Dose-adapted single-isocenter SRS for 4 to 10 brain mets (prospective) NCT02886572
For4-10 untreated brain mets, KPS ≥70, excluding SCLC, lymphoma, germ-cell, LMD
TL;DRmOS 8.1 mo, 1-yr OS 35.7%; 22/40 outlived DS-GPA prediction (≥17 needed for non-inferiority), 1-yr local failure 5%, no G3-5 toxicity.
The transferable piece is the dose-adaptation rule: switching every lesion to 5 Gy x5 when V12Gy exceeds 20 mL or a brainstem lesion is present gave 98% per-lesion LC with 1.3% necrosis, at aggregate PTV up to 32 mL. That supports hypofractionating the whole plan in high-volume cases instead of de-escalating single-fraction dose.
For a KPS ≥70 pt with 4 to 10 non-SCLC brain mets and aggregate volume pushing single-fraction V12Gy past 20 mL, this supports a single-isocenter 5 Gy x5 plan as a feasible SRS option; it does not address SCLC, leptomeningeal disease, or more than 10 lesions.
The transferable piece is the dose-adaptation rule: switching every lesion to 5 Gy x5 when V12Gy exceeds 20 mL or a brainstem lesion is present gave 98% per-lesion LC with 1.3% necrosis, at aggregate PTV up to 32 mL. That supports hypofractionating the whole plan in high-volume cases instead of de-escalating single-fraction dose.
14 details 4 trials watching
Prospective single-arm trial, N=40, enrolled Feb 2017 to Aug 2019, registered NCT02886572. Median follow-up 15.2 mo. Funded by Varian Medical Systems.
4 to 10 untreated brain mets, age ≥18, KPS ≥70, life expectancy >3 mo, DS-GPA ≥0.5, largest lesion ≤4 cm, >2 mm from optic apparatus; brainstem lesions allowed. Excluded SCLC, germ-cell, lymphoma, leukemia, myeloma, leptomeningeal disease. Median age 61, KPS 90, 6 mets; NSCLC 50%, breast 27.5%; 11 had prior SRS/WBRT.
Single-isocenter VMAT, 1 mm PTV margin. 20 Gy for PTV <2 cm, 18 Gy for 2.0 to <3.0 cm; 5 Gy x5 to all lesions if any PTV >3 cm, any brainstem lesion, or brain V12Gy >20 mL. 24 pts (125 lesions) got 1 fraction, 16 pts (79 lesions) got 5; median beam-on 8 min.
Primary: OS, judged non-inferior if ≥17 of 40 pts outlived DS-GPA-predicted survival. Secondary: local recurrence, distant brain recurrence, neurologic death, AEs, QOL, steroid use.
22 pts outlived prediction; mOS 8.1 mo, 1-yr OS 35.7%. 1-yr local failure 5% (10 of 204 lesions); distant brain recurrence 35.8% at 12 mo. Neurologic death 20% of deaths.
| Endpoint | Single fraction | 5 fractions |
|---|---|---|
| Local failure, lesions | 8 of 119 (no dose adjustment) | 2 of 79 |
| Local control | 93% | 98% |
| Radionecrosis per lesion | 3.4% (4 of 119) | 1.3% (1 of 79) |
AEs in 73%, all G1-2, mostly headache and fatigue; no G3-5 treatment-related AEs. Radionecrosis in 5 lesions in 4 pts; 2 pts needed surgery. MMSE, Trail Making, HVLT and FACT-Br unchanged at 3 mo.
mOS of 8.1 mo sits inside the range of prior multi-met SRS series summarized in the paper's Table 4 (medians 5.8 to 14.1 mo), most of them retrospective. The randomized SRS vs WBRT trials for 4 to 20 mets (NCT03550391, NCT03075072, NCT01592968) are the comparators this cannot replace.
The DS-GPA benchmark uses the original formulation, measures from first brain-met diagnosis, and was applied here from protocol SRS in a cohort where 11 had prior brain RT. LC and DBF exclude the 8 pts who died before first MRI, and the paper's own Discussion gives a 6-mo distant failure figure (58%) that conflicts with its Results (20%).
The OS endpoint mostly shows the technique does not shorten life against a prognostic index; it cannot show SRS beats WBRT. The more usable signal is the per-lesion one: 5-fraction plans held 98% LC with low necrosis despite larger volumes, which supports hypofractionating high-volume multi-met plans.
Single-arm, N=40, single institution; OS benchmarked against a DS-GPA prediction, not a randomized WBRT or multi-isocenter comparator. Too few events for LC or necrosis modeling.
- SRS vs WBRT for 4 to 15 brain mets: survival and neurocognition n=126 · primary completion 2026-12 · SRS alone vs SRS+WBRT, 5 to 30 mets, prospectiven=90 · primary completion 2029-04 · SRS vs HA-WBRT vs CS-WBRT, ≥6 mets, 6-mo cognition
- Single-fraction vs hypofractionated SRS for high aggregate volume intact mets active Comparing Single vs Multiple Dose Radiation for Cancer Patients With Brain Metastasis and Receiving Immunotherapy Phase NAn=58 · primary completion 2028-03 · SSRS vs FSRS for mets ≥2 cm or ≥4 cc on ICIn=504 · primary completion 2030-01 · randomised 3x10 Gy SRT vs single-dose SRS, ≤5 BM
📚 Sources · 📄 1 paper
The longer read
This trial asks a narrower question than its title suggests. Its primary endpoint is not whether single-isocenter SRS controls disease better than whole-brain RT, but whether treating 4 to 10 metastases this way shortens survival compared with what a prognostic index predicts. Clearing the bar (22 pts outlived their DS-GPA estimate against a threshold of 17) tells a reader the technique does not appear to cost survival in a selected, KPS-90 cohort. It does not tell them SRS is preferable to WBRT for this number of lesions. That question belongs to the randomized trials the authors name (NCT03550391, NCT03075072, NCT01592968), and this paper should be read as feasibility and safety groundwork alongside them.
The benchmark itself deserves skepticism. The original DS-GPA estimates survival from first diagnosis of brain metastases, while the primary analysis here starts the clock at protocol SRS, and 11 of 40 pts had already received brain RT. Measuring from a later time point biases toward shorter observed survival, so the comparison is probably conservative for the technique. On the other hand, the cohort was screened for life expectancy over 3 months, DS-GPA of at least 0.5 and KPS of at least 70, so the people enrolled were already the ones most likely to beat a population-level estimate. Those two biases pull in opposite directions, and with 40 pts there is no way to net them out.
The RT-relevant content sits in the dose-adaptation rule and the per-lesion outcomes. Rather than reducing single-fraction dose to fit a crowded plan, the protocol moved every lesion to 5 Gy x5 once any lesion exceeded 3 cm, any lesion sat in the brainstem, or brain V12Gy passed 20 mL. Those 5-fraction plans lost 2 of 79 lesions locally against 8 of 119 for unadjusted single-fraction lesions, with necrosis of 1.3% versus 3.4%. The groups are not randomized and differ by design (the 5-fraction group carried the larger volumes), so the contrast cannot be read as fractionation causing better control. What it does support is that hypofractionation absorbed the high-volume cases without an obvious penalty in control or necrosis, at aggregate PTVs up to 32 mL. That speaks directly to the Alliance question in resected cavities the authors cite and to the unanswered intact-lesion version of it.
Several reporting issues lower confidence in the secondary numbers. The local and distant failure analyses include only the 32 pts who reached a first MRI, so the 8 earliest deaths are not counted in the denominator. The Discussion quotes a 6-month distant failure risk of 58%, which cannot be reconciled with the 20% cumulative incidence at 6 months in Results. Lesion counts by fractionation (119 plus 79) do not sum to the 204 treated lesions. None of these overturns the main message, but together they suggest the secondary endpoints are descriptive, which the authors acknowledge when they note too few events for time-to-event modeling of local failure or necrosis.
The neurocognitive and QOL stability at 3 months is reassuring and fits the low hippocampal and integral brain doses reported, but testing attrition after 6 months, driven largely by death, means these data cannot speak to the late decline that defines the WBRT trade-off. For practice, the study supports single-isocenter, volume-adapted SRS as a workable option for 4 to 10 lesions in fit pts, and the V12Gy-triggered switch to 5 fractions as a planning rule worth adopting. It leaves the comparative question to the randomized trials.