Introduction: More than 200 treatments have been tested for COVID-19 in over 7000 clinical trials. Most of these treatments are repurposed generic drugs, many of which have been studied extensively for the treatment of cancer. As cancer patients are particularly vulnerable, there is a need to understand how COVID-19 treatments might affect a patient’s cancer. As part of the Reboot: COVID-Cancer Project, a living and freely available resource of clinical studies that report outcomes for cancer patients, we have developed a semi-automated pipeline to identify all relevant published clinical studies and registered clinical trials where COVID-19 drugs were tested for the treatment of cancer. Methods: Published clinical studies were assembled using targeted search queries in PubMed, rule-based approaches, and machine learning models. Machine learning models applied to natural language processing tasks were used to predict the drug, cancer type, study type, and therapeutic association. We used domain-specific rules and post-processing steps to further refine results, including determining whether a drug was used alone or in combination. Registered clinical trials were compiled from clinicaltrials.gov using targeted search queries, automated mapping, and rule-based screening. We extracted key information about each trial, such as the drug, cancer type, phase, location, trial status, age, gender, and availability of results. We applied our pipeline to a curated set of 202 drugs being tested for the treatment of COVID-19 in at least two interventional clinical trials worldwide, of which 27 are FDA-approved drugs that are standard of care for cancer, and 115 are FDA-approved drugs primarily used for non-cancer indications. Results: We found 28,138 published clinical studies and 9,118 registered clinical trials where the 202 drugs were tested for cancer. The published clinical studies include 5,286 case studies, 2,559 randomized controlled trials (RCTs), and 20,294 non-RCT clinical trials or observational studies. In 37% of the cases, the drug was used alone and not in combination. Lymphoid cancers were the most commonly tested, comprising 30% of studies. Possible benefit of the drug was found in 64% of publications. Of the 115 FDA-approved non-cancer drugs being tested for COVID-19, there is at least one published clinical study for 84 (73%) drugs. An additional 12 FDA-approved non-cancer drugs have been tested for the treatment of cancer in clinical trials, but have no results reported. Of the registered clinical trials, 39% are currently active, 66% are Phase 2 or later, and lymphoid cancers are again the most common, representing 29% of the trials. Discussion: Given the interconnection between COVID-19 and cancer, it is essential to understand how drugs used for COVID-19 might impact a patient’s cancer. We have created a living resource for rapid review of information. The datasets are updated monthly and are freely available via an interactive dashboard.

Citation Format: Allison Britt, Emily Yang, Devon Crittenden, Thomas Bhangdia, Ben Nye, Emily-Claire Duffy, Saradha Miriyala, Emily van der Veen, Sam Marchant, Kelly Fan, Ellie Strauss, Katherine McKinley, Kriti Sharda, Murshea Tuor, Ishita Mahajan, Noopur Ranganathan, Lailoo Perriello, Byron Wallace, Pradeep Mangalath, Laura B. Kleiman, Catherine Del Vecchio Fitz. Effects of COVID-19 treatments on cancer: A machine learning approach to synthesize clinical evidence at scale [abstract]. In: Proceedings of the AACR Virtual Meeting: COVID-19 and Cancer; 2021 Feb 3-5. Philadelphia (PA): AACR; Clin Cancer Res 2021;27(6_Suppl):Abstract nr P04.