Two pivotals, one enrollment gate: why RISE-3 can still read while RISE-2 waits

In Short. A new-enrollment-only partial hold splits Biohaven’s twin pivotal package in refractory focal epilepsy. Locked RISE-3 (NCT06309966) can keep dosing toward an H2 2026 seizure-frequency readout, while unfinished RISE-2 (NCT06132893) cannot recruit the patients its longer-window design still needs until the rodent-metabolite package clears.

---

Refractory focal epilepsy is a clinical problem with a blunt definition. Seizures begin in one region of the brain and keep coming after at least two suitable antiseizure medicines. For those adults, late-stage drug development still runs on seizure diaries. A patient spends weeks counting seizures before any study drug, then weeks on blinded drug or placebo, and the primary comparison is usually the change in average seizures per 28 days. That arithmetic sounds simple until a program needs two such studies at once, and a regulator freezes the door for new patients while leaving the people already inside on schedule.

Biohaven’s late-stage candidate for that population is opakalim, also known as BHV-7000, a selective opener of Kv7.2 and Kv7.3 potassium channels, given as an adjunct for adults with refractory focal onset epilepsy. The company built a pair of Phase 2/3 studies that share the same primary family, change in 28-day average seizure frequency from an observation phase, and differ in how the double-blind window is built. ClinicalTrials.gov lists them as RISE 2 (NCT06132893, company ID BHV7000-302) and RISE 3 (NCT06309966, BHV7000-303). An open-label extension (NCT06443463) invites completers of either parent study for longer dosing and safety follow-up. In a press release the pair looks like twins. In the protocols they are siblings with different geometries.

On 4 September 2026 the U.S. Food and Drug Administration issued a partial clinical hold letter on the BHV-7000 program. Biohaven reported the letter in an 8-K filed 10 September 2026 (accession 0001935979-26-000077, report date 2026-09-04). The Agency’s issue was nonclinical information. While characterizing metabolites of BHV-7000 seen in rodent testing, regulators judged that there was “insufficient information related to a specific metabolite to assess risks of that metabolite to human subjects at this time.” FDA instructed the company to pause new enrollment until clarifying nonclinical studies were done.

A partial clinical hold, under FDA’s clinical-hold framework in 21 CFR 312.42 and the Agency’s guidance on responses to holds, delays or suspends only part of the clinical work under an IND. Other parts may proceed. A complete hold freezes the IND’s clinical work more broadly. Biohaven’s 8-K states the restriction in plain English. “The partial clinical hold applies to new patient enrollment only.” Dosing continues for patients already randomized, “more than 600 patients.” The company also “voluntarily instituted this plan to its global sites outside of the U.S.” The enrollment gate is U.S.-ordered and globally mirrored by the sponsor. The human relevance of the rodent metabolite stays open on the company’s own words. The findings “may be specific to rodents and not relevant to human safety,” and the new nonclinical work is meant to answer that question.

That is the regulatory verb. The rest of the story is protocol geometry.

RISE-3 is the locked half of the pair. On ClinicalTrials.gov it is ACTIVE_NOT_RECRUITING. Design is parallel, randomized, and quadruple-masked, so patients, care providers, investigators, and outcomes assessors are all blinded. Estimated enrollment is 390. After an eight-week observation phase, the primary outcome compares that baseline to an eight-week double-blind treatment phase (Week 8 to Week 16) on change in 28-day average seizure frequency. The primary objective text on the registry names a 75 mg comparison to placebo, while the arm list also includes a 50 mg experimental arm. Estimated primary completion is October 2026. Biohaven’s 8-K says enrollment in BHV7000-303 completed in June 2026, dosing continues, and “topline data remain on track for the second half of 2026.” Because randomization finished before the hold letter, an enrollment-only gate leaves the observation and treatment windows already purchased still able to finish.

RISE-2 is the unfinished companion. The registry still lists it as RECRUITING, with status last verified in August 2026 and last update posted 24 August 2026, two weeks before the hold letter. Estimated enrollment is again 390. Design is sequential rather than parallel. Part A asks safety and tolerability questions over a double-blind window, counting deaths, serious adverse events, discontinuations, and laboratory abnormalities. Part B asks the efficacy question, comparing observation to a twelve-week double-blind phase (Week 8 to Week 20) on the same 28-day average seizure-frequency primary, across two doses versus placebo. Estimated primary completion is December 2026. A patient who enters Part B therefore spends a longer blinded efficacy stretch than a RISE-3 patient. The 8-K’s near-term implications list is explicit about what the hold does here. BHV7000-302 “will also continue dosing but new enrollment is paused pending nonclinical data.” Patients already inside keep their drug or placebo. Patients the study still needed cannot enter until the metabolite package satisfies FDA.

Read those facts as one machine with two chambers. Same disease, same drug, same primary measurement family, same estimated sample-size class. Different intervention models and different double-blind lengths. One chamber had already closed to new randomization. The other was still filling. An enrollment-only hold therefore leaves RISE-3’s seizure-frequency clock able to run on data already being collected, while RISE-2’s clock stretches by however long recruitment stays frozen. The delay lives in the unfinished design. Randomized patients keep dosing in both studies.

The open-label extension makes the same point from another angle. NCT06443463 is ENROLLING_BY_INVITATION for adults who completed the double-blind phase of either parent study. Its primaries are long-term safety counts over as much as 104 weeks. The 8-K says the OLE “continues dosing.” From a patient’s chair, a program under an enrollment-only hold can still look like medicine on schedule. The gate sits at the clinic door for new consenting subjects. It does not empty the pill bottle for people already assigned.

Registry literacy matters because ClinicalTrials.gov is often treated as a live dashboard. As of a fresh pull on 12 September 2026, RISE-2 still said RECRUITING, with an August update date. The 8-K says new enrollment is paused. Both statements can sit side by side for days or weeks when a sponsor has not yet submitted a status change. The check is simple. Read the hold letter’s enrollment language first, then ask whether the public registry has caught up. A stale RECRUITING flag after an enrollment pause is a paperwork lag. It is weak evidence that sites are still screening.

Company language and Street language about “two pivotals” should stay labeled separately. Inventing FDA’s eventual NDA answer would fake precision. Biohaven’s 8-K calls BHV7000-303 “one of two pivotal studies in refractory focal epilepsy,” and the January 2024 End-of-Phase 2 framing in the same filing describes “the first of two focal epilepsy trials.” That is the sponsor’s package design. BioPharma Dive’s hold coverage (10 September 2026) relayed RBC’s view that a RISE-2 delay may matter for an approval path that needs both studies. Those are expectations about a filing strategy. They are not a citation to an FDA letter that names a mandatory two-study rule for this IND. The honest operational sentence is narrower. When the commercial path is built as a twin package, RISE-2’s frozen enrollment is the binding constraint even while RISE-3 can still produce an H2 topline. A later regulatory conversation might put different weight on a single locked study. Which path FDA will accept is unknown from the September 4 letter.

What can be checked is how to read the next nonclinical submission. Biohaven says additional metabolite data are expected “in the coming weeks,” with the aim of clarifying human relevance. FDA’s hold-response process then gives the Agency a written reply clock once a complete response is in. When that package lands, three checks beat any “program on track” headline. First, does FDA lift the new-enrollment restriction, keep it, or convert it into a different partial limit. Second, does RISE-2’s estimated primary completion move once recruitment can restart, and by how many months relative to the December 2026 estimate still on the registry. Third, does ClinicalTrials.gov for NCT06132893 change from RECRUITING to a status that matches the hold or the lift. RISE-3’s H2 2026 claim can be judged on its own seizure-frequency readout when dosing and observation finish. RISE-2’s claim has to wait for an enrollment gate that the locked twin never faced.

The metabolite itself should stay where the 8-K leaves it. “Insufficient information” is an information gap. It is neither a demonstrated human toxicity finding nor a finished proof that the signal is rodent-only. More than 1,200 participants have received BHV-7000 in clinical studies to date, per the filing, and the company describes the clinical safety profile as generally safe and well tolerated so far. Those clinical counts leave FDA’s nonclinical question open. They also explain the shape of the hold FDA chose. The Agency could freeze new consent while leaving randomized patients on drug.

Put the pieces on one calendar. Observation and double-blind seizure counting only work if the right patients are already inside the study. RISE-3 filled that machine in June 2026 and kept it running under the September hold. RISE-2 still needed remaining slots when the hold arrived, and a twelve-week efficacy window remains longer than RISE-3’s eight-week window even after recruitment resumes. An enrollment-only partial hold therefore teaches a general clinic rule. Ask which protocols have finished randomization, which primary windows are measured in weeks of dosing already purchased, and which studies still depend on new consent. The verb “partial hold” names the legal instrument. The protocols name the clocks.

Focal epilepsy drug development still sells itself on seizure-frequency primaries. Sponsors still build twin Phase 2/3 packages that look alike in a headline and diverge in intervention model and double-blind length. Regulators still issue enrollment-only holds when nonclinical metabolite work is incomplete. A locked confirmatory clock can survive an enrollment gate. An unfinished companion cannot. The clinic rule is the same in any therapeutic area that runs paired late-stage studies under an enrollment-only hold. Ask who is already randomized, and who has not yet walked through the door.

---