Preparing doctors for the genomic tsunami

Mark Henderson has a great piece in the Times exploring the impact of personal genomics on the practice of medicine.

The basic theme should be familiar to anyone who has been following the emergence of the personal genomics industry: doctors are currently almost completely unprepared for the onslaught of genetic information they are about to experience. Here's the situation:

At present, genetic training focuses on Mendelian diseases - rare mutations in
single genes, which usually have severe effects. People who inherit the
Huntington's mutation, for example, will invariably develop the fatal brain
disorder, while 80 per cent of women who have a mutated BRCA1 gene will
contract breast cancer.

This focus is perfectly understandable. Doctors need to understand and
recognise these disorders, even if they will see few cases. Until recently,
too, Mendelian conditions were the only ones for which genetic roots had
been properly established. A knowledge of these rare diseases, however, is
not going to be much help when patients start to visit their GPs waving
printouts from genome scans like the one I took.

The genome scans Henderson describes are analyses of hundreds of thousands of common genetic variants scattered throughout the genome, currently offered by companies such as 23andMe, deCODEme and Navigenics
(Henderson had his own genome scan performed by deCODEme). Over the
last three years around 400 of these common variants have been
convincingly associated with almost 80 common diseases and complex
traits - everything from eye colour to prostate cancer - but each variant typically has a very small effect on disease risk, usually increasing it by 10-70% above baseline risk.

That means that the results of a modern genome scan are very different
to the clear-cut diagnostic genetic tests that clinicians are most
familiar with: instead of telling a patient that they will almost
certainly contract a serious, rare disease, clinicians are faced with a series of fuzzy probabilities: a lifetime risk of type 2 diabetes of 27% compared to the population average of 22%, for example.

The
task of sorting clinically relevant information from statistical noise
is daunting even for experts - let alone for GPs, lacking any training
in modern genomics, who might have a 10 or 15 minute consultation to
reassure a bewildered patient. In such cases the temptation must be
strong to simply shrug their shoulders, tell the patient that the whole
lot is garbage, and get back to the business of treating patients with
actual diseases.

Yet to do so would be a grave mistake. Even current genome scans can
yield clinically valuable data regarding the risk of some diseases,
such as Alzheimer's, or the risk of adverse reactions to drugs like warfarin. But it's also crucial to note that current genome scans represent just our first feeble steps into the world of predictive health genetics:
in less than five years large-scale DNA sequencing will be cheap enough
for whole-genome sequencing to become routine, and during that time our understanding of
the genetic basis of common diseases will continue to grow
exponentially. If individual doctors - and the medical establishment as
a whole - fail to adapt quickly to the approaching genomic era then
patients will miss out on the substantial benefits of genomic medicine.

The only possible solution is intensive clinician education: both incorporating genomic knowledge deeply into medical degrees, and offering continuing education to practising doctors. Of course it's hard to teach about a field that is developing so rapidly, and Henderson notes that course material will need to be broad and flexible:

Personal genomics is in its infancy, and with
new discoveries emerging all the time, we cannot yet know the detail of what
tomorrow's doctors will need to know. They do not need to learn about every
variant that has been linked to a disease risk or drug response: that
knowledge remains incomplete, and it can always be looked up.

What they do require, however, is an appreciation of how genetic discoveries
are likely to become integrated into medical practice, and basic skills to
make the most of them.

In other words, doctors won't need to memorise the fact that the T version of variant rs7903146 is associated with type 2 diabetes; but they will need to know the difference between a SNP and a CNV, or between a genome scan and a genome sequence. They will need to become familiar with the terminology and interpretation of probabilistic genetic risk estimates and with the use of available online resources.

There is genuine urgency here. Right now it is a simple fact that 23andMe explains the implications of a genome scan far more accurately than the vast majority of clinicians ever could, a fact that makes many of the lamentations of the medical establishment about the dangers of direct-to-consumer genomics (at least at the high end of the market, i.e. 23andMe and deCODEme) seem rather absurd.

If clinicians want to re-establish their centrality in the era of genomic medicine they have a lot of catching up to do - and they need to do it fast.

(Tangential note: I'll be away from my computer for most of the next two days, so if I don't respond to your outraged comments immediately - please be patient.)

 Subscribe to Genetic Future.