Page 368 - Withrow and MacEwen's Small Animal Clinical Oncology, 6th Edition

346 PART III Therapeutic Modalities for the Cancer Patient

Veterinary Contract Research Organizations That being said, adaptations based on blinded analyses at interim,
Contract research organizations (CROs) are generally privately if the blinding is strictly maintained, can largely reduce or com-
pletely avoid bias.
organized and run specialty networks of veterinary hospitals that
VetBooks.ir design, conduct, and report clinical studies for the animal health Stopping Rules
industry. Examples include Animal Clinical Investigation (ACI;
www.animalci.com), The One Health Company (www.theon- Stopping rules, which are rules that terminate a clinical trial ear-
ehealthcompany.com), and VetPharm (www.vetpharm.com). lier than originally projected or within a predetermined adap-
Although oncology drug development is within their portfolios, tive trial design, can be applied to randomized phase II or phase
other medical conditions, including inflammatory and metabolic III trials. Several methods and variations have been extensively
disease, cardiovascular disease, and arthritis, are also managed. reviewed. 75–77 Stopping rules are designed to protect treatment
CROs provide multisite, pivotal, or nonpivotal studies and com- subjects from unsafe drugs, to hasten the general availability of
mercialization support to help define effective novel veterinary superior drugs as soon as sufficient evidence has been collected,
therapeutics. and to help ensure the transfer of resources and patients to alterna-
tive trials. Trials are stopped for three reasons: the investigational
Current Challenges and Opportunities in treatment is clearly better than the control, the investigational
Oncology Drug Development treatment is clearly worse than the control (less activity or more
toxicity), or the investigational therapy is not likely to be better
Oncology drug development is a difficult and costly process. It is (so-called “stopping for futility” or “futility analysis”). The meth-
estimated that only 5% to 10% of drugs entering phase I oncology ods by which stopping rules are applied usually involve some type
clinical trials ultimately are approved by the FDA, with a cost of of interim analysis that evaluates the data (by a blinded individual
between 0.8 and 1.7 billion dollars per drug accrued through the or DSMB) generated to date and makes a determination based on
development process. 55,69 The most prevalent cause of cancer drug predetermined rules. The interim data are often analyzed for con-
failure is toxicity or inactivity. For every 1000 oncology agents ditional power, which is the probability of the final study result
in development, only 40% transition from preclinical studies to demonstrating statistical significance in the primary efficacy end-
phase I, 75% from phase I to phase II, 60% from phase II to point, conditional on the current data observed, and a specific
phase III, and 55% from phase III to approval. This means that for assumption about the pattern of the data to be observed in the
every 1000 preclinical candidates, only 99 new drugs will reach remainder of the study. If a study is designed a priori to involve
the clinic. conditional power calculations of interim data, the rules for early
termination are sometimes referred to as stochastic curtailing.
Model-Based (Adaptive) Drug Development Bayesian (Continuous Learning) Adaptive Designs
Adaptive Trial Designs and Stopping Rules Adaptive trial designs can also be used to change the randomization
Adaptive trial designs 70–74 allow investigators to modify trials while weight to better performing treatment arms, add new treatment
they are ongoing based on newly acquired data and, in some cases, arms, drop poorly performing arms, or extend accrual beyond the
taking into account data generated in other trials or past trials. In original target when more information is needed. With the avail-
adaptive trial designs, interim data from a trial is used to modify ability of advanced computational techniques, a new statistical
and improve the study design, in a preplanned manner and with- methodology, the Bayesian approach, was developed that makes
out undermining its validity or integrity. In the exploratory set- statistical inferences that focus on the probability that a hypothesis
ting, an adaptive trial can assign a larger proportion of the enrolled is true given the available evidence.
subjects to the treatment arms that are performing well, drop arms In contrast, Bayesian trials differ from the frequentist approach
that are performing poorly, and investigate a wider range of doses to statistics by using available patient outcome information,
so as to more effectively select doses that are most likely to succeed including biomarkers that accumulate data related to outcome (if
in the confirmatory phase. In the confirmatory phase, adaptive available and validated), and even historic information or results
design can facilitate the early identification of efficacious treat- from other relevant trials to adapt the current trial design continu-
ments, decisions to drop poorly performing trial arms, determin- ally based on newly informed probabilities. 78–80 An example that
ing whether the trial should be terminated for futility, and making illustrates the utility of the Bayesian approach involves interim
sample-size adjustments at interim time points to ensure that the analysis applied to a randomized phase II trial of neoadjuvant
trial is adequately powered. In some cases, it might even be pos- trastuzumab, paclitaxel, and epirubicin in epidermal growth fac-
81
sible to enrich the patient population by altering the eligibility cri- tor receptor 2 (HER2/neu)–positive breast cancer. In this trial
teria at an interim time point. These also include biomarker-based initially designed to enter 164 patients (based on the frequentist
adaptive designs which enable adaptations based on the response approach to power), a Bayesian approach was used to perform an
of biomarkers, biomarker/target screening designs (e.g., enrich- interim analysis after 34 patients were enrolled; 67% of patients in
ment designs such as “basket” or “umbrella” trials). 73 the investigational treatment arm experienced complete responses
Potential benefits of adaptive designs include (1) they allow compared with 25% in the standard treatment arm. The Bayes-
72
the investigator to correct wrong assumptions made at the begin- ian predictive probability of statistical significance if 164 patients
ning of the trial; (2) they help the investigator select the most were accrued, based on the data available from these 34 patients,
promising option early; (3) they make use of information that was calculated to be 95%, and the trial was stopped and the drug
emerges outside of the trial; (4) they enable the investigator to moved to phase III early because of these promising results. The
react earlier to surprises (either positive or negative); and (5) they very first Bayesian response adaptive trial involving companion
82
may shorten the development time and consequently speed up the animals was recently published. In this trial comparing toc-
development process. However, adaptive designs may undermine eranib versus vinblastine therapy for canine mast cell tumors,
validity and integrity of the trial if not managed appropriately. patients were randomized based on the “play-the-winner” rule;