NcXHF

STEPHEN T. SONIS
portantly, patients were able to make treatment preference decisions
based on their understanding of toxicity risk within the
context of cancer therapy treatment outcomes.
From the standpoint of integrating genomics into a comprehensive
management paradigm that includes supportive care, a
genomic test that addresses the inclusive risks of toxicities has
most value when it is interpreted in the context of a specifıc anticancer
regimen. For example, if a patient with breast cancer
can be effectively treated for her disease with either Regimen A
or Regimen B and the risk of the toxicity profıle associated with
each regimen can be defıned, the patient and the oncologist can
choose one over the other based on patient preference. If, on the
other hand, Regimen A is not as effective as Regimen B, but the
toxicity risk profıle is more tolerable to the patient, there may be
a trade-off point at which the patient determines that the risk
and scope of side effects outweigh the selection of the potentially
more effective cancer treatment.
A second value in understanding a patient’s toxicity risk focuses
on the chance to individualize and target the use of medications
that prevent or treat specifıc side effects. For example,
approximately 40% of patients treated with certain conditioning
regimens for hematopoietic stem cell transplant develop severe
oral mucositis with all of its consequent comorbidities. Palifermin,
keratinocyte growth factor-1, is approved to prevent its development.
8 To be effective, palifermin must be administered
for 3 days before the infusion of the conditioning regimen, a
time in which there is no evidence of mucosal injury. Not only
does this schedule add days to care, it also incurs substantial fınancial
costs. Both are well worth the price for the individual at
risk of mucositis, but there is more to be lost than gained for the
remaining 60% of patients. Thus to administer palifermin to everyone
does not make sense at multiple levels. However, if one
could predict with reasonable certainty which patients were at
risk for developing mucositis, the agent could be given selectively
and more economically.
In addition to risk prediction, at least two other potential applications
for genomics as it relates to personalization of supportive
cancer care are available. In the current paradigm of
drug development, success or failure is defıned as an assessment
of the mean. 9 Criteria for effıcacy are set around a one-size-fıtsall
vision in which it is presumed that all patients have an equivalent
response to a drug. Since the response rate for drugs ranges
dramatically (90% of drugs work in only 30% to 50% of patients),
10 we know this is not the case. There is clearly individuality
in how any population with cancer responds to a drug.
Clinicians often deal with this variance by adjusting dose or
through trial and error. However, genomic differences often defıne
response/nonresponse and the ability to dichotomize patients
prospectively offers a great opportunity to personalize
their care and create hierarchies for toxicity intervention.
For example, if multiple agents to treat chemotherapyinduced
nausea and vomiting are available, knowing
which one was most active in a prospective patient makes
prescribing more effıcient and cost-effective.
Finally, genomics provides an important tool in drug development—discovery
through clinical trials. Both radiation
and chemotherapy stimulate changes in gene expression that
trigger the pathobiologic events that produce toxicities. Identifying
and defıning the sequence of gene activation that underlies
regimen-related injury provides specifıc targets for
intervention. In addition, by mapping the inter-relationships
between cooperative groups of genes and organizing these
into networks, the hubs at the center of the network can be
targeted and disrupted. This approach mimics what happens
to the United Airlines flight schedule when there’s a snowstorm
in Chicago.
PHARMACOGENOMICS AND TOXICITY-RISK
PREDICTION AND TREATMENT
Pharmacokinetic Risk Assessment
Genomic contribution to chemotherapy-associated toxicity risk
is governed by two components: one associated with the concentration
and availability of the drug (pharmacokinetics [PK]),
and the other on how the drug affects the biology that underlies
the pathogenesis of the toxicity. Drug dosing typically is determined
on a one-size-fıts-all notion (recommended dose) that
typically has been determined by dose escalation in clinical trials
based on the mean response of the study population. Although
dosing is adjusted based on patient size, variable effıcacy and
toxicity outcomes are common. This should not be a surprise
because we now know that patients do not handle drugs in a
uniform way. Drug metabolism is affected by enzymes, and enzymes
are controlled by genes. Too much enzyme production
and the drug’s tumoricidal effects are minimized, but toxicity
risk is low. Too little enzyme production and there is an effective
overdose with a large toxicity risk.
A classic example of a PK-related pharmacogenomic
marker of toxicity-risk prediction is the catabolic enzyme, dihydropyrimidine
dehydrogenase (DPD), which plays a critical
role in fluorouracil (5-FU) metabolism. 11 Insuffıcient
DPD activity results in toxic levels of 5-FU and is associated
with increases in both hematologic and nonhematologic toxicities.
Variants in the DPYD gene affect DPD activity. At
least two variants, DPYD*2A and D949V, 12 have been identifıed,
largely through a candidate gene approach, as being
associated with increased toxicity risk. Furthermore, additional
DPYD variants have been uniquely described in black patients, a
fınding that emphasizes the importance of broad demographic
inclusion criteria in any study aimed at identifying genes affecting
PK. 13 DPD currently is the only genomic marker for 5-FU
risk prediction that the U.S. Food and Drug Administration
(FDA) recognizes and for which a commercial test exists. 14 Not
surprisingly, not all tests have equivalent value. 15
An association between gene-based changes, pharmacokinetics,
and toxicity risk also has been described for many chemotherapeutic
agents, 16 including methotrexate, platinum
drugs, taxanes, and anthracyclines. 17-19 Results of these studies
have been largely inconsistent. Whether the variance in
fındings is a consequence of small sample size, inconsistent
study design, differences in candidate gene or SNP selection,
or variability in toxicity defınitions is unclear. But the disparity
in conclusions is remarkable. For example, two studies of
the genomic PK toxicity association in pediatric patients
10 2015 ASCO EDUCATIONAL BOOK | asco.org/edbook