Welcome back to the journal review. This is the March 2026 issue of the Journal of the American Academy of Dermatology, and we've got four pieces worth your attention this time — a large pharmacoepidemiology study on niacinamide and cardiovascular risk, a Game Changer retrospective on head-and-neck melanoma in situ margins, a brief coding-pattern report out of Arizona Medicare data, and a multicenter dermoscopy study on ultraviolet-induced fluorescence. Let's get into it.

First up is an original retrospective cohort study titled "Decreased risk of cardiovascular disease in dermatology patients taking niacinamide: A TriNetX cohort study." You'll recognize the clinical backdrop immediately — we've been reaching for oral niacinamide, five hundred milligrams twice daily, as chemoprophylaxis in patients with multiple nonmelanoma skin cancers for years now, largely on the strength of the ONTRAC trial. But a recent prospective discovery cohort and meta-analysis raised a real concern: elevated levels of the niacin terminal metabolites, 2PY and 4PY, were associated with increased three-year major adverse cardiac events. Since niacinamide sits downstream in the same metabolic pathway as niacin, that finding cast a shadow over a supplement most of us hand out reflexively. This paper set out to directly test whether niacinamide use is associated with cardiovascular harm — and whether that risk differs based on preexisting atherosclerotic disease.

Methodologically, this is a retrospective cohort built on the TriNetX federated network, which pools de-identified electronic health record data across more than a hundred million patients from participating health systems. The authors identified patients with two or more nonmelanoma skin cancer diagnoses, or bullous pemphigoid, who were then started on oral niacinamide above five hundred milligrams — that dose threshold is important, since it excludes dialysis-related niacinamide formulations and other non-chemoprophylactic uses. These patients were propensity-matched one-to-one against niacinamide-naive controls, matching on age, sex, race, ethnicity, and a long list of cardiovascular risk factors — hypertension, diabetes, hyperlipidemia, chronic kidney disease, heart failure, prior MI or stroke, smoking, and statin use — with a standardized mean difference target under 0.1, which is the accepted threshold for adequate balance in propensity-matched work. A retrospective database design here makes obvious sense: you're never going to randomize thousands of patients to niacinamide versus placebo and wait years for STEMIs to accrue, so leveraging a massive real-world claims-and-EHR network is really the only feasible way to power a rare-outcome analysis like this. They also applied the Benjamini-Hochberg procedure with Storey's q-value method across the outcome panel, which is a sensible guardrail given they were testing nine separate cardiovascular endpoints and wanted to control the false discovery rate rather than let multiplicity inflate their positive findings.

After matching, each arm had just over three thousand two hundred patients, so about sixty-five hundred total. This was an older cohort — mean age right around seventy-two — predominantly white, about seventy-nine percent non-Hispanic, and fifty-nine percent male, with a heavy comorbidity burden: roughly six in ten had hypertension, about half had hyperlipidemia, a quarter had diabetes, and a quarter already had chronic ischemic heart disease. Here's the headline: niacinamide use was associated with a significant reduction in ST-elevation MI — 0.9 percent versus 1.6 percent in controls, roughly a 45 percent relative risk reduction, and this held up under multiplicity correction. Peripheral vascular disease was also significantly reduced, about a 25 percent relative risk reduction. Cardiac arrest showed a similar protective signal, roughly half the risk, but that one did not survive the Benjamini-Hochberg correction, so treat it as a trend rather than a confirmed finding. Non-ST-elevation MI and type 2 MI showed no difference between groups at all. And notably, this protective STEMI signal persisted specifically in patients with a prior ASCVD history, despite the fact that more than three-quarters of the whole cohort was already on a statin.

The authors' discussion leans into a mechanistic explanation for why this pattern makes sense rather than looking like noise: STEMI results from plaque rupture and acute thrombotic occlusion, an atherosclerotic process niacinamide has some translational biology supporting, whereas NSTEMI and type 2 MI are frequently secondary phenomena — supply-demand mismatch from sepsis or arrhythmia — where you wouldn't expect a plaque-stabilizing agent to move the needle. That's a reasonable argument for why splitting MACE into its component parts, rather than pooling everything into one composite endpoint the way prior studies did, uncovered a signal that a composite analysis would have washed out. They also contextualize this against the existing literature: a 2017 Cochrane review found niacin itself doesn't reduce cardiovascular risk and, after excluding biased trials, was actually linked to harm; the original ONTRAC trial found no significant cardiovascular safety signal but was underpowered for rare events with fewer than four hundred patients over eighteen months; and the most comparable prior retrospective study, a VA population analysis by Wheless and colleagues, found no increased MACE risk and a nonsignificant trend toward benefit — broadly consistent with what's shown here, and the authors even flag likely patient overlap between the two databases given shared institutional participation in TriNetX.

Limitations are the ones you'd expect from any observational EHR study: causality can't be established, there's the possibility of residual unmeasured confounding despite tight matching, events occurring outside participating institutions would be missed, the population skews heavily white which limits generalizability, and the three-year window may still be too short to capture the full cardiovascular risk-benefit picture. There's also some imprecision inherent in using ICD codes to define two-lesion NMSC history, though the authors note these codes carry high positive predictive value for pathology-confirmed skin cancer.

Practically, here's how I'd frame this for your own practice: this isn't the study that proves niacinamide protects the heart — it's an observational signal, not a randomized trial, and shouldn't be oversold to patients. But it is a meaningfully reassuring answer to a concern that had started circulating in response to that niacin metabolite paper. If you've been second-guessing whether to keep recommending niacinamide chemoprophylaxis in your high-risk NMSC patients because of cardiovascular metabolite worries, this data — alongside the VA cohort finding similar reassurance — supports continuing that practice, including in patients with established atherosclerotic disease. Call this reassuring-and-actionable for counseling purposes, not practice-changing in the sense of altering who you start on niacinamide or at what dose.

Next is a JAAD Game Changer piece, which as you know is a short retrospective commentary explaining how an already-published article shifted practice — the commentary authors here aren't the original investigators. This one revisits a 2024 paper by Tate and colleagues: a single-center, ten-year retrospective review of Mohs micrographic surgery for melanoma in situ on the head and neck, specifically looking at what excision margins were actually required for histologic clearance and which clinical variables predicted needing wider margins. The clinical problem here needs no introduction to this audience — there's still no real consensus on standardized margins for head and neck melanoma in situ, and cosmetically sensitive sites make the tension between adequate margin and tissue conservation especially sharp.

The original study's findings, as distilled here, were fairly specific and site-driven. Melanoma in situ on the cheek and the eyelid, along with lesions with larger preoperative diameters, were more likely to require margins wider than five millimeters for complete clearance. Conversely, lesions on the ear tended to clear within margins narrower than five millimeters. The Game Changer authors frame the significance of this as filling a real evidence gap — rather than applying a blanket five- or six-millimeter margin recommendation across the entire head and neck the way some consensus guidelines have suggested, this data lets you calibrate your surgical planning by anatomic subsite and lesion size going into the case.

There's no new methodology or limitations section to walk through here since this is a commentary rather than original data, but the practical takeaway is worth stating plainly: if you're doing Mohs for head and neck melanoma in situ, this supports counseling patients preoperatively that cheek, eyelid, and larger lesions are more likely to need multiple stages or wider initial margins, while ear lesions may reasonably clear with tighter margins — useful for setting expectations and possibly for initial margin planning, though it doesn't replace your own stage-by-stage histologic judgment.

Third is a brief report — a research letter, really — titled "Medicare data analysis of complex versus intermediate repairs following Mohs micrographic surgery in Arizona, 2018 to 2021." The backdrop: the CPT coding manual redefined intermediate and complex repair codes in January 2020, after concern that complex repair codes were being over-utilized nationally. A prior national analysis had already shown a twelve percent nationwide drop in complex repair coding among Mohs surgeons from 2019 to 2021, with the effect more pronounced in surgeons earlier in their careers and in academic practice. This group wanted to drill into a single state's data to see what's actually driving that aggregate national trend.

The design here is purely descriptive — they pulled publicly available CMS billing data for all dermatologists billing CPT 17311, the Mohs code for high-complexity anatomic sites, in Arizona across those four years, then looked at the ratio of intermediate repair codes to complex repair codes for each of the seventy identified physicians. A descriptive claims-based approach is really the only option available here, since CMS public data doesn't let you link the Mohs code to the specific repair performed with any more granularity, and there's no chart-level clinical detail to draw on.

The findings tell a story of uneven guideline adoption. Across the seventy Arizona Mohs surgeons, average intermediate repair coding increased by about eleven percent in the two years after the 2020 definition change compared to the two years before — consistent with the national trend. But that average masks real heterogeneity: in any given year, thirty to forty percent of these surgeons coded zero intermediate repairs at all. And most strikingly, twenty of the seventy — that's about twenty-nine percent — never billed a single intermediate repair across all four years spanning the coding change, while about ten percent actually decreased their intermediate coding proportion after the update, the opposite direction from what the new definitions would predict.

The authors are appropriately cautious about interpretation given the descriptive-only design — they can't tell from CMS data whether a physician's practice genuinely skews toward higher-complexity repairs justifying the coding pattern, whether there's a knowledge gap about the updated definitions, financial incentive at play, or whether billing staff rather than the physician are driving these coding choices independent of clinical documentation. Their limitations section is candid about all of this.

Practically, this isn't clinical practice-changing in the sense of altering how you operate, but it is squarely relevant to your practice management and compliance posture. If your own coding habits haven't shifted since January 2020 — if you're still defaulting heavily to complex repair codes across the board — this is a nudge to revisit the current CPT definitions with your billing staff, both to code accurately and to reduce audit risk, since persistent divergence from the post-2020 definitions is exactly the kind of outlier pattern that draws payer scrutiny.

Last, and worth the most time, is an original multicenter study: "Ultraviolet-induced fluorescent dermoscopy for the diagnosis of skin tumors." Background here is a nice quick tour through dermatoscope evolution — nonpolarized light requiring immersion fluid, then polarized light after 2007 revealing shiny white structures invisible under nonpolarized light, and now, since 2022, ultraviolet-induced fluorescent dermoscopy, or UVFD, built into pocket dermatoscopes. UVFD had only been validated in small case series up to this point — no one had looked at its performance across a large series of actual skin tumors. That's the gap this study fills.

The design is a retrospective, multicenter study across three tertiary referral clinics in Chile, Poland, and France, using consecutive biopsied lesions from a roughly fifteen-month window. Each lesion was imaged in polarized light, nonpolarized light, and UVFD — the UVFD component using a 365-nanometer UV LED dermatoscope attachment on a smartphone, with alcohol as the immersion fluid rather than the usual gel, since alcohol doesn't interfere with the fluorescence signal. Two experienced dermoscopists, blinded to histopathology, scored each image for standard dermoscopic criteria plus two UVFD-specific features — border definition and the presence of an "ochre" color — with disagreements adjudicated by a third reviewer. Worth flagging: because invasive melanomas were underrepresented in the initial consecutive series, the authors deliberately added extra invasive melanoma cases pulled from a public cancer center to bolster that subgroup. That's a reasonable move to ensure adequate representation of a clinically important but less common outcome, but it does mean the reported prevalence of invasive melanoma in this cohort isn't a natural, real-world base rate — it's enriched by design, which matters if you're trying to extrapolate diagnostic yield to your own everyday clinic mix.

Of five hundred sixty-seven lesions collected, five hundred fifty-one made the final analysis after excluding cases with missing data or poor image quality. Basal cell carcinoma was the single most common diagnosis at about twenty-eight percent, followed by nevi at about nineteen percent, melanoma at about thirteen percent — of which roughly two-thirds were in situ and one-third invasive, with a mean Breslow depth around one millimeter — then seborrheic keratosis and squamous cell carcinoma each around eleven to twelve percent.

The core finding is a genuine trade-off between modalities rather than UVFD simply outperforming polarized dermoscopy across the board. Keratin-related structures were dramatically better seen under UVFD — keratin itself roughly ten times more likely to be visualized, comedo-like openings essentially only appreciable under UVFD in discordant cases, and fissures and ridges also favoring UVFD. Border definition was the standout finding for this audience specifically: lesion margins were well-defined in about seventy-six percent of polarized images versus ninety-seven percent under UVFD, a substantial and clinically intuitive improvement given how UV fluorescence can highlight subclinical extension that visible-light dermoscopy misses. On the flip side, polarized light remained superior for vessels, pigment network, milia-like cysts, blue-white veil, shiny white structures, peppering, dots, and ulceration — which makes physical sense, since several of those structures depend on polarization physics or vascular color contrast that UV fluorescence doesn't capture. Agreement between the two modalities was high, above ninety percent, for most features, but notably lower for keratin and for border definition — which is really just another way of confirming that those are the two domains where UVFD is adding genuinely new information rather than just reproducing what polarized light already shows you. And mechanistically interesting for your invasive melanoma triage: that ochre-colored fluorescence signal was seen predominantly in invasive melanomas rather than melanoma in situ or benign lesions, hinting at a possible noninvasive clue to invasion, though the paper's own limitations section is upfront that this was a retrospective study with several diagnostic subgroups underrepresented, so this needs prospective validation before you'd lean on it clinically.

For a Mohs surgeon, I'd frame the practical takeaway in two tiers. The border-definition finding is the more immediately relevant one — better preoperative margin visualization under UVFD is directly applicable to presurgical mapping for ill-defined basal cell carcinomas or lentigo maligna, and it's plausible this could eventually inform biopsy or debulking planning, though this study didn't test whether UVFD-defined margins actually correlate with final histologic clearance or reduce stage counts, so that link remains to be proven. The ochre-color-and-invasion finding is the more speculative one — intriguing as a potential adjunct clue for triaging suspicious pigmented lesions, but it's an early, retrospective signal in an enriched melanoma sample, not something to act on yet. File this whole study under promising and worth watching for prospective data, particularly anything correlating UVFD-defined borders with actual surgical margin outcomes, rather than something that changes how you approach a case tomorrow.

That wraps up this month's review — a reassuring cardiovascular safety signal for niacinamide chemoprophylaxis, sharper site-specific margin guidance for head and neck melanoma in situ, a reminder to audit your own repair coding against the 2020 CPT definitions, and an emerging fluorescence-based dermoscopy technique that may eventually sharpen your presurgical margin planning. Thanks for listening, and I'll see you next month.