May 08, 2003
making things personal

Dean Williams, Ralph Snyderman, & Huntington Willard published this editorial in Science. The leadership in the Medical Center has chosen the concept of "individualized" medicine as a rallying point for research. From the angle of genomics, I published this editorial in the Chronicle, December 2001. I think such an approach will be the important next step in medical care. Everyone is very aware that assigning a patient to one prognostic bin versus another can be arbitrary. A woman with breast cancer and 0 positive lymph nodes is treated differently than a woman with 1 lymph node. These sorts of tools are very "low resolution"--on average they work, but there is significant room for improvement. And because these bins are large, a patient isn't treated as a distinct entity, but as an undifferentiated member of a bin. What we're harnessing genomic information for is to get away from discontinuous categories and move into a continuous space that allows us to predict a woman's individual prognosis as a point estimate for her, not as a member of a group.

Ultimately, this sort of individualization will revolutionize how we prescribe drugs, how we determine cancer treatment protocols. It will even allow us to "recover" drugs that had been shelved because on average they failed to work for large population, but we will now be able to tell specifically for whom that drug is actually efficacious.

Patients instinctively recoil from being treated as merely being a part of a population, they want to be addressed as a singular person with singular health characteristics...thus the wackiness of plying Whole Foods' herbal remedy aisles, which is symptomatic of Big Medicine's failure to answer personal needs.

Next week, our paper which demonstrates some first steps in this direction for High- and Low-Risk breast cancer patients as well as for predicting lymph node status will be appearing in The Lancet.

Posted by erich at May 08, 2003 02:31 PM
Post a comment

Email Address:



Remember info?