Welcome back to the journal review. This is the August 2026 issue of JAMA Dermatology, and we've got two pieces worth your time today — a neoadjuvant chemotherapy case series out of Montreal, and a large ecological study out of Germany asking a question that's been quietly unresolved for almost two decades: does population-based skin cancer screening actually save lives from melanoma. Let's get into it.

First up is a brief report titled "Neoadjuvant Capecitabine in Advanced Cutaneous Squamous Cell Carcinoma of the Head and Neck." This is a prospective case series, not a randomized trial, so keep your expectations calibrated accordingly — but it's addressing a real problem you've all felt downstream of: surgical wait times. The authors, a McGill-affiliated group working across two Montreal academic centers, note that cutaneous squamous cell carcinoma of the head and neck is rising sharply in incidence, and that delays beyond thirty days from diagnosis to treatment are associated with tumor progression and worse outcomes. In a publicly funded system with constrained OR access — and frankly this applies plenty of places outside Canada too — patients with bulky, high-risk head and neck squamous cell carcinoma can sit on a waitlist long enough for a resectable tumor to become a much harder surgical problem. Neoadjuvant cemiplimab has been explored for exactly this reason, with response rates up to around seventy percent, but immune-related toxicity in the range of fifteen to twenty percent limits its use, particularly in solid organ transplant recipients, who are precisely the population most likely to develop aggressive cutaneous squamous cell carcinoma in the first place. So the authors turned to capecitabine, an oral fluorouracil prodrug already used in colorectal and breast cancer, with the appeal that its metabolism is preferentially activated within tumor tissue, theoretically sparing systemic toxicity.

The design here is straightforward: fifteen consecutive patients with advanced disease who were facing anticipated surgical delays received two cycles of capecitabine, one gram twice daily for fourteen days with a one-week rest, following what they call the CAPTURE protocol. Surgery followed within one to two weeks of completing treatment. There's no comparator arm and no randomization — this is hypothesis-generating by design, essentially establishing feasibility and signal before anyone commits to a phase two trial, which the authors mention is already underway. The primary outcome was stage migration — comparing clinical stage going in to pathologic stage coming out — which is a reasonable surrogate when your real question is "does this shrink the tumor enough to matter for the surgeon."

The results are genuinely encouraging for a first look. Two-thirds of patients, ten of fifteen, had clinical tumor regression. Of the twelve who went on to surgery, downstaging occurred in about six in ten, and five patients — forty-two percent of those resected — had a complete pathologic response, meaning no viable tumor at all on the specimen. Two patients responded so completely on clinical exam that they withdrew from surgery altogether and went straight to radiotherapy instead. On the flip side, five patients simply didn't respond, and one of those progressed to metastatic disease and was switched to cemiplimab. Notably, both of the immunosuppressed kidney transplant recipients in the cohort were non-responders — a small but clinically relevant signal given that this population is exactly who you'd want an alternative to checkpoint inhibition for. Response was characterized using a modified Ryan tumor regression grade, and interestingly, the poor responders clustered specifically in scalp primaries and one parotid nodal metastasis, suggesting site or biology may matter here. On safety, there were zero grade three or higher capecitabine-related toxicities across the cohort, which is a meaningfully cleaner profile than what's been reported for neoadjuvant immunotherapy.

The authors frame this as the first description of neoadjuvant capecitabine specifically in cutaneous squamous cell carcinoma of the head and neck, and their argument is that it offers a self-administered, outpatient, comparatively low-toxicity bridge for patients facing surgical delay, and potentially a viable option for transplant recipients where immunotherapy carries graft-rejection risk — though again, the two transplant patients here didn't actually respond, so that's aspirational rather than demonstrated. The limitations are exactly what you'd expect from a fifteen-patient single-arm series: no control group to know what these tumors would have done without any intervention, short follow-up averaging about a year, and no ability to assess durability or long-term recurrence. There's no formal statistical testing here — this is a descriptive case series, and appropriately so given the sample size, so no statistics lesson attached to this one.

For your practice, here's the honest takeaway: this is interesting, not yet practice-changing. It's a reasonable option to have in your back pocket when you're staring down a patient with a bulky head and neck squamous cell carcinoma and a surgical date that's months out, particularly if they're a poor immunotherapy candidate — but you're going to want to see the phase two randomized data the authors say is already in motion before this becomes a default bridging strategy. The toxicity profile is the most actionable piece of information here; if you're in a multidisciplinary tumor board discussion where cemiplimab is being considered for a transplant patient and the team is nervous about rejection risk, capecitabine is now a documented, low-toxicity alternative worth raising, with the caveat that response is not guaranteed and both transplant patients in this series didn't respond.

Now to the second article, "Population Skin Cancer Screening and Melanoma Mortality Rates," and this one is a genuine comparative effectiveness study with real statistical machinery behind it, so we'll spend a little more time on the methods.

The background: Germany rolled out a nationwide skin cancer screening program in 2008, offering adults thirty-five and older a visual whole-body skin exam — no dermoscopy required — every two years, covered by statutory and private insurance. Participation has hovered around one-third of the eligible population. The program was greenlit on the strength of a pilot project in Schleswig-Holstein called SCREEN, run in 2003 and 2004, which reported a mortality benefit. But in the years since the nationwide rollout, multiple studies have failed to reproduce that benefit at a national level, and this paper is another attempt to settle the question, this time using a cross-country comparison design.

Methodologically, this is an ecological study — meaning the unit of analysis is entire populations and regions, not individual patients. The authors pulled age-standardized melanoma mortality data for fifteen German federal states and nine neighboring countries without population-based screening — Austria, Belgium, the Czech Republic, Denmark, France, Luxembourg, the Netherlands, Poland, and Switzerland — covering 2009 through 2022. They deliberately excluded Schleswig-Holstein from the main analysis because of a documented, likely artifactual dip in mortality there tied to the earlier SCREEN pilot, which they didn't want contaminating the national signal. The logic of the design is essentially a natural experiment: if the German screening program is doing what it's supposed to do, Germany's mortality trend should be measurably more favorable than trends in comparable neighboring populations that never adopted population-wide screening.

Here's the stats lesson embedded in this one. For each region, they fit what's called a log-linear model to the yearly mortality rates, which just means they modeled the logarithm of mortality rate as a linear function of time — the reason you log-transform first is that mortality rates tend to change by a roughly constant percentage each year rather than a constant absolute amount, so fitting a straight line to the logged values gives you a clean, interpretable annual percentage change, or APC, for each region, essentially the average yearly percent drop or rise in mortality. Once they had an APC and a standard error for every individual region, they needed to pool those into two summary numbers — one for Germany, one for the neighboring countries — while properly accounting for the fact that some regions have huge, precise populations and others are small and noisy. That's exactly what a random-effects meta-analysis does: it's the same statistical tool you'd use to combine effect sizes across separate clinical trials, weighting each region's contribution by the inverse of its variance so that more precise, more populous estimates count for more, while also allowing for genuine between-region heterogeneity rather than assuming every state or country should have had an identical underlying trend. Then, to formally test whether the German pooled trend differed from the non-German pooled trend, they used what's called an omnibus test of moderators — essentially treating "German" versus "non-German" as a grouping variable inside one combined meta-analysis and asking whether that grouping explains a statistically significant amount of the variation in APCs. This is a more rigorous alternative to just eyeballing two separate confidence intervals and calling it a day, because it directly tests the difference rather than the two estimates in isolation.

So what did they find? Mortality declined everywhere over the observation period — that part is unambiguous, and it's good news broadly. But when pooled, Germany's average annual decline was about two percent, and the non-German control regions' average annual decline was a bit steeper, also around two percent — the two numbers were close enough that the difference was not statistically significant. If anything, the point estimate favored the control countries having the slightly better trend, which is the opposite direction of what you'd want to see if the screening program were driving an incremental benefit. They also split the data into two windows, before and after 2013, roughly bracketing the arrival of modern immune checkpoint and targeted BRAF and MEK therapies for metastatic melanoma — mortality actually ticked up slightly in both regions from 2009 to 2013, then fell more steeply from 2013 onward, consistent with those new systemic therapies driving mortality improvement independent of any screening effect. Critically, the gap between Germany and its neighbors was essentially the same, small, and non-significant in both time windows, which argues against the differences being explained by unequal access to newer drugs.

In their discussion, the authors are candid that this result lines up with essentially every other post-implementation evaluation of the German program — none have shown a national mortality benefit — and it directly conflicts with the original SCREEN pilot data that justified the whole rollout in the first place. They walk through several plausible explanations for that discrepancy: possible undercounting or miscoding of melanoma deaths in the original Schleswig-Holstein data, differences in screening protocol quality — the pilot referred higher-risk patients on to dermatologists in a way the nationwide program doesn't systematically enforce — a younger eligible age cutoff in the pilot, a dedicated public awareness campaign that accompanied the pilot but not the national rollout, and possibly a straightforward efficacy-to-effectiveness gap, citing a survey of general practitioners showing that genital, perianal, and oral mucosal skin were frequently skipped entirely during real-world screening exams.

The limitations section is refreshingly honest, and worth relaying in full because they matter for how much weight to put on this. First and most fundamental: this is an ecological design, which cannot establish causation at the individual level — that's the ecological fallacy, and the only way to truly answer this question would be a randomized trial comparing screened versus unscreened individuals, which nobody has done and probably never will at this scale. Second, participation in the German program is only around one-third of eligible adults, which dilutes any population-level signal even if the screening itself works for the people who show up — though the authors note a microsimulation study estimated that even at thirty percent participation, you'd expect roughly a thirteen percent relative mortality reduction over ten years, corresponding to a meaningfully bigger APC difference than what they actually observed, so low participation alone doesn't fully explain away the null result. Third, informal screening activity in the control countries — specifically the pan-European Euromelanoma campaign, which offers free annual skin checks — may be diluting the contrast between "screened" Germany and "unscreened" neighbors, meaning this study can only detect a benefit that exceeds whatever background screening effect already exists next door.

Practical takeaways here. This isn't a US-practice-changing paper in the sense of altering your Mohs or excision technique, but it's squarely relevant to how you counsel patients and how you think about population screening policy debates that inevitably come up in dermatology circles, including here. The clinically meaningful message is this: a national, age-gated, low-intensity visual screening program without individual risk stratification has now repeatedly failed to demonstrate a melanoma mortality benefit at the population level, even over fourteen years of follow-up and even against a backdrop of major systemic therapy advances that should make any residual screening benefit easier, not harder, to detect. That's a genuinely important data point if you're asked to comment on U.S. screening policy, guideline committees, or if patients ask you whether Germany's model is something we should import wholesale. It's not evidence that skin exams in high-risk individuals are useless — this study can't speak to targeted, risk-stratified surveillance in your own transplant or syndromic patients, which is a completely different clinical question with different evidence behind it. What it does suggest is that low-participation, non-risk-stratified, one-size-fits-all population screening is a blunt instrument that may not move the mortality needle enough to justify itself as currently implemented, and the authors' own conclusion is basically a call to figure out why before deciding whether to keep funding it as is.

That wraps this episode. Two very different papers — one a small, promising signal for a new bridging strategy in advanced cutaneous squamous cell carcinoma, the other a sobering, well-powered null result on population melanoma screening. Thanks for listening, and we'll see you next month.