The first thing I looked for in an NDA or BLA submission wasn't the p-value.
I know that surprises people. Sponsors spend months — sometimes years — designing a trial to get a p-value <0.05. They agonize over it, refine the design, change the analysis plan. And when they finally submit, they expect the FDA statistician on the other side to open the package and go straight to the efficacy results.
That's not what happened. At least not for me.
Before I ever looked at the efficacy results, I needed to answer one question: did the Sponsor actually do what they said they were going to do?
That meant starting with the documents.
First, I compared the Statistical Analysis Plan against the protocol. The SAP is where the statistical methods are specified in detail — the exact analyses, the populations, the handling of missing data. The protocol is where the trial design is committed to. Together, they tell you what the Sponsor intended. Any differences between the two need to be understood: were they legitimate refinements made before the data were seen, or something else?
Then I compared the statistical section of the submission against the SAP and its amendments. Did the analyses that were reported match what was pre-specified? You'd be surprised how often the answer is no — not dramatically, but in the small ways that accumulate. A slightly different patient population. A handling of missing data that wasn't quite what was described. A sensitivity analysis that quietly became the primary analysis somewhere between the plan and the results.
The cleanest submissions told a coherent story across all three documents. Deviations were disclosed and justified. There was nothing to hunt for — and those submissions moved faster because of it.
There's a third thing I looked for, and it's the one sponsors least expect: the secondary endpoints.
FDA reviewers know that primary endpoints are managed carefully. What the secondary endpoints look like — how they're reported, whether they're all there, whether the effect sizes are consistent with the primary — tells a more candid story about how a drug is actually performing.
A drug that works tends to show coherent signals across endpoints. When the primary just clears the bar, and the secondaries are scattered, or selectively reported, or notably absent from the summary tables, that's information. It doesn't mean the drug doesn't work. But it means the reviewer is going to look harder.
What I want Sponsors to take from this is a shift in mindset.
The goal of your submission isn't to present your data in the best possible light. The goal is to make it easy for a reviewer to understand what you did and why it supports your conclusion. Transparency isn't a concession — it's a strategy. The submissions that tried to manage the narrative were the ones that generated the most questions. The ones that laid everything out clearly, including their limitations, were the ones that built trust quickly.
FDA statisticians are not looking for reasons to reject your drug. They are looking for reasons to believe the evidence. Your job is to give them those reasons — not by minimizing what went wrong, but by showing that you understood it and accounted for it.
That's what the strongest submissions had in common. Not perfect data. Honest data, clearly presented.
If you're preparing a submission and want a second set of eyes before it goes in, I'm happy to talk through where you stand. Book a free 30-minute call.
Thank you for reading!
Lisa