Detecting possible individuals with benefit from ovarian cancer screening: a counterfactual prediction modelling study
Principal Investigator
Name
David van Klaveren
Institution
Erasmus University Medical Center
Position Title
Associate professor
Email
About this CDAS Project
Study
PLCO
(Learn more about this study)
Project ID
PLCO-1606
Initial CDAS Request Approval
Jul 8, 2024
Title
Detecting possible individuals with benefit from ovarian cancer screening: a counterfactual prediction modelling study
Summary
Ovarian cancer is the most lethal gynecological cancer due to its non-specific symptoms, often leading to diagnosis in advanced stages. Despite that intensive efforts have been made to devise a screening method, two large randomized trials on ovarian cancer screening did not show a reduction in ovarian cancer-related deaths among screened women compared to those who were not screened.
The UK Collaborative Trial of Ovarian Cancer Screening (UKCTOCS) indicated that the benefit of early detection through screening may not be uniform across all cases. Researchers noted that cases detected through screening might have different progression pathways and oncological outcomes. This heterogeneity suggests that early detection does not universally translate to reduced mortality, highlighting the importance of considering the variability in screening effectiveness. An exploratory analysis of the UKCTOCS trial emphasized this point by showing that screening could identify high-grade serous cancers earlier, leading to better treatment outcomes for this particular subtype. This finding supports the necessity of targeting the heterogeneous effects of ovarian cancer screening, focusing on the types of cancer and patient characteristics that benefit most from early detection.
To address this, counterfactual prediction modeling has emerged as a promising method. This approach, often described as a blend of traditional prediction modeling and causal inference, allows researchers to predict the outcomes for individuals had they received a different treatment or intervention. By employing counterfactual models, it becomes possible to estimate the benefits of screening for specific subgroups of women based on their unique characteristics.
The proposed study aims to leverage counterfactual prediction models to predict all-cause mortality in the context of ovarian cancer screening trials. By treating screening or no screening as the intervention, the study seeks to identify subgroups based on baseline characteristics that derive the most benefit from screening. This approach promises to enhance the efficiency of ovarian cancer screening programs and improve outcomes for those most likely to benefit, thereby addressing the limitations observed in previous large-scale screening trials.
The UK Collaborative Trial of Ovarian Cancer Screening (UKCTOCS) indicated that the benefit of early detection through screening may not be uniform across all cases. Researchers noted that cases detected through screening might have different progression pathways and oncological outcomes. This heterogeneity suggests that early detection does not universally translate to reduced mortality, highlighting the importance of considering the variability in screening effectiveness. An exploratory analysis of the UKCTOCS trial emphasized this point by showing that screening could identify high-grade serous cancers earlier, leading to better treatment outcomes for this particular subtype. This finding supports the necessity of targeting the heterogeneous effects of ovarian cancer screening, focusing on the types of cancer and patient characteristics that benefit most from early detection.
To address this, counterfactual prediction modeling has emerged as a promising method. This approach, often described as a blend of traditional prediction modeling and causal inference, allows researchers to predict the outcomes for individuals had they received a different treatment or intervention. By employing counterfactual models, it becomes possible to estimate the benefits of screening for specific subgroups of women based on their unique characteristics.
The proposed study aims to leverage counterfactual prediction models to predict all-cause mortality in the context of ovarian cancer screening trials. By treating screening or no screening as the intervention, the study seeks to identify subgroups based on baseline characteristics that derive the most benefit from screening. This approach promises to enhance the efficiency of ovarian cancer screening programs and improve outcomes for those most likely to benefit, thereby addressing the limitations observed in previous large-scale screening trials.
Aims
1 Build a counterfactual prediction model for the all-cause/ovarian cancer mortality in the ovarian cancer screening trial setting with screening or not
2 Identify subgroups based on baseline characteristics that derive the most benefit from screening
3 Guide future ovarian cancer screening trials.
Collaborators
Keling Wang, Department of Epidemiology, Erasmus MC, University Medical Center Rotterdam, the Netherlands
Dianqin Sun, Department of Public Health, Erasmus MC, University Medical Center Rotterdam, the Netherlands
Jeremy A Labrecque, Department of Epidemiology, Erasmus MC, University Medical Center Rotterdam, the Netherlands
Iris Lansdorp-Vogelaar, Department of Public Health, Erasmus MC, University Medical Center Rotterdam, the Netherlands