Treatment of malignant tumors of the skull base with multi-session radiosurgery

We performed a retrospective review of 464 patients with intracranial tumors who were treated with CyberKnife^® stereotactic radiosurgery (CKS) at Georgetown University Hospital between January 2002 and December 2007. One hundred forty-five patients were classified as having tumors of the skull base, of which 108 were benign. Thirty-seven patients had 37 lesions that were classified as malignant skull base tumors. Six patients who had follow-up periods less than or equal to four weeks were eliminated from statistical consideration, thus leaving 31 patients for analysis.

For the purposes of this study, skull base lesions were defined as those that involved the osseous structures of the base of the skull, in close proximity to the critical neurovascular structures of the region. All the tumors included in this study either completely encircled, partially circumscribed, or directly contacted the brainstem, optic chiasm, or cranial nerves with meaningful remaining function. Primary brain tumors were excluded, unless they had the potential to metastasize and were thus considered malignant. An example of such a tumor is a hemagiopericytoma. Malignant orbital, sinus and head-and-neck tumors were included in this study only if there was intracranial extension.

This malignant skull base tumor group consisted of 21 men and 10 women, with a median age of 57 (range: 11 – 81) (Table 1). The histopathology of all tumors was either known from prior microsurgical resection, biopsy, or was presumed based on the intracranial extension of known head and neck cancers.

A multidisciplinary meeting of specialists that included neurosurgeons, otolaryngologists, radiation oncologists, medical oncologists, and neuroradiologists evaluated all patients. A collective decision to treat with radiosurgery was made for each individual patient. Radiosurgery was only offered to patients for whom conventional microsurgical resection was contraindicated because of high neurological risk, overwhelming medical comorbidities, poor prognosis with limited survival, or recurrent disease in the presence of prior microsurgical resection, chemotherapy and radiation therapy.

The CyberKnife^® radiosurgical system was used to administer cranial radiosurgery in every case. The technical aspects of CKS for cranial tumors have been described in detail [46, 50]. Briefly, the patient's head was immobilized by a malleable thermoplastic mask during the acquisition of a thin-sliced (1.25 mm) high-resolution computed tomography scan, which was used for treatment planning. The use of a contrast-enhanced MRI fused to the treatment planning CT scan was at the discretion of the treating physicians. This decision was influenced by various factors, such as previous radiation to the area, performance status, treatment intent and extent of contact and compression of critical neurological structures. The target volumes and critical structures were then delineated by the treating neurosurgeon. An inverse planning method with non-isocenteric technique was used for all cases, with specific dose constraints on critical structures such as the optic chiasm and brainstem. The planning software calculated the optimal solution for treatment, and the dose-volume histogram of each plan was evaluated until an acceptable plan was found. The treating neurosurgeon and radiation oncologist, who have a shared responsibility for all aspects of the treatment planning and procedure, determined the minimal tumor margin dose of the target volume, the treatment isodose and the number of treatment sessions into which the total dose was to be divided. This decision was influenced by various factors, such as previous radiation to the area, tumor volume, and extent of contact and compression of critical neurological structures. In most cases, the treatment dose was prescribed to the isodose surface that encompassed the margin of the tumor.

The delivery of radiosurgery by the CyberKnife^® was guided by real-time imaging. Using computed tomography planning, target volume locations were related to radiographic landmarks of the cranium. With the assumption that the target position is fixed within the cranium, cranial tracking allowed for anatomy based tracking relatively independent of patient's daily setup. Position verification was validated several times per minute during treatment using paired, orthogonal, x-ray images.

The homogeneity index and new conformity index were calculated for each treatment plan. The homogeneity index (HI) describes the uniformity of dose within a treated target volume, and is directly calculated from the prescription isodose line chosen to cover the margin of the tumor. It is calculated by the following equation:

The new conformity index (NCI) as formulated by Paddick, and modified by Nakamura, describes the degree to which the prescribed isodose volume conforms to the shape and size of the target volume [52, 53]. It also takes into account avoidance of surrounding normal tissue. It is calculated by the following equation:

Post-radiosurgical follow-up was typically performed in a multidisciplinary clinic of the treating neurosurgeon and radiation oncologist beginning one month after the conclusion of radiosurgery. Patients were subsequently followed in three-month intervals. During each follow-up visit, a clinical evaluation and physical examination were performed as well as a review of pertinent radiographic imaging. If a patient experienced deterioration in their clinical condition at any point during the follow-up period, an immediate evaluation was performed. The progress of all patients was discussed periodically at a multidisciplinary tumor conference of various specialists, ensuring precise interpretation of the available data. We analyzed tumor response, clinical outcome, treatment-related complications and survival during the follow-up period.

The characteristics of the study group including the distribution of gender, age, tumor histology and location are detailed below and summarized in Tables 1 and 2. The most frequent tumors in this series were squamous cell carcinoma (6 lesions), adenoid cystic carcinoma (5 lesions), rhabdomyosarcoma (2 lesions) and metastases of melanoma and renal cell carcinomas (3 lesions each). The median tumor volume was 18.3 cc (range: 3.2 – 206.5 cc).

Tumors varied in their skull base location, as illustrated in Table 2. A number of lesions, however, spanned multiple anatomical locations. CKS was the primary treatment to the malignant skull base tumor in 18 patients (58%). Of the 13 patients with previous treatment to the tumor involved in this study, 6 (46%) had previous craniofacial surgery, 4 (30%) had previous external beam radiation, and 1 (7%) had previous stereotactic radiotherapy. Four patients (13% of the entire series) had undergone biopsy only.

The specific dose and fractionation scheme for the tumors in this series was influenced by various factors, including previous radiation to the area, tumor volume, and extent of contact and compression of critical neurological structures. Details of the radiosurgical treatments are found in Table 3. A median treatment dose of 2500 cGy was delivered to the margins of the tumors in this study (range: 1260 – 3500 cGy). Radiosurgery was delivered during a median number of 5 sessions (range: 2 – 7) on a median isodose line of 75% (range: 68 – 88%) as defined at the margin of the treated tumor. The median homogeneity index (HI), a measure of dose homogeneity to the tumor, was 1.32 (range: 1.11 – 2.44). For the lesions where it was available (28 lesions), the median new conformity index (NCI) was 1.60 (range: 1.29 – 2.59).

The median follow-up was 37 weeks (range: 6 – 238 weeks) (Tables 4 &5). At last follow-up, or at the time of death from systemic disease, 5 tumors (16%) had regressed, and 18 (58%) exhibited stable local disease (Figure 1 and Table 4). Eight lesions (26%) progressed locally despite treatment (Figure 2). The overall tumor control rate in these 31 patients was 74%.

For those patients with local progression, the median time to progression was 24 weeks (range: 5 – 230 weeks). One patient with a renal cell carcinoma metastasis to the right jugular foramen/CPA who experienced local progression at 31 weeks underwent a second course of CKS, which halted further progression and resulted in subsequent local control at a follow-up of 72 weeks.

Ten patients (32%) were alive at the end of the follow-up period, having survived a median of 81 weeks (range: 18 – 238 weeks). For the 21 patients (68%) who died, the median time to death was 25 weeks (range: 6 – 142 weeks) (Tables 4 &5). Among those patients who died, 5 (25%) had local progression. However, no patients died specifically from radiosurgery-treated disease or treatment-related complications. The median progression-free survival of the cohort was 230 weeks (Figure 3). The median overall survival of the cohort was 39 weeks (Figure 4).

The follow-up clinical data were compared between the groups of patients for whom CKS was primary "stand-alone" treatment versus secondary treatment following surgery or external beam radiotherapy. Among the patients with adequate follow-up data, 18 patients were treated with CKS as a primary treatment. The median follow-up was 44 weeks (range: 7 – 238 weeks). Nine patients (50%) were alive at the end of the follow-up period, and 5 (27%) experienced local tumor progression, with a median time to progression of 31 weeks (range: 9 – 230 weeks).

For the 13 patients with previous treatments for their skull base lesion, the median follow-up was 35 weeks (range: 6 – 142 weeks). One patient (8%) was alive at the end of the follow-up period, and 3 (23%) experienced local tumor progression, with a median time to progression of 16 weeks (range: 5 – 32 weeks).

The neurological deficits before and after CKS are summarized in Table 6. Altered vision comprised the most common presenting symptom prior to radiosurgery, with 10 patients having reduced visual acuity, 13 patients having diplopia, and 1 patient having proptosis. Four patients (40%) experienced improved visual acuity and three patients (23%) experienced improvement from their diplopia following treatment. Otherwise, all symptoms remained stable at last follow-up. Of the 17 patients with facial weakness or facial pain on physical examination prior to CKS, 15 (88%) remained stable at last follow-up. One patient (6%) with facial weakness reported improvement. In one patient, facial weakness and swallowing difficulty worsened following CKS due to local disease progression involving all cranial nerves. Swallowing difficulties were found in four patients, 75% of which remained stable following treatment (Figure 5). In the absence of tumor progression, there were no cranial nerve, brainstem or vascular complications referable specifically to CyberKnife^® radiosurgery. Specifically, there were no new cranial nerve deficits observed following SRS in this series.

Radiosurgery may be uniquely suitable for treating these tumors, since it is non-invasive and can precisely target the tumor with minimal spread of radiation to surrounding normal neurological structures. Various investigators have reported their experience with stereotactic radiosurgery in the treatment of malignant skull base tumors. Cmelak et al. reported their data on 47 patients with 59 malignant skull base tumors [54]. Eleven patients with primary nasopharyngeal carcinoma were treated with Linac radiosurgery as a boost (7 – 16 Gy, median: 12 Gy) after a course of fractionated radiotherapy. None of the eleven had tumor progression during the follow-up period. The rest of the patients were treated for skull base metastases or local recurrences from primary head and neck cancers. Radiation doses of 7.0 Gy – 35.0 Gy (median 20.0 Gy) were delivered to these lesions, usually as a single fraction. A tumor control rate of 69% was reported for these patients during the study period (median: 36 weeks). Major toxicities occurred after 5 of 59 treatments. These included three cranial nerve palsies, one CSF leak, and one case of trismus. An important conclusion from their data was that local control did not correlate with lesion size, histology, or radiosurgical dose.

Two small studies from Japan showed similar results. Tanaka et al. reported on 19 malignant skull base tumors, which they treated with single fraction gamma knife radiosurgery [33]. The mean marginal dose utilized was 12.9 Gy. During a follow-up period of 22 months, a tumor control rate of 68% was recorded. The other study by Iwai and Yamanaka of 18 similar patients showed a tumor control rate of 67% during a median follow up of 10 months [31]. A local control rate as high as 95% at 2 years has been reported in one radiosurgery study, but the patient population in that series included 66% with skull base chordomas, chondrosarcomas and adenoid cystic carcinomas, which differ significantly from the cancer patient population studied in the other cited series and our own [55].

In the attempt to bring some order to a heterogenous group of skull base tumors, Morita et al. recently classified cranial base tumors by the degree of aggressiveness into benign, intermediate malignant (or low grade/slow growing), and highly malignant (or fast growing) [56]. Applying this strategy to our series, 31 tumors in our series (84%) would be classified as "highly malignant" or fast growing. Despite this unfavorable bias in our population, the tumor control rate in our series compared favorably to the rate reported in the literature [3, 31, 33, 54, 57]. We treated 31 malignant skull base tumors with a median marginal dose of 2500 cGy delivered in 2–7 sessions (median of 5) and achieved a local control rate of 74% during the follow-up period (median 37 weeks). The median progression-free survival was 230 weeks. In separate analysis of the patients with tumors classified as "highly malignant", the local control rate in this sub-group of patients did not differ significantly from the total study population (74% at 40 weeks), confirming the reported finding on metastatic tumors that response to radiosurgery may be independent of tumor characteristics [15]. Similarly, a comparison of patients who received radiosurgery as primary treatment versus adjunct treatment after surgery or radiotherapy did not reveal major differences in outcome.

Neurological deterioration occurred only in a minority of our patients and in each case, it was accompanied by local tumor progression. Neurological symptoms remained stable or improved in 94% of the patients. No neurological deficits were attributable to toxicity of radiosurgery. Although it is possible that a higher complication rate will emerge with longer follow-up, we believe that the lack of morbidity is largely the result of delivering radiosurgery in multiple sessions, with high conformality and homogeneity. Fractionation is a cornerstone principle in radiation oncology. The oncologist uses it to exploit the significantly different response to radiation of normal versus neoplastic tissue, for the protection of the former and ablation of the latter. It provides time for normal tissue repair between doses, and theoretically minimizes radiation toxicity. With the advent of frameless, image-guided radiosurgery, "hypofractionation" or multi-session treatment became possible. Adler et al. reported on their experience on multi-session radiosurgery for treating skull base, benign tumors situated within 2 mm of the optic apparatus. They achieved a high tumor control rate and found that 94% of the patients had stable or improved vision after treatment [46]. The authors believed that staging the treatment significantly contributed to the low incidence of radiosurgical toxicity. In addition to protective effects, the staging of radiosurgical treatments may have heretofore under-recognized tumor control benefits as well. A new report from Canada showed that patients who received staged radiosurgery to their brain metastases survived longer that those who received single-session treatment [58]. It is possible, that by allowing for a higher total dose delivery to the tumor, staging may lead to better tumor control.

A recent report out of our institution demonstrated that the CyberKnife^® radiosurgical system is capable of delivering a high dose of radiation to a well-defined clinical target volume with high conformity (median NCI 1.66) and homogeneity (median HI 1.26), regardless of irregular tumor shape, large tumor volume, or proximity to critical structures [59]. The median NCI in the present series was 1.60, and the median HI was 1.32. Although still controversial, it is our opinion that improved conformity and homogeneity may maintain high rates of local control while decreasing radiation-induced complications [53, 59‐61]. It seems intuitively evident that conformality and homogeneity are important in treating malignancies of the skull base, since all the tumors are in close proximity to, or entirely surround critical neurological structures that have limited radiation tolerance. In many instances, the encircled cranial nerve is not visible on the treatment-planning image, and one must assume that it received the maximum dose.

A significant majority of the patients in the present study received a does of 2500 cGy in 5 stages. The initial selection of the dose and staging regimen stemmed from our group's experience using the CyberKnife^® radiosurgical system to treat benign skull base lesions. Having encountered no neurological morbidity attributable to radiosurgery in this study, it is impossible to tell whether current treatment regimen represent the "ideal" dose to malignant skull base tumors. A higher average dose may lead to a better tumor control rate than the 74% seen in the present series, and still achieve an acceptably low rate of complications. It is also possible that the "ideal" dosing and staging is different for each patient, dependent on histopathology, previous treatments, tumor volume, neurological status and systemic tumor burden. Our confidence in raising the treatment dose, like the "true" complication rate, will no doubt come with time and further experience with these difficult tumors.