Welcome back to the journal review. This is your walkthrough of JAMA Dermatology for August twenty twenty-six, and we've got two pieces worth your attention this month — a striking single case report on targeted therapy in neurocutaneous melanosis, and a large genetic cohort study quantifying just how much extra skin cancer risk your porokeratosis patients are actually carrying. Let's get into it.

First up is a case report — a letter, really — out of the Hospital for Sick Children in Toronto, on BRAF inhibition in congenital melanocytic nevi and neurocutaneous melanosis. You know the background here better than most: congenital melanocytic nevi, when giant or multiple, raise the specter of neurocutaneous melanosis, meaning melanocytic infiltration of the central nervous system, and that can mean seizures, hydrocephalus, cord compression, a genuinely grim natural history. The overwhelming majority of these are driven by postzygotic NRAS variants, and when NRAS is the driver, the central nervous system disease has been notoriously hard to touch — MEK inhibition has been tried and the responses have been transient at best.

The patient here was a boy identified at birth with two giant congenital nevi, one lumbosacral and one on the thigh extending down the leg, plus over a hundred smaller satellite nevi scattered everywhere. Baseline MRI in infancy already showed the classic neurocutaneous melanosis picture — T1 hyperintense foci intracranially and thickening of the distal cord and conus medullaris — but he was asymptomatic, so he was simply followed. That changed at age seven, after a viral illness, when he developed a new foot drop and ankle clonus and lost the ability to walk independently. Repeat imaging showed the cord lesion progressing, thickening further with nodular enhancement, and critically there was no safe surgical corridor to that spinal lesion. So this was a scenario where standard tools were exhausted.

What they did next is the pivot point of the whole case: rather than assuming this was garden-variety NRAS disease, they sent a cutaneous lesion for targeted next-generation sequencing using a pan-cancer RNA panel. And it came back with an oncogenic BRAF V600E variant at a variant allele frequency around forty percent — no fusion transcripts, no other pathogenic hits beyond a couple of NF1 variants of uncertain significance. That single finding changed the entire treatment calculus, because BRAF V600E is actionable in a way NRAS simply is not with currently available inhibitors.

They started dabrafenib monotherapy after multidisciplinary discussion, titrated to eighty milligrams twice daily, and tolerance was excellent — the only adverse effect noted was mild acral hyperkeratosis. The results, honestly, are what make this worth your time. Over the following months the thickened, darkly pigmented nevi flattened and lightened visibly, and several of the acral satellite lesions cleared completely. By two years of continuous therapy, follow-up MRI showed reduction in cord caliber and stabilization of the nodular enhancement at the conus medullaris — and clinically, this kid regained independent ambulation and went back to playing ice hockey. That's not a surrogate imaging endpoint, that's a functional outcome that matters enormously to a family.

As a case report, there's no methods section to critique and no formal limitations paragraph — this is n of one, and the authors don't pretend otherwise. What they do offer is mechanistic reasoning: BRAF inhibition suppresses constitutive MAPK pathway signaling in the mutant melanocytes, which plausibly explains both the antiproliferative cutaneous effect and, given dabrafenib's known central nervous system penetration — already established in BRAF-mutant melanoma brain metastases and in pediatric low-grade glioma — the neurologic stabilization as well.

The practical takeaway for you: this is not practice-changing in the sense of altering how you manage the run-of-the-mill giant congenital nevus, but it is absolutely practice-changing in how you should think about molecular workup when a patient with extensive congenital nevi and neurocutaneous melanosis progresses despite surveillance. The message is simple and it's one you can act on directly — don't assume NRAS and don't assume nothing can be done. A BRAF V600E variant is present in a real minority of these cases, and finding it opens an actionable, well-tolerated, CNS-penetrant therapy where historically there was only surveillance. If you're the dermatologist in the room when one of these kids progresses, this case is your argument for pushing the oncology team toward tissue-based targeted sequencing before writing off further intervention.

Now to the second piece, and this one is a proper original study — a retrospective matched cohort analysis out of Yale, using the All of Us Research Program, looking at keratinocyte cancer risk in genetically defined porokeratosis.

The background gap here is worth stating clearly. We've known for a while, from clinical and histopathologic series, that porokeratosis carries excess skin cancer risk, with squamous cell carcinoma being the strongest association. But those prior studies were all built on clinical diagnosis of porokeratosis — meaning a dermatologist looked at a lesion and called it porokeratosis. The problem is that porokeratosis is a genodermatosis driven by germline pathogenic variants in mevalonate pathway genes — MVK, PMVK, MVD, FDPS, FDFT1 — plus a somatic second hit in the lesion itself, and a lot of people carrying these germline variants have minimal, easily overlooked disease and never present for care at all. So clinical-diagnosis-based studies were almost certainly missing a big chunk of the true at-risk population, and probably had substantial misclassification going the other direction too, since porokeratosis lesions get mistaken for other things all the time.

So the authors flipped the ascertainment strategy: instead of starting from a clinical diagnosis, they started from the genotype. Using All of Us, which links whole genome sequencing to electronic health record data, they identified everyone carrying a pathogenic or likely pathogenic variant in one of those five mevalonate genes — call that the genetic porokeratosis cohort — regardless of whether they'd ever been clinically labeled with porokeratosis. Controls were everyone else with whole genome data and no porokeratosis variant or diagnosis. That gives you a cohort defined by the causal lesion itself rather than by a sometimes-unreliable clinical read, which is a real methodological upgrade for this specific question.

To handle confounding, they used propensity score matching, five controls to every case, matched on age, sex, race, smoking, obesity, and immunosuppression status. This is worth pausing on because it's a technique you'll see constantly in dermatology-genomics papers going forward. The logic: in an observational cohort like this, cases and controls will differ on all sorts of things that independently affect skin cancer risk — an older, more immunosuppressed cohort will have more keratinocyte carcinoma regardless of porokeratosis genotype. Propensity score matching estimates, for every individual, their probability of being a "case" based on those covariates, then pairs each case with statistically similar controls so that the two groups end up balanced — which is exactly what you want, because randomization isn't possible here. It's a clean alternative to just throwing all the covariates into one giant regression, because it forces you to actually verify balance before you ever look at the outcome, rather than trusting the model to adjust it away after the fact. And they did verify it well — every standardized mean difference after matching was under one percent, meaning the matched groups were essentially indistinguishable on age, sex, race, smoking, obesity, and immunosuppression. That's about as clean a matched comparison as you'll see in registry data.

The actual outcome comparisons were then done with straightforward logistic regression, generating odds ratios for keratinocyte carcinoma overall and by subtype.

On to results. Out of over twenty-six hundred people with a genetic porokeratosis variant, genetic porokeratosis was significantly associated with keratinocyte carcinoma overall, with about a one-third increase in odds — clinically meaningful, not just statistically significant. Basal cell carcinoma odds were up by roughly a quarter, also significant. But squamous cell carcinoma is where the signal was strongest, with odds up by about sixty percent — and that tracks with everything we already knew clinically about porokeratosis being a preferential squamous cell carcinoma risk. Melanoma showed no difference at all between groups, which the authors read as evidence this is a keratinocyte-specific susceptibility rather than a general melanocytic cancer predisposition.

They ran a sensitivity analysis excluding anyone who also carried a variant in a known skin cancer predisposition gene, and the association didn't meaningfully change — so this isn't just piggybacking off some other cancer syndrome riding along in the same patients.

The stratified analyses are genuinely the most clinically interesting part. Among patients with no actinic keratosis diagnosis — so, less obvious chronic photodamage — genetic porokeratosis was still significantly associated with squamous cell carcinoma, with roughly an eighty percent increase in odds. But among patients who did have actinic keratosis, meaning presumably more lifetime UV exposure and more dermatologic surveillance already happening, the porokeratosis-associated odds bump disappeared. The authors' interpretation is that this argues for an intrinsic keratinocyte susceptibility from the mevalonate pathway defect itself, not just porokeratosis being a marker of cumulative sun damage — and they point to congenital and early-onset linear porokeratosis, which shows a high squamous cell carcinoma propensity even without much photodamage, as supporting precedent. When they broke it down by individual gene, MVD variants — which made up the large majority of the cohort — drove significant associations with both keratinocyte carcinoma and squamous cell carcinoma specifically. MVK and PMVK didn't reach significance, but the authors are appropriately honest that this is very likely underpowered given how few carriers there were, not necessarily a true null.

Limitations are stated plainly and are the ones you'd expect from any electronic-health-record-based study: diagnoses came from coded electronic health record data with the usual misclassification risk, and critically, there was no lesion-level data, so they cannot tell you whether these squamous cell carcinomas actually arose within a porokeratosis lesion itself versus elsewhere on a predisposed patient's skin. That's an important nuance — this tells you the patient is at higher risk, not necessarily that you need to biopsy every porokeratosis plaque with equal suspicion.

Practical takeaway, and I'd call this genuinely practice-relevant rather than just interesting: if you identify porokeratosis clinically — even limited, incidental-looking disease — this data supports counseling that patient on an elevated, disproportionately squamous cell carcinoma-weighted skin cancer risk, and it supports building in regular skin exams and photoprotection counseling as a standard part of managing porokeratosis, not an afterthought. It's not yet practice-changing to the point of dictating a specific biopsy threshold for individual porokeratosis lesions — the lack of lesion-level data means we can't make that leap — but as a rationale for surveillance intensity and patient counseling, particularly in patients with MVD-associated disease, this is solid, well-matched, genotype-first evidence you can actually use in the room.

That's the episode — a single dramatic case making the argument for molecular testing in progressive neurocutaneous melanosis, and a large genetic cohort study sharpening exactly how much extra keratinocyte carcinoma risk sits behind a porokeratosis diagnosis. Thanks for listening, and I'll see you next month.