Global Clinical Trial Operations
The implementation of CAPAs (Corrective and Preventive Actions) in Phase 3 clinical trials presents some considerations less common in the better-controlled areas of pharma. In the GCP environment, often what we are trying to “correct” is actually human behavior. The well-known phrase “to err is human” may have also been written in reference to the conduct of clinical trials. People make mistakes, and overworked study staff who are burdened by ambiguous rules, regulations, and often peculiar sponsor demands are no exception. Many study staff do research as part of their job, and the clinical trial represents only a fraction of their employment obligations. Since GCPs go above what is the common every day work habit, it is easy to understand how mistakes can happen. When a CAPA is written after an inspection, what results often amounts to little more than retraining of study personnel – essentially “don’t do it again” or “remember to do it this way.” Meaningful behavior change is not accomplished, yet meeting a commitment date and filing the paperwork is. The sponsor company’s management is happy, and business at the clinical trial site continues as usual.
When other procedural or document changes are made by the sponsor with the goal of reducing these errors, the change can create a ripple effect throughout the organization and across the clinical trial landscape which in turn creates more errors than were sought to remedy. Rather than perform a true Root Cause Analysis and collect valuable input from the people closest to the work; those in HQ make adjustments to procedures, forms, and templates based on what they think will result in the desired change. For example, additional instructions or reminders might be added to an already crowded and complex work instruction or form. In some cases this makes the form even more “busy,” and the additional instructions add even more information that the reader/user needs to process in order to do their work and record the data. Why not simplify the form, or eliminate it altogether? The more that is added; the increase in the likelihood that errors are made (think lean). The more data points collected, the more opportunities exist for error. The more instructions that are added to a form, the greater the likelihood is that at least some of those instructions will not be followed.
Take an overly simple example: a form has a space or a box for the name of the person completing it. Instinctively people will record their name.
If, however, we add an instruction “record the name of the person filling out the form” we have (1) slowed the time to completion if the person pauses to read the instruction, and (2) created an instruction that now can be followed/or not followed. If we then add “Last, first” to the form, we compound the chance of error because we have given an instruction to fill in the name (which people will naturally fill in the way they “always” record their name), and then created an inherent second instruction by dictating a format. If we then revise our instruction to reiterate “Record the name of the person filling in the form in the Last, First format” we then slow the time to fill the form because the filler is now expected to read the instruction, think of the format required, then mentally compare that format to the format required in the data field itself as they actually complete it. Since this is only an administrative form and we only want to know who filled out the form, neither the extra instruction nor the name formatting add any value and we’ve only complicated something as simple as writings one’s name.
Another factor to consider is that when new forms, templates, and procedures are implemented or revised, it creates a ripple effect across the population of users. Those who were performing the process correctly are “punished” by having to stop and learn a change, and then resume their work which was previously being performed within specifications. For them, what “ain’t broke” is being “fixed” anyway. Not only is this disruptive; it also diverts focus away from paying attention to some other process that may be inherently more error prone.
Perhaps a better approach is to make these changes more organically than a top-down CAPA approach. The author is reminded of sidewalks being installed at a university campus based on the paths in the grass created by the students heading to class; not based on a parallelogram design that fits nicely between square buildings and parking lots. So it should be with process change. The CAPA should be written to specifically include analysis, testing, and refinement of the change before it is implemented, and further measurement after it is implemented. The testing and refinement must include those who do the work and use the form, otherwise how will “fitness for use” be determined? Too often a new SOP is written, when indeed the process is not standard, nor did it even previously exist. The designers develop intricate instructions, and prepare FAQ’s (frequently asked questions) when indeed no questions have yet been asked, nor have they been collected, organized, and systematically counted to determine the frequency of their occurrence. The mere fact that FAQ’s are developed at the implementation of a new procedure (or change to existing procedure) is in itself a clear indication that the design is flawed. Why should people not understand the first time, and have to ask so many questions? It is either not easy to follow, or the instructions are unclear or onerous. Why implement something known to be flawed? It has been said that a good leader tells the underlings what needs to be done, then leaves it to them to figure out how to accomplish it. Likewise, if there are certain regulations to follow, and people understand what they are; let them create the tools and process for following them. The GCP environment is less exacting than the GMP and GLP world; there is room for interpretation and ways of ensuring and demonstrating compliance. If a form or template is useful and stands up to the scrutiny of inspection – others may wish to adapt it as well. After all, it is usually a sheet of paper with a header, footer, and some kind of a grid in the center, only the details are different. This is not to promote anarchy within the conduct of clinical trials, just to allow the people who are actually doing the work to make incremental adjustments as needed. What worked well before is often still good enough now. If a rule or regulation cannot prove it to be inadequate, why must it be changed? Another simple example; if a signature line is added to the bottom of a form, one might assume that a signature is required. Even a regulator may ask, “why do you have a space for a signature if you don’t expect people to sign it – are you not following your own process?” Even worse, a second form is produced that lists the printed names (Last, First, of course) and the corresponding signatures so that it can be matched up to the first form that was signed. Was a signature even required? It already had the person’s name on it.
Finally, there is the tendency to not measure the changes that are made by the folks at HQ. Procedures and forms are updated, revised, and communicated, then revised again without measuring the effect of the previous change. Did everyone receive the communication and did they understand the change? Did they receive the new form? Where do they get the form? If obtaining the new form is any more cumbersome than clicking “File...Save As…” we can be sure people will be using the old form that they completed the header of last time.
This is exactly the converse of the scientific maxim of changing only one variable at a time. It is difficult enough to have consistency across study sites without continually changing the substrate on which they operate. We now have several sets of forms and instructions that can be mismatched, resulting in an ambiguous implementation period and a compliance (small “c” if not big “C”) problem.
A CAPA should be viewed as an opportunity to improve, not a requirement to fix. Was the human error due to too little instruction/information, or perhaps too much? In daily life, how many people bother to read the instruction manual when they can rely on a “Quick Start” card? Yet appliances, cell phones, computers, and power tools manage to be used every day without incident. In the GCP world, what really needs to be taken into account in the writing and implementation of any CAPA are the ever-present potential for human error and the tendency of the sponsor to over-design.
Simplification should mean removing complexity, not repackaging it. Clinical trials are rooted in medicine and the endeavor to reduce or eliminate human suffering. Those that tinker with the systems and tools that are vital to this mission should also remember, “First, do no harm.”
Author Biography
Stuart Halasz has been at Merck for more than 25 years and has worked in the areas of data management, clinical research, clinical trial standards, and clinical research operations. He holds a master’s in quality assurance and regulatory affairs. In 2010 he transitioned from a CRA training role to Clinical Quality Management. He has extensive experience in CAPAs, audit response and inspection readiness for phase III trials.