If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Brain Tumor Center & Neuro-Oncology Unit, Beth Israel Deaconess Medical Center, Boston, MassachusettsDivision of Hematology/Oncology, Department of Medicine, Rhode Island Hospital, Providence, Rhode Island
Brain Tumor Center & Neuro-Oncology Unit, Beth Israel Deaconess Medical Center, Boston, MassachusettsDivision of Hematology/Oncology, Department of Medicine, Rhode Island Hospital, Providence, Rhode Island
Brain Tumor Center & Neuro-Oncology Unit, Beth Israel Deaconess Medical Center, Boston, MassachusettsDivision of Hematology/Oncology, Department of Medicine, Rhode Island Hospital, Providence, Rhode IslandDepartments of Neurology, Neurosurgery & Radiation Oncology, Rhode Island Hospital, Providence, Rhode Island
Tumor Treating Fields (TTFields) are alternating electric fields at 150-200 kHz that exert their anti-cancer effect by destroying tumor cells when they undergo mitosis. TTFields are currently being tested in non-small cell lung cancer patients with advanced disease (NCT02973789) and those with brain metastasis (NCT02831959). However, the distribution of these fields within the thoracic compartment remains poorly understood. Using PET-CT image datasets obtained from a series of 4 patients with poorly differentiated adenocarcinoma, the PET-positive gross tumor volume (GTV), clinical target volume (CTV) and structures from the chest surface to the intrathoracic compartment were manually segmented, followed by 3-dimensional physics simulation and computational modeling using finite element analysis. Electric field-volume histogram, specific absorption rate-volume histogram and current density-volume histogram were generated to produce plan quality metrics (95%, 50%, and 5% volumes) for quantitative comparisons between models. Unlike other organs in the body, the lungs have a large volume of air, which has a very low electric conductivity value, and this aspect is unique for the lung models. Our comprehensive and individualized models demonstrated heterogeneity in electric field penetration to the GTVs with differences upwards of 200% and yielded a diverse range of TTFields distributions. Target contact with the conductive pleura intensified TTFields at the GTV and CTV. Furthermore, in a sensitivity analysis, varying electric conductivity and mass density of the CTV altered TTFields coverage to both the CTV and GTV. Therefore, personalized modeling is important to accurately estimate target coverage at the tumor volumes and surrounding normal tissue structures in the thorax.
Introduction
Non-small cell lung cancer (NSCLC) accounts for the majority of lung cancers for both men and women in the United States. In 2017, there were 222,500 new cases of lung cancer and about 155,870 disease-related death (1) The standard methods of treatment for NSCLC include surgery, radiotherapy and drugs, which consist of cytotoxic chemotherapies, targeted agents or immunotherapy depending on the results of molecular profiling of the tumo (2) However, in a large number of patients, the adverse effects often limit certain clinical care options and reduce quality of life for them.
Tumor Treating Fields (TTFields) are loco-regional treatments that are pertinent to advanced NSCLC. TTFields are alternating electric fields at 150-200 kHz that are applied externally to the body via two pairs of transducer arrays positioned orthogonally from each other and are connected to a portable electric field generator. The efficacy of this type of treatment was initially investigated in glioblastoma and it was shown to prolong the survival of newly-diagnosed patient (3) Its efficacy is now being tested in randomized phase III clinical trials for stage IV non-small cell lung cancer patients after platinum failure (NCT02973789) and those with brain metastases following stereotactic radiosurgery (NCT02831959).
The anti-cancer efficacy of TTFields is based on its anti-mitotic activity. Septin and Tubulin are two intracellular proteins that have large dipole moments and are respectively disrupted by TTFields during cytokinesis and segregation of sister chromatids during anaphase (4) (5). In preclinical models of glioma cells, the extent of apoptosis depends on both the frequency and intensity of the applied alternating electric fields (6). For NSCLC cell lines, the optimal frequency is estimated near 150 kHz (7) (8) and both NovoTTF-100L (NCT02973789) and NovoTTF-100M (NCT02831959) devices currently being tested in clinical trials are also operating at this frequency. However, propagation of TTFields in the intrathoracic compartment is poorly understood. Unlike the brain in which it is immersed in highly conductive cerebrospinal fluid (9), the air within the pulmonary alveoli and bronchi is an electrically non-conductive medium that can potentially alter the behavior of TTFields. Therefore, we analyzed the electric field distribution, specific absorption rate (SAR), and current density in the thoracic cavity of 4 patients with poorly differentiated adenocarcinoma of the lung. Furthermore, the respective electric field-volume histograms (EVH), SAR-volume histograms (SARVH) and current density-volume histograms (CDVH) were constructed and the plan quality metrics (PQM) for the various segmented tissues including but not limited to, gross tumor volume (GTV), clinical tumor volume (CTV) and segmented normal intrathoracic anatomy were computed (10). Since TTFields therapy has recently emerged as a potential treatment for malignancies as in non-small cell lung cancers, we therefore investigated the penetrative ability and field distribution characteristics in silico.
Materials and Methods
Image Preparation for anatomical segmentation
DICOM imaging data for 4 non-small cell lung cancer patients were obtained according to an institutional review board-approved protocol from Dana Farber/Harvard Cancer Center for retrospective imaging analysis. Details regarding presentation and treatment histories of these patients can be found in Appendix 1. Computed tomography (CT) and positron-emission tomographic (PET) images were used in this study, where CT image segmentation for various tissues in the thorax was performed in ScanIP (Simpleware LTD., UK), and then co-registered with PET to delineate the gross tumor volume (GTV). Other computer modeling studies have indicated that the accuracy of a model is highly dependent on the input parameters and image resolution used to delineate various anatomic structures such as those described in Table 1 (11) (12) (13) (14) (15). Therefore, each image dataset was super-sampled to a 1 × 1 × 1 mm resolution prior to segmentation using nearest-neighbor interpolation between slices to minimize voxel edge effects.
Table 1Electric conductivity and mass density values for GTV, CTV, CTVp_ring, and other anatomic structures within the thorax.
Anatomical segmentation and preparation for mathematical mesh model
Intrathoracic tissues are not uniform across individual patients and they include, but are not limited to, various thoracic muscles, cortical and cancellous bone, blood vessels, air cavities, and cardiac structures. Airway structures such as carina and trachea, as well as vessels such as the inferior vena cava (IVC) and superior vena cava (SVC), were segmented separately to serve as anatomical landmarks for spatial reference. There is TTFields heterogeneity between models in the thoracic structures segmented due to the unique location of each gross tumor volume (GTV). A uniform 3-mm expansion around each GTV was performed and segmented as the clinical target volume (denoted as CTV) to include subclinical microscopic disease not necessarily detectable from radiological imaging. We chose a 3-mm expansion based on the fact that TTFields are not as conformal, and these field margins are much less well-defined, compared to ionizing radiation. Therefore, a smaller expansion of 3 mm is more appropriate than the 5-8 mm expansion for conventional radiotherapy. Furthermore, the CTV is defined as the expansion from and inclusive of the GTV to be consistent with the definition described in ICRU50 and 62 (16) (17). In this work, GTV and GTVp, as well as CTV and CTVp, will be used interchangeably to denote the primary tumor, because none of the models had nodal disease (GTVn and CTVn). A separate volume labeled CTVp_ring, which is a volumetric shell derived from the difference between CTVp and GTVp, was delineated in each model in order to apply various modeling conditions without disturbing the properties of the GTV within the CTV.
Transducer arrays with a layer of conductive hydrogel between each electrode and the skin surface were manually placed on the external skin contour, where arrays of 20 electrodes were placed anteriorly, posteriorly, right laterally and left laterally (Figure 1A-E). Within an array, each electrode was placed 2 cm apart radially from the center of adjacent electrodes (18). The array placement was arranged in a manner to reduce the radial distance of each array from the CTV in each patient model.
Figure 1Three-dimensional rendering of the thorax external contour, lungs and GTV (also denoted as GTVp) for each patient. The reconstructed thorax is displayed with the accompanying anterior arrays in turquoise color, left lateral arrays in magenta, posterior arrays in orange, and right lateral arrays in red (A-E). Original external contour without (inset of A) and with (inset of B) double mastectomy is shown for Subject 1. Axial electric field distribution maps for each patient are displayed, with GTV outlined in green and CTV (also denoted as CTVp) in magenta (F-J). Axial current density distribution maps for each patient are shown, with GTV shown in green, and CTV in magenta (K-O). Coronal SAR distribution maps for each patient are presented, with GTV displayed in green and CTV in magenta (P-T).
SUBJECT 1 has a much smaller GTV than other models in this study, and its borders are located far away from the chest wall, pericardium space, or diaphragm. Therefore, expansion models were generated to 32 and 64 times their original volume in order to determine whether the size of the GTV or its location was the primary determinant of electric fields intensity. Consequently, 64x GTV gained sufficient volume until it became partially in contact with the pleura. A 3-mm limited expansion around the GTV was also performed to maintain definition consistency but was not allowed to protrude outside the pleural cavity (Figure 3A-B). Four additional models where the GTV was virtually shifted in the axial plane while maintaining their original superior-inferior positions, with the nomenclature of (i) anterior shifted, (ii) posterior shifted, (iii) right-lateral shifted, and (iv) left-lateral shifted (Figure 3C-G). For CTVs, similar limited expansions by 3 mm were made where extrusion beyond the pleural cavity was forbidden.
Computational modeling of electric field distributions
A 3-dimensional finite element mesh was generated within ScanIP (Simpleware LTD, UK) for each model with a minimum tetrahedra edge length of 1.5 mm, and maximum tetrahedra edge length no greater than 5 mm in order to optimize the quality for smaller or irregular tissues without compromising computation time for larger or smooth structures in each model. Target maximum error of 0.05 mm was employed to force the adaptive surface remeshing algorithm to avoid the iterative remeshing from assigning finite elements too distant from the original mesh. COMSOL Multiphysics (COMSOL, Burlington, MA) was used to perform all physics simulations, where appropriate physical tissue properties (Table 1) and boundary conditions were applied, assuming negligible magnetic fields and acceptable quasi-static approximations (19) (20) (21). EVH, SARVH, and CDVH were generated to produce evaluation metrics for all quantitative comparisons between models. The initial conditions included an assumption that there was no pre-existing electric potential anywhere in the geometry prior to the delivery of the TTFields. A continuous 150 kHz sinusoidal wave and 50 volts peak to peak, was applied to the anterior-posterior (AP) and opposing posterior-anterior (PA) arrays, as well as another sinusoidal wave of equal potential was applied to the right-left lateral (RLAT) and opposing left-right lateral (LLAT) arrays. Plan quality metrics (PQM) were generated for a comprehensive analysis of the results including variations of those described in Supplemental Table 1-5. These PQMs include (i) the E95%, E50%, and E5% for the electric field strength received by respectively 95%, 50%, and 5% of a particular tissue structure, (ii) SAR95%, SAR50%, and SAR5% for the specific absorption rate received by respectively 95%, 50%, and 5% of a particular tissue structure, (iii) CD95%, CD50%, and CD5% for the current density received by respectively 95%, 50%, and 5% of a particular tissue structure.
Results
TTFields distribution differs among individual NSCLC patients
The computational modeling of TTFields distribution, based on electric field, SAR and current density PQMs, was performed in 4 patients with unresectable NSCLC, where the GTVs were identified in the right upper lung (SUBJECT 1 and SUBJECT 2), right middle lung (SUBJECT 3), and left upper lung (SUBJECT 4) (Figure 1F-T). Their imaging datasets were ideal for modeling analysis because none had intrathoracic pathology nor prior surgery that might have distorted baseline anatomic structures in the thorax. Our modeling revealed that the distribution of TTFields at the GTV varied by as much as 4 orders of magnitude among the models. SUBJECT 1 had the lowest PQMs, E95% (0.1 V/m), E50% (0.3 V/m), and hotspot E5% (0.4 V/m), while SUBJECT 2 had the highest E95% and E50% (137.0 V/m and 243.0 V/m, respectively), and SUBJECT 4 had the highest E5% (443.8 V/m) (Figure 2B). The lowest SAR95%, SAR50% and SAR5% were also found in SUBJECT 1 (<0.1 W/kg for all 3 metrics) while the highest SAR95% (10.6 W/kg) and SAR50% (36 W/kg) was in SUBJECT 2 and the highest SAR5% (114.7 W/kg) was in SUBJECT 4. Similarly, SUBJECT 1 also had the lowest CD95% (0.08 A/m2), CD50% (0.26 A/m2) and hotspot CD5% (0.43 A/m2), while the highest CD95% (13.8 A/m2) and CD50% (24.6 A/m2) were observed in SUBJECT 2, but the highest hotspot CD5% (44.9 A/m2) was found in SUBJECT 4. SUBJECT 1’s low PQMs may be related to its small GTV size and dissociation from the pleura. Therefore, the electric field, SAR and CD data at the GTV suggest that the location of the tumor within the thorax may be an important variable influencing the delivery of TTFields.
Figure 2PQMs for EVH, SARVH, and CDVH. Volume histograms EVH, SARVH, and CDVH are displayed for SUBJECT 2 for tumor targets as well as selected normal tissue structures (A). EVH, SARVH, and CDVH for targets only are shown for each patient model, including GTV (also denoted as GTVp), CTV (also denoted as CTVp), and CTVp_ring, revealing diversity in tumor coverage among individual patients (B).
To confirm our previous findings, we analyzed TTFields distribution in the CTV and revealed that SUBJECT 1 also had the lowest E95% (0.08 V/m), E50% (0.27 V/m) and hotspot E5% (0.48 V/m) compared to the highest E95% (105.72 V/m) and E50% (249.90 V/m) in SUBJECT 2 and E5% (582.19 V/m) in SUBJECT 4 (Figure 2B). A similar pattern was also observed in the SAR and CD metrics, where SUBJECT 1 had the lowest 95%, 50% and 5% metrics while SUBJECT 2 possessed the highest 95% and 50% metrics and SUBJECT 4 had the highest 5% metrics (Supplemental Tables 1-5). Therefore, our PQM analysis of the CTV confirmed our initial observations in the GTV.
Tumors dissociated from pleura has significantly reduced TTFields
We next investigated factors contributing to low TTFields in SUBJECT 1, with attention to the patient's breasts that increased the thoracic girth. The chest CT also revealed a large amount of fat and soft tissue between the skin and intrathoracic cavity compared to other models. Therefore, we sought to virtually remove the breasts in a separate model (denoted as SUBJECT 1_2xM) by subtracting the segmented breasts and leaving approximately 1 cm of soft tissue between chest wall and skin. Removing the breasts increased the TTFields strength marginally within the GTV by 4.3% in E95%, 1.4% in E50%, and 27.4% in E5% (Supplemental Table 1-2). No change in any of the SAR or current density PQM metrics were observed in the GTV. In contrast, the CTV, which encompasses the GTV, had a marked increase in CD95% by 128.8% while CD50% and CD5% increased minimally by 16.8% and 2.6%, respectively. A marginal increase was observed in E95% (3.6%), E50% (2.1%), and E5% (45.3%), as well as in SAR95% (0.2%), SAR50% (0.1%) and SAR5% (0.1%). Additionally, a 128.8% increase in CD95%, a 16.8% in CD50%, and 2.6% in CD5% was observed in the CTV. We then asked whether this increase in CD95% is due to the expanded volume shell of the CTV surrounding the GTV, or CTVp_ring. Indeed, there was a corresponding increase in CD95% of CTVp_ring by 216.1% while CD50% and CD5% increased minimally by 28.2% and 4.4%, respectively. Similarly, a marginal increase was also observed in E95% (3.3%), E50% (4.4%), and E5% (35.5%), as well as in SAR95% (0%) SAR50% (0%), and SAR5% (0.1%). Therefore, the virtual double mastectomy model revealed that breast tissue may be a contributing factor to the attenuation of TTFields intensity.
We then hypothesized that the radial distance from the tumor to transducer arrays may be an additional determinant of TTFields coverage of the GTV and CTV. Specifically, the distance from the centroid of each patient's GTV to the central electrode of each of the 4 arrays was calculated and averaged for each patient. None of the measurements revealed any significant difference in average distance from the center of each of the 4 arrays to chest wall despite the virtual double mastectomy (data not shown). We then performed additional distance measurements, including (i) the distance between the centers of each opposing array, (ii) distance from anteroposterior and lateral skin-skin distances at the central axial plane of each patient's respective GTVs, and (iii) circumference of each patient's model at the central axial plane. Still, none of these measurements revealed significant differences in the values of SUBJECT 1’s virtual double mastectomy model when compared to other models (data not shown). Therefore, additional modeling studies focused on altering the GTV and CTV were generated.
We focused on the SUBJECT 1 model by expanding the GTV volume from the original to 32 times (Figure 3B) and then to 64 times (Figure 3C). The 64x, but not the 32x, GTV had a significant increase in TTFields intensity (Figure 3H). However, it was still unclear whether this marked increase in field intensity was secondary to an increase in GTV volume or contact with the pleura as a consequence of the expansion. Therefore, using the original GTV, 4 virtual shifts of the tumor were made, each making partial contact with the pleural cavity (Figure 3D-G). The GTV shift producing the largest increase in TTFields intensity within all target volumes was a left lateral shift towards midline of the torso, while the smallest was an anterior shift, except for hotspot metrics E5%, SAR5%, and CD5% for the CTVp_ring, where a right lateral shift of the GTV produced the largest increase (Figure 3H). In all cases, shifting the GTV to make flush contact with the pleural cavity induced a dramatic increase in TTFields intensity compared to the original location. Therefore, we concluded that contact with the conductive pleura is paramount for intensifying TTFields at the GTV and CTV.
Figure 3Three-dimensional rendering of SUBJECT 1′s external contour, transducer arrays, lungs and GTV (also denoted as GTVp) with their respective GTV virtual spatial modifications. Rendering is shown as baseline (A), after 32x expansion (B) and after 64x expansion (C). The shift of the GTV is displayed as anterior shifted (D), posterior shifted (E), right shifted (F), and left shifted (G). Electric field, SAR, and current density comparisons for each modified SUBJECT 1 model's tumor targets, specifically GTV (green), CTV (also denoted as CTVp), and CTVp_ring (H). A significant increase in target coverage was noted when GTV was expanded to 64x, compared to baseline and expansion to 32x. Marked increase was also seen in the 95% and 50% volume metrics when the GTV was left shifted compared to anterior shifted, posterior shifted and right shifted models.
Varying electric conductivity and mass density of CTV alters TTFields coverage to CTV and GTV
The electric conductivity of patient lung tumors may vary significantly due to their heterogeneous tissue compositions. We then questioned to what extent does altering the conductivity affects TTFields coverage at the target volumes of GTV, CTV and CTVp_ring. Because CTV includes GTV and our prior analysis showed similar TTFields distribution in both, we therefore focused our modeling based on the conductivity of the CTV. In general, as the CTV conductivity increased, the field intensities at each target volume approached 0, most likely due to loss of charge retention. However, there are definite differences among the individual models. For example, in SUBJECT 4, the E95% metric for all 3 target volumes peaked with a CTV conductivity of 0.1 S/m, but neither E50% nor E5% exhibited a peak, indicating that this patient's tumor may have unique characteristics (Figure 4A). An opposite trend was observed for both SAR and current density, where less conductive CTV yielded low SAR and current density and vice versa for each of the 3 target volumes (Figures 4B and4C). However, SUBJECT 2 displayed a different pattern of SAR metrics, where SAR95%, SAR50% and SAR5% all exhibited a peak at a CTV conductivity of 0.1 S/m for each of the 3 target volumes. For SUBJECT 3 specifically, a peak was observed between 0.1 and 1 S/m for CTV conductivity, but the magnitude of SAR was much lower than that in SUBJECT 2 and SUBJECT 4. Together, these data showed a heterogeneous pattern of TTFields characteristics among each of the 4 lung models.
Figure 4Sensitivity analysis of CTVp conductivity on electric field, SAR and current density in all target volumes. (A) Varying electric conductivity of CTVp for each patient model revealed heterogeneity in tumor coverage in the GTV (also denoted as GTVp), CTV (also denoted as CTVp), and CTVp_ring. In general, as the conductivity of the CTVp increased, the electric fields coverage decreased. An opposite trend was demonstrated for SAR, except for SUBJECT 2, where a peak can be observed with conductivity of 1 S/m (B). Increasing conductivity of the CTVp_ring generally resulted in increased current density (C).
The CTV is an expansion of the GTV, and therefore the expanded volume shell or CTVp_ring is likely more heterogeneous than either GTV or CTV due to a mixture of tumor, lung tissue and air, all of which have very different conductivity values. We performed a sensitivity analysis by varying the conductivity of the CTVp_ring, while keeping that of the GTV constant at 0.1 S/m, to identify changes in TTFields distribution in each of the 3 target volumes (GTV, CTV and CTVp_ring). The resulting CTV E95% for SUBJECT 4 and SUBJECT 3 showed a peak near 0.1 S/m while the maximum for SUBJECT 2 was still near 0.001 S/m (Figure 5A). Similar peaks and maximum for the E95% metric were observed in CTVp_ring of the models but as expected the magnitude was lower than that of CTV. Notably, GTV E95% exhibited peaks near 0.1-1 S/m for all 3 models; 1 S/m for SUBJECT 4 and 0.1 S/m for both SUBJECT 2 and SUBJECT 3. Similar pattern of peaks was observed for GTV E50% and hotspot E5%, but no peak was observed for E50% and E5% of both CTV and CTVp_ring. For current density, peaks were observed in CTV CD95% and all 3 GTV CD metrics (CD95%, CD50% and CD5%) when the conductivity of CTVp_ring was near 0.1-1 S/m (Figure 5C). However, no change was apparent in plot characteristics in CD50% and CD5% for both CTV and CTVp_ring, other than changes in magnitude, at the higher end of the conductivity spectrum for CTVp_ring. Most importantly, SAR peaks were seen in every metric (SAR95%, SAR50% and SAR5%) across all models (SUBJECT 2, SUBJECT 3 and SUBJECT 4) and in each of the 3 target volumes (GTV, CTV and CTVp_ring), when the CTVp_ring conductivity was near 0.1-10 S/m (Figure 5B). Collectively, this sensitivity analysis revealed that CTVp_ring is a major determinant of TTFields coverage at the tumor and this is highly variable among individual patients.
Figure 5Sensitivity analysis of CTVp_ring conductivity on electric field, SAR and current density in all target volumes. Varying only the electric conductivity of CTVp_ring for each patient model revealed diversity in both sensitivity curve characteristics and tumor coverage of the GTV (also denoted as GTVp), CTV (also denoted as CTVp), and CTVp_rings (A). In general, as the conductivity of the CTVp increased, the electric fields coverage decreased. An opposite trend was noted for SAR, except for SUBJECT 2, where a peak can be observed with conductivity near 1 S/m. Increasing conductivity of the CTVp_ring generally resulted in increased current density (C).
SAR is inversely proportional to mass density. To further investigate this variable in SAR metrics, a sensitivity analysis was also performed by changing the mass density of CTVp_ring, ranging from values equivalent to air (1 kg/m3) to cortical bone (2000 kg/m3). For all 3 patient models (SUBJECT 2, SUBJECT 3 and SUBJECT 4), the 95%, 50% and 5% coverage metrics followed an expected decrease in SAR with increasing mass density (Supplemental Figure 1). However, the SAR coverage at the GTV remained constant, indicating that changes in the mass density of surrounding tissues do not affect SAR metrics within the GTV. As expected, the electric field and current density metrics were unchanged since neither of those are related to mass density.
Discussion
This is the first and largest comprehensive study of a lung cancer series for TTFields modeling. We used EVH, SARVH, CDVH metrics and their associated PQMs, including (i) E95%, SAR95%, and CD95%; (ii) E50%, SAR50%, and CD50%; (iii) E5%, SAR5%, and CD5%, to evaluate and compare TTFields coverage of target volumes among models. Although there is currently no unified method for quantifying the amount of TTFields delivered to target volumes in the body, these quantities are similar to the target coverage criteria for ionizing radiotherapy treatment plans recommended by the International Commission on Radiological Units 62, which includes D95% and other criteria. These coverage metrics allowed us to show, for the first time in silico using 4 non-small cell lung cancer patient models, that TTFields can penetrate thoracic cavities with highly heterogeneous distributions among individual patients. Second, proper assignment of physical properties for lung tumor volumes and surrounding normal tissues is critical for generating accurate prediction of TTFields distributions in the body. This is because changes in electric conductivity and mass density of the target volumes can have profound effects on their coverage. Lastly, we also found that tumor location, particularly when it is flushed against the pleura, can dramatically increase the penetration of TTFields.
The electric field distribution within the thorax is unique compared to other body compartments due to a large amount of non-conductive air in the lungs, which may cause a significant loss of electric field intensity in targets not bound to conducting tissues such as the pleura. Multiple muscle layers and numerous vasculatures also contribute to the heterogeneous intrathoracic anatomy among patients and this results in their highly variable TTFields distribution as shown in Table 1. Comparison of coverage between the lowest (SUBJECT 1) and highest (SUBJECT 2) at the GTV revealed a difference of nearly 200%. Even excluding the unique SUBJECT 1 model, the difference in GTV coverage metrics between the model with the next lowest coverage (SUBJECT 3) and the highest (SUBJECT 2) was still greater than 100%, and upwards of 200% in certain metrics. Therefore, personalized modeling is necessary to accurately determine TTFields distributions in individual patients.
There are pertinent examples of non-uniform TTFields distributions in our thoracic models. First, the entire skin layer is not saturated with TTFields, unlike the intense fields found at the scalp in head models as a result of the high impedance skull. Of note, the volume of skin directly underneath each active electrode still consistently has increased electric field intensity. In fact, our models showed that regions of higher TTFields intensity are distributed heterogeneously, though not randomly, throughout the thorax. For instance, tissues along the chest wall often exhibited much higher electric fields. This may be due to a buildup of charges before reaching the large air cavity in the lungs, which has low conductance and therefore electric current significantly decreases. It should be noted that although bronchial vessels are conductive due to the presence of blood, their small size and tubular structure likely function similar to electrical resistors with slightly lower impedance than the surrounding air and therefore cannot cause large changes in field strength. Second, it is acknowledged that mass densities of tissues are not necessarily homogeneous in the tumor or the lungs, and this can be distinguished on CT based on the Hounsfield units associated with differences in tissue absorption characteristics of X-rays. This is particularly important for lung tumors due to their variable histology, size and necrosis. However, at this time, we do not have the capability to incorporate variable mass density values of tissues in our models, and therefore we applied homogeneous values in GTV and CTV when we performed a sensitivity analysis of TTFields distribution in the thorax. Lastly, it can be observed in all of our models that electric fields are shunted away from deeper regions of the lung. This might be due to a fundamental property of electric currents to take the shortest path with the least resistance, and our current density maps for each model provided confirmation in Figures 1K-O. Together, these inherent variabilities provide a strong rationale for personalized modeling using each patient's individual radiological images when computing TTFields distribution in the thorax.
Analyses of SUBJECT 1’s GTV positioning and its expansion within the lung provided us the critical insight into factors that may decrease or increase TTFields’ target coverage. Air is a lossy conductive medium for electric currents and since TTFields propagation relies heavily on electrically conductive pathways, air in the lung has been shown to attenuate penetration into the tumor. Moreover, expansion analyses showed that TTFields coverage sharply increased when the GTV was enlarged to 64x, causing it to come into contact with the conductive pleura. These observations suggest that air or water adjacent to the GTV can attenuate or augment TTFields coverage.
A lung tumor suspended by vasculature and surrounded by air has low TTFields coverage but there are conditions that may alter this phenomenon. First, lobectomy or segmentectomy may leave behind a fluid fill space. The highly conductivity fluid may direct TTFields towards an adjacent tumor if it is partially in contact with the resection cavity. Second, if a bronchus or bronchiole is obstructed by the tumor, the distal lung may be consolidated with liquified cellular debris, pus or both, and these contents may have a conductivity value higher than air, which can potentially direct TTFields toward the active tumor. In this case, the shape of the consolidated lung or tumor may be important. This is because in the brain we have shown that the geometry of the modeled glioblastoma is a strong determinant of TTFields coverage at the GTV (9). Lastly, gold fiducials are often implanted near lung tumors for respiratory tracking during stereotactic body radiation therapy (22). This conductive metal may attract electric field lines toward the tumor and thereby augment TTFields coverage. However, all of the above predictions will require proper computer modeling studies and direct measurements in patients as validation.
There are a number of limitations in our models. First, suboptimal positioning of the arrays in the patient may limit TTFields coverage. However, the two anterior-posterior arrays and the right-left arrays are a common arrangement for lung cancer patients, and treatment efficacy will be determined by the randomized phase III LUNAR trial (NCT02973789) that is currently underway. Second, accurate estimation of mass density and electric conductivity is needed at the CTVp_ring, and this will require multiple tissue samplings of the tumor penumbra intraoperatively in a phase 0 study for measurements. Third, although we did not incorporate the use of a planning target volume (PTV) to account for potential technical variance, the source of TTFields is attached to the skin on the chest wall, and this moves in synchrony with the GTV and CTV when the patient breaths or moves. Therefore, the variance from relative motion between the source of TTFields and the GTV or CTV in the lungs are by comparison less than that compared to external beam radiotherapy. Fourth, although direct measurement of TTFields has not been done for a GTV or CTV in the lung, it was performed in the brain of a patient with a pineal region meningioma and the measured TTFields intensity was within 10% of the value derived from computer modeling (23). Lastly, patients with chronic obstructive pulmonary disease or emphysema have increased airspace in their lungs. Since air has low electric conductivity, we think that the increased volume of airspace may attenuate TTFields penetrance to the GTV or CTV. However, the extent of this attenuation will require a separate modeling study that is beyond the scope of our current investigation.
Conclusions
Our modeling study revealed a diverse range of TTFields distribution and demonstrated heterogeneity in electric field penetration to NSCLC patients’ target volumes with differences upwards of 200%. Conductive fluids in the pleura intensified TTFields at the GTV and CTV when they were in contact. In a sensitivity analysis, varying electric conductivity and mass density of the CTV altered coverage to both GTV and CTV. Therefore, personalized modeling is important to accurately estimate TTFields target coverage at the lung tumor volumes and surrounding normal tissue structures in the thorax.
References
1
Cancer Stat Facts: Lung and Bronchus Cancer. National Cancer Institute Surveillance, Epidemiology, and End Results Program. [Online] National Cancer Institute. https://seer.cancer.gov/statfacts/html/lungb.html.
2
Systemic therapy for stage IV non-small-cell lung cancer:American Society of Oncology Clinical Practice Guideline Update. Hanna, N, Johnson, D and Temin, S. 2017, J Clin Oncol., Vol. 35, pp. 3484-3515.
3
Maintenance Therapy With Tumor-Treating Fields Plus Temozolomide vs Temozolomide Alone for Glioblastoma: A Randomized Clinical Trial. Stupp, R, Taillibert, S and Kanner, A A. 23, 2015, JAMA, Vol. 314, pp. 2535-2543.
4
Tumor treating fields perturb the localization of septins and cause aberrant mitotic exit. Gera, N, et al., et al. 5, 2015, PloS One, Vol. 10.
5
A phase I/II trial of Tumor Treating Fields (TTFields) therapy in combination with pemetrexed for advanced non-small cell lung cancer. Pless, M, Droege, C and von Moos, R. 2013, Lung Cancer, Vol. 81, pp. 445-450.
6
Biological activity of tumor-treating fields in preclinical glioma models. Silginer, M, et al., et al. 4, 2017, Cell Death & Disease, Vol. 8.
7
Distruption of cancer cell repliation by alternating electric fields. Kirson, E D, Gurvich, Z and Schneiderman, R. 9, 2004, Cancer Research, Vol. 64, pp. 3288-3295.
8
Mitotic spindle disruption by alternating electric fields leads to improper chromosome segregation and mitotic catastrophe in cancer cells. Giladi, M, et al., et al. 2015, Sci Rep, Vol. 5.
9
Analysis of physical characteristics of Tumor Treating Fields for human glioblastoma. Lok, E, et al., et al. 6, 2017, Cancer Med, Vol. 6, pp. 1286-1300.
10
Tumor treating fields therapy device for glioblastoma: physics and clinical practice considerations. Lok, E, Swanson, K D and Wong, E T. 6, 20105, Expert review of medical devices, Vol. 12, pp. 717-726.
11
Conditions for numerically accurate TMS electric field simulation. Gomez, L, et al., et al. 1, 2019, Brain Stimulation, Vol. 13, pp. 157-166.
12
Measurements and models of electric fields in the in vivo human brain during transcranial electric stimulation. Huang, Y, et al., et al. 2017, eLife, Vol. 6.
13
Validation of a 3-d finite element human fingerpad model composed of anatomically accurate tissue layers. Wagner, M B, Gerling, G J and Scanlon, J. s.l. : IEEE, 2008, Symposium on Haptic Interfaces for Virtual Environment and Teleoperator Systems, pp. 101-105.
14
Electric Field Calculations in Brain Stimulation Based on Finite Elements: An Optimized Processing Pipeline for the Gneration and Usage of Accurate Individual Head Models. Windhoff, M, Opitz, A and Thielscher, A. 2013, Human Brain Mapping, Vol. 34, pp. 923-935.
15
Effect of micro-computed tomography voxel size and segmentation method on trabecular bone microstructure measures in mice. Christiansen, B. 2016, Bone Report, pp. 136-140.
16
International Commission on Radiological Units and Measurements (ICRU). ICRU Report 62, Prescribing, Recording and Reporting Photon Beam Therapy (Supplement to ICRU 50). Bethesda : s.n., 1999.
17
Internation Commission on Radiological Units and Measurements. ICRU Report 50—Prescribing, Recording and Reporting Photon Beam Therapy. Bethesda : s.n., 1994.
18
A phase I/II trial of Tumor Treating Fields (TTFields) therapy in combination with pemetrexed for advanced non-small cell lung cancer. Pless, M, et al., et al. 2013, Lung Cancer, Vol. 81, pp. 445-450.
19
Analysis of the quasi-static approximation for calculating potentials generated by neural stimulation. Bossetti, Chad, Birdno, Merrill and Grill, Warren. 1, 2007, Journal of Neural Engineering, Vol. 5.
20
Korshoej, A.R. Estimation of TTFields Intensity and Anisotropy with Singular Value Decomposition: A New and Comprehensive Method for Dosimetry of TTFields. [book auth.] S Makarov. Brain and Human Body Modeling. s.l. : Springer, 2019.
21
Tumour-treating fields (TTFields): Investigations on the mechanism of action by electromagnetic exposure of cells in telophase/cytokinesis. Berkelmann, Lukas, et al., et al. 7362, 2019, Scientific Reports, Vol. 9.
22
Assessment of lung tumor motion and setup uncertainties using implanted fiducials. Nelson, Christopher, et al., et al. 3, 2007, Int J Radiat Oncol Biol Phys., Vol. 67.
23
Alternating electric fields arrest cell proliferation in animal tumor models and human brain tumors. Kirson, Eilon, et al., et al. 104, 2007, PNAS, Vol. 12.
24
End-to-end workflow for finite element analysis of tumor treating fields in glioblastomas. Timmons, J J, et al., et al. 21, 2017, Phys Med Biol, Vol. 62, pp. 8264-8282.
25
Improving Tumor Treating Fields Treatment Efficacy in Patients With Glioblastoma Using Personalized Array Layouts. Wegner, C, et al., et al. 5, 2016, Int J Radiat Oncol Biol Phys, Vol. 94, pp. 1137-1143.
26
Personalizing Tumor Treating Fields (TTFields) Therapy With NovoTAL: Implications for Patterns of Local and Distal Recurrence in Glioblastoma (GB). Chaudhry, A, et al., et al. 2, 2016, Radiation Oncology, Vol. 96.
27
Alternating electric fields (TTFields) for treating glioblastomas: a modeling study on efficacy. Wenger, C, et al., et al. 2014, Neuro-Oncology, Vol. 16.
28
Enhancing Predicted Efficacy of Tumor Treating Fields Therapy of Glioblastoma Using Targeted Surgical Craniectomy: A Computer Modeling Study. Korshoej, A R, et al., et al. 10, 2016, PloS One, Vol. 11.
29
Impact of tumor position, conductivity distribution and tissue homogeneity on the distribution of tumor treating fields in a human brain: A computer modeling study. Korshoel, A R, et al., et al. 6, 2017, PloS One, Vol. 12.
30
Enhancing Predicted Efficacy of Tumor Treating Fields Therapy of Glioblastoma Using Targeted Surgical Craniectomy: A Computer Modeling Study. Korshoej, A R, et al., et al. 10, 2016, Plos One, Vol. 11.
31
Computed modeling of alternating electric fields therapy for recurrent glioblastoma. Lok, E, Hua, V and Wong, E T. 11, 2015, Cancer Medicine, Vol. 4.
32
Aquaporins: role in cerebral edema and brain water balance. Zador, Z, et al., et al. 2007, Prog in Brain Research, Vol. 161, pp. 185-194.
33
Blood–brain barrier breakdown in septic encephalopathy and brain tumours. Davies, D C. 6, 2002, J of Anatomy, Vol. 200, pp. 639-646.
34
Evolution of Brain Edema Concepts. Klatzo, I. 1994, Acta Neurochir, Vol. 60, pp. 3-6.
35
Development of realistic high-resolution whole-body voxel models of Japanese adult males and females of average height and weight, and application of models to radio-frequency electromagnetic-field dosimetry. Nagaoka, T, et al., et al. 2004, Phys. Med. Biol., Vol. 49, pp. 1-15.
36
A Numerical Implementation of a Modified Form of the Electric Field Integral Equation. Adams, R J and Champagne, N. 9, 2004, IEEE Transactions on Antennas and Propagation, Vol. 52, pp. 2262-2266.
Gabriel, C. Compilation of the Dielectric Properties of Body Tissues at RF and Microwave Frequencies. Armstrong Laboratory. Brooks Air Force Base : Occupational and Environmental Health Directorate, 1996.
39
Lok, E, Wong, E T and Sajo, E. Computer Simulation of Tumor Treating Fields. [book auth.] E Wong. Alternating Electric Fields Therapy in Oncology. s.l. : Springer, Cham, 2016, 2, pp. 15-27.
40
Gyri-precise head model of transcranial DC stimulation: Improved spatial focality using a ring electrode versus conventional rectangular pad. Datta, A, et al., et al. 4, 2009, Brain Stimul, Vol. 2, pp. 201-207.
41
Emerging treatment paradigms for brain metastasis in non-small-cell lung cancer: an overview of the current landscape and challenges ahead. Ulahannan, D, Khalifa, J and Faivre-Finn, C. 2017, Ann Oncol, Vol. 28, pp. 2923-2931.
42
Simulating pad-electrodes with high-definition arrays in transcranial electric stimulation. Kempe, R, Huang, Y and Parra, L. 2014, Journal of Neural Engineering, Vol. 11.
Funding Statement
This research is made possible by Musella Foundation and A Reason to Ride research fund.
Data Availability Statement
Research data may be shared upon request to the corresponding author(s)
Supplemental Figure 1: SAR is inversely proportional to mass density. Sensitivity analysis of SAR metric as a function of mass density of the CTVp_ring revealed decreasing SAR metrics for 95%, 50% and 5% volumes for CTV (also denoted as CTVp) (first row) and CTVp_ring (second row). However, there was no change in all 3 metrics for GTV (also denoted as GTVp) (third row).
Supplemental Table 1 Electric field, SAR, and current density PQM values for 95%, 50% and 5% volumes in SUBJECT 1.
Supplemental Table 2 Electric field, SAR, and current density PQM values for 95%, 50% and 5% volumes in SUBJECT 1_2xM.
Supplemental Table 3 Electric field, SAR, and current density PQM values for 95%, 50% and 5% volumes in SUBJECT 2.
Supplemental Table 4 Electric field, SAR, and current density PQM values for 95%, 50% and 5% volumes in SUBJECT 3.
Supplemental Table 5 Electric field, SAR, and current density PQM values for 95%, 50% and 5% volumes in SUBJECT 4.
Declaration of Competing Interest
ETW received research fundings from AstraZeneca, Five Prime Therapeutics, Immunocellular Therapeutics, Merck, Northwest Biotherapeutics, Novocure, Oblato, Orbus, Pfizer, and Vascular Biogenics. ETW also served as consultant and advisory board member for Gtree, Novocure, Orbus, Sumitomo Dainippon Pharma Oncology, and ZaiLab. ETW participated on data safety monitoring board for Turning Point Therapeutics, and OptimalTTF Phase II Trial (investigator initiated). ETW and EL have US Patent Application No.: 16/335,920 “System and Methods For Cancer Treatment Using Alternating Electric Fields”. EL received payment to perform clinical duties at US Oncology. OL and TM have no conflict of interest to report.
00032-5&id=gr1.jpg){kind=link}
00032-5&id=gr2.jpg){kind=link}
00032-5&id=gr3.jpg){kind=link}
00032-5&id=gr4.jpg){kind=link}
00032-5&id=gr5.jpg){kind=link}