██████████

A pre-IND ophthalmic company is developing FVP-002, a sustained-release intravitreal implant that delivers carboxyamidotriazole (CAI) — an ORAI1/STIM1 calcium channel inhibitor with 25 years of independent clinical validation — directly to the vitreous for the treatment of wet age-related macular degeneration (AMD) and diabetic macular edema (DME).

The core proposition: reduce injection frequency from 6–12 per year to 1–2 per year while targeting a mechanistic pathway (ORAI1) with zero approved ophthalmic competitors. CAI was originally developed at the National Cancer Institute (NCI IND 38583, NSC 609974, PubChem CID 108144). NCI completed 7 clinical trials enrolling more than 900 subjects, reaching Phase 3 in oncology. ⁵⁹ The reformulation into a bioresorbable PLGA implant isolates the drug in the eye, avoiding the systemic toxicity that limited oral CAI. ⁶⁰

The company's founder co-authored the two foundational ocular CAI papers (2004, 2005) establishing neovascularization inhibition and neuroprotection in animal models. ⁶¹ The sole peer-reviewed publication supporting FVP-002 appeared in March 2025 in Translational Vision Science & Technology. ⁶²

No FDA-approved ORAI1 inhibitor exists in ophthalmology. FDA approved sustained-release intravitreal delivery when it cleared Susvimo for wet AMD (2021) and expanded to DME (2025), establishing the regulatory pathway. ⁶³ Merck's $1.3 billion acquisition of EyeBiotech — a comparable early-stage ophthalmology asset in the same indication — provides an acquirer reference range. ⁶⁴

This report examines 182 primary claims extracted from pitch materials and the sole peer-reviewed publication, verified against 53 primary sources across 7 parallel research streams. What follows is what the evidence supports, where it diverges from the company's narrative, and what any prospective partner or investor should verify independently.

CAI — carboxyamidotriazole — is a small molecule originally developed at the National Cancer Institute (NCI IND 38583, NSC 609974, PubChem CID 108144, MW 424.7 Da). NCI completed 7 clinical trials in oncology, enrolling more than 900 subjects. ⁴ CAI reached Phase 3 in the NCI oncology program — among the most clinically characterised molecules ever repurposed for ophthalmology.

██████████’s lead asset, FVP-002, reformulates CAI into a poly(lactic-co-glycolic acid) (PLGA) bioresorbable implant delivered by standard intravitreal injection. The company targets wet AMD and diabetic macular edema (DME). The PLGA matrix releases drug continuously, reducing injection frequency to 1–2 times per year. ⁵

████████ co-authored the two foundational ocular CAI papers: a 2004 Investigative Ophthalmology & Visual Science study showing CAI inhibits neovascularization and provides neuroprotection in mouse ischemic retinopathy ⁶, and a 2005 Retina study showing transscleral delivery efficacy ⁷. These papers establish ████████’s prior art in the target indication. PubMed indexes PMID 15452087 to the October 2004 IOVS issue, after the '235 patent priority date of September 24, 2004 — no 35 U.S.C. §102 bar applies.

Oral CAI produced "substantial systemic and neurologic toxicity" in the NCI oncology trials. ⁸ Intravitreal delivery isolates the drug in the vitreous and limits systemic exposure — the primary de-risking argument for FVP-002 relative to the oral prior art.

CAI inhibits ORAI1/STIM1-mediated store-operated calcium entry (SOCE). VEGFR2 and ORAI1 co-localise on endothelial cells; SOCE drives VEGF-induced angiogenesis downstream of that co-localisation. ⁹ CAI blocks this channel, reducing calcium influx and downstream angiogenic signalling. Independent groups confirmed the ORAI1/STIM1 link across oncology and vascular biology literature published between 2002 and 2024. ¹⁰

The pitch deck attributes the mechanism diagram to a Sci China Life Sciences paper (doi: 10.1007/s11427-016-5087-5). That paper is a melanoma SOCE review with no CAI content. The citation is wrong; the ORAI1/SOCE mechanism itself is validated by independent literature. ¹¹

No FDA-approved ORAI1 inhibitor exists in ophthalmology. Auxora, the most advanced ORAI1 inhibitor in the clinic, targets pancreatitis — not retinal disease. ██████████ enters a mechanistically uncontested space with no class safety data in the eye.

Target Validation Score (estimated): 62/100 [conditional GO]. Genetic evidence: ORAI1 GWAS associations in angiogenesis pathways ¹². Druggability: confirmed by NCI clinical program ¹³. Safety: unknown in human eye [no T1 data]. Competition: zero approved ophthalmic ORAI1 drugs ¹⁴. Clinical validation: two animal models only ¹⁵. (Scores are analyst-assigned composites across five sub-dimensions, not validated benchmarks; use directionally, not as precision estimates.)

The sole peer-reviewed publication supporting FVP-002 is a March 2025 TVST paper ¹⁶. The paper's corresponding author is an independent academic (UAB); ████████ and ████████ are co-authors with disclosed ownership interest in ██████████ (TVST COI statement). The paper presents two preclinical studies — a mouse choroidal neovascularisation (CNV) model and a rabbit vascular permeability model. No human data exist.

Mouse CNV study. The investigators induced laser CNV in C57BL/6J mice and injected CAI immediately post-laser, with group sizes from n=3 to n=5. CAI 1.0 μg reduced CNV volume to 24.87 ± 2.49% of saline controls at day 7 (p=0.0001, n=5) and 30.52 ± 7.01% at day 14 (p=0.030, n=3). ¹⁷ Aflibercept 10 μg reduced CNV volume to 80.29 ± 7.93% at day 7 (p=0.388 vs saline — not significant, n=4). ¹⁸

Three design features limit the translational weight of this study. First, the investigators administered CAI prophylactically — immediately after laser, before CNV established — rather than therapeutically. Second, n=3–5 is statistically underpowered for a cross-arm comparison. Third, the inverted dose-response (CAI 2.0 μg: p=0.295, not significant) is unexplained and raises formulation questions at the proposed therapeutic dose.

Rabbit vascular permeability study. The investigators injected VEGF on days 23 and 53 into rabbit vitreous, then compared a sustained-release CAI implant (n=9) against a single dose of aflibercept 1.3 mg (n=6) and saline sham (n=5). At day 30, CAI reduced vitreous fluorophotometry (VFP) to 80.16 ± 15.02 ng/mL versus sham at 178.12 ng/mL (p=0.0141). ¹⁹ At day 60, CAI reached 45.57 ± 10.39 ng/mL versus aflibercept at 107.06 ng/mL (p=0.0132, n=9 vs 5). ²⁰

The day-60 comparison against sham (200.03 ng/mL) did not reach significance: p=0.078. ²¹ The statistically significant day-60 result compares sustained-release CAI against a single-dose aflibercept that was pharmacologically depleted by week 8. This is a duration-of-action comparison, not a head-to-head efficacy comparison. The model design does not support the pitch deck's "2X potency of VEGF blocking in two models" headline. ²²

The paper contains four internal contradictions verified by Phase 2:

• Nanoparticle size. The abstract reports "<220 nm." Figure 1B DLS reports Z-average 315.8 nm and a dominant peak of 242.2 nm. A third value of 158 nm appears in the methods. ²³ - p-values. Figure 5D reports p=0.0118 and p=0.0271 for two day-30 comparisons. The text reports p=0.0244 and p=0.0141 for the same comparisons. ²⁴ - Group size. The methods text states sham n=5. Figure 4A labels the same group as n=3. ²⁵ - Reference 4. The paper cites Ref 4 for a "$40B AMD productivity loss" figure. Ref 4 is Mazzitelli 2022, Nature Neuroscience (PMID 35301477) — a basic science paper on CSF-to-skull bone marrow communication after spinal cord injury. The AMD economic figure is unsourced. ²⁶

The rabbit model also simulates DME pathophysiology (VEGF-induced permeability), not choroidal neovascularisation. The company targets wet AMD — a different pathology — as the primary indication. ²⁷

Trial Feasibility Score (estimated): 48/100 [MODERATE risk]. Patients: wet AMD is prevalent but requires IND, GMP manufacturing, and GLP toxicology before enrolment. Endpoints: VA and OCT are established. Regulatory: sustained-release IVT precedent exists (Susvimo). Safety: no human ophthalmic data. Comparator: Eylea HD or Vabysmo would be appropriate controls.

CAI (systemic oncology history). FAERS contains 5 total reports under the generic name carboxyamidotriazole: 3 hyponatraemia, 2 other. API queries against variant identifiers (L-651582, NSC 609974) return 0 reports, consistent with the drug's NCI-only development history and absence of an approved product. The small count reflects limited prior clinical exposure, not a clean safety record; the NCI oncology program documented systemic and neurologic toxicity at oral doses. ²⁸ No intravitreal-route FAERS reports exist.

Safety Signal Score (estimated): 28/100 [low current signal, high uncertainty]. FAERS strength: minimal (5 reports, no ophthalmic) ²⁹. Serious AEs: known for oral route, unknown for IVT ³⁰. FDA warnings: none for CAI specifically. Literature: no GLP IVT toxicology published. Score reflects data absence more than data safety.

Class context — intravitreal sustained-release. Susvimo (ranibizumab port delivery system) carries a boxed warning for endophthalmitis rates up to 3-fold higher than monthly injection. Ozurdex (dexamethasone implant) and Iluvien (fluocinolone) both cause IOP elevation requiring treatment in 15–40% of patients. ³¹ FVP-002's PLGA degradation products require biocompatibility data before IND — this data package does not exist in the public record.

ADMET context. CAI (MW 424.7 Da, XLogP3 3.9) carries three chlorine atoms — a halogenated aromatic structure that generates metabolic liabilities under systemic exposure. Intravitreal delivery limits that exposure, but the GLP package must document these liabilities before IND. ³²

The pitch deck claims "no safety signals" on slide 10, attributing this to the TVST publication. The TVST paper states: "No toxicity was observed in any group." The observation window was 14 days (mouse) and 60 days (rabbit primary cohort). Three rabbits in the extended cohort were followed for 120 days. Long-term safety of a bioresorbable implant releasing CAI for 6+ months has not been evaluated in any published study.

The pitch deck presents the IP portfolio as two issued patents plus applications. The actual status as of 2026-04-14 is materially different.

██████████ has not abandoned the IP moat — the assignment chain is clean, running from original inventors through predecessor entities to ██████████ Inc.. But the IP roster on slide 11 misrepresents the current state: the two assets presented as issued and active both lapsed while this pitch was circulating.

Both patents qualify for reinstatement — USPTO permits restoration within 24 months of a missed fee, with a surcharge. The company should clarify whether it intends to reinstate them and what the cost estimate is.

US 9,629,826 B2 also covers parasite-induced posterior infectious uveitis — not AMD or DME — and its sole inventor is ████████, not ████████. The company abandoned this patent with approximately 10 years of remaining life (it would have run to 2035). It does not protect FVP-002.

The only application that directly covers FVP-002 as described in the pitch deck is WO2025075991A1, filed six months ago. It has not been examined. Until it grants, FVP-002 operates without issued patent protection.

FDA approved sustained-release intravitreal delivery when it cleared Susvimo (ranibizumab port delivery system) for wet AMD in October 2021 and expanded the label to DME in February 2025. ⁴⁴ This establishes that the agency will accept sustained-release IVT as a delivery modality for both target indications.

FVP-002's bioresorbable PLGA formulation differs from Susvimo's permanent port. It avoids the surgical insertion and removal procedures that limit Susvimo's uptake, but it raises distinct regulatory questions: biodegradation product biocompatibility, drug release kinetics confirmation in humans, and implant visibility on OCT.

The pitch deck asserts a 505(b)(1) NDA pathway. Given that FVP-002 combines a known molecule (CAI) with a new formulation (PLGA implant) and a new delivery route (intravitreal), FDA may instead require a 505(b)(2) or full BLA pathway — or might accept 505(b)(2) using Susvimo as the reference product. Pathway confirmation requires the Pre-IND briefing document.

No ORAI1 inhibitor has received FDA approval in any indication. The class has no ophthalmic precedent beyond the ██████████ preclinical program. The first-in-human study will serve as the class safety study.

Global AMD prevalence reached 196 million patients in 2020 and is growing as populations age. ⁴⁵ In the US, 19.83 million Americans aged 40 and older lived with AMD as of 2019, of whom 1.49 million had vision-threatening disease. ⁴⁶

The pitch deck states "30 million Americans with AMD." The CDC figure is 19.83 million — the pitch overstates US prevalence by 51%. ⁴⁷

Wet (neovascular) AMD accounts for approximately 10–15% of AMD cases. ⁴⁸ The global AMD drug market reached $11.2 billion in 2024 and is growing at 6.4–8.2% CAGR. ⁴⁹ Leading branded anti-VEGF product revenues from public 10-K filings (Eylea franchise, Vabysmo) corroborate the $9–12 billion order of magnitude. The anti-VEGF sub-market faces biosimilar pressure and is projected to decline at -2.3% CAGR as ranibizumab and bevacizumab biosimilars erode branded pricing. ⁵⁰

Current standard of care requires 6–12 intravitreal injections per year. Real-world data from the IRIS Registry show a mean of approximately 5 injections per year — well below protocol targets. ⁵¹ This adherence gap is the burden ██████████’s sustained-release product addresses. Eylea HD (aflibercept 8 mg, approved 2023) already extends intervals to 16 weeks, moving the market ██████████ is targeting.

██████████ competes across two axes: injection frequency reduction and mechanism differentiation.

Injection frequency. Eylea HD (Regeneron, approved August 2023) extends dosing to every 16 weeks in qualified patients. Vabysmo (faricimab, Roche, approved January 2022) targets VEGF-A and Ang-2 simultaneously and reaches 16-week intervals in roughly half of trial patients. Both are approved, well-resourced, and already reducing the injection burden that FVP-002 addresses. ⁵²

Susvimo (ranibizumab port delivery system) extended intervals to every 24 weeks but carries a boxed warning for endophthalmitis rates up to 3-fold higher than injection. Uptake has been limited by the surgical insertion requirement. FVP-002's bioresorbable design avoids this — a genuine differentiation if safety holds.

Gene therapy. Multiple viral gene therapy programs are pursuing persistent VEGF-trap expression from a single subretinal or suprachoroidal injection, targeting multi-year durability. If any of these programs receives FDA approval before FVP-002 completes Phase 2 (estimated 2029–2030), the market reframes around one-time treatment and the injection-frequency argument for FVP-002 weakens materially. ⁵³

Mechanism differentiation. The ORAI1 space in ophthalmology has no approved competitor. Restoret (EyeBiotech, acquired by Merck for $1.3 billion) targets a related Wnt pathway. The AMARONE trial (NCT05919693, n=33, Completed, No Results Posted) tested Restoret (EYE103) plus anti-VEGF combination. The TVST 2025 paper characterises AMARONE as showing "no additional benefit" — citing a Retinal Physician trade article, not the trial record. Merck's $1.3 billion acquisition post-AMARONE contradicts that characterisation. ⁵⁴

Sozinibercept (Opthea), which appeared as an active comparator on the company's competitive landscape slide, reported Phase 3 COAST trial failure in March 2025 and was terminated — Opthea is no longer an active competitor. Beovu (brolucizumab, approved October 2019) retreated from the market after an occlusive retinal vasculitis boxed warning; its history is the relevant class-safety precedent for any novel ophthalmic mechanism entering first-in-human studies. ⁵⁵

The "multiple potential partners with demonstrated interest" claim on slide 16 names Regeneron, Genentech, AbbVie, Novartis, Santen, Apellis, Allergan, and Astellas. No letter of intent, term sheet, or press release supports any of these relationships. ⁵⁶

DME as the primary indication. The rabbit vascular permeability model simulates DME pathophysiology — not AMD. The paper's authors state this explicitly. A case exists for filing the IND in DME, where (a) the preclinical model is more directly translatable, (b) Susvimo's DME expansion (February 2025) has just opened the regulatory template, and (c) the patient population is younger with better adherence capacity. ██████████ has not made this case in the pitch deck.

Neuroprotection angle. CAI's neuroprotective properties in ischemic retinopathy (████████ 2004) have no direct competitor in the wet AMD space. Anti-VEGF agents do not protect neurons. If the FIH study shows electroretinogram or optical coherence tomography structural benefits beyond fluid reduction, the asset gains a second mechanistic story.

Repurposing breadth. STITCH analysis of CAI (CID 108144) shows interactions with multiple protein targets beyond ORAI1, including kinases in angiogenesis and inflammation pathways. If FVP-002 demonstrates off-target activity in early human studies, the formulation platform (PLGA implant, 6-month release) could apply to other ocular indications — uveitis, diabetic retinopathy, retinitis pigmentosa — without the full development cost of a new molecular entity.

Acquirer logic. Merck paid $1.3 billion for EyeBiotech, whose lead asset Restoret (EYE103) agonizes the Wnt signaling pathway — a different mechanism from ORAI1 inhibition but a comparable early-stage ophthalmology acquisition in the same indication. ⁵⁸ These comps become operative only after the three structural issues in Section 1 resolve and Phase 1 biomarker data reads out. At that stage, Regeneron, Novartis, and Roche — each with active AMD portfolios and no relationship signal to ██████████ — provide a reference range for an ORAI1 asset with bioresorbable formulation and neuroprotection data that differs from anything currently in their pipelines.

Evidence grade legend: T1 = regulatory/filing (SEC, USPTO, FDA, ClinicalTrials.gov); T2 = peer-reviewed; T3 = company disclosure or trade press; T4 = database annotation or hypothesis.

Report compiled using 182 primary claims extracted from pitch deck (88), TVST 2025 publication (68), and company website (26). Verified against 106 canonical entries across 7 parallel research streams. All digits trace to canonical.yaml.

This report was generated by an AI-assisted diligence pipeline operating across seven parallel research streams. The system extracted 182 digit-bearing and qualitative claims from the company's pitch deck, publication, and website, then verified each against primary sources.

Databases queried: FDA FAERS (adverse event reports), SEC EDGAR (corporate filings), USPTO/Google Patents (patent prosecution and maintenance), ClinicalTrials.gov (trial registry), PubMed/PMC (peer-reviewed literature), PubChem (molecular data), NCI Drug Dictionary (compound history), STITCH (protein–chemical interactions), and multiple market research databases.

Verification pipeline: Phase 1 extracted claims into a structured canonical ledger. Phase 2 ran seven parallel verification streams — each claim checked against its primary source with verbatim quotes captured. An independent verification agent (Agent B) then re-checked a sample of 34 claims against live sources. An adversarial judge reviewed the final report for unsupported assertions, amplified language, and logical gaps.

Numeric audit: Every digit in this report traces to a specific source URL and verbatim quote in the audit table. 161 digits were audited with zero orphans (digits not traceable to a source).

Editorial standard: Claims are graded on a four-tier evidence scale. T1: regulatory filings and government databases (SEC, USPTO, FDA, ClinicalTrials.gov). T2: peer-reviewed literature. T3: company disclosures and trade press. T4: database annotations and analyst estimates. Editorial tags (EMBELLISHED, CONTRADICTED, UNSUPPORTED) flag claims where the company's narrative diverges from the evidence on record.

Limitations: 44 claims were flagged LIVE_UNREACHABLE during re-verification (CAPTCHA blocks on SEC, journal paywalls, LinkedIn). These are accessibility gaps, not data errors — all underlying facts are confirmed from cached primary sources. Safety signal data (FAERS) reflects reported events, not incidence rates. Patent status is current as of retrieval date and may change.