Why FDA Issues Complete Response Letters: 72% Cite Manufacturing

Of the 445 complete response letters FDA has published (newest dated 7 August 2026), 320 (72%) cite a manufacturing, facility or CMC deficiency and 115 (26%) cite a clinical one. 217 cite manufacturing and none of the other reasons counted here. 34 cite clinical alone. This is a count of published letters. It is not a count of every letter FDA has sent.

Source: FDA's openFDA CRL transparency API, counted by scripts/analyze-crl-reasons.py. A letter can cite several reasons, so the first four rows add to more than 100%. Download the data (CSV). The same letters are searchable in the CRL search. For what a CRL is and what follows one, see the glossary entry on complete response letters.

What the published letters cover

The dataset holds 458 records. 445 are complete response letters. The other 13 are tentative approvals, a refusal to file and similar actions. FDA put the letters out in three steps, and the steps decide which letters are here:

• On 10 July 2025 FDA published more than 200 letters for applications that were later approved.
• On 4 September 2025 it added 89 letters issued from 2024 onward for pending or withdrawn applications. It said it would release new letters promptly and publish all earlier letters for an application when it approves it.
• Letters since then arrive as FDA publishes them.

So 285 of the 287 letters dated 2023 or earlier are for applications that openFDA marks as approved. Of the 158 letters dated 2024 or later, 146 are for applications not approved when FDA published them. The years are not equally complete, and we cannot see how many letters FDA sent in any year. Read the yearly counts as published letters, not as a trend in how often FDA rejects applications.

Three more limits apply. FDA redacts trade secrets and confidential commercial information. The letters are scanned documents, so some headings are garbled: 48 letters have no substantive heading we could extract. And the data has no field for sponsor size, so this post does not split by it. Guessing size from a company name would add error, not information.

Reasons by year

The table is a dated copy of the figure above it, as of 11 November 2026. The figure reads the same openFDA data as the CRL search, so it moves when FDA publishes more letters and this table does not.

In every year with 20 or more letters, manufacturing is the most cited reason. It ranges from 64% (2025) to 84% (2018). Clinical sits between 16% and 45%. The mix is close in the two selection periods: manufacturing is cited in 71% of letters dated 2023 or earlier and 73% of letters dated 2024 or later. Clinical is 28% and 23%.

Labeling is a format effect, not a trend. 33 of the 35 letters with a labeling section are dated 2023 or earlier. Current letters put labeling comments under standing sections that nearly every letter carries, and this count leaves those out. We did not establish why the heading changed.

What a manufacturing citation looks like

217 letters have a facility inspection section. In 189, FDA says it conveyed deficiencies to the facility after an inspection. In 25, FDA says it had not completed an inspection or could not conduct one. The reasons FDA gives include travel restrictions in 2020 and a site that was not ready for inspection.

The letter does not list what the inspector found. Those observations are in the Form 483 the site received. A site's earlier inspection outcomes are public in the inspection database, and the most cited observations are in Most Cited FDA 483 Observations.

Named sites. 83 letters name a site by FEI number, for 59 distinct sites. 81 of those letters are dated 2023 or earlier. Only 2 of the 158 letters dated 2024 or later name a site this way. 221 of the 445 letters carry FDA's redaction mark for confidential commercial information. A letter does not say whether a named site is the sponsor's own or a contractor's. So this post does not count third-party manufacturing sites and does not rank any. The FEI number entry explains how to look a site up.

By application type and letter order

BLA applications are counted from the application number. 14 BLA numbers are scanned as "BL", and the count includes them. The first and later rows use the order of published letters, not FDA's review cycle. 82 applications have more than one published letter, and a "later" letter may follow a first cycle that FDA has not published.

Clinical problems are a smaller share of later letters (15%) than of the first letter we can see for an application (29%). The data cannot say why.

After the letter: approvals in Drugs@FDA

We joined each letter's application number to the Drugs@FDA bulk file (updated 11 November 2026) and looked for an original approval dated after the letter. Per application, the clock runs from the latest published letter before the approval. The 333 applications split like this:

• 225 have an original approval dated after a published letter.
• 85 have no original approval in Drugs@FDA.
• 12 are cell therapy, gene therapy or similar products that FDA's biologics center (CBER) reviews. Drugs@FDA does not list them, so they are left out of the counts below.
• 11 show an original approval on the same day as the letter or earlier. That happens when FDA splits an application into two originals and acts on them separately, and when a letter concerns a supplement. We cannot date these.

Read this table with three cautions.

• Selection. FDA's first release covered only applications that were later approved. The 2009 to 2023 row says nothing about how often a letter is followed by approval. It shows how long approval took where it came.
• Time. Recent letters have had little time. The 2025 median is short partly because slower approvals have not happened yet.
• Mixed releases. The letters dated 2024 include 12 that FDA published in the first release, for applications already approved.

Applications with a letter dated 2024 or 2025 split by the reasons in their latest letter:

These are counts of what has happened to 94 applications so far. They are not a rate to apply to a new application. Nine applications is too few to compare, and every number here changes as more approvals post.

For context, a 2014 study of 302 new molecular entity applications found a median delay of 435 days for resubmissions (Sacks et al., JAMA). A 2026 analysis of published letters found facilities cited in 65% and CMC in 51% of the letters it studied (Dilek et al., Therapeutic Innovation & Regulatory Science). Both used different samples and definitions from this post.

How we counted

Every letter is organized under FDA's own ALL-CAPS headings. We read the headings, not the prose, because most of a letter is standing text. Our earlier post explains this.

• Manufacturing, facility or CMC: a heading containing FACILIT, PRODUCT QUALITY, MICROBIOLOGY, CMC or MANUFACTUR. "Clinical microbiology" is excluded.
• Clinical: a heading starting with CLINICAL, or the headings SAFETY and STATISTICAL. This includes clinical pharmacology.
• Nonclinical: NONCLINICAL, NON-CLINICAL or TOXICOLOGY.
• Labeling: a LABELING heading. Prescribing information, carton and container labeling and proprietary name sections are standing sections and are not counted.

This count is wider than the earlier post's. That post counted exact headings and found 290 letters (65%) for manufacturing. This one also counts spelling variants such as "Product quality/facility inspections" and "Chemistry, manufacturing and controls", and finds 320 (72%). Both are right for their stated definitions.

Approval dates come from the first original submission marked approved in the Drugs@FDA bulk file from openFDA. A tentative approval does not count. The per-letter file with each later approval date is at crl-follow-up-approvals.csv.

How the numbers update

The figure above is computed on the server from the same openFDA feed as the CRL search, with the same committed fallback if openFDA is down, and refreshes at most daily. The tables and prose are dated. They come from scripts/analyze-crl-reasons.py, which re-reads openFDA and Drugs@FDA and rewrites both CSV files. We re-run it each quarter and update this post.

Frequently asked questions

What is the most common reason for an FDA complete response letter?

Manufacturing, facility or CMC problems. 320 of the 445 published letters (72%) cite one. Clinical deficiencies appear in 115 (26%), nonclinical in 56 (13%) and labeling in 35 (8%). A letter can cite several reasons.

How many complete response letters has FDA published?

445, with the newest dated 7 August 2026. This counts published letters, not every letter FDA has sent. FDA published them in batches, so the years are not equally complete.

Are most complete response letters about clinical data?

No. 217 letters (49%) cite manufacturing and none of the other reasons counted here. 34 (8%) cite clinical alone, and 69 (16%) cite both.

Does a manufacturing citation mean the inspection failed?

Not always. 217 letters have a facility inspection section. In 189, FDA says it conveyed deficiencies to the facility after an inspection. In 25, FDA says it had not completed an inspection or could not conduct one.

How long does approval take after a complete response letter?

For applications whose latest published letter is dated 2009 to 2023, the median was 423 days to approval, and the middle half ran from 300 to 662 days. That group was selected because FDA's first release covered only applications later approved, so it shows how long approval took where it came, not how often it came.

Where can I search the letters?

In the CRL search, which reads the same openFDA data. The counts in this post are also available as a CSV.