Figure: Ye olde FDA eCTD pyramid on vacation in paradise

Warning: Some facts and more opinions are found below. Use at your own risk!!

One day my manager came up told me and said, “Andrew, all this off-target work is going well but now we need to start writing research reports.” “What’s that?” I thought. I’ve heard of lab reports and school science (project) reports and journal articles but not a research report. Soon he delivered a nice summary of this mystery document type. I continued to piece together more details but wish I had a written summary to reference. When I search for information online I mostly found obtuse descriptions like the FDA’s Electronic Common Technical Document (eCTD) description which made my brain spin. But it has a cute pyramid that you can see in tropical paradise above! Oddly, I still can’t find this sort of summary so here’s my take which is, of course, just an opinion and comes with no warranties, guaranties of any sort.

The goal is to get that IND / CTA / NDA / BLA approved as soon as possible and not put on clinical hold or rejected. A lot of the struggle I see in organizations writing research reports stems from thinking of it like a journal publication. And the research report is like a journal publication except there is little or no research review in the introduction, the level of detail is often different, and there is no discussion. Let’s go through section by section highlighting how each differs from a journal article (if it does):

Abstract

This should summarize what the report describes and what it concludes.

Introduction (with minimal review of prior research)

Minimal background should be included here. If working with somewhat newer methods like long read sequencing a reference to a method or a paper is fine. Describing Next Generation Sequencing (NGS) or PCR, though, is generally not necessary.

Methods

Summarize the methods used. Though I don’t love the phraseology, this is common: “Briefly, the DNA was extracted X, then prepped with Y, and then sequenced via Z.” But this does get to the right level of abstraction and details will be ELNs as described below. (You’re not using a paper lab notebook, are you?)

Results

As you write, consider whether the result you are describing will have the corresponding ELN included. Agencies differ on whether ELNs are required, recommended, or should be available on request. Further, some sections may be deemed critical or contentious enough to need an accompanying ELN while others do not.

If the relevant result will have an ELN, a summary of the purpose and findings at a higher level of abstraction are sufficient since it’s easy for the agency to reference the ELN if necessary.
If ELNs for a corresponding to the experiment are not included in the filing but can be requested by the Agency, experiments are generally described in somewhat more detail than for EU filings.

Conclusion

Only state conclusions that are very well supported by the data. If some data is not significant but you think it is “trending”, don’t even think about writing this. This is a bad look in a journal and a much worse look for regulatory submissions. I generally suggest erring on the side of describing the data and guiding reviewers to the conclusion rather than trumpeting it.

Suggestions and tips:

In creating the report, your goal is to communicate that you have performed a well-designed, executed, and documented research study that is free of errors and comes to an appropriate conclusion about the efficacy, safety, or other aspects of your investigational drug or therapy. While in an academic publication it can be “fun” to speculate about mechanisms of action or connections to other research, this is not the place for fun. Sorry, not sorry! Ideally everything should point to the same conclusion. That may not always be the reality, but there’s no reason to make life harder and risk being put on clinical hold because you had a clever alternate idea about the data you wanted the world to know about. Bad scientist!
Use the same tense and voice (active or passive) throughout. Passive voice (no I or we statements is generally preferred.) An exception can be made for switching from past tense to present tense for work that is ongoing.
Write as simply as possible. Long sentences with long or multiple clauses should be broken up into simple sentences. If it’s spans more than 4 lines, break it up.
Use declarative sentences.
- BAD: When the threshold for inclusion in the dataset was set to a concentration of ABC that is tolerable to cells, the on-target editing frequency was observed to be 20%.
- GOOD: The threshold for inclusion in the analysis was a level that is tolerable to cells treated with ABC. The observed on-target editing frequency was 20%.
Use diagrams sparingly but do use them when they can illustrate process that would be hard to describe in words. Off-target gene editing pipelines are a good case since they often have multiple inputs being merged and aid the reviewer’s comprehension.
Make diagrams as simple as possible. Remove anything that is not strictly serving the central goal.
Reference earlier text rather than repeating it. That is, do not state your methods and then repeat the same method text in results - or as a subset of another more complex method. It is okay in later sections to greatly summarize the method in one sentence and reference the section number that describes it. This both shortens the document and prevents mistakes if one version gets changed and the other doesn’t. (It’s a similar to the coding practice Don’t Repeat Yourself where you shouldn’t making multiple copies of the same code since this can lead to bugs.)
Do not provide opinions. Write to your favorite newspaper or your favorite social media if you have opinions. Don’t write your opinions to the FDA, PEI, or MHRA. At best they’ll be insufficiently substantiated; and at worst the agency vehemently disagrees with you and has a poorer outlook on your submission.
At times it can be difficult to explain a method or a result. You must do your best to make it as clear as possible so that: a peer in the same field can understand it (e.g. the bioinformatics section should be understandable by a bioinformatician with no knowledge of the project) and ideally so that someone from a different field can understand it. If a capably / intelligent peer from your organization is left confused and unsure about something, STOP. Do not pass go. Re-write or re-diagram until that person can read it and understand. If they can’t understand it, an agency reviewer likely won’t understand it and you lose.
Make sure you are addressing the items that we know from experience regulatory bodies absolutely want in a submission. For example, gene editing / gene therapy with CRISPR / Cas9 or another nuclease (where official guidance from FDA is not yet available) some minimums for an off-target package are:
- computational homology search to nominate potential off-target sites based on similarity to the on-target sequence
- experimental search to nominate putative off-target cleavage sites
- sequencing to confirm each of the above nominated off-target sites
- other items, but this is long enough for an entire future post

That’s my current list that I’ve accumulated as I review submissions being prepared by sponsors. Hope this helps to provide some idea of how these reports are meant to be written.

Are there other key things you can think of? Please send me a message! Also get in touch if you’d like more detailed help with your research reports or IND submissions.