Monday, October 13, 2008

deCode Versus Arthur Caplan PhD.

To attempt to sort out the hype from the hope, Sherpa Style. I will review each and every SNP that deCode is using for it's Breast Cancer Test. This weekend I will cover the first 2. But first some hype from both sides.


Genetic testing for all sorts of conditions is all the rage these days. Everywhere you turn, some company is urging you to spit in a cup, take some blood or swab your cheek so your DNA can reveal your health risks, know who your long-dead ancestors are, pick the right mate or help you design a diet that is perfect for your genetic makeup. But, "spitomics" has gotten way ahead of genomics.

Sadly, the tests Decode and other companies are offering are more likely to empty family pocketbooks and leave women with a false sense of security than they are to prevent breast cancer.



Arthur Caplan stresses caution in the application of the new genetic risk tests for common diseases and I certainly agree that genetic testing should be applied with care. However, he goes too far when he says that the new deCODE BreastCancer genetic risk test is only useful for women who have two or more close relatives with breast cancer, is not based on large enough studies to be accurate, and is not regulated........Each of the genetic markers in this risk test have been replicated in between 5 and 30 different populations in studies by deCODE genetics, the National Cancer Institute, and UK Cancer Research.


These studies have been published in the most prestigious, peer-reviewed journals, including Nature Genetics and the New England Journal of Medicine. Altogether almost 100,000 patients and controls have been studied to define the marker risks. We made this test available for physicians to order for their patients through our reference laboratory which is regulated under CLIA by the US Federal government.


There are two major types of breast cancer: the rare, early onset form that occurs in certain families and for the detection (for which the Myriad Genetic test is well suited), and the common form which accounts for 95 percent of breast cancer.


Well, Dr. Jeff, you are making the case for BRCA smaller than the data shows.....So right off the bat I am a little suspect.....The number for familial breast cancer is 10-12% and BRCA has the majority of that......Oh and you do read the economist don't you? Famous is not always better buddy.

Ok, so who is right. The chances are high that they both are. But let's go to the tape!


What are the SNPs they test for? On limited weekend hours I will review 2 SNPs.


1. rs13387042 when looked up in a pubmed search only shows one article in Nature, by guess who? deCode.....still waiting for this "Well replicated SNP"

What does this SNP reveal and for whom?

Population: Icelanders, replicated on Northern Europeans (Sweden, Spain, Holland, US MEC)

Prevalence: 25% of Europeans, widely varied in other ethnicities

Function: Unknown, no genes known in this linkage disequilibrium block.

Penetrance: Unknown, OR is 1.2 for ER positive and 1.06 for ER negative

This study isolated 10 SNPs and only 2 SNPs replicated, I don't see the unreplicated ones in the paper. Importantly, this SNP did not replicate perfectly in the Swedes.... The SNP varied widely between ethnicities in the Multi-Ethnic Cohort in the US.......They observed no interaction between the 2 SNPs that replicated in Caucasians.....The effect seems that it could be recessive.....


The study even acknowledges that this was a limited effort in finding risk SNPs....."However as the relative risks remain low, they (these SNPs) can only account for a small fraction of the familial clustering of this disease.


2. rs4415084 guess what, the same thing happened with this SNP. Only one article, by guess who? deCode. This time in the Brief Communications of Nature Genetics......I look forward to the Independant Replication...


Population: Iceland first, then replication in (Sweden, Spain, Holland, US MEC). It flopped in Nigerians!

Prevalence: Not reported in this one and only paper on this SNP

Function: Unknown, but maybe linked to FGF10 which is amplified in 10% of breast cancers. Also MRPS30 is in this area....but really, we have no freakin' clue.

Penetrance: Unknown, but the OR for ER positive breast cancer is 1.27, for total breast cancer is OR of 1.14


So let me translate. These are 2 of the seven SNPs DeCode claims will help show risk of common (Read as not BRCA related) breast cancer. If this is any indication of the rest, I am sad to say that they are out on a limb here.


Several clinical questions come up when I want to order this test.

1. Who do I test? In the case of these SNPs, I would guess white women.

2. How common is this risk? Not very common in the case of Asians or Africans. In fact in one of these papers, a SNP being tested for didn't show increased risk, it actually showed protection from breast cancer.

3. How do I counsel a positive test? Tough to say. We don't know the penetrance, we know an odds ratio for a certain group.

4. Can I see deCode's data on their magical risk calculator thingy that includes these SNPs? Or is that a trade secret?

5. How reliable is this science? Not very, the studies were replicated by deCode and not by other groups.....

6. Is this "test" validated? I would love to review the seven panel test publication....oh wait, I couldn't find it.....


Not all of these SNPs are that thoroughly validated or as "replicated" as Dr Gulcher said. So this is the case of rushing to market with a test.

The Sherpa Says:

This test is Hype until proven otherwise. I await the future studies on these SNPs. Dr Jeff does a good job on deCode's blog of promoting it though. I know early detection is key and pre-disease is the new disease, but a clinician has to have confidence in their tools. This one is a little primitive for me.......and at worse could give patients a false sense of security. So guess what......Jeff and Art, you are both correct.



1 comment:

Andrew said...

Hm, looks like I'll have to get you tools ready for HelixGene so we can report this!