Retinoids, retinol to tretinoin.

Tretinoin is retinoic acid itself: no conversion, full receptor activity, pharmaceutical manufacturing standards and a stated strength. Its trials are summarised in Part 01 — the 1988 JAMA trial (Weiss), the 296-person emollient-cream trial (Olsen 1992), the 0.1%-versus-0.025% equivalence study (Griffiths 1995), the two-year trial (Kang 2005) and the maintenance study (Olsen 1997). The systematic review that compared tretinoin with 25 alternatives found its comparators "had variable efficacy" and tretinoin the gold standard they were measured against (Siddiqui 2024).

Which strength: 0.025% for photoaging in most people, 0.05% for those who adapt easily or have heavier pigment; 0.1% adds irritation, not results. Formulation: creams are gentler than gels; the emollient cream used in the photoaging trials is not sold everywhere, but any tretinoin cream at the same strength delivers the same molecule. Access in Europe: prescription-only, licensed for acne, prescribed off-label for aging by dermatologists and many GPs; €8–25 a tube, and a tube lasts two to three months at a pea a night. The one thing money cannot improve on.

Tazarotene is a synthetic retinoid with the largest modern photoaging trials of any retinoid. In the 349-person dose-ranging study, treatment success at 24 weeks was 67% on 0.1% tazarotene, 55% on 0.05% tretinoin emollient cream and 22% on vehicle (Kang 2001). In 563 patients it produced significantly more treatment success and at least one-grade improvements in fine wrinkling, mottled hyperpigmentation, lentigines, elastosis, pore size, roughness and coarse wrinkling at 24 weeks, with improvement that "had not plateaued by week 52" — and 20 of 283 tazarotene patients versus 1 of 280 on vehicle stopped for side effects (Phillips 2002). A second 568-patient trial confirmed it with significance "as early as week 2" (Kang 2005), and a direct comparison in 173 subjects found every significant difference in favour of tazarotene over tretinoin 0.05%, with a transient burning sensation in the first week (Lowe 2004).

The network meta-analysis put the trade-off in one sentence: tazarotene "most effective for coarse wrinkles" and, with glycolic acid, carrying "higher adverse event risks" (Lin 2025). In Europe it is licensed in a handful of countries for psoriasis and acne and otherwise obtained by import on a private prescription — the strongest rung, for a face that has already climbed the others.

Adapalene binds the receptors directly, is stable in light and is less irritating than tretinoin — which made it the acne retinoid of choice and, more slowly, a photoaging candidate. The key trial randomised photoaged subjects to adapalene 0.3% gel or tretinoin 0.05% cream once daily for 24 weeks and found no significant difference in global photoaging, periorbital or forehead wrinkles, melanosis or keratoses, concluding it "may be considered a safe and effective option for the treatment of mild or moderate photoaging" (Bagatin 2018). The earlier nine-month trial in 90 patients found keratosis counts down and lentigines lighter in 57% and 59% (0.1% and 0.3%) against 36% on vehicle, with blinded photographic review showing improved wrinkles (Kang 2003). A 2025 randomised trial of 0.1% cream against no treatment in 58 women found a five-point fall in a skin-aging score at six months, with pigmentation and wrinkles improved (Tran 2025).

Graded moderate: the trials are consistent but few, small and not placebo-controlled at the strength that matters for aging. A cosmetic prodrug, oleyl adapalenate, improved wrinkle severity 9.5% against 4.1% for 0.5% retinol in a 48-person 12-week trial — one manufacturer study (Nguyen 2024). In Europe adapalene remains prescription-only at both strengths; where a doctor will not prescribe tretinoin off-label, adapalene for "acne-prone skin" is often the easier conversation.

Retinaldehyde sits one enzyme away from retinoic acid, and the biology follows: in the human enzyme-induction study it was active at 0.01% against 0.025% for retinol (Duell 1997), and the Geneva group that introduced it showed dose-dependent epidermal thickening at 0.05–0.5% and, in 229 patients, that 0.05% and 0.1% "were well tolerated and allowed prolonged use (up to 3 years)" while 1% was tolerated by only 70% (Saurat 1994). The efficacy trial randomised 125 photodamaged patients to 0.05% retinaldehyde, 0.05% retinoic acid or vehicle: at week 18 both actives significantly reduced wrinkle and roughness features on silicone replicas of the crow's feet, the vehicle did nothing, and "retinoic acid caused more local irritation, and affected compliance" (Creidi 1998). A split-face study in 23 Korean volunteers found vesicle-encapsulated retinal 0.05% and 0.1% improving crow's-feet depth, hydration and elasticity significantly more than retinol at the same concentrations over eight weeks, with no adverse events (Kim 2021).

Graded moderate, not strong: one 125-patient randomised trial, one 23-person study and manufacturer work (Brown 2023) against tretinoin's dozens. But it is the best-evidenced thing you can buy without a prescription, its instability is managed by encapsulation in the serious products, and because Regulation 2024/996 does not name it, a 0.1% retinal cream is now the strongest retinoid legally on a European cosmetic shelf. Start at 0.03–0.05%, move to 0.1% after three months.

Retinol is two conversions from retinoic acid and penetrates well without occlusion (Duell 1997). Its independent evidence is the Michigan trial in two nursing homes: 0.4% retinol lotion applied by staff up to three times a week for 24 weeks to one arm of 36 elderly subjects (mean age 87) improved fine-wrinkle scores by 1.64 points against 0.08 on vehicle, with significantly increased glycosaminoglycan and procollagen I on biopsy (Kafi 2007). The manufacturer evidence is larger: a 52-week double-blind trial of 0.1% stabilised retinol in 62 subjects found crow's-feet fine lines improved 44% and mottled pigmentation 84%, with over half the subjects two grades better on many measures and increased procollagen I and hyaluronan on biopsy (Randhawa 2015); a pooled analysis of six vehicle-controlled studies (237 retinol, 234 vehicle, mean age 47) found improvement in all signs of photoaging from week 4 through 12 with only a handful of mild irritation events (Farris 2024). A 2024 systematic review of cosmeceuticals gave retinol its top recommendation grade (Lau 2024), and the network meta-analysis found it effective for fine wrinkles and pigment (Lin 2025).

Against tretinoin it is weaker weight for weight and gentler: a 1.1% retinol-plus-esters formula matched 0.02% tretinoin on photoaging scores in 20 women at 24 weeks with erythema six times less often (Chien, JAMA Dermatology 2022), and 1% retinol caused more redness than a conjugated retinoid in a split-face study (McDaniel 2017). The EU cap of 0.3% removes the 0.5–1% products from the market by May 2027; the 0.1–0.3% range is where most of the trial evidence sits anyway. What matters more than the number is the formula: stabilised, encapsulated, airless — because the stability study found plain retinol products losing most of their content on the shelf (stability study).

Bakuchiol is a meroterpene from the seeds of Psoralea corylifolia. Gene-expression profiling found it regulating retinoid-responsive genes in skin cells despite no structural resemblance to vitamin A (Chaudhuri 2014). The trial that made its name randomised 44 patients to bakuchiol 0.5% cream twice daily or retinol 0.5% once daily for twelve weeks, with blinded grading of high-resolution photographs: both "significantly decreased wrinkle surface area and hyperpigmentation, with no statistical difference between the compounds", and retinol users reported more scaling and stinging (Dhaliwal 2019). A systematic review found 30 articles — sixteen preclinical, seven clinical — and results "comparable to those achieved by topical retinoids" for photoaging, acne and post-inflammatory pigmentation (Puyana 2022).

Why emerging rather than moderate: one 44-person trial, twice-daily bakuchiol against once-daily retinol, twelve weeks, no vehicle, no biopsy. "As good as retinol" in that design is a real signal and a modest bar. Its practical niche is honest — the person who cannot use a retinoid at all, including in pregnancy, where bakuchiol has no known teratogenic signal but also no safety trial; and the person who wants the daytime product a photolabile retinoid cannot be.

Oral isotretinoin remodels skin from the inside — the same p53 reduction and collagen increase seen with topical tretinoin appear in biopsies — and Brazilian dermatologists have studied low doses for photoaging for twenty years. The randomised comparison enrolled 24 people aged 50–75 with advanced photoaging: 20 mg isotretinoin or 0.05% tretinoin cream, each every other day for six months, with moisturiser and sunscreen. Both groups improved clinically and histologically, with "no differences between groups": "low-dose ISO was not superior to 0.05% RA" (Bagatin 2014). The 2020 review found six studies and 251 patients in total, three randomised, "many hampered by methodological challenges", and concluded there is "currently insufficient evidence" (Honeybrook 2020). The network meta-analysis nonetheless ranked it first for fine wrinkles, on those few trials (Lin 2025).

The costs are systemic: absolute contraindication in pregnancy under a formal prevention programme, lipid and liver monitoring, dry eyes and a documented need for ocular surveillance in off-label photoaging use (ocular surveillance review). Emerging, and a conversation for a dermatologist who is already treating something else.

Retinyl esters are how the body stores vitamin A, and they are what many products mean when the front label says "retinol". They are stable, cheap and inert enough to sit in a moisturiser without irritating anyone — because very little of them becomes retinoic acid. The human enzyme study needed 0.6% retinyl palmitate to match 0.025% retinol (Duell 1997). The one independent, double-blind, placebo-controlled trial randomised 80 subjects to retinyl propionate cream or placebo for 24 weeks, 60 of them continuing to 48: "no statistically significant differences between the effects of the retinyl propionate cream and the placebo preparation were apparent for any of the clinical, histological or profilometric parameters" (Green 1998).

The manufacturer trials are kinder and confounded: a Procter & Gamble regimen of niacinamide, peptides and 0.3% retinyl propionate with SPF 30 improved wrinkles more than 0.02% tretinoin at eight weeks and comparably at 24 in 196 women (Fu 2010), and a Unilever retinyl propionate–climbazole combination beat 0.1% retinol on a split face (Hawkins 2017) — in each case the ester travelled with other actives and without a placebo arm. Since November 2025 the esters count toward the EU's 0.3% retinol-equivalent cap (0.55% retinyl palmitate ≈ 0.3% retinol). Graded limited: an independent negative trial and industry positives that cannot isolate the molecule.

Hydroxypinacolone retinoate is an ester of retinoic acid, sold as the "granactive retinoid" in a 2% or 5% complex that contains a tenth of that as the active. Its selling points are real in the laboratory: it was the most stable of sixteen retinoid derivatives in the six-month commercial stability study (stability study), it binds retinoid receptors in cell work, and a 2026 mouse study found a 9-cis version protecting against UV-induced matrix loss (Hu 2026). The clinical evidence is a Chinese cosmetic-science programme pairing it with retinyl propionate ("Gravi-A"): an optimal 5:9 ratio in fibroblasts, then an eight-week serum study reporting wrinkles improved 8.3%, smoothness 11.9% and elasticity 14.5–22.6% with fewer adverse effects than retinol at the same concentration — and no vehicle control (Wang 2023; nanoparticle formulation).

That is the whole human record as of 2026: no independent randomised trial against vehicle, retinol or tretinoin, and no biopsy evidence of collagen. It may well work — the chemistry is plausible — but a product whose advertising leans on stability and comfort rather than on a trial is telling you something. Graded limited on evidence, not on mechanism; the same rung as the classic esters until someone runs the trial.