Four years is 47 months, or 1,460 days, or 35,040 hours
July 25, 2020 10:32 AM   Subscribe

Experimental Blood Test Detects Cancer up to Four Years before Symptoms Appear

[Researchers] focused on developing a test for five of the most common types of cancer: stomach, esophageal, colorectal, lung and liver malignancies. The test they developed, called PanSeer, detects methylation patterns in which a chemical group is added to DNA to alter genetic activity. Past studies have shown that abnormal methylation can signal various types of cancer, including pancreatic and colon cancer.

The PanSeer test works by isolating DNA from a blood sample and measuring DNA methylation at 500 locations previously identified as having the greatest chance of signaling the presence of cancer. A machine-learning algorithm compiles the findings into a single score that indicates a person’s likelihood of having the disease. The researchers tested blood samples from 191 participants who eventually developed cancer, paired with the same number of matching healthy individuals. They were able to detect cancer up to four years before symptoms appeared with roughly 90 percent accuracy and a 5 percent false-positive rate. (Scientific American, July 21, 2020)

New Blood Test Detects 5 Types of Cancer Years Before Standard Diagnosis (ScienceAlert, July 22, 2020) A blood test has been shown to detect five types of cancer years before the diseases could be spotted using conventional diagnostic methods, according to a study published Tuesday.

"The immediate focus is to test people at higher risk, based on family history, age or other known risk factors," said co-author Kun Zhang, head of the bioengineering department at the University of California San Diego and an equity holder in Singlera Genomics, which developed the test. [...] More large-scale studies across long time periods are needed to confirm the potential of the test for early cancer detection, the authors cautioned. Half of the 32 authors either work for Singlera Genomics, are co-inventors of patents related to the test or hold equity in the company.

Non-invasive early detection of cancer four years before conventional diagnosis using a blood test (Nature.com, July 21, 2020) In the Taizhou Longitudinal Study (TZL), 123,115 healthy subjects provided plasma samples for long-term storage and were then monitored for cancer occurrence. Here we report the preliminary results of PanSeer, a noninvasive blood test based on circulating tumor DNA methylation, on TZL plasma samples from 605 asymptomatic individuals, 191 of whom were later diagnosed with stomach, esophageal, colorectal, lung or liver cancer within four years of blood draw. We also assay plasma samples from an additional 223 cancer patients, plus 200 primary tumor and normal tissues.

We show that PanSeer detects five common types of cancer in 88% (95% CI: 80–93%) of post-diagnosis patients with a specificity of 96% (95% CI: 93–98%), We also demonstrate that PanSeer detects cancer in 95% (95% CI: 89–98%) of asymptomatic individuals who were later diagnosed, though future longitudinal studies are required to confirm this result. These results demonstrate that cancer can be non-invasively detected up to four years before current standard of care.
posted by Iris Gambol (33 comments total) 27 users marked this as a favorite
 
The big question is, what can anyone do with that information?
posted by grumpybear69 at 10:35 AM on July 25, 2020 [10 favorites]


I mean, insurance companies could just outright cancel your coverage knowing you're about to get _real_ expensive.

Too cynical? I suppose the optimistic answer is, well, now we know you have a small colony of cancerous cells in your stomach, esophagus, colon, lung or liver, let's see if we can find it and kill it before it becomes more medically challenging?
posted by Kyol at 10:40 AM on July 25, 2020 [27 favorites]


From Nature.com link, study intro: Late stage cancers often lack an effective treatment option (1,2). Survival rates increase significantly when cancer is identified at early stages, as the tumor can be surgically removed or treated with milder drug regimens (3); average 5-year survival at early stage is 91%, while average 5-year survival at late stage is 26% (4). Detection of tumors at the earliest possible stage is therefore of paramount importance for cancer treatment.
posted by Iris Gambol at 10:40 AM on July 25, 2020 [26 favorites]


This seems like a very hopeful development, thank you for posting it!
posted by LobsterMitten at 10:43 AM on July 25, 2020 [8 favorites]


I mean, insurance companies could just outright cancel your coverage knowing you're about to get _real_ expensive.

Not since Obamacare, in the U.S. That's what it was like before Obamacare. Now, if Republicans are successful in repealing Obamacare, as they have been trying for 10 years and are still trying, then all bets are off.
posted by JackFlash at 10:45 AM on July 25, 2020 [23 favorites]


The big question is, what can anyone do with that information?

Some of these cancers can be very difficult to notice and don't get treated until they're at late or end stage. Pancreatic cancer especially is one of those ones that's most frequently detected at stage IV because it's usually clinically silent until it's ravaging everything.

If the levels of these tests are showing a possible cancer in one of these hard to detect places they can look harder for it with more detailed scans of an area, biopsies, and other diagnostic measures. Instead of having to fight a metastasized monster it may be possible to go for quick curatives with a minimal round of chemo or a smaller -ectomy.
posted by Your Childhood Pet Rock at 10:46 AM on July 25, 2020 [23 favorites]


This makes a lot of sense... Cancer growth is an exponential process, and exponential processes tend to have a bit of an elbow. And 'small, yet doubling' behavior prior to the elbow that you could presumably detect and deal with earlier - similar to, say, dealing with a pandemic.

I think the biggest danger with better cancer detection is overtreatment... Understanding the false positive rate (and benign positive rate) is key. But hopefully early detection can come with milder treatments.
posted by kaibutsu at 10:48 AM on July 25, 2020 [5 favorites]


I think the biggest danger with better cancer detection is overtreatment... Understanding the false positive rate (and benign positive rate) is key. But hopefully early detection can come with milder treatments.

Well that's one of the problem with full body scans for diagnostics. Every irregularity becomes a possible "is this cancer?" moment. If we can be more certain of where the probable problem areas are there's going to be a lot less overtreatment going after red herrings.
posted by Your Childhood Pet Rock at 10:54 AM on July 25, 2020 [4 favorites]


I guess I'd like to see these results replicated 10-20 times in other studies before putting any faith in the process. Misdiagnosis of stuff like esophageal cancer, which I have witnessed firsthand, can result in a severe quality of life degradation. There is no such thing as a happy post-esophagectomy patient. And count me among the people who think this will be used to justify decreased insurance coverage.

M4A now.
posted by grumpybear69 at 10:59 AM on July 25, 2020 [9 favorites]


Well that's one of the problem with full body scans for diagnostics. Every irregularity becomes a possible "is this cancer?" moment.

Yeah, but... IIRC, the body is constantly fighting off tiny cancers. It's easy to imagine this method finding stuff that the body eventually takes care of anyway, without intervention. And, indeed, they report a 96% specificity (true negative rate), implying a 4% false positive rate.

Suppose we get 88% True Positive rate (sensitivity) and 96% True Negative rate (specificity) on pancreatic cancer, the numbers reported in the paper. A quick google search says that ~60k people per year are diagnosed with it in the US, out of 210M adults. So, if we test everyone in the US, you catch ~53k of the 60k cases, but also trigger a false positive on 8.4 million people (4% of 210M). At which point you are happy to help the 53k, but also have to think hard about minimizing impact on the 8.4 million.
posted by kaibutsu at 11:18 AM on July 25, 2020 [17 favorites]


you catch ~53k of the 60k cases, but also trigger a false positive on 8.4 million people

So if the test is positive, you still only have a 0.6% chance of actually having pancreatic cancer.
posted by judgement day at 11:28 AM on July 25, 2020 [2 favorites]


Well there's going to be further attenuation of false positives as they go through other diagnostic pathways. If they find methylation but not a mass then obviously it's not going to be big enough to worry about treatment and it's one of those "keep an eye on it every 6-12 months" things. If they find a mass it goes to biopsy, and so on.

There's going to be false positives but as long as it's not the only tool in the toolbox it can be managed. Hell, if we get people with strong colon cancer markers but find nothing in a colonoscopy there might be time for intervention to change diet or other factors to minimize the future occurrence of precancerous polyps.
posted by Your Childhood Pet Rock at 11:30 AM on July 25, 2020 [19 favorites]


Yeah, YCPR, I was thinking something similar. Kind of like how I as a woman with abnormal cells from an hpv infection have yearly to check the status, and have had two colposcopies in the past two years to make sure it’s not gotten worse. I mean I’d rather it go away, but right now it’s just watch and wait.

I would imagine that depending on the point at when some very early cancer was caught, it would be “let’s observe and increase the frequency of observing” vs going straight to “nuke it from orbit”.
posted by [insert clever name here] at 11:36 AM on July 25, 2020 [4 favorites]


The big question is, what can anyone do with that information?

I work on a project that takes DNA methylation patterns from multiple biological samples, using some information theory to reduce them to comparable logos, similar to what people do with DNA sequence logos.

This of interest to me, as there may be utility in comparing methylation patterns of hundreds of known healthy and cancerous tissues, which our lab and other research consortia already generate, against samples drawn from patients, correlating that with variant and other datasets on hand. This could lead to refinements in what we do, answer other research questions, or assist in clinical applications.
posted by They sucked his brains out! at 11:41 AM on July 25, 2020 [7 favorites]


This appears to have been a retrospective analysis--please correct me if I'm wrong?

The next test is a long term study that uses it to actually predict cancer and see if it holds up.

Yeah, but... IIRC, the body is constantly fighting off tiny cancers. It's easy to imagine this method finding stuff that the body eventually takes care of anyway, without intervention.

Well, the endpoint was cancers that the body did *not* take care of--cancers that were detected by conventional means in their sample of 100,000+ people. There would have been more false positives if a significant number of the finds were cancers that would be spontaneously cured--they would have come up as failures, not successes.

So, if we test everyone in the US

They suggest as a possible utility only thinking about high risk populations, at least until there is some orthogonal diagnostic to rule out false positives.
posted by mark k at 11:54 AM on July 25, 2020 [1 favorite]


They suggest as a possible utility only thinking about high risk populations, at least until there is some orthogonal diagnostic to rule out false positives.

That kind of negates the use of detecting typically asymptomatic cancers though. One would hope a technique like this could reach the utility to be built into a screening panel conducted during the typical physical.
posted by Your Childhood Pet Rock at 11:58 AM on July 25, 2020


The big question is, what can anyone do with that information?

If you've ever watched somebody go through chemo it's pleasant to imagine a future where a lot less chemo is necessary to eliminate a strain of cancer a lot earlier.
posted by mhoye at 12:03 PM on July 25, 2020 [14 favorites]


you catch ~53k of the 60k cases, but also trigger a false positive on 8.4 million people

So if the test is positive, you still only have a 0.6% chance of actually having pancreatic cancer.


Yes, but that still represents a 21,000 fold increase in your chance of getting pancreatic cancer compared to the general population. If someone told you you had a greater than 1 in 200 chance of getting an incurable and deadly disease in the next 4 years that is useful information to have. Based on this information, you might be willing to undergo imaging every 6 months to catch it early for example. This is what screening tests are supposed to do--identify groups of people in whom additional and otherwise burdensome investigations may be justified and cost-effective.
posted by reformedjerk at 12:06 PM on July 25, 2020 [11 favorites]


If you've ever watched somebody go through chemo it's pleasant to imagine a future where a lot less chemo is necessary to eliminate a strain of cancer a lot earlier.

Some of these cancers are also of the kind that have very poor survival curves, so a method that can help with early diagnosis will almost surely improve odds, if not give people more time.
posted by They sucked his brains out! at 12:50 PM on July 25, 2020 [6 favorites]


Beloved bevy of beauties, pancreatic cancer is not one of the cancers addressed by the PanSeer test. PanSeer detects stomach, esophageal, colorectal, lung and liver malignancies.

"More large-scale studies across long time periods are needed to confirm the potential of the test for early cancer detection, the authors cautioned."

I miss the people I've lost to cancer, and worry about the ones who survived. Some of them will always be debilitated, courtesy of the disease and the painful treatments which saved their lives; they are terribly vulnerable to new illnesses. I'm taking my excitement over the study news elsewhere, while sincerely wishing all of you a pleasant weekend.
posted by Iris Gambol at 1:26 PM on July 25, 2020 [4 favorites]


There’s certainly a lot to be said for early diagnosis of cancer, but it is easy to overstate the importance of it. Part of the reason is that so much importance is placed on 5 year survival. Even without a corresponding improvement in treatment, earlier diagnosis means more patients will reach that 5 year mark; they just learned their diagnosis sooner than they would have previously. So this is certainly a positive development, but needs to be looked at in context.
posted by TedW at 1:33 PM on July 25, 2020 [4 favorites]


Beloved bevy of beauties, pancreatic cancer is not one of the cancers addressed by the PanSeer test.

My apologies... just heard news this morning that a friend of a friend has a pancreatic cancer diagnosis, and must have gotten my wires crossed (maybe additionally influenced by the name of the test). My bad. As penance, lemme run the numbers with the actual cancers targeted by the test.

Re-running my loose google-search-wth-rounding methodology, we get: Colorectal: ~150k, Esophogal: ~20k, Stomach: ~25k, Lung: ~250k, Liver ~50k, totaling almost 500k diagnoses per year. (FWIW, that's terrifying. I am terrified.) So we would see ~440k caught earlier, which is amazing, and still see about the same 8.36M false positives. The probability of a true positive given a positive diagnosis is about 5%.

All of the caveats about taking a careful response and adding extra observation is completely correct and exactly in line with the point I was making initially. It looks like a great result, but it's probably best for the scientists and doctors who work with this stuff full time to figure out how it fits into the overall 'user journey.' And not worry tooooo much yet about losing all of our medical coverage in the event of a result that is more likely than not a false positive.
posted by kaibutsu at 2:04 PM on July 25, 2020 [2 favorites]


The probability of a true positive given a positive diagnosis is about 5%.

Huh? Can you please elaborate on which cancer diagnoses you are referring to here, and how you arrived at the 5% number you cited?
posted by lazaruslong at 4:49 PM on July 25, 2020


Pretty sure early detection leading to better (more survival) outcomes is well established for most cancer types? Way to derail early with concern trolling :-/
posted by SaltySalticid at 6:23 PM on July 25, 2020


Huh? Can you please elaborate on which cancer diagnoses you are referring to here, and how you arrived at the 5% number you cited?

Probability of a true positive given a positive result is TP / (TP+FP). So (again, extrapolating a hypothetical all adult USians, with tp/fp rates as stated, I have a salt mine for sale), we have TP=440k, and FP=8360k. So 440/8800 = 0.045, which I rounded up to 5%.
posted by kaibutsu at 7:00 PM on July 25, 2020


Pretty sure early detection leading to better (more survival) outcomes is well established for most cancer types? Way to derail early with concern trolling :-/

This is not concern trolling and medical types are developing some real uncertainty about the usefulness of frequent, routine, suspicion-free testing for some types of cancer and some kinds of screening.

Consider breast cancer, only because I happen to have old numbers about it for undergraduate statistics classes. With at least old numbers, if you test 10000 women you will correctly state that 8950 don't have breast cancer, you will find 80 cancers, you will miss 20 cancers, and -- here is the part where the questions come in -- you will tell 950 cancer-free women that they might have cancer.

This is a problem because some few of those women are going to have cardiac events from the stress until more discriminatory tests can clear them.

This is a problem because some of those women are going to have either lumpectomies or full mastectomies that they don't need. Even beyond being disfigured for no reason, some few of those women are going to die on the table. Some few of them are going to get resistant MRSA or other hospital-borne pathogen and die, or die in some other way that's causally linked to their unnecessary surgery.

With some kinds of screening and some kinds of cancer, it's legitimately uncertain whether frequent, routine, suspicion-free testing would, net, save lives because there is a real danger of killing more cancer-free people than it can find early and treatable cancers.
posted by GCU Sweet and Full of Grace at 9:31 PM on July 25, 2020 [8 favorites]


"... (3); average 5-year survival at early stage is 91%, while average 5-year survival at late stage is 26% (4)..."

As TedW points out, an improvement in the five-year survival rate is a mathematical certainty. Let's say there's a cancer which, on average, takes two years to kill people without treatment. Moving the detection date to a year earlier in time, which starts that five-year survival rate clock a year earlier, will necessarily improve the five-year survival rate without changing anything at all (assuming no treatment). It's the definition of the statistic and math.

Certainly, as a general rule, treating a cancer earlier is more effective. But that's a generalization — some cancers just don't have effective treatments but they're treated anyway, and many cancers are only moderately slowed down by treatment. In all those cases, earlier detection will mathematically improve two- or five-year survival rates, but have little or no effect on the outcome.

This is important because sometimes, as GCU SaFoG just wrote, the earlier detection will result in earlier treatment but little or no change in outcome, and because cancer treatment is what it is, that ends up significantly increasing a patient's amount of time spent immiserated. And all treatment, especially surgery, has some fatality risk. In some cases, getting an earlier diagnosis can cause a worse outcome because cancer treatment is so aggressive.

This is something that it's becoming more important for people to understand. Especially with cancers, diagnostic tools are going to improve more quickly than therapeutics, just because diagnosis is easier because treatment is still such a difficult problem. And keep in mind we're not just talking about better diagnostics, this also applies to increased screening. This is why, for example, full-body MRI scans can be a net negative. It's why universal screening or just lowering the recommended age for a given cancer screen can do more harm than good.
posted by Ivan Fyodorovich at 4:33 AM on July 26, 2020 [2 favorites]


With some kinds of screening and some kinds of cancer, it's legitimately uncertain whether frequent, routine, suspicion-free testing would, net, save lives because there is a real danger of killing more cancer-free people than it can find early and treatable cancers.

This is a genuine problem with some cervical screening regimens. The folk narrative around cervical screening is that any abnormality found would definitely or probably turn into cancer. This is not true, most would not turn into cancer and would not cause problems. Your lifetime risk of finding abnormalities following cervical screening can be 50% under some regimens. Your lifetime risk of cervical cancer is less than 2%. At a population level, it probably makes sense to carry out cervical screening (nearly 2% is a lot, and early treatment prevents cancer developing at all), but testing at the right frequency makes a big difference in reducing the number of effective 'false positives' while minimising the health impact of false negatives.
posted by plonkee at 4:37 AM on July 26, 2020 [2 favorites]


It is good to read the examples above of early detection being followed by watchful waiting. That isn't a universal reaction though. Just in my own personal orbit, I know several people who received an early cancer diagnosis and whose response was "get this cancer out of me now!"

Several years later now for each of them, with a bit more time to have processed the whole experience and learn more about their particular cancers, as well as having undergone the rigors of treatment, several of them say they should not opted for such immediate treatment.
posted by PhineasGage at 8:06 AM on July 26, 2020 [1 favorite]


This is a problem because some few of those women are going to have cardiac events from the stress until more discriminatory tests can clear them.

That sounds to me more the fault of the way women are told of test results than the test itself. You don't tell them "This test says you might have cancer!!!", you tell them, "Hm. This test is a little off. Let's keep an eye on this in the future, it's probably nothing."
posted by JHarris at 1:16 PM on July 26, 2020


you tell them, "Hm. This test is a little off. Let's keep an eye on this in the future, it's probably nothing."

That's lying to your patient. You don't prescribe a test, get the results and then after the fact play it all down with happy talk.

If the test literally "probably means nothing", then why would you even prescribe it? Or else "probably means nothing" actually means "maybe something." One way or the other, you aren't being honest with your patient.

A doctor should have a detailed honest discussion with the patient before the test, describing why or why not to have it, and what the detailed statistics mean. And if the well-informed patient decides to proceed, you have the same detailed statistical discussion of the results. You don't just blow it off with "probably means nothing."
posted by JackFlash at 3:23 PM on July 26, 2020


Up above it was being said, "The probability of a true positive given a positive diagnosis is about 5%." That sounds like it's probably nothing to me.
posted by JHarris at 7:26 PM on July 26, 2020


I had a false positive test for a serious disease (not cancer and not fatal, but uncurable.) It was quite traumatic and took 3-6 months to get a doctor to look at everything and say "you're probably OK". I'm fairly educated on biochemistry and realized early the initial test should not have been done (because I was at high risk of a false positive). I seriously think the doctor who gave it to me was delinquent, unprofessional and lazy. But it turns out that I couldn't just mentally ignore the result on go on about my daily life normally--even if the initial doctor had said to do that, which they did not.

A five percent chance of having a serious cancer is absolutely something. No doctor would be blase about it and it would be a rare and arguably irrational person who wouldn't care--the odds of dodging the bullet are better than in Russian roulette, but not much better.

Tests for mass screening need almost perfect accuracy--like 99.99% or more--or some way to follow up quickly and cost effectively. This test has neither and the authors don't even discuss it in the context of screening. They talk about using it in high risk populations needing additional tests.

This actually *is* a huge scientific find (assuming it holds up) in a field that's seen immense activity with IMHO pretty limited success. It's the sort of thing a lot of people have been suggesting biomarkers could accomplish but have struggled to deliver. It's wholly unsuited for some of the suggestions people are making for it though, and those applications are several breakthroughs away.
posted by mark k at 8:36 PM on July 26, 2020 [5 favorites]


« Older The poet had just not considered that there are...   |   Median US Income is Three Metres a Year Newer »


This thread has been archived and is closed to new comments