Skip to Main Content
Principal Investigator
Paul Pinsky
National Cancer Institute
Position Title
About this CDAS Project
NLST (Learn more about this study)
Project ID
Initial CDAS Request Approval
Nov 19, 2010
Variability Among NLST Radiologists
Reader variability has been described for mammography and other radiologic modalities. Reader variability can be assessed both in a controlled setting, where readers all read the same cases, and in a clinical setting, where readers read different images and overall positivity rates for readers are compared. We propose to examine reader variability in NLST in both CT and CXR. Further, we propose to link the findings of the previous NLST controlled reader studies with those from actual NLST practice to compare variability assessed by the two methods. Finally, utilizing the NLST forms, which assess whether nodules were pre-existing, we can assess the rate at which nodules identified at one read were also identified at another read, by the same or a different radiologist.

Note: Additional work related to this project is being continued in <a href="/cdas/nlst/pubs_projects/project/476/">201112-0018</a>.

1. To compare the following rates across radiologists in NLST: T0 positivity in CT, T0 positivity in CXR, T1/2 initial positivity in CT/CXR, T1/2 final positivity in CT/CXR, rates of finding 4-9 and 10+ nodules in CT (at T0, and at T1/2), mean # of 4+ nodules identified in CT. Also, to examine whether readers' positivity rates correlate with the rates at which their positive subjects receive invasive f/u. We will also examine false positive rates for CT and CXR. To do this, we will identify any cancer diagnoses occuring within one year of the screen in question and eliminate these.

2. To assess the correlation between a radiologists' positivity rate in CT and their positivity rate in CXR. In other words, we will assess the following question: if a radiologist tends to have a high positivity (false positivity) rate on CT relative to the population of radiologists, does he/she tend to have a high postivity (false positivity) rate on CXR relative to the population of radiologists.

3. To compare variability in NLST rates with variability assessed in the NLST reader studies. Primarily, this would focus on the reader CXR variability study.

4. To examine the rate at which nodules are identified on two separate reads of the same case. THis will be done utilizing the section of the NLST form that examines whether a nodule was pre-existing or not.

Related Publications