Automated single-isocenter SRS for multiple brain mets: outcomes and margins
For4-10 brain mets <3 cm, frameless linac SRS, lung/breast/melanoma-dominant
TL;DR1y LC 89% per pt and 94% per lesion. Residual setup error caused coverage loss (V95 <95%) in 28 targets; simulated 1 mm GTV-PTV margin restored coverage.
Coverage loss clustered in small lesions far from isocenter: about 90% of V95 <95% misses were lesions <0.4 cm3 at ≥3.9 cm. Uncorrected intrafraction drift cost GTV coverage in 28% of lesions, so the 1 mm margin depends on 6D monitoring with mid-treatment correction.
For a pt with 4-10 small brain mets on a frameless linac with 6D ExacTrac and mid-treatment re-imaging, this supports a 1 mm GTV-PTV margin over 0 mm; it does not address mask setups without intrafraction correction or cumulative volumes ≥15 cm3.
Coverage loss clustered in small lesions far from isocenter: about 90% of V95 <95% misses were lesions <0.4 cm3 at ≥3.9 cm. Uncorrected intrafraction drift cost GTV coverage in 28% of lesions, so the 1 mm margin depends on 6D monitoring with mid-treatment correction.
16 details 4 trials watching
Retrospective two-center series (Rome, Negrar), 31 consecutive pts, 204 BM. Outcomes by Kaplan-Meier from SRS; dosimetric accuracy from offline CBCT fused to planning CT with residual shifts applied.
Age ≥18, 4-10 BM <3 cm. Median age 60; lung 14, melanoma 8, breast 5. Concurrent systemic therapy in 9 pts (IO 4, targeted 5).
Single-isocenter DCA via Brainlab Elements, 10 noncoplanar arcs at 5 couch positions. 20-22 Gy for <2 cm, 16-18 Gy for ≥2 cm or near critical structures; PTV = GTV + 1 mm (176 lesions) or 2 mm (28).
1y/2y LC 89%/77% per pt, 94%/86% per lesion; mOS 16.1 mo. Coverage loss (V95 <95%) in 28 targets, driven by distance ≥39 mm (P=.0001) and volume <0.40 cm3 (P=.015).
| Endpoint | 1 yr | 2 yr |
|---|---|---|
| LC, per patient | 89% | 77% |
| LC, per lesion | 94% | 86% |
| OS | 68% | 40% |
| Distant brain failure | 38% | n/r |
| Factor | Lesions losing coverage | P |
|---|---|---|
| Distance <39 mm vs ≥39 mm | 7 vs 21 | .0001 |
| Volume <0.40 cm3 vs ≥0.40 cm3 | 20 vs 8 | .015 |
| Diameter | 0 mm | 1 mm | 2 mm | 3 mm |
|---|---|---|---|---|
| 0.5 cm | 0.07 cm3 | 0.18 cm3 | 0.38 cm3 | 0.7 cm3 |
| 1 cm | 0.52 cm3 | 0.9 cm3 | 1.43 cm3 | 2.21 cm3 |
| 1.5 cm | 1.76 cm3 | 2.57 cm3 | 3.59 cm3 | 4.85 cm3 |
| 2 cm | 4.18 cm3 | 5.57 cm3 | 7.23 cm3 | 9.2 cm3 |
No G3-4 toxicity. Necrosis in 5 (16.1%) pts, 11% per lesion at 12 mo, 2 with G2 motor deficit; individual V12Gy >10 cm3 predicted it (P=.01).
Roper et al showed coverage held at 0.5° simulated rotation but failed in over one-third of plans at 2°; here measured residuals stayed within 0.5 mm / 0.5° tolerance, which is why 1 mm sufficed. Necrosis tracking V12Gy matches the Korytko gamma knife series the authors cite.
The paper reports pt counts inconsistently (4-6/7-10 vs 4-7/8-10; 10 vs 12 pts with coverage loss). LC is not stratified by margin or by distance from isocenter, so the clinical consequence of the dosimetric miss is untested.
The case for 1 mm is a trade-off: the margin rescues small distal targets while Table 4 shows each added millimetre inflates a 1 cm lesion's volume steeply, feeding V12Gy. That trade-off only holds if intrafraction drift is caught and corrected.
Retrospective 31-pt series, median f/u 11 mo, no comparator arm; margin conclusion rests on dosimetric re-computation, not lesion-level clinical failure by margin.
- Lesion-level local failure by distance from isocenter and margin size recruiting Trial of Differential Margins in Single Isocenter Radiosurgery of Brain Metastases Phase NAn=180 · primary completion 2028-06 · randomizes 0 vs 2 mm margin in single-isocenter SRS
- Neurocognitive outcome with formal testing after single-isocenter SRS for 5-10 BM n=80 · primary completion 2026-06 · SRS for ≥6 BM, ph2 1° EP neurocognitive declinen=110 · primary completion 2026-12 · prospective SRS cohort, 1-10 lesions, serial NCF testingn=90 · primary completion 2029-04 · rand ph2 SRS vs CS-WBRT vs HA-WBRT, ≥6 BM, 6-mo cognition
- Is 1 mm sufficient without intrafraction monitoring and correction?
📚 Sources · 📄 1 paper
The longer read
The question this paper answers is narrower than its title suggests. Its clinical outcomes, 89% per-patient and 94% per-lesion local control at one year with no grade 3-4 toxicity, are what one would expect from any competent single-fraction SRS series at 20-22 Gy and add little on their own: 31 patients, a median follow-up of 11 months, no comparator, and survival driven by extracranial control and concurrent systemic therapy. The contribution is the geometric one, and it addresses a specific worry about single-isocenter multi-target treatment: a rotational error at the shared isocenter becomes a translational miss at a distant target, and the miss grows with distance.
Roper et al had framed the problem in simulation, with coverage holding at 0.5° but failing in more than a third of plans at 2°. This series supplies the missing piece, the residual errors actually delivered on a frameless mask with 6D ExacTrac correction. Mean 3D residual was 0.34 mm, and with errors that small the losses concentrated where geometry predicts, in lesions under 0.4 cm3 sitting at or beyond 3.9 cm from isocenter. A 1 mm margin absorbed them in re-computation. The read for a linac SRS program is that the margin decision depends on the error budget your own system achieves, and this paper gives one measured budget rather than a universal answer.
The intrafraction data carry the more practical warning. Ten positionings in 7 patients drifted out of tolerance, and left uncorrected that drift cost GTV coverage in 28% of the 74 lesions involved; even with 1 mm, 2 lesions stayed short. So the 1 mm conclusion is conditional on mid-treatment imaging and correction. A department running mask-based single-isocenter plans without intrafraction monitoring cannot borrow it.
The skeptic's case rests on three points. First, adequacy is defined dosimetrically, by V95 on a CBCT-fused re-computation, and the paper never links coverage loss or margin size to lesion-level failure; 11 lesions recurred, and we are not told whether they were the distal small targets. Second, the reporting is internally inconsistent on patient counts by lesion number and on how many patients had coverage loss, which lowers confidence in the denominator behind the subgroup statistics. Third, the 28 lesions planned with 2 mm are folded into the outcomes, so the clinical series is not a clean 1 mm cohort.
The margin's cost side is the stronger argument. The authors' own volume table shows how quickly an isotropic expansion inflates a small sphere, and necrosis here tracked individual V12Gy above 10 cm3 rather than lesion count or cumulative volume, in line with the Korytko gamma knife experience they cite. That points toward the smallest margin that holds coverage, not toward the 2-3 mm some early single-isocenter planning studies suggested. What would overturn the conclusion is a series with lesion-level local failure mapped against distance and margin showing excess misses in distal small targets at 1 mm, or a center whose residual errors run materially above the 0.5 mm / 0.5° tolerance used here. Until then this is a reasonable, system-specific justification for 1 mm, not evidence that the technique matches per-lesion isocenter SRS on control.