Connections across cases

The most useful questions often sit between products, companies and decisions. How we build cross-case questions, and what five CRL histories show.

A question web reconstructs one company's history: a letter, a meeting, a resubmission, a second letter, an outcome. That is valuable on its own. But some of the most useful questions only appear when two histories are laid side by side. Did the FDA raise the same concern about two products? How did two companies respond to the same kind of letter? What usually happens after a Complete Response Letter, and what made the difference?

No company's record answers those questions. They exist only in the connections between records. This article explains how we approach them.

Three kinds of cross-case question#

Between products: Two therapies, related by indication, mechanism or regulatory pathway. How did the FDA evaluate the same endpoint in two products? Why was one single-arm trial considered sufficient and another not? What happened to the confirmatory-trial requirement after each accelerated approval? Did one product's history foreshadow another's?
Between companies: Two companies that need not be competitors. How did two companies respond to the same FDA concern? How did they describe the same kind of letter to investors, and how did those descriptions hold up? Did a later filing from one change how we read an earlier disclosure from the other?
Between events: One kind of regulatory event, traced across many cases. Instead of “What happened to Replimune?”, ask “What happens after a Complete Response Letter?” and answer it from every case in the record.

The third kind is the most powerful, because it turns individual histories into institutional knowledge.

The rule: every connection has an anchor#

Cross-case questions are easy to ask badly. “Company A versus Company B” invites a comparison of anything with anything, and the answer is opinion. So every cross-case question must name its anchor: a specific evidentiary connection that can be traced to primary records in each case. Anchors include:

The same FDA concern, in the FDA's words in each letter: “at least one additional adequate and well controlled study”, a facility named in an inspection.
The same procedural event: a Type A meeting, a Class 1 resubmission, formal dispute resolution, an advisory committee vote.
A comparable design element: a single-arm trial, an external control, a surrogate endpoint.
One record giving context to another: a later filing or decision that changes how an earlier one should be read.

This is the same rule that governs a single web, where two questions are joined only through a passage both answers cite (see Connections). Across cases, the anchor plays the role of the shared passage: if a cross-case question cannot point to its anchor in each record, it is not asked.

Worked example: what happens after a Complete Response Letter?#

Five companies have full question webs. Each received at least two Complete Response Letters for the same product. Read as one question, their histories look like this:

Figure 1. Five Complete Response Letter histories, read as one question. Each step is a dated event on the company's question web.

Company	What the letters turned on	After the last letter	Where the record ends
Aldeyra · reproxalap	Efficacy: a request for another adequate and well-controlled study; then the inconsistency of study results	Type A and Type D meetings; intent to file formal dispute resolution (Sep 2026)	Not approved; dispute pending
CorMedix · DefenCath	Manufacturing: conditions at the manufacturer; then the manufacturer, a heparin supplier and the product's name	Three manufacturing pathways; Type A meeting; resubmission	Approved Nov 2023, with the facility named in both letters
Outlook · ONS-5010	Efficacy: substantial evidence of effectiveness, after NORSE EIGHT missed its endpoint	Type A meeting; formal dispute resolution; appeal granted, as Outlook reports it	Class 1 resubmission, Jun 1, 2026
Replimune · RP1	Efficacy: the evidence from a single-arm trial	Path forward with the FDA; Class 1 resubmission; advisory committee 10–3	Accelerated approval, Aug 2026
Verrica · VP-102	The applicator's human factors; then a contract manufacturer's facility, inspected and reinspected	Type A meeting; a new bulk-solution manufacturer	Approved Jul 2023

What emerges#

Read one at a time, these are five stories. Read together, they raise questions none of them raises alone:

Two kinds of letter, two kinds of path. Both histories whose letters turned on manufacturing ended in approval, after two and three letters. The three that turned on efficacy each took a different route: an advisory committee and accelerated approval; formal dispute resolution and an appeal Outlook says was granted; and a dispute still pending. Does the kind of deficiency predict the kind of path out?
Concerns carry over. In every history, something in the first letter is still there in the next: the same request in the same words (Aldeyra), two of three deficiencies (Replimune), the same facility (CorMedix). Which concerns survive a resubmission, and what finally resolved them?
The same procedural steps recur. Every company met the FDA in a Type A meeting after a letter. Two filed Class 1 resubmissions in June 2026. Two turned to formal dispute resolution. When does a Type A meeting change the trajectory, and when does a dispute?
Companies frame letters in similar ways. CorMedix said its first letter raised no efficacy or safety deficiencies; Verrica's announcement said “No Clinical Safety or Efficacy Issues Identified”. Securities class actions followed letters to CorMedix, Replimune and Verrica. How did each company's first description of a letter hold up over the following year?

Published cross-investigations#

Each is grouped by the kind of connection it rests on (comparable evidence, divergent outcome, regulatory pathway, shared concern), and each company's question pages link to the cross-investigations that quote them.

More examples, with their anchors#

Question	Kind	Anchor in each record
When a company appeals a Complete Response Letter on efficacy, what evidence does the FDA's Office of New Drugs weigh?	Between companies	Outlook: the Office of New Drugs found NORSE TWO, with confirmatory evidence including NORSE EIGHT, established substantial evidence, as Outlook's 8-K reports it. Aldeyra: its intended appeal to the Office of New Drugs on the totality of evidence.
How did two companies describe a letter about a manufacturer to investors, and what did the next letter say?	Between companies	The 8-K after each letter, and the FDA's next letter on the same facility (CorMedix, Verrica).
What happened to the confirmatory-trial requirement after each accelerated approval?	Between products	RP1's accelerated approval on response rate and duration, and its Phase 3 trial; Aduhelm's accelerated approval on a surrogate endpoint, on the Alzheimer's record.
When the FDA says new evidence “does not alter” its conclusion, can the same evidence later support approval?	Between events	Outlook's 8-K on the December 2025 letter, and its 8-K on the May 2026 appeal decision about the same data; then the same pattern looked for in every other history.

The question map above the webs#

Each company web is a graph of questions joined by shared passages. Cross-case questions add a layer above them: a map whose nodes are kinds of event (a Complete Response Letter, a Type A meeting, a Class 1 resubmission, an advisory committee, an approval) and whose edges run down into every company web where that event occurs. A new company web does not just add one history; it adds a case to every cross-case question it touches.

This is why the method insists on anchors. A connection that can be traced to a passage can be reused indefinitely. A connection that cannot be traced has to be argued again every time. Built on anchors, cross-case questions become the institutional memory of the record: the part of the ephemeris of reasoning no single document, and no ordinary search, can supply.