By Leslie M. Sam, president, Leslie Sam and Associates

BOHEMIA was not a traditional clinical trial.

It was a fully decentralized trial designed conducted in accordance with GCP and enrolled approximately 50,000 participants across one site in Mozambique and one in Kenya. Participants did not come to the site for study visits. Instead, field teams went into participants’ homes to collect data and blood samples, conduct study procedures, and administer the study drug. On top of that, most site staff had never worked on a clinical trial. Originally, the sponsor planned to execute the same protocol and operating approach in both countries.

Near the end of the Mozambique trial, the sponsor hired me to assess how well the site staff executed the trial against GCP standards. My site visit coincided with the final on-site CRA monitoring visit.

The night before my visit to the Mozambique site, the CRA proudly offered me the 50-plus-page report from her previous on-site monitoring visit.

I didn’t take it.

“Is it good news or bad news?” I asked. “What are the top three things I should be concerned about?”

She couldn’t tell me.

She could describe what she had reviewed and the individual issues she had identified — many involving informed consent — but she couldn’t tell me what those findings meant for the quality of the trial or whether the processes were under control.

Compliance Isn’t the Same As Quality

Soon, I discovered three different groups had been performing 100% source data verification of informed consent forms but weren’t sharing their findings. At the same time, consent corrections — missing dates, signatures, and similar documentation issues — and other data corrections sometimes moved between field data collectors, on-site reviewers, and CRAs on handwritten sticky notes.

The laboratory presented yet another concern. Staff tracked consent status on an Excel spreadsheet, but the laboratory team did not have visibility of that information before samples were processed or shipped.

And while the site team was finding and fixing individual problems as they appeared, they couldn’t see the relationships among those findings — the patterns, anomalies, and signals that could say something more important about trial quality. My assessment revealed that traditional monitoring and oversight practices were generating plenty of compliance activity but less insight into what those findings collectively meant for trial quality.

What I saw wasn’t a lack of effort; it was a lot of activity that created an illusion of control.

Same Protocol, But A Chance To Design Execution Differently

By the time I completed the Mozambique assessment, the sponsor finalized the Kenya protocol and demographic mapping had begun. But key elements of trial execution had not yet been designed: the database, site and medical monitoring plans, drug-accountability processes, and sample-management plan.

Ideally, quality by design considerations begin during protocol development, when complexity can still be challenged and critical-to-quality (CtQ) factors can influence the design. That opportunity had passed, and what remained was the opportunity to build quality into execution.

The Mozambique assessment became the basis for how the trial would be executed in Kenya. Lisa Collins, BOHEMIA project manager, and I then worked with the site to put it into practice.

The execution redesign started with a basic question: What absolutely could not fail, i.e., what were the trial’s CtQ factors? Informed consent, adverse event reporting, investigational product accountability, and reliable endpoint data were among them.

In practice, the redesign ultimately touched people, processes, and systems. We mapped data flows so site and sponsor staff understood what happened upstream and downstream from their own work. We also established chain-of-custody processes, eliminated most paper source documents, and designed reports prospectively so site functions could review their own data as it became available. Because of this, they could identify anomalies, missing information, relationships, and patterns — often in near-real time rather than weeks or months later when a CRA came on-site.

At this volume, the trial could not rely on periodic CRA visits to bring disconnected activities together. Site functions needed visibility while information could still inform action. Oversight had to be systematic, integrated, and holistic.

Then We Had To Rethink The CRA Role

Site functions — data management, laboratory, pharmacy, consent review, medical monitoring, and others — used custom reports to review the data closest to their work, identify issues and patterns, and take action. We then directed CRAs to oversee whether those site functional oversight processes were working as intended.

We assumed experienced CRAs who said they understood RBQM would be well equipped for the planned monitoring approach Kenya required. We learned quickly that knowing the terminology and applying the thinking were different things.

We asked our first lead CRA to revise the Mozambique monitoring plan to make it risk-based for Kenya. What came back was essentially the same plan, with risk-based terminology sprinkled throughout and arbitrary percentages of SDV added.

Even after we explained that the role was different from that of a traditional CRA, these experienced CRAs kept asking when they would get access to participant-level data. They wouldn’t. By the time a CRA repeated data reviews already performed by site functions, much of the opportunity for timely action would have passed.

This change in responsibilities required a culture shift. CRAs were accustomed to finding problems, telling the site what needed to be corrected, and making sure the site actually corrected the issue. In Kenya, we trained sites to identify and address problems themselves; the CRA assessed whether they did.

CRA monitoring now centered on a line of questioning:

• Is this process in control?
• Were assigned reviews occurring?
• Were meaningful issues being identified?
• Were they documented and communicated? Escalated and reported when needed?
• Was follow-up occurring?
• Were issues actually corrected?

We also redesigned the monitoring reports to document the CRA’s assessment of whether each site function’s oversight activities were in control — not the volume of data checked.

Ultimately, the person who adapted best to the Kenya monitoring model was a study coordinator in Mozambique. We brought her to the Kenya site and transitioned her into the lead CRA role despite her having no previous CRA experience. We taught her the role we needed rather than asking someone else to unlearn years of traditional monitoring habits.

From Visibility To Timely Action

One of the clearest tests of the model came with informed consent.

More than 1,000 completed consent forms collected in the field did not arrive at the site as expected. The site knew they were missing because the chain-of-custody process had been prospectively designed to make the expected movement of those forms visible.

The team traced where the chain of custody had broken, located the forms on the transport vehicle, investigated what had happened, communicated and reported the issue as appropriate, and implemented a CAPA.

The revelation isn’t simply that the forms were found. It’s that the process itself revealed they were missing in the first place. The site saw the signal and acted without waiting for a CRA to detect the problem.

Safety oversight demonstrated another dimension of the same principle.

Adverse events were a primary safety outcome. With thousands of participants receiving repeated doses of study drug, medical monitors expected certain types and frequency of AEs. Too few AEs could be as meaningful a signal as too many.

When medical monitors saw no AEs in certain communities, they knew “no events” wasn’t automatically good news. Investigation showed that, in these communities, some participants feared reporting AEs might result in their removal from the trial, while some field data collectors viewed AEs as a reflection of poor performance.

The opposite pattern mattered, too. When medical monitors identified an unexpected concentration of AEs in one community, they investigated the signal and traced it to a localized flu outbreak rather than a relationship to the study drug.

Ultimately, 2,796 AEs were reported among 1,975 participants across more than 56,000 treatments.

In both cases, a departure from expectations was the signal; visibility and judgment turned that signal into timely action.

The Lesson

The Kenya site later underwent a GCP inspection with no major or critical findings. Published results reported a 26% lower incidence of malaria infection with ivermectin than with albendazole and no identified safety concerns.

The Kenya trial concluded before ICH finalized E6(R3), but R3 now gives familiar language for the principles we put into practice: QbD, CtQs, fit-for-purpose approaches, and risk-proportionate oversight.

The lesson is simple: Risk-proportionate oversight is not about doing less. It is about designing oversight differently — across people, processes, and systems — so the right people have the right information at the right time to recognize what matters, exercise meaningful judgment, and take timely action.

About The Author:

Leslie M. Sam is president of Leslie Sam and Associates, LLC, a clinical research and GCP quality consultancy. She has more than 30 years of experience spanning clinical operations, quality systems, risk-based quality management, inspection readiness, risk and issue management, and process improvement. She has contributed to industry initiatives through DIA, ACRP, TransCelerate, and the Metrics Champion Consortium and is a frequent speaker and trainer on risk-based quality management, ICH E6(R3), quality by design, and clinical trial oversight. Her publications include work in the New England Journal of Medicine, Therapeutic Innovation & Regulatory Science and DIA Global Forum on clinical trial quality, oversight, and culture. She is a Certified Six Sigma Black Belt and Certified Quality Improvement Associate.