onc brain

The longer read

The value of LAVA-CRLM is that it exists at all. Until now, choosing between thermal ablation and SBRT for a small colorectal liver metastasis rested on cross-series comparison, where MWA cohorts were reported with local control above 90% and SBRT cohorts spread across a range from 50% to 95%. That spread almost certainly reflected dose, lesion selection, and biology rather than modality, and a randomised comparison in lesions judged suitable for both is the only design that removes the selection that made those series incomparable. The eligibility rule requiring both modalities to be feasible is what preserved equipoise, and it is also what narrows the result to lesions any competent ablation service could reach.

The efficacy read should be held loosely. The sample size assumed a difference corresponding to HR 0.38, far larger than anything plausible between two ablative local therapies, so the trial was built to detect a gap it was never likely to find. The observed HR 0.87 with an interval from 0.43 to 1.74 is compatible with SBRT being clearly better, clearly worse, or equivalent. The authors are explicit that absence of difference is not evidence of equivalence, and a reader who takes the headline as 'SBRT equals MWA' is reading more than the design supports. The per-protocol estimate moving toward SBRT (86.0% vs 77.7%) is directionally reassuring after five crossovers were removed, but it is a secondary analysis on smaller numbers.

Two features cut in SBRT's favour when judging whether its local control estimate is a floor or a ceiling. First, dosing was uniformly ablative, with BED10 above 173 Gy to GTV, well past the threshold the authors cite from Kang for high 1-yr control, so this is not an underdosed SBRT arm losing to a well-executed ablation. Second, KRAS-mutant disease was nearly twice as common in the MWA arm, and the Hong data the authors cite put KRAS-mutant lesions at markedly worse SBRT local control. If that imbalance had run the other way, SBRT would likely have looked worse; as it ran, the MWA arm carried the harder biology. Neither point rescues the power problem, but both argue against reading the SBRT local control as flattered.

The firmer signal is toxicity. SBRT produced one G2+ acute event against thirteen after MWA, and every G3+ event, including a death from hepatic necrosis and perforation, followed ablation. The Fisher test at p=0.13 reflects small numbers rather than a trivial difference, and the mechanism is intuitive: MWA requires general anaesthesia and probe placement, with some cases needing open or laparoscopic access. Late G2+ cumulative incidence also favoured SBRT numerically (Fine-Gray HR 0.51), though the interval crosses 1. For an MDT weighing an elderly or comorbid pt, or a lesion near the capsule, this is the result that changes the conversation.

What the trial leaves open is transferability. Most SBRT was delivered on an MR-linac with daily adaptation, so centres treating on a conventional linac with fiducials cannot assume the same OAR-sparing or target coverage. No treated lesion exceeded 3.0 cm, leaving the intermediate-size question to COLLISION-XL. And the findings would be wrong if the true local control gap in favour of MWA sits near the lower end of the interval; only an adequately powered non-inferiority trial, ideally stratified by KRAS status, can close that.