Prognostic analysis of radiation pneumonitis: carbon-ion radiotherapy in patients with locally advanced lung cancer

This study was performed in accordance with the guidelines approved by the institutional review board of our institution. This study was a retrospective evaluation of all 141 patients who were treated with CIRT at the National Institute of Radiological Sciences (NIRS). Of these patients, 124 individuals who received CIRT between April 2000 and July 2015 were selected for the study. Eligibility criteria were as follows: 1) patients were diagnosed with non-small-cell carcinoma lung cancer by histology or cytology; 2) the clinical stage had been decided by imaging (computed tomography (CT), magnetic resonance imaging (MRI) or positron emission tomography (PET)); 3) the clinical stage ranged from II A to III B according to the Unio Internationalis Contra Cancrum TNM Classification 7^th edition, and the N stage was N0, N1 or N2 [15]; 4) the performance status was 0–2; 5) patients were not treated with chemotherapy within 4 weeks of initiation of CIRT; 6) patients were not suffering from other active cancers; 7) the estimated life expectancy was longer than 6 months; 8) patients had not received previous thoracic radiotherapy; 9) patients were not able to receive other curative therapy (surgery, chemotherapy, and/or radiotherapy) or refused it; 10) patients’ CT and planning data were available for this analysis; and 11) patients received follow-up for more than 6 months at the time of the analysis. This study included 92 patients who met our criteria. Of these patients, 27 of them were excluded because they underwent a second CIRT treatment for recurrent or newly developed lesions in the lung, mediastinal lymph nodes or bone. In addition, 5 patients were excluded because their follow-up periods were less than 6 months. Of these five patients, four patients died of causes unrelated to the treatment, and one patient did not attend the follow-up appointments. Consequently, we retrospectively analyzed the clinical results and treatment plans of 65 patients with locally advanced non-small-cell lung cancer.

The CT images for all patients, which were fixed by an individually tailored immobilization device (Moldcare; Alcare, Tokyo, Japan; and Shelfitter; Kuraray, Osaka, Japan), were taken in the supine or the prone position, with the respiratory system. The CT images were used to develop a 3-dimensional treatment plan using in-house HIPLAN software (NIRS, Chiba, Japan) until the end of 2011, and XiO-N (ELEKTA, Stockholm, Sweden and Mitsubishi Electric, Tokyo, Japan) was used beginning in 2012.

Carbon-ion beams were generated by the heavy ion medical accelerator in the Chiba (HIMAC) synchrotron and were delivered using the respiratory-gated irradiation system [16]. Irradiation was performed in 3–4 fields with 250 MeV or 290 MeV carbon ions. The most common prescribed dose was 72 Gy (RBE) in 16 fractions (45 patients, 69%), followed by 76 Gy (RBE) in 16 fractions (11 patients, 17%) and 68 Gy (RBE) in 16 fractions (9 patients, 14%). All doses were administered 4 days per week over 4 weeks.

Primary lesions and metastatic lymph nodes were contoured as gross tumor volume (GTV) on the CT images. The primary lesions with a 10-mm margin and any prophylactic lymph nodes were defined as the clinical target volume (CTV). For N0 cases, irradiation to prophylactic lymph nodes was omitted. Planning target volume (PTV) was defined as the CTV plus a 5-mm safety margin to account for position uncertainty. The dose was prescribed to the isocenter. The PTV was enclosed conformally at a minimum by the 95% isodose line with the prescribed dose. If there were metastatic lymph nodes, prophylactic irradiation to mediastinal lymph nodes was performed at a median dose of 49.5 Gy (RBE). Maximum dose constraints were as follows: main bronchus, 60 Gy (RBE); esophagus, 50 Gy (RBE); and spinal cord, 30 Gy (RBE).

The dose calculation algorithm at our hospital was updated from the broad-beam calculation using in-house HIPLAN software to the pencil-beam calculation using the XiO-N software in 2012. Different dose calculation algorithms cause different dose distributions for the same treatment. For the current study, the treatment plans calculated by the HIPLAN software were converted to DICOM format and imported into XiO-N. Their dose distributions were recalculated with XiO-N.

First, dosimetric parameters (PTV, MLD, V₅, V₁₀, V₂₀, V₃₀, V₄₀ and V₅₀) were calculated with XiO-N using the bilateral lung volume-GTV as the all-lung volume. Second, clinical parameters (patient characteristics and pulmonary function) and dosimetric parameters associated with ≥ grade 2 RP were identified using univariate analysis. Third, cutoff values of these parameters were determined using the receiver operating characteristic (ROC) curve. Finally, the predictors of ≥ grade 2 RP were identified using multivariate analysis.

The severity of RP was evaluated according to The National Cancer Institute’s Common Terminology Criteria for Adverse Events version 4.0 [17]. For example, we classified the initiation of required steroids as grade 2 and oxygen induction as grade 3.

After completion of treatment, follow-up observations were performed at 1, 3, 6, 9 and 12 months, and then every 3 or 6 months if serious complications did not occur. At follow-up, CT images, blood examination and respiratory function assessments were performed and, if necessary, MRI brain images and PET/CT were added.

We performed univariate analysis using Fisher’s exact test or Mann–Whitney U test. Multivariate analysis was performed using the Cox proportional hazard model. Statistical significance was set at p < 0.05. The ROC curve and the Youden index were calculated to determine cutoff values. We used JMP statistical software (version 11.0) for all statistical analyses.

The median follow-up period for all 65 patients was 22.0 months (6.0–145.7 months). The patients consisted of 51 males and 14 females (median age at treatment, 74 years; range, 46–88) (Table 1). Disease sites included 23 right upper lobes, 17 right lower lobes, 16 left upper lobes, 5 left lower lobes and 4 right middle lobes. The clinical stage was II A disease in 8 patients, II B disease in 29 patients, III A in 16 patients, and III B in 12 patients, according to the TNM classification system. Fourteen patients (22%) received chemotherapy before or after CIRT. No patients received CIRT and chemotherapy concurrently. Fifty-six patients (86%) were current or previous smokers.

Table 2 shows the number of patients according to RP grade. No patients developed ≥ grade 4 RP. Grade 3 RP occurred in 3 patients (5%) at approximately 4–6 months after the initiation of CIRT. These patients were prescribed steroids and home oxygen therapy. The prescribed doses were 72 Gy (RBE) for 2 patients and 76 Gy (RBE) for 1 patient. Grade 2 RP occurred in 6 patients (9%) at a median of 5 months (5–13 months), and all patients were treated with steroids. The prescribed doses were 76 Gy (RBE) for 1 patient, 72 Gy (RBE) for 4 patients, and 68 Gy (RBE) for 1 patient. In total, 9 patients (14%) suffered from ≥ grade 2 RP.

Univariate analysis results of patient characteristics with or without ≥ grade 2 RP are shown in Table 3. There were no patient characteristic factors associated with ≥ grade 2 RP. The incidence of ≥ grade 2 RP tended to be higher for the ≥75 years of age group than for patients <75 years of age, although the difference was not statistically significant.

Univariate analysis of the average pulmonary function was performed to explore potential prognosticators for ≥ grade 2 RP (Table 4). All data regarding pulmonary function were evaluated as continuous variables. The results showed that low percentage of vital capacity (%VC) and 1-second forced expiratory volume (FEV1) values were significant prognostic factors for ≥ grade 2 RP (%VC: p = 0.002; FEV1: p = 0.043). The ROC analysis was used to determine the cutoff values for %VC (86.9%) and FEV1 (1.16 L) for ≥ grade 2 RP. Patients were divided into two groups according to the cutoff values, and the actual incidences of ≥ grade 2 RP were 26.5% for %VC <86.9% and 0% for %VC ≥86.9% (p = 0.002). Similarly, the cutoff values were 6.8% for FEV1 ≥ 1.16 and 28.6% for FEV1 < 1.16 (p = 0.048).

The dose-volume parameters associated with ≥ grade 2 RP were analyzed (Table 5). The results illustrated that the mean lung dose (MLD), the volume of lung receiving ≥5 Gy (RBE) (V₅), V₁₀, V₂₀ and V₃₀ were significant predictive factors for ≥ grade 2 RP. The ROC analysis was used to determine the cutoff values for MLD, V₅, V₁₀, V₂₀ and V₃₀ for ≥ grade 2 RP, which were 12.5 Gy (RBE), 28.8, 29.9, 20.1 and 15.0%, respectively. The actual incidences of ≥ grade 2 RP were 35.7% vs. 7.8% (MLD, ≥12.5 Gy (RBE) vs. less than 12.5 Gy (RBE), respectively, p = 0.018), 24.1% vs. 5.6% (V₅, ≥28.8% vs. less than 28.8%, respectively, p = 0.066), 26.1% vs. 7.1% (V₁₀, ≥29.9% vs. less than 29.9%, respectively, p = 0.058), 21.9% vs. 5.9% (V₂₀, ≥20.1% vs. less than 20.1%, p = 0.074), and 28.0% vs. 0% (V₃₀, ≥15.0% vs. less than 15%, p = 0.022).

Multivariate analysis was performed for ≥ grade 2 RP using the two variables with the most significant p values from the univariate analysis of the clinical, pulmonary functional or dosimetric factors (Table 6). The results showed that %VC (odds ratio = 13.7; p = 0.0041) and V₃₀ (odds ratio = 6.1; p = 0.0221) were significant prognosticators for ≥ grade 2 RP.