2020 Issues in Design of Translational Research Trials

Susan Hilsenbeck, PhD, Baylor College of Medicine, Houston, Texas, USA

Fellow Summary

Authored by Alexander Bagley

This lecture focused on several important issues related to design of translational research studies of targeted therapies, including selection of biomarkers, outcomes, and study design. Regarding biomarker development, Dr. Hilsenbeck gave a helpful overview of biomarker development in the different phases of clinical trials, with exploration of biomarkers generally performed in Phase I (analytic validation), relationship of endpoints to biomarkers explored in Phase II (clinical validation), and using more established biomarker assays in Phase III trials (clinical utility). The co-development of a drug with a companion biomarker provided a helpful framework for discussing this topic. Ultimately, for a biomarker to be widely utilized it should result in a clinically actionable outcome and be “fit for purpose”, which could include aiding in diagnosis, prognosis, or prediction of treatment response or acting as a surrogate endpoint for response to therapy. Dr. Hilsenbeck then discussed the major roles for biomarkers in a clinical protocol: Integral, Integrated, and Exploratory. Whereas integral biomarkers must be assessed in order for a study to proceed (e.g., define eligibility or assign treatment), integrated biomarkers are included to help test pre-specified hypotheses related to biomarker/assay validation, and exploratory biomarkers are intended to help understand biological mechanisms or contribute to biomarker development. Dr. Hilsenbeck reviewed novel early-phase adaptive trial designs incorporating biomarkers to enrich selected populations under study.

Dr. Hilsenbeck offered a practical framework for thinking about how the availability of biospecimens in relation to treatment delivery (neoadjuvant, adjuvant, or metastatic settings) can help guide the appropriate selection of nominal (e.g., objective response, pathological complete response, toxicity) versus interval (e.g., disease-free survival, progression-free survival, overall survival) endpoints and objectives. One of the important considerations she emphasized was that interval endpoints such as progression-free survival provide more information per outcome in comparison to binary endpoints, and therefore trials using interval endpoints require smaller sample sizes. This is an important point that can help to minimize the number of patients needed for a given trial. 

She concluded with a discussion of several potential pitfalls in the design of clinical trials and translational research more generally: multiple looks for early-decision making, unplanned subgroup analyses, and correlative science using arbitrary cut-points for continuous variables. She emphasized that in order to avoid these pitfalls, careful upfront planning is required when designing your trial. Measures including pre-specifying hypotheses to be tested, planned interim analyses using Simon’s two-stage or Bayesian methodologies, adjustments for multiple comparisons, and seeking independent validation of post-hoc analyses should be attempted to strengthen the conclusions of a study.

Overall, Dr. Hilsenbeck provided a clear lecture that addressed several elements of clinical trial design that are important for clinical investigators to consider early in the development of their protocols. Biomarkers already play a significant role in oncology trials, and this is anticipated to continue not only for targeted therapies, but for immunotherapy, radiation therapy, and novel combination therapy approaches. This lecture provided a helpful way to think about how biomarkers can be meaningfully incorporated into clinical trials and how to navigate potential challenges when designing and interpreting such studies.

Fellow Summary

Authored by Charlie Kuang

There are many ways to define translational research, but some commonly agreed upon features are the focus on turning laboratory findings into meaningful health outcomes and the focus on targeted therapy.  Questions that frequently arise and need to be addressed are: how to include predictive biomarkers, what is the best endpoint, what are the best study designs and statistical characteristics, and how does one ensure rigor and reproducibility.

Much of the presentation and discussion focused on incorporation of biomarkers into translational studies.  The proposed levels of biomarker investigation are as follows: Phase 1 for exploring biomarkers, Phase 2 for determining relationship of biomarkers to outcomes, and phase 3 for locking down biomarker assays.

Good questions to ask when developing a biomarker:

  • What is a biomarker (a characteristic that is objectively measured as an indicator of normal or pathogenic biological processes, or pharmacological response to an intervention)?
  • How will the biomarker be used?
  • How good does the biomarker have to be?
  • What endpoints and study designs are appropriate?

Another way to look at the way biomarkers are tested at different phases of clinical trials: Phase 1 for analytic validation (demonstrate that the biomarker assay is robust), phase 2 for clinical validation (demonstrate an association between a biomarker test result and pathophysiologic state), phase 3 for clinical utility (demonstrate that biomarker use leads to improved clinical outcome).  Keep in mind that different biomarkers have different purposes; some are prognostic, predictive, indicate risk, suggest disease progression, or suggest a pharmacodynamic outcome.

Biomarkers may be incorporated in different ways into a study:

  • Integral: assessed in order for study to proceed, needs to be CLIA-certified and may require IND and IDE; could be used for trial eligibility, treatment assignment, or stratification
  • Integrated: intended to validate assay and biomarker, tests hypothesis with pre-specified plans
  • Ancillary/exploratory: for developing biomarker or assay, to understand the biology or the agent

One way to perform a biomarker driven trial is to use a Simon 2 stage with integral biomarker required for study participation and test only marker positive patients; however this runs the risk of missing out on evaluating all the biomarker negative patients, who may also benefit from this study/intervention.  One way to improve on this may be to use an adaptive/multistage design: 1. test for marker, randomize both biomarker+ and marker- groups to all treatments, and 2. allow for marker negative patients to drop out upon futility analysis based on an adaptive design.  In comparing the two designs in detail (as far as execution goes): 

  • Typical Simon 2 stage: one arm, all marker+, all treated with one arm, historical control, BUT there are no marker- patients.
  • Adaptive Simon 2 stage: enroll both marker- and + pts, have separate stopping rule for marker- group to determine if should keep enrolling marker- or drop out; may need different stopping rules for the early accrual of marker- cohort, marker+ cohort, and combined cohort if you end up accruing nonmarker selected population.

The difference between objectives and outcomes/endpoints (CT.gov now requires very clear documentation of these!)

Objective: purpose of trial (…to estimate the relationship between treatment X and response…)

Outcome/endpoint: measurement on participant (i.e. more specific and technical than objective)

Also consider comparative trials with biomarkers as outcomes: “window of opportunity” trials with short exposure to an agent, then a clinically indicated surgery or an additional procedure for biopsy/tissue.  The biomarker can be the primary or secondary outcome in this design and usually requires a much smaller sample size than for trials with a nominal outcome.

My take-home points: understand the ways to integrate biomarkers at different phases of clinical investigation, consider adaptive trial designs for phase II trials incorporating biomarkers and DON’T forget about accruing biomarker negative patients, and consider biomarker as primary or secondary outcomes in window of opportunity trials.

Fellow Summary

Authored by Abhishek Maiti

Dr. Hilsenbeck provided an excellent overview of statistical aspects of designing translational research trials. She discussed different aspects of incorporating biomarkers, focused outcomes, novel designs, sample size calculation, power, and common pitfalls. Translational research covers a spectrum of research with the goal of translating laboratory findings to clinical advances. The talk emphasized distinguishing between predictive versus prognostic biomarkers and the importance of determining the actionable potential of a biomarker. Predictive biomarkers essentially means biomarkers which predict the likelihood of a patient responding to a given treatment and hence help in guiding treatment; in contrast, prognostic biomarkers may provide insights into disease course and overall prognosis. It is crucial to clearly define the role a biomarker would play in the clinical trial, including whether it will determine eligibility or treatment, or whether the trial will validate that biomarker, or if that trial will help to further understand the role and value of the biomarker. Further steps are needed to understand the performance of the assay, optimal cut-offs, clinical validity, and the applicability of the biomarker in patient care beyond research. Given requirements of the clinical trial registries, biomarker-driven trials need better articulation of the purpose of a biomarker beyond simply an exploratory end point. Innovative adaptive designs can further improve such trials with capability to yield reproducible findings while minimizing sample size. One critical pitfall discussed was regarding multiple post-hoc testing and the potential for finding of false positive results by chance. There is a need for biomarker studies to pre-specify hypotheses to be tested and incorporate statistical methods to address such multiple hypothesis testing and account for the false discovery rate.

Fellow Summary

Authored by Makoto Nishino

Dr. Hilsenbeck’s lecture was held on Day3 of the STOFF meeting. As I knew her, and as she is a very well-known biostatistician, I was really looking forward her talk. She started her lecture by reviewing her life story: why she started to work as a biostatistician. The definition of translational research is that it aims to “translate” findings in fundamental research into medical practice and meaningful health outcomes. Next she discussed three fundamental questions from statistical points of view through phases of clinical trials; How to include a predictive biomarker, how to select outcome measure of the trial (ORR, PFS or OS), how to choose appropriate study designs, power, and sample size. As the phases of trial progress, the importance of biomarker development was discussed. There will be many questions during development, such as what is a biomarker itself, how to use the biomarker, and how much quality the biomarker has to have. At phase1, exploration of the biomarker is essential to finding the safety dose. At Phase2, the relationship between biomarker and efficacy will be measured. At phase3, not only do we find definitive evidence of efficacy, we also lock down biomarker assays.

A biomarker is a characteristic that is objectively measured as an indicator of normal biological processes, pathogenic processes, or pharmacological response to a therapeutic intervention. In order to validate the biomarker, we will consider two aspects, analytic and clinical; meaning to address the specific setting that is clinically actionable and to demonstrate its effectiveness, uptake, and improved clinical outcome. Biomarkers often have various purposes: to assess risks/susceptibility, to make diagnosis, to predict treatment benefit, to forecast prognosis, to show disease progression, and to guide dose adjustment. Often there is a discussion of the prognostic and predictive biomarker. The biomarker often plays an important role in clinical trials by integrally assessing in order for a study to proceed. Another role biomarkers play will be integrated; one that is to intend to validate assays to test hypotheses with pre-specified plans. Sometimes trial data are used to develop biomarker or assays to understand agents or biology; we often say refer to this process as reverse translational study. There was some interesting discussion on sample size. If we set alpha for one sided 5% and power for 80%, a traditional comparative trial needs almost 4 times compared to single arm study. As for the timing of biopsy, we often confront the “window of opportunity” in neoadjuvant settings that often have a short exposure to investigational agents before moving to surgery or standard therapies.

Fellow Summary

Authored by Gabriela Sanchez-Petitto

Translational research involves the integration of basic research, patient-oriented research, and population-based research with the aim of improving healthcare. Clinical trials are research studies that include human volunteers.

A biomarker is a biological molecule found in blood, other body fluids, or tissues that is a sign of a normal or abnormal condition or disease. A biomarker, also called molecular marker and signature molecule, may be used to evaluate how well the body responds to a treatment for a disease or condition. Biomarker validation goes through a process of analytic validation (assay performance), clinical validation (association between test results and pathophysiologic state), and clinical utility. Biomarkers have different purposes; some are useful for diagnosis (PSA), prognosis (oncotype), predictors of treatment benefit (ER+ status), disease progression (PSA), or surrogate endpoint (response rate), among others.

Phase I studies of a new drug are usually the first that involve people; the aim is to find the highest dose of the new treatment that can be given safely. It can also involve the exploration of new biomarkers. If a new treatment/biomarker is found to be reasonably safe/useful in phase I clinical trials, it can then be tested in a phase II clinical trial to evaluate for response and clinical benefit and the relationship of the response to the biomarker. Treatments that have been shown to work in phase II studies usually must succeed in phase III, which compares the safety and effectiveness of the new treatment against the current standard of care and also involves locking down biomarkers assay. Drugs approved by the FDA are often watched over a long period of time in phase IV studies.

Window of opportunity studies are trials in which patients receive one or more new compounds between their cancer diagnosis and standard treatment. Often times, tumor samples before and after the investigational treatment are collected for translational research, along with pre- and post-treatment imaging. Window of opportunity trials may expedite drug development, improve our understanding of pharmacodynamic parameters, and help to identify biomarkers for better patient selection. 


  • Biomarkers are measurable indicators of the presence of a specific condition, prognosis, or likelihood of response to treatment.
  • Clinical trials can explore the clinical benefit of the biomarkers.
  • Biomarkers help identify the ‘window of opportunity’ for cancer chemotherapy timing.

Fellow Summary

Authored by Elaine Walsh

Dr. Hilsenbeck discussed the concept and definition of translational research and described it as being a continuum incorporating laboratory findings, clinical trials, and the point of patient care. In particular, Dr. Hilsenbeck focused her talk on early translational research with a particular emphasis on targeted therapy. Dr. Hilsenbeck discussed how to include predictive biomarkers in translational research and what endpoints we should be using in these studies. In addition, she discussed study designs, power, and sample size as well as pitfalls and concerns pertaining to translational research of targeted therapy.

Dr. Hilsenbeck defined biomarkers and suggested that each investigator ask how the biomarker will be used, how good the biomarker has to be, and what study design and endpoints are appropriate. Dr. Hilsenbeck suggested that biomarkers should be incorporated at each phase of clinical trials. In the phase I setting, biomarkers can be explored; in the phase II setting, the relationship between therapies (or disease) and biomarkers can be studied, and in phase III studies, biomarker assays can be confirmed in the setting of therapies. Validation of biomarkers was discussed in three stages: analytic validation, where the assay performance is demonstrated, clinical validity, where the association between the test result and pathophysiologic state is demonstrated, and clinical utility, where effectiveness, uptake, and improved clinical outcomes are demonstrated with the use of the biomarker. Analytically and clinically valid biomarkers are shown to address a specific setting, are clinically actionable, and are “fit for purpose”. The topic of being fit for purpose was addressed more specifically, and examples were discussed for diagnostic, prognostic, and predictive biomarkers among others.

Dr. Hilsenbeck went on to discuss very practical examples of biomarker studies, such as looking at marker-positive versus marker-negative cases, differing treatments based on marker expression, and outcomes based on marker expression. Biomarkers play an important role in studies: studies may be integral, and biomarkers need to be assessed in order for a study to proceed, studies may be integrated and intended to validate assays, and biomarkers or studies may be exploratory, whereby the trial data is used to develop biomarker or assays, understand the drug in question, or the disease. In phase II studies in particular, biomarkers can be used to give further information about toxicity of targeted therapies, provide evidence of efficacy, select the right route and schedule, and may provide evidence for companion biomarkers. However, trial designs need to be cautious in expecting too much data to be derived from biomarker components.

Finally, Dr Hilsenbeck went on to discuss biospecimens, sample sizes, trial design, biomarker cut points, and potential downfalls in trial design and analysis. In summary, Dr. Hilsenbeck recommended that researchers pre-specify the hypothesis to be tested in order to plan appropriate sample sizes, plan for interim analyses, adjust for multiple comparisons (or estimate false discovery rates), and treat post-hoc analyses with caution.