A novel risk-based approach could be an effective way to triage women for mammography during times of limited capacity, such as the situation during the COVID-19 pandemic, according to new findings.
Researchers evaluated almost 2 million mammograms that had been performed at more than 90 radiology centers and found that 12% of mammograms with “high” and “very high” cancer risk rates accounted for 55% of detected cancers.
In contrast, 44% of mammograms with very low cancer risk rates accounted for 13% of detected cancers.The study was published online March 25 in JAMA Network Open.
Cancer screening programs dramatically slowed or even came to a screeching halt during 2020, when restrictions and lockdowns were in place. The American Cancer Society even recommended that “no one should go to a health care facility for routine cancer screening,” as part of COVID-19 precautions.
However, concern was voiced that the pause in screening would allow patients with asymptomatic cancers or precursor lesions to develop into a more serious disease state.
The authors point out that several professional associations had posted guidance for scheduling individuals for breast imaging services during the COVID-19 pandemic, but these recommendations were based on expert opinion. The investigators’ goal was to help imaging facilities optimize the number of breast cancers that could be detected during periods of reduced capacity using clinical indication and individual characteristics.
The result was a risk-based strategy for triaging mammograms during periods of decreased capacity, which lead author Diana L. Miglioretti, PhD, explained was feasible to implement. Miglioretti is division chief of biostatistics in the Department of Public Health Sciences at University of California Davis.
“Our risk model used information that is commonly collected by radiology facilities,” she told Medscape Medical News. “Vendors of electronic medical records could create tools that pull the information from the medical record, or could create fields in the scheduling system to efficiently collect this information when the mammogram is scheduled.”
Miglioretti emphasized that once the information is collected in a standardized manner, “it would be straightforward to use a computer program to apply our algorithm to rank women based on their likelihood of having a breast cancer detected.”
“I think it is worth the investment to create these electronic tools now given the potential for future shutdowns or periods of reduced capacity due to a variety of reasons, such as natural disasters and cyberattacks — or another pandemic,” she said.
Some facilities are still working through backlogs of mammograms that need to be rescheduled, which would be another way that this algorithm could be used. “They could use this approach to determine who should be scheduled first by using data available in the electronic medical record,” she added.
Five Risk Groups
Miglioretti and colleagues conducted a cohort study using data that was prospectively collected from mammography examinations performed from 2014-2019 at 92 radiology facilities in the Breast Cancer Surveillance Consortium. The cohort included 898,415 individuals who contributed to 1.8 million mammograms.
Information that included clinical indication for screening, breast symptoms, personal history of breast cancer, age, time since last mammogram/screening interval, family history of breast cancer, breast density, and history of high-risk breast lesion was collected from self-administered questionnaires at the time of mammography or extracted from electronic health records.
Following analysis, the data was categorized into five risk groups: very high
(> 50), high (22-50), moderate (10-22), low (5-10), and very low (< 5) cancer detection rate per 1000 mammograms. These thresholds were chosen based on the observed cancer detection rates and clinical expertise.
Of the group, about 1.7 million mammograms were from women without a personal history of breast cancer and 156,104 mammograms were from women with a breast cancer history. Most of the cohort were between ages 50-69 years at the time of imaging, and 67.9%were White (11.2% Black, 11.3% Asian or Pacific Islander, 7% Hispanic, and 2.2% were another race/ethnicity or mixed race/ethnicity).
Their results showed that 12% of mammograms with very high (89.6 to 122.3 cancers detected per 1000 mammograms) or high (36.1 to 47.5 cancers detected per 1000 mammograms) cancer detection rates accounted for 55% of all detected cancers. These included mammograms that were done to evaluate an abnormal test or breast lump in individuals of all ages regardless of breast cancer history.
On the opposite end, 44.2% of mammograms with very low cancer detection rates accounted for 13.1% of detected cancers and that included annual screening tests in women aged 50-69 years (3.8 cancers detected per 1000 mammograms) and all screening mammograms in individuals younger than 50 years regardless of screening interval (2.8 cancers detected per 1000 mammograms).
Treat with Caution
However, in an accompanying editorial, Sarah M. Friedewald, MD, and Dipti Gupta, MD, both from Northwestern University, Chicago, Illinois, point out that while the authors examined a large dataset to identify a subgroup of patients who would most likely benefit from breast imaging in a setting where capacity is limited, “these data should be used with caution as the only barometer for whether a patient merits cancer screening during a period of rationing.”
They note that in the context of an acute crisis, when patient volume needs to be reduced very quickly, it is often impractical for clinicians to sift through patient records in order to capture the information necessary for triage. In addition, asking nonclinical schedulers to accurately pull data at this level, at the time when the patient calls to make an appointment, is unrealistic.
In the context of the pandemic, the editorialists write that while this model uses risk for breast cancer to prioritize those to be seen in the clinic, the risk for complications from COVID-19 may also be an important factor to consider. For example, an older patient may be at a higher risk for breast cancer but may also face a higher risk for COVID-related complications. Conversely, a younger woman at a lower risk for serious COVID-related disease but who has breast cancer detected early will gain more life-years than an older patient.
There are also no algorithms to account for each patient’s perceived risk for breast cancer or COVID-19, and “the downstream effect of delaying cancer diagnosis may similarly lead to unintended consequences but may take longer to become apparent,” they write. “Focusing efforts on the operations of accommodating as many patients as possible, such as extending clinic hours, would be preferable.”
Finally, Friedewald and Gupta conclude that “the practicality of this process during the COVID-19 pandemic and extrapolation to other emergent settings are less obvious.”
The study was supported through a Patient-Centered Outcomes Research Institute program award. Miglioretti reported receiving royalties from Elsevier outside the submitted work; several co-authors report relationships with industry. Friedewald reported receiving grants from Hologic Research during the conduct of the study. Gupta has disclosed no relevant financial relationships.