Slashdot Mirror

← Back to Stories (view on slashdot.org)

Gene Testing Often Gets It Wrong

Posted by samzenpus on from the these-are-not-the-genes-you-are-looking-for dept.

BarbaraHudson writes: ABC is reporting that gene tests for risk of specific diseases are not as accurate as we'd like to think, with different labs giving different interpretations. Over 400 gene variants that could help one make medical decisions regarding breast and ovarian cancer or heart disease have different interpretations from different labs according to the study. "The magnitude of this problem is bigger than most people thought," said Michael Watson, executive director of the American College of Medical Genetics and Genomics, one of the study's authors. Researchers caution consumers to be careful when choosing where to have a gene test done and acting on the results.

37 comments

  1. That's wonderful! by Anonymous Coward · · Score: 0

    So many choices! Just pick the service which yields the results you want to hear!

  2. jolie did by turkeydance · · Score: 1

    sorry 'bout that.

  3. Medical Decisions by Anonymous Coward · · Score: 1

    My mom found out from a geneticist that she was at risk for breast cancer, so she preemptively had a double mastectomy. My father found out he was at risk for heart disease, so he had a cardiectomy.

    1. Re:Medical Decisions by Anonymous Coward · · Score: 0

      And obviously you've had a brainectomy.

    2. Re:Medical Decisions by Anonymous Coward · · Score: 1

      Fire the geneticist. Your genes can't tell if you're at risk for any cancer. They can only tell if you'll have a greater risk than the average human. Acting on anything found by your genes at this point is just stupid at best.

    3. Re:Medical Decisions by godel_56 · · Score: 1

      Fire the geneticist. Your genes can't tell if you're at risk for any cancer. They can only tell if you'll have a greater risk than the average human. Acting on anything found by your genes at this point is just stupid at best.

      The major breast cancer genes give the patient somewhere around an 80% chance of getting breast cancer, which in my book is close enough to certainty when death is a possible outcome.

      However the associated chance of ovarian cancer is only around 50% which makes for a much harder choice for a younger woman, as having your ovaries removed would cause immediate menopause, with its own serious life changing effects.

  4. I wouldn't hire him, then by wonkey_monkey · · Score: 4, Funny

    Gene Testing Often Gets It Wrong

    Maybe Mr Testing should pay more attention to his work.

    --
    systemd is Roko's Basilisk.
  5. Bayes rule by cowdung · · Score: 4, Interesting

    If a 99% accurate test is true, but the probability of the condition is only 0.0001%, it is still highly improbable that the person is afflicted by the condition on the basis of the test alone. Its important to narrow down the population before any testing is effective.

    1. Re:Bayes rule by lars_stefan_axelsson · · Score: 2

      If a 99% accurate test is true, but the probability of the condition is only 0.0001%, it is still highly improbable that the person is afflicted by the condition on the basis of the test alone. Its important to narrow down the population before any testing is effective.

      Yepp. It's not for nothing that the first thing a doctor will tell you as the answer to the "What do I do now"-question that inevitably results from a positive test is: "Have more tests".

      Initial screening tests are often less accurate, since that inevitably makes them quicker and cheaper. That's why they're called screening tests. The odd positive results is just confirmation that better, slower, more expensive tests should be done.

      --
      Stefan Axelsson
    2. Re:Bayes rule by BarbaraHudson · · Score: 1
      If you had bothered to read the article, you would have found that in the 11% that were compared by more than 1 company, 17% came back with differing interpretations. That's alarming.

      11 percent have been analyzed by more than one lab so results can be compared. In 17 percent of those cases, labs interpreted the findings differently, as either raising the risk of a disease, having no effect on it or having an unknown effect.

      So what do you do - get 3 tests and let the majority inform your decision?

      --
      "Transparent" is a shit show that trades on every stereotype going. A man in drag is NOT a transsexual.
    3. Re:Bayes rule by Anonymous Coward · · Score: 0

      Isn't that how DNA works in the first place?

    4. Re:Bayes rule by Anonymous Coward · · Score: 1

      From that we have an estimate that a minimum of 17% gene variants are misclassified by at least one lab. Just because the different labs agreed does not make those classifications correct. Also just because a risk is raised does not mean the correct response is some preventative action (raised how much?), so the correct action could still have been taken. Or not.

      Another thing is that from that news article we have no idea how many actual tests this would invalidate, we only know the percent of gene variants that must be in error, not the number of diagnoses made based on those (depends on the prevalence of the variants).

    5. Re:Bayes rule by Anonymous Coward · · Score: 0

      From that we have an estimate that a minimum of 17% gene variants are misclassified by at least one lab. Just because the different labs agreed does not make those classifications correct. Also just because a risk is raised does not mean the correct response is some preventative action (raised how much?), so the correct action could still have been taken. Or not.

      Another thing is that from that news article we have no idea how many actual tests this would invalidate, we only know the percent of gene variants that must be in error, not the number of diagnoses made based on those (depends on the prevalence of the variants).

      we actually don't even know that (at least not from the news story; I haven't gotten around to the original NEJM article yet). It SOUNDS like they just compared ClinVar enteries. The problem is that this data was accumulated over time and old entries are frequently not updated.

      Consider the following example:
      A patient with an uncommon disease A which is known to be associated with mutations in a gene B (for the sake of argument, suppose that 10% of cases of disease A are thought to be caused by mutations in Gene B but that of patients with known pathogenic mutations in Gene B, >90% go on to develop Disease A) has his/her Gene B sequenced and that the purpose of this testing was to see if other family members are at markedly increased risk of Disease A. A variant in the gene B is found but it is not yet in ClinVar. The lab uses some bioinformatic tools to make its best guess as to pathogenicity but cannot confidently make a prediction so they report this as a variant of unknown significance (VUS). Subsequently, another (unrelated) patient with disease A has the same testing in the same lab and they find the same variant in Gene B again. The lab now has two affected patients from different families with the same variant and this time they declare the variant pathogenic and file with ClinVar. You would now conclude that the lab disagrees with itself if you just looked at the way the variants were classified. The truth is that, if you use ClinVar correctly, there is not ambiguity about the interpretation. Now imagine a 3rd patient in the same situation but s/he goes to a different lab. They find the same variant and go to ClinVar and realize that it is likely pathogenic and report it as such. According the analysis that led to the 17% discordance statistic quoted in the article, the first and second labs disagree. The truth of course, is that, at the time the second lab reported its findings, it was in agreement with the first labs most current interpretation.

  6. Not the testing, the interpretation. by louic · · Score: 1

    There is absolutely nothing wrong with gene testing. And it is not "gene testing" who gets it wrong. The second part of the sentence explains it all: "... with different labs giving different interpretations." It is some of the interpretations that are wrong or in other words: bad science by incompetent "researchers".

    1. Re:Not the testing, the interpretation. by Anonymous Coward · · Score: 0

      There is absolutely nothing wrong with gene testing. And it is not "gene testing" who gets it wrong. The second part of the sentence explains it all: "... with different labs giving different interpretations." It is some of the interpretations that are wrong or in other words: bad science by incompetent "researchers".

      When a "whoops" diagnosis of cancer results in hundreds of thousands of dollars paid to cancer-curing institutions, people might tend to question just how incompetent the work really is.

    2. Re: Not the testing, the interpretation. by pao93 · · Score: 1

      http://www.nejm.org/doi/full/1... The article this abc story is based on.

    3. Re:Not the testing, the interpretation. by BarbaraHudson · · Score: 1

      Gene testing results are open to interpretation. What to do with the results is also open to interpretation, by both doctors and the patients. Times change, improved processes will result in different interpretations based on better knowledge of the weight each gene variant should be given.

      --
      "Transparent" is a shit show that trades on every stereotype going. A man in drag is NOT a transsexual.
    4. Re:Not the testing, the interpretation. by Rich0 · · Score: 3, Informative

      Agree. It seems like a simple solution is to unbundle the testing and interpretation.

      This is really no different from any other area of testing. A lab can assay the creatinine in my blood, or the microalbumin in my urine, or the concentration of glucose in my blood. Those results are likely to be very accurate and reproducible unless the lab is just criminally negligent.

      What those results mean is an entirely different matter. A doctor will certainly utilize those results as well as the results of many other tests, history, interviewing the patient, and so on to make a diagnosis, and refine it as more data comes in.

      Just make the labs, well, labs. Now you can certify them far more objectively.

    5. Re:Not the testing, the interpretation. by AK+Marc · · Score: 1

      Gene sequencing is not open to interpretation, thus never gets it wrong. Gene testing is not science.

      --
      Learn to love Alaska
    6. Re:Not the testing, the interpretation. by Anonymous Coward · · Score: 0

      After how many decades, we still seem pretty thickheaded about accepting that one, though.
      'Cause, "I F***ing Love Science!" (translation: I don't care about science but I like to show off my intellect to those who agree with me politically, and be a pawn in massive propaganda schemes that I don't begin to understand!)

  7. Headline versus article by tomhath · · Score: 3, Insightful

    The article says more data will improve accuracy, and advocates collecting and analyzing more data. Seems pretty reasonable compared to the scary sounding "Often Gets It Wrong" headline..

    1. Re:Headline versus article by BarbaraHudson · · Score: 1
      And what is the impetus for collecting more data? It's not because the tests and their interpretations are giving accurate results all the time, that's for sure.

      An investigative TV news program here did their own study - sent samples from the same people to different labs, and got back different results. The tests do often get it wrong.

      --
      "Transparent" is a shit show that trades on every stereotype going. A man in drag is NOT a transsexual.
    2. Re:Headline versus article by tomhath · · Score: 1

      Every test and every interpretation has a margin of error; obviously a person should understand the error rate for both false positive and false negative is before making a decision. The headline simply said "often gets it wrong" but I don't see any qualification of what "often" means. One in ten? One in a million? Who knows? Although they do state that some labs are more competent than others; no surprise there.

    3. Re:Headline versus article by BarbaraHudson · · Score: 1

      Read the article. It gives the numbers, as well as other info :-) Or if you don't feel like it, 17% of the tests that were run by different labs came back with different interpretations. So, about 1 in 6.

      --
      "Transparent" is a shit show that trades on every stereotype going. A man in drag is NOT a transsexual.
  8. A lot of people don't understand false positive by Anonymous Coward · · Score: 0

    The false positive rates, particular in comparison to true positive rates, for some of these tests can be quite high. In a study of the NonInvasive Prenatal Test (NIPT) they found that 2-3 of every 10 positive result of a particular - and particularly grim - genetic abnormality is a false positive. While NIPT is a different test in general, the underlying technology is next-generation sequencing and is likely shared with these gene tests as well.

  9. On the plus side... by Flavianoep · · Score: 1

    Every European is related to Charlemagne

    --
    Linux is for people who don't mind RTFM.
    1. Re:On the plus side... by wonkey_monkey · · Score: 1

      So is every single other human being on the planet.

      If you're talking about being descended from Charlemagne, it's only exceedingly (if not astronomically) likely - it's not an absolute certainty.

      --
      systemd is Roko's Basilisk.
    2. Re:On the plus side... by Flavianoep · · Score: 1

      It costs in excess of US$ 99 to be tested and be assured of being of Charlemagne ancestry, or you can stay with the exceeding likeliness for free. I'd choose free if I was white.

      --
      Linux is for people who don't mind RTFM.
  10. Which tests were more or less accurate? by generic_screenname · · Score: 1

    The article mentions a $250 cheek swab test for breast cancer risk. These tests are being sold to all sorts of doctors, and interpretation is included. Is this test more or less accurate than blood testing done for a BRCAplus panel done by a geneticist? Are *all* genetic test results prone to this kind of error in interpretation, or just the cheap commodity ones? Which ones are better? Is the test itself inaccurate, the interpretation, or both? Vague article is vague. Scaring people without actionable data is irresponsible and cruel, especially when we're talking about people who are concerned enough about these cancers to have sought out testing for risk.

    1. Re:Which tests were more or less accurate? by BarbaraHudson · · Score: 1
      The article says it's the interpretation. So does the summary. You can't say it's a vague article if you didn't read it. The actual article gives the hard numbers.

      This isn't the first time either - up here investigative tv journalists sent samples from each test subject to several different labs and the companies refused to be interviewed to explain why the results came back different depending on which company did the test.

      --
      "Transparent" is a shit show that trades on every stereotype going. A man in drag is NOT a transsexual.
  11. Testing for "risk" is risky by jbmartin6 · · Score: 1

    I am sure gene tests for conditions that are 100% due to genes are accurate. But how does one determine if an estimate of risk is accurate or not? If the lab tells me I have a 10% chance of getting cancer, and I do, that doesn't establish anything unless I can live the same life repeatedly. The article doesn't say anything about tests being 'often wrong' instead it is about different labs judging risk differently, and the need to share information to try to narrow that variance.

    --
    This posting is provided 'AS IS' without warranty of any kind, implied or otherwise.
  12. Skin in the game by mattwarden · · Score: 1

    Call me when someone is offering genetic testing with some consequence to them if they get it wrong. Until then, this should be treated as if it were a scam.

  13. But science KNOWS humans are causing global warm^H by PrimeWaveZ · · Score: 0

    Every day we read about how science doesn't have it's shit figured out as much as we thought, but damnit, CLIMATE CHANGE IS HUMAN CAUSED 100%.

    Never mind that giant ball of (deadly) nuclear waste in the middle of our solar system. (remember, nuclear power is bad, kids)

  14. Money by Tokolosh · · Score: 3, Interesting

    This is about the big-dog incumbents (Myriad Genetics, Quest Diagnostics), who charge upwards of $4,000 for testing. They are concerned to keep out small companies that are trying to disrupt the business with $250 testing. With modern technology there is no reason it should cost any more, including interpretation.

    The tactics are patents, lawsuits, lobbying and FUD.

    https://en.wikipedia.org/wiki/...
    https://www.aclu.org/cases/ass...

    --
    Prove anything by multiplying Huge Number times Tiny Number
  15. Headline grabbing Hype by Anonymous Coward · · Score: 0

    The intepretation of the Alphabet and the Digital number system are quite open to "interpretation" and "human error" too.. lol

  16. Re:Now he tells me by davester666 · · Score: 2

    no big. you weren't using it anyway.

    --
    Sleep your way to a whiter smile...date a dentist!