Impacts of lung and tumor volumes on lung dosimetry for nonsmall cell lung cancer

Abstract The purpose of this study was to determine the impacts of lung and tumor volumes on normal lung dosimetry in three‐dimensional conformal radiotherapy (3DCRT), step‐and‐shoot intensity‐modulated radiotherapy (ssIMRT), and single full‐arc volumetric‐modulated arc therapy (VMAT) in treatment of nonsmall cell lung cancers (NSCLC). All plans were designed to deliver a total dose of 66 Gy in 33 fractions to PTV for the 32 NSCLC patients with various total (bilateral) lung volumes, planning target volumes (PTVs), and PTV locations. The ratio of the lung volume (total lung volume excluding the PTV volume) to the PTV volume (LTR) was evaluated to represent the impacts in three steps. (a) The least squares method was used to fit mean lung doses (MLDs) to PTVs or LTRs with power‐law function in the population cohort (N = 32). (b) The population cohort was divided into three groups by LTRs based on first step and then by PTVs, respectively. The MLDs were compared among the three techniques in each LTR group (LG) and each PTV group (PG). (c) The power‐law correlation was tested by using the adaptive radiation therapy (ART) planning data of individual patients in the individual cohort (N = 4). Different curves of power‐law function with high R2 values were observed between averaged LTRs and averaged MLDs for 3DCRT, ssIMRT, and VMAT, respectively. In the individual cohort, high R2 values of fitting curves were also observed in individual patients in ART, although the trend was highly patient‐specific. There was a more obvious correlation between LTR and MLD than that between PTV and MLD.

groups by LTRs based on first step and then by PTVs, respectively. The MLDs were compared among the three techniques in each LTR group (LG) and each PTV group (PG). (c) The power-law correlation was tested by using the adaptive radiation therapy (ART) planning data of individual patients in the individual cohort (N = 4). Different curves of power-law function with high R 2 values were observed between averaged LTRs and averaged MLDs for 3DCRT, ssIMRT, and VMAT, respectively. In the individual cohort, high R 2 values of fitting curves were also observed in individual patients in ART, although the trend was highly patient-specific. There was a more obvious correlation between LTR and MLD than that between PTV and MLD.

| INTRODUCTION
In radiotherapy (RT) of NSCLC, patients are at risk of radiation pneumonitis (RP) which is sometimes fatal. 1 Patient characteristics, including tumor volume, shape, location, and lung volume often have a large range of variety, which increases the complexity and the level of challenging to decrease the lung dose. 2,3 In recent years, there has been a continuous increase in use of VMAT due to its rotational characteristic for much shorter treatment times and higher conformal dose distributions compared to ssIMRT often only collected all volume of tumors into one group (normally with a large tumor volume range). 9 Therefore, the detailed information could not be obtained on how the lung or tumor volume affected lung dosimetry in different techniques, which played an important role in pulmonary toxicity risk prediction.
In current RT, tumor volume is one of the commonly used patient parameters for decision of proper treatment modality in both conventional and unconventional fraction schemes. High tumor dose is normally inaccessible because of possible lung injury, 10   The PTV and the organs at risk (OAR), including ipsilateral lung, contralateral lung, spinal cord, heart, and esophagus, were outlined using lung window width and level settings (1600 HU,À600 HU) and mediastinal settings (400 HU, 20 HU) respectively followed by manually edits. The lung volume was defined as the volume of the total lung volume excluding the PTV.
In the case of inverse planning methods, a second volume was created and defined as the considered organ minus the PTV as an assistant area to avoid hot spots around PTV and to improve CI. Once the treatment planning was completed, the plan was normalized to cover 95% of the PTV by the prescription dose. The dose volume constraints were set as follows: V20 < 30%, V30 < 20%, and MLD < 16 Gy. The maximum dose point for the spinal cord was 45 Gy. In addition, the plan optimization was also performed to keep the esophagus dose of V50 < 25% and mean esophageal dose <25 Gy, and the heart dose of V40 to 30%.
3DCRT planning was performed with the superposition dose calculation algorithm using 3-5 coplanar beams. Beam angles were configured to avoid unnecessary radiation to the contralateral lung and this depended on the individualized anatomic structure, PTV, and PTV location. And the collimator or wedge was used to optimize dose distribution if necessary.
The ssIMRT plans consisted of five coplanar beams, and beams were configured to cover the PTV with nonfixed angles. The VMAT plan consisted of one single full-arc corresponded to a single 358°r otation, which started at the gantry angle of 179°and then counter-clockwise rotated to stop at the gantry angle of 181°.
The dose volume constraints and relative priorities for both ssIMRT and VMAT were the same at the start of the optimization.
The specific plan mode (ssIMRT or VMAT) was selected before optimization. The optimization was performed in two steps. The first step was performed by pencil beam dose calculation algorithm to obtain the optimal modulated fluence. During this process, objective parameters were adjusted to achieve optimal results until there was no gap between the goals which were previously adjusted and the results optimized. In the second step, the Monte Carlo dose calculation algorithm was used to optimize the segments aiming at small areas of targets. For ssIMRT, the minimum segment area was set to 2 cm 2 and the minimum machine output per segment remained LEI ET AL. | 23 constant 4 MU. For VMAT, the maximum control points were set to 120 and the minimum segment width was 0.5 cm.

2.D | Evaluation of LTRs
The evaluation was performed in three steps. Firstly, the least Thirdly, the power-law correlation was fitted to test if this correlation was also existed in individual patients in adaptive radiation therapy (ART) by using their planning data.

2.E | Statistics
All results were presented with mean value and standard deviation.
MLDs were compared using the Wilcoxon signed-rank test with P < 0.05 to be indicative of statistical significance for the groups in the population cohort with SPSS software version 19.0 (IBM, Chicago, USA). Spearman rank correlation was used to test the correlation between LTRs or PTVs and MLDs in the population cohort.

3.A | Patient characteristics
The details of 32 patients were shown in Table 1

3.B | Correlation in the population cohort
All the 144 plans achieved the planning objectives. Spearman rank test showed a significant correlation between MLD and LTR (P = 0.001) or PTV (P = 0.044). Different curves with high R 2 values were observed in Fig. 1 for 3DCRT, ssIMRT, and VMAT, respectively. The R 2 values in PTV and MLD were much lower than those in LTR and MLD, and the correlation between PTV and MLD also had no statistically significance (PTV data not shown). The fitting results showed that there was a more obvious correlation between LTR and MLD than that between PTV and MLD.
For certain LTR, the large and different error bars in Fig. 1 showed that the effects of individual characteristics (especially PTV

3.C | LTR and PTV
The MLD comparisons in PGs and LGs were shown in Table 2, which were divided according to the fitting results in the first step.    In all the three LGs, MLDs for the ssIMRT were also lower on average compared to VMAT (83 cGy, P = 0.028; 109 cGy, P = 0.007; 119 cGy, P = 0.001) with statistically significance. However, an obvious statistical difference was observed in the large-LTR and small-LTR groups compared to those in the corresponding groups in the PGs. MLDs for the 3DCRT was lower compared to VMAT (82 cGy, P = 0.006) in the large-LTR group, and was higher compared to ssIMRT (208 cGy, P = 0.043) in the small-LTR group with statistically significance. Additionally, for all the three techniques, MLDs in the large-LTR and middle-LTR groups were lower compared to the small-PTV and middle-PTV groups, respectively, and MLDs in the small-LTR group were higher compared to the large-PTV group. MLD was relatively more sensitive to LTR than to PTV.

Divisions of groups in the PGs and
LGs were according to the fitting results. If the amount in the groups became a little larger or smaller, slight variations would be observed for the statistical result without changing the statistical significance. The results in Table 2 and Fig. 1 were in agreement with each other for LTRs and MLDs.   ART for lung cancer could achieve clinically relevant reductions in MLD with obvious reduction in tumor volume. [18][19][20] In ART, plans were modified to be consistent with the tumor volume or shape variations due to treatment for a specific patient. 21 In that case, the MLDs. And it is necessary to make further efforts using more cases.

3.D | Correlation in the individual cohort
Thirdly, the ratio of the normal lung volume (excluding PTV) to PTV was used, and all our plans were designed to deliver a uniform dose to PTV. As some part of normal lung tissue was in CTV-to-PTV margins, it might be more proper to use GTV instead of PTV in the ratio which was actually correlated to tumor tissues. Nevertheless, that would not change the correlation between the ratio and MLD. Last but not least, the results were highly influenced by the planning techniques. The number and angle of beams in 3DCRT and ssIMRT were determined manually, which meant the experience of medical physicists had a great impact on results of the plans. Beam number and angle optimization 24 may provide a slight improvement in MLD for the 3DCRT and ssIMRT plans. This effect, however, implied to support the result that 3DCRT and ssIMRT were preferred for patients with large-LTRs.

| CONCLUSION
The impacts of lung and tumor volumes were different on MLD among the three techniques. There were power-law correlations between LTR and MLD in individual NSCLC patients who had replannings due to obvious reduction in tumor volume for the three planning techniques. To avoid irradiating the normal lung, 3DCRT and ssIMRT seemed to be preferred for a large-LTR, and ssIMRT seemed to be preferred for a small-LTR. The findings suggested that LTR was a useful patient characteristic and should be further evaluated in clinical investigations.

CONFLI CT OF INTEREST
The authors declare no conflict of interest.