One row per deficiency type. Applications is how many distinct applications cite it; eventually approved and months to approval show what happened next. The vertical line marks the all-application base rate — dots right of it beat the average, left of it lag. Click any row to filter the letter explorer below.
| Deficiency type | Apps | Share of applications | Eventually approved | Median months to approval |
|---|
The three-way rollup: every blocking deficiency mapped to Manufacturing, Clinical or Other. Routine labeling mark-ups are excluded — they appear in almost every letter and would swamp "Other".
Disease area comes from the approved label where one exists, otherwise the condition named in the letter. Applications still pending often have neither, so are unresolved and shown last — that is missing data, not a category.
| Disease area | Apps | Share | Eventually approved | Median months |
|---|
Select the deficiencies cited in a CRL. Every number below is computed live from the matching historical applications — no model, no extrapolation. Small matches are flagged, because a rate over five applications is not a rate.
Transparency coverage is deliberately uneven — FDA released a near-complete modern record and a thinner historical one, so pre-2013 counts reflect disclosure, not FDA behaviour.
A quarter of applications receive more than one CRL. Persistence is not obviously penalised — hover each bar for the approval rate at that cycle count.
Every letter, with the passage the classification was drawn from. Click a row for the evidence.