Scientists Develop 10-Minute Universal Cancer Test (theguardian.com)

An anonymous reader quotes a report from The Guardian: Scientists have developed a universal cancer test that can detect traces of the disease in a patient's bloodstream. The cheap and simple test uses a color-changing fluid to reveal the presence of malignant cells anywhere in the body and provides results in less than 10 minutes. The test has a sensitivity of about 90%, meaning it would detect about 90 in 100 cases of cancer. It would serve as an initial check for cancer, with doctors following up positive results with more focused investigations. The test was made possible by the Queensland team's discovery that cancer DNA and normal DNA stick to metal surfaces in markedly different ways. This allowed them to develop a test that distinguishes between healthy cells and cancerous ones, even from the tiny traces of DNA that find their way into the bloodstream.

Healthy cells ensure they function properly by patterning their DNA with molecules called methyl groups. These work like volume controls, silencing genes that are not needed and turning up others that are. In cancer cells, this patterning is hijacked so that only genes that help the cancer grow are switched on. While the DNA inside normal cells has methyl groups dotted all over it, the DNA inside cancer cells is largely bare, with methyl groups found only in small clusters at specific locations. Writing in the journal Nature Communications, the Queensland team described a series of tests that confirmed the telltale pattern of methyl groups in breast, prostate and colorectal cancer as well as lymphoma. They then showed that the patterns had a dramatic impact on the DNA's chemistry, making normal and cancer DNA behave very differently in water. The suspect DNA is added to water containing tiny gold nanoparticles, which turn the water pink. "If DNA from cancer cells is then added, it sticks to the nanoparticles in such a way that the water retains its original color," The Guardian reports. "But if DNA from healthy cells is added, the DNA binds to the particles differently, and turns the water blue."

Comments

how many false positives?

[140Mandak262Jamuna]

It looks like the false negative rate is 10%. any number on false positives?

Badly Oversold

[Anonymous Coward]

Disclosure: I'm a cancer biomarker researcher, and these types of studies are my bread & butter. I don't know this group, and they chose to look at tumour types that I'm largely not working on.

So this study is pretty uninteresting, and is getting a lot of unhelpful media attention. The core observation is that there are differences in methylation in tumour and normal, and that these can be detected in a pretty simple assay (10 min, blah blah). That's all either known, or pretty simple extensions of existing work. Nothing wrong with it, just not hype-worthy.

Media is then claiming an AUC of ~0.90 (and an operating point with an accuracy of ~85%). The problem is, their test situation is entirely irrelevant. Most of these analyses were based on a comparison of blood from:
1) healthy age- and sex-matched controls
2) patients with metastatic disease (cancer that has spread throughout the body)

This has a litany of problems:
1) almost all patients are symptomatic pre-metastasis, and thus only a small fraction (~5-25%) of cancers are diagnosed at this stage
2) patients diagnosed with metastatic disease have often elected to avoid testing (cost, access, personal decision, etc.)
3) sadly almost all metastatic disease is lethal -- we routinely cure patients with localized disease with surgery or radiotherapy
4) genomic and epigenomic changes accumulate over time, and metastatic tumours on average have significantly more
5) there is no health-care economics argument for screening for metastatic disease

So essentially, the paper says "we can distinguish black from white with 90% accuracy". That's fine, but the media reports are missing the fact that ~95% of real-life cases are gray, and the accuracy of this test will probably be lower in these white vs. gray comparisons. The likelihood is that this "90% accurate test" is actually going to be ~70-80% accurate in real-world settings. Which is fine, but, you know, matches existing cheap diagnostics in most tumour types anyways!

So in short, this study is over-hyped and goes far to creating a bad culture where people think we are closer to a "cure" than we really are. It's solid science, but in no way worthy of a slashdot article!

Re:Badly Oversold

[Anonymous Coward]

Well, my estimation is that there are probably 10-20 cancer biomarker papers of superior quality published each week. This paper won't be in the top 500 papers in its field this year. I personally believe Slashdot should do better, and highlight high-quality science, not mid-quality work. That of course is just an opinion! :)

Re:False positives?

[Luckyo]

In case of cancer, that's wrong. Many cancers are fairly harmless, while procedures to find them and put them into remission are extremely damaging, far more so than cancer itself.

A good example is the modern findings on prostate cancer, which in many cases is much less harmful than measures that clinicians used to put it into remission. Nowadays, certain prostate cancers aren't treated at all and instead merely monitored for example, and patients are likely to live with minimal to no symptoms for decades. Whereas treatment would cause severe symptoms immediately and to an extent permanently.

So in many cases, as shocking as it may sound, it's actually better in terms of health outcome to the patient to get a false negative on cancer test than an accurate positive one. Especially true if it's a generic case like this, where someone may go look for cancer that ends up as a small benign tumour with a series of exceedingly invasive biopsies that may cause severe damage to patient's health.

Oncology is really, REALLY difficult field of medicine, because not only is the illness effectively incurable in most cases, but oncologist must always contrast the harm caused by his actions vs harm caused by the problem he's looking for or trying to address.