Welcome back to the monthly deep dive into the Journal of the American Academy of Dermatology — this is the November twenty twenty-six issue, and we've got four brief reports worth your time today: a SEER registry analysis on radiation use in Merkel cell carcinoma, a Canadian cohort refining skin cancer risk prediction in thoracic transplant recipients, a single-center descriptive series on malignant transformation in mucosal lichen planus, and a retrospective look at dosing intervals for intralesional five-fluorouracil in low-risk squamous cell carcinoma. Let's get into it. First up is a retrospective registry study: adjuvant radiation therapy trends and predictors in Merkel cell carcinoma, using twenty years of SEER data — that's the Surveillance, Epidemiology, and End Results database. The clinical problem here is one you already live with — Merkel cell carcinoma is an aggressive neuroendocrine skin cancer, a third to half of patients recur or metastasize, and adjuvant radiation has reasonably good evidence behind it for improving disease control and survival. What the authors wanted to know is simple: over two decades, has adoption of adjuvant radiation actually kept pace with the evidence, and who's getting left out. Methodologically, this is a classic SEER pull — nearly nine thousand surgically treated Merkel cell patients from two thousand to twenty-twenty, radiation treatment defined from the registry's treatment-sequence variables. They used logistic regression with calendar year as a continuous predictor to test for a temporal trend, and then multivariable logistic regression adjusting for age, race, sex, income, urban residence, and stage to find independent predictors. Worth pausing on why logistic regression is the right tool here: the outcome — received radiation or not — is binary, so you need a model that estimates the odds of that binary event rather than a continuous outcome, and by putting year in as a continuous term rather than binning it into eras, they can ask a clean question — for every additional year, how much do the odds of receiving radiation shift — which is more statistically efficient than chopping time into arbitrary chunks and comparing them. The results: just under half of patients, about forty-five percent, received adjuvant radiation overall, and the rate did rise, going from roughly forty percent in the earliest years up to about forty-eight percent by the most recent era. That trend was statistically significant, but look at the absolute movement — eight percentage points over twenty years — that is a real increase, but it's a slow one, and the authors are blunt that this has not kept pace with the evidence. When they stratified by stage, the entire trend was being driven by regional-stage disease; localized disease — which is most of what you're managing — showed no significant change over two decades, and neither did distant disease, where the explanation is almost certainly the rise of systemic immunotherapy, since avelumab and pembrolizumab use nearly doubled over the same window. On predictors, age eighty and older was associated with roughly a two-thirds lower likelihood of getting radiation, male sex and regional disease both predicted higher radiation use, and income quartile and urban residence didn't matter. Within localized disease specifically — where guidelines matter most to you at the surgical margin — tumor size over two centimeters, head and neck primary site, and any nodal evaluation were independent predictors of radiation, but age eighty-plus patients were again significantly less likely to receive it. And here's the number I'd actually flag for practice: among localized patients with a head and neck primary — which NCCN explicitly lists as a risk factor pushing toward adjuvant radiation — only about forty percent actually received it. Importantly, refusal was rare, under three percent, so this isn't patients declining treatment; it's a systems and referral problem. Limitations are the usual SEER caveats — no granular data on comorbidities, immunosuppression status, or which specialty was actually managing the patient, all of which plausibly explain undertreatment in the elderly and in localized disease, and the dataset stops in twenty-twenty, so it may already be stale relative to current referral patterns. Practically, this is actionable for you right now, not just interesting: if you're managing a localized Merkel cell carcinoma with a head-and-neck primary, especially in an older patient, this data says there's a real and measurable gap between what guidelines recommend and what's happening — and since refusal isn't the driver, that's a referral-pathway and radiation-oncology-engagement problem you can influence directly in your own practice. Second article, and this one's for anyone running a transplant dermatology clinic: evaluating additional risk factors with the SUNTRAC calculator — that's the Skin and Ultraviolet Neoplasia Transplant Risk Assessment Calculator — in Canadian heart and lung transplant recipients. Background here: SUNTRAC was built to guide screening intensity in organ transplant recipients using age, sex, race, transplant type, and pretransplant skin cancer history, but it doesn't capture things like smoking or actinic keratosis burden or specific immunosuppressant exposure. This retrospective cohort of two hundred ninety-six thoracic transplant recipients from a single Edmonton center asked whether adding those factors actually sharpens risk stratification beyond SUNTRAC alone. The statistical workhorse here is a Fine and Gray competing-risks model, and this is worth teaching properly because it's exactly the right tool for this data structure and it's one you'll see again. The problem with a standard survival model in a transplant cohort is that patients can die before ever developing skin cancer, and death removes them from being able to have the event you're studying — that's a competing risk. If you just censor the deaths and run an ordinary Cox model, you'll actually overestimate the cumulative incidence of skin cancer, because the model pretends those patients were still "at risk" for cancer when in fact they were removed from the risk pool by a different, competing outcome. The Fine-Gray model instead directly models the cumulative incidence in the presence of that competing event, producing what's called a subdistribution hazard ratio — conceptually similar to an ordinary hazard ratio, but correctly accounting for the fact that death from other causes is pulling people out of the population who could have developed skin cancer. Results: about seventy-one percent of the cohort was male, median age fifty-six, the vast majority Fitzpatrick type one through three. Nearly a third of patients — ninety-three of them — developed skin cancer, split between squamous cell carcinoma at about nineteen percent, basal cell at about twelve percent, plus a couple of melanomas and one eccrine carcinoma. On multivariable analysis, age fifty or older, Caucasian race, smoking history, and pretransplant actinic keratoses were all independently associated with skin cancer risk, with pretransplant actinic keratoses carrying the strongest signal — roughly a three-and-a-half-fold increase in subdistribution hazard. Pretransplant skin cancer history itself was not significant, but that's almost certainly an underpowered result — only five patients in the whole cohort had that history. In secondary analyses, azathioprine exposure was associated with higher skin cancer rates, including, notably, in patients with darker skin types, though that subgroup was small and excluded from the core multivariable model, so treat it as hypothesis-generating. Voriconazole showed no signal. The piece I'd call clinically meaningful: when they stratified patients into risk tertiles using SUNTRAC alone versus SUNTRAC plus smoking and actinic keratoses, the augmented model separated risk groups more sharply. In the highest-risk tertile, ten-year cumulative incidence of skin cancer went from about forty-two percent with SUNTRAC alone up to roughly fifty-eight percent with the augmented model — meaning SUNTRAC alone was actually under-capturing just how high-risk that top tier really was. And at the other end, the lowest-risk tertile had a lower observed incidence with the augmented model than with SUNTRAC alone, meaning the richer model was also better at correctly reassuring your truly low-risk patients. The authors contrast this with prior attempts to improve SUNTRAC using things like BMI or HLA mismatch, which added little — the message is that simple, already-in-the-chart variables like smoking history and actinic keratosis burden may do more for risk stratification than fancier biologic variables. Limitations are what you'd expect: single center, retrospective, small subgroups for some of the secondary signals, and likely underpowered analyses for male sex and pretransplant cancer history. Practically, this isn't yet a validated replacement for SUNTRAC, but it is a reasonable argument for weighting smoking history and baseline actinic keratosis burden more heavily when you're deciding how aggressively to screen a given thoracic transplant recipient, and it's a nice confirmation that azathioprine-based regimens deserve extra vigilance, even outside the classically fair-skinned population. Third article is a descriptive cohort analysis, not a comparative trial, so it's a slightly different shape — malignant transformation in mucosal lichen planus, a ten-year single-center series. The clinical question is one that comes up constantly in tumor boards and consults: how often does mucosal lichen planus actually turn into squamous cell carcinoma, and does systemic immunomodulatory therapy change that risk one way or the other. Reported transformation rates in the literature are all over the map, from a fraction of a percent up to six percent, largely because of inconsistent diagnostic criteria and inclusion of lichenoid drug reactions. This was a single dermatologist's cohort at the University of Pennsylvania, a hundred forty-nine patients with oral, vulvovaginal, or esophageal lichen planus followed over about a decade, median follow-up just under four years, ninety-two percent biopsy-confirmed, predominantly older white women. Malignant transformation was defined strictly as biopsy-proven squamous cell carcinoma at a previously diagnosed lichen planus site occurring at least six months after diagnosis, with patients who already had atypia on their initial biopsy excluded — that's an important design choice, because it keeps this from just capturing missed dysplasia at baseline and instead isolates genuine progression over time. Because the number of malignant events is small, the authors used the Clopper-Pearson exact method to calculate confidence intervals rather than the usual normal-approximation method — that's worth knowing, because with small samples and rare events, the standard approximate formula can behave badly, even producing nonsensical bounds, whereas the exact method is built directly from the binomial distribution and stays accurate no matter how few events you have. That's exactly the right choice for a four-event outcome in a hundred thirty-five patients. The headline number, and this is one worth quoting to a patient: among a hundred thirty-five oral lichen planus patients, four developed oral squamous cell carcinoma, a rate of about three percent, with a confidence interval roughly between one and seven percent. Every single one of those four cases arose in patients with active, erosive disease — none in the reticular or purely erythematous subtypes — and the average time from lichen planus diagnosis to cancer was about three and a half years. In the esophageal lichen planus group, one of twenty-eight patients developed esophageal squamous cell carcinoma, about three and a half percent. Nobody in the vulvovaginal group developed cancer. And reassuringly, excluding anyone with less than six months of follow-up barely moved the oral number at all. The therapy comparison is the part that will generate discussion: zero malignancies occurred among the seventy-three oral lichen planus patients on systemic immunomodulatory therapy — methotrexate, mycophenolate, JAK inhibitors — versus four malignancies among the sixty-two patients managed with topical therapy alone, a rate of about six and a half percent, and that difference was statistically significant by Fisher's exact test. But I want to flag the authors' own caution here, and add my own: all four of the patients who developed cancer had persistent, poorly-controlled erosive disease despite topical treatment — so this isn't necessarily "systemic therapy prevents cancer," it may just as easily be "patients with refractory erosive disease who never got escalated to systemic therapy were the ones who progressed." With only four events total, you cannot separate a protective drug effect from simple selection — these were the hardest-to-control patients by definition, and this descriptive signal should not be read as proof that immunomodulators prevent transformation. Bottom line for practice: this supports what many of you probably already assume — persistent erosive oral lichen planus is the subtype to watch closely, transformation can take several years to manifest so long-term follow-up matters, and inadequately controlled disease on topicals alone may be a flag for escalation, though not because this study proves systemic therapy is protective. It's a genuinely useful descriptive data point to add to the literature, not yet a practice-changing treatment claim. Fourth and last article is a retrospective cohort study on treatment interval and efficacy of intralesional five-fluorouracil — IL5FU — for low-risk de novo squamous cell carcinoma and squamous cell carcinoma in situ. The background gap is narrow but practical: intralesional 5FU is a well-established surgical alternative for low-risk keratinocyte carcinoma, with reported resolution rates in the literature between seventy-five and ninety-one percent, but almost all of that prior data used treatment intervals under four weeks, so nobody really knew whether spacing injections out further helps, hurts, or doesn't matter. This was a single-institution retrospective review spanning October twenty-seventeen through April twenty twenty-five, two hundred eleven tumors across forty-eight mostly elderly, mostly female patients, median age seventy-eight. The lesion mix included squamous cell carcinoma in situ, well-differentiated invasive squamous cell carcinoma, and a chunk of purely clinically-diagnosed lesions without biopsy confirmation — that's worth remembering when you get to the limitations. They compared clinical resolution rates at follow-up for tumors treated at intervals under four weeks versus four weeks or longer, using chi-square tests for the categorical comparison and logistic regression with the treatment interval modeled as a continuous predictor as a complementary analysis. That combination is a nice teaching point on study design: dichotomizing at four weeks lets you report a clean, clinically interpretable comparison — this group versus that group — but it throws away information, since a six-week interval and a twelve-week interval both just get lumped into "four weeks or more." Running treatment interval as a continuous predictor in the logistic model lets them ask a more granular question — for every additional week of spacing, how much do the odds of resolution change — which tests for a genuine dose-response relationship rather than just a single cutpoint. When both the categorical comparison and the continuous dose-response signal point the same direction, that's reasonably reassuring that the finding is real and not an artifact of where they happened to draw the four-week line. Results: after a first injection, just over half of tumors — about fifty-one percent — resolved. Of the tumors that didn't resolve, most went on to a second injection, and just over half of those resolved by the next visit, bringing cumulative resolution after up to two treatments to about seventy-three percent, which lines up with the upper end of the previously reported range. Among tumors that did resolve, the resolution held for a median of about a year and a half of follow-up, and twelve tumors recurred, with recurrence showing up at a median of about six months. The key finding: at the first follow-up, resolution was significantly higher with intervals of four weeks or more compared to shorter intervals — fifty-seven percent versus thirty-six percent — and the continuous analysis confirmed a real dose-response, with each additional week of spacing meaningfully increasing the odds of resolution. The same pattern held at the second follow-up point after repeat treatment — sixty-four percent versus thirty-eight percent, again with a significant per-week trend. Adverse events were reported in twenty-three patients, mostly dermatitis, lower extremity edema, and post-inflammatory hyperpigmentation — nothing that changes the risk-benefit calculus substantially. Now, the honest limitation I want to add beyond what the authors state explicitly: because five-fluorouracil inhibits DNA synthesis, a treated lesion can keep visibly regressing for weeks after the injection — so there's a plausible alternative explanation for this finding that the paper gestures at but doesn't fully spell out. It's possible that longer intervals aren't causing better resolution so much as simply allowing more time for a lesion that was already going to resolve to actually finish resolving before you check on it. In other words, part of this signal may be about assessment timing rather than treatment spacing changing the biology — that's my interpretation, not the authors' explicit claim, but it's worth holding in mind before you change your injection schedule based on this alone. Other standard limitations apply too: retrospective, single institution, and no histologic confirmation of resolution — this is all clinical assessment. Practically, this is useful but I'd call it "interesting and reasonably actionable" rather than definitively practice-changing: if you're using intralesional 5FU for appropriate low-risk lesions, there's now reasonable retrospective evidence that waiting at least four weeks before your follow-up assessment — and plausibly before a repeat injection — may genuinely improve apparent resolution rates, or at minimum won't hurt them and will reduce unnecessary re-treatment of lesions still in the process of responding. It's a reasonable nudge toward longer intervals in your own protocols, not a mandate. That wraps up this issue — twenty years of underused radiation in Merkel cell carcinoma, a sharper transplant skin cancer risk tool built from variables already sitting in your intake form, a sobering but appropriately cautious look at malignant transformation in erosive oral lichen planus, and a practical argument for patience with intralesional 5FU dosing intervals. Thanks for listening, and I'll see you next month.