Dosimetric comparison of pipe-joint versus conventional VMAT for lattice radiotherapy in bulky tumors: a retrospective planning study Authors: Changyou Zhong, Meiling Xu, Shengzhen Dong, Cuiting Xu, Zhendong Jiang, Haidong Yu, Zhitao Dai Published: 2026 ABSTRACT AbstractBackgroundLattice radiotherapy (LRT) is a promising spatially fractionated technique for bulky tumors. However, when lattice vertices extend over a large superior–inferior distance, conventional volumetric modulated arc therapy (c-VMAT) may require broad jaw apertures, which may compromise longitudinal dose modulation, increase leakage-related low-dose exposure, and exacerbate island-blocking-like effects. To address these limitations, we developed pipe-joint VMAT (PJ-VMAT), a beam’s-eye-view-guided longitudinal segmentation strategy.MethodsNineteen consecutive patients with bulky tumors were retrospectively included. Spherical lattice vertices were generated within the gross tumor volume (GTV) using an in-house sphere-packing algorithm. For each patient, four plans were generated: lattice stereotactic body radiotherapy (L-SBRT) and standalone lattice radiotherapy (S-LRT), each implemented using c-VMAT or PJ-VMAT. Plans were normalized such that 95% of the lattice target volume received 100% of the prescription dose. The primary dosimetric endpoints were dose fall-off around the lattice vertices and the peak-to-valley dose ratio (PVDR). Secondary endpoints included background-dose reduction, normal-tissue sparing, plan complexity, and delivery efficiency. An exploratory geometry-based association analysis was performed to examine whether target-geometry descriptors were associated with the magnitude of dosimetric benefit from PJ-VMAT. No formal multiplicity adjustment was applied, and all reported values were nominal.ResultsCompared with c-VMAT, PJ-VMAT achieved steeper dose fall-off, with lower gradient index values in both L-SBRT (PJ-VMAT: 10.55 vs. c-VMAT: 12.72) and S-LRT (PJ-VMAT: 7.12 vs. c-VMAT: 8.33). PJ-VMAT improved most PVDR metrics, with larger gains observed in the S-LRT setting. In addition, PJ-VMAT reduced background dose to the GTV–LTV region, Body–GTV region, and GTV shell. These dosimetric advantages were accompanied by higher monitor units, greater plan complexity, and longer delivery times. Several associations between target geometry and dosimetric benefit were observed, particularly in the L-SBRT setting; however, these findings were exploratory and hypothesis-generating.ConclusionsIn this in silico planning study, PJ-VMAT demonstrated favorable dosimetric performance compared with c-VMAT for LRT in bulky tumors.The observed advantages may reflect improved longitudinal dose modulation through segment-specific jaw restriction, resulting in enhanced peak-to-valley dose separation and reduced background-dose exposure. The exploratory geometry-related findings should not be used to guide individual patient selection and require validation in larger cohorts. Supplementary InformationThe online version contains supplementary material available at https://doi.org/10.1186/s13014-026-02913-x.Keywords: Spatially fractionated radiation therapy, Lattice radiotherapy, Volumetric modulated arc therapy, Peak-to-valley dose ratio, Gradient index, Jaw restriction, Longitudinal segmentation, Bulky tumors Introduction Bulky tumors remain challenging in external beam radiotherapy because durable local control often requires dose escalation, whereas large target volumes and proximity to critical organs frequently limit the safe delivery of homogeneous high-dose irradiation. Spatially fractionated radiation therapy (SFRT) has emerged as a clinically relevant strategy in this setting by generating intratumoral high-dose peaks separated by lower-dose valleys, thereby enabling heterogeneous dose escalation while partially limiting dose exposure to adjacent normal structures. This unique combination may improve the therapeutic ratio compared with conventional homogeneous irradiation. Lattice radiotherapy (LRT), initially developed as a three-dimensional extension of GRID therapy [1, 2], further adapts this concept to modern inverse-planned radiotherapy and enables more conformal heterogeneous dose distributions for bulky lesions [3, 4]. Early clinical and translational studies, together with more recent systematic reviews, practice surveys, feasibility planning studies, and emerging clinical outcome reports, have supported the feasibility of LRT for bulky tumors and contributed to its increasing adoption in contemporary practice [5–11]. Lattice radiotherapy can be implemented using two principal planning approaches. In lattice stereotactic body radiotherapy (L-SBRT), high-dose lattice vertices are delivered together with a lower background dose to the remaining GTV using a simultaneous integrated boost strategy [12–15]. In standalone lattice radiotherapy (S-LRT), high-dose irradiation is delivered selectively to lattice vertices without a simultaneous background prescription to the entire GTV [16–18]. These approaches impose different dose-modulation requirements and were therefore evaluated separately. Despite the increasing use of volumetric modulated arc therapy (VMAT) for lattice radiotherapy (LRT), dose modulation remains technically challenging when lattice vertices extend over a long superior–inferior distance. Conventional VMAT (c-VMAT) may require a broad jaw aperture to cover multiple spatially separated vertices throughout gantry rotation, potentially increasing the modulation burden and unnecessary irradiation of the intervening valley regions. Xu et al. previously proposed a piecewise VMAT strategy for single-isocenter treatment of multiple brain metastases, in which different arcs were assigned to longitudinally separated target groups [19]. Related partial-target arc strategies have also been investigated for spatially discrete intracranial targets [20]. Both multiple-brain-metastasis planning and lattice planning require the simultaneous modulation of multiple spatially separated high-dose regions over a large beam’s-eye-view extent. Therefore, the underlying mechanical and optimization principle of piecewise VMAT may also be adapted to lattice radiotherapy. In the present study, we adapted this piecewise VMAT principle to lattice radiotherapy. Because the segment-specific treatment fields form a pipe-joint-like arrangement along the superior–inferior direction, we refer to this lattice-specific implementation as pipe-joint VMAT (PJ-VMAT). Each jaw-restricted arc is assigned to a limited number of lattice vertices and the corresponding local portion of the GTV. Although unrestricted c-VMAT may theoretically reproduce a similar dose distribution, practical optimizer and machine constraints may limit the achievable solution when all vertices and the intervening GTV are optimized simultaneously within a broad aperture. By reducing the field extent requiring simultaneous multileaf collimator modulation and decomposing the global optimization task into smaller regional problems, PJ-VMAT may mitigate island-blocking-like irradiation of the intervening valley regions. We therefore hypothesized that this adapted strategy would improve peak-to-valley dose separation and reduce background-dose exposure compared with c-VMAT. To our knowledge, although piecewise VMAT has previously been investigated for multiple brain metastases, its adaptation to lattice radiotherapy and its dosimetric performance under distinct L-SBRT and S-LRT planning paradigms have not been specifically evaluated. This study aimed to compare PJ-VMAT with c-VMAT in both the L-SBRT and S-LRT planning frameworks, with particular emphasis on dose fall-off around lattice vertices, peak-to-valley dose separation, suppression of background dose to non-vertex tumor regions and surrounding normal tissues, and the trade-off between plan complexity and delivery efficiency. In addition, we performed an exploratory geometry-based association analysis to examine whether target-geometry descriptors were associated with the magnitude of dosimetric benefit from PJ-VMAT. Materials and methods Patient cohort We retrospectively included all 19 consecutive patients who underwent lattice radiotherapy planning at our institution between January 2024 and December 2025. No patients were excluded from the final planning analysis. Patient and target characteristics are summarized in Table 1. This study was approved by the Ethics Committee of Meizhou People’s Hospital (IRB2025-C-191) and was conducted in accordance with the Declaration of Helsinki. The requirement for written informed consent was waived owing to the retrospective design. Table 1Patient and target characteristicsCharacteristicOverall cohortNumber of patients,n19Age, years, median (range)40 (35–77)Tumor site,n(%) Head and neck2 (10.5%) Thorax7 (36.8%) Abdomen9 (47.4%) Pelvis1 (5.3%)GTV volume, cc, median (range)845.1 (121.5–2734.1)Target geometry, cm, median (range) X 13.1 (4.70–25.32) Y12.4 (6.22–18.39) Z12.2 (7.03–28.13) \max(X,Z)15.8 (8.14–28.13)Number of lattice vertices (Nvert), median (range)9 (5–34)Note: GTV = gross tumor volume; max(X, Z) = maximum of X and Z Patient and target characteristics Note: GTV = gross tumor volume; max(X, Z) = maximum of X and Z Lattice vertex generation and verification The GTV and organs at risk (OARs) were exported as DICOM files and transferred to the smART Vision Pro auto-contouring platform (version 1.3.0; Shenying MedTech Co., Ltd., China). Spherical lattice vertices with diameters of 1.0–1.5 cm were generated within the GTV using an in-house close-packing algorithm. Vertices were placed 0.5–1.0 cm inside the GTV boundary and 1.0–2.0 cm away from the adjacent OARs. The minimum center-to-center distance between any two vertices was constrained to be at least three times the corresponding vertex diameter. Candidate positions that violated the predefined spacing, GTV-boundary, or OAR-exclusion criteria were rejected, and vertex placement continued until no additional eligible vertex could be accommodated. The automatically generated vertex configuration was subsequently reviewed by the radiation oncologists. When clinically warranted, the positions of individual vertices were manually adjusted based on patient-specific clinical considerations, while maintaining the predefined geometric constraints. Treatment planning Pipe-joint volumetric modulated arc therapy The longitudinal piecewise VMAT concept proposed by Xu et al. for multiple spatially separated targets served as the methodological basis for the present approach [19]. This concept was modified and adapted to the geometric and dosimetric characteristics of lattice radiotherapy. In contrast to the original piecewise VMAT implementation, which used a collimator angle of , all arcs in the present study used a collimator angle of . Because the resulting segment-specific fields formed a pipe-joint-like configuration along the superior–inferior direction, this modified lattice-specific implementation was termed pipe-joint VMAT (PJ-VMAT). Each PJ-VMAT plan consisted of four full coplanar arcs. The target was longitudinally divided into four partially overlapping pipe segments, with each jaw-restricted arc assigned to the corresponding segment. For L-SBRT, the jaw opening of each arc was defined according to the beam’s-eye-view (BEV) projection of the corresponding local portion of the GTV. For S-LRT, the jaw opening was defined according to the BEV projection of the corresponding portion of the LTV. Adjacent longitudinal segments overlapped by a fixed 1 cm margin to maintain continuous dose coverage and reduce the risk of underdosage in the junction regions. The collimator angle was fixed at for all four arcs, aligning the MLC leaf-motion direction with the superior–inferior target segmentation and facilitating segment-specific jaw restriction. A schematic illustration of the PJ-VMAT strategy is presented in Fig. 1. Schematic illustration of the proposed pipe-joint volumetric modulated arc therapy (PJ-VMAT) strategy. (a) Four coplanar arcs, each assigned to a specific longitudinal segment along the superior–inferior axis. Adjacent segments partially overlap to maintain continuous target coverage and a smooth dose transition. (b) For lattice stereotactic body radiotherapy (L-SBRT), the treatment-field boundary for each arc is defined according to the beam’s-eye-view projection of the corresponding local portion of the gross tumor volume (GTV). (c) For standalone lattice radiotherapy (S-LRT), the treatment-field boundary is defined according to the projected extent of the lattice target volume (LTV) within each longitudinal segment. In both planning strategies, segment-specific jaw restriction forms a pipe-joint-like configuration along the superior–inferior direction Four arcs were used for both PJ-VMAT and c-VMAT because four-arc c-VMAT is routinely used for lattice planning at our institution, and maintaining the same number of arcs enabled a controlled comparison between the two techniques. The four-pipe configuration represented a standardized, pragmatic balance among longitudinal field restriction, target-coverage continuity, plan complexity, and delivery efficiency and was not intended to represent an optimal patient-specific pipe number. Conventional volumetric modulated arc therapy The c-VMAT plans also consisted of four full coplanar arcs, with the same gantry start–stop angles and rotation directions as the corresponding PJ-VMAT plans. The collimator angles were set to , , , and , respectively, according to our institutional practice for VMAT-based lattice planning and to provide complementary MLC modulation directions. No longitudinal target segmentation or segment-specific jaw restriction was applied. For L-SBRT, the jaw opening of each arc encompassed the full longitudinal extent of the GTV, whereas for S-LRT, it encompassed the full longitudinal extent of the LTV. All lattice vertices and the relevant surrounding target volume were therefore optimized simultaneously within a broad jaw opening. Accordingly, the present comparison represents two practical planning implementations rather than an isolated assessment of jaw restriction alone. The principal delivery-design differences between PJ-VMAT and c-VMAT were BEV-guided longitudinal grouping, segment-specific jaw restriction, and collimator-angle arrangement. Plan design and optimization To compare the planning-level dosimetric performance of PJ-VMAT and c-VMAT, two lattice-based treatment planning paradigms were evaluated: L-SBRT and S-LRT. Each paradigm was implemented using both c-VMAT and PJ-VMAT. Accordingly, four plans were generated for each patient: L-SBRT, L-SBRT, S-LRT, and S-LRT The plan isocenter was automatically positioned at the geometric centroid of the LTV for all plans. For L-SBRT, 36 Gy in three fractions was prescribed to the LTV, together with a lower background prescription of 12 Gy in three fractions to the GTV. The GTV background prescription was implemented as a dose–volume objective requiring GTV Gy and was optimized simultaneously with the LTV dose objectives and OAR constraints. For S-LRT, 36 Gy in three fractions was prescribed to the LTV without a simultaneous whole-GTV background-dose objective. Within each planning paradigm, the c-VMAT and PJ-VMAT plans used identical target structures, lattice vertex configurations, dose prescriptions, optimization objectives, photon energy, and dose-calculation settings to enable a controlled within-patient comparison. All plans were designed for delivery on a TrueBeam linear accelerator (Varian Medical Systems, Palo Alto, CA, USA) equipped with a Millennium 120-leaf MLC with a 5-mm central leaf width at isocenter. All plans were generated in Eclipse version 15.5 (Varian Medical Systems, Palo Alto, CA, USA), using the Photon Optimizer for plan optimization and the Anisotropic Analytical Algorithm for final dose calculation [21]. Plan optimization was performed with a spatial resolution of 2.5 mm, whereas the final dose distributions were recalculated using a high-resolution dose grid of 1.25 mm. Plan evaluation After optimization, all plans were normalized such that 95% of the LTV received 100% of the prescribed dose, thereby enabling direct within-patient comparisons between PJ-VMAT and c-VMAT. LTV dosimetric parameters The evaluated LTV dosimetric parameters were , , , and the gradient index (GI). The GI was used to characterize dose fall-off outside the LTV and was calculated according to Eq. 1 [22, 23], where and denote the volumes receiving at least 50 and 100% of the prescription dose, respectively. Lower GI values indicate a steeper dose gradient. GTV dosimetric parameters The evaluated GTV dosimetric parameters were , , , , , and . The mean dose was also evaluated for the non-vertex tumor region, defined as the GTV after exclusion of the LTV (GTV–LTV). Because no standardized definition of the peak-to-valley dose ratio (PVDR) has been established for lattice radiotherapy, four complementary PVDR metrics were evaluated: (PVDR), (PVDR), (PVDR), and (PVDR) [16, 17, 24–26]. PVDR was defined as the ratio of the mean dose to the LTV to that of the entire GTV. OAR and background-dose parameters OARs located within 5 cm of the GTV in the superior–inferior direction were included in the dosimetric evaluation. The evaluated endpoints comprised: (1) for the heart and lungs; (2) for the rectum, esophagus, kidneys, duodenum, bladder, sigmoid colon, colon, small intestine, and spinal cord; and (3) for the liver after exclusion of the GTV (liver–GTV). Only patients for whom the corresponding OAR was located within the predefined evaluation range were included in each paired OAR comparison. A 1-cm-thick shell surrounding the GTV was generated as an exploratory geometric dose-spillage structure. The evaluated GTV-shell parameters were , , , and . The GTV shell was used solely to characterize dose exposure immediately surrounding the tumor and was not intended to represent a validated immunological compartment or a surrogate of immune modulation. For uninvolved normal tissue, defined as the external body contour after exclusion of the GTV (Body–GTV), both and integral dose (ID) were calculated. The ID was calculated according to Eq. 2: where and V denote the mean dose and volume of the evaluated structure, respectively, and and denote the dose and volume of voxel i derived from the differential dose–volume histogram [27]. Plan complexity and delivery-efficiency evaluation The total monitor units (MU) and delivery time were recorded for each plan as indicators of delivery efficiency. Plan complexity was characterized using leaf travel (LT), the modulation complexity score for VMAT (MCSv), average leaf gap (ALG), the proportions of apertures smaller than 10, 20, and 30 mm [, and ], and small-aperture scores calculated using thresholds of 5, 10, and 25 mm (SAS, SAS and SAS) [28, 29]. Lower MCSv values indicate greater modulation complexity, whereas higher and SAS values indicate a greater contribution from narrow MLC apertures. These metrics were extracted from the DICOM-RT plan files using the UCoMX package [30, 31] and were used to characterize plan complexity and estimates of delivery efficiency. Measurement-based patient-specific QA (PSQA) were not performed in this planning-only study. Exploratory geometry-based association analysis An exploratory geometry-based association analysis was performed to examine whether tumor geometry was associated with the magnitude of dosimetric benefit from PJ-VMAT. The evaluated geometric descriptors included the original GTV volume and the maximum target extents in the left–right (X), superior–inferior (Y), and anterior–posterior (Z) directions. The maximum nonlongitudinal target extent was additionally represented by \max(X,Z). A geometry burden index (GBI) was constructed as a composite descriptor of target size and spatial extent: where z(·) denotes z-score standardization. The cube root of GTV volume was used in the GBI calculation to express the volumetric term on a length scale comparable with those of the dimensional descriptors. In the individual-variable association analyses, however, the original untransformed GTV volume was used. GBI was calculated once for each patient and evaluated separately in the L-SBRT and S-LRT planning paradigms. These geometric descriptors were selected to characterize overall target size, longitudinal extent, and the maximum nonlongitudinal geometric burden relevant to BEV-guided field segmentation. The predefined dosimetric endpoint set comprised seven metrics: PVDR–PVDR, GTV–LTV , GTV-shell , and Body–GTV integral dose. To ensure a consistent direction of dosimetric benefit across all endpoints, the paired differences (Δ) for PVDR–PVDR were calculated as PJ-VMAT minus c-VMAT because higher PVDR values indicate better peak–valley dose separation. In contrast, the paired differences for GTV–LTV , GTV-shell , and Body–GTV integral dose were calculated as c-VMAT minus PJ-VMAT because lower values indicate better suppression of background or normal-tissue dose. Thus, a positive Δ consistently indicated a dosimetric advantage of PJ-VMAT for all seven endpoints, whereas a negative Δ indicated an advantage of c-VMAT. The same benefit-oriented difference definitions were used for the standardized paired effect-size analysis presented in Fig. 5 and the geometry–dosimetry correlation analysis presented in Fig. 6. The association analyses were considered descriptive and hypothesis-generating and were not intended to establish predictive thresholds or support individual patient selection. Forest plots of standardized paired effect sizes (Cohen’s ) for the seven predefined dosimetric endpoints in (a) L-SBRT and (b) S-LRT. Positive values favor PJ-VMAT, and horizontal lines represent 95% confidence intervals Spearman correlation heatmaps showing associations between tumor-geometric descriptors and the seven benefit-oriented dosimetric differences in (a) L-SBRT and (b) S-LRT. Positive coefficients indicate that the dosimetric advantage of PJ-VMAT increased with the corresponding geometric descriptor Statistical analysis Statistical analyses were performed using SPSS Statistics version 25.0 (IBM Corp., Chicago, IL, USA). Continuous variables were summarized as the mean ± standard deviation or median (range), as appropriate. For paired dosimetric comparisons, the distribution of the within-patient paired differences was assessed before hypothesis testing. Normally distributed paired differences were compared using paired t tests, whereas non-normally distributed paired differences were compared using Wilcoxon signed-rank tests. All statistical tests were two-sided, and p < 0.05 was considered nominally statistically significant. For the seven predefined dosimetric endpoints, standardized paired effect sizes were expressed as Cohen’s and calculated using the benefit-oriented paired differences defined in Sect. 2.5 where and denote the mean and standard deviation of the within-patient benefit-oriented paired differences, respectively. Positive Cohen’s values indicated a dosimetric advantage of PJ-VMAT, whereas negative values indicated an advantage of c-VMAT. For the exploratory geometry-based association analysis, Spearman’s rank correlation coefficients (ρ) were calculated between each geometric descriptor, including the original GTV volume, X, Y, Z, \max(X,Z), and GBI, and each of the seven benefit-oriented paired dosimetric differences defined in Sect. 2.5. Correlation analyses were performed separately for the L-SBRT and S-LRT planning paradigms. Because positive Δ values consistently represented greater benefit from PJ-VMAT, a positive correlation indicated that the dosimetric advantage of PJ-VMAT increased as the corresponding geometric descriptor increased, whereas a negative correlation indicated that this advantage decreased as the descriptor increased. Given the exploratory nature of this retrospective planning study, no formal multiplicity adjustment was applied to either the paired dosimetric comparisons or the exploratory correlation analyses. All reported p values are therefore nominal and should be interpreted cautiously, particularly for isolated findings close to the conventional significance threshold. Results Representative two-dimensional dose distributions and dose–volume histograms are shown in Figs. 2 and 3, respectively. In the representative case, PJ-VMAT showed tighter dose concentration around the lattice vertices and reduced background-dose spillage compared with c-VMAT in both planning paradigms. Quantitative target-related results are summarized in Table 2, detailed OAR results are provided in Supplementary Table S1, and plan-complexity and delivery-efficiency metrics are summarized in Table 3. Representative two-dimensional dose distributions of conventional volumetric modulated arc therapy (c-VMAT)- and pipe-joint volumetric modulated arc therapy (PJ-VMAT)-based lattice radiotherapy plans. Dose distributions for a representative patient are shown for (a) L-SBRT, (b) L-SBRT, (c) S-LRT, and (d) S-LRT. Axial, coronal, and sagittal views are presented for each plan to illustrate differences in dose concentration around the lattice vertices and background-dose distribution between c-VMAT and PJ-VMAT Dose–volume histograms of key target-related structures for the four treatment plans. Dose–volume histograms are shown for (a) The lattice target volume (LTV), (b) The gross tumor volume (GTV), (c) The non-vertex tumor region (GTV–LTV), and (d) The GTV shell. Red and blue curves represent the L-SBRT and S-LRT planning strategies, respectively, whereas solid and dash-dotted curves represent c-VMAT and PJ-VMAT, respectively. In this representative case, PJ-VMAT showed tighter dose concentration within the lattice vertices and reduced background dose to the non-vertex tumor region and surrounding shell Table 2Comparison of LTV, GTV, and PVDR dosimetric parameters between PJ-VMAT and c-VMAT in the L-SBRT and S-LRT planning strategiesp valuesMetricsL-SBRTL-SBRTS-LRTS-LRTLTV (Gy)40.11±1.1940.59±1.1641.14±1.3341.42±0.830.0020.040 (Gy)46.05±2.7047.41±2.7148.61±2.1948.85±1.340.0020.737 (%)51.52±16.3256.59±15.6361.44±13.4965.45±7.970.0050.040GI12.72±6.8810.55±5.028.33±3.617.12±2.210.001 < 0.001GTV (Gy)17.16±3.4516.64±3.5612.69±2.4211.87±2.290.001 < 0.001 (Gy)35.78±2.6435.81±2.7235.81±2.7635.86±2.910.6870.332 (Gy)28.59±3.7428.27±3.7427.27±3.7126.92±3.860.0570.108 (Gy)23.29±3.9722.71±3.9821.24±3.3720.55±3.250.0030.005 (Gy)15.55±4.0515.00±4.1111.61±2.3910.62±2.310.002 < 0.001 (Gy)12.94±2.9512.43±3.124.38±2.873.69±2.670.001 < 0.001 (Gy)12.47±2.7511.92±2.902.99±2.802.54±2.500.0010.001PVDR1.83±0.281.85±0.277.36±5.388.49±5.790.117 < 0.0012.34±0.412.43±0.3916.42±11.1318.69±11.760.003 < 0.0012.38±0.382.47±0.373.17±0.453.48±0.53 < 0.001 < 0.0012.41±0.352.51±0.373.35±0.633.61±0.68 < 0.001 < 0.001Data are presented as mean ± standard deviation. L-SBRTPJ and L-SBRTc denote lattice stereotactic body radiotherapy plans generated using PJ-VMAT and conventional VMAT, respectively; S-LRTPJ and S-LRTc denote standalone lattice radiotherapy plans generated using PJ-VMAT and conventional VMAT, respectively. LTV, lattice target volume; GTV, gross tumor volume; GI, gradient index; PVDR, peak-to-valley dose ratio. Dx% denotes the dose received by at least x% of the structure volume, and V110% denotes the percentage of the LTV receiving at least 110% of the prescription dose. a L-SBRTPJ versus L-SBRTc; b S-LRTPJ versus S-LRTc. All p values are nominal and were not adjusted for multiple comparisons Comparison of LTV, GTV, and PVDR dosimetric parameters between PJ-VMAT and c-VMAT in the L-SBRT and S-LRT planning strategies Data are presented as mean ± standard deviation. L-SBRTPJ and L-SBRTc denote lattice stereotactic body radiotherapy plans generated using PJ-VMAT and conventional VMAT, respectively; S-LRTPJ and S-LRTc denote standalone lattice radiotherapy plans generated using PJ-VMAT and conventional VMAT, respectively. LTV, lattice target volume; GTV, gross tumor volume; GI, gradient index; PVDR, peak-to-valley dose ratio. Dx% denotes the dose received by at least x% of the structure volume, and V110% denotes the percentage of the LTV receiving at least 110% of the prescription dose. a L-SBRTPJ versus L-SBRTc; b S-LRTPJ versus S-LRTc. All p values are nominal and were not adjusted for multiple comparisons Table 3Comparison of plan complexity and delivery efficiency metrics between PJ-VMAT and c-VMAT in the L-SBRT and S-LRT planning strategiesp valuesParameterL-SBRTL-SBRTS-LRTS-LRTMU7202.17±2391.3615281.22±3285.577692.03±1450.4316021.39±3095.68 < 0.001 < 0.001Delivery time (min)5.14±1.7110.86±2.355.49±1.0411.44±2.21 < 0.001 < 0.001LT172.86±78.55125.25±52.34104.06±50.3254.78±41.77 < 0.001 < 0.001MCSv0.13±0.040.08±0.020.12±0.030.06±0.01 < 0.001 < 0.001ALG12.25±5.217.74±1.978.74±3.677.79±1.99 < 0.0010.570 (%)11.86±20.9432.94±17.1925.85±19.8632.10±16.77 < 0.0010.107 (%)54.78±30.5586.26±14.5280.15±24.1493.38±12.94 < 0.001 < 0.001 (%)79.12±24.5697.87±3.5192.35±18.6198.98±3.63 < 0.001 < 0.001SAS0.35±0.130.51±0.080.42±0.100.41±0.07 < 0.0010.330SAS0.58±0.160.73±0.080.70±0.130.69±0.08 < 0.0010.180SAS0.87±0.110.96±0.040.94±0.060.97±0.03 < 0.001 < 0.001Data are presented as mean ± standard deviation. L-SBRTPJ and L-SBRTc denote lattice stereotactic body radiotherapy plans generated using PJ-VMAT and conventional VMAT, respectively; S-LRTPJ and S-LRTc denote standalone lattice radiotherapy plans generated using PJ-VMAT and conventional VMAT, respectively. MU, monitor units; LT, leaf travel; MCSv, modulation complexity score for VMAT; ALG, average leaf gap; psmall,x mm, percentage of apertures smaller than x mm; SASx mm, small-aperture score at an aperture threshold of x mm. Lower MCSv values indicate greater modulation complexity. a L-SBRTPJ versus L-SBRTc; b S-LRTPJ versus SLRTc.All p values are nominal and were not adjusted for multiple comparisons Comparison of plan complexity and delivery efficiency metrics between PJ-VMAT and c-VMAT in the L-SBRT and S-LRT planning strategies Data are presented as mean ± standard deviation. L-SBRTPJ and L-SBRTc denote lattice stereotactic body radiotherapy plans generated using PJ-VMAT and conventional VMAT, respectively; S-LRTPJ and S-LRTc denote standalone lattice radiotherapy plans generated using PJ-VMAT and conventional VMAT, respectively. MU, monitor units; LT, leaf travel; MCSv, modulation complexity score for VMAT; ALG, average leaf gap; psmall,x mm, percentage of apertures smaller than x mm; SASx mm, small-aperture score at an aperture threshold of x mm. Lower MCSv values indicate greater modulation complexity. a L-SBRTPJ versus L-SBRTc; b S-LRTPJ versus SLRTc.All p values are nominal and were not adjusted for multiple comparisons LTV and GTV dosimetric parameters As summarized in Table 2, PJ-VMAT increased LTV and in both planning paradigms (all p≤0.040). LTV increased in L-SBRT (46.05±2.70 vs. 47.41±2.71 Gy, p=0.002) but did not differ in S-LRT (48.61±2.19 vs. 48.85±1.34 Gy, p=0.737). Dose fall-off outside the LTV was steeper with PJ-VMAT, with GI decreasing from 12.72±6.88 to 10.55±5.02 in L-SBRT (p=0.001) and from 8.33±3.61 to 7.12±2.21 in S-LRT (p < 0.001). Within the GTV, the high-dose metrics and did not differ significantly between techniques (all p > 0.05), whereas , , , and were lower with PJ-VMAT in both planning paradigms (all p≤0.005). Thus, PJ-VMAT maintained the high-dose component of the GTV while reducing intermediate- and low-dose exposure and improving dose fall-off around the lattice vertices. Peak-to-valley dose ratio As summarized in Table 2 and illustrated in Fig. 4, the effect of PJ-VMAT on PVDR varied according to the PVDR definition and planning paradigm. In L-SBRT, PVDR increased from 2.34±0.41 with c-VMAT to 2.43±0.39 with PJ-VMAT (p=0.003), PVDR increased from 2.38±0.38 to 2.47±0.37 (p < 0.001), and PVDR increased from 2.41±0.35 to 2.51±0.37 (p < 0.001). PVDR did not differ significantly (1.83±0.28 vs. 1.85±0.27, p=0.117). In S-LRT, all four metrics increased with PJ-VMAT: PVDR from 7.36±5.38 to 8.49±5.79, PVDR from 16.42±11.13 to 18.69±11.76, PVDR from 3.17±0.45 to 3.48±0.53, and PVDR from 3.35±0.63 to 3.61±0.68 (all p < 0.001). Overall, PJ-VMAT improved three of the four PVDR metrics in L-SBRT and all four metrics in S-LRT. Paired comparisons of peak-to-valley dose ratio (PVDR) metrics between c-VMAT and PJ-VMAT in the L-SBRT and S-LRT planning paradigms. Panels (a)–(d) show L-SBRT, and panels (e)–(h) show S-LRT. PVDR–PVDR correspond to , , , and , respectively. Each point represents one patient, and paired points are connected to show the within-patient change from c-VMAT to PJ-VMAT Background dose and normal-tissue sparing PJ-VMAT generally produced OAR doses comparable to or slightly lower than those of c-VMAT, and the observed differences were mostly small. Detailed OAR results are provided in Supplementary Table S1. PJ-VMAT reduced background dose to the non-vertex tumor region, with lower GTV–LTV in both L-SBRT (p=0.001) and S-LRT (p < 0.001). GTV-shell , , , and were also lower with PJ-VMAT in both planning paradigms (all p < 0.001). For Body–GTV, PJ-VMAT reduced both and integral dose (all p < 0.005). Figure 5 summarizes the standardized paired effect sizes for the seven predefined dosimetric endpoints; all point estimates were positive and therefore favored PJ-VMAT. Plan complexity and delivery efficiency As shown in Table 3, PJ-VMAT increased monitor units and delivery time in both planning paradigms (all p < 0.001). Leaf travel was significantly shorter with PJ-VMAT, whereas MCSv was significantly lower in both L-SBRT and S-LRT (all p < 0.001), indicating reduced cumulative MLC travel but greater modulation complexity. The aperture-related changes differed between planning paradigms. In L-SBRT, PJ-VMAT produced a smaller ALG and higher and SAS values at all evaluated thresholds (all p < 0.001). In S-LRT, , and SAS were higher with PJ-VMAT (all p < 0.001), whereas ALG, , SAS, and SAS did not differ significantly. Overall, PJ-VMAT increased delivery burden in both planning paradigms, while its effect on fine-scale aperture modulation was more pronounced in L-SBRT. Exploratory geometry-based association analysis Figure 6 summarizes the exploratory associations between tumor-geometric descriptors and the seven benefit-oriented dosimetric differences. In L-SBRT, the largest positive coefficients were observed for PVDR and GTV–LTV . PVDR was positively associated with GTV volume (ρ=0.556), Y (ρ=0.554), X (ρ=0.667), \max(X,Z) (ρ=0.488), and GBI (ρ=0.549). The benefit in GTV–LTV was also positively associated with GTV volume (ρ=0.570), Y (ρ=0.545), X (ρ=0.530), \max(X,Z) (ρ=0.436), and GBI (ρ=0.531). PVDR showed a positive association with Y (ρ=0.467). In S-LRT, associations for most PVDR and regional-dose endpoints were weak or inconsistent. In contrast, the benefit in Body–GTV integral dose was positively associated with GTV volume (ρ=0.611), Y (ρ=0.419),X (ρ=0.589), Z (ρ=0.523), \max(X,Z) (ρ=0.596), and GBI (ρ=0.598). Because the paired differences were benefit-oriented, these positive coefficients indicate that the dosimetric advantage of PJ-VMAT increased with the corresponding geometric descriptor. These findings are descriptive and hypothesis-generating and should not be interpreted as establishing predictive thresholds or supporting individual patient selection. Discussion In this retrospective planning study of 19 patients with bulky tumors, PJ-VMAT demonstrated favorable planning-level dosimetric performance compared with c-VMAT in both the L-SBRT and S-LRT settings. PJ-VMAT produced steeper dose fall-off around the lattice vertices, improved three of the four PVDR metrics in L-SBRT and all four PVDR metrics in S-LRT, and reduced dose to the GTV–LTV region, GTV shell, and Body–GTV. These gains were achieved while largely preserving the high-dose component of the GTV, but at the cost of increased monitor units, longer delivery time, and greater modulation complexity. Taken together, the findings indicate that longitudinally segmented VMAT may improve peak–valley dose separation and background-dose suppression in lattice planning, although the present conclusions remain limited to treatment-planning performance. The longitudinal piecewise VMAT concept proposed by Xu et al. for multiple brain metastases provided the methodological basis for the present approach [19]. In this study, the concept was modified and adapted to lattice radiotherapy, including the use of a fixed collimator angle rather than the orientation used in the original implementation. Because the resulting segment-specific fields formed a pipe-joint-like arrangement, this lattice-specific implementation was termed PJ-VMAT. Its dosimetric advantages may be related to assigning each arc to a smaller longitudinal portion of the target, thereby reducing the beam’s-eye-view field extent and the number of spatially separated vertices requiring simultaneous MLC modulation. This localized optimization may reduce island-blocking-like irradiation of the intervening valley regions and improve dose confinement [19, 20]. The fixed collimator orientation also aligned MLC leaf motion with the longitudinal segmentation, which is consistent with the shorter leaf travel observed for PJ-VMAT. However, these mechanisms were inferred from the planning design and complexity metrics rather than directly measured. To our knowledge, this is the first study to systematically evaluate such a lattice-specific adaptation of longitudinal piecewise VMAT under both L-SBRT and S-LRT planning paradigms. The different responses of L-SBRT and S-LRT provide further insight into the role of the background-dose prescription. In L-SBRT, the simultaneous 12-Gy whole-GTV objective establishes a relatively high baseline valley dose, thereby limiting the extent to which additional valley-dose suppression can increase PVDR. Consequently, PJ-VMAT improved three of the four PVDR definitions, but the magnitude of change was generally modest. In S-LRT, no simultaneous whole-GTV background-dose objective was applied, so reductions in intermediate- and low-dose exposure translated more directly into higher PVDR values. Accordingly, all four PVDR metrics improved with PJ-VMAT in S-LRT. These findings indicate that PVDR improvements were more consistent in S-LRT, rather than demonstrating greater overall benefit across all dosimetric endpoints. The magnitude and dispersion of PVDR also depended strongly on the definition used (Table 2 and Fig. 4). The wide distributions of PVDR and PVDR in S-LRT should not be interpreted solely as inconsistent plan quality. Their denominators, and , are highly sensitive to small low-dose or unirradiated portions of the GTV, and this sensitivity is amplified when no whole-GTV background dose is prescribed. Tumor geometry, vertex distribution, and case-specific optimization difficulty may therefore produce substantial variation in these percentile-based ratios. By comparison, PVDR and PVDR, which use and the mean GTV dose as valley surrogates, were less dispersed. Figure 4 should therefore be interpreted as showing paired within-patient trends and interpatient variability rather than providing generalizable quantitative predictions. More broadly, current DVH-based PVDR metrics compress a three-dimensional heterogeneous dose distribution into one-dimensional summaries and cannot fully describe the spatial location, continuity, or connectivity of valley regions. Development of more reproducible and spatially resolved PVDR metrics remains an important direction for three-dimensional LRT [32]. The exploratory geometry-based association analysis provides an additional perspective on the variability of PJ-VMAT benefit. Because the paired differences were defined in a benefit-oriented direction, positive correlations in Figure 6 indicate that the dosimetric advantage of PJ-VMAT increased with the corresponding geometric descriptor. Geometry-related associations were more evident in L-SBRT, whereas the S-LRT associations were less consistent and were mainly observed for Body–GTV integral dose. This does not imply that L-SBRT derived greater absolute benefit; rather, the benefit in L-SBRT may have been more heterogeneous across patients, allowing geometric factors to explain a larger proportion of the observed variation. Although PJ-VMAT segments the target primarily along the superior–inferior direction, the BEV projection changes continuously during arc delivery. Overall GTV volume, transverse dimensions, and \max(X,Z) may therefore also influence aperture enlargement and simultaneous modulation burden in c-VMAT. The GBI was constructed to summarize these geometric features, but it remains an exploratory descriptor. Given the small and heterogeneous cohort, these associations are hypothesis-generating and should not be used to establish predictive thresholds or guide individual patient selection. The complexity findings should be interpreted as a benefit–cost trade-off. PJ-VMAT reduced total leaf travel, consistent with the localization of MLC motion within individual longitudinal segments, but produced lower MCSv values and a greater contribution from narrow apertures, indicating more intensive local modulation. This pattern was accompanied by approximately doubled monitor units and substantially longer estimated delivery times. Thus, longitudinal segmentation simplified the spatial range of leaf motion but increased the number and complexity of localized modulation tasks. Whether the resulting dosimetric gain justifies this additional delivery burden will depend on measurement-based delivery accuracy, treatment efficiency, and ultimately the clinical relevance of the achieved dose redistribution. Several limitations should be acknowledged. First, this retrospective study included only 19 patients with substantial heterogeneity in tumor site, volume, shape, and lattice configuration. The limited sample size may have contributed to the observed interpatient variability and reduced the stability of the exploratory geometry–dosimetry associations; validation in larger and more homogeneous cohorts is therefore warranted. Second, no formal multiplicity adjustment was applied to the paired dosimetric comparisons or exploratory correlation analyses. All reported p values are nominal, and isolated or marginally significant findings should be interpreted cautiously. The conclusions should be based primarily on the magnitude, direction, and consistency of effects across related endpoints. Third, a fixed four-pipe configuration was used for all patients as a standardized and pragmatic planning approach, but the optimal number, width, and overlap of pipe segments may vary according to individual target geometry. In addition, all plans were generated using the Eclipse treatment planning system and TrueBeam delivery platform; differences in optimization algorithms, MLC and jaw design, mechanical constraints, and delivery-control characteristics may limit direct generalization to other systems. Fourth, this was a planning-only study without measurement-based patient-specific quality assurance, portal dosimetry, or delivery-log verification. Consequently, the findings characterize planning-level dosimetric performance and complexity rather than experimentally verified delivery accuracy. Finally, the longer delivery time associated with PJ-VMAT may increase susceptibility to intrafraction motion, particularly for thoracic and abdominal tumors. Respiratory motion was not assessed using four-dimensional dose analysis, and it remains uncertain whether the observed dosimetric improvements translate into clinically meaningful gains in tumor control, toxicity reduction, or other patient outcomes [33]. Future studies should focus on four complementary directions. First, the planning-level findings and exploratory geometry–dosimetry associations should be validated in larger, preferably multicenter cohorts, followed by prospective evaluation of whether the observed dosimetric advantages translate into improvements in tumor control, toxicity, and other clinically meaningful outcomes. Second, patient-specific pipe configurations should be systematically investigated, including the number, width, and overlap of longitudinal segments, to determine whether individualized designs can further improve dosimetric performance while limiting plan complexity and delivery burden. Third, the delivery accuracy and reproducibility of PJ-VMAT should be evaluated using measurement-based patient-specific quality assurance and/or treatment delivery-log analysis. Finally, automated planning workflows should be developed to standardize target segmentation, jaw assignment, pipe-configuration selection, and plan optimization, thereby reducing manual workload and facilitating reproducible implementation of PJ-VMAT [34]. Conclusion This planning study evaluated pipe-joint volumetric modulated arc therapy (PJ-VMAT), a lattice-specific adaptation of longitudinal piecewise VMAT, for bulky tumors. Compared with conventional VMAT, PJ-VMAT produced steeper dose fall-off around the lattice vertices, improved all four PVDR metrics in the S-LRT setting and three of the four metrics in the L-SBRT setting, and reduced background dose to the GTV–LTV region, GTV shell, and Body–GTV. These dosimetric gains were accompanied by increased monitor units, greater modulation complexity, and longer delivery time. The findings support favorable planning-level dosimetric performance of PJ-VMAT but do not establish clinical deliverability or superiority. Exploratory geometry-based associations suggest that the magnitude of benefit may vary with tumor geometry, particularly in L-SBRT, but these findings should not be used for individual patient selection. Further validation through measurement-based quality assurance, individualized pipe-configuration studies, and larger prospective clinical investigations is required to determine whether the observed dosimetric advantages translate into meaningful clinical outcomes. Electronic supplementary material Below is the link to the electronic supplementary material. Supplementary Material 1 Acknowledgements No acknowledgements. Abbreviations Anisotropic Analytical Algorithm Average leaf gap Beam’s-eye view Conventional volumetric modulated arc therapy Digital Imaging and Communications in Medicine Geometry burden index Gradient index Gross tumor volume Integral dose Lattice radiotherapy Lattice stereotactic body radiotherapy Leaf travel Lattice target volume Modulation complexity score for VMAT Multileaf collimator Monitor unit Organ at risk Pipe-joint volumetric modulated arc therapy Peak-to-valley dose ratio Small-aperture score Standalone lattice radiotherapy Spatially fractionated radiation therapy Volumetric modulated arc therapy Author contributions Zhitao Dai designed the study framework. Changyou Zhong and Meiling Xu conducted data analysis and drafted the manuscript. Shengzhen Dong and Cuiting Xu prepared Figs. 1, 2, 3, 4, 5 and 6. Zhendong Jiang and Haidong Yu reviewed the treatment plans. All authors reviewed and approved the final manuscript. Funding This study was sponsored by Meizhou Medical Science Research Project (2025C0302006), Shenzhen Postdoctoral Research Funds (25005), Cancer Hospital Chinese Academy of Medical Sciences, Shenzhen Center (Hospital Research Project, E010322006), Sanming Project of Medicine in Shenzhen (SZSM201612063) and Shenzhen Key Medical Discipline Construction Fund (SZXK013). Data availability The datasets used and/or analyzed during the current study are available from the corresponding author upon reasonable request. Declarations Ethics approval and consent to participate The study was approved by the Ethics Committee of Meizhou People’s Hospital (IRB2025-C-191) and was conducted in accordance with the ethical principles of the Declaration of Helsinki.We confirm that all methods were carried out in accordance with relevant guidelines and regulations. Competing interests The authors declare no competing interests. Footnotes Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Changyou Zhong and Meiling Xu contributed equally to this work. References Associated Data Supplementary Materials Supplementary Material 1 Data Availability Statement The datasets used and/or analyzed during the current study are available from the corresponding author upon reasonable request.