Trump meets Zelenskiy, says Putin wants war to end, mulls trilateral talks
Investing.com -- Clinical trials are notorious for being expensive and slow, with rising costs and declining efficiency. According to Bernstein, industry R&D spending grew 44% between 2012 and 2022, while the number of novel drug approvals in the U.S. stayed flat.
“As of 2024, it was estimated that up to 80% of clinical trials overshoot their forecasted timelines,” the broker said in a recent report.
The challenges are long-standing. An NIH study of 640 phase 3 trials found that 54% failed, with 57% of those due to insufficient efficacy and 17% from safety issues. Financial and logistical barriers also weigh heavily: 22% of failed trials lacked adequate funding, while recruitment shortfalls and restrictive eligibility criteria often undermine results.
A McKinsey study cited by Bernstein estimated per-patient costs above $40,000, with the average span from Phase I to launch stretching roughly a decade. Success rates hover at just 10–12%.
Artificial intelligence (AI) is now being tested as a potential solution. In theory, AI could support every stage of the process, from design to analysis.
In trial design, for example, AI can draw on real-world data such as past studies, electronic health records, and patient registries to refine eligibility criteria and propose more measurable endpoints.
This, Bernstein analysts said, “allows for faster, more efficient trials with higher chances of success.”
Recruitment is another area where AI could play a major role. By analyzing electronic health records, lab results, and clinical notes, AI systems can match patients more precisely to trials.
Natural language processing can help uncover eligible candidates from unstructured medical data, while machine learning models forecast enrollment timelines and highlight effective recruitment channels.
Analysts said that “early detection of lagging enrollment allows sponsors to implement corrective actions before delays escalate.”
Once a trial is underway, AI can assist in monitoring by tracking site performance in real time. Advanced models can detect anomalies in protocol compliance, adverse event reporting, or data entry, enabling sponsors to address problems before they slow down a study.
Similarly, in the analysis phase, AI tools can detect subtle treatment effects, generate synthetic control arms from historical data, and accelerate statistical modeling.
A growing number of companies are already pursuing this opportunity.
Bernstein grouped them into three categories: traditional contract research organizations like IQVIA Holdings (NYSE:IQV), Icon (NASDAQ:ICLR), and Fortrea Holdings (NASDAQ:FTRE); health-tech firms such as Medidata, ConcertAI, Massive Bio, and Flatiron Health; and hybrid players like Tempus AI Inc (NASDAQ:TEM) and Caris Life Sciences Inc (NASDAQ:CAI), which combine diagnostics, sequencing, and AI-enabled trial matching.
Several have forged partnerships with technology leaders—IQVIA and ConcertAI both collaborate with NVIDIA (NASDAQ:NVDA) to develop AI agents for trial workflows.
The foundation of these efforts is data, and companies emphasize the scale of their collections. Caris describes its repository as “one of the largest, multi-modal databases of combined molecular and clinical outcomes data in the world,” while Medidata cites information from more than 36,000 trials and 11 million patients.
Still, Bernstein cautioned that despite heavy investment and rapid innovation, the role of AI in clinical trials remains uncertain.
“In five years, when the dust settles, the influence of AI may feel less like a revolution and more like plumbing: foundational, invisible, and absolutely essential.
Conversely, AI may struggle to make significant inroads into a highly-regulated and inefficient industry,” the analysts wrote.
What is clear, the report concludes, is that with costs escalating and trial timelines stretching, the industry is in urgent need of change—whether driven by AI or other forms of innovation.