Drug Development in Oncology in the Era of Precision Medicine

Wednesday, June 16, 2021

By: Prof. Christophe Le Tourneau, MD, PhD - Senior Medical Oncologist Institut Curie and Full Professor of Medicine - Paris Saclay University & Dr. Everardo Saad, MD - Medical Director - IDDI

Abstract

Cancer research is complex: trials require specific design expertise and often include biomarker and companion diagnostics, and failure rates tend to be higher than for drugs tested in other therapeutic areas. To mitigate risks associated with these complex studies, the right choice of endpoints, adequate definition of selection criteria, sensible use of safety and efficacy assessments, and state-of-the-art statistical planning, analysis and modelling tailored to oncology is vital.

Over the past decade, a better knowledge of cancer biology has led to a molecular segmentation of cancer. New trial designs have been set up to overcome the challenge of small, molecularly defined patient populations. These designs include basket trials (histologic stratification based on marker expression) and umbrella trials (molecular stratification of one specific histology), as well as strategy trials, which mix tumor types, targets and drugs in order to assess the efficiency of a treatment algorithm used to allocate drugs to patients.

Cancer can no longer be defined in conventional terms. The innovations in trial design mentioned above, alongside a better understanding of how to classify patients have the potential to accelerate research and drug development so the right therapies can be more rapidly delivered to the patients. However, new trial designs and the development of precision oncology require a change in the approach to research and data-gathering.

This article will highlight some of the challenges of new trial designs, as well as perspectives on how to speed up drug development in the era of precision medicine.

Subscribe to our e-Newsletters
Stay up to date with the latest news, articles, and events. Plus, get special offers
from American Pharmaceutical Review – all delivered right to your inbox! Sign up now!

Introduction

In oncology, it is critical to distinguish between patients with only a primary tumor and patients with distant metastases, as both require specific diagnosis and treatment. The ideal aim of any treatment is to cure patients. For patients with primary tumors, this predominantly entails the use of surgery and radiotherapy, with some drugs playing a supplemental role. The treatment of patients with distant metastases is a completely different story. Drugs play a very important role and are usually used for life, but the idea of a cure may no longer be realistic. Surgery and radiotherapy are still also used, but in a palliative way. If cure is not possible, the goal is often to transform the disease into a chronic illness, with patients receiving an effective treatment that allows for the best possible quality of life.

The ancestral paradigm of drug development in oncology started with chemotherapy agents that were discovered after the second world war, often developed by organ of origin, such as the lungs. Initially, targeted therapies generally followed the same path in that they were used in specific cancer types, often defined by organ of origin. Similarly, immunotherapy has followed suit, developed in specific cancer types without molecular selection.

Typically, the classical path of drug development has been through Phase I, Phase II and Phase III trials before market access. Phase I trials were usually non-randomized, with a primary endpoint of safety to establish the Phase II dose, in addition to looking at PK/PD parameters and preliminary evidence of efficacy. Phase II trials can be randomized or not, with a primary endpoint of efficacy and further characterization of safety. Phase II trials often include patients with a specific cancer type and in a specific clinical setting. Once the results are sufficiently promising, development progresses to Phase III, in which trials are randomized and have efficacy as their primary endpoint, while safety and quality of life are also assessed in relation to a standard of care.

In recent years, this pathway has changed to speed up the drug development process. Phase II clinical trials tend to disappear and are replaced by patient expansion cohorts performed in Phase I clinical trials.

Biomarker-Driven Drug Development

The development and qualification of biomarkers are keys to the future of drug development. Precision medicine relies on validated biomarkers that allow classification of patients by their probable disease risk, prognosis, or response to treatment. Surrogate endpoints and biomarker-based endpoints can be used to expedite drug development in precision medicine, notably in cancer research.

The move into the field of biomarkers began in the 90s, as some targets were found to be key in certain cancer types, such as HER2 in breast cancer. This led to the development of certain targeted therapies for specific cancer types amenable to molecular selection. Such drugs often move through the classical Phase I, II, and III pathway. One such example is crizotinib for lung cancer.1

A key question to ask in such trials is “should market access depend on confirmation in a randomized trial against standard of care”? Randomization is the gold standard to assess the efficacy of a drug, however it is challenging to run a randomized trial when the alteration is only present in a small number of cancer patients. In addition, some molecular alterations might be relevant in not only one cancer type but in several. In this setting, some drugs could be developed macross multiple cancer types. This highlighted the need for a different approach to clinical trials and led to the development of different types of trials, including basket and umbrella trials.

Basket Trials

Basket trials allow researchers to evaluate a drug concomitantly in multiple tumor types that have in common the expression of a given target. The main challenge is the low incidence of certain molecular alterations, as this requires screening a large number of patients in order to include only a few in the trial. This is also disappointing for patients who may be denied the drug. Another issue, even if efficacy is observed, is how to achieve a tissue-agnostic drug approval? Examples of basket trials of this type include those for the drugs larotrectinib and entrectinib, which target an alteration that is present in only 0.3% of cancer patients. Given the rarity of this and many other alterations, sufficiently powered randomized trials may be difficult to conduct. In some cases, it may be acceptable to rely on evidence from non-randomized trials for initial regulatory submission. For larotrectinib, for example, the response rate was very high, with most patients responding to the drug.2

Given that a randomized trial in every subgroup of cancer patients is not feasible, umbrella trials can be handy, as explained below.

Umbrella Trials

Umbrella trials usually offer different treatment options for patients with a given cancer type which may present with different targetable alterations. These trials are most often designed as parallel Phase II trials, but they can also include randomization. The main challenge is to prioritize among several potential alterations, especially because some cohorts might not be filled. The Up-stream trial is an example of umbrella trials for head and neck cancer patients which include biomarker-driven cohorts.3 Bias can sometimes be a problem here, with the enrolment of patients not representative of the whole corresponding population.

Ideally it would be beneficial to set up a basket-of-basket trial where molecular targets, drugs and cancer types are mixed to be able to answer relevant questions for every subgroup of patients. This is obviously impossible as it would require tens of thousands of patients. This leads to a solution known as algorithm-testing trials.

Algorithm-Testing Trials

Algorithm-testing trials mix multiple tumor types, molecular alterations and drugs. They can only conclude on the efficiency of the treatment algorithm to allocate treatments. Such trials cannot assess the efficacy of any of the drugs in any histologically or molecularly defined subgroup of patients. Given the vast molecular landscape of cancer, this is the basis of a precision medicine concept, in the sense that treatments are selected individually based on the alterations found.

Several clinical trials have evaluated the concept of precision medicine. For example, the pilot study published by von Hoff et al in JCO.4 This non-randomized study included patients with any kind of cancer with recurrent metastatic disease. Patients had to undergo tumor biopsy and molecular analysis. If a target was found, the patient received a matched therapy. The progression-free survival (PFS) ratio was used as an efficacy endpoint in that trial. However, use of that endpoint is based on the assumption that tumor growth is linear over time.

Bayesian outcome-adaptive randomization (OAR) designs for clinical trials are becoming popular. While traditional designs consider a fixed randomization probability during the trial, OAR designs make use of the outcome information obtained for patients already included in the trial to continuously update the probability. As this generally results in more patients being assigned to the ‘more promising’ treatment, based on all current information, the adaptation is suggested to increase patient-specific benefits in clinical trials. However, there is considerable controversy around the reliability of OAR. Implementation of the Bayesian (biomarker-driven) OAR designs is not trivial. Elements such as the selection of the prior distributions, early-stopping criteria, or biomarker assay accuracy strongly influence operational characteristics of the designs. Designs may offer advantages, such as a reduced total target sample size or a decrease in the variation of the accrued sample size. However, several issues have been identified, including statistical inefficiency due to imbalance in the number of patients assigned to different treatment arms and a non-trivial probability of ending up with a substantially larger number of patients assigned to the less-efficient treatment arm.

Randomized trials sit at the top of the evidence-based medicine pyramid and are the gold standard approach to establishing causal inference between treatment and outcomes. Both Phase III trials and randomized Phase II trials rely on properly conceived and implemented randomization methods, without which selection and accidental bias may preclude firm conclusions. Moreover, Phase 1 trials are increasingly randomized in this era of precision medicine and multiple competing treatment schedules that may be developed. Depending on trial size and goals, simple randomization, stratified randomization with permuted blocks, and dynamic methods of treatment allocation, including minimization, can be used to ensure properly balanced treatment arms with respect to known and unknown prognostic factors. The need for blinding adds a further layer of complexity, and additional issues come into play for randomized trials requiring this feature.

An example of a randomized trial is the SHIVA01 trial, published by Le Tourneau et al. in Lancet Oncol: Molecularly targeted therapy based on tumor molecular profiling versus conventional therapy for advanced cancer (SHIVA): a multicenter, open-label, proof-of-concept, randomized, controlled Phase II trial.5 This included patients with any kind of cancer, and patients had to undergo tumor biopsy to be sure that the molecular profile would reflect the disease biology at the time of enrolment.

The trial included patients for whom a molecular alteration was identified within one of three molecular pathways (hormone receptor, PI3K/AKT/mTOR, RAF/MEK), which could be matched to one of ten regimens including 11 available molecularly targeted agents (erlotinib, lapatinib plus trastuzumab, sorafenib, imatinib, dasatinib, vemurafenib, everolimus, abiraterone, letrozole, and tamoxifen). Patients were randomly assigned (1:1) to receive a matched molecularly targeted agent (experimental group) or treatment of physician’s choice (control group) by central blocked randomization (blocks of size six). The primary endpoint was PFS in the intention-to-treat population, which was not assessed by independent central review.

This trial demonstrated that the use of molecularly targeted agents outside their indications does not improve PFS compared with treatment of physician’s choice in heavily pre-treated patients with cancer. Off-label use of molecularly targeted agents should be discouraged, but enrollment in clinical trials should be encouraged to assess predictive biomarkers of efficacy.

Challenges and Perspectives

In treatment algorithm trials where cancer types are mixed and multiple drugs and alterations allowed, the most important factor is the treatment algorithm. Treatment algorithms must utilize innovative technology to identify molecular alterations and should not be changed throughout the trial. In addition, the same thresholds should be observed from the point of view of bioinformatics. Another crucial factor is molecular alterations and drug matching. Most trials use a unidimensional algorithm – one target per therapy. However, it has become clear that a multidimensional approach is better, considering co-existing molecular alterations, especially resistance mechanisms. Another important consideration is molecular alteration prioritization. But much more needs to be done to identify the best way to use multidimensional treatment algorithms.

Fundamentally, there is a need to improve treatment algorithms. One way to do this is to use functional assays, not only looking at the molecular alterations but also at the functional relevance of the molecular alteration. In addition, pathology may be more useful than previously thought. Importance should also be placed on molecular tumor boards for more discussion around patients to help inform analysis, next steps and identify any relevant clinical trials.

There is also a need to overcome the challenge of small patient populations. This may include utilizing novel endpoints and rely on real-world data to complement evidence obtained from clinical trials, which remain the most solid base on which to build therapeutic improvements. Data quality is crucial in that regard.

Conclusions

Cancer drug development requires innovative approaches coupled with deep knowledge of state-of-the-art conventional methodologies. Clinical cancer research is undergoing profound changes, mostly because of the advent of personalized cancer medicine. As we enter the era of precision medicine, new approaches to study design are required to quickly deliver the right therapy to the right patient. A comprehensive molecular profiling should be proposed to every patient with recurrent and/or metastatic cancer, provided patients have access to biomarker-driven drugs and clinical trials with matched therapies. Biomarker-driven cancer treatments demonstrate high efficacy, but still benefit few patients. Nevertheless, the concept of precision medicine is a reality with tissue agnostic drug approvals. On the other hand, there is not a single way to develop therapies, and the traditional development paradigm is still applicable for most drugs.

References

Kwak et al. NEJM 2010;363:1693-1703; Shaw et al. NEJM 2013;368:2385-94
Drilon et al. NEJM 2018;378:731 9; Doebele et al. Lancet Oncol 2020;21;271 82
Galot et al. ESMO 2019
Von Hoff et al. JCO 2010;28:4877-83
Le Tourneau et al. Lancet Oncol 2015;16:1324-34

About IDDI

International Drug Development Institute (IDDI) is an expert organization in biostatistical and integrated eClinical services that is committed to assisting pharmaceutical, biotech, medical devices, nutrition and academic companies in several disease areas, including oncology and ophthalmology. IDDI optimizes the clinical development of drugs, biologics and devices thanks to proven statistical expertise and operational excellence. Founded in 1991, IDDI has offices in Belgium, Boston (MA), Raleigh (NC) and San Francisco (CA).