Welcome back to the journal review. This is the November twenty twenty-six issue of the Journal of the American Academy of Dermatology, and we've got four pieces to get through today — one prospective diagnostic study, one original retrospective cohort study on radiotherapy for Merkel cell carcinoma, and a letter-and-response pair that's actually a terrific worked example of how to argue about attrition bias in a small trial. Let's get into it. First up is a prospective multicenter study out of Italy looking at clinical and dermoscopic triage of head and neck macules, specifically asking what reflectance confocal microscopy adds when you layer it onto dermoscopy for those maddening flat facial pigmented lesions where you're trying to rule out lentigo maligna without just biopsying everything on someone's face. The background here is familiar territory for anyone doing Mohs on the head and neck: solar lentigines, lichen planus-like keratosis, seborrheic keratosis, and flat nevi all sit in the same visual space as early lentigo maligna, and dermoscopy alone — per the meta-analysis they cite — runs around seventy percent sensitivity and eighty percent specificity for LM and lentigo maligna melanoma. That's not good enough when the alternative to a correct diagnosis is either an unnecessary disfiguring biopsy in a cosmetically sensitive zone or, worse, a missed melanoma treated with laser or cryotherapy that then delays real diagnosis. So the gap is clear: can confocal microscopy, used as an adjunct rather than a replacement, meaningfully improve that triage. Methodologically, this was a real-world prospective study across three Italian academic dermatology centers, enrolling consecutive equivocal flat facial and neck macules over about a two and a half year window, ultimately landing on just over two thousand lesions from eight hundred and one patients. Every lesion went through a standardized pipeline: clinical exam, dermoscopy with a grading scale from one — certainly benign — up through four — highly suspicious — then handheld confocal microscopy on anything equivocal, with excision or biopsy recommended for anything called malignant by either modality, and digital dermoscopic follow-up of at least a year for the double-negative lesions, meaning both dermoscopy and confocal called it benign. Histopathology served as the reference standard for anything biopsied, and the follow-up arm served as a pragmatic proxy reference standard for the lesions left alone. This is a sensible design choice given the clinical question — you obviously cannot ethically excise two thousand benign-appearing facial macules just to get a gold standard on everyone, so the follow-up-as-surrogate-truth approach is the practical compromise, though as the authors acknowledge, it does introduce verification bias, since lesions that look very benign are systematically less likely to ever get a histologic answer. Statistically this is mostly descriptive — sensitivities, specificities, predictive values, and a number-needed-to-excise calculation — but they do use Cohen's kappa to quantify agreement between dermoscopy and confocal, and separately between confocal and histology. Worth a quick refresher since kappa shows up constantly in diagnostic imaging papers: raw percent agreement between two raters or two tests is misleading because some agreement happens by chance alone, especially when one category — benign — is overwhelmingly common. Kappa corrects for that chance agreement mathematically, giving you a value from zero to one where anything above point six is generally considered substantial agreement and above point eight is near perfect. Here, dermoscopy and confocal agreed on about eighty-eight percent of lesions, with a kappa around point six seven, which lands in the substantial range — reassuring that the two modalities are largely reading from the same page, but the disagreement in that remaining twelve percent of lesions is exactly where confocal is earning its keep. So what did they find. After integrating confocal into the workflow, about twelve percent of all lesions ended up managed as malignant. Among the roughly one in eleven lesions dermoscopy had called equivocal or probably benign that confocal reclassified, you see the real value proposition: confocal pulled an additional two percent of all lesions out of the benign pile and correctly flagged them as LM or LMM that dermoscopy had been prepared to simply follow. Conversely, of the lesions dermoscopy flagged as probably BCC or squamous cell carcinoma, confocal downgraded nearly half back to benign — a huge reduction in planned biopsies. Of the double-negative lesions sent to a year of follow-up, about two in five hundred changed during that period, and only four lesions, a vanishingly small 0.4 percent, turned out to be malignant on eventual removal — three lentigo maligna and one basal cell carcinoma. On the lesions that did go to histology, dermoscopy alone caught only about fifty-five percent of LM and LMM, while adding confocal pushed sensitivity up to nearly ninety-two percent. For BCC and squamous cell carcinoma, dermoscopy was already decent at ninety percent sensitivity, and confocal nudged that further to about ninety-four percent. The authors' conclusion is that confocal integration meaningfully improves sensitivity for melanocytic malignancy detection while also improving specificity for benign lesions, which is the best of both worlds — fewer missed melanomas and fewer unnecessary excisions, with a number-needed-to-excise framework supporting that this combined pathway is efficient. The honest limitations are real though: this is a selected population of already-equivocal lesions from expert centers with investigators skilled in both dermoscopy and confocal interpretation, so this generalizes to a confocal-experienced practice, not necessarily to a general dermatology clinic without that training curve. There's substantial loss to follow-up in the conservatively managed group — over a third of the double-negative lesions sent for one-year monitoring never came back, which means that reassuring 0.4 percent residual malignancy rate among the ones we did see could be an underestimate if the lost patients were disproportionately the ones with something going on. And this remains a single-reference-standard problem where follow-up, not histology, serves as truth for the majority of lesions. Practically, for those of us doing Mohs and managing facial pigmented lesions, this is useful confirmatory data rather than purely novel: it reinforces that confocal microscopy, in experienced hands, is a legitimate noninvasive adjunct for exactly the population you're most reluctant to biopsy reflexively — thin facial macules in older patients where a scar matters and where LM is notoriously difficult to diagnose confidently on dermoscopy alone. If you have confocal access or work with a referring dermatologist who does, this supports using it specifically to triage the equivocal grade two and three lesions rather than as a blanket screening tool. It is not yet practice-changing in the sense of altering margin protocols or biopsy thresholds broadly, but it is a solid, real-world-scale validation that should increase your confidence in a negative confocal read before committing a patient to a year of watchful dermoscopic follow-up. Next, a genuine original study: a single-center retrospective cohort out of the University of Washington and Fred Hutch, asking whether a single eight gray fraction of postoperative radiotherapy is a viable alternative to conventional fractionated radiotherapy for Merkel cell carcinoma excised with narrow margins. The clinical problem is one you know well — Merkel cell carcinoma recurs locally in a meaningful minority of stage one and two patients, somewhere between two and twenty-seven percent depending on risk factors and surgical approach, and national guidelines recommend adjuvant radiotherapy when risk factors like head and neck location, tumor over a centimeter, immunosuppression, lymphovascular invasion, or positive margins are present. The problem is that conventional postoperative radiotherapy is roughly fifty gray delivered across twenty-five fractions over five weeks, which is brutal for a population that skews elderly, often frail, and often needing to travel repeatedly for treatment. The group had previously shown that single-fraction eight gray radiotherapy works well palliatively for symptomatic Merkel lesions, and a small prior series suggested it might work adjuvantly too. This study is the first to formally compare postoperative single-fraction radiotherapy against both observation and conventional radiotherapy for local control. Design-wise, this pulled from a long-running prospective Merkel cell carcinoma repository at Seattle, then applied a retrospective cohort analysis to three groups who weren't randomized but rather clinically selected — observation for lower-risk patients or those with contraindications to radiation, conventional radiotherapy for the standard pathway, and single-fraction radiotherapy offered specifically to patients who had risk factors but couldn't or wouldn't tolerate the full conventional course. That's an important asymmetry to flag before even getting to results: this is about as classic a confounding-by-indication setup as you'll see, since the single-fraction group was deliberately enriched for older, less fit, often head-and-neck patients who were steered away from the conventional regimen. For the statistics, they used the Aalen-Johansen estimator rather than a standard Kaplan-Meier curve, and this is worth actually understanding rather than glossing over. Kaplan-Meier assumes that when you don't observe the event of interest, the patient was simply censored — lost to follow-up or study ended — and it treats all non-events the same way. But in oncology, patients also die of other causes before they could ever recur locally, and plain Kaplan-Meier actually inflates the apparent cumulative risk of recurrence in that setting because it mathematically redistributes those competing deaths as if they were still at risk of recurring. The Aalen-Johansen estimator is the competing-risks-aware version — it explicitly partitions the probability space between the event you care about, local recurrence, and the competing event, death from any cause, so the cumulative incidence doesn't get artificially inflated by patients who died of something unrelated before they had the chance to recur. Given that Merkel cell patients are often elderly with competing mortality risks, this was the right tool rather than a cosmetic statistical choice. They also modeled treatment as a time-dependent covariate specifically to avoid immortal time bias — the trap where, if you count time retrospectively from surgery for someone who received radiotherapy forty days later, you accidentally credit the radiotherapy group with forty days of guaranteed recurrence-free survival they hadn't earned yet. Treating radiotherapy status as something that switches on at the actual treatment date, rather than baseline, is the statistically correct way to handle delayed treatment initiation, and they further confirmed robustness with a landmark sensitivity analysis. On to results. Despite having fewer risk factors per patient on average than either treatment group, the observation cohort still had worse three-year local control, around ninety-one percent, compared to about ninety-seven percent for conventional radiotherapy and a perfect one hundred percent for single-fraction radiotherapy — both differences statistically significant versus observation, and single-fraction was statistically indistinguishable from conventional. The effect was concentrated almost entirely in patients with narrow surgical margins of a centimeter or less: in that subgroup, local control with observation alone dropped to under ninety percent, while both radiotherapy approaches held at one hundred percent. In patients with wide margins, all three strategies performed similarly well above ninety-seven percent, meaning radiotherapy likely adds little when margins are already generous — an intuitive and clinically sensible finding. Toxicity with single-fraction radiotherapy was reassuringly mild: about four in five patients had no adverse events at all, and the remaining one in five had only grade one effects, mostly transient erythema — nothing resembling the mucositis, xerostomia, or ocular toxicity burden historically described with conventional courses. The authors' own interpretation is appropriately measured — they frame single-fraction radiotherapy as feasible, well tolerated, and achieving local control comparable to the conventional regimen, specifically positioning it for the narrow-margin population where adjuvant treatment seems to matter most. The limitations are the ones you'd expect from a single-center retrospective design with nonrandomized treatment selection: the single-fraction group was systematically older, more often head and neck primaries, and selected precisely because they were less able to tolerate conventional treatment, so there's real potential for the groups to differ on unmeasured factors like overall frailty or tumor biology that could influence both treatment choice and outcome. Follow-up duration also differed substantially between groups, longest for conventional radiotherapy and shortest for single-fraction, which matters when you're claiming comparable durability. And sample size in the single-fraction arm is modest at forty-three patients, with zero recurrences — reassuring, but a small-number result that deserves cautious interpretation rather than over-interpretation as definitively equivalent. For practical takeaway: this is genuinely interesting and practice-informing, but I'd stop short of calling it fully practice-changing yet, and the authors say as much themselves — they explicitly call for prospective multicenter validation. That said, for your narrow-margin Merkel cell patients who are elderly, frail, or logistically unable to complete five weeks of daily radiotherapy, this gives you real retrospective data to have an informed conversation about single-fraction radiotherapy as a reasonable, low-toxicity alternative to either observation or the full conventional course, particularly in a multidisciplinary tumor board context. It will not change what you do with wide-margin excisions, where this data suggests radiotherapy of any kind adds little. Finally, a short but genuinely instructive methodological back-and-forth — a letter to the editor and the authors' response, both concerning a previously published study on electronic skin surface brachytherapy for basal cell carcinoma, specifically its five-year cosmesis outcomes. The letter, from Li and Wang, raises a classic epidemiologic concern: attrition bias. In the original study, cosmesis data were available for ninety-four percent of patients at twelve weeks but only fifty-nine percent of patients by five years — meaning over forty percent of the original cohort simply didn't contribute to the headline finding that eighty percent of patients maintained good cosmesis at five years. The letter writers point out that late complications like hypopigmentation and telangiectasia in this cohort tended to emerge only after a year and persisted thereafter, raising the concern that patients with visible cosmetic problems might have been precisely the ones who stopped returning for follow-up, which would make the reported eighty percent an overestimate of true long-term cosmetic success. They run a best-case, worst-case sensitivity exercise showing the true five-year good-cosmesis rate could plausibly range anywhere from under half to just under ninety percent depending on what happened to the fourteen missing patients — a genuinely wide range — and they recommend, reasonably, that future studies report reasons for attrition, compare baseline characteristics between completers and dropouts, and run formal sensitivity analyses. The original authors, Barker and Zhang, respond with a pointed and worth-understanding rebuttal. Their core argument is that the five-year cosmesis figure was never a prespecified primary endpoint in the first place — the actual statistical design centered on patient-reported cosmesis at three months and local control at three years, both of which met their enrollment and attrition assumptions, with three-year attrition actually coming in lower than anticipated. The five-year numbers were explicitly descriptive, intended to add long-term context, not to support formal inference. They also push back on the letter's implied mechanism, noting that among the eleven patients who didn't contribute five-year data, most were accounted for by death from unrelated causes, administrative study closure, or relocation, with only four truly lost for unknown reasons — meaning the dropout wasn't obviously driven by patients hiding bad cosmetic outcomes. And they make a broader statistical point worth internalizing: a best-case-worst-case extreme-value sensitivity analysis defines the mathematical boundaries of what's possible, but it says nothing about what's plausible, and presenting a wide theoretical range isn't itself evidence that bias actually occurred or in which direction. There's no new patient data in either piece, so there's no results or takeaway beat in the usual sense — but as a pair, this exchange is a nice teaching moment in itself about the difference between a prespecified primary endpoint and a descriptive secondary observation, and about how to correctly weigh a theoretical sensitivity analysis against actual reasons for dropout. The practical lesson for your own reading of the literature: when you see a striking long-term descriptive percentage in a small study, your first move should be to check the denominator and ask whether that endpoint was designed and powered for, or just reported along the way — and if it's the latter, treat it as hypothesis-generating context rather than a number you'd quote to a patient. That wraps our four articles for this November issue. The confocal microscopy study reinforces an increasingly evidence-backed role for noninvasive imaging adjuncts in equivocal facial pigmented lesions; the Merkel cell carcinoma series offers a promising but not yet definitive low-toxicity radiotherapy alternative for your narrow-margin patients; and the brachytherapy letter exchange is a good reminder to read long-term outcome claims with an eye on exactly what was prespecified versus what was simply observed along the way. Thanks for listening, and I'll see you next month.