The dosimetric impact of stabilizing spinal implants in radiotherapy treatment planning with protons and photons: standard titanium alloy vs. radiolucent carbon‐fiber‐reinforced PEEK systems

Abstract Background Throughout the last years, carbon‐fibre‐reinforced PEEK (CFP) pedicle screw systems were introduced to replace standard titanium alloy (Ti) implants for spinal instrumentation, promising improved radiotherapy (RT) treatment planning accuracy. We compared the dosimetric impact of both implants for intensity modulated proton (IMPT) and volumetric arc photon therapy (VMAT), with the focus on uncertainties in Hounsfield unit assignment of titanium alloy. Methods Retrospective planning was performed on CT data of five patients with Ti and five with CFP implants. Carbon‐fibre‐reinforced PEEK systems comprised radiolucent pedicle screws with thin titanium‐coated regions and titanium tulips. For each patient, one IMPT and one VMAT plan were generated with a nominal relative stopping power (SP) (IMPT) and electron density (ρ) (VMAT) and recalculated onto the identical CT with increased and decreased SP or ρ by ±6% for the titanium components. Results Recalculated VMAT dose distributions hardly deviated from the nominal plans for both screw types. IMPT plans resulted in more heterogeneous target coverage, measured by the standard deviation σ inside the target, which increased on average by 7.6 ± 2.3% (Ti) vs 3.4 ± 1.2% (CFP). Larger SPs lead to lower target minimum doses, lower SPs to higher dose maxima, with a more pronounced effect for Ti screws. Conclusions While VMAT plans showed no relevant difference in dosimetric quality between both screw types, IMPT plans demonstrated the benefit of CFP screws through a smaller dosimetric impact of CT‐value uncertainties compared to Ti. Reducing metal components in implants will therefore improve dose calculation accuracy and lower the risk for tumor underdosage.


| INTRODUCTION
Postoperative radiotherapy (RT) after spinal decompression and stabilization presents a common treatment combination in the surgical treatment of spinal tumors, which frequently faces challenges due to the presence of metallic implants. Based on computed tomography (CT) images, RT treatment planning, including both the contouring process of tumor and organs at risk (OAR) and the dose calculation, relies on accurate image information. 1,2 Dose calculation algorithms integrated in treatment planning systems (TPS) depend on correctly assigned Hounsfield units (HU), which are converted into relative stopping powers (SP) for proton RT and into relative electron densities (ρ) for photon RT. 3,4 Streak artifacts and the acquired CT values of the implant, which in case of highly absorbing materials tend to show saturation effects, contain uncertainties, that can lead to substantial erroneous calculated dose distributions. 1 Moreover, developed for biological materials, clinical dose calculation algorithms are often insufficient to precisely model physical interactions associated with metallic implants. [5][6][7] The relevance of these uncertainties depends on the treatment modality and the applied RT technique, such as type of particles, degree of intensity modulation, and beam geometry. 1, 5 In general, Therefore, the application of protons in the presence of metallic hardware is controversially discussed. Dependent on the location, size of the implant, and institution, opinions range from "strict contraindication" to RT making use of clinical "workarounds" to improve treatment accuracy. 2,5,[9][10][11] The decision "pro" or "contra" proton treatment is challenging, considering on the one hand clinical studies which reported reduced tumor control rates for chordoma patients with titanium alloy implants compared to patients without surgical implants [9][10][11] and on the other hand the chance for improved clinical outcome compared to photon RT. The advantageous OAR sparing and possibility to apply higher tumor doses 9,10,12 indeed motivate to make use of the favorable particle properties.
For both proton and photon RT, the dosimetric impact of surgical implants as well as clinical measures to reduce associated treatment errors have been investigated broadly (e.g., Refs. [2,5,13]). The penetration of beams through implants should generally be avoided, 2,4,7 which isdependent on the tumor and implant locationnot always possible. Additionally, the manual assignment of CT values for artifacts and implants is frequently suggested and part of clinical practice. [8][9][10][11]14 While well-investigated methods exist to determine calibration curves which relate HUs to SPs, 3,4,15 the definition of metallic volumes is crucial and results in delineation and consequently dosimetric uncertainties. 1,2,5,16 Threshold-based auto-segmentation allows for delineation of CT regions within a specified range of HUs and promises hereby to reduce the subjectivity of the delineation process, but uncertainties in the CT scan due to the extension of the CT-value range of the scanner 17 and partial volume artifacts 18 limit the accuracy of volume definition. Improvements of contouring and planning accuracy are expected by the application of MV-CTs or dual energy CT for planning, as well as by advancing dose calculation and metal artifact reduction algorithms. 1, 6,14,[19][20][21][22][23][24] Recent developments in neurosurgery, the introduction of carbon-fiber-reinforced Polyetheretherketone (CFR-PEEK, or CFP for short) pedicle screw systems, may "solve" or at least improve the prescribed concerns in the future. 25-28 Dependent on the system and manufacturer, CFP systems feature a remarkably reduced amount of titanium or no metallic components, and thus decrease the impact of several correlated uncertainties. With its low atomic number, carbon-based materials have favorable radiation properties compared to stainless steel and titanium alloy (Ti) implants and therefore have raised researcher's interest in the field of RT. [28][29][30] Previously published measurements and simulations of proton and photon beam behavior through CFP compared to typical surgical metal implants underlined along with other investigations the promising properties of CFP implants. [28][29][30] Although carbonic screws are also affected by HU uncertainties, as any component of the human body is, especially dense bones, associated uncertainties are considered to be remarkably smaller than those of high Z materials.
We expand the investigations on CFP implants by a retrospective planning study for intensity modulated proton (IMPT) and photon therapy (volumetric arc technique -VMAT), which examines the dosimetric impact of uncertainties in CT values of Ti in CFP pedicle screw systems, which comprise minor Ti components, compared to standard Ti systems. Uncertainties in HU assignment of Ti were simulated by varying SP and density, for IMPT and VMAT plans, respectively.

2.A | Patient information and contouring
The study is based on CT data of 10 patients, who were previously treated at our institute for spinal metastases with postoperative photon therapy after spinal decompression and stabilization surgery. All planning CTs, performed with a Siemens Somatom Emotion 16 Scanner (Siemens, Erlangen, Germany), at a tube voltage of 130 kV and variable mAs, had a resolution of 1 mm × 1 mm and a slice thickness of 3 mm. Five patients featured standard titanium alloy (Ti) implants and five carbon-fibre-reinforced PEEK (CFP) pedicle screw systems (Icotec, Altstätten, Switzerland). 26 Carbon systems consisted of radiolucent non-metallic CFP pedicle screws (diameter: 6.5 mm), CFP rods (diameter: 5.5 mm), a thin titanium coating in the pedicle area and titanium tulips. Ti screws measured a diameter of 6 or 7 mm and rods of 5.7 mm.
Planning target volumes (PTV) were located in the lumbar vertebral spine (9 patients) and the sacrum (1 patient). All patients received a monosegmental posterior instrumentation comprising of four pedicle screws (2 × 2) and 2 rods. Artifacts and metal components were generated systematically by applying threshold-based auto-segmentation, followed by manual adaption of the derived structures. Contouring was performed by the same staff member to minimize interpersonal variations. Since the number of slices which contained Ti components was comparably small compared to the whole PTV, a contour, referred to as "PTV local," was defined as the PTV structure over the CT slices with the largest fraction of the screws. It consisted of 10 (2 × 5) PTV slices (Table 1) for all patients.
PTV local reflects the dosimetric situation in proximity of the implants and allows for a better analysis of dosimetric changes inside the PTV for all patients. As critical organs varied between patients, a ring structure ("PTV ring") of a 5 cm margin around the PTV extended by 5 mm was generated to evaluate dose to normal tissue ("PTV ring" = "PTV expanded by 5.5 cm" minus "PTV expanded by 0.5 cm").

2.B | Retrospective planning
Two initial plans were generated for each patient, one IMPT and one VMAT plan. IMPT plans were optimized and calculated in the research treatment planning system (TPS) matRad 31  To simulate uncertainties in the HU assignment of Ti components, varying relative stopping powers and electron densities were ascribed to the structure. Nominal plans were calculated with an SP and ρ of 3.2, for protons and photons, respectively. Both plans were recalculated onto the identical CT with increased and decreased SP or ρ by ±0.2 (~6%) to 3.0 and 3.4 (workflow, see Fig. 1).

2.C | Plan evaluation
Plan quality was evaluated by target coverage, homogeneity, and by several dose volume histogram (DVH) criteria of the PTV, PTV local, and PTV ring. The maximum dose was analyzed by D2, the maximum dose received by 2% of the associated structure, and the minimum dose by D98, the minimum dose received by 98% of the structure. The standard deviation σ within the PTV served as a measure for target dose homogeneity.

3.A | Volumes
The delineated structures of Ti components of both pedicle systems do not exactly reflect the real material outlines due to inaccurate representation in the CT scan; exemplary contours of one patient with CFP and one with titanium screws are presented in Fig. 2. Corresponding volumes of metallic components and image artifacts were smaller for CFP than for titanium alloy systems (Table 1).

3.B | Nominal plans
For both implants, dose to healthy tissue ( Table 2) was smaller for IMPT compared to VMAT plans, with hardly any difference between the screw types (Fig. 3). Photon dose distributions presented a more homogeneous and slightly superior PTV coverage compared to protons for both implants.
The comparison of CFP vs Ti for the initial VMAT plans by several dosimetric quality indicators (D2, D98, D95, σ of the PTV local) presented hardly any difference.
IMPT dose distributions showed reduced dose conformity and homogeneity in proximity to the Ti implants, caused by the large density gradient of penetrated materials (Fig. 3, left images of Fig. 4).
Coverage and PTV dose homogeneity were superior for CFP than for Ti implants.

3.C | Recalculated plans
Recalculated VMAT dose distributions of both pedicle screw types were hardly influenced by the simulated HU deviations. Changes of DVH criteria were slightly larger for Ti, but all deviations were smaller than 1%, except for one case where the recalculated σ (PTV local) deviated by 2% from the nominal plan (Fig. 5  One limitation, that has to be considered, is that our investiga-

Carbon
PTVlocal -initial plan PTVlocal -reduced SP PTVlocal -increased SP F I G . 7. Dose volume histograms of the "planning traget volumes (PTV) local," that is, the PTV region in proximity of the screw slices of one initial intensity modulated proton plan and two corresponding recalculated plans for Ti and CFP screws. Dose volume histogram correspond to the dose distributions of Fig. 4. The shape of the DVH curves of the target underlines the more stable coverage for C screws and indicates the regions of under-and overdosage by the reduced steepness for Ti. Generally, CFP pedicle screw systems present a reduction of metal compared to their metallic analogues such that obviously associated uncertainties will be reduced. Less artifacts and dose calculation errors, due to inaccurate simulated particle transport in metal, will improve dose calculations. Although not investigated here, the contouring process of targets and critical structures is expected to gain in precision from better visualization, with a particular benefit for more