The following article is an opinion piece written by Dr. Gen Li. The views and opinions expressed in this article are those of the author and do not necessarily reflect the official position of Technology Networks.
Humanity’s progress in the understanding, prevention and treatment of cancer has been nothing short of remarkable. Whether precision medicine driven by a deeper understanding of biomarkers, immunotherapy, cancer vaccines or next-generation sequencing, oncology is a beacon of healthcare innovation.
Now, we are witnessing another innovation in oncology, this time in the field of clinical data sciences. Predictive analytics and artificial intelligence (AI) are transforming oncology clinical trials, enabling quicker, more cost-effective clinical development, and helping more patients to gain timely access to essential cancer therapies.
Obstacles to development
But like all progress, breakthroughs are also accompanied by challenges. In clinical development, randomized controlled trials remain the gold standard for developing new therapies. But in oncology in particular, this model poses ethical questions. With new treatments potentially offering life-changing and even life-extending benefits, complex ethical questions arise. Is it right to offer trial participants, for example, late-stage cancer patients, placebo therapies or the current standards of care?
There is also a practical problem. Greater understanding of the genetic drivers of cancer is leading to more targeted therapies designed for smaller patient sub-populations. Oncology research is increasingly focusing on cancers with specific genetic markers/mutations. This precision in intervention inevitably also leads to much smaller patient populations suitable for the treatment. Cancer trials reached historically high levels in 2021, up 56% from 2016, with most focused on smaller, well-defined patients with specific genetic markers. With breast cancer the most studied disease in 2022, and a growing number of trials vying for the same patients, running large placebo-controlled studies is not only time-consuming and expensive, but this model is delaying access to much-needed therapies.
An already stretched research area is further plagued by the ongoing issue of trial terminations, with Phase II attrition 42% higher in 2022 than the five-year average. An analysis of trials over 20 years showed a termination rate of 22%, with 28% terminated due to poor enrolment. Terminations at Phase II and above are particularly costly and can have a significant impact on future industry investment and patient access to innovative treatments.
The rise of the single-arm trial
Working towards reducing and ultimately eliminating the use of placebo comparator arms should be an industry-wide goal. Single-arm trials – in which all patients receive the same investigational drug – will play a role in this. As well as solving the ethical dilemma, the benefits of single-arm trials include a shorter cycle time, a smaller sample size requirement, and the ability to detect efficacy signals early in development.
Oncology drug approvals are increasingly supported by single-arm trials. An analysis by JAMA Oncology revealed that between 2002 and 2021, the FDA approved 563 new oncology drug indications, and 176 (31%) of those approvals were based on data from single-arm trials.
Single-arm trials do have some limitations, with an inherent degree of uncertainty that can cause problems in trial design, patient recruitment and data interpretation – all of which can delay patient access. Small single-arm trials are not always representative of the real-world patient population, which brings complications around approval and reimbursement.
Filling the knowledge gaps
Now, the industry is embracing a new approach – using AI and predictive analytics to address the uncertainty of single-arm trials. Powered by a large enough clinical trials database, it is possible to produce digital patient profiles, digital trial arms and “digital twins”, all of which can support sponsors in designing and running successful single-arm trials.
Digital trial arms enable researchers to simulate and predict different clinical outcomes in a clinical trial with greater certainty. A digital twin, when planned and implemented in alignment with regulatory authorities, can function as a digital trial arm, or external control arm, as FDA calls it, eliminating the need for a placebo or a comparator arm. Data is collated from similar or identical trials using the same agent, with real-world patient data, to accurately model placebo/comparator outcomes while accelerating development.
Know your patients
Utilizing data science and predictive analytics can make a crucial difference to the success of new therapies. For example, researchers conducted a single-arm trial assessing dostarlimab, an anti–PD-1 monoclonal antibody, in a rare form of locally advanced rectal cancer. A total of 12 patients completed treatment with dostarlimab and all had a clinical complete response.
GSK built its case to support FDA review of a potential expansion of the dostarlimab label based on the results of this trial. However, a Phesi digital patient profile (Fig 1.) of patients with this rare form of cancer revealed that the patients in the trial were on average younger (52) than the typical patient for this disease (61). Also, the majority of the patients in this small trial were women (62%), while only 35% of the patients in the larger population are women. Both gender and age are important prognostic factors for this disease. This crucial information could help to interpret this small single-arm trial’s results more accurately, and better predict its efficacy in the broader target population.
Figure 1: Phesi Digital Patient Profile of patients with rectal cancer compared with the patient group in MSKCC 19-288 trial. The Phesi profile was generated using data from 133,324 real patients. Credit: Phesi.
This is just one example of how data-driven innovation can fill the gaps for single-arm trials, accelerating development and regulatory review and reimbursement decisions. Often a Phase II trial is conducted with a limited number of patients, with the results then used to implement a larger, randomized controlled trial. To find out that the positive signal seen in the Phase II trial was false is very costly and can result in futile Phase III trials. Through digital twins and digital trial arms, a clinical study can be modeled from early screening to protocol design to removing a comparator arm – addressing long-standing ethical questions in drug discovery and accelerating the route of treatments to patients.