Large-scale mammography screening has undoubtably saved lives, but there’s also a concern that it has led to breast cancer being over diagnosed.
As radiologists and scientists develop new ways of understanding how breast cancer presents and progresses, there is a hope that cancers that will progress into invasive cancer can be caught and treated, while reducing the amount of women who may undergo unnecessary procedures.
A session on Sunday included experts talking about the frequency of overdiagnosis, imaging trends of ductal carcinoma in situ (DCIS) and how data from radiomics can help get more information about tumors that may progress to invasive cancer.
How Often Is Breast Cancer Overdiagnosed?
Marc Ryser, PhD, assistant professor in the Department of Mathematics at Duke University, defined overdiagnosis as occurring in a patient with an asymptomatic tumor that she never would have known about without screening and where the patient would have died of another cause aside from breast cancer before the tumor became symptomatic.
Dr. Ryser said it has been hard to estimate how much overdiagnosis is occurring and estimates have ranged from 0 to 50%. However, his recent research worked with the Breast Cancer Surveillance Consortium and studied 36,000 women aged 50-74 who were screened between 2000 and 2018 using a mathematical model-based approach. The study found that the overall rate of overdiagnosis was 15%, with non-progressive tumors making up 6% and progressive tumors that would not have become symptomatic in the woman’s remaining lifetime making up the other 9%.
“This tells us that about one in seven screen-detected cancers will be overdiagnoses. That number is higher for women who are older and sicker, lower if younger and healthier,” Dr. Ryser said. “But it is lower than many of the influential prior studies that have dominated the discourse over the last decade or so.”
There is a concern that this is a cancer of screening and that the more screening we do, the more DCIS we find, according to Lars Grimm, MD, associate professor of radiology at Duke University and member of the Duke Cancer Institute.
But, showing images of four different types of DCIS that all evolved into invasive cancer, he highlighted the difficulty of knowing what tumors will progress into a dangerous cancer. Some qualities on imaging can help predict invasive disease, he said, such as asymmetry, extent of calcifications, and other suspicious features, but more research is needed.
“As radiologists we are typically trained to make a determination between cancer and not cancer, but we need to evolve our understanding to better predict invasive cancer versus not,” Dr. Grimm said.
Role of Radiomics in Understanding Tumor Progression
The session also included research done by Despina Kontos, PhD, associate vice-chair for research, Department of Radiology, University of Pennsylvania, on how radiomic signatures can help reduce overtreatment for DCIS.
“The amount of features you can extract from your images is in some ways only limited by your creativity,” said Dr. Grimm, who presented on Dr. Kontos’ behalf. For example, there are studies looking at breast MR imaging features associated with recurrence after treatment, and others looking at spatial tumor heterogeneity to make future predictions.
Radiomics is still developing and is often based on small datasets at individual institutions. However, Kontos’ presentation highlighted the Cancer Phenomics Toolkit, or CaPTk, a platform developed at Penn designed to be an open-source way to use image validated data to analyze more data, share information, and help spread knowledge.
“The goal of this line of work is to take these multiple inputs—the imaging, the clinical information, patient demographic information, the pathology information and any other molecular marker you can get— to put all this information together and the dream is that we would be able to provide our patients with the greatest precision and value in our guidance for their treatment going forward,” Dr. Grimm said.
Access the presentation, "Reducing Overtreatment of Breast Cancer: Leveraging Insights from Imaging, Population Data, and Radiomics," (S1-CBR03) on demand at Meeting.RSNA.org.