LINAC SRS Technique Comparison: DCA vs VMAT
ForLINAC-based single-target cranial SRS/SRT, Elekta Agility 5 mm MLC
TL;DRPaddick CI 0.78 VMAT vs 0.75 DCA+SSO+VDR vs 0.6 plain DCA across 11 replanned cases. VMAT needed 41 ± 17% more MU than DCA+SSO+VDR.
Against the optimized DCA+SSO+VDR variant, VMAT's dosimetric advantage shrinks to CI 0.78 vs 0.75 (p=0.03); GI (p=0.29) and V12Gy 11.14 vs 11.75 cc (p=0.38) did not differ, at the cost of MU 4040 vs 2818. On an Elekta 5 mm-leaf platform, SSO+VDR, not VMAT, captures most of the gain over plain DCA.
Against the optimized DCA+SSO+VDR variant, coplanar VMAT gains only CI 0.78 vs 0.75 (p=0.03), with no difference in GI or brain V12Gy (11.14 vs 11.75 cc, p=0.38) and MU 4040 vs 2818. On a 5 mm-leaf Elekta platform, SSO+VDR captures most of the improvement over plain DCA.
9 details
Retrospective dosimetric replanning of 11 brain cases, each planned four ways (DCA, DCA+SSO, DCA+SSO+VDR, VMAT). Paired t-tests between adjacent techniques, p<0.05 significant.
7 solitary metastases, 2 meningiomas, 1 acoustic schwannoma, 1 pituitary adenoma. PTV 0.93 to 22.5 cc, median 4.99cc.
9 SRS and 2 SRT, 18 to 30Gy total. PTV = GTV + 2 mm; objective 95% of PTV covered by 95% of prescription. Elekta Versa HD, Agility 5 mm leaves, 6 FFF, Monaco Monte Carlo. DCA plans noncoplanar (couch angles ≥30 degrees apart); VMAT coplanar single arc.
Paddick CI, RTOG HI, Paddick GI, MU as a treatment-time proxy, and healthy-brain V12Gy (brain minus PTV).
Each DCA refinement improved CI stepwise; VMAT edged DCA+SSO+VDR on CI only (p=0.03), with no difference in GI (p=0.29) or V12Gy (p=0.38) and higher MU (p=0.0007). DCA gave the most homogeneous dose (HI 1.17).
| Metric | DCA | DCA+SSO | DCA+SSO+VDR | VMAT | p (DCA vs SSO / SSO vs VDR / VDR vs VMAT) |
|---|---|---|---|---|---|
| Paddick CI | 0.6 | 0.69 | 0.75 | 0.78 | 0.0004 / 0.015 / 0.03 |
| RTOG HI | 1.17 | 1.22 | 1.26 | 1.28 | 0.008 / 0.04 / 0.06 |
| Paddick GI | 1.35 | 1.21 | 1.20 | 1.17 | 0.03 / 0.33 / 0.29 |
| MU | 2507 | 2724 | 2818 | 4040 | 0.001 / 0.20 / 0.0007 |
| Brain V12Gy (cc) | 16.75 | 14.13 | 11.75 | 11.14 | 0.10 / 0.01 / 0.38 |
Authors report CI direction matching Jessica et al and Lagerwaard et al (VMAT over DCA), and GI matching Chang et al. Chang et al reported VMAT needing higher MU than DCA by a factor of 2.37, larger than the gap here against SSO+VDR.
Coplanar VMAT vs noncoplanar DCA means geometry and modulation are not separable. No delivery QA, measured beam-on time, or setup-error sensitivity analysis, despite the authors raising both. Mixed benign and malignant targets across a wide volume range in N=11.
The defensible read is that SSO+VDR closes most of the DCA-to-VMAT gap; the authors' own claim that VMAT has shorter total treatment time rests on FFF dose rate and avoided couch entries, not a measured time.
Dosimetric surrogates only, N=11 mixed histologies; beam geometry confounded with technique and VMAT's edge over DCA+SSO+VDR rests on CI alone.
- Does noncoplanar VMAT outperform optimized noncoplanar DCA?
- Do dosimetric gains translate to lower radionecrosis rates?
- How sensitive are VMAT gradients to setup error versus DCA?
📚 Sources · 📄 1 paper
Abstract
The longer read
The paper's headline, that coplanar VMAT is the preferred LINAC SRS technique, is stronger than its own table supports. The stepwise structure of the comparison is the useful part: plain DCA to DCA+SSO to DCA+SSO+VDR to VMAT, each tested against its neighbour. Read that way, most of the conformity and gradient improvement is captured by segment shape optimization and variable dose rate, still within a forward-planned, noncoplanar arc framework. The final step to VMAT moves the Paddick CI from 0.75 to 0.78 (p=0.03) and nothing else that a radiosurgeon would weigh: gradient index and healthy-brain V12Gy are statistically indistinguishable (p=0.29 and p=0.38), homogeneity trends worse, and MU climbs to 4040 from 2818.
The design choice that most limits the conclusion is that technique and beam geometry travel together. Every DCA plan is noncoplanar and every VMAT plan is a single coplanar arc. Noncoplanar geometry is generally what buys gradient in cranial SRS, so a coplanar VMAT plan matching an optimized noncoplanar DCA plan on GI is interesting, but the study cannot say whether modulation or arc arrangement drives any of the differences. A noncoplanar VMAT arm, which most current LINAC SRS workflows would consider, is absent, so the comparison a reader actually faces in planning is not tested.
Platform matters for transfer. The Agility MLC here has 5 mm leaves, and the smallest targets were under 1 cc. Coarse leaves penalize forward-planned conformal shapes more than inversely optimized ones, which may flatter VMAT on CI for the smallest lesions; the authors show a CI by PTV volume correlation figure but the source text does not report the slope. Centres on finer-leaf MLCs, cones, or dedicated platforms should not assume the same ordering.
Statistically, eleven paired cases spanning metastases, meningiomas, a schwannoma and a pituitary adenoma, 18 to 30Gy, SRS and SRT, is a heterogeneous sample for detecting small index differences. Only adjacent pairs were tested, without multiplicity correction, so VMAT was never directly compared against plain DCA, and the p=0.03 CI result would not survive a conservative correction across the fifteen comparisons in the table.
The treatment-time argument is also unresolved. MU is used as the time proxy, yet the conclusion states VMAT delivery was faster overall because of FFF dose rate and fewer couch rotations; no beam-on or room time was measured. The authors also flag that a steeper VMAT gradient is more sensitive to setup error and that VMAT carries heavier QA demands, but neither was quantified.
What would make the paper wrong is plausible: a noncoplanar VMAT arm or a finer MLC could widen the VMAT lead, while setup-error modelling could erase it. Practically, the result supports DCA+SSO+VDR as a defensible single-lesion SRS technique on this platform, and gives no dosimetric reason to abandon it for coplanar VMAT on V12Gy grounds.