Two days before its Presidential Symposium presentation, the INTerpath-001 melanoma vaccine data leaked. Merck and Moderna released the results on October 22, a day after ESMO introduced its Principles for Humanising Scientific Reporting. So I walked into the Alicante Auditorium on Saturday with the suspense gone and one question left: what would survive contact with Monday morning clinic?
Three Trials That Change My Monday Conversations
DeLLphi-305 was the result that changed my thinking the most. In extensive-stage small cell lung cancer without progression after durvalumab plus platinum-etoposide, adding tarlatamab to durvalumab maintenance improved median overall survival (OS) from 12.7 to 21.1 months at the interim analysis, measured from maintenance randomization (hazard ratio [HR], 0.42). That changes what I can reasonably tell a patient who reaches maintenance. Cytokine release syndrome occurred in roughly half of patients, almost entirely grade 1 or 2. The result gives us a reason to build the monitoring pathway now. Across fifteen hospitals, I know the distance between a trial's observation protocol and a community infusion center. Closing that distance will decide who benefits.
Krascendo 1 brought a direct comparison in previously treated KRAS G12C-mutated non-small cell lung cancer. Investigational divarasib improved median OS to 17.7 versus 12.0 months (HR, 0.56), with fewer severe treatment-related events than the pooled sotorasib/adagrasib control. As a site principal investigator on a next-generation RAS inhibitor trial, I see the promise. Access remains the problem I cannot solve from the clinic alone.
TroFuse-005 addressed a familiar dilemma: endometrial cancer progressing after platinum and immunotherapy. Sacituzumab tirumotecan improved median OS to 19.0 versus 12.1 months with doxorubicin or paclitaxel (HR, 0.68). Stomatitis and cytopenias require preparation. Its FDA target action date is June 15, 2027. These results change counseling and planning now; prescribing must still account for approvals and access.
Two Results That Made Me Slower
INTerpath-001 delivered a meaningful result: adding personalized intismeran autogene to pembrolizumab after melanoma resection improved two-year recurrence-free survival from 70.0% to 82.2% (HR, 0.58). OS remains immature. The presentation data also showed that inadequate tissue excluded 17.3% of screened patients; median time to the first vaccine dose was 64 days. I would gladly enroll eligible patients in trials. But adequate tumor tissue and timely manufacturing remain practical requirements. A positive Kaplan-Meier curve does not build the pathway that gets an individualized vaccine to a patient.
MOIO asked whether patients doing well after six months of immunotherapy could move to infusions every 12 weeks. The interim analysis did not establish noninferiority, and a concerning survival signal prompted early stopping. It did not formally establish inferiority either. The practical conclusion is narrower, and sufficient: these data do not support that schedule in this population. Convenience still has to earn its evidence.
The Comparator and the Biomarker Threshold
CLARITY-Gastric01 deserves both enthusiasm and precision. Sonesitatug vedotin improved OS in third-line-or-later CLDN18.2-expressing gastric, gastroesophageal junction, or esophageal adenocarcinoma (HR, 0.47). Its other primary endpoint, progression-free survival (PFS), missed significance in the broader second-line-or-later population. Those are different populations; the second-line question remains less settled.
The testing lesson matters just as much. CLARITY enrolled tumors with membranous CLDN18.2 staining in at least 25% of tumor cells at any intensity versus zolbetuximab's approved threshold of at least 75% with moderate-to-strong staining. This echoes the HER2-low story in breast cancer: eligibility depends on the treatment and its evidence. I want the percentage and staining intensity in the pathology report. A binary label may no longer tell us enough. The lower CLARITY cutoff is trial evidence, not an interchangeable approved threshold.
KEYNOTE-C93 improved PFS with pembrolizumab alone versus chemotherapy in first-line dMMR endometrial cancer, without establishing an OS advantage at this analysis. But it did not compare monotherapy with today's chemo-immunotherapy standard. The question I need answered is whether we can safely omit chemotherapy from that combination. This trial does not settle it.
What I Take Back to Clinic
My benchmark is meaningful patient benefit against an appropriate comparator, with a delivery pathway that works. ESMO's inaugural AI track made that question relevant to my own field; the point here is accountability. The AI presentations I reviewed left substantial questions about prospective clinical benefit. I want our own systems tested against the care they are meant to improve, using outcomes that matter to patients.
Monday's conversations change, and so do the questions I ask pathology and our infusion teams. Fewer visits cannot simply be assumed safer. A biomarker threshold cannot be separated from the treatment it selects. The standard I brought home is straightforward: show the benefit, then show how it reaches the patient.
Dr. Loaiza-Bonilla has received consulting fees or honorarium from Astra Zeneca and Exelixis. He has received payment for lectures from Guardant Health and CARIS. He has received stock/stock options from Massive Bio.
