Is there an optimal mammography recall rate?

By Kate Madden Yee, staff writer
March 1, 2019

There's an optimal range for mammography screening recalls that maximizes the detection of life-threatening cancers and minimizes the harms of false positives and overdiagnosis, according to a presentation delivered on Wednesday at ECR 2019 in Vienna.

Finding the recall rate "sweet spot" is important, especially when the range is wide, said presenter Dr. Matthew Wallis of Cambridge University Hospitals NHS Trust.

"One of the major screening challenges, at least in the U.K., is that we have very tight definitions and standards for recall rates, but we're not very good at enforcing them," he said. "There's a very wide range of recall rates in the U.K., particularly for first screening mammograms -- from 3.5% to 15%. If you've got that kind of range, something must be going on that needs to be addressed."

Low recall rates are associated with low cancer detection and low sensitivity. But high recall rates can also be problematic, Wallis and colleagues noted. As recall rates rise, the cancer detection rate for invasive cancers and high/intermediate-grade ductal carcinoma in situ (DCIS) reaches a plateau above which almost all recalls are false positives.

The group modeled the association between cancer detection and recall rates to better understand the optimal balance of harm and benefit of screening mammography. The study included data from 11.3 million screening exams taken between April 2009 and March 2016. Of these, 2.3 million were prevalent (first) screens in women between the ages of 45 and 52, and 8.9 million were incident (further screening) exams in women between the ages of 53 and 70. The team also included previously published data from a Dutch screening program.

From their model, Wallis and colleagues found that the following recall rate ranges appear to optimize the benefit/harm ratio of mammography screening:

  • Prevalent screening: 4.6% to 7%
  • Incident screening: 2.6% to 4%

"If your recall rate is less than 4.6% for first screening mammograms, you're probably not finding enough cancers, but if your rate is over 7%, you're probably doing more harm than good," Wallis said.

He and his colleagues hope other European countries will model their recall rates as well so the data can be assessed in the context of other screening programs.

"For all cancers, both prevalent and incident, the more you recall, the more false positives," he concluded. "Our model predicts there is a recall range that optimizes detection of important cancers while minimizing harm of false positives and potential overdiagnosis of low-risk cancers."

AI-based CAD software improves breast cancer detection
VIENNA - Computer-aided detection (CAD) software based on artificial intelligence (AI) technology can help radiologists of all levels of experience detect...
High mammography recall rates show diminishing returns
A new analysis of more than 11 million breast screening exams from the English National Health Service program has shown that increases in recall...
Swedish trial shows DBT detects more breast cancer
Digital breast tomosynthesis (DBT) with a single view detects over a third more breast cancers than traditional mammography, a large Swedish study published...
DBT: Longer reading time, lower breast cancer recall rate
Interim analysis of a randomized controlled trial on digital breast tomosynthesis (DBT) showed the modality took longer to read than digital mammography...
DBT + mammo improves cancer detection, increases recalls
In a substudy of a large digital breast tomosynthesis (DBT) trial, Italian researchers found screen reading using DBT with either acquired or synthesized...

Copyright © 2019

Last Updated np 3/7/2019 2:23:18 PM

2 comments so far ...
3/1/2019 3:03:17 AM
Philip W
[The message below was sent anonymously and posted by Philip Ward, editor in chief,, on 1 March 2019] 
11 million cases do not necessarily give quality to the authors' conclusions, as mentioned last month. They are rambling about low grade "DCIS", but do they even know what "DCIS" actually is? Low grade "DCIS" is localized within the "lobules", not in the "ducts". In addition, there are 12 different mammographic presentations of carcinoma in situ, so the term "DCIS" is extremely misleading, since it is a single name for 12 different diseases! The U.K. really needs to publish the mortality results of their screening -- since their three-year long (or often longer) inter-screening interval does not help at all in detecting the rapidly growing, high-grade cancers. So what is the point of always talking about harm and benefit when they deliberately enhance the "harm" and undervalue the benefit? Of all the harms associated with breast cancer screening, the greatest harm comes from non-attendance (J Malmgren's article, Radiology, 2012; 262:797-806). The choice of age limit by Wallis et al. for screening is archaic and their loooong inter-screening interval is actually allowing ("encouraging") the potentially killing cancers grow to an undesirable size and stage. 

3/1/2019 3:57:30 PM
John Keen, MD
As the previous article on this topic,
the optimal rate depends on the selection of marginal ppv.
See reply to Grabler, AJR 217 209 (2).
BTW, Malmgren's article is junk science.  Increasing screening intervals does little for the advance cancer rate.