The Much-Feared FDA Form 483, Part 2
By Robert Califf, MD
In Part 1 of this series on the FDA’s use of a regulatory instrument known as Form 483, I put forward my view of the form and how I hope it will be used to improve quality across the board rather than as a single summary estimate of a firm’s quality. In Parts 2 and 3 I will provide opinions about the use of 483s in two specific circumstances: clinical evidence generation (AKA clinical trials) and generic drugs.
As with all of my posts here, I want to acknowledge that I’m articulating a personal perspective that’s the result of decades of working on all sides of clinical practice and industries regulated by the FDA. I recognize that experts may have better ideas and that I could be wrong in these opinions, but I hope these posts will prompt reflection and generate enthusiasm for improving the system.
One theme that I’ve visited multiple times in posts on this blog is my belief that our system of evidence generation is poised to deliver a much more comprehensive base from which better decisions can be made about medical care, medical product development, and public health policies. Over time, however, efforts to close the gap between the evidence we need and the evidence we have were hindered by the accrual of complex methods and interpretations of regulations that have gummed up the system, rendering it too slow, inefficient, and expensive. The FDA’s Form 483 is one of many important examples of a tool that could either contribute significant benefit or add to systemic inefficiency across the entire spectrum of evidence generation, depending on how it is used as a tool for ensuring quality.
Evidence about medical products is expanding beyond traditional clinical trials to include real-world data (RWD) and real-world evidence (RWE). At any one time, there are thousands of trials taking place; with the addition of RWD and RWE and the vast global expansion of trials, the remit of FDA oversight of clinical trials involving medical products is enormous. Over 5000 new trials per year fall under FDA oversight, and thanks to the legal requirement to register applicable clinical trials and report their results, we can be confident that these numbers are accurate.
The workforce at FDA specifically devoted to conducting inspections and investigations of clinical research is small. The budget determined by Congress only allowed 109 positions in BIMO (Bioresearch Monitoring) during my second tenure at the agency. Its remit is vast, covering the integrity and quality of research, the protection of human research subjects/participants, and animal welfare. And it’s not just medical products, but also food and food additives (including things like artificial colors), animal products, and tobacco and tobacco-related products. In fairness, the product Centers work collaboratively so that the total workforce concerned with trial quality is larger than just BIMO. Still, the fact is that even with this caveat, the workforce is entirely too small for “hands on” oversight of a vast universe of research.
The investigators assigned to BIMO are hardworking and dedicated, but not all are experts in the science of clinical trials and observational studies. They often learn on the job, and a huge amount of effort is put into the training system. (This is, by the way, typical of most who work in the clinical trials industry—they come from various disciplines and backgrounds and learn operational methods through a variety of training and educational programs and on-the-job training.)
A ROBUST SYSTEM FOR ENSURING QUALITY
None of this should be taken as discrediting the existing system for evidence generation about medical products. The system itself and the accompanying requirements for evidence in product development and labeling represent major advances in human civilization, with well-documented impacts on health. The system’s quality is ensured by multiple levels of oversight, including review of protocols by investigators, sponsors, regulators, and institutional review boards (IRBs), which also oversee trial conduct.
In studies funded by the medical products industry, it’s routine to have plans for monitoring and auditing to assure that the protocol is working and that the data meet standards outlined in FDA and Good Clinical Practice (GCP) guidance. (In this context, monitoring generally refers to ongoing assessment of protocol adherence, data management, and human research participant protections, while auditing generally refers to periodic independent assessment of compliance with regulations.)
For studies that are important to product labeling, the FDA is the only regulatory agency that performs an independent analysis of the trial data. And in fact, a method called statistical process control, in which patterns in the data are assessed, has become a standard approach for detecting fraud or sloppiness in the conduct of research. In the future, AI will play an important role in surveilling data for evidence that it either looks “too good” or shows patterns that are not plausible.
Given this extensive system of quality control for clinical studies, it’s reasonable to ask: What is the role of Form 483, and why is it still needed?
One answer is that the form continues to be a valuable indicator of a need for closer scrutiny. When an FDA investigator leaves a 483 at the coordinating center for a clinical trial, or with a study sponsor or a research site, serious attention is merited. However, the question of context remains. I see two major problems with the way the 483 has been used in the clinical research setting:
THE POTENTIAL FOR UNWARRANTED REPUTATIONAL HARM
Early in my career, I was shocked to learn that issuance of a 483 is often enough for a clinical research site to be “blackballed” from participation in future trials. I’ll never forget that during a major push to make large clinical trials more representative of the populations in which therapies would eventually be used, we recruited a site that cared for a large proportion of immigrants. When the site was inspected, the records were a mess. But even though the problems in no way impacted the interpretation of the trial’s data or findings, not only was a 483 issued, but the site investigator was also barred from participating in FDA-regulated trials for a period of time.
CREATING PERVERSE INCENTIVES
The second issue is how individuals and institutions respond to the role of inspection in the overall research system. There is substantial literature pointing out that when quality improvement is seen as a positive mission, collaboration with intent to improve quality leads to real improvement. On the other hand, when it is seen as punitive, deviant behavior can occur, including hiding problems and punishing employees for honestly identifying and admitting issues.
PROLIFERATING COMPLEXITY
Another serious concern is the tendency to react to a problem or a reported problem with a fortified set of standard operating procedures that adds substantial human labor and may or may not improve quality and are typically never assessed to determine their value by an objective standard. To make it even more difficult, when someone suggests taking away a layer of checking, concerns are usually raised about “lowering quality” or making the studies less safe.
FUTURE DIRECTIONS
As the system evolves, we will see more RWD and RWE incorporated into regulatory submissions. The remarkable progress in ascertainment and curation of data collected in clinical practice and from devices used in daily life (e.g., smartphones, wearables, monitors, etc.) could cause a dramatic improvement in evidence generation at a much lower cost. It will be important for the research community to have significant input into FDA guidance that will provide the basis for agency investigators to draw conclusions about regulatory compliance in this rapidly evolving data environment. A practical approach that improves quality in FDA-regulated research could support this potential revolution in the pace and efficiency of the evidence generation system.
I believe the basics of such a system have been articulated in the combination of the recent update to GCP guidelines, the Good Trials document from the Wellcome Trust and Gates Foundation, and the Quality by Design approach described by both the Clinical Trials Transformation Initiative and the FDA. The main point is that one size does not fit all: placing regulatory requirements in context requires expertise in the methods and rationale for clinical investigation, an essential component of regulatory science.
Finally, I’ll stress again the potential role of AI in identifying discrepancies and irregularities in trial data and conduct. We can expect sponsors to augment trial monitoring and auditing with AI-driven statistical process control while the FDA will use AI technologies to guide inspections. The veracity and quality of clinical studies will increasingly be assessed in real time rather than through intermittent monitoring and inspections conducted at the end of study alone, although these time-honored methods will remain important.
I hope we’ll change the ethos of the 483 so that it’s understood less as a punitive instrument (as it is too often regarded now) and more as a positive element of a broad quality ecosystem in evidence generation that signals a need for improvement.
In upcoming third and final installment of this series, I’ll be taking a look at the role of Form 483 in the specific context of generic drug manufacturing.
EDITOR’S NOTE: This article originally appeared on Dr. Califf’s Substack on March 16, 2026. It has been reprinted here with his permission.
Bio:
Dr. Robert Califf practiced intensive care cardiology, outpatient cardiology, and did clinical research for more than 30 years. He founded the Duke Clinical Research Institute and later served as 23rd and 26th FDA Commissioner. He also worked at Alphabet as a senior advisor from 2017-2022.