Information Requests and Deficiency Letters: Answering Well

During review, FDA asks questions. How you answer them is one of the few remaining levers on the outcome, and the range of formality is wide.

The forms it arrives in

Information Request (IR). The everyday one. A reviewer needs a clarification, a reanalysis, or a document you did not include. Often informal, often with a short response window measured in days or weeks.

Discipline Review Letter (DRL). Conveys deficiencies from one review discipline before the overall review is complete, so you can begin addressing them.

Complete Response Letter (CRL). Not a question. FDA's decision not to approve the application in its present form. That is a different document with different consequences, covered in the CRL entry.

For devices, the equivalent during a 510(k) or PMA review is an Additional Information (AI) request, and it stops the review clock until you respond.

The clock matters

For device submissions an AI request places the file on hold. The review clock stops, and you have a defined period, commonly 180 days, to respond before the submission is considered withdrawn. Elapsed time is yours, not FDA's.

For drug applications the interaction is different but the principle holds: a major amendment in response to a request can extend the goal date by three months. An answer that arrives as a major amendment has cost you a quarter, which is worth knowing before you decide how much new data to include.

Answering well

Answer the question asked. A response that reframes the question to one you would rather answer reads as evasion and produces a second request.

Answer completely the first time. A partial answer buys another round. Each round is elapsed time on a review clock and another reviewer context switch.

Do not volunteer new problems. Being forthcoming is right; attaching unrequested analyses that raise fresh questions is not the same thing, and it lengthens review.

Be consistent with the file. The answer has to agree with what is already in the application. Where it does not, say so explicitly and explain, because a reviewer noticing the discrepancy themselves is much worse.

Say when you disagree, with data. A reviewer's premise can be wrong, and saying so with evidence is legitimate. Silence reads as agreement.

What to do before answering

Read the question against your own file first. Frequently the answer is already in the submission and the real problem is that it was hard to find, in which case the response is a pointer plus a clearer presentation rather than new work.

Where the answer genuinely is not in the file, that gap tells you something about the submission that probably applies elsewhere.