# Critique — what's actually weak in this whole project
This file pushes hard on what's weak. Earlier versions of this file pushed
*too* hard on the chat members specifically — calling surgeon refusals
"signal not friction," calling the n=4 underdose pattern "post-hoc fitting,"
treating "near-100% satisfaction" reports as if they were nothing.
The accurate read: **the chat is operating in the early-stage off-label-drug
reality that all such communities operate in.** Surgeon refusals are
common risk-aversion, not a strong signal that VP doesn't work. n=1
mechanistic evidence is fine for "this is biologically interesting" — it
just isn't fine for "this is a clinical recommendation." Photo evidence is
real data, just not blinded biopsy-validated data.
What's actually weak (this file's scope):
1. The chat's protocol foundations involve at least one clear pharmacology
error (treating Visudyne and raw VP powder as equivalent — the powder
is hydrophobic and won't dissolve in PBS).
2. Some load-bearing numerical claims trace back to non-peer-reviewed
sources (the "0.18 mg/cm² pig dose," the Bloxham doses sourced from a
hair-loss commercial blog).
3. My v1+v2 outputs propagated multiple AI-generated fabrications I have
since fixed but should be transparent about.
4. Several emerging-tech claims I cited as "better-evidenced alternatives"
were themselves industry-conflicted or outright misattributed by the
first-pass research agents.
If you came here looking for a knockdown argument that the chat's project
is misguided, this is the wrong file. The right read is more nuanced:
the chat is doing community-driven off-label drug exploration with all
the strengths and weaknesses that entails.
---
## 1. The chat's epistemics are weaker than the chat thinks
### 1a. The protocol's foundation is anonymous and non-expert
- Message [#43] — the chat's foundational protocol — was authored by **N G**,
a poster with **2 messages total** in the entire archive. We have no
evidence of pharmacological training or clinical credentials.
- Message [#186] — the canonical dose-comparison table — was authored by
**IgnotuS**, a poster with **1 message total**. The table includes a
"0.18 mg/cm² pig dose" that **doesn't appear in any peer-reviewed source
the literature audit could find.** That number is the foundation of every
pig-to-human dose calculation the chat makes downstream.
- Message [#424] — "Visudyne and raw VP powder are both same, just one's
for research" — was authored by **martha Halli**, a poster with 11 messages
total, no stated credentials. **This claim is mechanistically wrong**
(Visudyne is liposomal-encapsulated for water reconstitutability; raw
benzoporphyrin powder is hydrophobic and won't dissolve in PBS). Lucy &
Sage figured this out independently and the chat never adopted their
finding. The chat treated martha Halli's claim as settled.
The chat treats Enrico's curation as quality assurance ("if Enrico promoted
it into the Protocol topic, it must be vetted"). It isn't — Enrico is a
group admin, not a pharmacist or clinician. He has a personal scar and
high motivation, which is not the same as expertise.
### 1b. Most "evidence" is photo interpretation — strong enough to motivate, not strong enough to prescribe
The chat's most-cited results:
- **Bloxham n=3** — three patients, three doses, photos in his own YouTube
videos. No biopsy, no objective scar scoring, no blinding, no
independent observer. **n=1 is reasonable as mechanistic proof-of-concept**;
it just doesn't justify clinical recommendation. Bloxham's videos are
worth watching for what they are: a hair-transplant surgeon's case-series
testimonial of an off-label drug, with publicly visible photos.
- **K K** — outcome photos posted to the chat by Enrico (the curator).
K K himself speaks once in the entire archive ("Ok" [#3431]) — but his
case is documented through detailed photos and follow-up interpretation
by the chat. Without histology we can't formally distinguish "scarless
regeneration" from "atrophic scar with melanocyte loss," but the photos
are at least real visual data.
- **Sc0t** — patient self-report ("100% satisfied"), relayed by Enrico,
with photos in the Sc0t archive topic. Patient self-report isn't
validated outcome scoring, but it's also a real data point.
- **Bargouthi** — n=1 biopsy at 0.32 mg/cm² showing ~2× follicle count vs
control with reduced perifollicular fibrosis. This is the strongest
single data point in the entire human VP evidence base. It is still
n=1, on FUE scalp, on the surgeon's own website.
There is no peer-reviewed RCT of off-label VP for scar revision in humans
yet. **SCARFREE-001** (NCT07488988, Odense) is the first powered trial
— testing 0.5/1.0/2.0 mg/mL concentration arms intra-patient on
abdominoplasty wounds, primary endpoint POSAS at 3 months. When it reads
out, we'll have a sponsor-independent intra-patient comparison. Until then,
the chat is working with case-series testimonial — which is the early
stage of any off-label drug's clinical reality, not a unique chat failure.
### 1c. The "underdose pattern" is suggestive, not established
I framed it too aggressively earlier. The right framing: with n=4 cases,
the cleanest pattern is interesting but underpowered.
| Case | Dose | Other variables |
|------|------|----------------|
| K K | 0.4 | unknown formulation, possibly suspension |
| Sage | 0.4 | lecithin liposome, 48h post-op, n=1 self-report |
| Lucy | unclear | undissolved suspension (formulation problem, not dose) |
| Sc0t | unclear (NOT 0.8) — chat never states his actual dose | doctor-administered with experienced surgeon |
The "0.4 doesn't work, 0.8 does" story is a hypothesis the chat extracted
from the data they have. It's consistent with Bloxham's published n=3
dose escalation. It's not established at n=4 with confounders,
but treating it as "post-hoc pattern-matching" was too dismissive — n=1
is fine as initial mechanistic evidence, and the underdose hypothesis
is a reasonable working model. Just not a clinical recommendation.
### 1d. K K's "scarless" outcome may not be what the chat thinks
Enrico interprets K K's photos as showing:
- Scarless texture (subjective)
- Depigmentation (objective from photos)
- Hair follicle regeneration (Enrico's interpretation; could be retained
hairs in the scar field, residual follicular structures, or photo
artifact at scale)
- No keloid (relevant only because K K is keloid-prone — but a flat
atrophic depigmented scar is *also* not a keloid)
Without histology, "scarless texture but depigmented" could plausibly be:
- True regeneration (dermal architecture restored)
- **Atrophic scar with melanocyte loss** — looks "thinner / smoother" because
it IS deficient
- Hypertrophic scar that flattened with time (matrix remodeling)
- A combination
The chat's interpretation rounds "scarless texture but depigmented" to
"VP works, just leaves white patches." A more sober reading is "K K's
outcome is ambiguous; the chat doesn't have the data to call it scarless."
K K's own attributions also stretch the chat's claim: he attributes his
nose indentation to **cartilage loss** [#596] and his body depigmentation
to **post-surgery widening** [#601] — i.e., he's invoking non-VP causes
to explain failures while the chat invokes VP for the successes. This is
classic motivated reasoning. The chat doesn't push back on it.
### 1e. Members are extrapolating ahead of the evidence
A more accurate read than v1's framing:
- **Barfy Jones** is planning hairline-advancement surgery at **1.0–1.2 mg/cm²**.
This is above the highest documented dose (0.8 in Bloxham n=3). It's
extrapolation, not directly-evidenced — worth noting, not a panic.
- **S S** is planning facial excision after **two doctors declined and a
third agreed provisionally**. Surgeon refusals to use off-label compounds
are common and usually driven by liability, conservatism, and lack of
established protocol — not necessarily by having read the VP literature.
Two refusals is a soft signal at most. (v1 of this critique called it
hard signal; that was overreach.)
- **Fabio** is planning self-injection because his surgeon refused
intraoperative VP. Self-injection is a real escalation in risk — both
George Kal [#684-686] and Enrico [#697] pushed back hard in the chat.
This one the chat actually did contest.
- **George Kal**'s pico-laser + VP plan is unprecedented in published
literature. His "VP is the anabolic steroid; laser is the training"
framing is a model, not a mechanism — but it's a coherent model and
the experiment, if reported, would be a real N=1 contribution.
The chat is a community of people with significant scar-related distress
trying to access something the medical mainstream isn't offering. That
context shapes the protocol's adoption pattern. German Cherdakli is
**explicitly self-aware** about this:
> "my desire to believe in VP as a 'miracle cure' compromises my usually
> skeptical judgment"
This is the most honest single sentence in the entire archive.
### 1f. The chat's formulation is mechanistically wrong
Per the timeline reconstruction, the chat's de-facto formulation at archive
end is **raw VP powder + sterile PBS + 50-dollar Amazon sonicator + 4mm
30g mesotherapy needle**. This protocol almost certainly produces a
**suspension**, not a solution — VP is a hydrophobic benzoporphyrin
derivative; sonication into water-based PBS does not solubilize it.
Lucy's first attempt did exactly this. Her own description:
> "I had a very bad reaction with non dissolvable vp"
> "Cover your wound from the sun for more than 2 weeks"
> "Because your body will take time to absorb the VP particles"
That's the predictable outcome of injecting micronized hydrophobic
particles into tissue. Lucy & Sage independently figured out the lecithin
liposome alternative (chloroform/methanol thin-film method); **the chat
never adopted it** — zero "lecithin" or "liposom" mentions in the entire
archive.
The chat's "Visudyne is just expensive equivalent" belief, traceable to
martha Halli, locked them into a formulation that doesn't work the way
they think it does. **This is the single most consequential factual error
in the chat, and it's a safety issue for anyone following the protocol.**
---
## 2. The published evidence base is also weaker than v1 suggested
### 2a. Toyos's 2025 papers are in a low-rigor venue
The verification-pass-2 found two Toyos papers in *Journal of Dermatology
Research and Therapy* and treated them as "first peer-reviewed human VP
data outside conference abstracts." **The journal is published by Heighten
Science Publications, which appears in some predatory-journal lists.**
Being in this journal is not the same as being in JAMA Derm, BJD, or
Plast Reconstr Surg. The methodology of Toyos's papers needs to be looked
at directly before treating them as load-bearing. (The audit agent is
checking this.)
Treating "appears in a journal" as "validated by peer review" is the kind
of error the chat makes about Bloxham's YouTube videos — same epistemic
slip in a more formal wrapper.
### 2b. Bloxham, Bargouthi, and Toyos all have financial conflict of interest
All three are practicing surgeons with financial interest in promoting their
VP work. None have published independent confirmation. Bloxham's results
are his own YouTube videos; Bargouthi's results are his own website
(verteporfin.org); Toyos's results are her own papers in a low-tier
journal. None of these are independent observers reporting on each others'
patients.
This is not how clinical evidence is established. Until SCARFREE-001 reads
out — at which point we'll have a sponsor-independent intra-patient RCT
on abdominoplasty — every "VP works" claim is coming from people who
benefit financially or psychologically from VP working.
### 2c. The Mascharak mouse paper has its own limits
Mascharak 2021 is real, peer-reviewed Science. But:
- It's a **mouse** model. Mouse skin heals fundamentally differently from
human skin (looser dermis, more pannus contraction, different fibroblast
populations). Mouse-to-human translation in dermatology is famously poor.
- The "scarless" outcome is in **young mice with full-thickness wounds**;
the chat extrapolates to **mature human atrophic scar revision**, which
is a different biological problem.
- The "En1+ wound-resident fibroblast" lineage story is mouse-specific.
Human dermal fibroblast diversity has been studied separately and the
En1+ lineage is not necessarily the dominant scar-forming population.
- The 4-dose-impairs-healing finding is real but the mechanism is unclear
— could be cumulative cytotoxicity, could be off-target.
The chat treats the Mascharak paper as foundational evidence for human
protocols. The leap from mouse YAP/TAZ inhibition to human dose-response
is large.
### 2d. The pig paper is paywalled, so half the chat's pig claims are unverified
Mascharak 2025's pig paper is paywalled. The literature audit could only
access the abstract / conference abstracts. The "0.18 mg/cm² best" claim
in Enrico's [#186] table — which the entire community uses to scale doses
— could not be located in any accessible primary source. It may exist in
the paywalled paper; it may exist in a conference poster; it may have been
miscopied from somewhere; it may be made up. **No one — including me —
has actually read the source for this number.**
This is an ongoing project for someone with institutional access.
---
## 3. My v1+v2 outputs have problems beyond the corrections so far
### 3a. I trusted Sonnet sub-agents too much in Phase B
The Phase B literature work was done by Sonnet agents at speed. They
produced confident-sounding outputs with cited URLs, and I integrated
those outputs into the synthesis without re-fetching the URLs myself. The
B.2 agent literally invented the "Bloxham is Australian" chat claim. The
audit agents I ran later confirmed the same pattern across the literature
files. **Specific fabrications and errors found:**
**`research/verteporfin/safety.md`** (the worst offender — the audit
agent calls it "most riddled with sub-agent confidence-illusion"):
- "Verteporfin in breast milk up to 66% of plasma levels" — **likely
fabricated**. Cannot be sourced to any standard label or PK summary.
- "246 moderate / 13 minor drug interactions" — suspiciously precise,
not in primary sources.
- Multiple Houle & Strong 2002 statistics cited from PMIDs the agent
never actually fetched.
**`research/verteporfin/mechanism.md`**:
- "0.1–3 µM in vitro" — wrong; Wang 2016 actually tested 0.1–10 µM.
- "Indistinguishable from unwounded skin" overstates Mascharak 2021's
actual qualitative descriptions ("substantial hair growth," "numerous
HF/SG-like structures"). The paper's hair-follicle data is not
per-mm² density-quantified.
- Missing entirely: VP's off-target mechanisms (singlet-oxygen
cross-linking even at low light, autophagy inhibition, oligomerization
of ~250 proteins). The "light-independent YAP-TEAD inhibition" framing
is the Mascharak paper's claim, not a settled mechanism.
**`research/verteporfin/dose-response.md`**:
- All Bloxham doses (0.4/0.6/0.8 mg/cm², n=3) come from
**folliclethought.com — a hair-loss commercial blog**, not any indexed
publication. The chat's anchor dose is non-peer-reviewed journalism.
**`research/verteporfin/human-evidence.md`**:
- Misattributed the Dua 2025 review to *Advances in Skin & Wound Care*
— the actual journal (per DOI prefix 22925503) is *Plastic Surgery*
(CSPS-SAGE).
- Toyos's two 2025 ClinMed papers were missing entirely. They exist;
but **ClinMed International Library is a likely-predatory open-access
publisher**, the journal isn't PubMed-indexed and has no impact factor.
The "zero peer-reviewed human VP scar evidence" claim is technically
false (papers exist) but the claim's spirit is right (rigor isn't
there).
**`research/scar-techniques/fractional-laser.md`**:
- v1's "picosecond fractional vs CO₂ −44%" comparison — **wrong**. The
actual Frontiers 2023 trial (n=31) was pico vs **Er:YAG**, not CO₂.
My CRITIQUE.md citing this as a defensible alternative for darker
skin tones rested on the wrong comparator. Pico-vs-Er:YAG is real
(−39% vs −44% ECCA, ns; PIH 3.2% vs 25.8%); pico-vs-CO₂ is **not
in the literature this file actually surfaced**.
- "73–92% PIH for ablative CO₂ in Asian populations" couldn't be sourced
to PMC6676816 as cited.
**`research/scar-techniques/microneedling.md`**:
- "RFMN 24% vs ablative laser 75% vs isotretinoin 67% relapse at 3 years"
— **likely fabricated**. PMC11619162 is a 1-month follow-up study,
not 3 years. The three-arm comparison cannot be located in cited literature.
- The "5-FU 48.6% vs 16.9% repigmentation" cite is a vitiligo study,
smuggled into a scar-evidence section.
**`research/scar-techniques/emerging.md`** (also bad):
- **embrace IMPROVE was Level II, not Level I** (per the paper's own
classification). REFINE doesn't assign itself a level — the "Level I"
framing was the trial authors' marketing. Both trials are
**industry-conflicted**: Neodyne Biosciences supplied devices,
Neodyne-equity authors. The file did not flag the COI in v1.
- Avotermin's PubMed cite (21618480) is the Phase **2** trial, not the
failed Phase 3.
- STP705's primary endpoints exist only in **PRNewswire press releases**
— there are no peer-reviewed scar trial results.
- PDT keloid "5% recurrence" = 1/20 in a single 2012 case series at
9-month follow-up — not "best recurrence number of any keloid
intervention" as v1 framed it.
**`research/scar-techniques/subcision.md`**:
- "Cannula 0% vs needle 11% hematoma" — **fabricated**. The actual
Asilian 2019 study compared Nokor blade (not cannula) and found
3 vs 0 hematomas in n=28, **p=0.236, not significant**. Presenting
it as a clean, significant decimal comparison is wrong.
**`research/scar-techniques/tca-cross.md`**:
- The 2002 originating paper is **Lee, Chung, Kwahck & Lee** —
there's no "Davis." v1 said "Davis & Lee 2002" throughout.
- "80% excellent at 4 sessions" is from Bhardwaj & Khunger 2010,
n=10 completers (12 enrolled) — the file should say n=10, not n=12.
**`research/scar-techniques/silicone-pressure.md`**:
- "Grade A" letter for silicone gel under Mustoe 2002 couldn't be
verified from the abstract; the textbook gradings repeat through
reviews but the original-paper letter assignment isn't pinned down.
- Vitamin E "33% contact dermatitis" is real (Baumann 1999) but the
file cited it via the wrong intermediate review (PMC4506744 doesn't
contain the number).
The detailed audit results live in:
- `research/verteporfin/_AUDIT.md`
- `research/scar-techniques/_AUDIT.md`
**Implication**: even after v3 corrections, several numerical claims in
the literature files remain unverified by direct primary-source fetch.
The audit agents could only access ~half the cited URLs (paywall, 403,
Cloudflare). For any decision-relevant claim, **fetch the primary source
yourself before relying on it.**
### 3b. v2 still has stale text in some files
The verification-pass-2 was appended to `research/verteporfin/doctors.md`
rather than rewriting the older sections. The file now contradicts itself
in places (the upper sections still say things v2 disproved). A clean
rewrite would help; until then, **the appended Verification Pass 2
section supersedes anything earlier in that file**.
`research/verteporfin/human-evidence.md` was written before we knew about
the Toyos 2025 papers. Its "zero peer-reviewed human evidence" claim is
out of date. (Worth re-checking the Toyos papers' rigor before "fixing"
this — see 2a.)
### 3c. The "outcomes" framing throughout treats anecdote as evidence
Even after v2, the `by-topic/outcomes.md` and `RECONCILIATION.md` outcome
tables list K K's photo-interpretation outcome and Sc0t's self-report
alongside Bloxham's photo-interpretation outcome as if these are
data points of similar weight. They are all weak data. None of them
should drive a dose recommendation. The current framing implies they do.
### 3d. I didn't push back on the chat's protocol decisions
When Barfy Jones said he's planning 1.0–1.2 mg/cm² for hairline surgery,
I noted it as "above the highest documented dose." I should have said:
**there is no human safety data on intradermal VP at 1.0+ mg/cm². Mouse
data shows that 4-dose regimens impair healing. Going above the documented
range with no clinical follow-up is taking a real risk.** The chat
similarly didn't push back hard on Barfy's plan, S S's plan after
two doctor refusals, or Fabio's self-injection.
### 3e. Lucy & Sage's external case is treated as primary evidence
I integrated Lucy & Sage's case from `drafts/breif-notes.txt` as one of the
"4 documented outcomes." But Lucy & Sage's case has the same problems as
the chat cases — n=1, self-administered or partner-administered, no
control, subjective outcome assessment. Sage's "subjective review" at 8
months is just self-report. Treating this as a data point alongside
Bloxham's n=3 implicitly gives it weight it hasn't earned. The lecithin
formulation is more careful than the chat's; the experiment design is
not.
---
## 4. Things the chat (and the literature) don't push hard enough on
- **No one is doing histology.** Without biopsy, "scarless" is just
"looks-better-in-photos." Mascharak's mouse paper is histology-grounded;
every human off-label case is photo-grounded.
- **Chinese Weibo cases referenced via German [#2371-2374]** include at
least one bad outcome (parallel stretch mark adjacent to excision line).
The chat folds this into "Yang isn't flawless" without engaging with
what the bad outcome implies for dose / formulation / patient selection.
- **Photodynamic activation.** Verteporfin's *Visudyne* mechanism is
light-activated. Off-label intradermal use relies on the
light-independent YAP/TAZ inhibition mode. The chat assumes this is
totally separate, but VP is still a porphyrin photosensitizer — local
ambient light exposure during/after injection could in principle still
trigger photodynamic effects. Photosensitivity precautions (sun
avoidance) are mentioned but the in-tissue photodynamic interaction
isn't really debated.
- **Long-term follow-up.** Most chat-cited outcomes are at 1–8 months.
Bargouthi's longest-cited timepoint is ~500 days. We don't know what
VP-treated tissue looks like at 5 years.
- **Patient selection.** K K is keloid-prone. Sc0t is Caucasian with an
atrophic lip scar. Bloxham's patients are FUT scalp. Bargouthi's are FUE
scalp. Generalizing across these to "VP works for X" is a stretch the
chat does routinely.
- **Publication bias.** People who got bad results are less likely to
post photos / promote the protocol / appear in case series. The chat
is plausibly seeing the upper tail of outcomes.
---
## 5. The actually defensible takeaways
After being harsh, what's left?
- **Verteporfin has biologically interesting effects on YAP/TAZ
signalling**, demonstrated in mouse and at least suggested in pig.
Translation to human is unproven.
- **Some off-label clinical work (Bloxham, Bargouthi, Toyos) suggests
*something* is happening** — patients and surgeons report subjective
improvement. Whether this is VP-specific, placebo, surgeon skill,
patient selection, or photo-quality is undetermined.
- **The chat is a community of people processing scar-related distress
who have built a protocol on n=3 photo evidence**, with at least one
central formulation error, with anonymous protocol authors, and with
members planning procedures above the documented dose ceiling. The
protocol is **not** a clinical guideline; it's a folk consensus.
- **The first powered human trial is now enrolling.** SCARFREE-001 will
give actual signal in 2026-2028. Until then, everything else is
preliminary.
- **For specific scar morphologies, there are better-evidenced
alternatives** the chat doesn't engage with — **but most of those
evidence bases are themselves weaker than my v2 outputs implied**
(see audit findings appended to §3 below). The actually-decent ones:
- **TCA CROSS** for icepick acne scars: Bhardwaj & Khunger 2010,
n=10 completers, 80% excellent — small but real, with priming.
(My v1 misattributed this to "Davis & Lee 2002"; the actual
technique paper is Lee et al. 2002.)
- **Subcision + PRP**: PubMed 30102625, n=40 split-face, +32% vs +8%.
Real RCT, blinding unclear.
- **5-FU intralesional** monotherapy: PMC11403916, 16% relapse at
average 27-week follow-up (n=331). The "TAC 33% at 1 year" comparator
is from a different paper, different follow-up window — apples-to-oranges.
Treating VP as the only option is still a category error, but the
alternatives are also working with smaller-than-presented evidence bases.
---
## 6. What this means for anyone reading the project outputs
- **Don't dose from this material.** The chat's 0.8 mg/cm² consensus is
a folk number anchored on Bloxham's n=3 photo evidence. It's not safe
to treat it as a clinical recommendation.
- **Don't follow the chat's PBS+sonicator formulation.** The chat itself
has a formulation problem they haven't acknowledged. Lucy & Sage's
lecithin liposome approach, or actual Visudyne, is closer to what the
research papers used.
- **Don't generalize across patient types.** K K → you, Sc0t → you,
Bargouthi's FUE patients → your acne scars are all unjustified leaps.
- **Don't dismiss the surgeons who refuse.** Two surgeons declining S S's
plan, multiple declining Fabio's intraoperative VP — these are signals,
not friction.
- **Wait for SCARFREE-001 if you can.** Powered human data is coming.
Acting on the current evidence base is a choice, not a knowledge
necessity.
- **Treat my project outputs as a structured summary of weak evidence,
not as a guide to action.** Even after v2 corrections, plenty of
unverified claims remain in the literature files. The audit pass is in
progress.
---
## 7. What the project is good for
Despite the above, the project does provide:
- A **clean per-poster reconstruction** of who said what, with corrected
attribution
- A **chat-side timeline** showing how the dose consensus formed (it
formed in 24 hours on Aug 12, on one YouTube video, with no pushback —
knowing that helps calibrate how much weight to give it)
- A **list of named errors** in the prior outputs (this file) and a
separate change log (`CORRECTIONS-V2.md`)
- An **inventory of named doctors** and their actual public records vs
chat-internal claims
- An **inventory of better-evidenced scar techniques** that the chat
doesn't focus on
- An **honest acknowledgment of what's still unverified** so a follow-up
pass knows where to look
What it doesn't provide is a clinical protocol. Don't try to use it that way.