Call us: + 31 (0)20 - 4350 580

Home > News > Blog > The changing landscape of oncology clinical tri...

The changing landscape of oncology clinical trials - AACR 2017

The changing landscape of oncology clinical trials - AACR 2017

The 2017 annual meeting of the American Association for Cancer Research (AACR) could not start any better. A fascinating plenary session included Angela Belcher (Massachusetts Institute of Technology) and Bert Vogelstein (Sidney Kimmel Comprehensive Cancer Centre). The former, discussed how understanding natural design and harnessing directed evolution can help generate nanomaterials that can be applied to the detection of tiny ovarian tumours. The latter, made an impassionate case that irrespective of the source of cancer-causing mutations, primary prevention is the best way to combat cancer; nevertheless chemoprevention and earlier detection and intervention, will bring about the 2nd revolution in treating the disease, in the (near) future.

A great proportion of remarkable research presented during the meeting concerned translational and clinical research and emphasis was placed on how the clinical trial landscape is changing, or better, is evolving. 

Clinical trial methodology is increasingly adapting to address the questions posed by targeted and immune-therapies. At the beginning of the clinical pipeline and in early clinical trials, traditional rule based (e.g. 3+3), dose escalation designs still hold prominent place. It is however widely accepted that designs based on Continual Reassessment Method (CRM), such as the modified CRM and the Escalation With Overdose Control (EWOC), and newer adaptive designs such as modified toxicity probability interval (mTPI) are more suited to the development of these novel agents (Ji and Wang 2013, Iasonos and O'Quigley 2014, Wages at el., 2016). 

Key advantages of these approaches include the ability to determine the Maximum Tolerated Dose (MTD) while taking into account the whole of the patient population rather than only those experiencing Dose Limiting Toxicities (DLTs) and allowing fast dose escalation in sub-therapeutic dose ranges. The mTPI design more specifically was born out of the need to handle late onset DLTs, as is often the case with immunotherapies.

Jiawen Zhu (Roche) described two examples of using such approaches highlighting how they allow for flexible cohort size with most cohorts including 3-8 patients. Moreover, even though model recommendations for dose escalation where found in one study to be initially too aggressive, clinical data input allowed for timely adjustments. Paul Chapman (Memorial Sloan Kettering) argued that currently accepted norms, such as the definition of DLTs and subsequently MTD, which are based on the Common Terminology Criteria for Adverse Events (CTCAE) and focus mainly on grade 3 and 4 events, should not be used for clinical decision making with immunotherapies. Similarly, exclusion criteria such as previous malignancies, brain metastasis and levels of creatinine (> 1.5 ULN) and platelets (<100,000) are there for historical reasons with limited relevance to todays clinical trials. 

Further, down the clinical research pipeline, basket and umbrella trials searching for signs of efficacy are ever expanding and increasing in complexity. Shivaani Kummar (Stanford University) reviewed some key examples. The Investigational New Drug (IND) Application for pembrolizumab (Phase I) was submitted in December 2010. Two and half years later and after 8 protocol amendments, over 1,100 patients were recruited in 9 expansion cohorts. In September 2014, data from this trial supported the accelerated approval of pembrolizumab in melanoma and in October 2015 the approval in Non-Small Cell Lung Cancer. Amendment 9 of the study will be carried out in 10 parts with each having two phases and patient numbers increasing from 40 to 400. She cautioned that the failure of the SHIVA umbrella trial should be strong reminder that “studies need to be designed to protect patients from failures”. One of the ways to address this is to employ statistical methodologies that allow information sharing between trial arms. Mithat Gonen (Memorial Sloan Kettering Cancer Centre) suggested that this could reduce samples sizes by 10-30%. It could however require 5-10% more patients if the treatment works in only one of the arms but this would be a modest price to pay considering the potential benefits. 

Platform trials such as GBM AGILE represent the (r)evolution in adaptive clinical trials. In its essence, GBM AGILE aims to use the biology of the disease to develop as many therapies for the glioblastoma as possible. The trial has employed a crowdsourcing model to collect knowledge of the disease and potential therapies, and it currently comprises a global “force” of over 130 neurosurgeons, neuro-oncologists, pathologists, imagers, basic and clinical neuroscientists and other professionals dedicated to improving the survival of patients with GBM. In terms of methodology, it is a Bayesian adaptively-randomized, multi-arm, seamless, two-stage platform trial for primary and recurrent isocitrate dehydrogenase 1 (IDH-1) wild type glioblastoma. It is being performed under a master protocol and consists of six therapy subtypes (including enrichment biomarkers) and 10 possible biomarker signatures stages. The first of the two stages of the study serves as efficacy screening whereby multiple therapies are tested in patient subtypes compared to a common control arm. Better performing therapies are assigned to that patient subtype with higher probability and graduate to stage 2 provided they perform sufficiently well in a particular signature. Seven products have graduated to stage 2 to date. The second stage is a fixed sample size, fixed randomization expansion cohort. The primary objective is overall survival and for its calculation all patients from both stages and respective controls are used. It is worth noting that all data is registration quality from the outset. The trial design ensures efficient use of eligible patients whereby each cohort competes for patients and once a signal is identified from a specific therapy, then patients flow is adjusted towards that cohort. Donald Berry (Berry Consulting) likened GBM AGILE to a time machine, as it ensures that all patient data across time is being used (longitudinal model). GBM AGILE does not only represent a paradigm shift in the management of glioblastoma but also the future of clinical trials and oncology drug development, according to Ann Barker, the Director and President of the National Biomarker Development Alliance (NBDA), the collaborative performing the study.

The ability to draw insights from emerging data on an ongoing basis appears to be central across the spectrum of adaptive clinical trials. As Shivaani Kummar (Stanford University) concluded, the importance of real-time data analysis during each trial phase is paramount in adapting to the changing landscape of oncology clinical trials.




Ji Y, Wang SJ. Modified toxicity probability interval design: a safer and more reliable method than the 3 + 3 design for practical phase I trials. J Clin Oncol. 2013 May 10;31(14):1785-91. doi: 10.1200/JCO.2012.45.7903.

Iasonos A, O'Quigley J. Adaptive dose-finding studies: a review of model-guided phase I clinical trials. J Clin Oncol. 2014 Aug 10;32(23):2505-11. doi: 10.1200/JCO.2013.54.6051.

Wages NA, Ivanova A, Marchenko O. Practical designs for Phase I combination studies in oncology. J Biopharm Stat. 2016;26(1):150-66. doi: 10.1080/10543406.2015.1092029.