What FDA cites compounders for
We parsed the FDA compounding action-page snapshot collected September 3, 2026, and extracted observations from 39 selected documents. This is an analysis of published records, not a census of inspections or a prediction of your facility’s risk.
Ten things the record shows
Compounders are cited for how people behave in the room
Personnel garbing and aseptic technique accounts for 11.3% of compounder observations, against 0.4% of all FDA drug-program citations in fiscal years 2021 through 2025. That is a 28-fold over-representation. Facility design and airflow runs 9.5 times over-represented, environmental monitoring 4 times. These are ratios of citation shares in differently selected and categorized datasets, not relative inspection or failure rates.
“An operator blocked first air by placing gloved hands and/or forearms directly over open tray of stoppers intended to be sterile during critical in-process operations within the ISO 5 production areas.”
Cleaning and paperwork appear in seven of every ten documents
Counting how many documents mention a topic at all, rather than counting individual lines, cleaning and documentation each appear in 71.8% of the 39 selected documents, including 483s, warning letters and untitled letters. Sterility assurance appears in 69.2%. These sample proportions cannot estimate the probability of a finding at another facility.
The largest single bucket is still a missing record
Documentation and records, meaning batch records, standard operating procedures, the quality control unit, investigations, deviations, and corrective actions, is 27.8% of all observations. Most of what an inspector writes down is about a record that does not exist, was not followed, or was never reviewed. Not about a contaminated product.
“The responsibilities and procedures applicable to the quality control unit are not in writing and fully followed.”
Every warning letter used the same two words
All 12 compounder warning letters in the sample cite insanitary conditions under section 501(a)(2)(A) of the Food, Drug, and Cosmetic Act. It is the standard legal hook, and it is assembled almost entirely out of ISO 5 behavior, disinfection practice, and the physical condition of the facility. The categories in finding 01 are what the phrase is made of.
49 facilities had both a 483 and a warning letter
Of the 174 facilities that received at least one Form 483 between 2021 and 2026, 49 also received a warning letter. That is 28.2%. This facility-level overlap does not establish that a particular 483 escalated, the order of events, or an escalation probability. FDA generally allows 15 business days for a voluntary Form 483 response; check the correspondence received.
Outsourcing facilities account for 153 of 278 posted 483s
503B outsourcing facilities are 69 of the 246 cited firms, about 28%, but they collect 153 of 278 Form 483s. The comparison is limited to facilities represented in this dataset. Without the total inspected and operating populations, it does not measure inspection likelihood or attribute the difference to a single cause.
Enforcement dipped, then turned back up
Warning letters fell every year from 2021 through 2024, then more than doubled in 2025 to a five-year high. The 2026 row is partial and FDA backfills its posting page, so recent-year counts may change.
| Year | Form 483s | Warning letters | All actions |
|---|---|---|---|
| 2021 | 70 | 15 | 173 |
| 2022 | 60 | 12 | 218 |
| 2023 | 37 | 10 | 147 |
| 2024 | 41 | 8 | 77 |
| 2025 | 54 | 17 (five-year high) | 98 |
| 2026 partial | 16 | 6 | 39 |
Three states carry a third of everything
Texas, Florida, and California together account for 35.5% of the 744 state-coded actions across 41 states. Add New Jersey and Arizona and five states reach 47.4%. These counts have no denominator for the number or output of facilities in each state, so they cannot rank state-level risk.
The same facilities keep getting written up
Of 174 facilities with a 483, 66 received two or more and 28 received three or more. One outsourcing facility in Allentown, Pennsylvania has ten separate enforcement events on the public record. Repeat entries do not, by themselves, show whether the same deficiency recurred or what happened between inspections.
Training is rarely the citation, but it is all over the detail
Training and competency is the headline of only 0.6% of observations, yet training language appears somewhere in 17.9% of the documents. This difference reflects headline categorization versus a text search across the document. Review the underlying observation before inferring that training caused the event or that a particular corrective action was inadequate.
“Personnel infrequently changed and sanitized gloves to prevent contamination.”
Which records need closer review?
A self-assessment against the areas above. It summarizes your answers, shows selected gaps by category, and suggests supporting records for at each one. No call, no salesperson, and you keep the findings either way.
Run the readiness checkMethod and limits
Actions were parsed from FDA’s compounding inspections, recalls, and other actions pages covering 2021 through September 2026, giving 752 actions against 246 distinct facilities in 41 states. Observations were extracted from 39 documents, 24 Form 483s, 12 warning letters, and 3 untitled letters, issued between July 2023 and June 2026, producing 309 individual observations. Each was categorized by hand-written rules over FDA’s templated citation language, with a weighted keyword fallback and the method recorded per row.
Primary sources: FDA compounding actions, FDA inspection citation datasets, and FDA’s explanation of Form 483. A Form 483 is not a final agency determination. The snapshot analysis and automated categorization are CompliRx’s work, not FDA conclusions.
What this analysis cannot tell you:
- Thirty-nine documents is a sample, not a census. The smallest categories, hazardous drug handling and training, rest on two observations each and are anecdotes rather than rates.
- The document sample skews toward 503B outsourcing facilities, which carry a CGMP obligation a 503A does not, so the category mix leans that way.
- FDA posts selectively and backfills its pages, so recent years understate the eventual totals.
- The action count includes posted recall entries but is not a complete recall census. Use FDA’s separate recall resources to investigate recalls.
- Many Form 483s are scanned images, and some observation text carries optical character recognition errors. Quotations above were selected from clean passages.
The comparison population in finding 01 is FDA’s published inspection citation data for all drug firms, fiscal years 2021 through 2025, used for a descriptive citation-share comparison. Selection, category definitions and time windows differ; this is not an estimate of relative facility risk.
Review your own documentation gaps.
The twelve-question readiness self-assessment covers the same areas this analysis ranks: cleaning, environmental monitoring, personnel competency, certification, and documentation.