BIVI Drops 50% on Bezisterim SUNRISE-PD Phase 2 Data

BioVie (BIVI) announced topline results from the Phase 2 SUNRISE-PD trial of bezisterim in early Parkinson’s disease on August 6. The stock crashed 50%, closing at $0.9993 against a $7.5 million market cap, down from roughly $2.00 before the readout. The market’s message was unambiguous: the data did not support the company’s framing.

SUNRISE-PD enrolled 57 patients with early-stage Parkinson’s who had never taken carbidopa/levodopa. Patients received 20 mg of bezisterim or placebo twice daily for 12 weeks. Bezisterim (NE3107) is an oral small molecule that crosses the blood-brain barrier and modulates NF-κB, ERK, and TNFα pathways to reduce neuroinflammation.

The study’s primary endpoint listed on ClinicalTrials.gov was change from baseline on MDS-UPDRS Part III, the standard motor exam. But the press release led with EPNIC-15, a composite of 15 motor and non-motor measures BioVie constructed itself, citing a 2025 paper on composite scales. On this composite, bezisterim showed a -0.04 change versus +0.18 for placebo (Cohen’s d = -0.94, p=0.0006).

The MDS-UPDRS Part III data, buried in a corporate presentation rather than the press release, told a different story. Placebo performed numerically better than bezisterim at lower baseline platelet counts. Bezisterim showed an advantage only in patients with higher platelet counts, roughly half the trial population, and the 95% confidence interval’s upper boundary remained above zero.

Every p-value in the release was “nominal and unadjusted for multiplicity,” meaning none were corrected for multiple comparisons. The disclaimer acknowledged the results are “exploratory” and that “the FDA has not validated these biomarkers as surrogate endpoints for Parkinson’s disease.” BioVie also filed to offer stock at $1.33 per share alongside the data release, adding to the selloff.

The biomarker data was directionally interesting. Bezisterim reduced a composite neuroinflammation measure (Cohen’s d = -1.06, p=0.0018) and lowered neurofilament light chain, a marker of neuronal damage (Cohen’s d = -0.746, p=0.008). In the high-platelet subgroup, bezisterim showed statistically significant advantages across all clinical measures. But biomarkers do not substitute for clinical endpoints, and the 57-patient, 12-week study was designed for signal-finding, not proof of disease modification.

Bezisterim’s safety profile was clean. Adverse events were reported in 39.3% of bezisterim patients versus 51.7% on placebo, with no severe or serious events and one treatment-related AE in each group. Patients on placebo had more total AEs and a higher proportion of moderate events.

The company hosts a conference call on August 12 to discuss Phase 3 design. Bezisterim is also in the ADDRESS-LC trial for Long COVID, with data expected late summer 2026. That readout is now the nearest catalyst for a company with a $7.5 million market cap.

For a framework on reading clinical trial press releases and spotting endpoint spin, see the guide to reading clinical trial press releases. For the disease-specific scoring scales behind Parkinson’s endpoints, see the guide to neuroscience clinical trial endpoints.

What It Means for the Trade

The 50% cut to a $7.5 million market cap is the market grading an endpoint maneuver in real time, and the grade is an F. When a company constructs its own 15-item composite (EPNIC-15), leads the press release with it, and buries the protocol-specified primary endpoint — MDS-UPDRS Part III, the standard motor exam — in a corporate deck, the read is not that the data is ambiguous; the read is that the company already knows what the standard endpoint says. That inference, not the p-values, is what destroyed the equity. The Part III story as disclosed — placebo numerically better in low-platelet patients, drug advantage confined to the high-platelet half with a confidence interval still crossing zero — is a subgroup-dependent, statistically unconvincing result for a drug that needs to eventually win a registrational trial on standard endpoints.

For anyone still trading BIVI, the honest framing is that the equity now trades near option value on two things: the ADDRESS-LC Long COVID readout expected late summer 2026, and whatever salvage the company executes from the SUNRISE-PD biomarker data. Both are worth examining on their own merits, but the $1.33 stock offering filed alongside the data tells you the balance sheet came first in management’s priorities — and dilution at that price level means existing holders are funding the optionality themselves.

Mechanism and Data Context

The scientific story underneath the spin is worth separating from the presentation. Bezisterim’s premise — oral, brain-penetrant modulation of NF-κB/ERK/TNFα neuroinflammation in Parkinson’s — is a legitimate hypothesis in a field that badly needs disease-modifying options, and the biomarker data was genuinely directionally supportive: a neuroinflammation composite improvement (Cohen’s d = -1.06, p=0.0018) and a neurofilament light chain reduction (d = -0.746, p=0.008) suggest the drug engages its target biology. But the trial’s own disclaimers do the analytical work: every p-value nominal and unadjusted, the FDA having validated none of these biomarkers as Parkinson’s surrogates, and the company itself labeling results exploratory. In a 57-patient, 12-week signal-finding study, an unadjusted p=0.0006 on a self-built composite is a hypothesis generator, not a efficacy claim.

The platelet stratification is the interesting scientific wrinkle. If neuroinflammation severity genuinely tracks with baseline platelet counts — platelets transport inflammatory mediators and interact with microglial activation — then the high-platelet subgroup advantage could be a real enrichment finding that shapes a Phase 3. That is the most generous read, and the August 12 conference call was presumably where the company tried to make it. But enrichment hypotheses built post hoc from half of a 57-patient trial carry an enormous replication risk, and the FDA has never accepted a biomarker-stratified Parkinson’s design without far stronger precedent.

Risks and What Could Go Wrong

1. A Phase 3 in Parkinson’s powered on standard motor endpoints is now a very hard fundraise at a $7.5 million market cap — the more likely path is the company never runs one, and SUNRISE-PD becomes the terminal dataset for the indication.
2. ADDRESS-LC in Long COVID carries the same composite-endpoint risk profile; if that readout also leans on unvalidated measures, the market will discount it identically.
3. The dilution spiral: successive offerings near $1 compress the shareholder base until the optionality belongs mostly to new money.
4. Delisting/nasdaq-compliance mechanics at a sub-$1 share price add timeline risk that has nothing to do with science.

Position read: this is a lottery ticket with a negative recent print on its biggest data event. The biomarker engagement data is real enough to keep the story alive into ADDRESS-LC, but nothing in SUNRISE-PD supports a Parkinson’s thesis at standard evidentiary bars. If you’re in it, you’re in it for the Long COVID readout — size it as such.

Source: BioVie SUNRISE-PD topline results press release, August 6, 2026 · SUNRISE-PD on ClinicalTrials.gov (NCT06757010)