Full Body MRIs vs. liquid biopsies for early cancer detection
I often get asked how Full Body MRIs compare to liquid biopsies (also known as circulating-tumor DNA tests). For those not familiar with these types of screening procedures, full body MRIs take an imaging approach to cancer detection - we’re trying to “see” tumors in the body. Liquid biopsies take a molecular approach, trying to find genetic material in the bloodstream that originated from a malignant tumor.
The accuracy of a screening test is measured using sensitivity and specificity. Before we look at how sensitivity & specificity compare for these two types of screening exams, let’s start with some definitions.
Sensitivity is the percentage of true positives. If 100 people with cancer get a screening test and it comes back positive for 90 of them, the test has a 90% sensitivity. The remaining 10% are false negatives, as the test incorrectly determined that these 10 people don’t have cancer.
Specificity is the percentage of true negatives. If 100 people without cancer get a screening test and it comes back negative for 90 of them, the test has a 90% specificity. The remaining 10% are false positives, as the test incorrectly determined that these 10 people have cancer.
Sensitivity and specificity cannot be taken in isolation: they are flip sides of the same coin. The higher the sensitivity, the lower the specificity, and vice versa. A useful analogy is thinking about metal detectors at the airport, calibrated to detect firearms. If the metal detector is too sensitive, it will also go off for belts and watches. If it’s too specific, it might miss a gun. For firearms, as well as for cancer, we want our tests to be highly sensitive (as opposed to specific), as false negatives are much worse to deal with than false positives. That’s because early-stage cancers are more curable when detected early.
So what are the acceptable levels of sensitivity and specificity in cancer screening? If we look at USPSTF-mandated screening guidelines, we find four screening procedures that are considered to have acceptable sensitivity & specificity: mammogram for breast cancer, low-dose chest CT for lung cancer in smokers, colonoscopy for colon cancer, and pap smear for cervical cancer.
Organ | Modality | Sensitivity | Specificity |
Breast | Mammogram | 85% | 90% |
Lung | Low-dose CT | 88% | 92% |
Colon | Colonoscopy | 95% | 86% |
Cervical | Pap-smear | 75% | 98% |
Sources: mammogram, low-dose CT, colonoscopy, pap-smear.
What’s clear from the table above is that the currently accepted screening procedures were selected because they have both relatively high sensitivity (~80%) and high specificity (~90%). So how do Full Body MRIs and liquid biopsies compare?
Full Body MRI
The table below shows sensitivity & specificity for Stages I - III as per the public literature in the organs covered by the Ezra Full Body MRI scan. These numbers do not necessarily apply to any arbitrary Full Body MRI. They only apply to scans that have been calibrated carefully for high sensitivity and specificity in cancer detection, such as the best-practice protocols we use at Ezra in each body area of interest.
Organ | Sensitivity | Specificity |
Brain | 92% | 85% |
Pancreas | 85% | 63% |
Prostate | 72% | 96% |
Liver | 84% | 94% |
Brain | 92% | 85% |
Bladder | 80% | 78% |
Esophagus | 92% | 67% |
Kidney | 91% | 89% |
Sarcoma | 65% | 85% |
Ovaries | 97% | 84% |
Uterus | 95% | 96% |
Stomach | 75% | 84% |
Larynx | 90% | 76% |
Thyroid | 84% | 92% |
Gallbladder | 91% | 87% |
Average | 87% | 85% |
The sensitivity and specificity of Full Body MRIs across our organs of interest is on par with what is expected from a high-quality screening procedure (87% sensitivity, 85% specificity). Sources: scroll to the bottom of the full body MRI page on the Ezra website.
ctDNA
For liquid biopsy data, we will use public research published by GRAIL, the leading company in the liquid biopsy space. GRAIL reports overall sensitivity across all stages as well as Stages I - III. The table below shows sensitivity & specificity across Stages I - III.
Cancer Type | Sensitivity (%) | Specificity |
Prostate | 6% | 99.5% |
Breast | 28% | 99.5% |
Colon/Rectum | 77% | 99.5% |
Cervix | 79% | 99.5% |
Melanoma | 0% | 99.5% |
Thyroid | 0% | 99.5% |
Urothelial Tract | 0% | 99.5% |
Kidney | 8% | 99.5% |
Uterus | 26% | 99.5% |
Bladder | 29% | 99.5% |
Gallbladder | 44% | 99.5% |
Stomach | 47% | 99.5% |
Sarcoma | 50% | 99.5% |
Lung | 63% | 99.5% |
Pancreas | 69% | 99.5% |
Esophagus | 75% | 99.5% |
Head and Neck | 76% | 99.5% |
Ovary | 78% | 99.5% |
Anus | 81% | 99.5% |
Liver/Bile-duct | 88% | 99.5% |
Multiple primaries | 100% | 99.5% |
Myeloid Neoplasm | 20% | 99.5% |
Lymphoid Leukemia | 41% | 99.5% |
Lymphoma | 55% | 99.5% |
Plasma Cell Neoplasm | 72% | 99.5% |
Total | 48% | 99.5% |
I was pleasantly surprised to see the high sensitivity in some of the organs GRAIL covers. The specificity in the table above is the same across all organs as GRAIL only reports aggregate specificity. Source: GRAIL manuscript.
Comparison
Now that we have looked at individual accuracy data for Full Body MRIs and ctDNA tests, let’s bring it all together in a table and see how they compare. For the purpose of the analysis, we’ll also exclude the four organs that already have well established screening guidelines, and do a join between the remaining organs so we can do an apples-to-apples comparison.
To make it easy to visualize, I’ve color coded the table so we can see which modality has higher sensitivity / specificity.
Conclusion
When it comes to cancer we want a test that minimizes false negatives (high sensitivity) to avoid giving people a false sense of security. We also want a test that has high specificity, to reduce false positives and minimize unnecessary follow-up procedures.
Looking at the table above, it’s obvious that Full Body MRIs have superior sensitivity when compared to ctDNA. Conversely, liquid biopsies (at least in the aggregate) have higher specificity across the board. This makes the two tests highly complementary. Therefore, a screening protocol that includes both a Full Body MRI as well as a ctDNA test would obtain the best mix of sensitivity & specificity for early cancer detection.
It’s important to note that these tests are not cheap. The Ezra Full Body Flash, our most affordable full body MRI, costs $1,350. A GRAIL test costs $1,000. For those who can afford it, I do recommend both tests. At Ezra, we are hard at work bringing future costs of our Full Body MRI down to $500.
For those who can only purchase one test, pairing them serially starting with a Full Body MRI would be optimal. The Ezra Full Body MRI’s high sensitivity will ensure that every relevant finding is identified. If a finding in the MRI needs further evaluation, a GRAIL ctDNA test could help determine if the finding is malignant, increasing the overall specificity of the protocol.
My view of the future is that accurate cancer detection will require a multi-modality approach. I’m excited to see so much progress being made across the board, both by Ezra and by our peers in the industry.
Thank you to Dr. Daniel K. Sodickson and Dr. Carol DerSarkissian for reviewing drafts of this post.