Posted on April 20, 2017 at 4:18 PM
The news release Monday morning grabbed my attention:
“Study finds wide gap in quality of BRCA1/2 variant
classification between Myriad Genetics and a common public database.”
Myriad Genetics had been exclusively providing tests, for
$3000+ a pop for full BRCA gene sequencing, for 17 years before the Supreme
Court invalidated key gene patents back in 2013. Since the ruling a dozen or so
competitors have been offering tests for much lower prices. Meanwhile, Myriad
has amassed a far deeper database than anyone else, having been in the business
so much longer. And it’s proprietary.
CLASSIFYING GENE VARIANTS
Public databases of variants of health-related genes have
been around for years too. The best known, ClinVar, collects and curates data
from the biomedical literature, expert panels, reports at meetings, testing
laboratories, and individual researchers, without access to Myriad’s database.
ClinVar uses several standard technical criteria to classify variants as
“pathogenic,” “benign,” or “of uncertain significance.” (“Likely pathogenic”
and “likely benign” were used more in the past.)
ClinVar lists 5400 variants just for BRCA1. The criteria
come from population statistics, how a particular mutation alters the encoded
protein, effects on the phenotype (symptoms), and other information.
Bioinformatics meets biochemistry to predict susceptibility. The BRCA1 protein
acts as a hub of sorts where many other proteins that control DNA repair
gather. DNA Science discussed the genes behind breast and ovarian cancers here.
As gene sequences accumulate in the databases and troops of
geneticists and genetic counselors annotate them, the proportion of pathogenic
and benign entries will increase as that of the unsettling “variants of
uncertain significance” — VUS — will decrease. Knowing will outpace not
knowing. In the meantime, a woman with a VUS for a gene that confers
susceptibility to develop breast and/or ovarian cancer, concerned that her DNA
sequence isn’t “normal,” might have her at-risk organs removed. And then she
might learn, as the databases grow, that her variant is really benign after
all, because while she was being treated for a disease that she didn’t actually
have, more patients emerged who have the gene variant but not the related
The reclassification of a VUS as something more meaningful
depends on access to as much information as possible.
How many unnecessary surgeries are happening? Have happened?
Myriad estimates hundreds to thousands.
The news release announced a paper from Myriad just
published in The Oncologist that compared 4,250 unique BRCA1 and BRCA2 gene
variants in the company’s database to entries in ClinVar, and found that 26.7%
of the ratings do not fully agree. (ClinVar lists multiple sources for many of
the variants, and 14.5% didn’t agree on all counts and 12.3% on only some.)
According to Myriad, most of the VUS listings were in ClinVar. In fact, only
0.5% of Myriad’s BRCA1 sequences and 1.1% of their BRCA2 sequences are VUS.
They’re clearly ahead of the game.
A breast cancer cell (NHGRI).
The discordance isn’t really news. It echoes earlier reports
both from Myriad researchers for the BRCA genes and from a non-Myriad, multi-center group that also
found 26% discordance, but their analysis included several cancer
predisposition genes; the most conflicting were CHEK2 and ATM. And another
recent study found that about half of women with VUS undergo bilateral
mastectomy, concluding that “many surgeons managed patients with BRCA1/2 VUS
the same as patients with BRCA1/2 pathogenic mutations.” That’s terrifying. It
means that the problem isn’t due just to overanxious patients, but also to
surgeons who might not be genetics-savvy.
The news release doesn’t go to the obvious conclusion, that
privatization of the database may be fueling inappropriate surgeries, or at
least not preventing them, a point that Sharon Begley made elegantly in Stat
News. Many of her expert sources countered the implication that variant
classifications from ClinVar are just accepted and delivered to patients
without further and often intense research and validation by health care
In the news release, William Gradishar, MD, from the Feinberg
School of Medicine at Northwestern University and lead author on the new study
from Myriad, provided the company view (slightly edited):
“The high degree of discordance seen in this study signals a
cautionary note … it means that different labs are providing different results
to patients for the same genetic mutation … some patients are receiving
incorrect results that may have life-changing or -threatening implications…
Although efforts are underway to resolve the quality problems within public
databases, it is unlikely the issue will be resolved soon and users of public
databases likely will continue to encounter discrepancies. At this time, labs
should not use public databases in any way in clinical variant classification.”
Actress and activist Angelina Jolie had her breasts and
ovaries removed to lower her risk of BRCA1-associated cancers.
Ron Rogers, spokesperson for Myriad, wrote in an email that
the company provides database access for certain scientific collaborations, but
then provided the business view:
“Myriad has invested hundreds of millions of dollars to
develop our database over two decades, which is used to classify variants as
part of the Company’s specific process. We encourage other commercial labs to
make their own investments, which are required to offer a quality product. The
problem is that some other commercial labs refuse to make the needed
investments. Instead, they are trying to short cut the process by relying on
inaccurate public databases, which has the potential to deliver inaccurate
results and harm real people.”
If anyone remembers the race to sequence the first human
genome, which pitted NHGRI and their public database against Celera Genomics
and their proprietary one, the battle over BRCA is a little deja vu all over
Rogers added that if patients who have had bilateral
mastectomies following a report of a VUS had come to Myriad, “we could have
provided many of them with definitive answers and helped to prevent a
significant number of these life-changing procedures.” About 85% of the BRCA
testing market indeed uses Myriads products.
Myriad’s tests as well as their support (genetic counseling)
are excellent. The company helped a friend of mine interpret and confirm BRCA1
test results from 23andMe, which her son had taken on a whim but the unexpected
finding of a mutation had sent her family into a tailspin. 23andMe, the
direct-to-consumer genetic information company, coaxes customers to provide
their results for research (“Be part of something bigger”) and charges investigators
for access to the data. Like Myriad, 23andMe also suffered a setback in 2013.
FDA yanked some of their health-related carrier tests, but the agency reversed
the ban on April 6, 2017.
23andMe and Myriad serve different types of individuals. Many
of 23andMe’s customers, like my friend’s son, aren’t distraught over cancer,
just curious and acting on their own. So here’s another comparison: Is Myriad’s
protecting of their data similar to my reaction to people downloading my books
for free? No. I’m just losing royalties.
My mother’s lost battle with breast cancer can’t help but
influence my view that impeding accurate diagnosis for profit, to protect
assets, may make sense in the business world but is not the right thing to do.
I have trouble commoditizing patient tests and sequestering information that
might help, and the many rock stars of genetics who have joined the Free the
Data movement would agree.
Because the entire discussion makes me queasy, I thought
instead I’d illustrate the distinction between a harmful mutation and a VUS, by
comparing the BRCA1 gene sequence to the English language.
THE HANDMAID’S TALE
Most genes are thousands of DNA bases long, and therefore
can vary in many ways. Gene variants – mutations – can affect health or not
depending upon how they disrupt the proteins that they encode.
BRCA1 (the gene) is 125,951 bases long, only 5,589 of which
are represented in the protein. Mutations near its ends are more damaging than
those in the middle. One of the two Ashkenazi (Jewish) “founder” mutations,
185delAG, removes an A and a G at position 185, near the beginning. Because a
gene sequence is read in triplets – every 3 DNA bases encodes an amino acid –
adding or deleting a number that isn’t 3 or a multiple of it devastates the
protein’s function. 185delAG is, then, a gene variant of very high
significance. (My analogy below is not quite on target because the words in a
sentence vary in length.)
I entered 125,951 letters into a publishing calculator that
translates numbers of alphabet characters into a book of a particular size, and
wound up with a 6” by 9” volume of 265 pages. That view of the BRCA1 gene’s
enormity corresponds to approximately the length of Margaret Atwood’s dystopian
masterpiece “The Handmaid’s Tale.” I’m rereading it, inspired by the “Make
Margaret Atwood Fiction Again” signs at the March on Washington January 21.
I took a sentence about as far into the book as the 185delAG
mutation is in the BRCA1 gene:
Everything except the wings around my face is red: the color
of blood, which defines us.
Removing two adjacent letters – the “ex” in “except” – and
then moving the other letters into the word-sizes, clearly disrupts the
Everything ceptth ewi ngsar oundmy fa ceis re dth: eco lorof
bl oodwh ichde finesus
Gibberish, and it would continue to the end of the book. In
a gene, the offset sequence would likely lead to a stunted protein, halted when
reaching a “stop” signal in the mRNA. And when a BRCA1 gene loses meaning,
control of DNA repair falters and other mutations go unchecked. Cancer results.
A VUS would do something less drastic, such as replacing a
letter in a way that doesn’t obscure the overall meaning of the sentence (or
gene). Substituting the “g” in “everything” with a “j,” for example, doesn’t
alter the meaning. And many changes to a DNA sequence have no effect on the
encoded amino acid sequence at all.
So that’s what a VUS is, a DNA sequence of a particular gene
of interest that’s rather rare in a population, and that may or may not mean
something. And their numbers will decline and then vanish as the databases
build and their entries are annotated and validated. But until they do, I think
that sharing all that we know about variants of possibly disease-causing genes,
to prevent unnecessary treatments, is the only ethical path to take.