A study of the dosimetric impact of daily setup variations measured with cone‐beam CT on three‐dimensional conformal radiotherapy for early‐stage breast cancer delivered in the prone position

Abstract Purpose To evaluate the dosimetric impact of daily positioning variations measured with cone‐beam computed tomography (CBCT) on whole‐breast radiotherapy patients treated in the prone position. Methods Daily CBCT was prospectively acquired for 30 consecutive patients positioned prone. Treatment for early‐stage (≤II) breast cancer was prescribed with standard dose (50 Gy/25 fractions) or hypofractionation (42.56 Gy/16 fractions) for 13 and 17 patients, respectively. Systematic and random errors were calculated from the translational CBCT shifts and used to determine population‐based setup margins. Mean translations (±one standard deviation) for each patient were used to simulate the dosimetric impact on targets (PTV_eval and lumpectomy cavity), heart, and lung. Paired Student’s t tests at α = 0.01 were used to compare dose metrics after correction for multiple testing (P < 0.002). Significant correlation coefficients were used to identify associations (P < 0.01). Results Of 597 total fractions, 20 ± 13% required patient rotation. Mean translations were 0.29 ± 0.27 cm, 0.41 ± 0.34 cm, and 0.48 ± 0.33 cm in the anterior–posterior, superior–inferior, and lateral directions leading to calculated setup margins of 0.63, 0.88, and 1.10 cm, respectively. Average three‐dimensional (3D) shifts correlated with the maximum distance of breast tissue from the sternum (r = 0.62) but not with body‐mass index. Simulated shifts showed significant, but minor, changes in dose metrics for PTV_eval, lung, and heart. For left‐sided treatments (n = 18), mean heart dose increased from 109 ± 75 cGy to 148 ± 115 cGy. Shifts from the original plan caused PTV_eval hotspots (V105%) to increase by 5.2% ± 3.8%, which correlated with the total MU of wedged fields (r = 0.59). No significant change in V95% to the cavity was found. Conclusions Large translational variations that occur when positioning prone breast patients had small but significant dosimetric effects on 3DCRT plans. Daily CBCT may still be necessary to correct for rotational variations that occur in 20% of treatments. To maintain planned dose metrics, unintended beam shifts toward the heart and the contribution of wedged fields should be minimized.

treatments. To maintain planned dose metrics, unintended beam shifts toward the heart and the contribution of wedged fields should be minimized. Whole-breast radiotherapy (WBRT) after BCS is historically delivered to the patient in the supine position. In patients with large breast separation, however, dose inhomogeneity can result in worse fibrosis. 3 Additionally, pendulous breast anatomy could exhibit larger daily setup variability due to arbitrary shifts of breast position by gravity. 4 Attempts to mitigate this problem with angled breast boards can result in increased breast to skin contact over the inframammary folds 5,6 and higher heart and lung tissue doses. 4,6 To better address these problems, prone positioning has been evaluated for treatments during which large breast size or pendulous breast tissue is of concern. This approach has improved dose homogeneity leading to excellent cosmetic outcomes, [7][8][9] while also providing dosimetric advantages by reducing intrafraction setup variation from respiratory motion, 4,10 minimizing lung volumes receiving 10 and 20 Gy, and increasing displacement of breast tissue from the chest wall, heart, and contralateral breast. 11,12 Prone positioning is particularly powerful for reducing the heart dose to as low as possible, 5,6 which is especially important in light of the study by Darby et al that demonstrated a no-threshold relationship of dose to chronic heart toxicities in breast cancer patients. 13 Despite these demonstrated dosimetric advantages, interfraction variability of breast tissue and the necessity of daily setup image guidance remain uncertain. Prone WBRT is additionally challenging because the anatomical site is not directly visualized and the patient lies on deformable tissue (i.e., the contralateral breast). 14 Studies evaluating prone breast setup for WBRT using cone-beam CT (CBCT) have demonstrated daily setup variations necessitating a clinical target volume (CTV) to planning target volume (PTV) margin of 1.0-2.2 cm. 10,14,15 While margins are not typically utilized for WBRT plans, they provide a consistent way to compare both systematic and random setup uncertainties among various institutions. Previous studies focused on comparison of prone to supine positioning 10,11,14,16 or on the impact of setup error on partial breast irradiation. 4,17 More recent studies have estimated the dosimetric effects of positioning variations as quantified from planar images but not from CBCT. 18,19 We hypothesized that daily CBCT is necessary in these patients to achieve accurate positioning, particularly to correct for rotations and deformations that we had regularly observed, in order to ensure accurate dose delivery during treatment. To our knowledge, this study is the first to evaluate the dosimetric impact of daily positioning variations measured with CBCT on WBRT. Here, we report on the impact that both setup margins and patient characteristics have on the dosimetry of three-dimensional conformal radiotherapy (3DCRT) plans and on the role of CBCT for prone WBRT. or lumpectomy with sentinel lymph node evaluation for those with invasive breast cancer. Prone WBRT was limited to patients with node-negative disease requiring breast-only treatment. Hypofractionation was utilized for patients who did not require chemotherapy, lumpectomy cavity boost, or who were >50 yr as it has been shown to provide acceptable acute toxicity rates at our institution. 8,9

2.A | Radiotherapy planning and delivery
Patients were simulated using a CDR Systems prone breast board with indexable handles (CDR Systems Inc., Calgary, AB, Canada) with the addition of upper and lower alpha cradles. While patients lie flat on the prone device, the board underneath the contralateral breast was either flat (n = 5) or angled (n = 25). Simulations were performed on a Brilliance BigBore CT scanner (Philips Healthcare, Andover, MA) using 3-mm slice thickness. Three-dimensional conformal radiotherapy plans were created using high-energy photon beams (6)(7)(8)(9)(10)(11)(12)(13)(14)(15)(16)(17)(18), physical or computerized wedges, and/or large segments for optimization of dose homogeneity and target coverage as described previously. 8 A structure to evaluate target dose (i.e., PTV_eval) was generated to encompass all breast tissue within the tangential beams but excluding the tissue in the buildup region (i.e., 6 mm from the skin). At our institution, this structure is generated automatically using the 70% isodose line calculated from the unmodified tangential beams and edited by physicians if necessary. The lumpectomy cavity was contoured to include the seroma (n = 29) plus any visible surgical clips when present (n = 24).

2.B | Margin calculations
Margins to account for setup uncertainty in each dimension were calculated using the van Herk recipe: 2.5 × Σ + 0.7 × σ, where Σ and σ are standard deviations for systematic and random errors, respectively. While the calculated margin does not account for rotations, it ensures coverage by at least 95% of the prescribed dose to the CTV for 90% of patients in whom translations are not corrected daily via image guidance. 20 First, translational shifts from CBCT in the anterior/posterior (A/P), superior/inferior (S/I), and right/left (R/L) dimensions were tabulated and the mean shift (M) and standard deviation (STD) were calculated for each patient. Note that the mean shift M includes the direction of the shift depending upon the sign associated with the shift magnitude. Then, total setup error was calculated from the mean of M across patients, Σ was calculated from the standard deviation of M across patients, and σ was calculated from the root mean square of STD across patients. This methodology has been applied to breast patients treated in different fractionation regimens by others. 14 2.B.1 | Dose calculation for plans with simulated shifts To simulate the dosimetric effects of uncorrected systematic and random setup errors, the original treatment plan was recalculated using the mean setup errors (i.e., M ± STD) for each patient in every dimension and each direction. Doses were tabulated and maximum deviations from the intended treatment plan were compared to institutional constraints for the following dose metrics: V95% (volume receiving 95% dose) for PTV_eval and lumpectomy cavity targets, V105% (volume receiving 105% dose) for PTV_eval, mean heart dose, and V20 Gy (volume receiving 20 Gy) for ipsilateral lung.

2.B.2 | Statistical analysis
Linear regression using R statistical package (http://www.R-project. org) was used to identify significant correlations (P < 0.01) between: Anterior breast distance from the sternum as a measure of breast pendulousness for two patients, with PTV_eval (red contour) and lumpectomy cavity (green contour) overlaid. Breast tissue contacted the Styrofoam base for the patient on the right. random errors were compared using a two-tailed t test and systematic errors were compared using an F test (P < 0.01) as reported by Feng et al. 21 Paired Student's t tests at α = 0.01 were used to compare dose metrics across plans after Bonferroni correction for multiple testing resulting in a P < 0.002 (i.e., P < 0.01/5).

3.A | Patient characteristics
Patient characteristics are shown in Table 1 Table 2 shows the means of absolute shifts and total setup errors, as well as standard deviations of systematic and random errors, calculated from the translations indicated by the terminal CBCT (i.e., after all postural corrections had been implemented) in 597 treatment fractions. The total setup error combines the positive and negative shifts, resulting in smaller total error values than absolute shifts. After excluding the 19.6% of fractions with a priori information (i.e., from the initial CBCT which indicated that postural correction was necessary), the random and systematic errors calculated from the remaining 479 fractions did not differ significantly.

3.B | Daily inter-fractional setup error
The random and systematic errors were also not significantly different for the subset of patients treated with 16 or 25 fractions compared to those from all 597 fractions.

3.C | Margins for prone whole-breast radiotherapy
The systematic and random errors from CBCT measurements were used to estimate a setup margin using the van Herk's population-

3.D | Effect of patient characteristics on interfractional setup error
We evaluated the associations between patient age, BMI, the pendulousness of breast tissue via ABDS, and the average 3D shifts. Only  | 149 significant correlations are reported here (P < 0.01). Data from three patients were excluded from the correlation coefficient analysis as their breast tissue was distorted due to contact with the Styrofoam base as shown in Fig. 1(b). These three patients had ABDS >16 cm, which potentially would have been larger in the absence of the treatment Significant correlations including their 95% confidence intervals are summarized in Table 3. Correlations were additionally evaluated for the following subsets of data: (a) patients treated with differing fractionation regimens, and (b) CBCT shift data without a priori information (n = 479). We found that patients with a larger ABDS required larger average 3D shifts than those with less pendulous breasts, with a correlation coefficient of 0.62 [ Fig. 2(a)]. Although average 3D shifts correlated with ABDS, they did not correlate with BMI despite the fact that ABDS and BMI were moderately corre-   Fig. 3, which reduced the distance between the beam edge and the heart, were the cause of the increased heart dose for the majority of plans (16 of 18) and for ≥1% increases in lung V20 in four plans. The 105% hotspots (i.e., V105%) to the PTV_eval increased by 5.2% ± 3.8%, and correlated (r = 0.59) with the total monitor units (MU) of wedged fields as shown in Fig. 2(b). This correlation increased to 0.77 for the patients treated in 16 fractions.
Despite statistically significant changes in dose, simulated plans exceeded institutional dose constraints in only one plan (V105% for PTV_eval) in which all fields were wedged.

| DISCUSSION
Prone positioning has been successfully used for WBRT following breast-conserving surgery, but formal guidelines for use of IGRT for setup verification have yet to be outlined. 12  design. 14 Other studies using different prone devices found that the T A B L E 3 Correlation coefficients and their 95% confidence intervals for various patient characteristics including body-mass index (BMI), anterior breast distance from sternum (ABDS), and average three-dimensional (3D) shifts calculated from daily conebeam computed tomography (CBCT) with (n = 597) and without a priori information (n = 479 largest setup margin is required in the A/P, 16 R/L, 14,15 or the S/I dimension. 10 In our study, in which patients lie flat on a prone board and daily CBCT was used, we found that the A/P margin is smaller than in other dimensions.  18,19 Although minor dosimetric degradation of plans was found in our simulations, the goal at our institution is to minimize cardiac doses as much as possible in order to reduce potential cardiac toxicities. 1,13 By using CBCT daily, we can accomplish this goal while also ensuring that the breast shape is reproduced for treatment in order to limit dose inhomogeneity which has been associated with acute radiation dermatitis. 26 Although daily CBCT had a small but significant effect on the planned dose, there are other reasons for its use, such as for quality control 18 or for an in-house investigation of setup variability 27 as in the current study. Similarly to the rationale presented by Mulliez et al., 14 we use CBCT to correct the patient's roll and ensure that the contralateral breast tissue is outside of the treatment field. In our study, postural adjustments requiring rotations were necessary in approximately 20% of treated fractions (19.6 ± 13%) and cannot be accounted for using PTV margins. 20 Moreover, the use of PTV margins is not typical for 3DCRT in breast radiotherapy due to the increased dose that would be delivered to OARs although such margin calculations would be applicable for prone partial breast irradiation. 17 Instead, daily CBCT allowed us to correct the patient's posture and provided our therapy team with valuable information about patient positioning.
(a) Three-dimensional average shift correlates significantly (r = 0.62) with the anterior breast distance from sternum after excluding data points for patients whose breast was in contact with a Styrofoam base. (b) Increases in V105% to the PTV_eval in simulated plans correlate significantly (r = 0.59) with the total monitor units from wedged fields. In both plots, the linear fit is shown in a black line while the 95% confidence interval is shown in light gray shading.
T A B L E 4 Dose metrics for intended and simulated plans using patient-specific mean shifts with one standard deviation (M ± STD).  deformations that could occur in breast shape on a daily basis. This effect is expected to be small in our study compared to others who have made this assumption without the added benefit of 3D imaging 18,19 since CBCT led our therapy team to correct the patient's posture and subsequently any gross deformations in breast shape prior to treatment. Another limitation is that the CBCT scans acquired for positioning could not be used to perform dose calculations because of significant artifacts that stem from truncation of patient anatomy, much of which extend outside of the CBCT field-of-view. These artifacts are similar to those shown in the study by Jozsef et al. 17 Also, the use of a mean shift to simulate the day-to-day variations could have overestimated the true dosimetric changes. To temper this potential overestimate, one STD was added to the mean shift rather than two STD. Finally, the simulated dosimetric changes in V95% and V105% to PTV_eval may not necessarily reflect changes in dose to breast tissue. The PTV_eval structure represents breast tissue within the tangential fields and has been used by RTOG protocols (e.g., RTOG1005) to limit the target from extending outside of the treatment fields or into the thorax/lungs. As a result, this structure served as a surrogate for the target in our clinic and did provide insight into the dosimetric impact of shifts detected by our study. As shown in Fig. 1, PTV_eval predominantly encompasses breast tissue.

| CONCLUSION
Our work demonstrates that in 3DCRT WBRT plans without intensity modulation, setup variations caused small but significant dosimetric changes although further gains could be achieved by minimizing anterior shifts of the patient toward the beam to limit increases to the heart dose and by reducing the total monitor units of wedged fields to limit increases in target hotspots. While these results appear to downplay the importance of daily CBCT for reproducing the planned dose to targets, forgoing CBCT may have occasionally delivered higher than planned dose to the heart and would have risked treating patients with uncorrected postural rotations/deformations in 20 ± 13% of fractions. Based on the results of our study, daily CBCT is recommended for prone positioning for 3DCRT plans.

CONF LICT OF I NTEREST
No conflict exists.