With clinical trials having become extremely expensive to conduct, industry buzz around the potential of data and artificial intelligence (AI) to allay these costs has grown. However, experts are split as to the true potential of digital tools to reduce clinical development outlay and provide better patient outcomes in the near future.
As Professor Jackie Hunter, chief executive of clinical and strategic partnerships at BenevolentAI, has written for PharmaBoardroom, “Clinical trial outlays over the past two decades have soared – it costs an average of USD 2.6 billion and 12 years to bring a new therapy to market – but the industry has not delivered commensurate high levels of innovation in return.” Indeed, only a mere 13.8 percent of drugs nowadays actually successfully navigate clinical trials according to the latest analysis from MIT.
Conducting clinical trials is becoming impossibly expensive
“Over 80 percent of trials are delayed,” adds Christina Busmalis, IBM Watson Health’s director for global life sciences & go-to market leader for Europe & Asia Pacific. “Over 50 percent of sites never recruit a patient. An amendment costs USD 40,000 a day and up to USD eight million in potential revenue. All statistics show that the clinical trial process is currently flawed.”
Janet Woodcock, director of the US FDA’s Center for Drug Evaluation and Research (CDER), notes that “Conducting clinical trials is becoming impossibly expensive. It is the most expensive part of drug development. There is a hope that by using real-world data or evidence, that [pharma companies] can make their trials smaller or substitute data for trials. There is a lot of hope that data can provide a solution to this.”
Barbara Lopez Kunz, CEO of the Drug Information Association (DIA) in the USA is, however, cautiously optimistic, noting the incredible progress that has already been made in drug development. “We are living in a time of incredible advances in science and technology, and we have seen the positive impact in many ways in healthcare. New therapies, new diagnostics, and cures are available today that we never imagined possible.”
These new approaches may provide ways to accelerate our processes and reduce risk, with the potential to lower costs
She adds, “Advances are being made with the integration and use of data, and tools are available that accelerate insights – such as in clinical research – support regulatory decision-making and increase efficiencies and the effectiveness of processes. People ask – why does it take so long and cost so much to develop a new therapy? These new approaches may provide ways to accelerate our processes and reduce risk, with the potential to lower costs.”
For Tom O’Leary, chief information officer of clinical research organisation (CRO), ICON, data has the potential to revolutionise patient identification and recruitment with the global biopharmaceutical industry increasingly moving towards niche and targeted therapies. “The reason we are looking at data is that the therapies that are being developed are much more targeted and specific than ever before,” he asserts.
“For example, cancer is a collection of rare diseases, so finding patients that are going to see benefits from the therapies being developed is key. To do this patient identification, you need to look at genomic data. Gone are the days of broad-spectrum antibiotic drugs and cardiovascular therapies. Today it is very niche, looking at cohorts of patients that are going to get benefits from very specific conditions. That requires us to find the patients that have the gene mutations and so forth that are going to get the greatest benefit from the therapy.”
Most [start-up companies looking at AI in drug discovery] will not survive, but some are already adding value
In terms of recruitment, O’Leary notes that “We [currently] do not have enough data to help identify from where to recruit those patients. The way patients have been recruited traditionally is going out to a network of physicians, explaining to them that you have a clinical development program in a specific therapeutic area and asking whether they have patients in their practice that would benefit from this particular drug.”
He continues, “Now, we are looking to heatmap the world and find where there are concentrations of patients suffering from specific conditions or diseases with the help of electronic health and medical record data.”
IBM Watson’s Busmalis, noting that the cost of bringing a drug to market has dropped below USD two billion thanks to an increasing industry focus on developing orphan drugs for rare diseases, asks “Imagine how technology can improve this even further? From a technology perspective, in drug discovery, there are over 150 start-up companies looking at AI in drug discovery, which is an overwhelming amount. Most will not survive, but some are already adding value. For example, a company called FDNA looks at phenotypic and genomic data to understand precision medicine. Another, Insilico Medicine, looks at identifying new biomarkers around ageing-associated diseases.”
Standards to be Set
However, most stakeholders agree that data will not be a cure-all for the expensive and lengthy clinical trial process and that real change may still be a distant goal. “I believe that data and AI may be game-changers in the future, but not in the immediate future,” states the FDA’s Woodcock. “The old maxim about data – garbage in, garbage out – still holds … [For example] US electronic healthcare records are not interoperable. Even within electronic health records, a lot of the terms are not standardized. There is a huge problem of noise in the system.”
The buzzword of ‘big data’ and its potential is being discussed a lot and fills headlines. However, the quality, consistency, and currency of that data is still very weak
ICON’s O’Leary strikes a similar tone. “The buzzword of ‘big data’ and its potential is being discussed a lot and fills headlines. However, the quality, consistency, and currency of that data is still very weak. I have seen situations where you may need to look at 50 million patient records to identify only 20 patients for a clinical trial. Beyond that, the data is not always complete, the patient may not still be alive, or the patient’s medical health record may have the wrong address listed.”
He continues, “Standardisation is vital. The challenge for the medical community is that the data has not been standardised, it is not captured consistently, and a lot of it exists in unstructured formats such as patient notes. The tech tools are advancing to be able to read that unstructured data, but we are not there yet.”