How Useful is Personal Genomics? A Case Study

How much can you help yourself by getting your genome sequenced? A lot, a little, not at all? Scenario 1 (big help): You discover you have a greatly elevated risk of Disease X. You do various things to reduce that risk that actually reduce it. Scenario 2: (a little help): You discover you have a greatly elevated risk of Rare Disease X. You do various things to reduce that risk but they don’t help. At least, when Disease X starts, you will be less upset. Scenario 3 (no help): You discover that you have a greatly elevated risk for a common easily-noticed disease (such as obesity). You already watched your weight, this changes nothing. Scenario 4 (harm): You discover that you have a greatly elevated risk of Scary Disease X (e.g., bipolar disorder). It is depressing news. Later studies show that the gene/disease association was a mistake. (Many gene/disease associations have failed to replicate.)

A recent Wired article tries to answer this question for one person: Raymond McCauley, a bioinformatics scientist who had his genome sequenced four years ago and learned he was “four or five times more likely than most people to develop age-related macular degeneration (AMD)”. The article says “of all the ailments described in the 23andme profile, AMD has one of the strongest genetic associations”. If I found this in my genetic profile, I would want to know the confidence interval of the increased risk. Is it a factor of 4.5 plus or minus 1? Or 4.5 plus or minus 8? This isn’t easy to figure out. In addition to the question of variability, there can easily be bias (= estimate is too high). Let’s say I do 100 gene/disease association studies. Then I scan these studies to pick the one with the strongest gene/disease association. It should be obvious that this particular association is likely to be too high and, depending on the details, could plausibly be pure chance (i.e., true association is zero).  I have been unable to find out how replicable the gene/AMD association is. According to Wikipedia, “the lifetime risk of developing late-stage macular degeneration is 50% for people that have a relative with macular degeneration, versus 12% for people that do not have relatives with macular degeneration.” (Until it was eliminated via better diet, pellagra also ran in families.) The Wired article does not say whether any of McCauley’s relatives have/had AMD — a huge omission, given the uncertainty of gene/disease associations.

It wasn’t obvious what McCauley should do, according to the article:

McCauley read that there were a few preventative measures he could take to reduce the chances of AMD one day rendering him blind: don’t smoke and avoid ultraviolet light, for instance. Also, it seemed, he could try taking a special combination of vitamins, including B12 and lutein. But when he consulted the research, he could find little evidence to support the effectiveness of the regime, based on his genotype.

The article says nothing about quitting smoking but he does wear glasses that reduce ultraviolet light and takes certain vitamins. It is very hard for him to determine whether they help.

Here is a study that found greater omega-3 consumption associated with lower risk of AMD. Here is a study that found AMD associated with inflammation (too little omega-3 increases inflammation). Here is a study that found no association between vitamin and mineral intake and AMD. Based on this, if 23andme told me I had an increased risk of AMD, I would make sure to optimize my intake of flaxseed oil (or other omega-3 source) using some sort of brain test. I have documented in other posts that brain function is sensitive to omega-3 intake and (probably) most people don’t get enough. Of course, just as it is foolish to smoke (a lot) regardless of whether you have genetic risk of AMD, it is foolish to not optimize one’s omega-3 intake, whether or not you have genetic risk of AMD. In other words: everyone should optimize their omega-3 intake.  If the 23andme results cause McCauley to do something wise like this that he would otherwise not have done, they have helped him.

The omega-3 study appeared after the Wired article so I don’t know how McCauley reacted to it. A puzzle about the  story is that it isn’t even clear that the gene/AMD associations are true. Consider McCauley’s older relatives: parents, grandparents. Did/do any of them have AMD? If not, it is more plausible that all of them were at 12% risk of the disease than at 50% risk. Suppose all of them had, according to 23andme, the same increased risk as McCauley (at least some of them have the risk-bearing genes). Now it becomes more plausible that something is wrong with the 23andme risk estimate. If some of McCauley’s older relatives do have AMD, it is not clear why the 23andme results would make much difference. He should have already have known he was at increased risk of AMD.

The upshot is that in this particular case, I cannot even rule out Scenario 4 (does harm). All four scenarios strike me as plausible.  Based on this article, we are a long way from learning the value of personal genomics.

Previously I used the example of Aaron Blaisdell to make the possibly counter-intuitive point that if you have a genetic disease something is wrong with your environment. Well, I do not have any obvious genetic disease. But I discovered, via self-experimentation, that my environment was terrible — meaning it could be improved in all sorts of ways: stop eating breakfast, drink flaxseed oil, eat butter, look at faces in the morning, take Vitamin D in the morning, and so on, not to mention eat fermented foods (which I figured out via psychology, not self-experimentation). My findings about what is optimal are so different than the way anyone now lives (except people who read this blog) that I believe everyone‘s environment can be vastly improved. If so, the value of discovering you have a genetically elevated risk of this or that is not obvious — you should already be trying to improve your environment. At least that is what my data has taught me. On the other hand, maybe genetic info (even wrong genetic info!) will give you a kick in the pants. Maybe that has happened with McCauley.

 

12 Responses to “How Useful is Personal Genomics? A Case Study”

  1. gwern Says:

    > The upshot is that in this particular case, I cannot even rule out Scenario 4 (does harm). All four scenarios strike me as plausible. Based on this article, we are a long way from learning the value of personal genomics.

    The ultimate value could be as propaganda – fake information – like the evopsych theory of placebos. The value of information for a recommendation like ‘quitting smoking is 1% better for you than the average person’ is exactly 0: you should already be doing it! But the impressive circumstances – whoo *genomics* – may be more likely to make one do it.

    As far as paying voodoo priests to impress you with theatricality, $1-1000 isn’t that expensive.

  2. Greg Says:

    I have found 23andMe to be a very big help. It told me three years ago that I have elevated risk for Celiac Disease. I had never heard of it before. When I read about the symptoms, they described my health situation at the time. I had been suffering for years. Additional testing confirmed that I should not consume gluten. My symptoms resolved with a gluten free diet and I continue to be symptom free today.

  3. Tom Says:

    Off-topic, but any idea why women are feel cold when most males are comfortable?

    I’m sitting in a Starbucks in a short-sleeved shirt and the AC just switched on, and as a result I’m finally getting comfortable. Yet I noticed that every woman near me has just put on a sweater.

    It doesn’t make sense to me that the genders have always been this way. Any idea what might have changed in the modern environment to create this discrepancy?

    Seth: As a way of defending body fat. It is clearly more difficult for women to lose weight than for men to lose weight. Perhaps the evolutionary reason is to give women a greater chance of surviving a famine.

  4. Sheila Buff Says:

    Hi Seth,
    Risk factors for AMD are indeed well established from epidemiological studies, but for the personal component you would have to know family history, and not everyone does. If McCauley’s parents died young, for example, their genetic tendency for AMD wouldn’t have been expressed and he wouldn’t know he’s at risk. In such circumstances, having the genome info would be helpful, since some preventive measures, such as not smoking, can prevent or slow AMD. Solid research shows that the main candidate gene for AMD is HF1. Assuming that 23andme tested accurately for SNPs of HF1, McCauley’s increased risk is highly likely. In addition to not smoking, he should consider taking ocular supplements. The AREDS study showed that taking high levels of antioxidant vitamins (C, E, beta carotene) and zinc could reduce the risk of progression to advanced AMD by about 25%. The AREDS II study, which is expected to end in 2013, adds fish oil, lutein, and zeaxanthin to the formula.

    Seth: You write: “Assuming that 23andme tested accurately for SNPs of HF1, McCauley’s increased risk is highly likely.” I’m less sure. In the genome-wide association studies that found the association that 23andme relies on, a huge number of tests were done, looking for gene/disease associations. I have yet to read anything that convinces me that the researchers involved understand how to correct for the number of tests done. That leaves replication as the only test of accuracy and these gene/disease associations often don’t replicate. I believe the 23andme statement about increased risk could easily be wrong.

  5. Tom Says:

    Seth: As a way of defending body fat. It is clearly more difficult for women to lose weight than for men to lose weight. Perhaps the evolutionary reason is to give women a greater chance of surviving a famine.

    Interesting, and makes sense. I understand that, compared to men, women store a higher percentage of Omega 3 fatty acids (vis a vis Omega 6 fats), enabling them to continue to supply the developing fetal brain should food supplies falter during pregnancy.

  6. Nancy Lebovitz Says:

    Off topic: I’ve raised the question of whether there’s any evidence about the health effects of organically raised vs. conventional food, and turned up practically nothing. One anecdote about someone who ended their celiac problems by switching to organic food, and no experiments on multi-cellular animals.

    It occurred to me that folks here might be doing some self-experimentation on the subject, or know about institutional research.

  7. Kirk Says:

    @Tom,

    In the past few years I talked with two women (at separate times) who each said she wanted to move to a colder climate. Each woman was overweight, comfortable with her size, and was eating high-carbohydrate foods during the conversation.

    In terms of my own experience, I have felt excessively cold during several periods in the past ten years. The first was when I was trialing a low-carb way of eating. The second was when I attempted 16-hour true fasts.

    Based upon my experience, my conclusion is that the women you noticed were in the midst of a restricted-calorie, low-carbohydrate, or intermittent-fasting regime.

  8. LemmusLemmus Says:

    Nancy: I believe the British government published a report on that some months ago, finding essentially no health effect. There may have been a link to that study from the Marginal Revolution blog. Hope this helps somewhat.

  9. Kirk Says:

    @Nancy,

    I definitely noticed a huge difference between non-organic coffee and organic coffee. Regular coffee generates muscle knots in my upper back.

    Usually I buy organic apples, potatoes, and sweet potatoes, based upon the recommendations of several smart bloggers in the paleo field, but I have not noticed any difference.

  10. Evolutionarily Says:

    This appears relevant to the discussion: http://www.fooledbyrandomness.com/NEJM.pdf

    NEJM
    Whole-Genome Analysis
    In their article on a whole genome scan looking for SNPs associated with sporadic ALS,
    Dunckley et al. (NEJM August 3 2007 issue)
    1
    consider finding ten SNPs that replicate
    over three studies unusual, based on unadjusted p-values of <0.05 in each study. We
    question the non-randomness of the result as presented, as it recalls the following
    “mysterious letter” effect
    2
    . In study one with 766,995 SNPs, one expects ~38,350 SNPs
    to meet the criterion by chance; 5% of that gives ~1,917 SNPs; 5% of that gives ~95 SNP
    expected to replicate over the three studies; the authors report ten. A major assertion is a
    Bonferroni adjusted p-value for marker FLJ10986 less than 0.05 in the first study.
    Computation of a Bonferroni p-value is the simple multiplication of the unadjusted pvalue,1.8×10
    -5
    , by the number of tests under consideration, 766,995, which, being greater
    than 1, is taken to be not significant. We are skeptical that any of the claims (See Figure
    1) would replicate and the biological conclusions could be ex-post explanations to what
    appear to be random data

    Seth: Yes, this is what I mean when I say the scientists involved seem to have no idea how to handle the false positive problem produced by doing so many tests.

  11. J Q Simon Says:

    Re: Macular degeneration – both my parents had it, which puts me at high risk. I started taking lutein every day and increased my intake of purple and orange foods (zeaxanthin and beta-carotene). I am over 50, which also increases the risk. After one year I had to change my eyeglass and contact lens prescription to a lower one – my vision has improved.

  12. J Q Simon Says:

    Re: Women feeling colder than men. I was always the one (woman) wearing a sweater when everyone else was in T shirts. It turns out I was hypo thyroid. The standard medication, Synthroid, did nothing to change this, or the other symptoms. I finally found a doctor who would listen to me instead of just taking a TSH test, and I started taking compounded T3 only. Every symptom improved, including temperature. Many people, especially women, are subclinical hypothyroid. TSH test alone does not find this. (Subclinical really means TSH alone does not find hypothyroidism) The doctor should also test Total T4, Free T4, Total T3 and Free T3.