Welcome back to the journal review. This is the January twenty twenty-six issue of the Journal of the American Academy of Dermatology, and we've got four items to get through today — two original brief reports, and a letter-and-response exchange that's really worth sitting with as a case study in how to critique a big database study. Let's get into it.

First up is a retrospective cohort study out of a Bronx tertiary center looking at cutaneous metastasis in non-Hispanic Black and Hispanic and Latino patients. The setup here is straightforward — cutaneous metastasis from an internal malignancy is uncommon, it usually signals advanced disease, and historically carries a median survival in the three-to-five-month range. What's genuinely missing from the literature is any meaningful data on how this presents in non-Hispanic Black and Hispanic populations specifically, since most of the foundational cohorts — Lookingbill's classic series, the Taiwanese cohort, the Indian series — weren't built to answer that question. So this group set out to fill that gap using their own institutional experience.

Methodologically, this is a retrospective chart review pulling from the electronic medical record over a ten-year window, two thousand thirteen through twenty twenty-three, identifying every Hispanic or Latino and non-Hispanic Black or African American patient with a confirmed cutaneous metastasis. They ran standard comparative statistics — t-tests, chi-squared, Wilcoxon rank-sum — and built a Kaplan-Meier survival curve. The authors don't spell out their rationale for the retrospective design, but it's the obvious choice here, and I'd flag that as my own inference rather than theirs — cutaneous metastasis is rare enough, and this specific demographic slice rare enough within that, that you simply couldn't accrue a meaningful prospective cohort at a single center in any reasonable timeframe. A retrospective EMR pull across a decade is really the only feasible way to get to a sample of any size.

And that sample was sixty-five patients — about two-thirds Hispanic or Latino, one-third non-Hispanic Black. Mean age at the primary cancer diagnosis was fifty-nine, climbing to sixty-four by the time the cutaneous metastasis was recognized, with a median gap of about a year and a half between the two diagnoses. Every single case was biopsy-confirmed, which is reassuring methodologically — no cases were included on clinical impression alone. Adenocarcinoma was the dominant histologic subtype, showing up in nearly half. On primary tumor type, breast cancer led at forty-five percent, lung a distant second at sixteen percent, and — notably — melanoma was rare, only about one in twenty. Most patients, roughly eighty-five percent, developed the skin metastasis after their primary cancer was already known, but in about one in seven, the skin lesion was actually the first sign that led to the cancer diagnosis — which is the clinically important reminder buried in this paper. The anterior chest wall was the most common site at over forty percent, with the abdominal wall and back each accounting for about one in six. And when they compared survival between the group presenting with skin metastasis first versus those with a known primary already, there was no significant difference — so cutaneous metastasis as a presenting sign doesn't appear to carry worse or better prognosis than developing it later in the disease course.

In their discussion, the authors note their findings largely mirror the existing international literature — breast cancer as the leading primary, subcutaneous nodules as the predominant morphology, similar favored anatomic sites. Two things stood out as different from prior cohorts: median survival in this group was about twenty-three months, meaningfully longer than the roughly nine-month median reported in a Thai cohort, and melanoma was distinctly underrepresented compared to a prior US series — which the authors reasonably attribute to the lower baseline incidence of melanoma in patients with darker skin types. They're upfront about the limitations — small sample, single center, and dependence on EMR documentation, which is really just an acknowledgment that this reflects the inherent rarity of the entity rather than a fixable design flaw.

Practically, I'd call this descriptive and hypothesis-generating rather than practice-changing — there's no new diagnostic algorithm or treatment signal here. But it is a useful reminder for anyone doing skin exams or biopsying atypical nodules in Black and Hispanic patients: keep breast and lung adenocarcinoma on the differential for a new chest wall, abdominal, or back nodule, remember that in about one in seven cases the skin finding comes before any cancer diagnosis exists, and don't let the rarity of melanoma in this population make you complacent about biopsying anything that doesn't fit a benign pattern.

Second article: a retrospective study looking at germline pathogenic variants and immune-related adverse events in melanoma patients on checkpoint inhibitors, out of Cleveland Clinic's familial melanoma registry. The background here is a nice bridge between genetics and immunotherapy toxicity — we already know checkpoint inhibitors carry a real burden of immune-related adverse events, and prior work has mostly looked at tumor-intrinsic features driving that risk, with a few early signals — like variants in IL7 — suggesting host genetics might matter too. This same group had previously shown that patients carrying germline pathogenic variants in cancer-predisposition genes actually respond better to checkpoint inhibitors and live longer with melanoma. The open question they're tackling now is whether that same germline signal that helps drive better tumor control also drives more toxicity.

The design is a retrospective comparison, twenty-seven germline-variant-positive patients against one hundred twenty-four variant-negative patients with similar baseline characteristics, drawn from a cohort they'd already characterized in prior publications. The authors don't explicitly justify the retrospective design, but my read is that this is essentially a planned secondary analysis of an existing, carefully phenotyped registry — since they'd already established the efficacy findings in this same population, going back to look at toxicity in the same patients is the logical and efficient next step rather than building a new cohort from scratch.

The variant spectrum itself is worth noting for a moment — BRCA2 was the most frequent single gene involved, followed by CHEK2 and MUTYH, with smaller numbers of BAP1, MITF, and several DNA-repair and tumor-suppressor genes scattered through the rest.

On to the results, and this is where it gets clinically interesting. Variant-positive patients had significantly more total immune-related adverse events than variant-negative patients — about nine in ten versus roughly seven in ten — a statistically significant and clinically meaningful gap. When you split that by severity, the low-grade and high-grade events were each numerically higher in the variant-positive group, but neither reached significance on its own — likely an underpowered-subgroup issue given the small numbers, so I wouldn't read too much into that split. What was significant and clinically important: variant-positive patients were more than twice as likely to discontinue checkpoint inhibitor therapy because of toxicity — about forty percent versus seventeen percent. That's a number that matters at the bedside, because it means these patients are meaningfully more likely to lose access to a therapy that, per this group's prior work, actually works better for them.

Breaking it down by treatment setting: in the adjuvant setting, variant-positive patients had significantly more adverse events overall. In the metastatic or unresectable setting, the overall event rate was similar between groups, but the variant-positive patients hit their first immune-related adverse event significantly sooner — a median of about six weeks versus roughly nine weeks — suggesting earlier immune activation. And organ-by-organ, variant-positive patients had significantly more cutaneous and gastrointestinal events specifically, which is directly relevant to us as dermatologists, since rash and dermatitis were the leading cutaneous manifestation driving that difference.

The authors tie this back to their earlier mechanistic work, proposing that a reduction in circulating myeloid-derived suppressor cells in variant-positive patients may explain both the improved tumor response and the increased toxicity — essentially the same enhanced immune activation cutting both ways. They're appropriately cautious about limitations: the variant-positive group is small, which limits how far you can push subgroup analyses, and adverse event documentation across a retrospective chart review is never perfectly consistent.

From a practice standpoint, I'd frame this as interesting and biologically plausible, but not yet something that should change how you counsel patients before starting immunotherapy — this isn't powered or designed to generate a validated risk-prediction tool. What it should do is raise your index of suspicion: if you're following a melanoma patient on checkpoint inhibitors who's known to carry a germline cancer-predisposition variant, especially in the adjuvant setting, have a lower threshold for early skin and gastrointestinal symptom surveillance, since this group is both more likely to flare and more likely to flare sooner.

Now the third and fourth pieces belong together — a letter to the editor and the original authors' response, both concerning a large retrospective cohort study on mastocytosis and skin cancer risk that was published earlier this year. I won't re-litigate the original study in full, but the short version, as described by the letter writers, is that it used a large United States health care network database, propensity-score matched more than twenty thousand patients, and found an increased risk of skin cancer associated with mastocytosis.

The letter raises four methodologic points, and they're worth walking through because they're the kind of critique you should be running yourself whenever you read a large registry-based study. First, the control group was selected based on having had a general health checkup — coded under the relevant International Classification of Diseases code for routine exam — and the letter writers argue this could make the control group systematically healthier or more health-engaged than the mastocytosis group, introducing what's called healthy-user bias. They suggest a dermatologically matched control group, such as patients with benign adnexal tumors, would better control for surveillance and detection bias. Second, they point out the study lumped cutaneous and systemic mastocytosis together, despite these being biologically distinct entities with different malignancy risk profiles — citing Swedish registry data suggesting risk varies by whether the disease is benign, advanced, or mixed. Third, race and ethnicity weren't included in the matching, which matters given well-documented disparities in skin cancer incidence and presentation across racial groups. And fourth, the original study looked only at incidence — melanoma, basal cell carcinoma, small cell carcinoma — without any data on tumor aggressiveness, metastatic spread, or mortality, which the letter writers argue limits how much clinical weight you can put on the finding, particularly for melanoma.

The response from the original authors, Bergström and colleagues, addresses each point directly, and it's a good model of how to defend a real-world data study without being defensive. On the control group, they argue the bias isn't necessarily one-directional — arguably, patients getting the routine checkup are more likely to have incidental skin cancers caught, which if anything would understate rather than inflate their reported risk — and they note that multiple sensitivity analyses, including an unmatched analysis, produced similar risk estimates, supporting robustness. On the cutaneous-versus-systemic subtype question, they concede the point but explain that International Classification of Diseases coding limitations made that stratification impractical in this dataset, and they mention a more granular follow-up study is already underway. On race, they actually went back and ran an additional matched analysis incorporating age, sex, and race and ethnicity categories — White, Black, Hispanic, Asian — and report that the significant associations for melanoma, squamous cell carcinoma, and basal cell carcinoma all held up, though that data wasn't shown in the letter itself. On broader subgroup stratification, they explain their choice was to preserve statistical power given how rare mastocytosis is, and they explicitly invite future pooled or multinational studies to enable that kind of subgroup work. And on tumor aggressiveness and mortality, they acknowledge that granularity simply isn't available in the TriNetX platform they used, but note that mortality specifically is the planned focus of a dedicated follow-up study.

There's no results or takeaway beat to extract here in the usual sense — this is a methods debate, not new data. But the practical value for us is really about how to read big real-world-data dermatology-oncology studies going forward: watch control group selection closely, ask whether disease subtypes were lumped inappropriately, notice whether race and ethnicity were accounted for, and remember that incidence signals without severity or mortality data are a first step, not the final word. Keep an eye out for the promised subtype-stratified and mortality-focused follow-up studies from this same group — those will be the ones that actually move the needle on how closely we should be screening mastocytosis patients for skin cancer.

That wraps our four articles for January. To summarize the practical bottom line: the cutaneous metastasis paper is a useful descriptive reminder to keep breast and lung primaries on your differential for new trunk nodules in Black and Hispanic patients; the germline variant paper should heighten your vigilance for early cutaneous and gastrointestinal immune toxicity in checkpoint inhibitor patients with known cancer-predisposition genes, without yet changing consent conversations; and the mastocytosis exchange is less about a specific action item and more about sharpening how critically we read large database oncology-dermatology studies. Thanks for listening, and we'll see you next month.