Dose-Response Modeling of the Visual Pathway Tolerance to Single-Fraction and Hypofractionated Stereotactic Radiosurgery

Patients with tumors adjacent to the optic nerves and chiasm are frequently not candidates for single-fraction stereotactic radiosurgery (SRS) due to concern for radiation-induced optic neuropathy. However, these patients have been successfully treated with hypofractionated SRS over 2-5 days, though dose constraints have not yet been well defined. We reviewed the literature on optic tolerance to radiation and constructed a dose-response model for visual pathway tolerance to SRS delivered in 1-5 fractions. We analyzed optic nerve and chiasm dose-volume histogram (DVH) data from perioptic tumors, defined as those within 3mm of the optic nerves or chiasm, treated with SRS from 2000-2013 at our institution. Tumors with subsequent local progression were excluded from the primary analysis of vision outcome. A total of 262 evaluable cases (26 with malignant and 236 with benign tumors) with visual field and clinical outcomes were analyzed. Median patient follow-up was 37 months (range: 2-142 months). The median number of fractions was 3 (1 fraction n = 47, 2 fraction n = 28, 3 fraction n = 111, 4 fraction n = 10, and 5 fraction n = 66); doses were converted to 3-fraction equivalent doses with the linear quadratic model using α/β = 2Gy prior to modeling. Optic structure dose parameters analyzed included Dmin, Dmedian, Dmean, Dmax, V30Gy, V25Gy, V20Gy, V15Gy, V10Gy, V5Gy, D50%, D10%, D5%, D1%, D1cc, D0.50cc, D0.25cc, D0.20cc, D0.10cc, D0.05cc, D0.03cc. From the plan DVHs, a maximum-likelihood parameter fitting of the probit dose-response model was performed using DVH Evaluator software. The 68% CIs, corresponding to one standard deviation, were calculated using the profile likelihood method. Of the 262 analyzed, 2 (0.8%) patients experienced common terminology criteria for adverse events grade 4 vision loss in one eye, defined as vision of 20/200 or worse in the affected eye. One of these patients had received 2 previous courses of radiotherapy to the optic structures. Both cases were meningiomas treated with 25Gy in 5 fractions, with a 3-fraction equivalent optic nerve Dmax of 19.2 and 22.2Gy. Fitting these data to a probit dose-response model enabled risk estimates to be made for these previously unvalidated optic pathway constraints: the Dmax limits of 12Gy in 1 fraction from QUANTEC, 19.5Gy in 3 fractions from Timmerman 2008, and 25Gy in 5 fractions from AAPM Task Group 101 all had less than 1% risk. In 262 patients with perioptic tumors treated with SRS, we found a risk of optic complications of less than 1%. These data support previously unvalidated estimates as safe guidelines, which may in fact underestimate the tolerance of the optic structures, particularly in patients without prior radiation. Further investigation would refine the estimated normal tissue complication probability for SRS near the optic apparatus.