Jian-Lun Xu

Ph.D

NCI

Mathematical Statistician

PLCO (Learn more about this study)

PLCO-600

Mar 25, 2020

Estimating the probability of overdiagnosis in PLCO trial

Screening procedures are widely recommended by various professional and advocacy groups and utilized for early detection of diseases such as cancer with the goal of detecting disease early, before it becomes symptomatic. The hope is that treatment may be easier and more effective, increasing patient life span when the disease is detected at an earlier stage. While the potential value of early detection is clear, it is offset in part by the occurrence of false (positive or negative) results and overdiagnosis. Compared to false results, overdiagnosis has received little attention because it is a challenging concept to understand and communicate and cannot be directly measured. In this project, we develop a model to estimate the probability of overdiagnosis in a repeated screening program. The asymptotic normality of the estimate is proved. In addition, we apply the model and method to the data from PLCO trial.

• Develop a model to estimate the probability of overdiagnosis.
• Study the asymptotic property of the estimate.
• Apply the model and method to each screening modality in the PLCO trial.

I need the following data from each of screening modalities:
1. Control group:
nc = number of cancers diagnosed clinically,
Nc = sample size of control group.
2. Intervention group:
nrc = number of cancers diagnosed clinically in the refusal group,
Nrc = number of individuals in the refusal group,
nsc = number of cancers detected by a screening modality,
nic = number of interval cancers from a screening modality,
Ni = sample size of intervention group.
pn = post screen cases in the intervention arm.