Welcome back to the journal review. This is your walkthrough of the September 2026 issue of the Journal of the American Academy of Dermatology, and today we're covering four pieces from the Brief Reports and consensus sections — a SEER-based melanoma epidemiology study, a National Cancer Database analysis of Merkel cell carcinoma, a single-institution comparison of nail unit versus non-nail acral melanoma, and an international Delphi consensus on perioperative nail surgery. Let's get into it.

First up is a brief report using the Surveillance, Epidemiology, and End Results program — SEER — to look at melanoma incidence by tumor thickness before and during the COVID-19 pandemic. You'll recall the well-publicized finding that melanoma incidence dropped by more than fifteen percent in 2020, driven mostly by missed in situ and thin diagnoses during lockdown-era care disruption. This group wanted to know what happened after that — did early-stage diagnoses rebound, and did the pandemic leave any fingerprint on thick, T4 disease specifically.

Methodologically, they pulled histologically confirmed melanomas from seventeen SEER registries between 2012 and 2022, deliberately excluding two registries — Los Angeles and Connecticut — because more than half of their melanoma cases had unknown thickness, which would have contaminated the T-category analysis. For the long-term trend, they used Joinpoint regression to calculate an average annual percent change, or AAPC, across the full eleven-year window, excluding 2020 itself per SEER's own recommendation since that single pandemic year behaves like a statistical outlier that would distort a smooth trend line. Joinpoint regression is worth pausing on: instead of fitting one straight trend line across the whole period, it tests whether the data are better explained by two or more line segments joined at a breakpoint — so it can detect a change in trajectory, like a trend accelerating or reversing, rather than assuming incidence moves at a constant rate for a decade. For the short-term pandemic effect, they instead used simple pairwise percent changes, comparing 2020, 2021, and 2022 each against a 2019 pre-pandemic baseline, with Bonferroni correction applied since they were running multiple comparisons and didn't want to inflate their false-positive rate.

The results tell a two-part story. Over the full decade, in situ melanoma incidence rose significantly, with an average annual increase of about three and a half percent, while invasive melanoma overall stayed essentially flat. T4 tumors, however, also rose significantly, at roughly three percent per year — a trend the authors note actually predates the pandemic by years. Now the short-term pandemic comparisons: 2020 showed the expected across-the-board drop, with in situ down about eighteen percent, invasive down sixteen percent, and even T2 and T3 tumors down by seven to fourteen percent relative to 2019 — consistent with diagnostic disruption touching essentially every thickness category that year. By 2021, in situ melanoma had already rebounded, up nearly nine percent over 2019, and by 2022 it was up a full twenty percent over baseline, with even T1 tumors showing a modest but significant four percent increase. Critically, though, T4 tumors did not show that same short-term rebound pattern in the 2021 or 2022 comparisons against 2019 reaching Bonferroni significance in the same way — despite the long-term decade-long upward trend continuing underneath.

So the takeaway the authors draw is that early-stage detection — thin and in situ disease — has largely caught back up and even overshot pre-pandemic rates, consistent with either catch-up screening or a resumption of the pre-existing rising trend in indolent, screen-detected disease. Thick tumor incidence, meanwhile, continues its slow pre-pandemic climb, but there's no clean signal of a discrete pandemic-driven surge of delayed, thick melanomas — it looks more like continuation of a trend that was already there before COVID than a new wave of missed diagnoses presenting late. The limitations are real: excluding two large registries because of missing thickness data limits generalizability, there's no longer-term follow-up beyond 2022, and the design can't distinguish postponed care from other drivers of the thick-tumor trend. For your practice, this isn't practice-changing, but it is reassuring background epidemiology — it suggests the anxiety about a coming wave of pandemic-delayed thick melanomas hasn't materialized as a distinct bump, though the underlying secular rise in thick tumors remains a real trend worth continued vigilance and patient education around skin checks.

Next, a National Cancer Database study looking at immunotherapy use and survival in metastatic Merkel cell carcinoma, comparing patients with a known primary cutaneous lesion against those with metastatic disease of unknown primary — abbreviated here as mUPMCC. Metastatic Merkel cell carcinoma of unknown primary accounts for roughly one in ten to one in seven cases, and the background gap here is that population-level outcome data in the immune checkpoint inhibitor era comparing these two presentations has been limited.

The authors queried the National Cancer Database from 2004 through 2023, identified stage IV Merkel cell carcinoma by histology code, excluded cases missing clinical T-stage, and defined unknown primary as cT0. Because immune checkpoint inhibitors weren't approved for metastatic Merkel cell carcinoma until 2017, they restricted the treatment-utilization and survival analyses to patients diagnosed from 2017 onward — a sensible design choice to avoid diluting the immunotherapy-era signal with pre-approval patients who simply couldn't have received the drug. They used three complementary statistical tools: logistic regression to model the odds of receiving immunotherapy as a function of patient and disease factors, Kaplan-Meier estimation to describe raw survival curves over time, and Cox proportional hazards modeling to compare survival between groups while adjusting for confounders. It's worth teaching the Cox model briefly since it does a lot of work in this paper — rather than comparing survival at one fixed time point, it compares the instantaneous risk of death between groups at every moment someone remains under observation, which is what lets it handle patients enrolled at different times and followed for different lengths of follow-up without throwing away partial information from patients who are lost to follow-up or still alive at data cutoff. The output, a hazard ratio, tells you the relative risk of death in one group compared to the reference group across the whole follow-up period, assuming that relative risk stays roughly constant over time — that "proportional hazards" assumption is the main thing you're buying into with this technique versus, say, comparing simple two-year survival percentages.

On the results: unknown-primary patients were younger and more often treated at academic centers than known-primary patients. Immune checkpoint inhibitor utilization was similar between groups — about sixty-three percent in unknown primary versus sixty-one percent in known primary — and in the adjusted logistic regression, primary tumor status itself was not independently associated with receiving immunotherapy. What did predict lower odds of receiving immunotherapy: being over eighty, having a higher Charlson-Deyo comorbidity score, and having received radiotherapy or chemotherapy, the latter cutting the odds by roughly two-thirds. The survival story is the more interesting part. Among patients who did not receive immunotherapy, overall survival was statistically similar between unknown and known primary — two-year survival around thirty-nine versus thirty percent, not a significant difference. But among patients who did receive immunotherapy, unknown-primary patients did meaningfully better — two-year survival of fifty-six percent versus forty percent for known primary, with median survival roughly double, at about thirty months versus fifteen months, a difference that was statistically significant with an adjusted hazard ratio favoring unknown primary. When the authors formally tested for an interaction between primary tumor status and immunotherapy receipt in the Cox model, the interaction trended toward a greater benefit in unknown primary but did not reach statistical significance — so this is a real, numerically striking finding that stops short of statistical proof that the treatment effect itself differs by group.

The biological rationale offered is interesting: unknown-primary Merkel cell carcinoma is hypothesized to arise from an immune-mediated clearance of the primary tumor, and these tumors show more CD8 tumor-infiltrating lymphocytes and higher ultraviolet-associated mutational burden — both plausible substrates for better checkpoint inhibitor responsiveness. The authors also contrast their findings with a prior population-based study that found no survival improvement for unknown-primary disease across the pre- versus post-checkpoint-inhibitor eras, and they reasonably caution that era-based comparisons conflate immunotherapy availability with general improvements in supportive care and management over time, which may explain the discrepancy. Limitations here are the usual registry-study issues — selection bias inherent to Commission on Cancer-accredited centers reporting to the database, no cause-specific mortality or recurrence data, and no granularity on which checkpoint inhibitor was used, at what dose, or in what sequence. For your practice, this is genuinely interesting but not yet practice-changing — it doesn't argue for withholding or preferentially selecting immunotherapy based on primary tumor status, since checkpoint inhibitors already benefited both groups. What it does do is offer a physiologically plausible signal that unknown-primary patients may derive an even larger benefit, which is useful for counseling and hypothesis-generating for future prospective work, but the interaction term not reaching significance means you shouldn't treat this as established differential efficacy yet.

Third, a retrospective comparison from two San Francisco health systems looking at nail unit acral melanoma against non-nail acral melanoma. Nail unit melanoma is the rarer, more aggressive-behaving cousin of acral melanoma, disproportionately affecting patients with skin of color, and it comes with real diagnostic friction — distinguishing it from benign longitudinal melanonychia, getting an adequate biopsy from the nail matrix, and dermoscopic features that are less reliable than on glabrous skin. Most prior literature comparing nail unit and non-nail acral melanoma has focused narrowly on survival; this group wanted a broader clinicopathologic comparison.

They retrospectively identified all acral melanomas diagnosed between 2007 and 2022 across University of California San Francisco Health and Zuckerberg San Francisco General Hospital — deliberately pairing an academic tertiary center with a safety-net hospital to broaden racial and ethnic representation. Diagnoses came from dermatopathology databases, cross-checked against clinical images and notes, and a fellowship-trained Mohs surgeon on the team personally adjudicated any case where the pathologic label and clinical picture disagreed on nail involvement — a nice quality-control step given how easily nail unit and adjacent digital acral disease can be conflated in a chart. For statistics, they used Fisher's exact test for categorical comparisons and the Mann-Whitney U test for continuous variables. Both choices make sense for this dataset: Fisher's exact test is the right tool rather than a standard chi-square test when you have relatively small or sparse categorical subgroups — which they did here, especially within race and ethnicity strata — because chi-square relies on a large-sample approximation that can misbehave with small cell counts, while Fisher's exact test calculates the exact probability of the observed table directly. The Mann-Whitney U test, meanwhile, compares whether one group's values tend to rank higher than another's without assuming the data follow a normal distribution — appropriate here because Breslow depth is classically right-skewed rather than bell-shaped.

Out of two hundred eighty-one acral melanoma patients, fifty-eight had nail unit disease and two hundred twenty-three did not. Age, sex, race and ethnicity distributions, clinical stage, and median Breslow depth were all statistically similar between the two groups — so on paper, these looked like comparable tumors at comparable depths. The one variable that stood apart was ulceration: present in thirty-six percent of nail unit melanomas versus eighteen percent of non-nail acral melanomas, a statistically significant and clinically substantial two-fold difference.

The authors' interpretation is the interesting part. Because Breslow depth was statistically indistinguishable between the two groups, the excess ulceration in nail unit disease can't be fully explained by nail tumors simply presenting deeper due to diagnostic delay — if delay were the whole story, you'd expect deeper tumors to track with more ulceration. Instead, they raise the possibility that nail unit melanoma has some intrinsic biological tendency toward ulceration independent of depth, which would be a genuinely novel observation if it holds up. They're appropriately cautious, noting this needs further study to see whether ulceration alone carries independent prognostic weight specific to the nail unit, separate from its established role as a general acral melanoma survival predictor. Limitations are straightforward: retrospective design, a fairly small nail unit sample of just fifty-eight cases, and restriction to two urban San Francisco health systems, which limits generalizability. For your practice, this is a genuinely thought-provoking observation rather than an immediately actionable one — it doesn't change staging or margins, but it does support maintaining a low threshold of suspicion for nail unit pigmented lesions and considering that ulceration may be showing up at earlier depths in this subsite than you'd otherwise expect.

Last is a different kind of article entirely — an international Delphi consensus statement on perioperative management in nail surgery, which many of you performing digit and nail unit procedures will find directly useful. This isn't an original data study, so there's no results-versus-limitations statistical scaffold here; it's a structured expert-opinion document, and it's worth understanding the Delphi method itself since it's used throughout dermatology guideline development. The rationale the authors give is straightforward: nail surgery practice is heterogeneous because high-quality comparative evidence is sparse, so rather than waiting for trials that may never be funded, a Delphi process systematically aggregates expert opinion toward areas of genuine agreement while explicitly flagging where consensus can't be reached.

Mechanically, this involved ten dermatologists from nine countries, all with substantial nail surgery experience — a median of seventeen and a half years in practice and fifteen years focused on nail surgery specifically, mostly in academic settings. Round one was fifty open-ended questions used to generate candidate statements. Rounds two and three converted those into two hundred sixty Likert-scale items, with consensus predefined as seventy percent or greater agreement or disagreement on a five-point scale. Of the two hundred thirty items evaluated in round two, one hundred fifty-eight reached consensus outright, thirty-three were revised for round three, and thirty-nine were dropped entirely. Round three then reassessed the revised items, approving twenty-two more, dropping eight, and leaving eight without consensus — for a final tally of one hundred eighty consensus statements, an iterative structure worth noting because it shows the panel genuinely converged rather than just voting once.

On substance, a few domains stand out. On antibiotic prophylaxis: the panel did not support routine use, reserving it for situations like bone exposure, active infection, significant ulceration, reoperation after infection, or field contamination risk, plus patient factors like uncontrolled diabetes, immunosuppression, peripheral vascular disease, or specific cardiac risk conditions — notably, anatomic site of surgery alone was not considered sufficient justification. Penicillins and cephalosporins were favored first-line, with lincosamides or macrolides for penicillin allergy, and a single dose one to two hours preoperatively was deemed sufficient. On bleeding risk, the consensus generally favored continuing antithrombotic therapy through most nail procedures rather than stopping it, with INR checks two to three days ahead for patients on vitamin K antagonists, specialist consultation for higher-risk direct oral anticoagulant cases, and a clear risk stratification — total nail unit excision, malignant tumor surgery, and procedures with bone exposure were classified as higher bleeding risk, while punch biopsies for longitudinal melanonychia, spicule excision, phenolization, and nail plate avulsion were lower risk. On anesthesia, the distal digital block was preferred as first-line over proximal block for faster onset and less discomfort, with infiltration volumes kept small — roughly half a milliliter to one and a half milliliters per side for fingers, up to two and a half for the great toe — and multimodal pain-reduction strategies like slow injection rates, thin needles, distraction techniques, and topical or physical adjuncts were favored over any single approach; notably, there was no consensus on routine use of epinephrine-containing anesthetics, which is worth knowing given ongoing practice variation on that exact question. On antisepsis, chlorhexidine and povidone-iodine were both endorsed as first-line, without a routine need for a second antiseptic step outside high-contamination or immunosuppressed scenarios. On tourniquet use, the panel emphasized that pressure, width, and correct technique matter more than device sophistication — a simple rolled glove finger was considered adequate for most cases — with duration generally capped around fifteen minutes and periodic release recommended for longer procedures, with extra caution in vascular disease, advanced age, or smoking history. And for postoperative pain, nonopioid regimens were favored across the board, with acetaminophen alone sufficient for low-pain procedures, acetaminophen plus a nonsteroidal anti-inflammatory for moderate-to-high pain procedures, and scheduled dosing in the first twenty-four hours transitioning to as-needed use afterward.

The authors are upfront that this whole document rests on expert opinion in an area where high-quality comparative evidence is genuinely lacking — so treat it as a practical scaffold for standardizing practice rather than a substitute for trial-level evidence. For those of you doing nail surgery regularly, this is one of the more directly usable pieces in the issue: it won't overturn your existing routine if you're already following sound principles, but it gives you a citable, internationally vetted reference point for exactly the kinds of perioperative questions — prophylaxis thresholds, anticoagulation management, tourniquet timing — that come up constantly and rarely have a clean answer in the primary literature.

That wraps this month's review — a reassuring early-stage rebound in melanoma incidence with a persistent but pre-existing thick-tumor trend, a hypothesis-generating survival signal favoring immunotherapy in unknown-primary Merkel cell carcinoma, a provocative ulceration finding in nail unit melanoma worth watching rather than acting on yet, and a genuinely practical international consensus for anyone doing nail surgery. Thanks for listening, and we'll see you next month.