Welcome back to the journal review. This month we're in the June twenty twenty-six issue of the Journal of the American Academy of Dermatology, and we've got four pieces from the Notes and Comments and Brief Reports sections — two of them are actually a paired exchange of letters on the same paper, so we'll take those together, then a letter on artificial intelligence and patient education, and finally a brief report on pediatric dermatofibrosarcoma protuberans. Let's get into it.

First up is a correspondence exchange — a letter to the editor and the original authors' response — both centered on a study you may already know: Akaike and colleagues' multicenter paper showing that circulating tumor DNA, or ctDNA, correlates with time to clinical recurrence in Merkel cell carcinoma. The letter comes from Gao and Chen, and it's a thoughtful methodologic critique rather than a new dataset, so think of this as expert commentary sharpening the interpretation of results you may be using clinically already.

Their central point is that ctDNA behaves as a continuous, log-linear biomarker rather than something you can collapse into a simple positive-or-negative call. The original study reported an adjusted hazard ratio of about two and a half for every tenfold rise in ctDNA — so each order-of-magnitude jump roughly doubles to triples recurrence risk. But Gao and Chen highlight something clinically sharper than the hazard ratio itself: once ctDNA turned detectable at all, about seven in ten patients recurred within a year, regardless of how high that initial level was. In other words, positivity itself seems to mark a biological switch from occult to active disease, and the exact titer beyond that point matters less than you'd think.

They also flag the kinetics. Median ctDNA levels rose roughly nineteen-fold between first positivity and clinical recurrence, with an estimated doubling time of only about three weeks. That's fast, and the practical implication they draw is that a surveillance interval longer than about a month after a positive ctDNA result risks missing meaningful progression.

On sensitivity, they raise a fair point: four biopsy-confirmed local or in-transit recurrences occurred while ctDNA stayed negative, meaning small or dermal-only lesions can shed too little tumor DNA into plasma to be caught. Their takeaway is that physical exam and imaging still matter even when the liquid biopsy is reassuring, and that combining ctDNA with viral oncoprotein antibody titers might sharpen sensitivity further, though antibody status alone didn't correlate with recurrence in this cohort.

Methodologically, they push back a bit harder — arguing that the study's own design may have inflated the apparent lead-time benefit of ctDNA, since a positive result triggered imaging within about four weeks, and since testing intervals weren't uniform, with a median around three months but ranging from roughly two to three and a half months. That kind of informative censoring, they argue, could make ctDNA look like it's buying more lead time than it truly is. They also note the ctDNA cutoffs used — under one, one to ten, and over ten molecules per milliliter — were descriptive categories, not statistically validated thresholds.

Now, the authors' response, from Akaike, Zaba, Nghiem and colleagues, engages point by point and is worth hearing because it clarifies how they actually intend this to be used. They agree ctDNA is fundamentally continuous — they note a strong correlation between ctDNA level and tumor size in prior work — and they explain the three discrete strata weren't meant to replace that continuous relationship, just to translate it into something a clinician can act on at the bedside, acknowledging that the thresholds themselves need further refinement.

Where they push back is on surveillance cadence. They're hesitant to endorse a blanket monthly retesting rule after any positive result, because among patients with low-positive ctDNA, roughly a quarter hadn't recurred by twelve months — so an indiscriminate monthly interval for everyone positive may be overkill, and their study wasn't actually designed to tell us the right interval after a positive result kicks in. That's flagged as an open research question, not a solved one.

On the false-negative recurrences, they clarify: four of fifty total recurrences were ctDNA-negative, two of which were only caught on physical exam and missed even by standard imaging, likely because of small lesion size and the technical limitation of PET-CT contrast in skin. Interestingly, three of the four were in patients who don't produce detectable viral antibodies at all — a reminder that only about half of Merkel cell patients are antibody producers to begin with, so pairing ctDNA with antibody testing won't help everyone. Their bottom-line message is that ctDNA negativity, combined with a physical exam, catches nearly everything imaging would catch — which is actually an argument for cautiously de-escalating imaging frequency in ctDNA-negative patients, not escalating it. And on the lead-time critique, they simply disagree, arguing that rapid imaging follow-up after a positive result would if anything shorten the apparent lead time, not inflate it.

For your practice, the practical signal from this whole exchange is: treat any ctDNA positivity as a meaningful event that warrants tightened follow-up, know that the kinetics can move fast, but don't take this as a mandate for automatic monthly retesting on every positive patient — and don't fully substitute ctDNA for your eyes and hands on exam, since small locoregional recurrences can still slip through molecularly. This is refinement of an existing biomarker's interpretation, not yet a new algorithm — genuinely useful context, not yet practice-changing on its own.

Next, a letter to the editor addressing a prior JAAD paper on AI-based Spanish translation of Mohs surgical instructions. This is a commentary piece laying out guiding principles for using AI to draft and translate patient education materials, and it's squarely relevant if your practice generates any handout, portal message, or post-op instruction sheet.

The core principle is simple: AI can speed up drafting, but a certified interpreter needs to verify accuracy and intent before anything goes out the door — that's non-negotiable, since unvetted AI output risks misinformation. They also bring in readability science. The American Medical Association recommends patient materials sit around a sixth-grade reading level, and the authors score the three institutional Mohs handouts referenced in the original paper using the Flesch-Kincaid scale, which runs zero to one hundred, higher meaning easier to read. Brigham and Women's material scored about 34, which lands at college reading level; University of Mississippi's scored 58, closer to ninth or tenth grade; and Rochester's scored 37, again college level. Their point is that if your source English document is already written above a sixth-grade level, the AI-translated output is likely to inherit that same complexity in the target language — so fixing readability has to happen before translation, not after.

They also flag a real equity issue: AI translation accuracy drops for languages that aren't Roman-alphabet-based or that carry complex formality structures — they specifically mention Korean, lower-resourced languages like Swahili or Bengali, and right-to-left languages like Arabic — and they recommend extra scrutiny of AI output for these groups. Finally, they raise the idea of simply asking patients how they'd prefer to receive information — written versus audio, for instance — as a low-friction way to surface literacy or format needs without singling anyone out, and they point toward emerging tools like avatar-based multilingual video generation as a future direction.

There's no original data here, no methods or results — this is pure guidance and commentary. The actionable piece for your practice is concrete, though: before you hand any AI-translated instruction sheet to a patient, run the source text through a readability check first, have a certified interpreter verify the translation, and build in a low-key way to ask patients about their preferred format.

Last, a brief report — a single-institution retrospective chart review of pediatric dermatofibrosarcoma protuberans, or DFSP, out of Cleveland Clinic. This is a genuinely rare-tumor case series, and the framing is straightforward: DFSP is fundamentally an adult disease, pediatric and congenital cases are scarce enough that only around seventy have ever been published, so any single-center experience adds meaningfully to the literature.

Methodologically, this is a retrospective chart review spanning 2008 to 2024, yielding eleven pediatric cases, four of them congenital. A retrospective design here makes sense for the obvious reason the authors imply — with a tumor this rare, you simply can't run a prospective trial; you have to accumulate experience over a long institutional window and describe it. Descriptive statistics only, no inferential testing, which is appropriate given the tiny sample.

On findings — mean age at diagnosis was about twelve years, younger than the roughly fourteen years reported in prior literature, and congenital cases specifically averaged under eleven years. Most tumors, over ninety percent, were first noticed by the patient themselves rather than found on exam or imaging, which reinforces how easily these lesions get mistaken early on for keloids, hypertrophic scars, cysts, or lipomas. All eleven cases were treated with surgery alone, no adjuvant therapy — split roughly evenly between Mohs micrographic surgery, in about five cases, and wide local excision with at least two-centimeter margins down to fascia in the remainder. Seven of the eleven underwent preoperative MRI, which the authors use specifically to map tumor extent and decide whether a multidisciplinary team is needed — in one striking case, MRI suggested possible cranial bone and dural invasion, prompting a coordinated resection with neurosurgery and plastic surgery, though final pathology showed the periosteum and outer table weren't actually involved. Imaging changed the surgical plan in one of the eleven cases.

The headline outcome: zero recurrences across the entire cohort, including every Mohs case, over an average follow-up of about four years, ranging from under two months up to fifteen years. That's a clean result supporting what NCCN guidelines already say — Mohs is the preferred modality for DFSP, with wide excision and complete margin assessment as the fallback when Mohs isn't available or feasible in a child.

The authors don't spell out a lengthy limitations section, but it's implicit and worth naming plainly for you: this is eleven patients at one center, so recurrence rates near zero should be read as reassuring rather than definitive, and subtype- or variant-specific conclusions simply aren't possible at this sample size. They do explicitly call for more data on the role of imaging in postoperative surveillance, which remains an open question.

For your practice, the actionable takeaways are: keep DFSP on the differential for a pediatric lesion masquerading as a keloid or cyst, especially since patients themselves are the ones typically noticing these tumors first; use preoperative MRI liberally in pediatric cases to gauge extent and to flag when you need a multidisciplinary team; and this data adds further reassurance, on top of existing guidelines, that Mohs is a safe and effective first-line approach even in children, with a proposed surveillance schedule of exams every six months for two years and then annually after that. It's not a change in the standard of care — it's confirmatory, real-world support for the approach most of us are already using.

That wraps our four pieces for June. The ctDNA correspondence is a great example of how post-publication dialogue refines biomarker interpretation without yet changing the algorithm, the AI translation letter gives you a concrete quality-control checklist for patient materials, and the pediatric DFSP series reinforces Mohs as the preferred approach in a setting where level-one evidence will likely never exist. Thanks for listening, and we'll see you next month.