This is JAMA Dermatology, March 2026, and we're covering one article this time — but it's a substantive one, so we're going to take our time with it. It's called "Identification of Risk Factors for Acral Melanoma in US Veterans," out of the Hartman group at Brigham and Women's, with VA collaborators, published as an original investigation with an accompanying editorial in the same issue.

Let's start with why this paper exists at all. Acral melanoma — palms, soles, nail unit — is the one melanoma subtype we all know behaves differently from garden-variety cutaneous melanoma, presents later, has a worse prognosis, and doesn't track with UV exposure the way superficial spreading or lentigo maligna melanoma do. The frustrating part clinically has always been that we don't have a real risk-factor profile for it the way we do for UV-driven melanoma. Most of what drives our risk counseling — fair skin, nevus count, sun exposure history — was built on non-acral cutaneous melanoma cohorts and doesn't obviously apply here. And because acral melanoma is rare, building a cohort large enough to say anything statistically meaningful is hard. That's the gap this paper is trying to fill, and they chose an unusually good population to do it in: US veterans, using the Veterans Affairs health system's national data infrastructure, which lets them assemble a rare-disease cohort at a scale most single-institution studies could never touch. Veterans also carry a specific exposure history worth interrogating — Agent Orange and other dioxin-containing herbicides from the Vietnam era — which gives this study a hypothesis that's genuinely unique to this population.

Methodologically, this is a nested case-control study spanning cancer diagnoses from 2000 to 2024. Worth walking through why they built it this way. Acral melanoma is rare enough that a cohort study looking forward from exposure to outcome would need an enormous and impractically long-followed population to accumulate enough cases. Case-control design flips that — you start with your rare outcome, identified here through the VA Cancer Registry plus a validated natural-language-processing pipeline run against pathology reports, which they report performed with an F1 score of point nine-two, so quite good precision — and then you look backward at exposures. Every acral melanoma case was matched four-to-one against two different control groups: veterans with non-acral cutaneous melanoma, and veterans with no melanoma diagnosis at all. That dual-control design is the smartest part of the methodology, and it's worth appreciating why they did it. Comparing acral melanoma against melanoma-free controls tells you what's associated with developing acral melanoma at all. But comparing acral melanoma against non-acral cutaneous melanoma controls does something more specific — it tells you what distinguishes the acral subtype from melanoma in general, stripping out factors that just predispose to melanoma broadly. That second comparison is really the intellectually important one, because it's testing whether acral melanoma is mechanistically distinct, not just a rarer flavor of the same disease.

They matched on diagnosis year and VA outpatient utilization to control for surveillance intensity and health-system engagement — a real strength, since more clinic visits mechanically means more chances to biopsy incidental lesions, which could otherwise confound any risk factor tied to health-care contact. Exposures pulled from the data warehouse included the usual demographic and comorbidity variables, but also body mass index, smoking, an alcohol-use screening score, prior actinic keratosis and keratinocyte carcinoma, prior nevi, photosensitizing medications, dermatology visit frequency in the two years before diagnosis, and — the star exposure here — Agent Orange exposure, captured through a validated VA flag that the authors note has been checked against Vietnam-era records with about ninety-five percent correct identification. They ran conditional logistic regression to get adjusted odds ratios, which is the right regression choice given the matched design — it respects the matched sets rather than treating everyone as independent. They also pre-specified two sensitivity analyses: restricting to Vietnam-era veterans specifically, and restricting to palmoplantar and subungual acral melanoma sites, both sensible ways to check whether the main findings hold in the subgroups where you'd most expect a herbicide signal or where site-specific biology might differ.

Now the numbers. The final cohort was one thousand two hundred ninety-two veterans with acral melanoma, matched against a little over five thousand melanoma-free controls, and separately against a little over five thousand non-acral cutaneous melanoma controls. As you'd expect from a veteran population, this was overwhelmingly male, around ninety-four percent, with a median age in the seventies.

The headline finding is Agent Orange exposure. It was significantly associated with higher odds of acral melanoma in both comparisons — about a thirty percent increase in odds versus non-acral cutaneous melanoma controls, and a similar roughly thirty percent increase versus melanoma-free controls. Both were statistically significant, and importantly, this effect showed up in the acral-versus-cutaneous-melanoma comparison, not just the acral-versus-no-melanoma comparison — meaning Agent Orange exposure doesn't just track with melanoma risk generally, it appears specifically associated with the acral subtype over and above other melanoma. That's the finding that makes this paper newsworthy rather than incidental. That said, the authors themselves translate the absolute magnitude for you, and it's worth saying out loud because it recalibrates the excitement appropriately: using published US incidence figures for acral melanoma, this effect size works out to somewhere between about zero point four and zero point eight additional cases per million veterans per year. So this is a real, statistically robust, biologically interesting signal — but the absolute population impact is small. Statistically significant and clinically real as an epidemiologic association, but not something that's going to reshape a screening program.

The second major finding cuts the other direction and is honestly just as interesting: current smoking was associated with lower odds of acral melanoma — about a thirty-five percent reduction versus cutaneous melanoma controls, and a full fifty percent reduction versus melanoma-free controls. That's a substantial, statistically significant inverse association, and it's the kind of counterintuitive finding that's easy to over-read. It doesn't mean smoking protects against melanoma — smoking is obviously not protective against anything you'd want to promote clinically — but it's an association worth flagging because it mirrors similar paradoxical smoking-melanoma associations reported elsewhere, and possibly reflects unmeasured confounding around health behaviors, surveillance patterns, or survivor-type biases in a registry population, rather than true biological protection.

The other findings had an interesting split personality depending on which control group you used. Prior actinic keratosis and prior keratinocyte carcinoma were each associated with higher odds of acral melanoma when compared with melanoma-free controls — which makes sense, since these are both markers of a person who's engaged with dermatologic care and has some baseline skin cancer risk. But both were associated with lower odds of acral melanoma when compared against cutaneous melanoma controls. That pattern is exactly what you'd predict if actinic keratosis and keratinocyte carcinoma are markers of cumulative UV damage — they push you toward melanoma risk in general, but cutaneous melanoma much more than acral melanoma, because acral melanoma isn't UV-driven. So relative to other melanoma patients, acral melanoma patients look like they carry less of that UV-damage signature. Prior nevus followed a similar logic — associated with higher odds of acral melanoma versus melanoma-free controls, which is unsurprising, but the paper's supplemental framing again underscores that this doesn't distinguish acral melanoma from cutaneous melanoما as cleanly.

On limitations, the authors are appropriately restrained in their conclusions section, and it's worth being honest about what this design can't do. It's a retrospective registry-based study, so exposure ascertainment quality depends on how well things like Agent Orange flags, smoking status, and comorbidity variables are captured in the data warehouse — and while the Agent Orange flag has published validation, there's always some misclassification risk with administrative exposure variables. The population is overwhelmingly older and male, so generalizability to women, younger patients, and non-veteran populations is genuinely limited — and given that acral melanoma disproportionately affects Black, Asian, and Hispanic patients in the general population, a cohort that's still majority white, even though this study captured meaningfully higher proportions of Black and Hispanic patients in the acral melanoma group than in the cutaneous melanoma comparison group, doesn't fully mirror the demographic reality of acral melanoma nationally. Registry-based case ascertainment, even with a well-validated NLP pipeline, could still miss cases diagnosed outside VA pathology capture. And fundamentally, this is a case-control association study — it establishes correlation with appropriately conservative causal language, and the authors explicitly frame their conclusion as identifying a need for continued investigation rather than establishing causation between herbicide exposure and acral melanoma.

So what do you actually do with this as a practicing Mohs surgeon and dermatologic oncologist? The Agent Orange finding is scientifically important and worth knowing — it's a legitimate, biologically plausible signal that deserves follow-up mechanistic and epidemiologic work, and if you have veteran patients presenting with acral pigmented lesions, it's reasonable to have herbicide exposure history in the back of your mind as part of a fuller clinical picture, and to know that the literature is starting to substantiate a concern many veterans themselves have raised. But this is not practice-changing in the sense of altering your biopsy threshold, your surveillance intervals, or how you counsel patients today — the absolute risk increment is tiny, well under one extra case per million veterans annually. What is more conceptually useful for daily practice is the broader message running under this whole paper: this study adds real evidence to something you probably already intuited, which is that acral melanoma has a distinct risk-factor architecture from UV-driven cutaneous melanoma, with actinic keratosis, keratinocyte carcinoma, and smoking all behaving differently in acral melanoma than they do in ordinary cutaneous melanoma. That reinforces why we shouldn't be extrapolating UV-based risk stratification or screening logic onto acral lesions, and why acral melanoma continues to deserve its own risk model, its own research pipeline, and its own index of suspicion independent of a patient's overall UV risk profile. Interesting and hypothesis-generating on the herbicide front, confirmatory and clinically grounding on the "acral melanoma is not just sun-driven melanoma in a weird spot" front.

That's the one article for this episode. Thanks for listening, and we'll pick back up with the next issue.