Comparative toxicity of synchrotron and conventional radiation therapy based on total and partial body irradiation in a murine model

Synchrotron radiation can facilitate novel radiation therapy modalities such as microbeam radiation therapy (MRT) and high dose-rate synchrotron broad-beam radiation therapy (SBBR). Both of these modalities have unique physical properties that could be exploited for an improved therapeutic effect. While pre-clinical studies report promising normal tissue sparing phenomena, systematic toxicity data are still required. Our objective was to characterise the toxicity of SBBR and MRT and to calculate equivalent doses of conventional radiation therapy (CRT). A dose-escalation study was performed on C57BLJ/6 mice using total body and partial body irradiations. Dose-response curves and TD50 values were subsequently calculated using PROBIT analysis. For SBBR at dose-rates of 37 to 41 Gy/s, we found no evidence of a normal tissue sparing effect relative to CRT. Our findings also show that the MRT valley dose, rather than the peak dose, best correlates with CRT doses for acute toxicity. Importantly, longer-term weight tracking of irradiated animals revealed more pronounced growth impairment following MRT compared to both SBBR and CRT. Overall, this study provides the first in vivo dose-equivalence data between MRT, SBBR and CRT and presents systematic toxicity data for a range of organs that can be used as a reference point for future pre-clinical work.


Further explanation of dosimetry method, traceability and uncertainty
In all three delivery modes (SBBR, MRT and CRT), the absorbed dose to water was determined at a point on the beam axis at a depth of 5 mm in a theoretical phantom designed to mimic the scatter conditions of a mouse in the custom plastic (polymethyl-methacrylate; PMMA) holder used during the irradiations (Supplementary Fig.   2). In this phantom the mouse is modelled as a rectangular water slab (2 cm wide x 10 cm high x 1.5 cm thick) with a 0.5 cm air gap, and PMMA holder of 7 cm wide x 16 long and 2.4 cm thick. While this model does not take into account any details of the mouse itself, it does include the overall scatter conditions, and therefore is a more accurate choice than using the dose in a full-scatter water phantom, or the incident air kerma. Initially the phantom calculations were performed to estimate the uncertainty in using the full-scatter water dose as a surrogate for the mouse dose.
However, when the magnitude of the difference between full-scatter water and the mouse phantom became apparent, it became clear that this was a better surrogate for the actual mouse dose. Importantly, the mouse plastic phantom (with a lack of backscatter) has a pronounced effect on MRT valley dose compared to the fullscatter water phantom.

SBBR
Dosimetry for SBBR was performed using a PTW (Freiburg, Germany) model 31014 pinpoint ionization chamber in a virtual water phantom according to the protocols previously described by Lye et al. 1 and Livingstone et al 2 . The chamber was calibrated by the National Measurement Institute of Germany, Physikalisch-Technische Bundesanstalt (PTB), for absorbed dose to water in conventional X-ray beams, and the calibration coefficient was interpolated in HVL to the synchrotron beam spectrum to obtain N D,w = 2.70 x 10 9 Gy/C with an uncertainty of 2.2% (k=1).
We estimate an additional 0.5% uncertainty arises from the difference in the field size, depth and spectrum of the synchrotron beam compared to the PTB calibration beams. Charge from the Pinpoint chamber was integrated as it and the phantom were scanned through the 1 mm high x 30 mm wide synchrotron beam to deliver a uniform rectangular field. After correction for recombination, the dose at the measurement depth (20 mm) was multiplied by a measured depth-dose relationship (PDD) to obtain the dose at 5 mm depth. Dosimetry was performed for the four field sizes (30 mm x 20 mm, 30 mm x 30 mm, 30 mm x 60 mm, 30 mm x 100 mm) which produce slightly different doses due to increased backscatter in the larger fields. In this dynamic delivery mode, the incident air kerma and beam size is constant. The dose is controlled by the velocity of the sample as it is scanned through the beam.
Uncertainties in all of these factors give rise to a combined uncertainty in the delivered dose at 5 mm depth in a water phantom of 2.4% (k=1). An additional uncertainty due to air around the mouse and loss of backscatter in the mouse phantom was calculated from Monte Carlo simulations to be in the range of 0.94-0.90 leading to combined uncertainties of 4.8 % for SBBR.

MRT
Dosimetry for MRT was determined relative to the measurements made for SBBR.
Previous work by Poole et al. 3 used GEANT4 Monte Carlo models to establish the output factor (OF) for the peak dose and the peak to valley dose ratio (PVDR). The OF is the ratio of the dose in the peaks to the dose in the broad beam when the MRT collimator is removed (i.e. the SBBR case). The PVDR is the ratio of the dose in the peaks to the dose in the valleys. In the current study, the PVDR varied from 31.8 (TBI) to 41.3 (Thoracic PBI). The uncertainty in the OF is estimated to be 1.7% and in the PVDR is 6.9% (k=1). These uncertainties are added in quadrature with the SBBR uncertainty to obtain a combined uncertainty of 5.1% for the peak doses and 8.6% for the valley doses in MRT.  Combined D_(w,z=0.5) uncertainty (k=1) 6.07 [1] Inherent calibration uncertainty, from ARPANSA calibratkion for air kerma at this beam quality (NXE250)

Uncertainty calculation for mouse irradiation dosimetry -CRT
Approximate 1/r^2 uncertainty in positioning accuracy of 0.5-1mm in 300mm. Also note that SSD was d -5 mm [3] Type A is ESDM in typical current measurement, Type B is calculated in worksheet "IC current type B", U8 spreadsheet [4] Beam non-uniformity for the largest exposed site (measured with film) is a 26% correction with significant asymmetry at one end (down to 40% for TBI).
Estimate uncertainty is 100% of largest correction, which is 6% (DB -I've kept this at 6% but I think its defintion should include more geometry of the mouse) [5] See U26 -different lab and detector but the estimate is still valid and dominated by possible difference between temperatures at chamber and thermistor [6] Based on therapy QC history Guess, chamber NK changes 0.5% with HVL change of ~30%, unlikely HVL will change more than that with distance, treated as rectangular distribution [8] from Monte Carlo calculation Monte Carlo calculation based on jig irradiated at centre (actually moved depending on exposed site) and no lead shielding included in model which affects backscatter Rough calculation based on ratios of irradiated field sizes and TG-61 backscatter factors (which assumes full scatter conditions) gives up to 4% correction for smallest exposed sites (head and thorax, see 'backscatter with lead' tab), estimate uncertainty is 100% of this correction DB: Even when including the full MC calculation of the dose, there is about a 4% uncertainty due to the air around the mouse. I think this cannot be got rid of.
[10] Estimated difference between FWHM of 9.5 cm and 10.3 cm (approximate measured value and value used in model) Ionization current 0.05 0.05 [3] temperature/pressure correction (kTP) 0.50 [4] Beam stability (no monitor used) 0.13 [5] Interpolation to synchrotron spectrum 0. Inherent calibration uncertainty, from PTB Certificate [2] ApproximatePDD uncertainty in positioning accuracy of 0.5-1mm in water PDD [3] Type A is ESDM in typical current measurement, Type B is calculated in worksheet "IC current type B", U8 spreadsheet [4] TP variation mostly due to possible temperatre and pressure lags. Larger in the synchrotron than ARPANSA. [5] Synchrotron output variation in top up mode: max 0.5 mA in 200 mA -so 0.25% max (divide by 2) [6] Estimate of interpolating to Mo/Mo spectrum -kQ from PTB is a slowly varying function with energy [7] Possible variation due to different fields sizes etc. But this is an ion chamber -insensitive to such things [8] Ion chamber measurements performed at 5mm depth in soid water. Positional uncertainty accounted for in [2].
[9] A correction is applied to account for the difference between fullscatter conditions and the simplified mouse model.
The uncertainty is 4% due to variable field size due to Pb strips during exp.
[10] An additional 1% since the backscatter correction factor is applicable to the Cu,Cu spectrum only.
[11] The beam is almost parallel when the BDA defines the beam's static height. Small beam divergence accounted for in simulation. Field size variation conservatively estimated to be +-4mm Corresponds to a 0.5% change in dose