This year’s Outsourcing in Clinical Trials (OCT) event focused on the challenges and opportunities in current and future clinical operations, as well as the state of AI and tech transformation in clinical trials. The keynote speech was delivered by Ken Getz, Executive Director and Professor at Tufts University School of Medicine’s Center for the Study of Drug Development. Getz addressed how pivotal phase 3 protocol designs have evolved from 2013 to 2025, with increasing complexity from:

• 14 to 18 total endpoints
• 9 to 13 total countries
• 65 to 109 total investigative sites
• 597 to 737 total patients randomized

Additionally, these protocol complexities have prolonged the duration of:

• Trial initiation (27.2 % from approval to First Patient First Visit)
• Participant enrollment (36.9%)
• Substantial amendments (increasing by over 52.2%)

On top of the pressure this complexity places upon trial sponsors and investigative sites, the burden on study participants has increased. The mean number of procedures per patient visit increased from 11.1 in 2009-2012 to 13.9 visits in 2017-2020. 64.3% of premature terminations from trial participation from 2019-2023 were due to patient choice.

A clear takeaway from the keynote was that as trial protocols become more complex, trial execution becomes less efficient, and site burden increases. Below are three trends in clinical trials and clinical trial services we took away from the event.

Clinical Trial Services Takeaway #1: Human-in-the-Loop to Human-on-the Loop

A notable leading sponsor discussed the lack of an AI operating model among life sciences companies. This is especially notable despite the industry’s extensive use of AI and pilot programs. According to a 2025 McKinsey & Company article, most life sciences companies use AI. However, due to factors such as weak talent plans and change management or loose governance, only:

• 32% have managed to scale it
• 5% realized competitive value

Despite the challenges in realizing ROI in the current AI environment, the sponsor foresees that, in a few years, the human role in clinical research and AI-generated content will increasingly transition from a Human-in-the-Loop (HITL) model to a Human-on-the-Loop (HOTL) model (depending on the use case and its risk). In HITL models for clinical trial services, a human is involved in every decision and validates the AI system’s output. In HOTL models, the human becomes a supervisor who monitors the system or trends without reviewing or validating every output or decision. In HOTL, human intervention happens only when risks or anomalies are detected. Therefore, the model is unlikely to be used anytime soon in high-risk, highly regulated GxP areas or where patient safety is at risk. The transition to a HOTL model is likely when credibility can be established toward regulatory authorities, and models are robust and explainable.

Clinical Trial Services Takeaway #2: AI Emerges in eCOAs for Select Use Cases

It was clear from the discussions at OCT that AI is gradually being considered and implemented for electronic Clinical Outcome Assessments (eCOAs), but only for select purposes. Translation of questionnaires, instructions, and patient-facing text was presented as suitable AI clinical trial services use cases. Quick translation drafts can be generated with AI for review by linguists or clinical teams and for suggestions of culturally appropriate phrasing. Additionally, AI can be used in text-level comparisons and screenshot reviews. Many copyright holders, however, are now including in their agreements that no AI can be used on their scales. For more insights into the use of AI in eCOAs, visit Lionbridge’s recent update on the newest industry regulatory compliance on AI in COA Translation.