In the industry-wide campaign to improve the efficiency of clinical trials, risk management has emerged as a leading strategy. In this context, it is useful to recognize that, to a large degree, the study site is where study conduct happens, errors happen, and indeed, risk happens. It would seem obvious that, in a strategic campaign to lower risk, tactics to improve study site performance would be paramount. In spite of this, a formal, comprehensive initiative to optimize study site performance and lower errors is lacking among most sponsors.

Complexity

There is widespread consensus backed by research that clinical trials have been increasing in complexity for decades, threatening the sustainability of the clinical trials. Yet most study sponsors continue add complexity to their clinical trial designs along several dimensions.

• • The architecture of protocols increasingly incorporates new designs such as adaptive trials, decentralized trials, risk-based strategies, and real-world evidence.
  • Data demands have steadily escalated, with more primary and secondary endpoints, more assessments, continuous data capture, synthetic data, and use of massive external healthcare datasets.
  • New technologies continue to proliferate, including various wearable devices, digital biomarkers, and rapidly-evolving AI technologies.

Site Capabilities

At the same time that study sites are confronted with more demanding studies, they face challenges in attracting and retaining qualified staff. Across the industry, this has been characterized as a crisis in availability of professional staff.

Levels of qualified study site staff started to wane prior to the COVID-19, but the pandemic served as an inflection point, with large numbers of research professionals retiring or changing employment. Adequate numbers of qualified staff have failed replace this loss of professional talent.

As a result, many study sites face a shortage of experienced staff, adding stress to the remaining site members, who are additionally stressed by factors such as intense competition for qualified patients, and relentless pressure to meet study timelines and objectives.

The increasing complexity of clinical trials, together with less capable staff and other pressures on sites, combine to create a perfect storm increased errors and risk in the conduct of clinical trials. To protect the quality, expense and timelines of trials, sponsors must take action to de-risk against a rising tide of site errors.

Experience has shown that the most likely sources of risk are relatively consistent across trials. And of course, every trial has special risk elements that need to be anticipated.

Based on risk assessment, we focus resources on the areas of study site performance that are likely to yield errors. As the study progresses, we continue to track errors and re-direct resources to emerging areas of risk.

Operational risk in clinical trials tends to cluster in known areas such as:

• • Patient screening and eligibility
  • Informed consent
  • Investigational product (IP) management
  • Source data integrity
  • Patient visit assessments
  • Adverse event (AE/SAE) reporting
  • Study documentation/ISF

The quest for simplicity in clinical trials start with establishing a solid foundational knowledge of the study without “over-training.” We do this with streamlined initial training that does not delve into myriad technical details for all parts of the study. Instead, training starts with a strong understanding of study goals, protocol structure, patient enrollment, protection of patient safety, and other priority topics.

A separate level of training is conducted based on the roles of site members and the skills they will need. Follow-up training is timed to coincide with tasks such as effective data capture and study documentation. And training continues over the life of the trial to update sites and reinforce performance that drives down errors.

Brief training resources and special study tools provide sites with the support they need to excel at the point of performance. Throughout the trial, sites gain rapid access to accurate information in several formats, including a dynamic document center, comprehensive FAQs, and an AI tool that provides instant, accurate answers to site questions.

The problem is, sites face greater challenges than 20 years ago, such as the boom in study complexity and the need to enroll narrowly defined patient cohorts. It’s not surprising that legacy practices generally fail to achieve effective site readiness.

Legacy practices in preparing study sites span an array of tactics that are simply insufficient today. These include:

• • On-site investigator meetings that are expensive, rushed, and partially attended,
  • A front-loaded “data dump” of information with minimal downstream reinforcement,
  • Online training modules that are not concise, role-based or engaging
  • Study documents that are distributed ad hoc,
  • Minimal activities to reinforce performance over time, and
  • Lagging use of advanced digital solutions, including AI.

This traditional mix of tactics falls far short of an integrated Performance Optimization program that fully prepares sites for the challenge of executing complex trials.

Because errors are widely known to be expensive, the various cost components stemming from errors have been thoroughly studied and quantified, as follows:

Direct Study Costs

The direct impact of errors on study budgets occurs across a range of error types and is almost entirely due to error remediation and not error prevention. Remediation costs include detection, investigation, corrective measures and documentation, and they vary widely depending on the seriousness of the error and intensity of remediation.

Trial Delays

The cost of remediating errors in clinical trials is significant, however, this cost is dwarfed by the high cost of delays in study timelines that occur due to errors. Delays are not as easily observed, but due to their known significance, they have been extensively studied.

The first cost of delay above directly impacts the study budget. The second type of cost occurs after the drug is approved, and although it falls outside the study budget, it has a profound effect on company-wide financial performance.

And finally, as costly as these factors are, almost all efforts to manage errors are related to remediation after the fact. RASPberry AI targets prevention of errors, which is an overlooked area and a major opportunity for sponsors to employ a new risk-based tactic.

Expedited Timelines

• • Fewer errors result in shorter gaps in time between training, go-live, enrollment, first patient visit, last patient visit and site closeout.
  • Sites complete tasks easier, faster and with greater confidence. In doing so, they avoid errors that can delay the time to site closeout and database lock.

Cost Savings

Referring to data in the prior section, and assuming we can prevent 1 error out of 5 (20%) by applying a comprehensive program of performance optimization, we arrive at the projected cost savings below.

This cost savings projection suggests that, based on an error prevention estimate of 20%, a RASPberry AI performance optimization initiative program can lower the direct costs of s clinical trial by over $2M, prevent between 20% and 25% of errors by study sites, a budget of $2 million could produce an ROI in a range of 20:1 to 40:1.

Note: All data on the cost and incidence of various types of errors are based on published data. To compute the total number of errors, we use an average phase III trial of 100 investigator sites, 1,000 patients and 30 months duration. For average revenue per NME per month, we use PhRMA published figures. To see references for data herein, please see the RASPberry AI Performance Optimization whitepaper, which can be downloaded from this website.

Unified Knowledge Base

The core of RASPberry AI Trial Hub is a unified knowledge base 100% comprised by vetted study documents. This knowledge base ensures that AI-generated content is accurate and linked directly to trusted reference sources. The knowledge base is curated by AI project consultants to be comprehensive and updated, and the entire platform is shielded within our proprietary perimeter, with state-of-the-art security, compliance and fail-over provisions. RASPberry AI Trial Hub blocks access by outside AI platforms, search engines, or other third parties, which protects the platform from false information and prevents AI hallucinations.

Our approach to training starts by applying generative AI to produce training content for an investigator meeting that is more focused and engaging. Topics often include:

• • Protocol design and goals
  • Drug background and clinical profile
  • Screening and enrolling patients
  • Patient safety protections
  • Special issues unique to each trial

Our training modules are built around known areas of risk in study conduct, so that key learning points directly prepare sites to make fewer consequential errors. Our approach to initial training does not attempt to describe all the detailed technical information surrounding each component of the trial, which can be overwhelming for site members. Instead, initial training is streamlined and learnable, with references to the tools sites will use to perform optimally over the life of the trial.

In the post-activation phase of the trial, additional training is role-based and timed to coincide with the application of skills, such as data capture, document management, and patient visit

assessments. As studies become more complex, it is more challenging to prepare sites to execute myriad technical tasks. We address this challenge by directing site members to access information sources and study tools that provide guidance as sites conduct trial procedures. Essentially, we break down complex processes into a series of manageable tasks, and provide well-crafted guidelines, work tools and checklists to simplify tasks. This “point of performance” approach is less taxing on study sites and drives more rapid execution of trial activities with fewer errors.

It’s vital that sites have access to the full set of study documents, but even then, the sheer volume of information can be intimidating for sites to manage. The Document Center incorporates special agentic AI tools that enhance accuracy and productivity along several dimensions:

• • Advanced search tools enable sites to rapidly find documents, as well as the ability to extract the exact information being sought within documents, with full references and links to pages within documents.
  • Document version control and archiving is automated, consistent and reliable.
  • The entire library of documents can be scanned, flagged and revised to reflect updated information in the study. In the case of a protocol amendment, agentic AI can drive the process of updating all documents, which expedites amendment timelines and yields tremendous productivity benefits for study teams.
  • The Exchange Place in the document center incorporates automated workflows and a clear, structured environment that enables sponsors, CRAs and sites to exchange a multitude of documents with ease and accuracy.
  • Documents can be uploaded and downloaded to systems such as sponsor eTMF, site eISF (aka eReg Binder), and IRB/EC document platforms.

RASPberry AI Answers Center is an evolved, risk-sensitive AI platform, backed by extensive guardrails and quality-assurance provisions. It is nothing like a generic chatbot, but rather a best-in-class, mission-aligned agentic AI model.

The Answers Center has the potential to be a true game-changer in empowering sites to proceed through complex trials without stalling due to uncertainty. Currently sites rely on CRAs to provide guidance as they encounter issues, and that can take 24 hours or longer. The Answers Center provides reliable guidance instantly, ensuring correct performance while keeping investigator sites on track with study timelines.

Frictionless, accuate execution of complex protocols is the future of clinical research, and it is available today with RASPberry AI.

Study correspondence has long been conducted by using general email inboxes, which is an intrinsically chaotic, high-risk environment. RASPberry AI Correspondence operates within a proprietary platform that is significantly more secure, compliant, and de-risked. All messages are tagged by topic, enhancing search capabilities and supporting automated workflows, including automated escalations.

RASPberry AI Trial Hub supports automated filing of correspondence into the sponsor TMF and site ISF, with RASPberry AI doing the heavy lifting to distinguish between essential records that should be filed, versus non-essential communications that are not saved. This eliminates a task that is notoriously susceptible to procrastination and rushed efforts, and often riddled with errors such as lost, misfiled, or deleted essential records.

RASPberry AI continuously monitors all correspondence and detects discrepancies in facts or actions that are not aligned with the protocol. In such cases, RASPberry AI alerts sponsor/CRA users and proposes changes to the flagged content. This capability to mitigate against errors before they become actionable can have profound benefits in lowering risk during clinical trials.

At trial initiation, RASPberry AI applies an extensive set of standard FAQs against the trial knowledge base, generating study-specific answers and a comprehensive FAQ library from the very start. As the trial proceeds, the AI engine scans activity in the Training and Correspondence modules, as well as questions being submitted in the Answers module, to identify new FAQs and to autogenerate appropriate answers.

RASPberry AI tracks and reports what questions are arising most often across study sites. Such metrics can provide valuable insights about the most significant challenges sites are encountering. This can lead sponsors to make adjustments to the study plan, or to provide more in-depth training and guidance on critical aspects of the study.

Based on an extensive set of templates, RASPberry AI generates study-specific versions of tools that streamline work, especially for tasks that are known sources of error and risk. Our AI consultants drive this effort by curating a cleanly organized Tools Center. Examples of tools are:

• • Patient Eligibility Assessment Aid
  • Informed Consent Version Control
  • Inclusion/Exclusion Determination Form
  • Visit Window Calculator
  • Source Documentation Guidelines
  • Monitoring Visit Preparation Checklist
  • Patient Visit SOA Worksheets
  • Amendment Implementation Sheets
  • SAE Reporting and Reconciliation Prompts
  • IP Dispensing Checklist
  • Lab Processing Guides
  • Protocol Deviation Assessment Form
  • Regulatory Binder Job Aid
  • Study Close-Out Checklist