BACKGROUND: Since the 1800s, pathologists have reported seeing malignant cells floating free in the blood stream of cancer patients. However, the technology to find extremely low numbers of circulating tumor cells (CTCs) is more recent, starting about 15 years ago when the term “liquid biopsy” emerged. The next step was new technology that allowed detection of “tumor DNA” (tDNA), or fragments of DNA that “leak” from cancer cells and can be detected in the bloodstream. The implication is that cell death has occurred, releasing the tDNA. (Note: a patient can have “cell free DNA” which is not from a malignant tumor, rather from cells that die normally). It became apparent that tDNA in the bloodstream might be better than CTCs when it comes to using “liquid biopsy” for cancer screening. (NOTE: there are other uses for tDNA besides screening — monitoring for recurrences, residual disease checks, guiding specific therapies. However, I am focused here entirely on asymptomatic screening).
In 2016, the company GRAIL was formed through financing strategies that included Jeff Bezos and Bill Gates, eventually raising over 2 BILLION dollars, with multiple studies launched using this new technology in several hundred thousand patients. One of these studies resulted in the Galleri blood test that can detect more than 50 different types of cancer. Galleri is not the only test available for tDNA, but for practical purposes, we’ll focus on Galleri.
FULL DISCLOSURE: Since I have been heavily involved in development of a blood test for breast cancer (only 1 of the 50+ cancer types from Galleri), I have had to follow the developments in this space closely ever since the inception of the technology. Furthermore, I am a consultant for Syantra, Inc., a Canadian company that is in the process of releasing their breast cancer blood test across Canada. And while it might seem that our blood test is in direct competition with Galleri, this is not the case. In fact, there is no contest between our single cancer approach versus the breast cancer data from Galleri. When the tDNA data started to emerge for breast cancer, it became quite clear that breast cancer does not “leak” its DNA reliably enough for early detection. So, we forged ahead using a different, albeit longstanding, approach — that is, circulating biomarkers — proteins that reflect the presence of cancer cells, in this case, specific proteins for breast cancer. So far, we chose the correct approach, though who knows what technology might be in the works? So, is Galleri the long-anticipated blood test for breast cancer I’ve been working on for nearly 30 years? No. (Other cancer types are another story — see Addendum).
MEDIA ANNOUNCEMENT: The reason I’m writing about this is that Mercy Hospital announced through a multi-media approach that it would be offering the Galleri test. This came from Mercy Central in Missouri, where many policies are established for the network hospitals and physicians. As you might imagine, I had people contacting me to see if the Galleri test had any relation to my research at Mercy where we had shipped 10,000 samples to researchers around the world (the Galleri test, by the way, is based on 20,000). The answer is: No. But my question is: “Why the middleman?” Galleri doesn’t need a hospital-based program. It’s been available by home kit for nearly a year. All you need is a registered physician-provider who agrees to write the prescription for the test, then the kit is sent, blood drawn at any lab (22 approved sites within a 20-mile radius of my home), then the physician gets the report. What does Mercy add? I don’t know, but the Galleri folks are experimenting with new business models, so I suspect there’s a “deal” that’s been made, and the question will be — who benefits? Does the patient get a discount on the nearly $1,000 cost if they go through a Mercy physician? If so, then great. But call me a “doubter.”
ABOUT GALLERI: The test is valid and extremely accurate — that is, extremely accurate for one of the two properties we need for early cancer detection using a blood test. Those 2 properties are SPECIFICITY and SENSITIVITY. In lay terms, Specificity is how well a test can identify true negatives through a low false-positive rate (confusing, I know, but hang in there). This is where Galleri shines. In medicine, Sensitivity and Specificity are often a teeter-totter — improve one side, and you lose on the other side. We like to see values above 90% for both Specificity and Sensitivity. Galleri boasts the remarkable Specificity of 99.5%! When you think about it, though, if you identify circulating tumor DNA, then there ought to be a cancer somewhere. So, in terms of “Positive Predictive Value” (PPV), a function of Specificity, Galleri is basically saying, “If you test positive, then cancer is present.” This is a luxury in screening because the great bugaboo in screening is the high “false-positive rate” that sends us scrambling on wild goose chases looking for what caused the “positive.” And then, there might be nothing, except continued anxiety for the patient. (Example: the callback and/or biopsy after a mammogram that turns out to be nothing.) With Galleri…virtually no false-positives!
But on the other side of the teeter-totter…things are not as bright. Sensitivity is how well a test can identify true positives through a low false-negative rate. Simply stated, Sensitivity is percentage of cancers that are detected by the screening tool. You can have very high Specificity and low Sensitivity, or vice versa, or both high, both low. Usually, the more the teeter-totter moves up, the other end moves down. And it’s the Sensitivity that is lacking in Galleri for many of the types of cancer being tested.
Across the board, for all tumor types, the Sensitivity is a rather low 51.5%. That is, Galleri is only picking up half the cancers. “That’s better than nothing,” you say. Yes, but that’s not the whole story. You see, there is another variable at work here — How early is the test picking up cancer? Are we talking about 51.5% Sensitivity for tiny breast tumors found on MRI? No. The 51.5% is for all stages of cancer, including advanced Stage 4 disease. So, breaking down into Stage at the time of detection, Galleri only detects 16.8% of Stage I cancers, the stage where most likely lives will be saved. Stage 4 — 90.1% Sensitivity, but that doesn’t help anyone. (That 16.8% is probably the most important number to remember in this post.)
These concepts are foreign to most, sometimes even confused by clinicians. So, let me explain by illustration the difference between Sensitivity and Specificity: Imagine 100 circles, 90 of them empty, and 10 of them colored in. The colored circles represent 10 patients with cancer, unknown to anyone, and no screening planned. Our job is to use a blood test to find those 10 cancers from the group of 100. We take samples of blood from all 100, and we have 5 positives, and all 5 turn out to have cancer. That gives us a Positive Predictive Value (derived from Specificity) of l00%. That is, when the test was positive, cancer was present 100% of the time. This is perfect Specificity at work. Great news that there were ZERO false-positives, so no wild good chases. However, notice that we missed 5 of the cancers. Enter a different blood test that has better Sensitivity, and in this case, positive results return for 9 of the 10 patients with cancer, for a Sensitivity of 90% (9 out of 10 colored circles identified). But the very properties that allowed the test to find 9 of 10 cancers spills over to the patients without cancer (those with empty circles). And, as it turns out, there were 20 positive results, with only the 9 having cancer. Thus, 11 patients get a false-positive prompting medical imaging or other testing trying to find the cancer. But there is no cancer. Even after a clean bill of health, the patient will likely worry. The false-positive rate is the great downside of screening for any type of cancer.
So, Galleri is in a bind as to how they market this test. Certainly, there is different Sensitivity for different types of tumors, with 18.2% for kidney cancers detected, but 93.5% for lung cancers detected. But then break down into Stage, and the lack of early detection is bothersome, the very thing for which Galleri is promoted. Breast cancer is one of the 50, but is not featured in marketing efforts. Why? The Sensitivity for Stage 1 disease is very low. In an effort to combat this result, the PR points out that early breast cancers must not leak very much tDNA, and this is probably indicative of a low-grade tumor that might be “overdiagnosed” through mammography, that is, a tumor that would never harm the patient. It might be true that only high-grade breast cancers leak their DNA early, but this is a long way from scientific corroboration. And when using such conjecture to cover a poor result, one is left with the unsettling interpretation — “With our test you won’t have to worry about overdiagnosis because our test doesn’t diagnose very many breast cancers anyway.”
Once upon a time, in order to introduce a screening modality, you had to prove a mortality reduction. This is why the U.S. Preventive Services Task Force won’t endorse much of anything when it comes to customizing breast cancer screening. But those days have softened by allowing surrogates, such as fewer interval cancers, fewer advanced stage cancers, etc. But what we have devolved to now is the mere detection of early cancer, the standard that many laypersons have held all along. But when it comes to early breast cancer, the tDNA tests (remember, there are other companies with similar results) are mostly finding Stage II and above, and even then, with lower sensitivities than other cancers. If the tumor cells don’t leak DNA, the cancer won’t be found by using blood.
Back to the overall Sensitivity of an unacceptably low 51.5% with Galleri…the PR has been presented cleverly to minimize this disconcerting number. Instead of the overall number, the Sensitivity figure is presented on the web site as 76%. And this would be good if it were for early-stage disease in all cancer types. But it depends on the type. It depends on the stage. And to show how the numbers game is played, the 76% has a qualifier — it applies only to those cancers which cause 2/3 of the cancer deaths. Okay. Fair enough. But those same cancers are more aggressive, and using the “conjecture approach” one could theorize these aggressive cancers are least likely to be affected by so-called early detection. Example: Does finding an unknown pancreatic cancer at 2.0cm yield a better outcome than a 4.0cm pancreatic cancer? Maybe. Or, maybe not. Pancreatic cancer often spreads very early in its course. This question is only settled through prospective, randomized screening trials that are decades away, if they happen at all.
Another maneuver in the numbers game is to focus on ACCURACY. “Accuracy” is not a lay term in this context. Accuracy is one of the 5 major performance characteristics of any screening test — Sensitivity, Specificity, Positive Predictive Value, Negative Predictive Value, and Accuracy (note: Galleri has the data for all cancer types, each type with 5 or more characteristics). Accuracy is the COMBINATION in formula form, of Sensitivity and Specificity, each providing half the weight. Thus, if a test is very strong in one or the other, it can “hold up” its sister concept and smooth things over. Never a better example than Galleri. An incredibly high Specificity holds up a very low Sensitivity, and we get an official Accuracy of 89%. Great! Only it’s misleading. PR knows that Accuracy is assumed to be Sensitivity by the lay public who really want to know: “If I have cancer, what are the chances it will be found by this test?” That’s not Accuracy the public wants to know…. it’s Sensitivity. Yet, when a layperson reads 89%, they would never guess that Galleri is only finding half of the cancers, and only 16.8% of Stage 1 cancers.
A blood test should be used in tailor made fashion based on its performance characteristics. High Specificity implies different utilization than High Sensitivity. This has been so confusing in my 30 years in this space that I wrote what might be the only “What if” article ever published on breast cancer screening and how a blood test fits into the picture. Not only do we have to deal with performance characteristics with a blood test for breast cancer, but also how does it fit into our current algorithms that are based on 3 different forms of imaging. (Reference: Hollingsworth et al, Modeling the Impact of a Screening Blood Test on the Use of Adjunct Breast Imaging. Breast J 2019; 25:1045-1046. doi: 10.1111/tbj.13397 )
And one final and very important point — Saving lives through screening (morality reduction) is based on a deceivingly simple formula: Vulnerable Biology + Sensitivity = Mortality Reduction. The cancer you are trying to find early must have a biology that lends itself to early detection. Some cancers are so indolent, they will be cured with or without screening. Others are so aggressive, they will not be cured even if found “early.” But those with a Vulnerable Biology lend themselves to early detection. That said, your screening tool must have good Sensitivity, that is, must find a large percentage of the cancers. And the greater that percentage…the greater the mortality reduction on the other side of the equation above.
NOTE: Specificity is NOT PART OF THE EQUATION. Specificity and Positive Predictive Value and Disease Prevalence are critical for screening feasibility, cost effectiveness, minimizing false-positives, etc. but they have NO impact on whether or not lives are saved. So, when I read this sentence from Galleri justification for its low Sensitivity numbers, I groan: A multicancer test like Galleri “with only moderate sensitivity but very high specificity and aggregate prevalence could detect more cancers than a single-cancer test with very high sensitivity but lower specificity and disease prevalence.”
I would love to see that last sentence translated to English, then run the numbers and show me how more lives will be saved by diagnosing fewer cancers. Again, the equation is remarkably simple — Sensitivity Alone dictates saved lives once a vulnerable biology has been established. For breast cancer, the “vulnerable” biology was established in the historical mammography screening trials. So, from here on, anything we do to improve sensitivity should save more lives. And for now, our best hope is a single cancer test that, when positive, tells the woman with normal mammograms that she should also have an ultrasound or MRI.
Ironically, my collaboration with 13 different biotech companies and academic institutions over the course of the past 30 years culminated in a large blood biobank at Mercy–OKC. Because we followed protocols to every detail, our samples were repeatedly noted as “superior” to commercial biobanks. But what really set us apart from the rest of the world was the fact that our blood was tied to MRI results. So, a “normal” was really normal. Word spread over the years such that our blood and our data was in great demand. And, after repeated failures in 3 clinical trials, we seem now to be on the verge of a blood test revolution. And the irony? Mercy–OKC shut down my research and my high-risk program two years ago, shortly after I was asked to play a leadership role for Syantra, introducing their test in the United States. Hopefully, we’ll get things going again in the U.S. with the emergence of a new breast center under construction now at SW 89th and I-44 in OKC.
ADDENDUM: It didn’t take long for someone to contact me and ask, “What about the other 49 types of cancer?” Here, it’s a different matter, especially if we’re talking about types of cancer for which there are no routine screening recommendations (as the company points out).
I think guidelines will emerge for “liquid biopsy” screening (and they’ll change as the technology improves), but it’s a personal decision to undergo this type of testing in the absence of a family history, with the understanding that a “negative” result does NOT assure one that cancer is NOT present. On the other hand, a “positive” result means cancer is very likely present (99.5%). Now, if a healthy person at baseline risk decides to do the test every year, or every other year, I think the benefit might become more evident over time.
However, for now, I think the test is a best used in those patients with a strong family history for various types of cancer where there are no current screening guidelines. We see this in our genetic testing program where families will have 3 or 4 or more types of cancer in the family, yet the genetic test is NEGATIVE. To me, this is the ideal use for the Galleri test.
As for those who test gene-POSITIVE for a cancer predisposition, it would depend on which organ systems are involved by the mutation. If breast cancer is the only site where a specific mutation increases risk, then there’s no use for Galleri if the patient complies with screening recommendations. But most of our gene-positive patients have more than one organ system for cancer predisposition. Again, if covered by current screening recommendations, then no need for Galleri. But if a genetic predisposition is identified, and the mutation involves a wide variety of possible cancer types, then Galleri is going to help, and we would make that recommendation (in fact, we would order the test for the patient as part of our high-risk program).