CANCER COMMONS: BIOMEDICINE IN THE INTERNET AGE - Collaborative Computational Technologies for Biomedical Research - page 170

together or disseminated effi ciently because medicine relies upon the publica-
tion of review papers or guidelines which are only occasionally updated, and
their results are not computationally accessible and so cannot be effi ciently
utilized in clinical settings. Moreover, there is enormous variability in treat-
ments and outcomes, even given the guidelines. What is needed in all of these
cases is to capture and integrate data and evidence not only from large-scale
controlled trials but also from the thousands of ad hoc N - of - 1 experiments
that occur every day in the practice of oncology. These experiments test
hypotheses based on the creativity, knowledge, and experience of practioners
seeking the best possible outcomes for their patients, not pharmaceutical
companies seeking regulatory approval for a particular drug.
The concepts of adaptive trials and of N - of - 1 medicine are, independently,
fairly well understood. Indeed, there are entire topics dedicated to the design
of adaptive trials [e.g., 6], although they remain somewhat controversial and
must be designed with great care [7]. There is also a long tradition in medicine
of publishing case histories as a way of communicating treatment experiments
to others.
Technically, an N - of - 1 study is a specifi c sort of experimental design where
the subject, for example, an individual patient, acts as his or her own control.
For example, a baseline measurement may be made, serving as a control, and
then a treatment applied and a response observed. The treatment may then
be removed, the patient observed returning to baseline, and so forth, in accord
with a protocol that can be replicated to get the required sample. In diseases
like cancer such a design could sometimes be applied, but usually what one
means by N -of-1 in these settings is not such a specifi c experimental design,
but rather just the assumption that each patient's disease is different from
every other patient's disease. Although in a trivial sense this is always true, it
is clear that in some cases it is more practically relevant than in others. For
example, if your friend were to acquire a bacterial infection from you, it is
likely that you functionally have the same disease—that is, an antibiotic that
works on you is very likely to work on your friend as well. Of course, there is
a small chance that the organism could have mutated along the way or that
your friend is allergic to the antibiotic, so adjustments will always have to be
made from one person to the next, but generally speaking bacterially transmit-
ted diseases do not require a great deal of intricate tailoring of treatments to
each individual. This is not the case for many other diseases, especially ones
where there is a great deal of variability at the genomic level, the paradigmatic
example of this being HIV/AIDS, which is known to mutate rapidly. Cancer
is a particularly complex case because, whereas we know that all cancers are
due to genomic mutations, we do not know what the practical number of effec-
tive carcinogenic mutations is; and the number of functional subtypes requir-
ing different treatments could range from tens to hundreds or even thousands.
As a practical matter, in the case of cancer N - of - 1 treatment has come to mean
profi ling the specifi c genomic mutations that drive an individual's cancer and
then choosing a treatment based upon that specifi c genomic profi le. Of course,