Welcome to this November twenty twenty-six review from the Journal of the American Academy of Dermatology. We've got four pieces on the docket this time — a clinical review on drug interactions with immune checkpoint inhibitors, a retrospective National Cancer Database study on surgical timing in vulvar melanoma, a brief report on recurrence risk stratification for high-risk head and neck squamous cell carcinoma treated with Mohs, and an ethics journal club piece on diagnostic discordance in melanocytic lesions. Let's get into it. First up is a clinical review out of Albert Einstein and Montefiore on medications commonly prescribed by dermatologists in patients receiving immune checkpoint inhibitors — so this is a synthesis piece, not a primary study, pulling together the existing evidence base rather than generating new data. The setup is one we all live with now: checkpoint inhibitors have transformed oncology, but response rates swing anywhere from fifteen to seventy percent depending on tumor type, and a growing literature suggests that's not just about the tumor — it's about the host and, critically, about concurrent medications. The authors' point is that dermatologists sit right in the crosshairs of this because so many of our bread-and-butter prescriptions — antibiotics, systemic steroids, JAK inhibitors, biologics — are exactly the drugs under scrutiny for blunting checkpoint inhibitor efficacy, and we're often the ones reaching for them to manage immune-related adverse events or incidental skin disease in these very patients. The antibiotics section is the most mature literature here, built on the idea that the gut microbiome shapes checkpoint inhibitor response — germ-free mice getting fecal transplants from responders do better, non-responder transplants do worse, and antibiotics strip out the favorable bacterial diversity. Across a systematic review of over forty thousand patients, any antibiotic exposure was linked to significantly worse overall and progression-free survival, and that signal was strongest when antibiotics were given within the sixty days before or after checkpoint inhibitor initiation — a finding replicated in an independent meta-analysis of over five thousand solid tumor patients. Outside that window, the effect seems to fade; one retrospective study found no survival difference when antibiotics were given more than forty-two days after initiation. Class matters too — fluoroquinolones and piperacillin-tazobactam look worse than other classes, plausibly because they hit the favorable bacterial strains harder — and duration matters, with courses longer than fourteen days showing impact while shorter courses didn't. Even indication seems to matter, with prophylactic use looking worse than treatment of actual infection in most but not all datasets. And interestingly, probiotics seem to do the opposite, correlating with improved overall and progression-free survival, presumably by protecting microbial diversity. The clinical relevance for us is concrete: doxycycline for bullous pemphigoid, or antibiotics paired with EGFR-inhibitor-associated acneiform eruptions, are both scenarios where an ICI may be running concurrently, and the authors' practical point is that sequencing antibiotic courses to avoid that critical two-month window around ICI initiation, when feasible, is a reasonable harm-reduction strategy. Corticosteroids get a similar dose-duration-timing breakdown. The historical exclusion threshold in checkpoint inhibitor trials was prednisone at ten milligrams a day or above, and a large retrospective study of over thirteen thousand patients bore this out with real granularity — more than five milligrams daily was associated with roughly a twenty percent reduction in overall survival, and more than seventy-five milligrams daily with roughly a forty percent reduction. Duration tracked similarly, two to five days costing ten to twenty-five percent of survival, a week or more costing closer to thirty-five percent. Timing mattered too, with the sharpest impact — roughly a halving of the survival-related hazard — when steroids were given within a month of ICI initiation, an effect that attenuated the further out you got, essentially disappearing beyond six months. The genuinely tricky part clinically is cutaneous immune-related adverse events themselves, which happen in thirty to sixty percent of ICI patients and are paradoxically associated with better survival — yet the data on whether treating them with systemic steroids erodes that survival benefit is mixed, with some studies showing no effect and others implicating steroids given early or at high peak dose. Current NCCN guidance still leans toward systemic steroids for most grade one and above cutaneous toxicities, but the authors flag that steroid-sparing alternatives are increasingly reflected in updated guidance, and topical steroids remain the safe default given minimal systemic absorption on intact skin. The JAK inhibitor section is more mechanistic and earlier-stage — the hypothesis is a bimodal relationship where JAK signaling is actually necessary early in the antitumor immune response, helping antigen presentation and direct tumor cytotoxicity, but sustained JAK signaling later drives T-cell exhaustion and blunts response. That's led to trial designs deliberately delaying JAK inhibition until after ICI priming — a phase two trial added itacitinib six weeks into immunotherapy for metastatic non-small cell lung cancer to try to thread that needle. The review's full scope also extends into biologics and other immunomodulators, though the detail there isn't fully captured in what we're working from here. The overarching takeaway for practice is this: none of this is practice-changing in the sense of a new mandate, but it should sharpen your threshold for reaching for systemic steroids or antibiotics in ICI patients — favor the lowest effective dose, the shortest course, topical or steroid-sparing alternatives where reasonable, and an awareness of timing relative to ICI initiation when you have any flexibility in scheduling. Next, a retrospective cohort study using the National Cancer Database looking at surgical timing in early-stage vulvar melanoma. This is a rare, biologically distinct entity from cutaneous melanoma with worse baseline outcomes, and the specific gap here is that while surgery is standard of care for early-stage disease, nobody had really quantified what delay in getting to that surgery costs you. Methodologically, the authors pulled stage one and two vulvar melanoma patients diagnosed between two thousand four and twenty twenty-two, defining time to treatment initiation as the interval from diagnostic biopsy to definitive excision, and asked two questions: does delay predict positive margins, and does it predict overall survival. For the margin question they used multivariable logistic regression, which is the right tool because the outcome — margin positive or negative — is binary, and logistic regression estimates the odds of that outcome while adjusting for competing variables like age, comorbidity burden, tumor thickness, and ulceration. For the survival question they turned to Cox regression, which instead of predicting a yes-or-no outcome, models the instantaneous risk of death at any given moment among those still being followed, which is the standard approach whenever you have patients followed for different lengths of time and you care about *when* events happen, not just whether they happen. Here's the genuinely instructive wrinkle: the Cox model has a baked-in assumption called proportional hazards — that the relative risk between groups stays constant over the entire follow-up period. The authors checked this using Schoenfeld residuals, a standard diagnostic test, and found the assumption was violated for their treatment-delay variable. Rather than forcing a single hazard ratio across the whole follow-up period, which would have been misleading, they split the analysis at a prespecified thirty-six-month cut point and estimated separate hazard ratios for the early and late intervals — a time-stratified Cox approach that's a nice example of a model being adapted to what the data are actually telling you rather than the other way around. The results bore out exactly that kind of two-phase story. Among just under nineteen hundred patients, delay of ninety-one days or more from biopsy to excision was independently associated with about a seventy percent increase in the odds of a positive margin compared with excision within thirty days — statistically significant and clinically meaningful, since margin status is something we act on directly. Shorter delays, up to ninety days, showed no significant margin effect. Survival followed the same pattern you'd expect from that biology: Kaplan-Meier curves were essentially superimposed for the first three years regardless of delay, then diverged, and the adjusted Cox models confirmed this — no survival difference within the first thirty-six months for delayed patients, but beyond thirty-six months, a delay of ninety-one days or more carried roughly a sixty percent increase in the hazard of death, which was both statistically and clinically significant. Breslow depth and ulceration were also independently tied to positive margins, which is reassuring from a biologic-plausibility standpoint. The authors are appropriately circumspect about limitations — this is retrospective registry data, the National Cancer Database doesn't capture melanoma-specific survival or recurrence, so everything here is overall survival rather than disease-specific, there's no information on biopsy technique, surgeon experience, or the actual reasons behind each delay, and residual confounding is always possible despite the adjustment. One finding worth flagging on its own: delays were getting longer in more recent years and were more common among non-White patients, which is a system-level equity signal worth sitting with. For practice, this one leans toward actionable: it reinforces that for vulvar melanoma specifically — a site where biopsy-to-excision logistics can be slower given multidisciplinary coordination with gynecologic oncology — pushing to get definitive excision within roughly thirty to sixty days, and certainly under the ninety-day mark, has a real margin and long-term survival rationale behind it, not just generic urgency. Third, a brief report from the University of Vermont on high-risk head and neck cutaneous squamous cell carcinoma treated with Mohs. The clinical problem is a familiar one: NCCN guidelines recommend considering adjuvant radiation for high-risk tumors after Mohs, but disease-specific survival with Mohs alone is already excellent — up to ninety-nine percent in some series, over ninety-five percent even for high-risk tumors — so blanket escalation to adjuvant radiation risks over-treating a lot of patients who were never going to recur. The design is a retrospective single-institution cohort, two hundred seventy-one consecutive non-irradiated high-risk tumors in two hundred thirty-one patients, all treated with Mohs alone, restricted to tumors at least two centimeters in diameter — which by itself already puts them at T2 by AJCC eighth edition and T2a by the Brigham staging system. The analytic approach here is a classic two-stage screen: univariate analysis first to flag candidate risk factors, then multivariable analysis to see which of those survive adjustment for everything else. This is worth pausing on because it's a very common but imperfect strategy — the logic is that you don't want to throw every variable into a multivariable Cox model at once with a modest sample size, since that risks overfitting and unstable estimates, so you use univariate screening to narrow the field first. The tradeoff, which the authors don't dwell on but is worth knowing, is that this two-step approach can miss variables that only matter in combination with others, and the choice of which univariate threshold earns a spot in the multivariable model is somewhat arbitrary. The multivariable piece itself is a Cox proportional hazards model, estimating hazard ratios for local recurrence, regional recurrence, and relapse-free survival. On multivariable analysis, three factors held up as independent recurrence risks: poor differentiation, perineural invasion, and diameter greater than four centimeters — each roughly tripling to sixfold increasing the hazard depending on the outcome examined. Despite that, the overall cohort did remarkably well — only five of two hundred thirty-one patients, about two percent, actually died of their squamous cell carcinoma. That set up the paper's key proposal: stratify patients by how many of those three factors they carry. Patients with zero or one of these risk factors — the overwhelming majority, two hundred sixty-two of the two hundred seventy-one tumors — had a two-year relapse-free survival above ninety percent, with no meaningful difference between having zero versus one factor. But the nine patients who had two or more of these three factors together had a two-year relapse-free survival of only twenty-six percent, and roughly a seventy-four percent two-year cumulative incidence of local recurrence, in-transit metastasis, or nodal metastasis — compared with under ten percent incidence when only one risk factor was present. That's a genuinely dramatic separation, and the authors line it up against existing high-risk criteria from NCCN, from Gupta and colleagues, and from Ruiz and colleagues, essentially arguing their combination rule is simpler and sharper. The obvious caveat, and the authors would surely agree, is that nine patients is an extremely small denominator to be building a risk-stratification threshold on, even though the effect size is large and the two-year rate differential is striking. This is single-institution, retrospective, and the group is already running a prospective phase two trial — the MOHSAHRTSS study — to test a risk-adapted approach of active surveillance versus adjuvant radiation using exactly these kinds of thresholds. So for your practice: this is useful, intuitively-aligned decision support — a two-centimeter-plus head and neck squamous cell carcinoma with zero or one of poor differentiation, perineural invasion, or diameter over four centimeters appears to do excellently with Mohs alone and probably doesn't need you to reflexively invoke adjuvant radiation per NCCN's broader high-risk language. But having two or more of those three factors together is the signal that should trigger a real conversation about adjuvant therapy. Call this informative and reinforcing of a risk-adapted mindset rather than fully practice-changing yet, given how thin that high-risk subgroup is — the prospective trial will be the one to watch. Last is an ethics journal club piece, so this is a different animal entirely — no methods, no data, just a case-based ethical discussion in a Dear-Doctor format. The scenario: a forty-two-year-old man's changing mole gets read by three different dermatopathologists who land in three different places — severely dysplastic nevus, atypical intraepidermal melanocytic proliferation associated with a nevus, and outright invasive melanoma. The question posed is how the dermatologist should ethically navigate that kind of three-way split. The response grounds this in real numbers — histologic discordance among dermatopathologists reading atypical melanocytic lesions produces equivocal diagnoses in roughly a quarter of cases in published series — and walks through why: melanocytic lesions sit on a genuine biologic spectrum, the classification criteria are inherently nuanced, and even gene expression profiling, while commercially available, isn't yet established as a standard adjunct, so it adds data without fully resolving ambiguity. The ethical tension is laid out plainly — leaning toward the most severe read means wider excision, heavier surveillance, and psychological burden for a patient who might not have needed any of it, while leaning conservative risks under-treating a lesion that could carry real mortality risk. The piece acknowledges that clinicians often default to whichever dermatopathologist they trust most, which provides psychological comfort but isn't itself a rigorous arbitration method, and that medicolegal anxiety pushes many toward over-treating the worst read even when that's not proportionate care. The core ethical guidance centers on informed consent and transparency: telling a patient that experts disagree is uncomfortable, but withholding that disagreement undermines genuine informed consent, and the right move is to disclose the range of interpretations honestly without being alarmist, framing the management rationale and the risk-benefit tradeoffs of each path, and leaning into shared decision-making precisely because there may be no single objectively correct answer here. The piece points to systemic fixes in progress — consensus conferences, second-opinion pathways, subspecialty consultation, and standardized frameworks like MPATH-Dx version two, a four-class hierarchy meant to improve interobserver reporting consistency — while being honest that none of these eliminate discordance entirely. The closing message is really a professionalism point for both clinicians and pathologists: the goal isn't to manufacture false certainty, it's to communicate uncertainty honestly, avoid overstating any single read, and clearly state which interpretation you're acting on and why. That's the November lineup — a practical reminder to think twice about antibiotic and steroid choices around checkpoint inhibitor timing, solid evidence that surgical delay in vulvar melanoma costs you margins and long-term survival, a sharper risk-stratification lens for adjuvant therapy decisions in high-risk head and neck squamous cell carcinoma after Mohs, and a grounding reminder that diagnostic uncertainty in melanocytic lesions is a communication challenge as much as a pathology one. Thanks for listening, and I'll see you next month.