Sciencedirect
Sciencedirect
Sciencedirect
available at www.sciencedirect.com
ScienceDirect
www.elsevier.com/locate/molonc
Review
Donald A. Berry
Division of Quantitative Sciences, The University of Texas MD Anderson Cancer Center, Houston, TX, 2006-2010,
USA
A R T I C L E I N F O A B S T R A C T
Article history: Clinical trials are the final links in the chains of knowledge and for determining the roles of
Received 24 February 2015 therapeutic advances. Unfortunately, in an important sense they are the weakest links.
Accepted 24 February 2015 This article describes two designs that are being explored today: platform trials and basket
Available online 11 March 2015 trials. Both are attempting to merge clinical research and clinical practice.
ª 2015 Federation of European Biochemical Societies. Published by Elsevier B.V. All rights
Keywords: reserved.
Biomarker-driven clinical trials
Platform trials
Basket trials
Bayesian adaptive clinical trials
Molecular biomarkers point the way to unraveling cancer. Bi- Stretching the analogy with pioneering, the rocketships of
ologists are cracking cancer codes at an unprecedented rate modern biology culminate their final stage of delivery in a
and are learning how biomarker profiles interact with therapy. wagon train.
Even so, future historians will view today’s cancer biology as Traditional clinical trials are straightforward and pur-
nascent. What is still unknown dwarfs the known. posely simple. Each trial addresses a single scientific question,
Clinical trials are the final links in the chains of knowledge one that barely scratches the surface in unraveling the nature
and for determining the roles of therapeutic advances. Unfor- of complex diseases such as cancer. Answering a single ques-
tunately, in an important sense they are the weakest links. tion is a baby step in the big picture of understanding and
Moreover, as biology advances still further traditional clinical curing cancer. Actually, some steps are backwards because
trials will increasingly limit progress. In the 1940s A. Bradford that they occupy resources that would better serve if allocated
Hill designed and conducted a randomized clinical trial (RCT) more wisely. Simple, one-question trials were satisfactory or
of streptomycin in tuberculosis. That pioneering achievement at least tolerable when we had a small number of therapies
turned medicine from anecdote and case study into legitimate to investigate for a disease that was regarded to be
science. RCTs have changed little in the last 70 years. homogeneous.
5
This is a contribution to the special issue edited by Drs. Ringborg, Mendelsohn and Schilsky.
E-mail address: [email protected]
http://dx.doi.org/10.1016/j.molonc.2015.02.011
1574-7891/ª 2015 Federation of European Biochemical Societies. Published by Elsevier B.V. All rights reserved.
18780261, 2015, 5, Downloaded from https://febs.onlinelibrary.wiley.com/doi/10.1016/j.molonc.2015.02.011 by Reprints Desk Inc, Wiley Online Library on [18/06/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
952 M O L E C U L A R O N C O L O G Y 9 ( 2 0 1 5 ) 9 5 1 e9 5 9
The problem is that cancer is many, many diseases. Our to patients who have the disease and present after the trial.
erstwhile taxonomy by organ of origin may have hindered But the world has changed, and it continues to rapidly evolve.
progress as much as it helped. A lung cancer is more like other Traditional, large clinical trials take many years to design and
lung cancers than it is like breast cancer, say. But the hetero- run. And they can be straightjackets for clinical practice by
geneity of lung cancer is fundamental to understanding it. providing answers to outdated questions. In the interim cancer
And modeling across organ sites is fundamental to curing it. biology will have asked many new questions, perhaps
Coupled with the greater understanding of the drivers and including which combinations of the experimental agents un-
backseat drivers of cancer is the burgeoning number of small der investigation in the trial are appropriate for which patient
molecules that have anti-cancer activity. Combinations of these subsets. These subsets will likely be defined by biomarkers that
molecules make for an essentially infinite number of thera- were unknown at the start of the trial.
piesdfar more therapies than patients. Researchers cannot
address more than a small proportion of them in clinical trials.
We speak of false negatives and false positives, but both are 2. Clinical trials in the Brave New World
dwarfed by false neutralsdtherapies that have not been and
may never be evaluated in clinical trials. Resources have been The focus of traditional clinical trial design is hypothesis
allocated traditionally to evaluating a small number of thera- testing, including control of type I error rate and statistical po-
pies, ensuring that they are neither false positives nor false neg- wer. In a typical scenario, investigators propose a clinical trial
atives. Like a myopic gold miner we sift the dirt in our tiny claim comparing an experimental arm to a control. Assuming the
while nearby mountain ranges harbor the real treasures. primary endpoint is time to a particular type of negative event
In view of the huge number of and the complicated nature they might want to be able to detect a 25% reduction in hazard
of pathways involved in cancer, the “gold in them thar hills” with 90% power and 5% two-sided type I error rate. Suppose
involves combinations and sequential therapies. Consider the median time to event in the control arm is expected to
the space of single agents that have potential anti-cancer ac- be 3 years, accrual rate is 4 patients per month, and minimum
tivity, the “therapy space.” Its dimension is the number of subsequent follow-up is 3 years then they would calculate
possible single agents. Suppose 1015 just to be specific. Dose that about 650 patients would be required.
is a whole other set of dimensions, at least another 1015. Such a trial would report results 16 years hence, when the
Now consider the tumor/patient biomarkers that are (or may trial’s conclusion may be irrelevant. Nobody would run such a
be) associated with cancer, perhaps 1010. They constitute the trial. Instead they would pretend that they expect a 65%
“patient space.” The ways in which the therapy space inter- reduction for which moderate sample size would give reason-
acts with the patient space are hugely complicated, at least able power. It’s a joke. The punchline is that the same hypoth-
1040. We have barely begun to learn which of the myriad of esis testing approach is the status quo for both common
combinations of therapies are best for which patients. conditions such as hyperlipidemia and rare diseases such as
Guardians of yesteryear cling to what they know and shun high risk relapsed Wilms’ tumor. It may work for the former
the new and the unknown. Tradition is a crutch, one that im- but it does not work without pretenses such as the above for
pedes progress. The advances in medicine made by RCTs the latter. The rub is that the latter is a glimpse into the future
confirm their value. But their number is too small. We must of all cancer research. The ability to run definitive clinical tri-
take the RCT to new levels. And we must consider a variety als that address questions biologists are asking is dwindling,
of approaches to clinical trials as the Brave New World un- including for diseases that were common when they were
folds. It is ironic that we take the same clinical trial approach categorized by organ of origin.
to evaluate all manner of potentially amazing transformative Future trials in cancer (and more generally) must explicitly
experimental therapies and yet we don’t experiment with the consider disease prevalence. They must also consider the
design of the clinical trial itself. rapidity of advances in biology and the rate at which alterna-
tive therapies are being developed. This will mean smaller,
shorter, and more focused trials. The new clinical trial para-
1. Experimenting with experiments digm should have the goal of delivering good medicine to pa-
tients who have or will have the narrowly defined disease in
Traditional clinical trials focus on large populations. Clinical question. Randomization will continue to play a role, albeit
trials keep getting bigger, making therapies more expensive one that is more refined, as I will describe. But hypothesis test-
and delaying their availability to patients. The rub is that mod- ing’s role will be at most ancillary. Trial designs in the Brave
ern biology is slicing and dicing cancer into ever smaller sub- New World will be radically different. No one knows what
sets. Soon every cancer patient will have an ultra-orphan drug development and drug regulation will be like, but dra-
disease. Traditional approaches to designing clinical trials matic change is inevitable.
are helpless and hopeless when evaluating therapies for dis- In later sections I will describe some trial designs and ap-
eases with an incidence of a few hundred patients per year. proaches that are being investigated in the face of a burgeon-
And yet this is the future of all cancer. ing number of patient subpopulations that are ever shrinking
The Belmont Report of 1979 draws a clear distinction be- in size. Some of these designs may presage the future. Some
tween clinical research and clinical practice, stressing that are radically different from the usual approaches to designing
the former is “designed to test an hypothesis . and thereby clinical trials. But they are not different enough. All the de-
to . contribute to generalizable knowledge.” (US Department signs I consider are anchored in the past in the sense that
of Health (1979)) “Generalizable knowledge” is that applicable they take a standard hypothesis testing perspective.
18780261, 2015, 5, Downloaded from https://febs.onlinelibrary.wiley.com/doi/10.1016/j.molonc.2015.02.011 by Reprints Desk Inc, Wiley Online Library on [18/06/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
M O L E C U L A R O N C O L O G Y 9 ( 2 0 1 5 ) 9 5 1 e9 5 9 953
Name Description
Platform trial Evaluates many therapies in a particular disease or group of diseases. Therapies usually have different
sponsors and may be combinations or sequences.
Standing trial Platform trial in which therapies enter and leave over time.
Master protocol A trial with multiple treatment options requiring separate protocols but under the same aegis. Informed
consent is usually required for both the master protocol and the respective individual protocol.
Indication finder Evaluates a particular therapy across multiple cancers that are defined by organ type, or across subtypes
within a specific organ type. The goal is to determine which diseases or which biomarker subtypes are
appropriate for further development.
Basket (or bucket) trial Evaluates the effect of a particular targeted therapy on a particular genetic or molecular aberration across
cancer organ types. Variant of indication finder but the therapy is not evaluated for its off-target effects.
Umbrella trial This term may be useless because it is used for very different designs by different researchers and reporters.
I use it for platform trials (many therapies) that are indication finders for each therapy.
Adaptive trial Trials in which unblinded data are monitored and used to determine the future course of the trial based on
prospectively defined decision rules.
Seamless phasesa A particular kind of adaptive trial that moves from one phase of drug development to another without
pausing accrual. Decisions at the phase switch usually involve greater focus. Examples include dropping arms,
dropping doses or schedules, dropping patient subsets, changing randomization proportions, estimating the
sample size for the next phase, and there are many other possibilities. (I do not include changing primary
endpoint for reasons indicated in the text.)
a FDA’s February 2010 draft “Guidance for Industry, Adaptive Design Clinical Trials for Drugs and Biologics” <http://www.fda.gov/downloads/
Drugs/Guidances/ucm201790.pdf> is focused on phase 3 trials. It avoids the term “seamless phase 2/3” because the term provides “no additional
meaning beyond the term adaptive.”
18780261, 2015, 5, Downloaded from https://febs.onlinelibrary.wiley.com/doi/10.1016/j.molonc.2015.02.011 by Reprints Desk Inc, Wiley Online Library on [18/06/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
954 M O L E C U L A R O N C O L O G Y 9 ( 2 0 1 5 ) 9 5 1 e9 5 9
the same control arm in a single “platform trial.” Having two therapies in metastatic disease and then, if the therapy is suc-
experimental arms and one control in a single trial reduces to- cessful, evaluate it in primary disease. The problem with the
tal sample size by 25% in comparison with two two-armed tri- traditional approach is that it misses effective agents that pre-
als. For a trial with a greater number of experimental arms the vent metastases but have no or little effect once metastases
savings approaches 50%. Not all experimental arms need be have occurred.
included in the randomization scheme at the start of the trial Table 2 shows the 8 disease subtypes considered in I-SPY 2.
but instead arms can be added as they become available: a These are defined by tumor hormone-receptor status (estro-
“standing trial.” Indeed the “trial” may be an unending gen- or progesterone-receptor positive versus neither),
screening process. HER2-positive versus not, and MammaPrint high-plus versus
Even greater efficiency can be achieved by including a other. Table 2 also shows the 10 signatures (or indications)
range of types of patients while adaptively identifying and/ considered for the therapies in the trial. These are biologically
or confirming which patient/biomarker subsetsdif anydbe- plausible combinations of the 8 basic subtypes.
nefit from which of the therapies: an “indication finder.” The numbers of patients assigned to each therapy both
Another efficiency is to modify the randomization probabil- within tumor subtypes and overall are determined by the trial
ities based on the data accumulating in the trial to increase results. The totality of results determine when a therapy grad-
the probability of assigning better performing therapies uates from the trial or drops for futility. These results are used
within patient subtype: an “adaptive design.” This has the to find a Bayesian predictive probability of success in a future
effect of moving better performing therapies through the phase 3 clinical trial (Berry, 2006). In particular, a therapy grad-
trial faster. Finally, since primary clinical endpoint informa- uates from I-SPY 2 if the predictive probability that it will show
tion may be delayed, statistical models of tumor burden statistical superiority over control in an equally randomized
over time can enable more informed adaptive decisions 300-patient phase 3 trial restricted to its graduating signa-
about therapeutic benefits, again depending on patient ture(s) is at least 85%. Upon graduation accrual stops in all pa-
subtype. tient subtypes. Stopping accrual is easy to effect by setting all
A prototypic example of the above description I-SPY 2. the therapy’s randomization probabilities to 0. This happens
(Berry, 2012; I-SPY 2 Trial, 2015; ClinicalTrials, 2015; Barker silently with only selected individuals at the therapy’s owner
et al., 2009) It is a phase 2 adaptively randomized “umbrella being informed. Announcements of graduations to the trial’s
trial” in high-risk neoadjuvant breast cancer. Specifically, it investigators and the general public occur after all the patients
is a standing platform trial that seeks to identify each ther- assigned to the therapy and the concurrently randomized
apy’s indication, or biomarker “signature.” This is subset of control patients have had surgery, which is approximately 6
patients who benefit from the therapy. The primary endpoint months after graduation. Similarly, accrual stops for futility
in I-SPY 2 is pathologic complete response (pCR), an endpoint if the predictive probability of statistical superiority in a future
that the U.S. FDA has come to accept for accelerated approval trial is less than 10% for all 10 signatures.
in neoadjuvant breast cancer. (fda, 2015; Driving Biomedical Figure 1 is a flow chart showing the adaptive process used
Innovation, October 2011) Moreover, to get an early read in I-SPY 2.
regarding the likelihood that each patient will be a pCR on As of this writing I-SPY 2 has evaluated or is still in the pro-
her assigned therapy, we use an adaptive statistical longitudi- cess of evaluating 8 experimental therapies from 6 different
nal model of tumor burden based on MRI volume measure- pharmaceutical companies, and more therapies are in the
ments at 3 weeks and 12 weeks after the start of therapy. queue. Two therapies have graduated to phase 3: veliparib/
I-SPY 2 considers experimental therapies in primary dis- carboplatin with a signature of triple-negative disease, (Rugo
ease. It bucks the tradition of first using experimental et al., 2013; Helwick, 2014) which is Signature 37 in Table 2,
Table 2 e There are 8 biomarker subtypes in I-SPY 2. These are defined by hormone-receptor (HR), HER2, and MammaPrint (MP) statuses.
There are 10 biomarker signatures. These are combinations of the 8 subtypes as shown. Signature prevalences are estimated from I-SPY 1,
(Esserman et al., 2012) which like I-SPY 2 focused on high-risk primary breast cancer treated neoadjuvantly.
Biomarker Subtypes of breast cancer: HR, HER2, MP Estimated
signature prevalence
þþþ þþ þþ þ þþ þ þ
1: All X X X X X X X X 100%
2: HRþ X X X X 49%
3: HR X X X X 51%
4: HER2þ X X X X 37%
5: HER2 X X X X 63%
6: MPþ X X X X 48%
7: a X X 34%
8: þa X X 17%
9: þþa X X 20%
10: þa X X 29%
Characteristics Trials
a b c d
I-SPY 2 BATTLE Lung-MAP UK Matrix NCI-MATCHe NCI-MPACTf IMPACT 2g MICATh
M O L E C U L A R O N C O L O G Y 9 ( 2 0 1 5 ) 9 5 1 e9 5 9
Treats all comers X X X X
Combination X X ? X X
therapies
Sequential therapies X
Regimens enter & X X X
leave trial
Learn off-target X X X
effects
Pair regimens with X X X
biomarkers
Common control arm X X X
Adaptive X X X X
randomization
Adaptive sample size X X X
Early “curable” disease X
Registration endpoint X X X X ?
Longitudinal disease X X
modeling
Seamless phases 2/3 X
a Berry, 2012; I-SPY 2 Trial, 2015; ClinicalTrials, 2015; Barker et al., 2009.
b Kim et al., 2011; Berry et al., 2012.
c LUNG-MAP, January 2015; Lung-MAP, January 2015.
d Cancer Research UK Science, 2015.
e NCI-MATCH, January 2015.
f NCI-MPACT, January 2015.
g IMPACT 2, January 2015.
h MICAT, January 2015.
18780261, 2015, 5, Downloaded from https://febs.onlinelibrary.wiley.com/doi/10.1016/j.molonc.2015.02.011 by Reprints Desk Inc, Wiley Online Library on [18/06/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
M O L E C U L A R O N C O L O G Y 9 ( 2 0 1 5 ) 9 5 1 e9 5 9 957
the latter including 12 from AstraZeneca and 2 from Pfizer types may have different historical response rates because
(Torjesen, 2014; Harrison, 2014). An agent that “shows prom- standard therapies and their effectiveness differ by disease.
ise” in 15e30 patients will lead to a separate, larger trial. Spe- Simply pooling trial results across tumor types as in NCI-
cifics of the UK Matrix design are not publicly available. MATCH is not reasonable. For example, two trials with iden-
MICAT is the Melanoma International Consortium for tical results by tumor type could have very different overall
Adaptive Trials. As indicated in Table 3, a highlight of the trial response rates because they included different proportions
is its evaluation of sequential use of therapies as well as ther- of the various tumor types.
apies in combination. The trial is under development and is In the other extreme, it makes no sense to regard two
not yet accruing patients. different tumor types that have the same molecular aberra-
NCI-MATCH uses Next Generation Sequencing of solid tu- tion as being completely distinct entities. The two tumor types
mors and lymphoma to identify patients with molecular aber- may respond similarly to targeted therapy. Suppose one tumor
rations who have progressed on at least one therapy. Patients type has a 15% historical response rate and a second type has
are excluded if they have an aberration and tumor type for a 30% historical rate. Suppose there are 9 (60%) responders of
which a targetable agent has been FDA-approved. Just as for 15 patients in the first type and 6 (30%) responders of 20 pa-
Lung-MAP and UK Matrix the MATCH master protocol in- tients in the second type. The sample sizes are small. But
volves triaging, with no borrowing across sub-protocols. There doubling both historical rates is more suggestive of an
will be up to 25 sub-protocols, each of which will be traditional improvement in both types than when viewing the treatment
phase 2 single-arm assessments with approximately 30 pa- effects in the two types separately. Put another way, the con-
tients irrespective of the organ types of the tumors. The fidence intervals of treatment effects in both tumor types
reason for question mark in Table 3 is that combinations are should be narrow to account for the similarity of treatment ef-
possible but there are no specific plans as yet for such thera- fects in the two tumor types.
pies. The primary endpoint of each sub-protocol is overall In another type of example, suppose three tumor types with
response rate, with power to distinguish between 25% and the same historical rate of 20%, suppose there are, respectively,
5% (NCI-MATCH, January 2015; Abrams et al., 2014). So each 12, 8, and 8 responders out of 20 patients of each type. Given the
sub-protocol is a basket trial in which all patients are regarded small sample sizes the observations 60%, 40%, and 40% are not
to be exchangeable across tumor types, and in particular as very different, and the three true rates are probably closer
having the same underlying response rate irrespective of tu- together than the observed rates. The first one might be 55%,
mor type. (The next section deals with basket trials and does say, and the other two might be 43%.
not make this highly questionable assumption.) This kind of estimation is called “shrinkage” or “regression
MPACT is an ongoing “pilot trial” considering four targeted to the mean.” It has a surprisingly powerful characteristic. In
therapies within three molecular pathways. The objective is to the 1950s Charles Stein showed that when there are at least
test the targeting strategy. The co-primary endpoints are 3 groups shrinkage estimates improve the na€ıve approach of
response rate (CR þ PR) and PFS (NCI-MATCH, January 2015). no shrinkage regardless of the true state of nature (Stein, 1956;
The goal is to have 180 patients overall, with 2:1 stratified James and Stein, 1961). In particular, the usual (no shrinkage)
randomization within molecular pathway, favoring targeted estimator of group means has greater mean squared error
therapy. regardless of the true values. Expressed colloquially, if some-
IMPACT 2 is a single-institution trial that was motivated by one goes through life and never shrinks their estimates, there
a non-randomized comparison of phase-1 patients of any tu- are a countless number of people who do shrink and whose
mor type treated with targeted agents whose cancers estimates are ordained to be better overall regardless of the
harbored the targets with matching patients who were not underlying state of nature.
treated with a targeted agent. (Tsimberidou et al., 2012) The shrinkage approach has led to Bayesian hierarchical
IMPACT 2 employs randomization to targeted versus non- modeling in which there is “partial borrowing” of results in
targeted therapy in a phase-1 setting similar to NCI-MPACT, the various groups, and where the amount of borrowing is
but with a much broader range of what may be targetable, greater across groups that have similar results. This method
similar to NCI-MATCH. The sample size is 200 patients, with has been shown to enable building adaptive “indication-
the first 100 assigned using balanced randomization. Subse- finder” clinical trials that are efficient in terms of saving re-
quent patients will be assigned using adaptive randomization sources while at same time providing better estimates (Berry
within the subtypes of patients defined by biomarker type and et al., 2013). The method applies for randomized trials and
tumor type. The analysis and the adaptive features of the other types of endpoints, including times to an event.
design uses Bayesian hierarchical modeling, an extension of Figure 2 shows hypothetical results for a clustering hierar-
the modeling presented in the next section. chical borrowing method that is being used in industry basket
trials. It exhibits regression toward the overall mean as well as
toward the cluster means for positives and negatives
6. Basket trials separately.
R E F E R E N C E S
Mullard, A., 2014. Multicompany trials adapt to disciplines drug development. Cancer Discov. 1, 17e20. http://dx.doi.org/
beyond cancer. Nat. Med 20, 3. http://dx.doi.org/10.1038/ 10.1158/2159-8274.CD-11-0036.
nm0114-3. Rugo HS, Olopade O, DeMichele A, et al. Veliparib/carboplatin
NCI-MPACT: Molecular Profiling-Based Assignment of Cancer plus standard neoadjuvant therapy for high-risk breast
Therapy for Patients With Advanced Solid Tumors. https:// cancer: First efficacy results from the I-SPY 2 trial. 2013, San
clinicaltrials.gov/ct2/show/NCT01827384 (accessed 1 January Antonio Breast Cancer Symposium. Abstract S5e02. Presented
2015). December 13, 2013. http://sabcs.org/Documents/
Neratinib graduates to I-SPY 3. Cancer Discovery, OnlineFirst 2013SABCSProgramWeb.pdf (accessed 1 January 2015).
April 7, 2014;http://dx.doi.org/10.1158/2159-8290.CD-NB2014- Stein, C., 1956. Inadmissibility of the usual estimator for the mean
055; http://cancerdiscovery.aacrjournals.org/content/early/ of a multivariate normal distribution. Proc. Third Berkeley
2014/04/04/2159-8290.CD-NB2014-055.full (accessed 1 January Symp. Math. Statist Prob 1, 197e206.
2015). Stratified medicine and the lung cancer ‘Matrix’ trial e part of a
NCI Molecular Analysis for Therapy Choice Program (MATCH) & cancer care revolution. Cancer Research UK Science Blog.
Pediatric MATCH. http://www.cancer.gov/clinicaltrials/ http://scienceblog.cancerresearchuk.org/2014/04/17/stratified-
noteworthy-trials/match (accessed 1 January 2015). medicine-and-the-lung-cancer-matrix-trial-part-of-a-cancer-
Park, J.W., Liu, M.C., Yee, D., et al., 2014. Neratinib plus standard care-revolution/(accessed 1 January 2015).
neoadjuvant therapy for high-risk breast cancer: efficacy Torjesen, I., 2014. Large personalised medicine trial in lung cancer
results from the I-SPY 2 TRIAL. In: Proceedings of the 105th heralds new research partnership. BMJ 348, g2837. http://
Annual Meeting of the American Association for Cancer www.bmj.com/content/348/bmj.g2837 (accessed 1 January 2015).
Research; 2014 Apr 5e9; San Diego, CA. Philadelphia (PA). Tsimberidou, A., Iskander, N.G., Hong, D.S., et al., 2012.
AACR. Abstract nr CT227. http://www.abstractsonline.com/ Personalized medicine in a phase I clinical trials program: the
Plan/ViewAbstract.aspx?mID¼3404&sKey¼ed5341ec-ef13- MD Anderson Cancer Center Initiative. Clin. Cancer Res. 18,
493e-8844-e14be322679c&cKey¼d9b44ad7-1673-4ec9-b360- 6373e6383.
11254fbb92be&mKey¼6ffe1446-a164-476a-92e7-c26446874d93 US Department of Health, Education, and Welfare. The Belmont
(accessed 1 January 2015). Report: Ethical principles and guidelines for the protection of
Rubin, E.H., Anderson, K.M., Gause, C.K., 2011. The BATTLE trial: a human subjects of research, hhs.gov/ohrp/humansubjects/
bold step toward improving the efficiency of biomarker-based guidance/belmont.html (1979, accessed 1 January 2015).