A medicine needs more than clinical evidence.

Explore four review gates, then read the scope behind two published sets of complete response letters.

Teaching model
Readiness requirements, scoped historical datasets and candidate formulation checksFour separate requirementsClinicalevidenceReadyQuality /processReadyFacilityinspectionReadyLabelingReadyAll four marked readyTeaching checklist; no approval prediction.202 letters · trade analysis2020–24 letters · later-approved applicationsInitial July 2025 release; selected set.150 flagged quality / manufacturing74.3% of those lettersCoding not fully published.Not an all-drug failure rate.Candidate checksDissolvesMeets checkRemains stableMeets checkRepeatable processMeets checkIllustrative formulation / process flags.

Enlarge for bigger labels, then swipe sideways →

Open requirements0
Candidate checks needing work0
Dataset scope202 letters

All four teaching gates are marked ready; this is not an approval prediction. The illustrative candidate passes the solubility, stability and repeatability checks. Changing a gate does not change the historical dataset.

What this model assumes

The readiness gates and candidate flags are teaching examples, not FDA approval rules or a probability model. Marking every item ready does not establish that an application is approvable. The trade analysis reported quality/manufacturing issues in 150 of 202 letters in the initial July 2025 release of 2020–24 letters for later-approved applications, with coding methods not fully published. It is not an all-drug failure rate. The separate peer-reviewed study examined letters for 43 ultimately approved therapeutics and reported overlapping categories. Its categories, denominator and period differ; these percentages must not be combined or added. Candidate flags describe formulation/process questions and do not estimate clinical efficacy, toxicity or commercial viability.

A review letter identifies work still to do.

A complete response letter means FDA cannot approve an application in its current form. The initial July 2025 public release contained letters from 2020–24 associated with applications that were ultimately approved. These are not all drugs that failed development.

The trade analysis reported quality or manufacturing issues in 150 of 202 letters: 74.3%. Its coding method was not fully published. The count describes that release, not the probability that a new drug will fail for manufacturing.

A separate peer-reviewed analysis covered 43 ultimately approved novel therapeutics. Facilities appeared in 65%, product or process quality in 51%, labeling in 44%, efficacy and safety in 26% each, and pharmacokinetics or pharmacodynamics in 14%. Categories overlap; adding them is not meaningful.

The four gates and candidate checklist are teaching examples, not an FDA decision algorithm. Adequate clinical evidence, quality, facilities and labeling all matter. A candidate that dissolves, stays stable and can be made consistently still needs extensive development and review.

Why does 150 of 202 not mean that most drugs fail because of manufacturing?

Taking a real process
beyond the lab?

Jon and I are building RARE Labs toward an autonomous pilot plant builder using AI and robotics. Join the waitlist for updates, and tell us what you’re trying to scale.

Join the RARE Labs waitlist ↗

Go a little deeper.