Drug repurposing advantages include shorter research and development times, potentially lower development costs, and reduced risk of failure because an approved drug typically has an established safety profile. However, after successful proof-of-concept clinical trials – often conducted by researchers from academic or independent research organisations – large, randomised Phase III clinical trials are necessary to gain regulatory approval for an intended use. As these are expensive, labour-intensive, and time consuming to complete, pharmaceutical firms have little economic incentive to finance those trials, especially if the drugs are out of basic patent or regulatory protection. Result: The hurdles before successfully repurposing an established drug are challenging.
Using AI and real-world data to replicate large clinical trials
Researchers at the Ohio State University in Columbus have successfully developed a deep learning framework for drug repurposing emulating clinical trials using real-world databases of millions of patients. The first test of this artificial intelligence (AI) tool focused on USA Food and Drug Administration (FDA) approved prescription drugs that were successfully used ‘off label’ to treat coronary artery disease (CAD). The database included 107.5 million patients diagnosed for coronary artery disease between 2012 and 2017.
Out of 55 candidate drugs included in the study, the researchers identified six prescription drugs not indicated for CAD, which might be repurposed, and three new drug combinations using the AI tool. These included metformin, escitalopram, and the combination of lisinopril and atorvastatin.
Writing in Nature Machine Intelligence, principal investigator Ping Zhang PhD, director of the Artificial Intelligence in Medicine Laboratory, and colleagues, explained that they followed the protocols of randomised clinical trial design and computationally screened repurposing candidates for beneficial effect by explicitly emulating the corresponding clinical trials using real world longitudinal observational data. They applied deep learning and causal inference methods to control real-world data confounders, and systematically estimated the drug effects on various disease outcomes. (The process is explained in detail in the article doi.org/10.1038/s42256-020-00276-w)
In addition to identifying repurposing drug candidates to treat CAD, the authors developed an innovative study design to estimate the drug class or drug combination’s effect. They also compared their framework with three existing pre-clinical drug repurposing methods and showed that their AI framework was superior.
Funding independent clinical research with innovative mechanisms
In a newly published article in Trials, researchers from Belgium describe and evaluate potential funding sources for Phase III clinical trials of promising repurposed drugs, based on extensive research and interviews with 16 experts representing university hospitals and academia, banks, the pharmaceutical industry, consulting, and health technology assessment organizations.
First author Ciska Verbaanderd PhD and co-authors at the Department of Pharmaceutical and Pharmacological Sciences of KU Leuven, identify four potential funding mechanisms, discussing in detail the advantages and disadvantages of each. They also discuss the need for increased harmonisation and centralisation of clinical research and funding at European and international levels to reduce fragmentation and maximise the value of limited resources.
- Traditional grant funding by non-profit organisations, government agencies and pharmaceutical companies for clinical drug repurposing research is increasing, but not to the extent needed. Experts suggest that government agencies identify the most important unmet needs in healthcare and target additional funding to these research areas.
- Crowdfunding, using social media and digital platforms with global outreach, provides the opportunity to raise funds for innovative projects with a potentially high societal or patient impact but low commercial return. Crowdfunding was used to finance a Phase II clinical trial investigating the use of a 40-year-old anti-malarial agent as a colorectal cancer treatment.
- Establishing and managing a crowdfunding campaign tends to be time-consuming, labour intensive, and expensive. The experts believe it is best suited for early-stage research projects but inappropriate as a way to raise significant funds needed for large clinical trials.
- Public-private partnerships (PPPs) are collaborations between, among others, academic researchers, non-profit public funders, and industry. One successful initiative repurposed a drug that was initially developed to treat rheumatoid arthritis in the mid 1980’s to treat chronic lymphocytic leukaemia.
The authors advise that most current PPPs focus on the repurposing of experimental or investigational assets that went through several stages of clinical development for another indication, but were ‘shelved’ due to a lack of efficacy, commercial interest, or other reasons. In general, PPPs do not offer a sustainable solution for the repurposing of drugs that are out of basic patent or regulatory protection due to the lack of incentives for pharmaceutical industry partners. The exceptions cited by experts are initiatives to find new treatment options for neglected diseases in low- to middle-income countries, in situations where companies may be motivated by social responsibility.
- Social impact bonds (SIBs) are new, leveraging private investments to develop public health services or interventions. This innovative pay-for-success financing is a formal agreement between a payer, such as a government or health insurance company, and a provider, where investors providing funds are repaid if the goal is reached.
Currently Findacure, a Cambridge, UK charity dedicated to treatment of rare diseases, is collaborating with other charities to incentivise investment with an SIB into drug repurposing clinical trials into rare diseases.
At the outset, goals and clinical outcome measures must be specifically defined, and start-up costs are high, according to experts, but SIBs may offer promising new funding sources.
Ping Zhang PhD is an Assistant Professor at Ohio State University (OSU), with joint appointments at the Department of Biomedical Informatics (BMI), and the Department of Computer Science and Engineering (CSE), where he leads the AIMed (Artificial Intelligence in Medicine) Lab. His research focuses on data mining, machine learning, causal inference and their applications to biomedical informatics.
Ciska Verbaanderd PhD is a research assistant at the Clinical Pharmacology and Pharmacotherapy department of the University of Leuven. Her focus lies on clinical studies and regulatory science, especially in oncology, work carried out in collaboration with the Anti-cancer Fund, a Belgian Foundation of Public Utility dedicated to expanding the range of treatment options for cancer patients, regardless of commercial value.