onc brain

The longer read

The question MAVERICK answers is narrower and more useful than 'does PCI work.' PCI reduces brain metastases, and the conclusions slide confirms that again here. What was never settled is whether that reduction is worth its cognitive cost once staging MRI and serial MRI surveillance can find brain metastases early enough for effective salvage. In ES-SCLC, the Takahashi Japanese trial had already shifted many practices toward surveillance. In LS-SCLC, the PCI standard still rests on randomized data and meta-analysis from an era of CT staging without routine brain MRI. MAVERICK is the first randomized trial to put LS-SCLC with MRI staging into that comparison, and LS-SCLC made up 68% of the enrolled population.

The primary endpoint needs to be read carefully, because it largely determines the result. Cognitive failure-free survival counts death and cognitive failure as events. If survival does not differ between arms, and the preliminary OS analysis after 128 events suggests it does not, then the CFFS difference is almost entirely a cognitive-toxicity difference. That is a legitimate endpoint for a de-escalation question. But the trial is a demonstration that PCI costs cognition without a detectable survival return, not a demonstration that surveillance is oncologically superior. The statistics follow the same logic: a 90% CI and a one-sided p=0.0005 are appropriate for a directional hypothesis, and with an HR of 0.60 the result is not fragile.

For radiation oncologists, the most consequential detail is that hippocampal avoidance did not rescue PCI. 77% of the PCI arm received HA-PCI, and the HA-PCI-only subgroup still favored MRI alone. That removes the most common counterargument, that modern technique would close the gap. Stage also did not modify the effect (LS HR 0.59, ES HR 0.65, interaction p=0.79), which argues against keeping PCI for LS-SCLC on the grounds that curative intent raises the value of preventing intracranial relapse.

Several features should temper confidence. OS is immature, and the final analysis is planned at 190 events. A late OS divergence favoring PCI in the LS subgroup, where long-term survivors live long enough for intracranial control to matter, would reopen the question. The source does not quantify the brain metastasis reduction, and it does not describe the symptomatic burden, SRS versus WBRT use, or neurologic morbidity of salvage in the surveillance arm. Those numbers determine whether surveillance trades toxicity now for toxicity later. ADRIATIC established consolidation durvalumab for LS-SCLC during enrollment, and only 41% of pts received immunotherapy upfront. How the result carries over to a durvalumab-consolidated LS population, whose intracranial relapse pattern may differ, is inferred rather than tested.

The practical upshot: for a pt with a clean post-induction brain MRI, the burden of proof now sits with PCI rather than with surveillance, in both stages. That assumes a surveillance program the trial actually delivered, with MRI at 3, 6, 9, 12, 18 and 24 months. A practice that cannot reliably deliver that schedule is not running the tested strategy. The final OS analysis, with PFS and salvage-RT patterns reported in full, is what would turn this from a strong challenge to the PCI standard into a settled replacement.