Welcome back to the journal review. This is the September 2026 issue of the Journal of the American Academy of Dermatology, and we've got four pieces worth your time this episode — a technology pearl on total body photography, a pharmacovigilance letter on sonidegib, a methodological commentary on electronic health record research, and a letter responding to a stage II melanoma risk-stratification study. Let's get into it.

First up is a JAAD Online technology pearl out of Trento, Italy, from Pagliarello and Girardelli, addressing a problem you've probably run into yourself if you run a pigmented lesion clinic: total body photography is excellent for pattern change detection, but even purpose-built systems don't let you go back to an archived image and pull an absolute lesion measurement. So when a patient swears a nevus has grown, or you eyeball apparent enlargement on comparison, you're stuck making a subjective call — comparing apparent sizes across images that may have been taken at different distances, different zoom, different positioning, with no way to anchor them to a true scale.

Their solution is refreshingly low-tech: place a reference marker of known size — they use a 20 millimeter white adhesive sticker — on the skin within each photographed sector before you shoot. Because calibration has to be image specific, they recommend one marker per sector, not just one per session, since camera distance and magnification will vary sector to sector even within the same visit. Once you've got that marker in frame, you can go back retrospectively, measure the marker's pixel diameter in something like ImageJ or Fiji, and use simple proportional scaling to convert any lesion's pixel dimensions into an absolute millimeter measurement. They've also built a free, offline, single-file HTML tool that does this calibration and measurement step for you, no installation required, and it's posted on Mendeley for anyone who wants it.

There's no results or limitations section here — this is a pearl, not a study — so the value proposition is really just practical: minimal added time per visit, works across camera systems including handheld devices, and it converts a purely subjective "this looks bigger to me" conversation into an objective, quantifiable one you can show the patient. For your practice, this is a nice one to actually adopt tomorrow if you or your referring dermatologists do serial total body photography for dysplastic nevus syndrome or melanoma-prone families — it costs you a sticker and a few seconds per photograph, and it directly strengthens the evidentiary basis for a "stable, no biopsy needed" conversation, or conversely gives you real numbers when you're deciding whether to intervene.

Next, a letter to the editor presenting a real-world pharmacovigilance analysis of sonidegib, the smoothened inhibitor approved back in 2015 for locally advanced basal cell carcinoma not amenable to surgery or radiation. The gap they're addressing is that despite sonidegib's efficacy, hedgehog pathway inhibitors as a class are notorious for adverse events that drive treatment discontinuation, and there hadn't previously been a dedicated FAERS-based characterization of sonidegib's real-world safety profile specifically in basal cell carcinoma.

Methodologically, this is a straightforward descriptive pharmacovigilance study — they queried the FDA Adverse Event Reporting System from 2015 through mid-2025, restricted to cases explicitly coded for basal cell carcinoma, removed duplicates, and excluded pure administrative or treatment-failure reports unless there was an accompanying physiologic adverse event. This is not a comparative or inferential study, so there's no hypothesis test to teach here — it's just frequency counting and ranking by MedDRA organ system category, which is the appropriate first-pass tool for signal detection in a spontaneous reporting database. It's worth flagging for the audience, as the authors themselves do, that FAERS has no denominator — you know how many adverse events were reported, but not how many patients were actually exposed to the drug, so you cannot calculate a true incidence rate from this design, only a proportion among reports.

With that caveat, out of two hundred ninety-six adverse event reports, the average patient was mid-sixties, with a slight male predominance, and about three-quarters of reports came from healthcare professionals rather than consumers, which lends some credibility to clinical accuracy. There were five deaths and roughly one in five reports involved hospitalization. The most common physiologic signals were musculoskeletal — muscle spasms in about one in nine reports, along with fatigue, alopecia, asthenia, and elevated creatine phosphokinase — plus a gastrointestinal cluster of diarrhea, nausea, and vomiting, and weight loss showing up in about one in seventeen reports. The signal I'd flag as clinically important: rhabdomyolysis and renal failure each appeared in about two percent of reports, and cerebrovascular accidents in a bit over two percent — notably more neurologic dysfunction, including taste disturbance, dizziness, and gait disturbance, than prior trial data would have predicted.

The authors' discussion ties this back to hedgehog pathway biology — the pathway is involved in tissue repair, stem cell proliferation, and maintenance of muscle, hair follicle, and taste bud function, so off-target inhibition plausibly explains the muscle, alopecia, and dysgeusia signals mechanistically. They also note that trial data suggested sonidegib might have a more favorable adverse event profile than vismodegib for things like muscle spasms and dysgeusia, but this real-world dataset picked up a broader spectrum of serious events, including rhabdomyolysis and stroke, than the controlled trials captured. Limitations are the ones inherent to any FAERS study: voluntary reporting means underreporting and reporting bias, no denominator means no true incidence, and no causality can be established from spontaneous reports alone.

Practical takeaway: this isn't practice-changing in the sense of altering who you select for sonidegib, but it is a useful reminder to build creatine phosphokinase monitoring and a low threshold for asking about weight loss, gait change, or taste disturbance into your longitudinal management of patients on hedgehog inhibitor therapy — real-world surveillance suggests the neurologic and musculoskeletal tail risk is broader than trial data alone would lead you to expect.

Third is a commentary — Notes and Comments — from the Mount Sinai group, Patel, Karpoff, and Lebwohl, and this one is squarely aimed at how we should read every database study you'll encounter in this and every other journal issue going forward. Their thesis: electronic health record and claims-based "real-world evidence" databases feel objective because they're large and standardized, but diagnostic and procedural codes reflect billing and administrative necessity, not verified clinical diagnosis — and this problem is magnified enormously for rare diseases, where even a small absolute rate of miscoding can swamp the true disease signal.

There's no original methods or results section here — it's an argument built on two illustrative examples, both using generalized pustular psoriasis as the case study. In a previously published TriNetX analysis by another group, over forty-four thousand patients were coded with generalized pustular psoriasis prior to matching, used to derive a hazard ratio linking that diagnosis to pyoderma gangrenosum. The authors point out that population-based registry studies put true generalized pustular psoriasis prevalence at roughly one to nine per hundred thousand people — meaning a federated network cohort of over forty thousand adults with this coded diagnosis is epidemiologically implausible for a disease this rare and typically requiring tertiary-center, biologic-level management. Their own follow-up TriNetX query for adults with a generalized pustular psoriasis code returned over one hundred ninety-one thousand patients — again wildly out of proportion to Swedish registry data suggesting a prevalence around nine per hundred thousand. Their point isn't subtle: if the denominator disease cohort is inflated by miscoding under a nonspecific ICD-10 code, then any downstream hazard ratio built on that cohort — even one that reaches statistical significance — may simply be reflecting diagnostic noise rather than a real biological association.

Their call to action is directed at investigators, journals, and IRBs: disclose coding validation methodology, avoid overstating causal claims from code-based cohorts, and require transparency about diagnostic fidelity before publication. For your own practice, this isn't a clinical takeaway so much as a critical-appraisal takeaway — the next time you see a splashy database study, especially one built around a rare dermatologic diagnosis, your first question should be whether the reported cohort size is even epidemiologically plausible for that disease, because if it isn't, everything downstream — hazard ratios, odds ratios, adjusted risk estimates — is standing on sand.

Last, a letter to the editor responding to the Adigbli et al. study on tumor-infiltrating lymphocyte scoring in stage II melanoma, so we're getting the commentary lens on that original retrospective cohort work rather than the primary paper itself. The clinical gap being addressed is real and familiar to you: American Joint Committee on Cancer eighth edition staging for sentinel-node-negative stage II melanoma is purely anatomical, and stage IIA patients in particular are currently excluded from adjuvant immunotherapy trials despite the fact that, as this letter highlights, over forty percent of progression events in the original cohort actually occurred in that supposedly lower-risk IIA group.

The underlying study the letter is discussing had real methodological heft worth noting — multicenter, median follow-up of nine and a half years, with substantial interobserver agreement on tumor-infiltrating lymphocyte scoring, using the standard three-tier classification of brisk, nonbrisk, and absent infiltrate. The letter writers highlight the counterintuitive core finding: it wasn't absent infiltrate that predicted the worst outcomes, it was the nonbrisk pattern — present in about eighty percent of progressing tumors, compared to roughly six in ten stage IIB or IIC tumors overall. And critically, stage IIA patients with a nonbrisk pattern had ten-year progression-free survival numbers, around eighty percent, that looked much more like the higher, formally staged IIB and IIC group with favorable absent-or-brisk patterns, at around eighty-nine percent, than like other IIA patients — meaning anatomical staging alone was misclassifying risk in a clinically meaningful way, not just a statistically detectable one.

The letter's discussion section adds a mechanistic interpretation worth remembering: the paradox that both the complete absence of lymphocytic infiltrate and a vigorous brisk infiltrate both predicted a favorable course, while the intermediate nonbrisk pattern predicted progression, suggests that nonbrisk infiltration may represent a dysfunctional, exhausted T-cell state — active but ineffective immune engagement — rather than simply weak or strong immune surveillance in a linear sense. That's a biologically coherent story, but the letter is appropriately cautious about translation: the underlying cohort is retrospective with selection bias, there's no standardized tumor-infiltrating lymphocyte quantification protocol across institutions yet, and the cohort predates modern adjuvant and neoadjuvant immunotherapy, so it's unclear whether these same tumor-infiltrating lymphocyte patterns and outcomes hold in patients now routinely getting checkpoint inhibitor therapy after diagnosis.

Practical takeaway here: this is genuinely interesting and potentially important, but it is not yet practice-changing for you as a treating clinician. The letter itself calls for prospective validation with standardized tumor-infiltrating lymphocyte assessment before clinical adoption. What it should do is sharpen your index of suspicion when you're reading a pathology report on a thin-to-intermediate, sentinel-node-negative melanoma with a nonbrisk infiltrate pattern noted — that's a biologic clue worth flagging in tumor board discussion and multidisciplinary planning, even though it isn't yet incorporated into formal staging or adjuvant therapy eligibility criteria.

That wraps our four articles for this September 2026 issue — a practical measurement pearl you can implement immediately, a real-world safety signal worth folding into your hedgehog inhibitor monitoring, a sobering methodological check on how much to trust big database studies, and a biomarker story in stage II melanoma that's promising but still awaiting prospective validation. Thanks for listening, and I'll see you next issue.