In the days when standardization was absent, researchers were not obliged to share clinical trial data and whether participants were voluntarily recruited into their studies. These inadequate and poor-quality outcome reporting challenged succeeding researchers to evaluate, replicate and build upon previous findings and ensure patients’ health will not suffer from any sub-optimal treatment.
In 1993, 30 experts from the realms of epidemiology, editing, clinical trial and methodology decided to step up for a change. They met in Ottawa, Canada and created a 32-item checklist called the Standardized Reporting of Trials (SORT). Coincidentally, around the same period of time, a separate group of experts with similar concerns met in California and came up with the Asilomar proposals.
At the suggestion of Dr. Drummond Rennie, nephrologist, physiologist, and contributing deputy editor of The Journal of the American Medical Association (JAMA), the two groups decided to merge their evidence-based recommendations into one, resulting in what known as the Consolidated Standards of Reporting Trials (CONSORT) today. CONSORT was first published in 1996 and was subsequently revised in 2001 and 2010.
Two weeks ago, the new CONSORT-AI extension was announced in Nature Medicine, the British Medical Journal, and the Lancet. This additional checklist augments the core CONSORT 2010 with 14 new items, guiding investigators on the skills required to handle and evaluate artificial intelligence; human-AI interaction; error analyses, and so on. Specifically, investigators are expected to indicate the inclusion and exclusion criteria for input data or the data to be used for training the AI.
For example, if investigators are developing a breast cancer diagnostic system, the input data will be the pre-processed mammography scans and patients’ health information. Investigators should detail why a particular patient was chosen and if the quality of the patient’s scans happened to be compromised in one way or another, investigators will also have to include reasons for excluding the scans.
It’s very likely the AI algorithm is not generalizable for a wider environment other than to serve the purpose it was built; thus, investigators will have to state if the AI intervention is run on any vendor-specific devices or if any particular hardware is needed. Similarly, the AI algorithm may undergo multiple updates or changes during its lifespan. Hence, investigators will have to specify which version of the AI system was deployed in the clinical trial. More importantly, a description of the requirement for successful human-AI interaction and training or instructions that users ought to acquire needed to be provided too.
Accelerating clinical trials with AI
CONSORT-AI was developed to complement SPIRIT-AI (Standard Protocol Items: Recommendations for Interventional Trials-Artificial Intelligence) for high-quality protocol reporting of AI-trials and is supported by EQUATOR (Enhancing the Quality and Transparency of Health Research). It hopes to build transparency and completeness in reporting clinical trial results and assist fellow editors, peer-reviewers and readers to understand and appraise these trials in critical manner.
All along, conducting clinical trials have been a tedious and often fruitless process. An analysis which examined clinical trial data obtained between January 2000 and April 2019 found that only an estimated of 12% drug development programs would come to success because many trials do not necessarily demonstrate that a regimen is safe or efficient. Other reasons such as the inability to recruit desired participants; lack of funding; high drop-out rates, and poor study designs are also common reasons for clinical trials to fail.
Researchers in both academia and pharmaceutical industry are gradually looking upon AI to improve traditional randomized trials. They believe AI bears the potential to save billions of dollars by improving the protocol design; speeding up the process of finding eligible participants or bringing interested patients closer to clinical trials. The ongoing COVID-19 pandemic also witnessed the use of AI in repurposing existing drugs to search for novel coronavirus cures. Thus, CONSORT-AI comes just in time, as AI tries to enhance itself as well as clinical trials.