Stereotactic radiosurgery (SRS) is a term originally created by Lars Leksell to describe a new approach to radiotherapy for brain tumors using multiple convergent beams, precise localization via a stereotactic coordinate system, rigid patient immobilization, and singlefraction treatment. The technique contrasts sharply with conventional three-dimensional (3D) conformal external-beam radiotherapy (CRT), in which the total radiation dose is given in numerous-often 25 or more- small daily doses or fractions. The dose-fractionation allows for repair in normal tissues between fractions, but the need to reproduce the patient's initial positioning on multiple occasions necessitates the inclusion of an extra margin of normal tissue around the tumor in the treatment field to account for daily set-up variations. With SRS, a stereotactic immobilization frame is tightly pinned to the patient's skull, allowing the spatial precision for treating with zero to minimal margin of surrounding tissue. The resulting reduction in volume of normal tissue irradiated leads to the ability to deliver higher doses, thereby potentially improving tumor control. Positive clinical experience accumulated over the past 2 decades has affirmed the utility of SRS in the treatment of a variety of intracranial tumors. For brain metastases, treatment with SRS lengthens median survival compared to conventional radiotherapy,[ 1] and for other clinical entities (eg, arteriovenous malformations, trigeminal neuralgia, acoustic schwannomas), SRS has offered a valuable new alternative to surgery.[2] Recent advances in 3D-CRT include alterations in the treatment schedule to shorten the length of treatment as well as modulation of the intensity of the radiation beam profile, either to provide selective boost doses to gross disease within a larger field or to minimize the radiation dose to irregularly shaped normal tissues adjacent to the area treated. However,these approaches have changed the daily or fractional dose by a very limited proportion relative to conventional doses, and the capability of delivering higher fractional doses in the radiosurgical range have not been previously tested. The anatomic characteristics of the skull and stability of cranial contents make SRS feasible for intracranial tumors, but the lack of a similar fixed bony reference structure as well as respiration-induced movement creates difficulty for application outside the cranium. In 1994, Lax et al of the Karolinska Institute reported a method for performing what is now called stereotactic body radiation therapy (SBRT) for abdominal malignancies, using a custom body cast with stereotactic coordinates.[3] Blomgren and Lax subsequently reported an 80% progression- free rate in 42 tumors of the lung, liver, or abdomen treated with this method to mean doses of 30 Gy delivered in one to four fractions.[4] Since this report, there has been increasing interest in and reports on the use of a similar approach of spatially precise, single-dose treatment or very brief regimens involving typically three to five individual doses (hypofractionation). The principles that have generally been used in the selective application of SBRT mirror those of SRS. Treatment is limited to small- to moderatevolume discrete tumor targets. Prophylactic coverage of the surrounding region of potential microscopic disease is not performed, thereby maintaining a simple volume and eliminating dose gradients between spatially separated target structures. Targets are within radiosensitive normal tissue or surrounding structure with parallel (as opposed to serial) architecture, where small portions of normal tissue surrounding the structure can receive high doses of radiation without clinical consequences due to functional reserve of unaffected organ. A dose-response relationship should be exploited using the capacity for increased dose delivery offered by SBRT. In this paper, we describe the methods used in treating patients with "extracranial radiosurgery," "extracranial radiotherapy," or SBRT, the clinical scenarios in which it is applicable and why, and unresolved issues and future directions for investigation. Physics and Technology of SBRT For single high-dose or hypofractionated SBRT, it is necessary to maintain a higher confidence in radiation dose delivery than with 3D-CRT. Several methods have been used to achieve spatial accuracy during SBRT, and herein are discussed methods that have been described in published reports of patients treated with hypofractionated, high-dose radiation and multiple beams. The term stereotactic implies the use of an external frame of reference indexed to internal volumes to be targeted. Not all of the methods necessarily utilize strict stereotactic localization methods, but the term stereotactic has been a common denominator when describing this literature as a whole. In general, the sequence of events for patients undergoing SBRT include the following: (1) computed tomography (CT) simulation, (2) immobilization, (3) planning, (4) repositioning, (5) relocalization, and (6) treatment delivery. The CT simulation is used to assess tumor size, location, and range of motion and to determine if the patient can tolerate the planned immobilization. During the initial CT, the range of tumor motion observed provides the necessary data for the required margin to be added around the gross target volume to define the planning target volume-the actual volume to be targeted in order to be certain that the gross target volume receives the full intended dose during each treatment. Immobilization includes a custom-fitted device to minimize motion and breathing effects, and provide a reproducible setup. The planning must address the small-field dosimetry issues common to cranial stereotactic radiosurgery but with the additional focus of inhomogeneity corrections for lung fields, and volumetric considerations for normal tissue complications of the critical organ being treated (ie, lung or liver). Repositioning addresses the accurate setup of the patient in the planned position, whereas relocalization addresses the specific identification of the tumor and planned isocenter in the treatment field. Finally, treatment delivery is performed using any of a wide array of high-precision beam delivery techniques, including mini-multileaf collimation, gantry mounted linear accelerators (linacs), and combined imaging and treatment units. Patient Immobilization, Positioning, and Relocalization All of the reports on SBRT have included a CT simulation to assess tumor location, size, range of motion, and feasibility of treatment within the established parameters of the particular institute. Lax's initial report involved the use of a body cast within a rigid box frame with radio-opaque scale markers for acquisition of imaging data. The scales mounted on the frame, corresponding to fiducials, were used to establish a coordinate system in the 3D space of the treatment room. Diaphragmatic movement was limited by using a plate applying pressure on the abdomen.[3] Blomgren's mean reproducibility for 75 evaluable tumors was reported as 3.7 mm in the transverse plane and 5.7 mm longitudinal; a margin of 5 mm was added to the gross target volume in the transverse plane and a margin of 10 mm was added in the longitudinal direction to define the planning target volume.[4] Since the report of the Karolinska's treatment method, some institutions have used the same technique of a body frame with diaphragmatic pressure plate and stereotactic coordinates to treat lung and liver tumors (eg, Leibinger frame, Elekta frame), while others have used a variety of immobilization and repositioning methods in attempts to achieve similar accuracy. Overall, the reported accuracy is within 5 mm for the various methods utilized. A comprehensive evaluation of these methods is beyond the scope of this paper, but Table 1 lists some of the methods used; select methods are described in more detail below.[3,5-17] Sato et al described a frameless stereotactic technique for metastatic liver cancer in which 14 patients were instructed to keep shallow respiration with an oxygen mask and abdominal belt on the treatment table.[5] Patients received transcatheter arterial chemoembolization with lipiodol, which played a role as a radio-opaque marker for x-ray simulation and CT monitoring. Relocalization accuracy was performed with a multifunctional unit incorporating CT scanner, x-ray simulator, and linac sharing one treatment couch. Patients underwent imaging and target localization prior to each daily treatment without getting off the couch between scan and treatment. No overall statistics on the patient immobilization and tumor relocalization were reported.[5] Nakagawa et al utilized a similar approach of immediate pretreatment image verification with a megavoltage (6-MV) CT scanner mounted on a linac.[6] Hara et al employed a method of transferring the patient to the treatment unit within a custom bed after a CT scan was performed to place the isocenter; treatment was performed with respiratory gating.[7] The CyberKnife (Accuray, Sunnyvale, Calif) consists of a gantrymounted linac coupled with a realtime image guidance system. The treatment site is imaged by two x-ray fluoroscopes and bony landmarks are referenced to the components of the hardware. Once the location of the bony landmark has been determined relative to the robot arm, the position of the tumor is accounted for and adjusted accordingly. The cameras can frequently update the body position and accommodate for patient motion and setup changes, and in that regard, this device has six degrees of freedom allowing for volumes with more complex shapes. Murphy reported a mean total radial error of 1.6 mm and found that the accuracy of the system was equal to or better than that of frame-based systems currently in use.[8] Whereas the Lax system was designed to treat soft-tissue tumors, Hamilton et al described a system for spinal cord irradiation for extracranial stereotactic radiosurgery using a screw fixation technique.[9] The Hamilton system was based on the principles of rigid skeletal fixation. A rigid box is used for the patient to lie prone and a clamp is fixated on the surgically exposed spinous process above and below the target areas. Once the coordinates of the target have been acquired from the CT images, the box is transferred to the linac treatment table and aligned using standard lasers. The clinical experience with this frame has shown reproducibility on the order of 2 mm. Primary Lung Tumors Several reports have described the treatment of early-stage primary non- small-cell lung cancer (NSCLC) with SBRT. Although surgical resection is the primary modality of treatment, radiotherapy has been used in patients who are not medically fit to undergo surgery. Survival rates among patients who received 3D-CRT are generally inferior to those of surgery, but selection bias likely plays a role on the observed difference, since patients referred for 3D-CRT usually have significant intercurrent illness. The use of higher doses has been associated with improved outcomes.[18,19] Sibley summarized the results of CRT in patients with T1/2, N0 tumors, concluding that elective nodal irradiation was unwarranted and higher doses were associated with improved tumor control and disease-free survival.[20] Although 3D-CRT doses of 60 to 66 Gy were not associated with radiation pneumonitis in this review, a study by Langendijk et al suggests that 3D-CRT may negatively affect quality of life.[21] Patient-reported quality of life was assessed at intervals starting prior to radiotherapy in 46 patients, and showed worsening trends in dyspnea, appetite, and fatigue up to 24 months posttherapy. Despite the fact that progression of medical comorbidities is common in this population, the worsening trend was seen to begin at 2 weeks into therapy. Dysphagia was the only symptom that was statistically different between patients treated to locoregional fields vs those receiving treatment to the lung primary alone. Reports of SBRT in early-stage NSCLC describe doses that are biologically higher than doses used in 3D-CRT. Based on tumor control probability calculations derived from a database of patients treated with conventional techniques, Martel et al predicted that a dose of 84 Gy (in 2-Gy fractions) was necessary to achieve local progression-free survival in 50% of patients.[22] However, any doseescalation scheme using convention-al fractionation requires an increase in overall treatment time, which has been negatively correlated with tumor control rates and patient outcomes.[ 23] Randomized clinical evidence supports the use of a shortened treatment duration (accelerated fractionation) in NSCLC, albeit at a cost of increased pneumonitis with CRT.[24] Numerous series have documented the efficacy and safety of SBRT in the treatment of early-stage NSCLC (see Table 2).[7,13,14,16,25-27] Although median follow-up durations are short, local control rates are high and compare favorably with those for conventional treatment. Due to the large differences in fraction sizes between SBRT and conventional treatment, direct comparison of total nominal doses is not feasible. The linear quadratic method is generally accepted for drawing comparisons between different fractionation schemes. Table 3 compares the biologic effective dose of the two types of treatment, assuming an alpha/beta ratio of 10.[7,13,16,27,28] One can readily see that the biologic effective doses for SBRT are higher than the standard 60 to 66 Gy for conventional therapy. Despite the use of higher doses than typically given, SBRT has not been associated with an increased rate of complications. Table 4 summarizes the toxicities described in the literature for SBRT in patients with lung tumors (both NSCLC and metastatic), with the reported incidence of grade 3 toxicities generally less than 5%.[4,6,7,13,14,25-27,29] Fukumoto performed routine pulmonary function testing before and after treatment and found no decline in median forced expiratory volume in 1 second (FEV1) or diffusing capacity of the lung for carbon monoxide (DLCO) values following therapy.[15] Given that a dose-response relationship exists for NSCLC, and SBRT has been found to have little associated toxicity, dose-escalation studies are needed. Timmerman et al have performed the only dose-finding study for SBRT thus far.[16] A total of 36 patients with medically inoperable stage I NSCLC were treated in a phase I study using a three-fraction regimen, starting at 8 Gy per fraction. The technique employed was the same as that of the Karolinska Institute. Dose was safely escalated to 20 Gy per fraction (total dose: 60 Gy), with the maximal tolerated dose not reached. Toxicity consisted of one case of grade 3 pneumonitis and one case of grade 3 hypoxia; pulmonary function testing revealed no statistically significant changes following treatment. Radiographic response was seen in 87% of patients. At a median follow-up of 15 months, there were 6 local failures, all occurring in patients treated with less than 18 Gy per fraction.[16] Despite the stated potential advantages of SBRT, a similar dose-escalation study using CRT has not been performed in this patient population. Thus, one cannot exclude the possibility of similar results being achievable with higher doses of CRT. Lung Metastases Most series describing SBRT for lung tumors include patients treated for metastatic disease. Dose-fractionation schemes and techniques employed are similar to those for primary lung tumors, and control rates range from 66% to 100% with minimal reported toxicitiy (see Table 5).[4,6,7,13,25-27] Although radiotherapy has not traditionally been employed in curative fashion for these patients due to perceived poor prognosis, data exist to support aggressive local therapy for selected patients with lung metastases.[30] The International Registry of Lung Metastases reported the largest series to date of surgical metastasectomy in 5,206 cases from 18 institutions in Europe and North America. Primary histology was epithelial in 2,260 cases, sarcoma in 2,173, germ cell in 363, and melanoma in 328. With a median follow-up of 35 months, actuarial survival after complete metastasectomy (achieved in 88% of patients) was 36% at 5 years and 26% at 10 years. Multivariate analysis showed better prognosis for patients with a single metastasis, disease-free intervals of 36 months or more, and germ cell tumors, although long-term survivors were seen in all histologic types.[31] Some series of surgical metastasectomy have also found tumor size and nonmelanoma histology to be of prognostic significance, although others report no differences.[30,32] Given the high rates of local control and low toxicity demonstrated for SBRT, similarly selected patients who are not otherwise eligible for surgery may receive comparable benefit from treatment with SBRT and should be considered for treatment.