The Verteporfin Hub
A structured synthesis of a Telegram community building a folk protocol for off-label intradermal verteporfin, paired with an independent literature audit. Case studies expandable with chat quotes and inline photos.
What this is
The defensible read in five paragraphs.
Verteporfin (VP), an FDA-approved photodynamic agent for wet macular degeneration, has a separately-studied light-independent ability to inhibit YAP/TAZ signalling. In mouse and pig wound models from the Longaker / Mascharak group (Stanford), this produces healing with reduced scarring. Clinical translation is unproven in any peer-reviewed RCT as of April 2026; SCARFREE-001 (NCT07488988, Odense) is the first powered human trial — enrolling, not yet read out.
The Verteporfin Hub is a public Telegram supergroup of ~22 active posters who, in the August 2024 archive used here, drafted and refined a folk protocol for self-administering VP at the time of surgical wounding. Their working consensus settled at 0.8 mg/cm² in a 2 mg/mL solution, intraoperative, intradermal via 4mm 30g mesotherapy needle. This number propagates through their planning conversations and members' surgical plans.
The chat is doing community-driven off-label drug exploration with all the strengths and weaknesses that entails. Most "evidence" they work from is photo interpretation and patient self-report — which is real data, just not blinded biopsy-validated data. Surgeon refusals (when they happen) are common risk-aversion from liability and conservatism, not strong signal that VP doesn't work. n=1 mechanistic evidence is fine for "this is biologically interesting"; it just isn't enough for a clinical recommendation.
The chat does have specific factual problems worth knowing about. Its de-facto formulation (raw VP powder + sterile PBS + sonicator) treats Visudyne and raw VP as equivalent — but raw VP is a hydrophobic benzoporphyrin and won't dissolve in water. Lucy & Sage (external to the chat) figured this out independently and developed a soy-lecithin liposome alternative. The chat never adopted it. The "0.18 mg/cm² pig dose" the chat uses for translation calculations is not findable in any peer-reviewed source; targeted searching of the 2025 STM pig paper, the 2021 mouse paper, the 2022 Cell Stem Cell paper, the PSRC 2024 abstract, and the C-DOCTOR proposal all came up empty. The closest published mg/cm² value in the Mascharak corpus is 0.106 mg/cm² in mice; the actual pig per-cm² figure lives in supplementary methods that are paywall-gated. Separately: 4 of 5 of Bloxham's YouTube update videos are now Private, with no Wayback snapshots — the chat's primary visual anchor for the 0.8 mg/cm² consensus is partially retracted at the source, though Bloxham himself confirms the dose range in a Hair Restoration Network forum post: "Most seem to be falling into the 0.8 mg/cm² – 1 mg/cm² range."
This synthesis exists in three layers. Case studies reconstruct individual VP experiments with chat quotes and inline photos — click any card to expand. What's weak is the harsh-but-fair read of what the evidence supports and what it doesn't. Corrections documents the bugs I found in earlier drafts of this synthesis (sub-agents fabricated content I had to fix in multiple passes). Read in roughly that order.
Things that are actually weak
Pushed harder than the cheerful read; less hard than v1 of this section, which was unfair to the chat members.
The chat's protocol has a real pharmacology error
The chat's working formulation is raw VP powder + sterile PBS + ~$50 amazon sonicator + 4mm 30g mesotherapy needle. This protocol almost certainly produces a suspension, not a solution: VP is a hydrophobic benzoporphyrin derivative, and sonication into water-based PBS does not solubilize it. The chat-internal claim that drove this — "Both [Visudyne and raw VP] are same. Visudyne is made by Bauch and Laumb [Bausch & Lomb], but they ran out of patent. The other is for research purpose" [#424] — is mechanistically wrong. Visudyne is liposomal-encapsulated for water reconstitutability; raw powder needs lipid encapsulation or organic solvent to actually dissolve. Lucy & Sage (external) figured out a soy-lecithin liposome alternative; the chat never adopted it.
Some load-bearing numerical claims aren't peer-reviewed
The "0.18 mg/cm² pig dose" the chat uses to translate Stanford's pig data into human dose calculations isn't findable in any peer-reviewed paper. A targeted hunt across the 2025 STM pig paper, the 2021 Science mouse paper, the 2022 Cell Stem Cell paper, the PSRC 2024 abstract, the C-DOCTOR cleft-lip proposal, and ClinicalTrials.gov NCT06944249 found no source. The closest published mg/cm² value in the Mascharak corpus is 0.106 mg/cm² (mouse: 30 µL × 1 mg/mL into 6 mm wounds, per the 2022 STAR Methods). The pig paper specifies only concentration (2 mg/mL was best vs 1, 4, 8 tested) and per-animal cumulative totals; per-site volumes are in supplementary methods that are paywall-gated. The 0.18 figure may be someone's back-of-envelope from a paywalled supplement, or a different derivation no public source uses.
Bloxham's three-patient dose-escalation results (the source of the chat's 0.8 mg/cm² anchor) live in his own YouTube videos plus a hair-loss commercial blog. 4 of his 5 Update videos are now Private on YouTube, and Wayback has no snapshot — the original visual evidence is partially retracted at the source. The 0.4 and 0.8 endpoints are confirmed via Follicle Thought's January 2024 summary; the 0.6 middle dose is plausible-by-interpolation but only "medium" is publicly named. Bloxham's own statement on his current practice, from a Hair Restoration Network forum post:
"Most seem to be falling into the 0.8 mg/cm² – 1 mg/cm² range."
Bargouthi's published doses on his own website (verteporfin.org) cap at 0.4 mg/cm²; the chat's "0.6/0.8/1.0 mg/cm²" claims for him [#387] are a chat-internal relay (with Enrico's "seemingly" hedge) that the public record doesn't confirm.
The "underdose pattern" is suggestive, not established
Across n=4 cases (K K, Sage, Lucy, Sc0t), the cleanest reading is "low-dose attempts produce mixed/null outcomes; the higher-dose case worked." That's consistent with Bloxham's published n=3. It's a reasonable working hypothesis for the chat. It's not a clinical recommendation — n=4 with formulation, timing, and administrator confounders doesn't demonstrate dose-response cleanly. Treating it as evidence in the strong sense is post-hoc; treating it as a working hypothesis is fine.
Members are extrapolating beyond the documented dose ceiling
Barfy Jones is planning hairline-advancement at 1.0–1.2 mg/cm², above any documented dose. Barfy's rationale (more tissue tension on the scalp warrants higher dose) is biologically reasonable but the safety data isn't there yet. S S found a third willing surgeon after two declined his facial excision request. Surgeon refusals are common risk-aversion, not evidence that VP fails — but they do mean S S's plan is going forward with a less-experienced operator than the original two. Fabio is planning self-injection because his surgeon refused intraoperative VP; both George Kal and Enrico pushed back hard against this in the chat [#684-#697]. Self-injection of an undissolved porphyrin suspension is a meaningfully riskier proposition than the rest.
K K's "scarless" outcome is real photo data, but ambiguous
Without histology, "scarless texture but depigmented" can be true regeneration, atrophic scar with melanocyte loss, hypertrophic scar that flattened with time, or some mixture. K K's case has good photographic documentation (see the K K case study below) — the photos show something and the chat's depigmentation observation is novel data. Whether it's "scarless regeneration" specifically requires histology no one has done.
My v1+v2 outputs had specific fabrications too
Sub-agents I used to research the literature produced confident-sounding outputs that, on audit, contained: a fabricated "VP in breast milk up to 66% of plasma levels" figure (no primary source); a fabricated three-arm "RFMN 24% vs ablative 75% vs isotretinoin 67% relapse at 3 years" comparison (cited paper has 1-month follow-up); a misattribution of a key picosecond-fractional study from "vs CO₂" to "vs Er:YAG" (the actual comparator); a "Level I" framing of the embrace device's RCTs that the trials' own classification calls Level II, with manufacturer-equity authors. These propagated into earlier drafts. They are corrected in current outputs but the fact that they appeared at all should temper trust in any AI-generated literature synthesis. Full corrections log →
Case studies
Click any card to expand. Each case is reconstructed with direct chat quotes, inline photos and videos where available, and a short critique. The chat is public; quotes reproduced with message IDs.
Who's in the chat
22 active posters. Most are processing scar-related distress; a handful have completed or planned VP experiments. Counts use original-author attribution (not forwarder).
Named doctors
Where the chat's claims about each clinician hold up against independent verification.
Cross-cut by theme
The chat's discussion organized by what it's about.
How the consensus formed
The 0.4 → 0.8 mg/cm² pivot happened in 24 hours on August 12 2024, on one Bloxham YouTube video, with no pushback in the chat.
Corrections log
The synthesis went through three audit passes. Each pass found bugs in the previous one.
How this was made
The synthesis was built across multiple passes (April 24–27 2026). All work was done by Claude (Opus 4.7 + Sonnet sub-agents). No clinical input. The Telegram archive is public; quotes are reproduced with message IDs. Inline photos and videos are direct exports from the chat.
- Phase A.1 — per-poster extraction. Six Sonnet subagents in parallel, one profile per heavy contributor.
- Phase A.2 — cross-cut by theme. One Sonnet subagent producing 13 topic files.
- Phase B — independent literature. Four Sonnet subagents researching VP and non-VP scar techniques.
- Phase C — corrections and audits. Multiple Opus verification passes found systematic errors: fabricated chat claims; the Telegram-export forwarded-message attribution bug; misattributed papers; fabricated numerical claims in the literature files. See
CORRECTIONS-V2.md. - Phase D — webapp. Single self-contained HTML file with inline JSON for case-study quotes and photo manifest.
What this synthesis cannot tell you
- Whether VP works for any specific scar morphology in any specific person.
- What dose, formulation, or timing is right for a planned procedure.
- What happened after August 27 2024 (this archive's cutoff).
How to read AI-generated medical synthesis honestly
- For any quantitative claim that would change a decision, fetch the primary source yourself.
- "Confirmed" tags from one AI agent are not reliable when another agent cited a URL but didn't fetch it.
- "Peer-reviewed" is a starting point, not a conclusion — check the journal's review process and indexing.
- Industry-funded clinical trials are not the same as independent confirmation.
- Photo interpretation is real data — just not biopsy.