We’re often called in to help respond to the FDA following an inspection that results in observations or an enforcement action, such as a warning letter or consent decree. After an inspection, we’re sometimes handed a draft response to a Form 483 that urgently needs an expert to review before it’s sent off to the agency.
Many of those “first draft” responses hurriedly describe plenty of work the firm intends to do without showing whether the problem is actually under control, or at least that steps are being taken to get there. Procedures may have been revised, employees may have been trained, suppliers may have been contacted, but the reader (the FDA) still can’t tell what actually changed in the operation, what evidence supports the change, or what’s left to do. That’s the difference between reporting activity and demonstrating control.
One of the Form 483 response projects we were involved in clearly shows this difference. Its self-authored response to the FDA’s follow-up covered Quality Unit oversight, production and incoming inspection, labeling, and oversight of outsourced manufacturing and testing.
The FDA laid out the commitments and records it wanted from the firm, and that list is an incredibly useful guide to what a strong response needs to prove. Here’s a quick look at how we approach this.
What a Form 483 response needs to establish
Let’s first back up and level-set with the basics:
A 483 lists inspectional observations. According to the FDA’s Form 483 FAQ, it “does not constitute a final Agency determination,” and there “may be other objectionable conditions that exist at the firm that are not cited.”
The FAQ also says companies are responsible for addressing “the cited objectionable conditions” and “any related non-cited objectionable conditions that might exist.” This point is so often missed. A response that fixes only the example the investigator wrote down can often fall short.
Earlier this year, the FDA published a draft guidance, Responding to FDA Form 483 Observations at the Conclusion of a Drug CGMP Inspection, that lays out the agency’s current thinking on this topic.
Among other things, it recommends a single, complete response within 15 business days of the 483’s issuance. When the work is too complex to finish in that window, the response should include at least a CAPA plan and a proposed timeframe. It also recommends that a member of executive management with the authority to carry out the commitments sign the response and places significant emphasis on verifying the effectiveness of CAPAs.
Here are a few lessons from the project we worked on that put those principles into practice.
1. A revised SOP isn’t, on its own, evidence of implementation
The firm’s Form 483 cited quality procedures that were missing or weren’t being followed. When the FDA followed up, it didn’t stop at asking for procedures. It asked for in-process inspection records covering a defined operating period, the nonconformances and their associated investigations, and a complete list of complaint records.
In other words, the agency wanted to see the process running, not just written down.
Updating an SOP may be very necessary, but it doesn’t show that the new process is operating or that it prevents recurrence. While every situation is different and it’s reckless to issue any universal rules here, we tend to organize each response item around four stages because each one calls for different evidence:
Immediate correction or containment: what you did to address the condition and manage the current risk.
Systemic action: what changed to address the investigated cause.
Implementation evidence: the executed operational records that show the revised process is in use.
Effectiveness evaluation: how you’ll show, over an appropriate period, that the action solved the problem.
An approved SOP supports a claim that a procedure was revised. Training records support a claim that training happened. But that’s where most teams tend to stop. Only executed records support a claim that the process is implemented.
For every sentence in your response that says “completed,” identify the attachment that proves that exact claim. If effectiveness hasn’t been evaluated yet, say so and describe the plan and the interim controls. Again, the FDA recognizes there’s a time crunch here. If you don’t have enough time to get something done now, give them your plan with a timeline.
2. Explain why the procedure and the work didn’t match
The inspection in this case documented two concrete disconnects.
In one case, a procedure specified a sampling plan, but the people performing the inspection first conducted a single-container check and then random checks.
In the other, incoming size verification relied on reference samples stored outside the receiving area, and verification wasn’t consistently performed.
“We retrained staff” doesn’t answer either one. Retraining doesn’t move the reference samples to receiving. The response has to explain how the prescribed method will be performed, including the materials, access, instructions, and records available at the work site.
Try to investigate three questions separately here:
Execution: Could people perform the step as written with the tools, information, staffing, and access they had?
Procedure design: Was the method clear, technically appropriate, and aligned with the real operation?
Oversight: Why didn’t routine review, deviation management, or internal audits catch the mismatch?
Don’t fall into the trap of calling “human error” or “failure to follow the SOP” the root cause just because it describes what it looked like on the surface. The FDA wants you to test plausible causes against records, interviews, and operating conditions. If the investigation is still open, separate confirmed facts from working hypotheses/theories you’re looking at.
Then walk through the revised process at the point of use. Confirm that materials and information are available, that responsibilities are understood, that exceptions have an escalation path, and that the records capture what happened. Document that verification.
3. Fix the connected controls, not just the visible failure
A few labeling observations covered several linked problems:
Obsolete and current artwork stored together.
No established way to confirm the current revision.
Unrestricted access to printed labels.
No documentation of artwork review.
One labeling observation was explicitly identified as a repeat.



