Welcome back to the journal review. This is the June twenty twenty-six issue of the Journal of the American Academy of Dermatology, and we've got four pieces to get through today — a brief report on AI documentation tools, a meaty retrospective on cartilage grafting in nasolabial interpolation flaps, a short reply letter on AI and patient education materials, and a large cohort study on TNF-alpha inhibitors and skin cancer risk. Let's get into it.

First up is a brief report, essentially a workflow study, looking at AI-driven digital scribes in clinical documentation, comparing high-volume versus low-volume specialties against dermatology. This is a follow-up to the same group's earlier pilot in dermatology alone, and now they've broadened out to a multi-specialty cohort at a single tertiary academic center. The setup is straightforward: clinicians who'd never used a scribe before did a one-month trial with an AI digital scribe — this was the Nuance DAX platform — and then researchers compared documentation metrics for the months before versus after adoption, excluding that training window. Each clinician served as their own control, which is a sensible design choice for something like this, since documentation habits vary enormously between individual physicians and you'd otherwise be fighting a huge amount of between-person noise. They split everyone into high-volume clinicians, more than ten visits per half-day, versus low-volume, using the institutional median as the cutoff.

The headline finding: benefit tracked with volume, not specialty. Dermatologists, who were all classified as high-volume in this cohort, saved a clinically meaningful chunk of time — about a quarter hour off daily documentation burden, which was statistically significant. Internal medicine, family medicine, and internal medicine subspecialties saw even larger absolute reductions, on the order of half an hour to over thirty minutes a day. But specialties like orthopedics, neurology, oncology, and urology showed no significant change. When they stratified purely by volume rather than specialty label, high-volume clinicians overall saved about two minutes per appointment and something like half an hour a day — translating to roughly one to two and a half hours saved per week. Low-volume clinicians, interestingly, didn't fare as well — some measures of their after-hours EMR burden actually went up. The authors speculate this might reflect longer, more complex visits, lower scribe accuracy for those encounter types, or just a bigger editing burden relative to the time saved. Worth noting: note length went up for dermatology's high-volume users without a corresponding increase in time spent, which the authors read as evidence the extra verbiage was low-effort templated content rather than something clinicians had to labor over.

Limitations are honestly stated — modest sample size, single institution, single scribe platform, and incomplete temporal pairing meaning they had to rely on independent samples t-tests rather than a paired design. Practical takeaway for you: if you're a high-volume Mohs and dermatologic oncology practice, this adds another data point suggesting digital scribes are likely to pay off for you specifically, more so than for lower-volume cognitive specialties. It's not fully practice-changing on its own — it's a modest, single-center brief report — but combined with their earlier dermatology-specific pilot, it's a reasonably actionable signal if you're deciding whether to invest in one of these tools.

Now the main course — a retrospective cohort study out of Penn on cartilage grafts in nasolabial interpolation flap reconstruction of alar defects, and this one has real management implications. The clinical question is one a lot of us have never had rigorously answered: we routinely tuck in a free auricular cartilage graft to support the ala during NLIF reconstruction, but nobody has actually shown that doing so improves outcomes, and the downsides — donor site morbidity, added bulk, potential nostril obstruction — have been more theoretical than measured.

Methodologically, this is a single-center retrospective cohort spanning an impressively long window, September 2008 through May 2024, pulled from a prospectively maintained Mohs surgery database at a tertiary academic center. A retrospective design here makes sense for the obvious reason — you're never going to randomize cartilage graft versus no graft in a business this dependent on individual surgeon judgment and defect-specific anatomy, so the pragmatic approach is to look at what surgeons already do across a large volume of cases and control statistically for the differences between groups. They used univariable analysis to find what predicts cartilage graft use, then built multivariable logistic regression models for the outcomes of interest — compromised breathing and surgical scar revision — adjusting for confounders and checking model fit with standard tools like the variance inflation factor and the Akaike information criterion, which is a nice touch showing they were being careful about overfitting with so many candidate variables.

The results are compelling. Out of two hundred seventy-eight NLIFs, cartilage grafts were used in a little over four in ten cases. What predicted graft use wasn't patient factors — demographics were essentially identical between groups — it was surgeon experience and defect anatomy. Less experienced surgeons used cartilage far more often, with usage dropping from nearly nine in ten cases early in the study period down to about one in four by the most recent years, suggesting a real practice shift over time even independent of this study. Grafts were also more common when the defect was confined to the alar subunit alone, and when the tumor invaded subcutaneous fat or muscle.

Now the outcomes that actually matter to your patients. Donor site complications from the ear harvest were low, about three percent, mostly infection or bleeding — not nothing, but not the main story. The main story is what happened at the ala itself. Compromised breathing was reported in roughly one in fourteen patients overall, but it was three times more common with a cartilage graft than without, a difference that held up on multivariable analysis as an independent, nearly three-fold increased risk. Surgical scar revision told the same story — needed in about one in eight patients overall, but again roughly three times more likely when a cartilage graft had been placed, and this stayed significant after adjustment. Most of those revisions were for cosmetic appearance rather than function, though functional revision was also numerically more common in the graft group.

The authors' discussion is refreshingly self-critical of a widespread practice — they note the cartilage graft rate in prior literature ranges enormously, from about thirteen percent up to nearly universal use, which itself signals that indications are not evidence-based but surgeon-preference-based. They're careful to state they're unaware of robust data actually showing cartilage grafts improve outcomes, which this study certainly doesn't provide — if anything it points the opposite direction. Limitations are the ones you'd expect: single center, retrospective, and there's an inherent confounding-by-indication problem here that multivariable adjustment can only partially solve — surgeons may be grafting the more anatomically challenging defects to begin with, even after adjusting for the variables they measured.

Practical takeaway — this one is close to practice-changing, or at minimum practice-questioning, for anyone doing routine cartilage grafting with NLIFs. The signal is consistent, the effect sizes are not trivial — three-fold increases in both a functional and a cosmetic bad outcome — and it lines up with the biomechanical rationale that native alar cartilage doesn't exist and that added bulk and stiffness could genuinely impair the muscular flaring that keeps the nostril patent. This should prompt at least a case-by-case reconsideration of routine grafting, reserving it for more selective indications rather than defaulting to its use, particularly recognizing that the trend at this institution over sixteen years was already toward less frequent use.

Next, a short one — a reply letter to the editor, responding to a prior commentary titled "Guiding Principles for Using Artificial Intelligence to Simplify and Translate Mohs Micrographic Surgery Educational Materials." The original authors, responding here, had published work on using AI translation models to convert Mohs surgical instructions into Spanish. The correspondents who wrote the guiding-principles piece raised points about readability analysis for translated materials and about addressing patient literacy gaps with audio-visual tools. In this reply, the authors thank them for those points and update the record with their own subsequent work — they note they've since studied ChatGPT specifically for improving readability of English-language Mohs postoperative instructions, finding that readability scores did improve across multiple indices, but with a real caveat: key information about postoperative complications was sometimes lost during simplification. They acknowledge they haven't yet applied formal readability indices to the Spanish-language translations specifically, and they agree this is worth doing. They also mention a separate project using a chatbot to deliver postoperative instructions and discuss the idea of audio-visual tools, while conceding the correspondents' point that internet connectivity and access issues are a legitimate barrier to that approach. There's no new data here — it's a scholarly exchange — but the throughline for practice is a caution worth internalizing directly: if you're using AI tools like ChatGPT to simplify patient-facing Mohs instructions for readability or translation, don't assume simplification is safe by default — verify that critical postoperative complication information survives the process, since these authors have already documented that it doesn't always.

Last article, a large retrospective cohort study examining whether TNF-alpha inhibitor therapy is associated with increased long-term risk of nonmelanoma skin cancer and melanoma. This addresses a genuine, longstanding gray zone — prior studies on this question have been mixed, often underpowered, and complicated by confounding from concurrent immunosuppressants and short trial follow-up periods, so a big real-world cohort with long observation time is exactly what's been missing.

This was a single health-system retrospective cohort out of Northwestern, spanning nineteen ninety-six through twenty twenty, pulling patients with chronic inflammatory disease — psoriasis and psoriatic arthritis, hidradenitis suppurativa, Crohn's, ulcerative colitis, rheumatoid arthritis, ankylosing spondylitis, and uveitis — and comparing those with TNF inhibitor exposure against those without. Retrospective is really the only feasible design for a question like this: you need decades of follow-up and thousands of patients to catch a modest-effect, low-incidence outcome like skin cancer, and nobody's randomizing patients to a known effective therapy versus placebo for twenty years to study a secondary cancer signal. They used manual chart review to confirm every skin cancer diagnosis and every TNF inhibitor exposure, rather than trusting diagnosis codes alone — an important quality-control step given how messy administrative coding can be for something like histologic subtype. The core analysis was a multivariable Cox proportional hazards model treating TNF inhibitor exposure as time-dependent, adjusted for age, race, sex, smoking, underlying diagnosis, and duration since first exposure, with sensitivity analyses further adjusting for concurrent immunosuppressants and phototherapy — a smart move given that plenty of these patients were also on methotrexate or similar agents, which are their own confounders for skin cancer risk.

The numbers: of just over fifty-six thousand patients, about thirteen and a half thousand had TNF inhibitor exposure. Follow-up was long — a median of about eight years total for exposed patients, with roughly four years of that being post-exposure follow-up specifically. Skin cancer occurred in about three percent of exposed patients versus two percent of unexposed — sounds small, but on adjusted analysis this translated into a consistent, statistically significant, and clinically meaningful signal across the board: roughly double the risk for basal cell carcinoma, about a seventy percent increase in risk for squamous cell carcinoma, and about a fifty percent increase in melanoma risk, with any skin cancer combined showing roughly a two-fold increased hazard. Reassuringly, though, when they looked at tumor characteristics among those who did develop cancer — histologic aggressiveness for basal cell and squamous cell, and clinical stage for melanoma — there was no significant difference between exposed and unexposed patients. In other words, TNF inhibitors appear to raise the odds you'll develop a skin cancer, but not the odds that the cancer you get will be a nastier one.

Other independent risk factors that fell out of the model were unsurprising — older age, White race, male sex, and a Crohn's disease diagnosis all tracked with higher risk across the board, while Black race, other race, and female sex were protective, and psoriasis or psoriatic arthritis specifically was associated with lower risk of basal cell carcinoma. The authors are upfront about limitations — it's retrospective, from a single health system, and while they tried to account for concurrent immunosuppression and phototherapy, documentation of individual-level risk factors like lifetime UV exposure or tanning bed use simply wasn't reliably captured, so residual confounding is a real possibility.

Practical takeaway here leans toward practice-relevant, though with appropriate caution given the observational design. This is a large, long-follow-up, chart-validated cohort that reinforces what many of us already suspected clinically — TNF inhibitor exposure appears to meaningfully raise both nonmelanoma skin cancer and melanoma risk, though without pushing patients toward more aggressive tumor biology. For those of us doing skin cancer screening and surveillance in patients on long-term TNF inhibitor therapy, particularly older, White, male patients with Crohn's disease, this supports being more vigilant about screening intervals, even though it doesn't yet tell us the optimal interval or justify wholesale changes to national guidelines on its own.

That wraps our four articles for this issue — a scribe workflow report favoring high-volume practices like ours, a strong retrospective challenge to routine cartilage grafting in alar reconstruction, a thoughtful exchange on the limits of AI-simplified patient education, and a large cohort reinforcing the skin cancer risk of TNF inhibitor therapy. Thanks for listening, and we'll see you next month.