This study was approved by the Hungarian Ethical Committee, ETT TUKEB (23,546–3/201) (23,546–3/2017/EKU) and was conducted following the principles of the Helsinki Declaration. The five subjects were informed about the aim of the study and gave their written informed consent.

Blood sample

Venous blood samples were collected from five healthy non-smoking volunteers (three females, two males (age 37.6 ± 12.3 years) in Li-heparinized vacutainers. The blood test was preceded by a routine occupational medical examination. They filled a questionnaire specifying, date of birth, smoking status, dietary habits, alcohol consumption, exposure to diagnostic X-ray, use of therapeutic drugs. All were healthy and non-smokers. The maximum blood volume obtained was 20 ml per volunteer. Spontaneous chromosome aberration baseline values were determined before starting the experiments. We chose healthy subjects for our study because we wanted to measure the absolute effect of radiation on the human chromosome as accurately as possible. However, in a healthy population the age, smoking, long-term medication, drugs, alcohol, toxic chemicals, some viruses and bacteria could cause baseline shift in the number of total aberration.

If we had used blood from cancer patients, this may have affected the aberrations we wanted to measure.

Phantoms for measurements

In our study, we used measurement setups similar to the clinical routine of oART. We used two different phantoms for the measurements (Fig. 1). The first phantom was an in-house made small box phantom filled with clean water, cylindrical in shape (small cylindrical phantom), with a volume of approximately 380 cm3, and 80 mm in diameter and 80 mm in height. The phantom contained only one hole for a 2 ml cryotube, 11 mm in diameter. The second phantom was a Pure Imaging CTDI PMMA phantom (Pure Imaging Phantoms, Spring Court, Farnham Royal,UK) [9] with a body cylinder, approximately 11,965 cm3 in volume, 320 mm in diameter and 145 mm in length. This phantom contained nine rods, which could be exchanged for perfectly fitting cryotubes. At our institute, oART treatments started in the pelvic region, so the choice of this body phantom was appropriate.

Fig. 1

Small cylindrical water-filled phantom with cryotube (A) and PMMA CTDI phantom with body cylinder and nine cryotubes (B)

CT simulation

To reproduce a realistic treatment situation, we prepared CT-based external treatment plans for both phantoms. We used radiotherapy CT protocols depending on the width of the phantom. The used protocols were the pelvis for CTDI phantom with 120 kVp voltage and 210 mAs current and lung for water-filled with 120 kVp voltage and 35 mAs current. Water-filled cryotubes (2 ml) were inserted into the phantoms before CT imaging. CT images were imported into the Ethos Treatment Planning System 2.1, (TPS) (Varian, Palo Alto, CA, USA).

Irradiation planning

The cryotube was considered as the gross tumor volume (GTV), but we expanded it by a uniform margin of 1 cm to create the planning target volume (PTV). This volume was used to prescribe the dose. With this PTV, we were able to account for possible phantom shifts and also for the scattered dose from phantom materials. The small cylindrical phantom had only one readout point in the center of the PTV. The CTDI phantom contained nine holes, so we could use all of them for readout points. The PTV volume was placed in the center of the phantom and the volume of the other eight cryotubes was outlined and then used to read the dose.

There are several options to create a plan for this phantom. It is possible to add only one direct field to the phantom and normalize the dose to the cryotube [10]. In our investigation, “real” treatment plans were created, which means that intensity-modulated radiotherapy plans (IMRT) using the sliding window technique (SW) were used. This was an important criterion in our investigation because IMRT SW plans were also generated for oART. We used 12 equidistant fields with 10-degree collimator rotation. The isocenter was placed in the middle of the central cryotube. We used 6 MV FFF beam energy with a maximum dose of 800 MU/min (therapeutic ray) (Fig. 2).

Fig. 2

Small cylindrical (A) with one, and CTDI phantom (B) with the nine read-out volumes (small circles), and dose distributions in three views

We kept the 100% average dose requirement for GTV volume, and 100% of the PTV was to be covered by 95% of the prescribed dose, ensuring accurate dose coverage. The monitor units delivered varied depending on the dose prescribed per tube, 268 MU at 0.5 Gy, 414 MU at 1 Gy, 761 MU at 2 Gy, 1543 MU at 4 Gy, 2575 MU at 6 Gy, and 3416 MU at 8 Gy, respectively. The dose volume histogram (DVH) integrated into the planning system was used for post-processing of the dose data. With this tool, the data could be analyzed visually and numerically in the entire study volume.

Irradiation of blood samples

Reproducing a real-life situation, we set up the phantom using the same procedure as we set up patients in clinical practice. In the Ethos system, the first step is to position the phantom using external lasers. The phantom on the couch is then automatically shifted into the isocenter by the LINAC using precalculated translation values.

We placed a cryotube in the small cylindrical phantom with 2 ml of blood in the central hole. In the CTDI phantom, we placed one cryotube with the blood sample in the central position of the phantom, the other eight holes were filled with water-filled cryotubes as in planning CT imaging.

The same type of imaging protocol was used for the both phantoms, namely pelvis large protocol, for exact comparison. The high voltage was 140 kV, and the exposure was 1068 mAs, the calculated CTDIvol based on TPS (volumetric Computer Tomography Dose Index) was 0.038 Gy, the DLP (Dose Length Product) was 0.852 Gy*cm, and the range was 19.4 cm. However, the actual physical dose may differ from that presented by TPS because of the uncertainty of the input data for dose modelling, dose calculation, commissioning heterogeneity changes, CT calibration, which can cause errors of up to 3–5 percent so it is important to know the real biological effect of CBCT [11]. The use of this imaging protocol was essential because online adaptive therapy requires high image quality and good spatial resolution, independently of the irradiated region.

The use of an imaging guidance is mandatory before starting a treatment with Ethos. Therefore, when we used the therapeutic beams for the calibration curves, we had to scan the phantom with water-filled cryotube instead of a blood sample. After imaging and registration, the tube was replaced with the blood-filled cryotube.

Separate irradiated blood sample measurement

We investigated several dose delivery setups to measure clinically relevant chromosome aberrations from the kV-CBCT and the therapeutic beams. We generated calibration curves with predefined dose steps (0.5, 1, 2, 4, 6, 8 Gy) with both phantoms. Henceforth, the prescribed dose of 2 Gy was clinically relevant for us. We chose this therapeutic fraction dose because it is the most commonly used in both conventional and adaptive radiotherapy. Afterwards, we irradiated blood samples with 1 to 5 consecutive CBCT. Measurements were performed by placing blood samples in each hole of the CTDI phantom one by one and irradiating with 3 or 5 CBCT to evaluate the peripheral aberration.

Co-irradiated blood sample measurement

We then mixed the CBCT frequency and therapeutic beam frequencies to evaluate real situations. The dose from the therapeutic beams was the same in every situation (2 Gy), the number of the CBCTs were 1, 2, 3, 4 and 5. The most important configuration investigated was the 3 CBCTs with 2 Gy therapeutic beam because this was our online adaptation schema. We also paid attention to the sequence of delivery. We performed 2 CBCTs first, followed by therapeutic beams, and finished with one CBCT. We performed measurement in the center of the CTDI phantom with only 3 CBCTs and only 2 Gy therapeutic dose to investigate whether the effects of the two types of radiation could be biologically combined and what additive biological effects the combined radiation had.

Lymphocyte cultures

Blood samples were irradiated at dose rates ranging from 1 to 800 MU/min and 6 MV FFF energy with a dose of 0.5–8 Gy. Culture and chromosome preparation were performed using standard cytogenetic techniques after exposure: 0.8 ml of blood was added to 9 ml of RPMI-1640 culture medium containing 15% bovine serum albumin and penicillin/streptomycin (0.5 ml/L). Cell proliferation was induced with phytohaemagglutinin M (0.2%). Incubation time was 52 h at 37 °C. Lymphocyte proliferation was inhibited with 0.1 μg/ml Colcemid (Gibco) during the last 2 h of culture. Cell cultures were then centrifuged, treated with a hypotonic solution of 0.075 M KCl at 37°Cfor 15 min and fixed with a 3:1 solution of cold methanol-acetic acid. After several washes in fresh fixative, the cells were resuspended in a small volume (0.5 ml) of fresh fixative, then this suspension was dropped on glass slides, dried and stained with 3% Giemsa.

Study of chromosomal aberrations

Between 100 and 200 metaphases were analyzed per experimental points in manual mode with a light microscope at × 1500 magnification. Chromosome analysis was performed at the first cell division. Only clear oval metaphase cells were counted. All aberration types were recorded. On the basis of structural differences, chromatid-type breaks (chromatid break, exchange) and chromosomal-type breaks and rearrangements (fragments, dicentrics, rings, translocations) were identified. One dicentric or ring and one acentric fragment observed in the examined cell were counted as one chromosome aberration (CA). One tricentric chromosome was counted as two dicentric equivalents. Excess fragments were not distinguished as terminal or interstitial deletions according to the position of the chromatin loss. Acentric fragments without dicentric or ring aberrations were counted as excess fragments. The evaluation was performed according to ICPEMC requirements [12]. Slides were coded and metaphases were analyzed by three well-trained scorers. For scoring, 100–300 complete metaphase cells were evaluated per dose point per donor in all conditions. A total of 158 samples were irradiated and a minimum of 31,600 metaphases were evaluated.

Statistical analysis

The yields of aberrant cells and aberrations (Y) were expressed per 100 cells scored. Standard errors (SE) for the mean aberration yield were calculated from the dispersion (σ2) of aberration among cell distributions. At each experimental point, aberration among cell distributions was checked for consistency with Poisson model using the variance-to-mean ratio (σ2 / Y) and Papworth’s u-test [13]. The dose response was fitted to a linear- quadratic model using the iteratively reweighted least squares method. Student's t-test was used for statistical analyses. Significant differences were determined at 95% confidence interval, with a P value of < 0.05 considered as the limit of significance. Dose response was fitted using CABAS-2 (Chromosomal Aberration Calculation) software [14] and GraphPad Prism 5 was used for calculations and data presentation [15]. We did not perform low-dose calibration curves because we wanted to evaluate the summarized aberrations from low-energy and high-energy radiation relatively. For this consideration, we fitted the low-dose energy values in the high-energy calibration curves. In case when a low energy calibration curve was used, it would not had been possible to evaluate the effect of the therapeutic dose, since it is known that it has a larger impact. We know that the most precise procedure would be to evaluate the low energy CBCT dose separately using a low energy calibration curve and the therapeutic dose separately using a high energy calibration curve, but in this case we would not be able to measure the combined effect of the two types of radiation. In addition, in a prospective human study, we would not be able to take blood samples after CBCT radiation and before the start of treatment, so the most feasible approach would be to use a high-energy calibration curve, as we done.