening up new avenues for research and improving our understanding of how these tests can impact patient outcomes,” the team explains.
In conclusion, this study contributes a valuable new tool to cancer research: a model that enables shorter, more efficient screening trials without sacrificing scientific rigor. Their model advances trial design by providing a practical, achievable endpoint—late-stage cancer incidence reduction—that balances the urgency of early detection with the ethical considerations of overdiagnosis. As technologies like liquid biopsy advance, this model could fundamentally transform how we assess and implement screening programs, accelerating the development of public health policies that emphasize both patient safety and the life-saving potential of early cancer detection
The work is supported by grants awarded by the National Institutes of Health.
Kehao Zhu is a research associate at Fred Hutch and this work is part of his dissertation for the Ph.D. in Biostatistics from University of Washington. Fred Hutch/University of Washington/Seattle Children’s Cancer Consortium members Drs. Ying-Qi Zhao, and Yingye Zheng contributed to this study.
Zhu, K., Zhao, Y. Q., & Zheng, Y. (2024). Designing cancer screening trials for reduction in late-stage cancer incidence. Biometrics, 80(3), ujae097.