This Week in Oncology — Sep 30, 2026

This AudioScholar briefing is generated by artificial intelligence for healthcare professionals and trainees. It is not medical advice.

Welcome to This Week in Oncology. This week we're covering 10 notable papers spanning local therapy de-escalation and delivery in breast and brain disease, immunotherapy biomarkers and combinations, and a set of population-level studies on prevention, screening intensity, and how we report what treatment costs patients. Let's dive in.

We start with local therapy, where two randomised trials ask how radiation should be delivered and one prospective cohort asks how much surgery the axilla still needs. In The Lancet Oncology, Sun and colleagues report ten-year outcomes from the Chinese hypofractionated postmastectomy radiotherapy trial, which randomised 810 analysable women with high-risk disease, almost all of them stage III, to either 50 Gray in 25 fractions over five weeks or 43.5 Gray in 15 fractions over three weeks, covering chest wall and supraclavicular region after modified radical mastectomy and axillary dissection [1]. With a median follow-up of about eleven and a half years, cumulative locoregional recurrence was about 10 percent with conventional fractionation and about 12 percent with the shorter course, a difference the trial could not distinguish from chance, and the authors are explicit that this long-term analysis carried no additional confirmatory non-inferiority testing. Severe late toxicity was rare and essentially identical between arms, with ischaemic heart disease and lymphoedema each affecting roughly one percent of patients, and no grade 4 or 5 events and no brachial plexopathy in either group. This is the durable safety and efficacy signal that regional nodal hypofractionation has been waiting for, and it supports the three-week schedule as a time-efficient option in this high-risk postmastectomy population.

Staying with radiation, the ROADS trial in the Journal of Clinical Oncology tested whether radiation delivered at the moment of resection beats waiting for stereotactic radiotherapy afterwards. Weinberg and colleagues randomised 230 patients across 32 United States centres with a newly diagnosed brain metastasis requiring resection to either caesium-131 collagen tile implantation at surgery or post-operative stereotactic radiation, with 204 patients in the modified intention-to-treat analysis and a median follow-up of about thirteen months [2]. Surgical bed recurrence was dramatically less frequent with the tiles, with the median time to recurrence not reached versus just over seventeen months in the stereotactic arm, and surgical bed recurrence-free survival roughly halved the event rate in favour of brachytherapy, with a median not reached versus about eleven months. Overall adverse events looked similar, at roughly four in five patients in each arm. The trial met both non-inferiority and superiority, which is a strong result, though follow-up is still short, overall survival is not reported here, and the technique requires a centre equipped to implant the tiles intraoperatively.

Also in the Journal of Clinical Oncology, Barrio and colleagues asked whether sentinel node biopsy can stage the axilla in locally advanced breast cancer that converts to clinically node-negative after neoadjuvant chemotherapy, a group historically excluded from de-escalation studies [3]. Among 145 enrolled patients, close to half of the cohort had inflammatory breast cancer, and all went on to completion axillary dissection. At least one sentinel node was found in about three quarters of patients, and among the 87 patients with adequate mapping there was a single false negative, giving a false-negative rate of about 2 percent. That is reassuring, but the estimate rests on 47 patients with residual nodal disease, the uncertainty around it is wide, and the identification rate of 74 percent means a quarter of patients still cannot be staged this way. The authors frame it as suggesting sentinel node biopsy may be appropriate after downstaging, not as settled practice.

Turning to immunotherapy, three papers address who benefits, what to combine, and an unexpected potential modifier. In Annals of Oncology, McDermott and colleagues report the final analysis of KEYNOTE-427, the single-arm phase two study of first-line pembrolizumab monotherapy in advanced renal cell carcinoma [4]. After about five years of follow-up, the objective response rate was roughly 36 percent in clear cell disease and about 27 percent in non-clear-cell disease. The biomarker work is the real contribution: a T-cell-inflamed gene expression profile and PD-L1 combined positive score were each positively associated with response, with PD-L1 also tracking with progression-free and overall survival in the non-clear-cell cohort, while driver gene mutations showed no consistent pattern. Patients with detectable circulating tumour DNA at baseline had a median overall survival of about 30 months compared with nearly 54 months in those without. These are exploratory analyses from a single-arm study, so they generate hypotheses for biomarker-driven selection rather than justifying it.

A more cautionary combination result comes from Clinical Cancer Research, where Fidelman and colleagues combined pembrolizumab with lutetium-177 DOTATATE peptide receptor radionuclide therapy in 26 patients with high-risk, well-differentiated neuroendocrine tumours, most of them grade 3 [5]. The response rate was about 35 percent with a median progression-free survival of just over 13 months, but serious adverse events occurred in just over half of patients, higher than expected for radionuclide therapy alone, and the authors conclude plainly that the added benefit of checkpoint inhibition here remains unclear. Responders did show higher baseline proliferating CD8 T cells and fewer myeloid-derived suppressor cells, which is mechanistically interesting in a 26-patient pilot and nothing more.

Then there is the week's most provocative association, in PLOS Medicine. Chen and colleagues used a target trial emulation in a federated electronic health record network, comparing patients on checkpoint inhibitors with depression or anxiety who were taking a selective serotonin reuptake inhibitor against those taking a benzodiazepine, matching 1,567 pairs on 49 covariates [6]. Two-year all-cause mortality was about 24 percent with an SSRI versus about 34 percent with a benzodiazepine, roughly a 37 percent lower hazard, consistent across all five individually analysable drugs, with slightly more immune-related adverse events driven by thyroid dysfunction and no excess hepatitis, pneumonitis, or colitis. Tumour transcriptomic data showed serotonin transporter expression inversely correlated with T-cell inflammation in 13 of 20 cancer types. The authors are appropriately blunt about the central weakness: performance status is not captured, and benzodiazepines are prescribed in quite different clinical circumstances, so confounding by indication could plausibly generate an effect of this size. This is a hypothesis for prospective testing, not evidence for repurposing.

Finally, four papers on prevention, screening intensity, and the quality of our evidence. In The Lancet Oncology, Rumgay and colleagues estimate that about 2.3 million new cancer cases in 2024, roughly 12 percent of the global total, were attributable to twelve infectious agents, led by Helicobacter pylori at around 760,000 cases and human papillomavirus at about 750,000, with eastern Asia accounting for well over 40 percent of the worldwide burden [7]. The practical message is that a large share of this burden is addressable with existing vaccination, testing, and treatment tools, with Epstein-Barr virus standing out as the agent lacking any preventive option. On the screening side, JAMA Internal Medicine published a retrospective cohort by Bryant and colleagues covering 1.41 million male veterans dually enrolled in Veterans Affairs and fee-for-service Medicare [8]. Among men aged 80 and older, prostate-specific antigen testing rates in Veterans Affairs primary care were about 30 percent lower than in fee-for-service Medicare and biopsy rates about a third lower, with lower testing also seen in men with under five years of predicted life expectancy, while in men aged 65 to 69 Veterans Affairs testing was modestly higher. That pattern suggests an integrated system aligning screening intensity with likelihood of benefit, though this is observational and cannot isolate the mechanism.

Two papers address the infrastructure of cancer care. In the Journal of Clinical Oncology, Wilke and colleagues cluster-randomised probands with BRCA1 or BRCA2 pathogenic variants to facilitated cascade testing with navigation support versus a standard family letter, and found that testing uptake among first-degree relatives at six months rose from about 51 percent to about 73 percent, reaching 90 percent in the intervention arm by 18 months, with close to half of all tested relatives carrying a pathogenic or likely pathogenic variant [9]. Free testing was available to both arms, so the gain came from navigation, not cost. And in The Lancet Oncology, Tannock and colleagues, writing for Common-Sense Oncology and the European Organisation for Research and Treatment of Cancer, recommend that adjuvant therapy trials report the proportion of patients experiencing a prespecified deterioration in validated quality-of-life scales and how long that deterioration lasts, included in the primary report rather than a later companion paper [10]. Given that many adjuvant regimens buy small absolute survival gains, this is a reasonable argument that the harm side of the ledger is currently under-reported.

If you only have time for one paper this week, make it the ten-year analysis of the Chinese hypofractionated postmastectomy radiotherapy trial in The Lancet Oncology [1]. It addresses the question that has kept regional nodal hypofractionation contested — whether shortening the course quietly costs late cardiac, brachial, or lymphatic toxicity a decade out — and the answer, in a high-risk, largely stage III population, is that it does not appear to.

Here is what this week's evidence adds up to in Oncology. Long-term randomised data now support three-week postmastectomy radiotherapy with nodal coverage as durably effective and no more toxic than five weeks, at least in a predominantly stage III Chinese cohort [1]. Intraoperative caesium tile brachytherapy outperformed post-operative stereotactic radiation for surgical bed control in resected brain metastases, a positive superiority result that still needs longer follow-up and survival data before it displaces the standard [2]. In immunotherapy, gene expression and PD-L1 measures remain associated with benefit in renal cancer but only exploratorily [4], adding checkpoint blockade to radionuclide therapy in neuroendocrine tumours produced more toxicity without clear added benefit [5], and the striking survival association with selective serotonin reuptake inhibitors is observational and vulnerable to confounding by performance status [6]. On the system side, roughly one in eight cancers worldwide is infection-attributable and largely preventable with existing tools [7], navigation more than cost explains who completes cascade genetic testing [9], and both the Veterans Affairs screening data [8] and the quality-of-life reporting recommendations [10] point at the same theme, which is matching the intensity of what we do to the benefit patients can actually expect.

That's your roundup for This Week in Oncology. The full transcript and references are available on the episode page in your AudioScholar library. This is an AI-curated summary — for clinical decisions, always consult primary sources and current guidelines. See you next week.

If this weekly briefing is useful, follow the show in your podcast app so new episodes arrive automatically. And if it saves you time, a quick rating in your podcast app helps other physicians find the show.

This is an automated summary generated by artificial intelligence, which can make mistakes. Always review the original source materials.