Welcome back to this month's journal review — November twenty twenty-six issue of the Journal of the American Academy of Dermatology. We've got four pieces for you this time, and interestingly two of them are a connected pair of letters on sunscreen regulation that bookend each other nicely, plus a brief report on melanoma genomics and an ethics column on artificial intelligence in the electronic health record. Let's get into it. First up is a letter to the editor — correspondence responding to the Abdel Azim et al CME piece on sunscreen regulation and safety. This is a cross-sectional comparison, not a clinical trial, so think of it as a regulatory audit rather than a patient-outcomes study. The authors' starting premise is one most of you already know informally: the FDA treats sunscreens as over-the-counter drugs, which means a slow, monograph-based approval pathway, while the European Union regulates them as cosmetics, which allows much faster adoption of new filters. The numbers bear this out — the EU has approved twenty-nine ultraviolet filters against just sixteen in the US. The novel contribution here is Korea. The authors did a structured Google search exercise, pulling the top twenty sunscreens marketed in the US and in Korea, and cross-referenced every active filter against FDA and EU approval databases. Methodologically, this is about as simple as it gets — two independent searches on separate machines, converging on identical product lists, then ingredient lookups via manufacturer sites and a public cosmetic-ingredient database. There's no statistical test here, just a straightforward tabulation, so no stats lesson needed for this one. What they found: the top US sunscreens cluster around familiar FDA-approved filters — butyl methoxydibenzoylmethane, octocrylene, zinc oxide, titanium dioxide. The Korean top twenty, by contrast, are dominated by bemotrizinol, bisoctrizole, and diethylamino hydroxybenzoyl hexyl benzoate — all EU-approved, none FDA-approved at the time of writing. And critically, those Korean products aren't confined to Korea. They're sitting on Amazon and YesStyle, fully accessible to your US patients right now. The authors trace the regulatory lag back to the 1999 sunscreen monograph, note that the 2020 CARES Act tried to modernize the approval pathway but progress has been slow, and conclude with a fairly modest ask: FDA should prioritize evaluating internationally-approved filters, and dermatologists should be ready to counsel patients navigating these imported products. Practical takeaway here: this is mostly background awareness rather than a change to how you practice. Know that patients may be using imported sunscreens with filters you won't find in the US monograph, be ready to reassure them these filters have long safety track records abroad, and recognize that broader UVA coverage — which many of these EU filters offer over our avobenzone-dependent formulations — may actually be a clinical plus for your photosensitive and pigmentary-disorder patients. Which brings us directly to article two, a response letter that updates the story in real time — this is the fun part of reading journals in sequence. Brandon Adler writes in to report that between the original CME article and this responding correspondence, the regulatory landscape actually moved. In December twenty twenty-five, the FDA issued a proposed order to add bemotrizinol — also known as BEMT — to the sunscreen monograph, and by the time of this letter, June ninth twenty twenty-six, that action was finalized. This is the first new sunscreen active ingredient approved in the US since nineteen ninety-nine. Clinically, why does this matter? BEMT offers true broad-spectrum UVA and UVB coverage including long-wave UVA, it's photostable — unlike avobenzone, which has been our only broad-spectrum organic filter and which degrades in sunlight unless stabilized — and it plays well in combination with other approved filters. It also cleared FDA's bar for minimal systemic absorption, earning it GRASE status, generally recognized as safe and effective. The sobering footnote: even with two decades of international use behind it, getting BEMT through the US process took over twenty years of manufacturer effort and something on the order of twenty million dollars. The author flags a genuinely practice-relevant development here too — the SAFE Sunscreen Standards Act, signed into law in late twenty twenty-five, which directs FDA to allow real-world evidence, including observational data, to support safety and efficacy going forward. That's a structural change that could meaningfully speed up future filter approvals. Practical takeaway for you: this one actually is practice-changing, in a small but real way — you now have a legitimate, US-approved rationale to look for bemotrizinol-containing formulations as they reach the US market, particularly for patients who need superior photostable UVA coverage, and you can tell patients currently importing Korean sunscreens for that specific filter that domestic options should be coming. Now to a genuine original study — a brief report on driver mutations, tumor mutational burden, and survival in stage four melanoma. The background here is conceptually elegant: the same oncogenic pathways that drive melanocytic proliferation produce very different biological behavior depending on which gene is hit — GNAQ and GNA11 driving blue and congenital nevi, NRAS behind common acquired nevi, BRAF in conventional melanoma, while KIT and NF1 alterations are more classically associated with mastocytoma and neurofibroma respectively. The authors' question is whether these divergent growth biologies translate into divergent survival once a patient has progressed to stage four disease. This was a retrospective cohort using standardized hundred-seventy-gene sequencing on patients collected over about five years, ultimately yielding three hundred eighty-nine evaluable patients after excluding incomplete records and non-melanoma deaths. They stratified tumors by driver mutation — GNAQ or GNA11, NRAS, BRAF, KIT, NF1, or wild type — and layered on tumor mutational burden as a continuous variable. The primary analytic tool was Kaplan-Meier survival estimation compared across these mutation groups, followed by Cox proportional hazards regression to adjust for tumor mutational burden, immunotherapy use, and targeted therapy use. Worth pausing on why Cox regression is the right tool here: you have patients followed for different lengths of time, some who died, some who were still alive at data lock, and you want to know whether mutation group predicts the instantaneous risk of death at any given moment while adjusting for other variables simultaneously — ordinary comparison of medians can't do that adjustment, and a Cox model handles it by comparing hazard, the instantaneous event rate among those still at risk, between groups across the entire follow-up period rather than just at one fixed time point. That's why it's the standard here rather than, say, a simple t-test on survival times. The results are the clinically interesting part. Median survival from stage four diagnosis to death was about four years overall, but this varied enormously by driver gene. NF1-mutated tumors had the longest median survival, over six and a half years, followed by BRAF at about four and three-quarter years. NRAS and wild-type tumors sat in the middle, both a little over four years. GNAQ and GNA11-mutated tumors fared worst, with median survival of only about fifteen months, and KIT-mutated tumors landed in between at around two and a half years. This separation was statistically significant and held up even when the analysis was restricted to five- and ten-year follow-up windows, and remained significant in the Cox model after adjusting for tumor mutational burden, immunotherapy, and targeted therapy — so this isn't just an artifact of who got which treatment. Tumor mutational burden itself tracked with mutation type: NF1 tumors ran the hottest, with the highest mutational burden by a wide margin, while GNAQ and GNA11 tumors were the coldest. And higher tumor mutational burden independently correlated with better overall survival. Interestingly, immunotherapy use didn't correlate with tumor mutational burden itself, though it did vary somewhat by mutation group, with BRAF-mutated patients being slightly less likely to receive immunotherapy — intuitive, given targeted therapy is often the first-line alternative there. The authors' own limitation, stated plainly, is that survival associations here will shift as therapies evolve — this is a snapshot of outcomes under roughly twenty eighteen to twenty twenty-two treatment paradigms, and I'd add on top of that the usual retrospective-registry caveats: single or limited-institution sequencing cohort, potential selection bias toward patients who got comprehensive genomic profiling in the first place, and the GNAQ and GNA11 group was quite small, under thirty patients, so that dramatic survival difference deserves some caution before it's treated as gospel. Practical takeaway: this isn't practice-changing in the sense of altering your surgical management, but it's a genuinely useful prognostic framework to carry into tumor board discussions — when you see a GNAQ or GNA11-mutated stage four melanoma, recognize that historically those patients have had a more aggressive course independent of treatment received, which should inform urgency of referral and monitoring intensity, while NF1 and BRAF-mutated patients, particularly V600E, have historically had more favorable trajectories. Last piece this month is from the Ethics Journal Club, so no methods or results to walk through — this is a structured ethical opinion piece in advice-column format, addressing a dermatologist's question about adopting AI tools embedded in the electronic health record. The question posed is whether it's ethical to adopt AI-powered documentation assistants, inbox triage, prior authorization support, and diagnostic image analysis tools, given rising healthcare data breaches and dependence on third-party vendors. The response walks through this using a classic four-principles framework. On beneficence and justice, the argument is that efficiency gains from AI could free up clinician time for direct patient care and improve access. But the piece quickly pivots to the harder issues. Dermatology data are uniquely sensitive — facial photographs, pediatric images, genital or stigmatizing-condition photographs — and once AI tools are embedded in the EHR, that data can be accessed passively and continuously, raising real risk of unauthorized access, breach, or secondary commercial use of de-identified datasets. The autonomy concern is subtle but important: unlike a discrete clinical intervention a patient consents to, EHR-embedded AI often operates invisibly, and existing informed consent frameworks weren't built for this kind of continuous, passive data processing. The piece calls for explicit disclosure of when AI contributes to care and transparency about data handling, while candidly noting that doing this properly costs time, potentially eating into the very efficiency gains that motivated adoption in the first place. On nonmaleficence, the reminder is unambiguous — AI can generate documentation errors or misleading prior authorization content, and the clinician remains responsible for verifying every output before it touches patient care. And on justice, the piece flags a real disparity risk: large health systems have the resources to vet vendors and audit AI performance, while smaller practices are often stuck taking vendor claims on faith, which is where the authors suggest organizations like the American Academy of Dermatology should step in with standardized guidance. There's no data and no limitations section here because it's an opinion piece by design, so the practical takeaway is really about institutional diligence rather than clinical management: before adopting any EHR-embedded AI tool, push your practice or health system to get clarity in the vendor contract on data ownership, permitted secondary uses, and bias or performance validation, and treat every AI output — especially anything touching diagnosis or documentation — as a draft requiring your review, not a final answer. That wraps up this month's four pieces — a regulatory letter-and-response pair on sunscreen filters that's worth filing away for patient conversations about imported products, a melanoma genomics brief report that sharpens your prognostic intuition without changing surgical management, and an ethics column that's a good prompt to ask your own practice some pointed questions before the next EHR AI vendor demo. Thanks for listening, and we'll see you next month.