Personalized Medicine - Translational Informatics: Realizing the Promise of Knowledge-Driven Healthcare - page 49

expense requires signifi cant evidence that outcomes are improved. As an example,
CMS will currently only approve genetic testing for warfarin metabolism if the test-
ing is performed in the context of a clinical study. In addition, all of the benefi ts
described above apply on average . Individual cases may raise ethical quandaries.
As an example, consider pharmacogenomics. If a terminally ill patient has a 50 %
chance of responding to a drug, should insurance cover it? What about 5 %? 0.05 %?
What if another drug is available, but causes more unpleasant side effects? What if
the patient in question is your child?
One stakeholder for whom personalized medicine is mixed news is the pharma-
ceutical industry. On one hand, stratifi cation of the patient population makes block-
buster drugs less likely. Drug companies stand to make greater profi ts when a drug
is prescribed across the largest possible number of people. On the fl ip side, diagnos-
tic companion tests may enable FDA approval for whole classes of drugs that would
not have been seen as successful across the population. In some cases, a test may
help rule out patients who are likely to have adverse events. In other cases, the test
may help single out people who are particularly likely to respond well to a drug.
Focusing on the right segment of the population can make clear the benefi ts of a
drug that, on average, would not outperform the current standard of care.

3.5.2 Ethical, Legal, and Social Issues

3.5.2.1 Data Sharing

From the data generation and policy perspective, personalized medicine necessi-
tates “big data” approaches. The sheer number of variables means that larger num-
bers of subjects (i.e. “bigger N”) are required to support scientifi c conclusions. In
light of the need for larger N, data sharing and re-use becomes increasingly impor-
tant. Research funding cannot sustain the sample size that is required in the increas-
ingly high-dimensional domain of biomedical research [ 41 ]. Those generating data
must share the data in a manner that makes them accessible and comprehensible to
those who would use them to further biomedical discovery. Makers of policy,
including publishers and funders, have already established a number of guidelines
for good data sharing practices. And increasingly these policies are actually be
enforced.
Of course, a mandate to share data necessitates the creation of somewhere to put
the data. Publicly available databases are also needed as repositories for researchers
to deposit and access data. Free, unobstructed access to well-annotated, high quality
data enables collaborations and data re-use and reduces obstacles to research. dbGap
and TCGA (The Cancer Genome Atlas) are two examples of such repositories,
though arguably there is some room for improvement in ease of use of the interfaces
for data access [ 42 ]. In addition, after researchers demonstrated that it was possible
to identify the presence or absence of an individual in a complex mixture of DNA
samples, the NIH limited access to GWAS data to eligible researchers who are