No, Tretinoin Isn't Killing People. Reading the Papers.
A close look at the claim that tretinoin causes deaths, thins your skin, lets pollutants into your bloodstream, and is the same compound used in chemotherapy. And what the actual papers say.
Hi Complexions,
A lot of you sent me this week a TikTok by a creator named Taylor, making a list of fairly serious claims about retinoids. ( I don’t want to link it, so it doesn’t get any extra boost from the algorithm. )
Among them: that the VATTC trial was terminated early because tretinoin caused excess deaths.
That topical retinoids convert to retinoic acid, which is the same compound used in chemotherapy.
That they force the skin to turn over faster than it’s biologically designed to.
That after 12 months your epidermis is thinner and your barrier compromised.
That this barrier compromise lets pollutants and chemicals enter your bloodstream more readily.
That this is why photosensitivity gets mentioned. And then a separate set of claims about Accutane: that it’s pregnancy category X, that it raises LDL, triglycerides, and causes insulin resistance regardless of lipid levels within six months.
These are specific factual claims with specific cited evidence.
I have opinions about all of this. I have a over decade in biochemistry and cosmetic chemistry and a particular relationship with retinoids that goes back years. So I want to do something specific today: I want to put my opinions, my clinical instincts, and my personal experience to one side, and look at the evidence with you.
This is how I do all my research, actually. I don’t go in trying to confirm what I already believe. I go in trying to find out. If the evidence says I’m wrong, I’d rather know now than carry on being confidently wrong for another decade. That’s the whole game.
If Taylor is right, I’ll say so. If the evidence diverges from the framing, I’ll say that too. The aim isn’t to win against him. It’s to find out what the evidence actually says and put it in your hands so you can decide for yourself.
So today: no priors, no allegiances. Just the papers.
What’s actually being claimed
There are eight separate claims in the video and they need separating because they don’t share an evidence base. Some are about topical retinoids. Some are about oral isotretinoin. Some are about a specific 2009 trial called VATTC. The unifying problem, which I’ll come back to, is that the video repeatedly treats findings from one of these categories as evidence for another. Topical retinoid users are not isotretinoin patients. Cosmetic dosing is not chemotherapy dosing.
Hold that thought. Let’s start with the most serious claim.
Claim 1: The VATTC trial was terminated because tretinoin caused deaths
This one is technically true. But the way it’s being used isn’t. The combination of those two things ‘yes, but’ is exactly the shape of how scientific findings get weaponised online, so it’s worth slowing down here.
The trial is Weinstock et al., Archives of Dermatology, 2009. VATTC stands for Veterans Affairs Topical Tretinoin Chemoprevention trial.
Here’s what it actually was.
It was a vehicle-controlled randomised trial run across six US Department of Veterans Affairs sites between 1998 and 2004. It recruited 1,131 veterans with a mean age of 71, approximately 97% male, all of whom had a history of at least two skin cancers (basal cell or squamous cell) in the previous five years.
So the population was elderly, overwhelmingly male, heavily sun-damaged, and already at high risk for skin cancer and the kinds of health problems that come with decades of UV and smoking.
The treatment was 0.1% tretinoin cream applied twice daily to the face and ears, the strongest concentration on the market, applied twice as often as most people apply it, on a bigger area than most people use. The primary endpoint was new skin cancer. Death was not part of the plan.
The trial was halted around six months early because the safety monitoring board saw more deaths in the tretinoin arm than the vehicle arm. At end-of-study: 122 deaths out of 566 in the tretinoin arm (22%) versus 90 out of 565 (16%) in the vehicle arm.
That’s a real, statistically significant difference, and it’s the fact Taylor is pointing to.
Here’s what he doesn’t mention, which is the entire interpretation of that fact by the people who actually ran the trial.
First, the investigators themselves wrote, precisely, in the abstract: ”We observed an association of topical tretinoin therapy with death, but we do not infer a causal association that current evidence suggests is unlikely.” And in the body of the paper: ”we found it difficult to construct biologically plausible mechanisms that would explain a direct causal link... and we were unable to conceive of a plausible mechanism by which tretinoin could indirectly lead to a fatal outcome.”*
This is the paper Taylor is citing. The paper’s own authors rejected the causal interpretation he’s repeating.
Second, there was no dose-response. In fact, the strongest mortality signal was in the once-daily stratum (P = 0.005), not the twice-daily stratum. If tretinoin were actually toxic, you’d expect more drug to produce more death. The data show the opposite. The paper explicitly states there was ”no dose-response relationship between exposure to topical tretinoin and death risk.”
Third, the deaths clustered in causes that match what you’d expect in this population, not in retinoid pharmacology. The excess was in lung disease (lung cancer, COPD, respiratory issues) and vascular events. That’s the death pattern of decades of smoking in elderly men, not of vitamin A applied to the face. Smoking data in the trial were incomplete, a limitation the editorial and independent commentators both flagged, which means the smoking imbalance between the arms couldn’t be fully adjusted for.
Fourth, and Weinstock himself wrote this, verbatim, in his own reply: ”this trial does not provide appropriate grounds for hesitating to use topical tretinoin in clinical practice in the absence of additional evidence.”
Topical tretinoin has been on the US market since 1971. If there were a real population-level mortality signal in cosmetic-dose users, fifty-plus years of pharmacovigilance in millions of women would have produced it by now.
It hasn’t.
So the actual shape of the VATTC story is this. A real mortality signal in a frail elderly male veteran population. High cosmetic dose on a wide treatment area. No dose-response. Cause-of-death categories matching the population’s smoking history. And the trial’s own investigators concluding that a causal link is unlikely and the finding does not generalise.
That’s the paper. That’s what it says. It is not what Taylor is using it to say.
Claim 2: Retinoic acid is the same compound used in chemotherapy
Yes, and so what. Let me unpack why this framing fails.
All-trans retinoic acid (ATRA, tretinoin) is the active molecule in topical tretinoin. It’s also a drug used to treat acute promyelocytic leukaemia (APL), a specific type of blood cancer. So at the molecular level, yes, same compound in both places.
The issue is dose. ATRA for APL is given orally at 45 mg/m²/day, which works out to roughly 80 mg per day for an average adult. Nearly all of that gets absorbed into the bloodstream. The point of treatment is to push the leukaemic cells to differentiate, which needs plasma levels orders of magnitude above what the body normally encounters.
Topical tretinoin is a completely different exposure. The classic absorption study Latriano et al., JAAD, 1997 measured how much of the tretinoin you put on your face actually makes it into your bloodstream. With long-term use, the answer is around 1.1% of the applied dose. A pea-sized application of 0.05% tretinoin on a whole face, about 0.5 mg of tretinoin, works out to roughly 5 to 25 micrograms per day reaching the bloodstream.
The authors found that plasma tretinoin levels in long-term users were ”not significantly changed when compared with the corresponding endogenous concentrations before treatment.” In plain English: topical use doesn’t measurably move your baseline retinoic acid level out of the normal range your body already runs at.
The arithmetic: 5–25 micrograms vs 80,000 micrograms. That’s roughly three to four orders of magnitude. A 3,000–15,000-fold difference in systemic load.
The “same compound as chemotherapy” line is the skincare version of objecting to paracetamol because, at 50 grams, the same molecule will destroy your liver. True at the molecule. Meaningless at the dose. The reason topical tretinoin is approved as a skincare treatment and not a chemotherapy drug is that the amount actually reaching your bloodstream sits within the range your body already handles every day.
Claim 3: Retinoids force the skin to turn over faster than it’s biologically designed to
This is the accelerated-turnover framing again, dressed in slightly different language. The teleology that the skin was “designed” for a particular turnover rate and retinoids violate that design doesn’t survive contact with biology.
Retinoic acid receptors exist in your skin specifically to respond to retinoic acid. They’re not vestigial. Your body makes small amounts of retinoic acid as part of normal skin function. Topical retinoids work by binding receptors that are already there, ready, waiting for them. The molecular machinery is physiological. The mechanism isn’t alien to the skin. It’s a dose escalation of something the skin already does.
What retinoids do drive is a real proliferative response above the baseline. The pathway runs through RXR/RARγ in the suprabasal keratinocytes, which produce HB-EGF and amphiregulin, which then activate EGFR on the basal keratinocytes Chapellier et al., EMBO J, 2002 ; Rittié et al., JID, 2006. That’s the same EGFR pathway your skin uses for normal wound healing and growth factor responses. It’s not a chemotherapy-style toxic insult. It’s an amplified version of a normal cell-signalling cascade.
And functionally, the proliferation produces better skin, not worse. Cho et al. 2005 found reduced keratinocyte and melanocyte atypia in long-term users, meaning the cells coming up through the epidermis look more normal, not more disordered. Bhawan’s four-year follow-up found the same. If retinoids were forcing pathological turnover, you’d see atypia rising over time. The opposite happens.
The “biologically designed” framing is also doing rhetorical work it shouldn’t. By the same reasoning, exercise forces your muscles to grow faster than they were “designed” to. Sunscreen prevents melanin production your skin was “designed” to do. The argument proves too much.
Claim 4: After 12 months, the epidermis is thinner and the barrier is compromised
This is the factual error I want to be cleanest about, because it’s the one most likely to get repeated in comments and the one a reader can most easily verify if they want to.
There is no peer-reviewed study showing the epidermis is thinner than baseline at 12 months of cosmetic-dose topical retinoid use. None. I went looking specifically. It doesn’t exist.
What does exist, and what I think Taylor is misreading, is a well-documented finding that goes like this. The acute thickening of months 1 to 6, the hyperplasia that happens when retinoid signalling first kicks the basal keratinocytes into higher gear, regresses back toward baseline by month 12. That’s tissue normalisation. Not thinning. The skin is returning to its normal thickness from a temporarily elevated thickness. Those are very different things.
The actual long-term histology, which I’ll cite in full because the claim keeps getting repeated:
Kang et al. 2005, the two-year placebo-controlled trial in 204 subjects, with 101 in the tretinoin arm and 103 in vehicle. Verbatim from the abstract: ”Histologic evaluation showed no increase in keratinocytic or melanocytic atypia, dermal elastosis, or untoward effects on stratum corneum following treatment with tretinoin compared with placebo.” Procollagen 1C terminal, a direct marker of new collagen synthesis, was significantly elevated at month 12 (p = 0.0074). The stratum corneum, which is the actual barrier layer, showed no abnormalities.
Bhawan et al. 1996, the four-year histological follow-up. No cytologic atypia. Decreased dermal elastin (which is good, elastosis is a hallmark of photoaging). Reduced perivascular inflammation. The epidermis normalised from its acute hyperplastic peak but did not drop below baseline.
Cho et al. 2005. 34 subjects, at least six months of daily 0.05% retinoic acid. Papillary dermal collagen band roughly doubled in thickness. Atypia reduced.
So an honest summary of what 12+ months of daily topical retinoid use looks like on biopsy:
the epidermis returns to its normal thickness after an acute hyperplastic peak,
collagen continues to accumulate in the dermis,
atypia goes down rather than up,
the stratum corneum is intact,
and elastosis improves.
The “thinner and compromised at 12 months” claim is the opposite of what the three landmark long-term studies actually show.
Claim 5: Retinoid-damaged skin lets pollutants and chemicals into the bloodstream
Two things in this claim: the barrier compromise, and the pollutants-in-bloodstream consequence. They need separating.
The first half is grounded in something real.
During retinisation (the first 2–12 weeks of use), transepidermal water loss (TEWL) does rise. Effendy and Maibach showed that topical retinoic acid produces a TEWL increase comparable to 12% glycolic acid during the acute phase. Retinoid dermatitis is a real thing. Your barrier is genuinely more permeable in the early weeks. So the acute part of the claim isn’t made up.
But the next point doesn't hold.
First, the barrier disruption is transient. Kang’s two-year study explicitly found no effects on the stratum corneum at the long-term end of treatment. The barrier rebuilds as the new corneocytes, the ones differentiating under retinoid signalling, replace the old ones. TEWL comes back toward baseline. Chronic users are not in a state of ongoing barrier breakdown.
Second, the “pollutants enter the bloodstream” framing has no empirical foundation. There is no peer-reviewed study showing that topical retinoid use measurably increases skin absorption of environmental pollutants like heavy metals, phthalates, polycyclic aromatic hydrocarbons, anything you’d care about.
TEWL measures water moving out of the skin. It’s an indirect surrogate for barrier function. It doesn’t translate one-to-one into pollutants getting in and reaching the bloodstream. Even in atopic dermatitis, which causes dramatically elevated TEWL across years of life, there’s no recognised toxicity syndrome linked to pollutants getting in through the skin. The pollutants of public health concern reach the bloodstream mainly through the lungs and gut, not through retinoid-treated cheeks.
The acute irritation during retinisation is real. The systemic toxicity story is invented.
Claim 6: Photosensitivity is mentioned because retinoids damage the epidermis
This one is wrong in both halves, and the actual reason retinoids are recommended at night is interesting enough to be worth knowing.
Slade et al., 2009, in Photodermatology, Photoimmunology and Photomedicine, ran four prospective controlled trials looking at whether topical 0.05% tretinoin makes skin more sensitive to UV. The verbatim conclusion: ”tretinoin appears to be neither phototoxic nor photoallergenic in vivo.” The minimal erythema dose, the amount of UV needed to produce a sunburn, is not lowered on tretinoin-treated skin.
Smit et al. 2000 in Acta Dermato-Venereologica confirmed the same thing independently. So the photosensitivity claim, in the sense of “your skin becomes more vulnerable to sun damage,” is just not what the controlled trials show.
What is actually true: tretinoin and retinol are photolabile. UVA breaks them down. So if you apply them in the morning, you’re losing the molecule to sun before it can do its job. Nightly application is recommended because the drug doesn’t survive daytime UV, not because your skin is now sun-sensitive. The advice is about preserving the active ingredient, not protecting damaged skin.
Sunscreen during retinoid use is still recommended, but for two reasons that aren’t “your skin is now photosensitive.”
First, retinoids reverse photoaging that UV is currently causing. If you’re investing in repair, it makes sense to stop the ongoing damage.
Second, retinised skin in the irritation phase is more uncomfortable to burn, but that’s discomfort, not formal photosensitisation.
Claim 7: Oral isotretinoin (Accutane) is pregnancy category X
Correct. This is the one claim in the video where I have nothing to push back on, and it’s worth saying clearly because the rest of the piece has been corrective.
Oral isotretinoin is a known and serious teratogen. The category X designation is real, iPLEDGE in the US is real, the EMA Pregnancy Prevention Programme in the UK and EU is real. The malformation pattern, characterised in Lammer et al., NEJM, 1985, includes craniofacial defects, heart defects, CNS malformations, and thymic underdevelopment, all reflecting disruption of how the embryo’s neural crest cells migrate in early development. The relative risk for major malformations in exposed pregnancies was 25.6 (95% CI 11.4–57.5).
Any prescriber putting a patient on oral isotretinoin handles this with serious protocols. Two negative pregnancy tests before prescribing, monthly testing during, two forms of contraception, narrow prescription windows. The teratogenicity is actually one of the strongest pieces of evidence for how systemic retinoid pharmacology is genuinely different from topical use, millions of women have used topical tretinoin without producing the retinoic acid embryopathy pattern, because the systemic exposure simply isn’t there.
So on this one, he is right. Worth saying. And worth holding onto, because it makes the next claim, and the broader topical-vs-oral confusion, easier to navigate.
Claim 8: Oral isotretinoin raises LDL, triglycerides, and causes insulin resistance regardless of lipid levels within 6 months
Partially correct. Lipids: yes. Insulin resistance: the evidence is much weaker than the claim.
On lipids, the literature is clear and old. Bershad et al., NEJM, 1985 ran a 20-week study in 60 patients and found significant rises in triglycerides (men +46.3 mg/dL, women +52.3 mg/dL), LDL cholesterol, and total cholesterol, plus HDL drops. About 17% of completers had triglycerides between 200 and 600 mg/dL.
The key clinical detail, also from Bershad: ”Plasma lipid and lipoprotein levels returned to baseline by 8 weeks after discontinuation of the drug.” Larger population data (Zane et al. 2006, n=13,772) confirmed the same picture: transient, reversible, well-characterised. This is why patients on isotretinoin get blood lipids checked at baseline and during treatment.
On insulin resistance, the claim is much shakier than the lipid claim, and “regardless of lipid levels within six months” overstates what the studies actually show. The relevant primary data:
- Stoll et al. 2004 in Metabolism: isotretinoin raised plasma triglycerides but ”did not change whole body insulin-mediated glucose disposal and lipolysis.”
- Ertugrul, Karadag et al. 2011 in Clinical and Experimental Dermatology, 48 patients on isotretinoin for 3 months: triglycerides, AST, ALT, total cholesterol and LDL all rose. ”No significant change in fasting blood glucose, insulin, C-peptide levels or HOMA-IR.” The conclusion: ”Three months of isotretinoin treatment did not change insulin sensitivity in patients with AV.”
- Paschalidou et al. 2024, a systematic review and meta-analysis of 15 studies on isotretinoin and insulin resistance: ”no statistically significant outcomes were observed for insulin, glucose levels, and the HOMA-IR.” The only statistically significant metabolic outcome across the 15 studies was a rise in adiponectin, which is insulin-sensitising, the opposite direction from Taylor’s claim.
So the honest summary on Claim 8 is: lipid rises are real, expected, monitored, and reversible. The insulin resistance claim is not supported by the largest pooled analysis available, and some of the literature points in the opposite direction.
The line “isotretinoin causes insulin resistance regardless of lipid levels within six months” makes a contested question sound like a closed one.
Two other systemic effects of isotretinoin worth briefly naming, because they come up in this conversation.
The first is inflammatory bowel disease, where the historical signal that drove a decade of litigation has been substantially walked back by recent meta-analyses (Yu et al. 2023, pooled odds ratio 1.01 across 2.5 million participants, basically no association).
The second is depression and suicide risk, where the meta-analytic data are mostly null or slightly favourable for isotretinoin (Huang & Cheng 2017 found depression scores actually went down from baseline). Some people do experience mood worsening on it, and that's worth watching for.
So is Taylor right?
Let me work through it piece by piece, and let the evidence speak rather than me.
On VATTC, no. The trial’s own investigators rejected the causal interpretation he’s repeating. The population was 71-year-old male veterans on twice-daily 0.1% tretinoin to face and ears, the deaths clustered in smoking-pattern causes, and there was no dose-response. Weinstock himself wrote that the trial does not provide grounds for hesitating to use topical tretinoin in clinical practice.
On retinoic acid being “the same as chemotherapy,” technically true at the molecule, meaningless at the dose. 5–25 micrograms a day topically versus 80 milligrams a day orally is a three to four order of magnitude difference. The framing is dose-blind.
On forced unphysiological turnover, the framing fails the biology. RAR receptors exist in skin to respond to retinoic acid. The proliferation pathway uses normal EGFR signalling. Long-term histology shows reduced atypia, not increased.
On 12-month epidermal thinning, false. The three landmark long-term studies: Kang two years, Bhawan four years, Cho six-plus months, all show the opposite. No primary study shows net thinning below baseline at 12 months of cosmetic-dose use. The likely source of the claim is a misreading of acute-hyperplasia normalisation as atrophy. They aren’t the same thing.
On pollutants entering the bloodstream, the acute TEWL rise during retinisation is real, but the systemic-toxicity leap has no empirical basis. The pollutants of public health concern come through lungs and gut, not retinoid-treated skin.
On photosensitivity, false. Slade 2009 found no phototoxicity or photoallergenicity. The MED is unchanged. Retinoids are applied at night because the molecule is photolabile, not because the skin is photosensitised.
On pregnancy category X, correct. The teratogenicity of oral isotretinoin is real, serious, and well-managed by iPLEDGE and the EU Pregnancy Prevention Programme.
On lipids and insulin resistance with oral isotretinoin, partially correct. Lipid rises are real and reversible. The insulin resistance claim is overstated. The largest pooled analysis available shows no significant change in HOMA-IR, and the only significant metabolic outcome across 15 studies was an increase in insulin-sensitising adiponectin.
There's one problem running through the whole video, and I want to call it out, because it's the thing that produces almost everything else.
Taylor keeps mixing up two different worlds. On one side: chemotherapy ATRA, oral isotretinoin, twice-daily 0.1% prescription tretinoin in elderly male veterans.
On the other: a normal person putting cosmetic-strength retinoid on their face at night. He treats findings from the first world as if they tell us something about the second. They don't. Once you collapse the two together, you can make almost any prescription drug sound terrifying just by comparing it to its over-the-counter cousin.
But the dose changes everything. The route changes everything. The population changes everything.
None of that comes through in the video.
What I do with this
A note on the framing before I close, because I think it matters.
I opened this piece saying I’d put my priors aside. I want to be honest with you: I came in with a decade of biochemistry already forming an opinion about retinoids, and what I found in the literature largely agreed with that opinion. That could read as confirmation bias if I left it there.
Here’s the distinction I’d draw.
Confirmation bias is selecting evidence that confirms what you already believe and ignoring evidence that doesn't.
A calibrated prior is something different. It's when your existing view is itself the product of years of working through the actual evidence, so when you go back to the papers, you tend to agree with them because your view was already built on them.
The honest test isn't whether the conclusion matched what I expected. It's whether I would have changed my mind if the evidence had pointed the other way. And I would have. If the VATTC investigators had themselves concluded a causal link, or if there were a long-term study showing real epidermal thinning, or if the insulin resistance literature were consistent rather than contested, this article would be a different article. None of that is what the evidence shows. So the position I came in with is the position I leave with. That's not the same as not having looked.
So in practice, here's where I land.
The dose mix-up is the problem underneath everything Taylor argued.
Topical retinoids at cosmetic strengths, applied at night to a clean dry face by an adult who isn't pregnant, sit within the range of vitamin A your body already handles every day.
They don't compromise the barrier long-term.
They don't make your skin sun-sensitive.
They don't thin the epidermis at 12 months.
And they have a 54-year track record on the consumer market.
Oral isotretinoin is a different drug, with a different exposure, and a real list of side effects that needs to be managed by a prescriber, not feared into avoidance. For the people who genuinely need it, it's life-changing.
The chemotherapy framing is doing the most rhetorical damage in the video. Once you know that the systemic exposure from topical use is roughly 3,000 to 15,000 times lower than from oral ATRA, the comparison stops being scary and starts being misleading. If you take one thing from this piece, take that one.
If you’re spiralling after watching the video, take a breath. Look at the papers if you want to look at the papers, I’ve linked them all. Look at your own skin. The evidence is more boring and more reassuring than viral TikTok summaries tend to allow.
One last thing. If you want me to write the next piece, how to actually get through retinisation without chronic inflammation, what to layer with retinoids, which moisturisers and barrier supports have evidence behind them, leave a comment and tell me. I’ll write it.
Until next time, Complexions
Marina x
*If you found this useful, you can subscribe below. Free for every post, paid for those who want to support the time these pieces take. Either way, I'm glad you're here.
This article is educational and reflects my interpretation of the published literature as a cosmetic chemist and biochemist. It is not medical advice and is not a substitute for consultation with a qualified dermatologist or other healthcare professional. If you have a skin condition, are pregnant or breastfeeding, are taking prescription medications, or are considering starting a prescription retinoid, please speak with your GP, dermatologist, or pharmacist about what is appropriate for you.
Individual skin responds differently to topical retinoids based on skin type, barrier status, concurrent products, climate, age, and genetic factors. The aggregate findings of clinical trials describe what happens on average across a study population. They do not predict how any particular individual will respond. If your skin is not tolerating a regimen the literature would otherwise consider safe, your skin’s signal is the one to trust.
All papers cited are linked directly so you can read them yourself. I encourage you to.
I was wondering if you would comment on this haha. Im so tired of uneducated influencers thinking that because they read a few studies the wrong way, they suddenly know more than board certified professionals