Thursday, November 11, 2010

Personal Genetics has a Family History of getting beaten up

Over the last few days personal genetics has come in for a bit of a bashing, first it was knocked out by family history then it was clearly nailed into the coffin by traditional risk factors.

Also have a look at Genesherpa’s blog for some more putting the boot in here and here.

Update: Nov 12th, here is the 23andMe blog on the subject

But what is behind the hype and the headlines? In this post I will look at family history and deal with EGAPP in the next.

The latest attack was sparked by a press release of some work presented at the ASHG. We don’t know too much because all we have is an abstract and a video but we have enough. FH was compared to the Navigenics genetic profile (PGS). From the abstract

“None of the 3 hereditary prostate cancer subjects were assessed as high risk on PGS. Based on FHRA, 10 subjects had hereditary breast cancer risk and PGS only identified 1 as high risk (K=0.12). None of the 9 hereditary colon cancer subjects were high risk on PGS.”

Sounds awful and this was the general message in the press, but it was not really so bad, not bad at all in fact. Why? Beacause Navigenics do not include the rare highly penetrant mutations that cause hereditary cancer, it concentrates on the common versions (no BRCA and no MMR genes), e.g. hereditary no polyposis colorectal cancer (HNPCC) makes up only about 3-5% of colorectal cancers. These are the cancers that FH picked up but which Navigenics missed, maybe because it was not looking for them?? (Maybe Navi should include them, maybe they don’t for regulatory/FDA reasons, but that’s a different issue).

What would really be much more informative would be to compare FH and genetics for the remaining 95-97% of cancers.

To be fair to Dr Eng, although I believe the study is flawed, she did say thatused in concert with family health history assessments, those tests could become more effective and accurate

We can only really conclude that FH was better than Navigenics at finding something that FH was looking for but Navi was not. If Navi included the BRCA and MMR genes then maybe they would have at least been on the same playing field and playing the same game. (This BTW is not a correct conclusion from ABC News: Family History Better at Predicting Disease Risk Than Screening)

In any case it’s a false battle, why choose one over the other? Although proper FH is not so straightforward, it requires a lot of effort on the part of the patient, maybe weeks to contact and collect info from all the relatives – it’s not a 20 min questionnaire, there is no doubt that it is useful when available.

Family history measures genetics to a certain extent and it measures environment as well, but it’s not a full genetic analysis and of course will not detect all genetic risks, it cannot. Genetic profiles say nothing about environment. So what does this tell us? Don’t go looking for 100% concordance because you won’t find it, but while there is overlap there are also independent contributions to be made by both – they are complementary…

The Eng study was not helpful at all in comparing FH and genetics for the vast majority of the common complex diseases, it could not be with only 44 subjects. It would be interesting and useful to do a similar but much wider study on CVD or diabetes, not to see which is best but to show how and where they are complementary and how to get the best out of both.


  1. If you listen to the abstract authors' comments in the press briefing session webcast, you will find that they are not criticizing one side or the other, but rather they point out that personalized genetic screening still needs to be examined more for validity, accuracy, utility, etc. and could be used as a complementary tool alongside family health history in clinical practice in the future.

  2. @anonymous Yes I agree the authors were more circumspect than the press and others. Still I don't think the study was a fair comparison for the reasons above. The Navi panel simply was not designed to look for highly penetrant cancers so the result was a foregone conclusion!

    What I find interesting is that Family History is often held up as the gold standard but I can find no examples of proof of clinical utility via RCTs, all I find is and - genetics on the other hand is absolutely required to have such proof, called for by great FH proponents such as Khoury et al (see

  3. Thanks to Bob West who pointed out that not all can see red links due to color blindness, so I have changed them to blue and underlined. This was reported before on the Genome Law blog but I did not follow the advice!

  4. @Keith,

    I think the point was
    1. A consumer could easily be deluded into thinking these tests cover ALL cancer risk.
    Obviously the do not.
    2. The Eng study was designed as an example of that.
    3. 9p21.3 was added to traditional risk factors, it added nothing. TCF7L2 likewise. Even a recent panel came up short.

    So, it is not one or the other. Nor is it one vs the other. Even when added together it is not too helpful.

    I maintain, only PGx is worthwhile for now....

    AKA Bootstomper/Howard Stern of Genomics/Firebomber

  5. Hi Steve

    1. Yes a reasonable point, but...
    2. Eng abstract: "...there has been limited study of the accuracy and concordance of these 2 methods of risk assessment for common diseases". So I disagree with you, it was not designed to do that and it didn't. If it was designed to do that it would be following up the several people said to be a raised risk of common cancer, detected by PGS but not FH. Abstract ends: "evaluation of family history is still the proven gold standard, and this should be used to clinically evaluate an individual’s risk of developing cancer"

    3. Discuss later in separate post on trad risk factors.

    Maybe for now it is only PGx, that's not my point, I'm not trying to claim PGS is worthwhile, I'm saying we need to do the right studies.

    Back to the Eng study - it should have been a comparison of panels of rare cancer variants vs. FH. In fact that would be interesting, given they are highly penetrant and that proper FH is very difficult and underused, it could be cost effective, and simple more effective, at rooting out people with extremely high risk, to use genetics rather than FH. Unfortunately we don't know because the opportunity was missed to test it.

  6. good post. have to add: you can have a strong FH of cancer (or anything) and yet didn´t inherit the risk variants (or most of them) by chance. in that case is obvious you´ll be misclasified by FH alone and better clasified by genome profiling. so, big news: the two of them work better in combination. really? I never would have thougth that!!! having more information about a patient is better than having less, that´s new.