FDA writes in a careful, load-bearing register. We agree, we do not object, we recommend, and at this time, we do not agree are not interchangeable — each carries a precise weight, and getting the weight wrong can cost a program years. Most people never get to learn the code, because complete INDs are almost never public.
This guide is drawn from two complete dossiers in the archive — the Reloxaliase (IND 100,169) and ALLN-346 (IND 143,480) files donated by Lumen Bioscience. They let us do something rare: not just translate the FDA's language, but validate the translation against what happened next. Every entry below pairs a verbatim FDA quote with the downstream event in the same file that confirms what it really meant. (All quotes are taken verbatim from the letters; each FDA letter carries a Reference ID for verification.)
1. The agreement thermometer
FDA's approval verbs form a precise gradient. The program's fate tracks them exactly.
| What FDA wrote (verbatim) | Register | What it means | Confirmed by |
|---|---|---|---|
| "We agree that rodent carcinogenicity studies will not be needed to support a BLA for ALLN-177." (Reloxaliase, Oct 2016) | Genuine yes | A settled concession; the issue is closed. | The carcinogenicity question never reappears in the file. |
| "We generally agree … however … the final determination of the adequacy of the comparability data will be made during the assessment of the BLA." (Reloxaliase CMC, Jul 2021) | Deferred yes | Agreement in principle that binds FDA to nothing; the real decision is reserved. | Every one of the five CMC answers in that letter ends by routing the decision to "BLA review." |
| "We conditionally agree with the amendment to Study 202 to enroll subjects in Cohort C … As this is contingent on the interim safety analysis…" (ALLN-346, Dec 2021) | Conditional yes (yes-if) | Permission unlocked only when you meet a named, pre-specified gate. | The same sentence names the gate (Cohort A/B interim safety) the sponsor must build. |
| "We do not object to your proposal to reduce the sample size of URIROX-2 … We note, however, that maintaining your original sample size … could increase your power…" (Reloxaliase, Jan 2020) | Permission, risk on sponsor | You may proceed, but you own the consequences — and FDA has logged its reservation. | Six weeks later, FDA "reminds you" the approval standard is unchanged (see §3). |
| "At this time, we do not agree." (ALLN-346 reproductive tox, Dec 2021) | Firm no, door ajar | No today; the "at this time" invites a future yes only with new evidence. | Followed immediately by a list of studies required before the answer could change. |
| "No, we do not agree. … these data are not sufficient to support the use of small changes in 24-hour urinary oxalate excretion as a surrogate endpoint…" (Reloxaliase, Sep 2015) | Hard no | A flat rejection that will hold until the premise changes. | FDA repeats this "No" at the 2017 EOP2 and again in 2018 (see set piece A). |
Takeaway: "We do not object" is the weakest yes in the language — it means you carry the risk. "We generally agree … determined at BLA" defers everything. Only a clean "We agree" closes an issue.
2. Soft verbs that are hard requirements
When recommend, should, or encourage appear inside an agreement or an action list, they are usually requirements wearing a polite costume.
- "We recommend a reproductive toxicology program … and an embryo-fetal development (EFD) study in rats (or rabbits) be completed prior to the initiation of any phase 3 clinical trials." (ALLN-346, Dec 2021) — "recommend" + "prior to Phase 3" = a hard gate on the entire program.
- "kcat and Km specifications will be a requirement for licensure and … ideally the assay should be put in place prior to the release of Phase II material." (Reloxaliase CMC, Feb 2013) — stated outright as a requirement; "ideally" sets the deadline.
- "The Agency strongly recommended including peptide mapping for identity testing." (Reloxaliase, Feb 2013) — Confirmed: in the same minutes, "the sponsor agreed to use peptide mapping for drug substance release specifications." A strong recommendation the sponsor treated as an order.
- "…immunogenicity/antidrug antibodies (ADA) should be assessed in all patients." (ALLN-346, Dec 2021) — overrides the PK waiver granted two sentences earlier; ADA testing is now mandatory in every cohort.
Takeaway: parse the verb and its consequence. "Recommend … before Phase 3" is not advice — it's a precondition.
3. "We remind you of our [date] letter" — the position is locked
When FDA cites its own earlier correspondence, it is doing two things: signaling the position will not be reopened, and building an administrative record that you were on notice. The clearest example is the accelerated-approval standard for Reloxaliase, stated once and then repeated verbatim as a reminder — three times in eighteen months:
- Dec 21, 2018 (the agreement): "assessment of the adequacy of the pre-approval study … will be based on both the size of the effect seen on UOx and the model relating UOx levels to stone formation rates … accelerated approval will not be based simply on demonstration of a statistically significant effect on UOx…"
- Jan 24, 2020: "We remind you of our December 21, 2018 Advice letter, in which we stated…" (verbatim repeat).
- Jul 6, 2020: "We remind you of our December 21, 2018 Advice letter and January 24, 2020 Written Responses, in which we stated…" (verbatim repeat, now citing both).
Takeaway: a citation to a prior letter means don't relitigate this — and that FDA is creating a paper trail. If you see your own past commitments quoted back to you, the door is shut.
4. The "at this time" hedge predicts a reopening
"At this time" and "we do not contemplate" reserve FDA's optionality. They are not waivers.
- 2006 Pre-IND: "At this time, we do not contemplate asking for genetic toxicology or carcinogenicity studies."
- 2015: the question is reopened — "The need for rodent carcinogenicity studies will depend on the results of the chronic 6-month toxicity studies in rats."
- 2016: finally resolved, only after clean histopathology — "… we agree that rodent carcinogenicity studies will not be needed."
Takeaway: when FDA hedges with "at this time," budget for the issue to come back.
5. An "Information Request" can be a clinical hold in disguise
FDA can constrain a trial without issuing a formal hold. In January 2020, an ALLN-346 "Clinical/Nonclinical Information Request" directed:
"Limit your proposed clinical protocol to a single ascending dose study. You can amend your currently proposed protocol to remove parts II and III." — and — "the current nonclinical data does not support clinical dosing of ALLN-346 beyond a single dose."
That is a partial clinical hold delivered as a routine information request. Separately, a 2021 CMC letter opened with "these requests are not clinical hold issues. However, response … is requested by April 12, 2021" — explicitly distinguishing a record-building request from a hold, which tells you the distinction is deliberate and worth tracking.
Takeaway: read what a letter does, not what it's titled. A request that removes protocol arms is a hold.
Three stories the dossiers tell in full
A. Years to "yes," and the yes had strings
The spine of the Reloxaliase program is a single question — can lowering urinary oxalate (a lab value) support approval without first proving fewer kidney stones? — answered "no" repeatedly across nine reviewers and three companies' worth of history, then "yes" with conditions:
- 2006: "We do not believe that small changes in urinary oxalate would serve … as surrogate for subpart H approval."
- 2015: "No, we do not agree … not sufficient to support … a surrogate endpoint." (door cracked: "we might be willing to accept a substantial change … greater than … your single arm phase 2 trial.")
- 2017 (EOP2): "We would not accept a small change in 24-hour urinary oxalate excretion as a basis for full or accelerated approval."
- 2018 (Type C): "we question whether changes in UOx of the magnitude expected would be reasonably likely to predict clinical benefit."
- Dec 2018 (the turn): "we agree with the overall strategy to obtain accelerated approval" — but only coupled to a UOx-to-stone model and a confirmatory trial, a standard then "reminded" repeatedly (see §3).
Lesson: a surrogate endpoint is a multi-year negotiation, not a checkbox. Even the eventual "yes" kept the burden of proof squarely on the sponsor.
B. Two versions of the same meeting
After the January 2018 Type C teleconference, two sets of minutes exist in the file. They read like different meetings:
- Allena's draft (Jan 30, 2018): "the Division agreed in principle with the program proposed by Allena … Demonstration of a trend toward fewer KS events in the ALLN-177 group in Study 302" would support Accelerated Approval.
- FDA's final (Feb 22, 2018): all three questions answered "No," and FDA instead re-raises the harder path — "we would like to revisit … a trial enriched for patients at high risk of kidney stone events … establishing effectiveness using a stone-based endpoint."
Lesson: a sponsor's meeting minutes are an advocacy document; the FDA's official minutes govern. Read both — and plan to the FDA's.
C. A designation denied on evidence, not need
The Breakthrough Therapy request (2020) split cleanly along the two statutory prongs:
"while we have determined that [the condition] meets the criteria for a serious or life-threatening disease … the preliminary clinical evidence you submitted does not indicate that the drug may demonstrate substantial improvement…"
FDA then dismantled the data: "oxalate is removed by dialysis, making it difficult to interpret plasma oxalate findings in a single-arm trial in dialysis patients"; only one of seven patients crossed the threshold; and "the findings are based on an exploratory, post hoc, subgroup analysis, raising questions about the reliability."
Lesson: designations turn on evidence quality, not unmet need. Controlled, prospective, pre-specified data clear the bar; a serious disease alone does not. (Contrast: ALLN-346's Fast Track request — tied to a clear clinical endpoint — was granted in 55 days.)
A note on judgment, not formulas
The same sponsor got opposite reproductive-tox answers for its two non-absorbed drugs: for Reloxaliase, "we agree that reproductive toxicology studies are not needed for a drug that is not absorbed" (2006); for ALLN-346, FDA required an embryo-fetal study anyway, because "due to the biological action of ALLN-346 … it may confer some indirect effects on a mother and/or fetus" (2021). Non-absorption was not a get-out-of-jail card — FDA reasoned about mechanism.
Patterns worth watching for
- Rank every FDA "yes": agree (settled) > conditionally agree (yes-if) > generally agree, determined at BLA (deferred) > do not object (your risk).
- Treat recommend / should / strongly encourage — especially next to "prior to Phase 3" or "for licensure" — as requirements.
- A citation to a prior letter means the position is closed and the record is being built.
- "At this time" forecasts a future re-examination.
- Read what a letter does; an information request can constrain a trial like a hold.
- Write your own meeting minutes precisely — but the agency's version is the one that counts.
Sourced entirely from the Reloxaliase (IND 100,169) and ALLN-346 (IND 143,480) dossiers, both from the same sponsor. Quotes are verbatim from FDA correspondence; the original letters — with FDA Reference IDs — are in each dossier's Module 1. Examples were selected because the file contained a clear downstream confirmation, so they illustrate the patterns rather than measure how often they hold. Nothing here is regulatory or legal advice. Create a free account to read the source letters and judge for yourself.