Quality by Design (QbD) and Risk-Based Monitoring (RBM) are frequently discussed in modern clinical research. Rather than attempting to eliminate every minor error, the emphasis has shifted toward protecting participant safety and ensuring the reliability of critical data through a risk-based approach.
A recent discussion published by RAPS following the DIA Global Annual Meeting explored these principles in the context of FDA inspections and Form 483 observations. One example used during the discussion was particularly memorable.
A clinical trial was compared to a field of corn. Each patient represented a stalk, each data point a kernel, and the CRA was expected to inspect every kernel on every cob in every row of the field. The message was that expecting a CRA to examine everything is unrealistic, and criticizing them for missing only a few "kernels" among billions is equally unreasonable. (https://www.raps.org/resource/fda-investigator-experts-seek-to-dispel-misperceptions-around-483s.html)
The analogy is memorable, but I found it difficult to interpret. Were billions of kernels actually examined or re-examined? Without knowing how many kernels were assessed, it is impossible to estimate what error rate the seven missed kernels represent or what they tell us about the quality of the entire field.
A more practical way to think about CRA workload comes from everyday clinical operations.
Imagine a CRA responsible for managing 1,000 data queries during a study.
Over the course of the trial, the CRA successfully resolves 999 of them.
One final query remains open.
That single unresolved query delays database lock.
The critical question becomes:
"Why wasn't this last query closed? It only takes about 15 minutes."
The more difficult question is rarely asked.
But what workload was required to resolve the other 999?
Assume, optimistically, that each query requires only 15 minutes to review, communicate with the site, document, verify, and close.
That represents:
1,000 queries
× 15 minutes
= 250 hours of work (approximately 1.5 months of full-time effort)
In reality, many queries require repeated follow-up with study coordinators, investigators, data managers, or sponsors. Some remain open for days or weeks while waiting for information that is completely outside the CRA's control.
This raises another question: where were those 250 hours planned and budgeted?
Were they considered when allocating CRA capacity?
Sometimes the last unresolved query is not evidence of poor CRA performance.
It may simply reflect that the operational effort behind the previous 999 resolved queries becomes largely invisible once the project is delayed by the last one.
P.S: While the cornfield analogy illustrates the scale of clinical trial data, the unresolved-query example illustrates something different - the often invisible workload required to produce and maintain that quality.
