The digital twin trial arm demonstrated a 52.7% overall response rate at 6 months for prednisone in chronic GVHD, aligning with real-world data.
In a recent study, researchers investigated the effectiveness of using a digital twin (DT) trial arm, based on a real-world clinical trial database, to evaluate the efficacy of the standard-of-care (SOC) treatment, prednisone, for graft-vs-host disease (GVHD). Using thePhesi Trial Accelerator platform, they enrolled patients with chronic GVHD undergoing first-line treatment into a DT cohort.Their findings showed a 52.7% overall response rate at 6 months, aligning with the existing literature.1
“These findings demonstrate that reliable and robust data can be obtained from diverse sources to construct DT arms and support the potential utility of DT SOC arms in real-world clinical trials,” according to study investigators.1
“In a clinical trial, there is a protocol, where you start going to hospitals and such to recruit patients, and we do the same. We need to have this guidance to confirm the design of the trial according to protocol,” Gen Li, PhD, MBA, president and founder of Phesi and lead author of the study, told Targeted Therapies in Oncology, in an interview. “However, instead of [recruiting, for example,] from a hospital, we recruit from our own database, where we now hold over 120 million patient data.”
To evaluate the SOC for chronic GVHD in the first-line setting,investigators choose from a patient population within this same context, following the inclusion and exclusion criteria. The DT arm is then evaluated int he same way as a comparator arm in a clinical trial. “We have similar challenges found in clinical trials,” Li explained. “Sometimes we may not be able to find the right patient, or get as quality data as we may want, but these are all part of the typical challenges associated with clinical trials.”
Of the 61 million patients in the Trial Accelerator database at the time of analysis, researchers identified 106,183 patients (770 cohorts) with any type of GVHD. From this group, they selected 17,769 adult patients (209 cohorts) with chronic GVHD. Ultimately, 2042 patients (32 cohorts) with chronic GVHD who received first-line treatment were chosen to construct the first-line GVHD DT cohort. These patients were enrolled at centers in over 16 countries, including the United States, Australia, China, Germany, France, Italy, Taiwan, Japan, South Korea, and Austria, between 2004 and 2021, with one cohort from a 1997 study.
“This process [has been] 20 years in the making,” Li said; however,technology is only now catching up to allow for the concept of DT arms in clinical trials.2 Over the past 2 decades, the rise of the Internet of Things (IoT), sensors, and connected devices has revolutionized data exchange, leading to an explosion in big data production. Advances in big data analytics, cloud computing, and artificial intelligence now enable the storage and processing of IoT data, paving the way for the widespread adoption of DTs.“It’s now the right time, and we have started to publish digital twins over the past 2 or 3 years,” Li said.“This is number 3,and we have several others in the making.”
Li et al have been evaluating clinical developments to find avenues where clinical data science can help to reduce the time and economic resources required, as well as facilitate patient participation in clinical trials and enhance clinician engagement.Other trials Li et al conducted involved creating a DT for KRAS G12C non–small cell lung cancer. This trial assessed the SOC treatment and measured outcomes such as median progression-free survival. Another trial focused on chimeric antigen receptor (CAR) T-cell therapy for cytokine release syndrome (CRS), also using SOC as a benchmark. In this case, outcomes were evaluated based on the distribution of CRS severity grades.3,4
If not fully replacing standard methods, DTs provide incremental advantages in trial planning and execution by facilitating the understanding of patient needs and characteristics, leading to better-aligned trial designs. DTs also help to identify design issues before the trial begins, minimizing avoidable protocol amendments. In many standard cases, investigators find errors in the trial design and need to amend the protocol before they can proceed, Li explained. These issues can incur significant costs, impacting the budget, the patients enrolled, and the trial’s timeline.
Other incremental advantages include reducing the sample size of the control arm by 25%, 50%, or 75% instead of completely replacing it. “This is still a tremendous benefit for everyone involved, including reduced cycle times, lower costs, and decreased burdens on patients,” Li stated.
Using historical controls is seen as an enhancement from a regulatory perspective, supporting the credibility of trial designs, Li explained. For example, regarding the FDA, “you do not need to have FDA approval to put in a historical control in your submission for approval,” Li said.
Although there is potential to replace traditional control arms in phase 3 trials with digital alternatives, this remains a significant step that requires careful consideration and regulatory engagement, Li explained. In some scenarios, such as those concerning rare diseases or specific cancer strains, the application of DTs can provide valuable insights and justify their use in clinical trials. Achieving these advancements will require concerted efforts among regulatory bodies, researchers, and industry stakeholders to navigate the evolving field of clinical trial design.
The effective use of DTs in healthcare necessitates extensive patient data collection and storage, which poses ethical challenges regarding the confidentiality and security of sensitive information.2 The cybersecurity of DT databases is a significant concern, and to mitigate these issues, laws such as the European Union’s General Data Protection Regulation enforce new legal requirements, including the right to withdraw consent and the right to be forgotten.
Gathering comprehensive and high-quality data is also a challenge due to fragmentation across healthcare institutions and inherent biases such as sensor inaccuracies and data entry errors, according to Katsoulakis et al.5 Longitudinal data, necessary for evolving DTs to reflect health changes accurately, can be incomplete and sparse, and can cause challenges that hinder precise DT development and data labeling. Bias toward a certain demographic or condition also remains a threat in this context and ensuring that DT models are free from biases is vital.
Despite concerns, the concept of DTs offers an interesting conundrum: whether to“accept the variability between the real human being and its DT more or the variability between different human beings having similar baseline characteristics,” noted Patrizio Armeni and colleagues in a critical review of DT technology.2
“People may think the idea is just to replicate some of the things already known about human beings, but that’s not true. Knowledge can be gained from accumulated data, and it’s quite common for us when analyzing existing patient data to gain different insights and understanding of the disease. This can present us with new treatment opportunities and reveal nuances of the disease that were previously unknown,” Li said.
Lipsky Discusses Second-Generation BTK Inhibitors in Relapsed/Refractory CLL
October 12th 2024During a Case-Based Roundtable® event, Andrew H. Lipsky, MD, moderated a discussion on the efficacy and safety of newer BTK inhibitors used to treat patients with chronic lymphocytic leukemia.
Read More