This retrospective study was approved by the institutional review board of the National Center for Global Health and Medicine Ethics Committee. Informed consent was waived by the Ethics Committee due to the retrospective nature of the study. According to the approval, informed consent from the patients was not required. Inclusion criteria were histologically/cytologically confirmed NSCLC or the tumor size enlarged more than 25% on sequential CT examination, clinical stage I (T1/2N0M0) staged by CT examination, age ≥20 years, Eastern Cooperative Oncology Group performance status 0-2 and no history of radiation therapy. Exclusion criteria were interstitial pneumonitis or huge bullae on CT examination, centrally located tumor, history of chemotherapy, sever psychologically disease and active infectious disease. Between November 2003 and October 2015, patients with stage I NSCLC receiving SBRT at a dose of 50 or 60 Gy over five fractions at our hospital were studied. Eighty consecutive patients were identified. The median age was 75.5 years (range, 29-95 years), 54 and 26 patients were male and female, respectively, and there were 49 and 31 T1 and T2 tumors, respectively (Table I). There were no patients with interstitial pneumonitis. Patients with interstitial pneumonitis did not undergo SBRT because of a risk of pneumonitis in our treatment policy. Centrally located tumor was treated by a different dose fraction. None underwent chemotherapy or molecular targeting therapy during the observation period. Two experienced radiologists diagnosed 33 patients with emphysema based on CT examination before radiotherapy. RP was determined according to common terminology criteria for adverse events (CTCAE v. 4.0). All patients underwent physical examination, blood tests, and chest X-rays at least every 4 months up to 3 years, and then every 6 months. CT examinations were performed at least every 6 months up to 3 years and then every year or if tumor enlargement was suspected upon chest X-ray.

A planning CT was acquired with patients in the supine position on a vacuum cushion with their arms raised. To assess intra-fraction tumor movement, four-dimensional (4-D) CT images were obtained using commercial software (Real-Time Position Management System [RPM] system^®, Varian Medical System, Inc.). The respiratory cycle was controlled by a self-monitoring device for respiratory movement (Abches^®, APEX Medical, Inc.) Ten equally spaced respiratory time bins from CT images were reconstructed from 4-D CT. An inspiratory phase CT image was also obtained using a self-monitoring device. Typically, three bins were used to create the gross tumor volume (GTV) in each bin. Internal target volume (ITV) was created from each GTV and contoured on an inspiratory phase CT image. Planning target volume (PTV) was generated by adding 5 mm to the ITV. Respiratory-gated irradiation or breath-hold technique was used according to patients' respiratory cycles and tumor movement. CT images were obtained within at least 10 days before the initiation of radiotherapy. Dose calculation was performed by a planning system (Eclipse ver. 10^®, Varian Medical Systems). A dose of 50 or 60 Gy given in five fractions was prescribed to the isocenter. Five to 10 (median, seven) ports to optimize the directions for non-coplanar static beams or dynamic arcs using 6 MV energy were planned. Mean lung dose and the percentage of normal lung volume receiving ≥20 Gy (V20) and >5 Gy (V5) were evaluated in the inspiratory phase of CT images. On-board cone beam CT was performed to reduce inter-fraction error before every irradiation session.

Parameters of the 64-detector CT (Aquilion TSX-101HA, Toshiba Medical System, Inc.) were as follows: Auto exposure control mA (tube current range, 40-200 mA); tube voltage, 120 KV; and gantry rotation time, 0.5 sec. CT data were reconstructed in an axial CT image with 1.6-mm section thickness. CT values varied with the CT scanner, and thus all patients underwent CT using the same machine. The details of quantitative assessment by CT have been reported elsewhere (16). CT images of the inspiratory phase were obtained from a planning CT, and quantitative assessment of the percentage of lung CT voxels below the threshold of -910 HU were analyzed (Synapse Vincent^® software, Fujifilm Co.). The percentage of low attenuation areas (LAA%) in both whole lungs and the average density value of both lungs were obtained.

Quantitative CT value and dosimetric parameters were analyzed by Wilcoxon's rank-sum test. The time to RP development was calculated from the first day of SBRT. The time to RP was estimated by Kaplan-Meier analysis and compared with log-rank tests. Cox's proportional hazard model was used to analyze risk factors for RP. The CT values assessed for emphysema were evaluated by a receiver operating characteristic (ROC) curve. Most cases of radiation pneumonitis occurred within 1 year, and hence ROC analysis is an acceptable approach to clarify radiation pneumonitis. A two-sided probability value (P<0.05) was considered statistically significant. Statistical analysis was performed with commercial software (STATA v. 13^®, StataCorp).

During the median observation period of 18.8 (1.8-106.8) months, 26 (33%) and three (4%) patients experienced Grade 1 and Grade 2 RP, respectively. Three patients with Grade 2 RP were successfully treated with steroids. No patients suffered from ≥Grade 3 RP. The median time to development of RP was 14 months. Actuarial RP rates at 6, 12 and 18 months were 16.9% [95% confidence interval (CI), 9.7-28.5%)], 49.0% (95% CI, 35.1-65.0%), and 74.1% (95% CI, 51.6-92.0%), respectively.

LAA% (<-910 HU) in patients with emphysema was significantly higher than that in patients with non-emphysema (P<0.001, Table II). The average HU of the whole lung in patients with emphysema was significantly lower than that in patients without emphysema (P<0.001). Fig. 1 shows the number of patients with an average HU of the whole lung and LAA% of -910 HU and less. No patients who had an average HU of ≤-850 HU experienced RP (Fig. 1).

In ROC analysis of patients with and without emphysema, the AUC of LAA% (<-910 HU) and average HU were 0.948 and 0.942, respectively. Fig. 2 shows a ROC curve comparing patients with and without RP. The AUC of average HU, mean lung dose, LAA% (<-910 HU), V20, and V5 were 0.692, 0.643, 0.641, 0.619, and 0.591, respectively.

Table III shows the difference in the quantitative CT value and dosimetric parameters between RP and non-RP. A LAA% (<-910 HU) in patients without RP was significantly higher than that in patients with RP (P=0.037). The average density of the whole lung in patients without RP was significantly lower than that in patients with RP (P=0.004). V20 (P=0.078) and V5 (P=0.178) were not significantly different between subjects with and without RP. The mean lung dose in patients with RP was higher than that in patients without RP (P=0.034). Fig. 3 shows RP rates analyzed by the Kaplan-Meier method. The RP rate with a mean lung dose of <4 Gy was not significantly different compared with doses ≥4 Gy (P=0.11). The RP rate in LAA% (<-910 HU) of ≤25% was significantly higher compared with that of subjects with an LAA% (<-910 HU) >25% (P=0.037). The RP rate in subjects with an average HU >-790 HU was significantly higher compared with those with ≤-790 HU (P=0.036). The RP rate at an average HU of >-850 HU was significantly higher compared with those with an average HU of ≤-850 HU (P=0.045).

Table IV shows age [hazard ratio (HR)=2.46, P=0.03] and average HU (HR=3.39, P=0.02) were significant risk factors for RP by multivariate analysis. Sex, T factor, and mean lung dose were not significant.