Welcome back to the JAAD Rounds review — this is the September 2026 issue of the Journal of the American Academy of Dermatology, and we've got four pieces to get through today: a large multinational study on nail unit squamous cell carcinoma, a SEER-based brief report on Mohs versus wide excision for cutaneous leiomyosarcoma, a letter reanalyzing sebaceous carcinoma survival outcomes with propensity matching, and a National Cancer Database study on vulvar dermatofibrosarcoma protuberans. Let's get into it.

First up is the nail unit squamous cell carcinoma paper, and this is a full original study — a multicenter, retrospective cohort spanning multiple countries, led by Lipner and colleagues. The clinical problem here is one we all recognize: nail squamous cell carcinoma is the most common nail malignancy, but it's persistently under-recognized because it mimics so many benign nail conditions — warts, chronic paronychia, onychomycosis. The specific gap they're addressing is that nobody had adequately characterized what distinguishes in-situ disease from invasive disease clinically, before you even get a biopsy back, and nobody had pooled enough cases to look at treatment outcomes with any real power, given how rare this tumor is institution by institution.

Methodologically, this is a retrospective, multicenter pooling of two hundred sixty-nine tumors in two hundred sixty-one patients — and a retrospective multicenter design is really the only feasible way to study a tumor this uncommon; you're not going to accrue a meaningful prospective cohort of nail squamous cell carcinoma at any single center in a reasonable timeframe, so pooling across more than a dozen international sites is the pragmatic tradeoff for sample size at the cost of standardization. They used t-tests and chi-square tests for univariate comparisons between in-situ and invasive disease, which is standard for comparing means and proportions, but the more interesting analytic choice is their use of multinomial logistic regression. Here's the concept: logistic regression normally handles a binary outcome — yes or no. Multinomial logistic regression extends that when you want to simultaneously estimate the odds of belonging to one category versus a reference category, adjusting for multiple predictors at once — in this case, estimating the odds of invasive disease versus in-situ disease, while holding age, smoking status, and morphology constant, so that a factor like subungual ooze isn't just correlated with invasion because it's also correlated with older age. This is the right tool here because their outcome — in situ versus invasive — is categorical, and they had several candidate clinical predictors they suspected were intertwined.

On results: the cohort was mean age sixty-one, seventy-two percent men, and the tumors clustered heavily on the right index and middle fingers and the thumbs bilaterally — about six in ten cases — which is a pattern the authors interpret as supporting a transmissible, likely human papillomavirus-related origin on the digits people use most, or perhaps related to cigarette holding. Koilocytosis, their histologic proxy for HPV, was present in about a third of tumors where data were available, confirmed by PCR in a subset, and notably ten of fifteen PCR-positive cases were HPV-16. Koilocytosis was significantly more common in in-situ disease than invasive disease. On the clinical side, subungual ooze independently predicted invasive disease with roughly a four-fold increase in odds — that's both statistically and clinically meaningful, since ooze is something you can actually see in clinic — while pain and loss of nail plate production tracked with invasion on univariate analysis but didn't hold up as independent predictors after adjustment. Treatment-wise, Mohs surgery was the most common approach at about forty percent, followed by local excision and en bloc nail unit excision, with amputation reserved for a small minority, mostly bone-invasive disease. Overall recurrence was under ten percent, and — this is the actionable point — there was no statistically significant difference in recurrence between Mohs, local excision, en bloc excision, or amputation.

The authors' honest limitations are the retrospective design across many centers with inevitable variability in workup and follow-up, and the fact that HPV status relied mostly on histologic koilocytosis rather than universal PCR confirmation, so the true HPV association is likely under-captured. Their conceptual takeaway is that periungual tumors with a warty morphology tend to be in-situ and HPV-associated, essentially behaving like a sexually transmitted, self-inoculated process, while subungual tumors presenting with ooze, pain, and nail plate loss behave like a distinct, more dermally invasive phenotype.

For practice, the practice-relevant takeaway is this: digit-sparing surgery, Mohs included, is validated as first-line therapy with recurrence rates that are comparable across modalities, so there's no oncologic argument for reflexive amputation in the absence of bone invasion. The morphologic pattern-recognition piece — subungual ooze as a red flag for invasion — is genuinely useful triage information you can use at the first visit, before pathology ever comes back. What's interesting but not yet practice-changing is the HPV angle; it raises the question of whether we should be counseling patients about partner screening or considering vaccination status, but the data aren't mature enough to generate a formal recommendation.

Next, a brief report from the SEER database looking at cutaneous leiomyosarcoma, comparing Mohs surgery to wide local excision on disease-specific survival. Quick framing: cutaneous leiomyosarcoma is rare, arises from dermal or subcutaneous smooth muscle, and wide excision has historically carried recurrence rates as high as thirty to fifty percent, while small case series and a prior meta-analysis suggested Mohs could bring that down into the single digits. What's been missing is survival data comparing the two approaches, since recurrence rate and disease-specific survival aren't the same thing.

This is a registry-based retrospective analysis using SEER 17, identified by ICD-oncology histology codes, comparing three hundred seventeen wide excision patients to seventy-two Mohs patients. The statistical workhorse here is the Cox proportional hazards model, and it's worth pausing on why. A Cox model doesn't ask "what percentage survived at five years" the way a simple Kaplan-Meier curve does — it compares the instantaneous risk, or hazard, of dying from the disease between two groups at every moment someone remains at risk, which lets you include patients who were followed for very different lengths of time without throwing away information, and lets you adjust for other variables simultaneously. The authors also mention checking Schoenfeld residuals — that's a standard diagnostic to confirm the model's core assumption, that the hazard ratio between groups stays constant over time, actually holds. It didn't hold for tumor location, so they stratified by primary site instead of estimating a single hazard ratio for it, which is the technically correct move rather than reporting a misleading pooled number.

Results: the Mohs cohort skewed toward head and neck tumors — traditionally a higher-risk location — yet on both univariate and multivariable Cox regression, there was no statistically significant difference in disease-specific survival between Mohs and wide excision. Age over sixty was the one factor independently associated with worse survival, roughly a three-fold increase in hazard, and unsurprisingly distant-stage disease at presentation carried a dramatically worse prognosis.

The authors frame this as reassuring for Mohs — that even with a higher-risk anatomic distribution in that cohort, survival was equivalent, which they argue supports Mohs as an effective option even at difficult sites. But they're upfront about real limitations: retrospective registry design, no granular surgical detail, small sample sizes limiting statistical power, and — importantly — SEER's surgery coding excludes subcutaneous tumors from this comparison, which are known to behave differently. There's also an unmeasured confounder the authors flag honestly: Mohs cases in a database like this likely cluster at academic or well-resourced centers, so you can't fully separate "modality" from "who was doing it and where."

Practically: this is interesting and supportive of the existing practice pattern — using Mohs for cutaneous leiomyosarcoma, particularly at cosmetically or functionally sensitive sites — but it's not practice-changing on its own, given the modest sample size and the equivalence finding rather than superiority. It adds a survival-outcome data point to a literature that previously only had recurrence-rate comparisons.

Third is a letter to the editor, and it's a direct methodological rebuttal to a previously published SEER study on sebaceous carcinoma. The original paper by Zhou and colleagues found no significant difference in five-year mortality between Mohs and wide excision. This letter's authors reran the SEER data themselves, but added propensity score matching, and reached the opposite conclusion — so this is worth slowing down on methodologically.

Propensity score matching is worth teaching because you'll see it constantly in registry-based surgical literature going forward. The problem it solves: in a retrospective database, treatment assignment isn't random — patients selected for Mohs might systematically differ from those selected for wide excision in ways that predict survival independent of the surgery itself, things like tumor size, comorbidity burden, or facility type. A simple multivariable Cox model tries to adjust for those variables, but if the two groups don't overlap well on those characteristics, that adjustment can be unstable. Propensity score matching instead calculates, for every patient, a predicted probability of having received Mohs based on their measured characteristics, then pairs Mohs patients with wide excision patients who have similar predicted probabilities — essentially constructing two groups that look statistically comparable on paper, closer to what you'd get from randomization, before comparing outcomes. The authors confirmed good balance after matching, with standardized differences under the usual threshold of point one.

After this matching, the picture flipped: wide excision was associated with significantly better overall survival and better cancer-specific survival, and this held up on multivariable Cox regression within the matched cohort, with a hazard ratio for cancer-specific survival suggesting roughly a two-thirds reduction in risk of dying from disease favoring wide excision. That's both statistically significant and, if it's real, clinically substantial.

But — and the letter's own authors say this — this is a letter, not a definitive study, and it carries the same fundamental SEER limitations as the paper it's critiquing: no data on margin status, no recurrence data, no detail on surgical technique or why a given patient was steered toward one modality over the other. Propensity score matching can only balance measured confounders; it can't correct for something like tumor aggressiveness or anatomic complexity that never made it into the database fields at all. The authors' own hypothesis — that wider margins in wide excision might better eradicate subclinical extension in sebaceous carcinoma — is speculation, not something the data can directly test.

The practical takeaway here is genuinely "interesting, not yet actionable, and actually somewhat unsettling." This letter doesn't overturn the equipoise around Mohs versus wide excision for sebaceous carcinoma — it reopens the question that a prior paper had seemingly closed, using a technique that specifically targets the confounding problem inherent to registry comparisons. If you're counseling a patient with sebaceous carcinoma right now, this shouldn't change your practice, but it should make you appropriately humble about citing "no survival difference" as settled fact, and it flags that this is a space needing prospective or at least better-controlled registry work.

Last is a brief report on vulvar dermatofibrosarcoma protuberans using the National Cancer Database, covering 2004 through 2023. The rationale is simple: vulvar DFSP is essentially only reported as isolated cases, so anything resembling a cohort is valuable, and the question is whether the same margin-and-recurrence principles that guide cutaneous DFSP elsewhere on the body apply at this anatomically constrained, function-and-cosmesis-sensitive site.

Two hundred sixty-two patients were identified, median age forty-seven, median tumor size three centimeters, and reassuringly, distant metastasis was essentially absent — one patient in the whole cohort. Surgery was performed in virtually everyone, split between partial vulvectomy, total vulvectomy, and organ-sparing approaches like local excision. The analytic tools were multivariable logistic regression and modified Poisson regression to identify predictors of positive margins, plus Cox regression and Kaplan-Meier estimates for survival.

It's worth teaching the modified Poisson regression choice here, because it's a subtler statistical point than we usually cover. Logistic regression gives you an odds ratio, which is a fine metric mathematically but gets misinterpreted constantly as if it were a risk ratio — and when an outcome is fairly common, as positive margins were here, odds ratios can meaningfully overstate the actual relative risk. Modified Poisson regression with robust variance estimation is a workaround that directly estimates a risk ratio — the straightforward relative probability of the outcome between groups — while still handling a binary outcome correctly, which is why the authors ran both models in parallel; you'll notice the risk ratios in their table are systematically smaller than the paired odds ratios, which is exactly what you'd expect and is a nice illustration of that odds-versus-risk distinction in real data.

The results: organ-sparing surgery carried roughly double to triple the odds, and about a two-fold relative risk, of a positive margin compared with partial vulvectomy. More recent year of diagnosis and treatment at a nonacademic facility were also independently associated with positive margins. Despite that, five-year overall survival across the entire cohort was excellent — around ninety-seven percent — and critically, surgical approach was not significantly associated with overall survival on either Kaplan-Meier or multivariable Cox analysis. Organ-sparing surgery, partial vulvectomy, and total vulvectomy all produced statistically indistinguishable survival curves.

The authors' honest limitation, and it's the big one: the National Cancer Database doesn't capture local recurrence, doesn't capture Mohs micrographic surgery as a distinct variable, doesn't capture fibrosarcomatous transformation, and doesn't capture cause-specific mortality separate from margin status. So while margin positivity didn't translate into a measurable survival hit in this dataset, the authors themselves note that's likely because overall survival is simply the wrong endpoint to detect the clinical consequence of a positive margin in a slow, locally recurrent tumor like DFSP — recurrence is the outcome that actually matters here, and it's invisible to this database.

Practically, this is not practice-changing, and the authors are clear-eyed about that themselves — they explicitly call for prospective evaluation of Mohs surgery in this population rather than drawing conclusions from what they have. What it does usefully establish is the epidemiologic backdrop: vulvar DFSP has an excellent overall prognosis, metastasis is rare, and organ-sparing approaches carry a real, measurable margin-positivity penalty relative to more extensive vulvectomy — which is exactly the kind of finding that builds the rationale for a margin-controlled technique like Mohs at this site, even though this particular dataset can't test that hypothesis directly.

That wraps our four articles for this September issue — a nail squamous cell carcinoma cohort reinforcing digit-sparing surgery and HPV biology, a reassuring survival equivalence for Mohs in cutaneous leiomyosarcoma, a propensity-matched challenge to the sebaceous carcinoma equipoise that deserves close attention rather than a verdict, and a vulvar DFSP registry study that quietly makes the case for margin-controlled surgery without being able to prove it. Thanks for listening, and we'll see you next month.