Inside an Oncology Statistician's Toolkit

February 9, 2017



 In this blog, Adam Hamm, PhD, Director Biostatistics at Cytel shares some of the most important knowledge he uses in  his day to day work as a biostatistician working extensively in oncology research. Adam has broad experience with statistical analysis and methodology over all phases (I-IV) of development, in particular working in the oncology arena. 

 Adam Hamm Head Shots.jpgAs a Director of Biostatistics at Cytel, I work on design, statistical analysis and reporting projects for a range of biotechnology and pharmaceutical sponsors. During my career, I’ve developed a particular focus on oncology trials, so in this blog I’ll share some insights into the  knowledge which I have found particularly vital  as a biostatistician working in this area. This knowledge spans specific statistical methodologies and understanding of the clinical issues across the phases of clinical development. The summary is not exhaustive, but provides a glimpse into the broad exposure which is needed for a biostatistician to develop a fully rounded understanding in the area. To learn more, read on...


Early Phase Oncology Research

With Maximum Tolerated Dose (MTD)  Finding a primary goal at this early stage of development, it’s critical for the biostatistician to have a strong understanding of what is meant by a “Dose Limiting Toxicity” and how this is defined across different oncology indications and types of studies, as its interpretation can be very broad in the clinical space.

There are a range of design approaches used to determine MTD and the biostatistician needs to have a secure knowledge of these designs and, importantly,  their implementation.

Some approaches are:


  • A routine, rule based approach which is very common in this setting. The decision criteria are based on the number of toxicities seen at a particular dose
Continual Reassessment Method or Bayesian Logistic Regression Modeling (BLRM).
  • Bayesian-based modeling approaches which require statistical knowledge in Bayesian methods
Modified Toxicity Probability Interval (mTPI)
  • A “hybrid” approach which holds some similiarities with 3+3 but has a basis in Bayesian methodology.

 Read more about adaptive dose escalation approaches

Phase 2 Oncology (MTD Dose Expansion/Dose Response)

At this stage, from a clinical point of view, it is important for the statistician to thoroughly understand how efficacy responses are defined in oncology trials, as opposed to early phase when the focus is on dose-toxicity responses.  For solid tumor types, the RECIST (Response Evaluation Criteria in Solid Tumors) guidelines were originally introduced in 2000  to provide a standard for evaluating tumor response.  Other clinical endpoints (composite response/remission endpoints for example) are used in hematological cancers such as Acute Myeloid Leukemia and it is important for the biostatistician to be familiar with the components used to derive these endpoints.

Phase 2 oncology studies often involve a multi-stage approach. Knowledge of multi-stage designs such as Simon two-stage designs, which are commonly used in this area, is critical. Perhaps the most crucial reference in this regard is Simon’s paper on the subject ( 1 ) (Simon, 1989). This paper helps the oncology statistician understand the decision making process between stages and so is a critical reference when designing multi-stage studies. Phase 2 oncology studies may also employ a design that utilizes Bayesian Predictive Probability. This can be very useful for sponsors who want to perform continuous monitoring to end futile studies earlier.


Phase 3 (Confirmatory) Oncology

At Phase 3, the requirements for clinical understanding may be somewhat similar to Phase 2, although additional endpoints are also employed. A detailed knowledge of  specific time-to-event endpoints used in oncology ( for example, Overall Survival, Progression Free Survival) and other endpoints such as Objective Response Rate is clearly critical.  It’s increasingly common for Phase 3 oncology studies to have an adaptive component, and therefore the biostatistician working in an oncology setting needs to have developed a robust understanding of interim analyses in order to design and implement these types of trials. A knowledge of the theory behind sample size reassessment approaches such as the promising zone design (2 ) (Mehta and Pocock, 2011)  is also required, along with an understanding of how conditional power is derived and calculated.


 Read more about the promising zone design and VALOR case study


Cytel's  biostatistics, data management and medical writing groups have extensive experience designing and implementing oncology trials and understand the specific trial design, and data collection and analysis requirements within this space. To learn  out more about our clinical research services click the button below.

Clinical Research

 Cytel will be attending ASCO June 2-6, 2017 in Chicago, Illinois.  To arrange a meeting with one of our experts at the event contact us via the button below.

Contact Us

1) Simon, R. (1989) ‘Optimal two-stage designs for phase II clinical trials’, Controlled clinical trials., 10(1), pp. 1–10.

2)  Mehta, C. and Pocock, S. (2011) ‘Adaptive increase in sample size when interim results are promising: A practical guide with examples’, Statistics in medicine., 30(28), pp. 3267–84.


About the Author

Adam Hamm, PhD is a Director Biostatistics at Cytel and has extensive experience with statistical analysis and methodology over all phases (I-IV) of clinical research. Dr. Hamm brings more than 12 years of experience in clinical trials. His areas of expertise include strategic statistical consulting, development of statistical sections of protocols, development of adaptive design plans including simulations and modeling, development of sample size reports, randomization schemes and statistical analysis plans, and clinical study reports and standalone statistical reports in support of clinical research in numerous therapeutic areas, including oncology. Adam has extensive experience in early phase oncology trials with adaptive designs. He implemented the CRM in a phase I trial studying solid tumors. He has also analyzed studies utilizing Simon two-stage designs, and consulted on studies that used the mTPI method and Bayesian predictive probability with continuous monitoring. In addition, Adam has worked on two phase 3 breast cancer studies with interim analyses triggered by event milestones.