Clinical outcomes of 130 patients with primary and secondary lung tumors treated with Cyberknife robotic stereotactic body radiotherapy

Cyberknife® (Accuray Inc. Sunnyvale, USA) robotic SBRT treatments were started at the Liege University Hospital in April 2010. Ordinary indications for SBRT treatment include T1-T2 primary NSCLC, recurrent primary lung tumors, and solitary-, or oligometastases. However, more rarely this treatment is appied on T3 tumors or soliter lymph node metastases.^16-19 The majority of primary and recurrent lung tumors in our cohort were considered ineligible for surgical resection because of poor lung functions or severe comorbidities. For metastatic lesions medical inoperability; > 1 lesions in different lobes or lungs; prior lobectomy and patient preference were the major causes leading to the choice of SBRT. Based on individual medical consideration and absence of realistic therapeutic alternatives a small number of unusual indications were also included, like patients harboring T3N0 or T1N1 disease. In the present study 130 consecutive patients treated with BED_10Gy ≥ 72 Gy were evaluated. Central or large tumors were not excluded, but the dose and number of fractions were adapted. Central lesions were defined as lesions located whithin 2 cm from the pulmonary hilum, heart, great vessels, or trachea. Indications for each individual patient were discussed and approved in multi-disciplinary tumor boards. Especially for primary tumors, pathological confirmation was requested either by bronchofiberoscopy or transthoracic biopsy. For patients considered not eligible for histological confirmation (due to technical or medical reasons), the indication was based on strong clinical suspicion supported by positron emission tomography (PET).

One hundred and thirty patients, with a total of 160 lung lesions were treated between April 2010 and June 2012. Patient and tumor characteristics are listed in Table 1.

Table 1

Patient, tumor and treatment characteristcs

Median age of patients at treatment was 71 years (range 40–93), and 59% (n = 77) were males. Distribution of lesions were: 53% (n = 86) primary, 22% (n = 35) recurrent tumor/intrapulmonary metastasis of a lung tumor and 25% (n = 39) metastases from other cancer. Cancer of origin for metastatatic lesions were: colorectal (49%; n = 19), salivary gland (13%; n = 5), breast (10%; n = 4), melanoma (5%; n = 2), kidney (5%; n = 2), neuroendocrin (3%; n = 1), multiple primary (13%; n = 5), unkown (2%; n = 1). Distribution of histological types for the patient group with pathologically confirmed primary lung cancer was: 47% adenocarcinoma (n = 29), 33% squamosus cell carcinoma (n = 21), 15% NSCLC (n = 9), 5% undifferentiated (n = 3). The maximal number of lesions treated by SBRT in the same patient was four. Four patients were presented with stage T1N1 disease. For these patients the affected lymph node(s) were also treated with SBRT. Positivity of these lymph nodes were based on high SUVmax value on PET CT, without cytological confirmation, but usually the histology of the belonging primary tumors were known.

One patient was categorized as T3 for tumor size, the other four T3 patients had mediastinal pleura invasion or separate nodule in the same lobe.

Distribution of the 113 peripheral lesions was : n = 66 primary, n = 21 recurrent, n = 26 metastasis. The group of 47 central lesions was composed of 17 primary tumors plus 3 synchronous N1 lymph nodes, 14 recurrent cancers and 13 metastases.

Technical characteristics and tracking options of the Cyberknife robotic SBRT system have been exhaustively detailed elsewhere.^16-17 For thoracic tumors there are three different tracking types, which can be appropriately selected according to each clinical case. Synchrony® is a real-time tumor tracking algorithm which requires fiducial markers to be previously inserted inside or near to the target. The fiducials are detected by orthogonal X-rays at the treatment room. The system includes an infrared camera that monitors the movement of the chest. During treatment, spatial information on the location of the fiducials and data of the respiratory cycle are connected to redirect the robot, and realize real-time tracking. Fiducial insertion can be contraindicated for some patients because of the inherent risk of pneumothorax. For selected cases, when tumor silhouette is sharply identified on both ortogonal X-ray detector panels, the algorithm of Xsight Lung® can be used for tracking the target, without the need for implanted markers. When none of these two previously mentioned tracking algorithm is feasible, tracking is performed on the vertebra (XsightSpine®).

Planning CT images were obtained with a slice thickness of 1 mm. Patients were immobilised using an individual vacuum bag “in supine position, with arms next to the body. Four-dimensional (4D) CT simulation was not introduced for Cyberknife treatment, thus expiration and moderate inspiration CT scans were acquired to estimate magnitude of respiratory releated tumor movement. In case of fiducial-, or direct tumor tracking, only expiration CTs were used for delineation. For patients with fiducial markers, CT simulation was delayed with a minimum of 10 days after implantation to minimise uncertainty linked to the potential marker migration.

The vast majority of patients (n = 125; 96%) had PET CT scans in treatment position using the same individual vacuum bag used at the CT simulation, to optimize target volume definition.

For patients with real-time tumor tracking generally a margin of 3 mm was applied around the gross tumor volume (GTV) to achieve clinical target volume (CTV). CTV contours were then, corrected manually when overlapping with ribs or mediastinal structures. An additional 2 mm was added to create planning target volume (PTV).

When real time tumor tracking was not feasible, we used an internal target volume of GTV, large enough to cover all possible tumor positions during the respiratory cycle. After that, the method and the margins for creating CTV-internal target volume (ITV) and PTV was similar to real time tracked patients.

Treatment plans were implemented with Multiplan treatment planning system (TPS) version 5.1 (Accuray Inc. Sunnyvale, USA), using Ray Tracing calculation algorythm. Prescription doses varied between 40 to 60 Gy in 3 to 5 fractions, depending on proximity to organs at risk (OAR) and on tumor size. Dose was typically prescribed to the 80% isodose line (75–82%) encompassing the PTV. Dose constrains to OARs were applied according to a class solution (Table 2) which was based on published data of Timmerman and AAPM Taskgroup 101 guidelines.^18-19

SBRT treatments were performed by Cyberknife Robotic Radiosurgery treatment unit Accuray Inc. Sunnyvale, USA). Treatment consisted of typically 100–200 non-coplanar beams using Iris® various aperture collimator in a range between 15 to 60 mm with a dose rate of 600 MU/min.

Patients were followed up by the treating radiation oncologist and/or by referring pulmonologist or oncologist. In addition to regular CT-scans, metabolic follow up of treatment effect by PET CT was performed in 118 patients (91%) to make distinction between local disease progression and localized pulmonary fibrosis.^20,21 Acute and late toxicities were evaluated using the Common Terminology Criteria for Adverse Events (CTCAE v4.0). Toxicity was classified as acute up to 3 months after SBRT.

The patient follow up time was defined as the period between the first day of Cyberknife treatment to the date of last visit or death.

Patient and lesion characteristics and toxicities were described in terms of means or medians (range) or in terms of numbers (%). A descriptive analysis was used to present patient and treatment characteristics and toxicity data. Local control (LC), overall survival (OS) and cause-specific survival (CSS) rates were estimated by the Kaplan-Meier method. The prognostic value of patient and tumor characteristics on LC was determined using uni- and multivariate Cox regression models. Results were considered to be statisticallysignificant at p-values ≤ 0.05. Statistical analysis was performed with the SAS version 9.3 software (SAS Institute, Cary, NC, USA).

This retrospective cohort study was approved by institutional review board.

For the whole cohort the actuarial 1-, 2-, and 3-year LC rates were 86%, 75%, and 62%, respectively.

In univariate Cox regression model, a higher BED_10Gy was associated with better LC (p = 0.008). Analysis of the different dose levels found a cutoff value between BED 112.5 Gy and 132 Gy. Treatments using doses higher than 112.5 Gy showed a significant advantage in terms of LC, resulting 1-, 2-, and 3-year actuarial LC rates of 93% vs 73%, 80% vs 61%, and 63% vs 54%, respectively (p = 0.0061, HR = 0.384; Figure 1).

In univariate analysis there were no significant differences between primary (P), recurrent (R) and metastatic (M) lesions in terms of actuarial LC (p = 0.091). However in pairwise comparison primary tumors provided improved results compared to metastases: 1-, 2-, and 3-year LC rates were 89% vs 84%, 80% vs 59%, and 64% vs 53%, respectively (p = 0.035; Figure 2). Other factors such as tumor tracking (inclusive tracking modality), and histological confirmation of malignancy had no significant effect on LC.

During separate analysis of the primary tumor group there was no significant difference in LC according to T-stage. However, there was a a nonsignificant trend favoring LC in T1 compared to T2 (p = 0.063).

Local control was significantly higher for peripheral lesions, compared to central lesions (p = 0.025), resulting in 1-, 2-, and 3-year LC rates of 91% vs 74%, 79% vs 63%, and 60% vs 56%, respectively (Figure 3).

In univariate Cox regression model, larger GTV and PTV volumes were associated with a higher risk of local relapse (p = 0.0034 and p = 0.0013, respectively).

The variables tested in multivariate analysis were tumor type (primary/recurrent/metastasis), tracking (yes/no), confirmed histology (yes/no), location of lesions (central/peripheral), BED_10Gy, GTV, PTV, and PTV coverage. These variables were selected in order to determine treatment and tumor factors influencing LC. In multivariate analysis the metastatic origin of lesions (HR = 7.3; p < 0.0001), the histological confirmation of malignancy (HR = 4.1; p = 0.0052) and larger PTV (HR = 1.03; p < 0.0001) were associated with significantly lower LC rate.

One-, two-, and three-year actuarial rates of OS were 85%, 74% and 62%, respetively, whereas the respective rates of CSS were 93%, 90%, and 80%.

Treatment related Grade (G) 3 or higher acute and late toxicities were observed at 3 (2%) and 6 patients (5%), respectively. Acute toxicities included 2 cases (1.5%) of G3 pneumonitis and a single case (0.8%) of G5 pulmonary haemorrhage. This latter elderly (85 years old) patient had a fatal ipsilateral pulmonary haemorrhage 1 month after the completion of his SBRT (45 Gy in 5 fractions) for a right sided central tumor recurrence, and was classified as a possible treatment related adverse event. The patient had already been treated with chemotherapy 4 years earlier for his primary lung tumor, and 1 year earlier by Cyberknife for a contralateral upper lobe relapse without progression until the time of death.

Late toxicities were G3 dyspnea (n = 3; 2.3%, all presenting with chronic obstructive pulmonary disease [COPD] Global Initiative for Chronic Obstructive Lung Disease [GOLD] III prior to SBRT), G3 sick sinus syndrome (n = 1; 0.8%) requiring pacemaker implantation 8 months after SBRT, G3 pain due to a rib fracture requiring major analgesic (n = 1; 0.8%). One patient (0.8%) suffered a fatal haemorrhage (G5) 7 months after SBRT for a centrally located recurrent tumor mass, which invaded vascular structures already at the time of detection, and had shown progression after Cyberknife treatment (40 Gy in 5 fracions). This case was encoded as a treatment related adverse event, although local tumor progression could not be formally excluded.

Grade 2 late toxicities were also recorded such as asymptomatic or moderately painful rib fractures (n = 5; 3.8%), recurrent laryngeal nerve palsy (n = 1; 0.8%), late radiation pneumonitis (n = 14; 10.8%) and pneumothorax (n = 6; 4.6%) after transthoracic marker placement requiring tube placement for a few days.