Welcome back to the Journal Review, your monthly walkthrough of the Journal of the American Academy of Dermatology. This is the April 2026 issue, and we've got three pieces on deck — an ethics journal club entry on artificial intelligence in diagnostic dermatology, a surgical pearl on tray organization, and a letter from the editor on desmoplastic melanoma staging. Let's get into it.

First up is the Ethics Journal Club piece, titled "Artificial Intelligence in Dermatology: Ethical Dilemmas and Diagnostic Decisions." This is framed as an advice-column style case discussion — a "Dear Dermatoethicist" letter — so there's no methods or results section here, just a case vignette worked through the standard bioethical framework of autonomy, beneficence, nonmaleficence, and justice. The scenario is one you've almost certainly lived: a patient runs a photo of a mole through a consumer AI app, gets flagged as concerning for malignancy, and shows up requesting a biopsy of something that looks to you like a garden-variety seborrheic keratosis. The authors note there are now over thirty patient-facing AI dermatology apps in circulation, and they frame the core tension as "algorithmic authority" — the tendency for patients to treat an algorithm's output as more objective or trustworthy than a clinician's exam, which can quietly erode their own autonomous decision-making even as it looks like patient empowerment.

On the diagnostic accuracy point, they cite a sobering number worth remembering when you're counseling patients: a Cochrane review found smartphone melanoma-triage apps with specificity as low as thirty-seven percent in some studies. That's a lot of false positives feeding into exactly this kind of visit. The ethical balancing act is real — beneficence pushes toward correcting the AI's misconception and steering the patient to biopsy and histopathology as the actual gold standard when clinical suspicion warrants it, while nonmaleficence cautions against reflexively biopsying every AI-flagged lesion just to appease the algorithm, given the costs of unnecessary scarring and psychological distress. The authors are candid that declining to biopsy carries its own liability exposure, especially if multiple apps concur on a melanoma flag, and they recommend the practical medicolegal hedge you're probably already doing: document your clinical rationale and engage in shared decision-making rather than unilaterally overriding the patient's request.

The justice angle is the part I'd flag as most clinically relevant going forward — these AI tools are well documented to underperform in patients with darker skin tones, largely a function of biased training datasets. The piece mentions industry responses worth knowing about, like Google's adoption of the Monk Skin Tone Scale and Meta's Casual Conversations dataset, both attempts to build more racially inclusive training data than the traditional Fitzpatrick scale allows. The authors also push for a transparency norm — that these apps should be required to disclose their actual specificity to users, which right now mostly isn't happening. The bottom line they land on, and it's a reasonable one to internalize: your role isn't to compete with the algorithm, it's to contextualize it, and validating the patient's proactive instinct while correcting the diagnostic record is how you preserve trust rather than erode it.

Next, a quick and genuinely practical one — a JAAD Online surgical pearl titled "Baker's Rack for Surgical Tray Organization." This is a pure technique piece, no methods, no data, just a problem and a fix, so I'll keep it brief. The problem: commercially marketed surgical instrument racks for Mohs and dermatologic surgery trays are excellent for organization, mobility, and protecting delicate instruments from the damage that comes with stacking trays — but they run anywhere from about fifteen hundred to ten thousand dollars. The proposed solution is refreshingly low-tech: commercial bakery pan racks, the kind used in restaurant kitchens, which run somewhere between one hundred and four hundred dollars. The authors note these are typically built from durable stainless steel or aluminum, rated to hold more than five hundred pounds, and — just like the purpose-built surgical racks — can be outfitted with locking casters for the same mobility benefit. It's a straightforward cost-saving substitution for anyone setting up or expanding a procedure room, and the takeaway is simply operational: if you're pricing out tray storage for a Mohs suite, a restaurant-supply baker's rack gets you the same functional benefit at a fraction of the cost.

Last, and the meatiest of the three, is a Letter from the Editor by Dirk Elston, responding to a companion article by Gibson and colleagues in the same issue, on desmoplastic melanoma. This is an editorial commentary, not original data, so there's no methods section to walk through — it's Elston synthesizing the accumulating evidence that desmoplastic melanoma, and more broadly several rare melanoma subtypes, behave biologically differently enough from conventional melanoma that lumping them into a single American Joint Committee on Cancer staging framework may be actively misleading clinicians who aren't steeped in the subtype literature.

The central biological argument is the pure-versus-mixed distinction. Pure desmoplastic melanoma, Elston argues, behaves more like dermatofibrosarcoma protuberans than like conventional melanoma — locally aggressive, prone to extensive perineural spread and high locoregional recurrence, but with a genuinely low rate of distant metastasis. Mixed or hybrid desmoplastic tumors, containing a conventional epithelioid melanoma component, behave very differently and carry substantially higher metastatic risk. He cites a Dutch cohort of almost two hundred forty patients with desmoplastic melanoma in which a Cox model found the mixed variant significantly associated with a positive sentinel node, and — this is the point worth remembering clinically — Breslow thickness only predicted outcome in the mixed subtype, not in pure desmoplastic disease. That's a meaningful challenge to reflexively applying depth-based staging logic across the board.

Therapeutically, the letter notes that anti-PD-1 immunotherapy has not meaningfully moved outcomes in pure desmoplastic melanoma, and that differential PD-L1 expression between pure and hybrid tumors suggests the mixed subtype patients are the better candidates for checkpoint inhibition — in other words, immunotherapy response may track with the same pure-versus-mixed biology rather than with stage alone. He also flags differences in neurofibromin expression and TERT promoter mutation frequency between subtypes as evidence of genuinely distinct molecular drivers, not just histologic variants of the same disease.

Elston broadens the argument to the other rare melanoma subtypes — acral, mucosal, uveal, and blue-nevus-like melanoma — noting these together make up only about five percent of all melanoma diagnoses, which is exactly why they're underpowered and underrepresented in the large randomized trials that current staging and treatment paradigms are built on. He draws quick contrasts: acral melanoma also responds poorly to immunotherapy but, unlike desmoplastic melanoma, carries worse overall survival; mucosal melanoma is marked by wide variability in presentation, poor prognosis, and low mutational burden; uveal and blue-nevus-like melanomas share GNAQ and BAP1 mutations distinct from other subtypes entirely.

There's a genuinely practical Mohs-relevant pearl buried in here on distinguishing perineural melanoma spread from normal perineurium on immunostains — normal perineurium stains EMA-positive, that's epithelial membrane antigen, and is negative for S-100 and SOX-10, while perineural melanoma extension shows the reverse pattern, S-100 and SOX-10 positive, EMA negative. Elston notes this staining reversal works well in Mohs sections and helps pathologists reliably separate pure desmoplastic tumors from mixed variants that carry the higher metastatic risk — which matters directly for how you interpret margins and counsel these patients postoperatively.

Since this is an editorial rather than a study, there's no results or limitations section to walk through — the piece is Elston's argument, built on cited literature, that AJCC staging in its current one-size-fits-all form doesn't capture this biological heterogeneity, and his closing point is essentially a call to action for staging reform rather than a new finding of his own.

So, three very different kinds of pieces this month. The AI ethics discussion is conceptually useful for how you frame conversations with app-driven patients, but it's not changing your biopsy threshold — treat it as a communication and documentation framework, not new evidence. The baker's rack pearl is immediately actionable if you're outfitting a suite and want to save real money without sacrificing function. And the desmoplastic melanoma letter is the one worth sitting with — it's not itself practice-changing since AJCC staging hasn't formally changed, but the pure-versus-mixed distinction, the caveat about Breslow thickness only tracking outcome in mixed tumors, and that EMA versus S-100/SOX-10 staining trick for perineural spread are all things you can fold into how you evaluate margins and counsel patients on desmoplastic melanoma today, ahead of any formal staging revision. That's the issue — thanks for listening, and I'll see you next month.