Patterns of Failure seen in the SBRT Treatment of Paraspinal Disease Michael Lovelock Ph.D Medical Physics Department, Memorial Sloan-Kettering Cancer Center, New York City, NY Acknowledgements • • • • • • Josh Yamada MD – Radiation Oncology Mark Bilsky MD - Neurosurgery Eric Lis MD - Neuroradiology Tom Losasso PhD – Physics Jennifer Keam MD - Radiation Oncology Joan Zatcky N.P. – Radiation Oncology Treatment of Paraspinal Disease at MSKCC • Single or hypo-fractionated radiotherapy has been given to patients with metastatic disease to the spine or paraspinal sites since April 2000. • Two patients cohorts: 1. Single or oligo-metastatic patient with no prior radiation 2. Patients who have been previously treated and have experienced local failure Cohort #1: • Patients with no prior radiation to the site, and with either a single metastasis or oligo-metastatic condition • Initially prescription dose 18 Gy, (maximum cord point dose 12 Gy). • Prescription dose now 24 Gy (maximum cord point dose 14 Gy) Cohort # 2: • Patients who have had prior treatment to the site, and have experienced local failure • Initially, patients received 20 Gy in 5 fractions • (more recently, Not part of this analysis, this has been changed to 30 Gy in 5 frac._ • Maximum dose to the myelogram defined cord is 2.8 Gy/.frac Local Control – Cohort 1 –single frac A significant difference in dose response is seen between 18-23 Gy and 24 Gy prescription Under-dosed sub-volume 24 Gy 18 - 23 Gy Yamada et al, I.J.R.O.B.P 2008 71(2) p. 484-90. Local Control – Cohort 2 –hypo-frac Wright et al, Am J Clin Oncol. 2006 29(5) 495-502 Objective: • Examine the patterns of failures seen in the two patient cohorts to see what can be learned about: – Prescribed dose sufficiency – Margins / geometrical miss – Cold spots due to proximity of spinal cord Delivery Technique • Treatment Planning: – All plans developed using a inverse-planning technique – Dose delivered using ‘sliding window’ intensity-modulation • Delivery – Image guidance (formerly MV + implanted markers, now kv radiographs or cone beam without implants) used for all fractions Immobilization Cradle • Designed to maximize patient comfort • 4 lateral paddles help maintain patient position Real time monitoring of patient position Stereoscopic infra-red camera Marker Displacement (cm) • Infra-red reflectors taped to patient’s skin • Positions monitored in real time to check patient still and breathing regularly Marker Locations - Left chest - Right chest - Belly Time (seconds) Characteristics of the Recurrence Volumes • Recurrence was identified by a radiologist on followup (every 3-4 months) MR scans • Radiologist delineated the MR volume of failure • After registration, the volumes were transferred to the planning CT, permitting the spatial and dosimetric characteristics to be evaluated. Failure volumes of the 5 fraction patient cohort 1 2 4 5 3 6 Recurrence GTV PTV Recurrence volumes – 5 frac -continued 7 8 9 10 11 12 Recurrence volumes 5-frac -continued 13 Recurrence volumes singlefraction treatment cohort 1 4 2 5 3 6 Recurrence volumes – single fraction cohort - continued 7 Characteristics of Failure Volumes 5 Fraction Cohort 1 Fraction Cohort • Volumes Location with respect to the GTV Generally large, covering the vertebral body Generally at the edge of the GTV, smaller Volume Median: 67.1 cc Range: 14.6 – 341 cc Median: 10.3 cc Range: 2.9 – 52.1 cc Dosimetric Characteristics of the Failure Volumes Failure Volume • Was the existence of a cold spot in the GTV associated GTV with the failure volume – Did the failure volume encompass the cold spot in the GTV? – compare the mean dose of the GTV with the mean dose of the GTV – failure volume overlap Failure Volume GTV Mean Dose in GTV – Failure volume overlap (cGy) Mean Dose in the GTV (cGy) Difference: Overlap – GTV (cGy) 1 fraction cohort 2102 2280 -178 5 fraction cohort 2304 2327 -23 One interpretation • For the single fraction treatments, there is evidence that the cold spots in the target are associated with recurrence • This is not observed in the 5-fraction patients. The analysis is confounded by the large size of these failure volumes. • This may indicate that the 5-fraction doses are simply too low, resulting in many surviving clonogens. This leads to rapid tumor regrowth, not necessarily associated with a GTV cold spot Cold spot exists because of the proximity of the spinal cord The maximum dose to any point on the spinal cord is limited to 12 – 14 Gy Under-dosed sub-volume The cord, visualized with either a co-registered MR scan, or a with a myelogram, may be in close proximity to the target volume The steepest dose gradient achievable with IMRT is around 10% per mm. Cord Tumor (gross target volume) Prescription isodose This may lead to a cold spot if the target is within 3-4 mm of the cord The dependence of local failure on dose insufficiency with the tumor was investigated by comparing treatment plans of patients with long term local control with those who experienced local failure Cold Spot Analysis • The minimum doses received by the hottest 95%, 98%, and 100% (D95, D98, and Dmin) of the gross target volume (GTV) were computed for 91 consecutively treated lesions seen in 79 patients • A Wilcoxan rank-sum statistic was used to assess difference between local failure and local control. Dmin, D98, and D95 • Dose distributions computed by calculating the dose at 5000 points placed in the target quasirandomly • In paraspinal targets with an adjacent region with a steep dose gradient, Dmin, a single point dose, may be affected by small clinically insignificant changes, such as a 1 pixel shift in a contour • D98 and D95 are more stable measures of low dose, although less sensitive to the presence of a small region of under-dose. Results 1:renal 2:GI 3:melanoma 4:prostate 5:headandneck 6:sarcoma 7:breast 8:bladder 9:lung 10:leydig Target (GTV) Minimum Dose 3000 Dose (cG y) 2500 2000 1500 1000 500 0 2 4 6 8 10 12 Histology • With a 15 month median followup, 7 local failures have occurred P-Values Distributions of D95, D98, and Dmin for the Gross Target Volumes of treatments that resulted in local failure were found to be statistically different from the treatments that resulted in local control Dmin Histology Ignored 0.005 Histology accounted for 0.012 D98 0.012 0.012 D95 0.044 0.040 Correlation between local failure and GTV volume Local Failure as a Function of GTV Volume and Vertebral Position 120 100 GTV Volume (cc) Local control 80 Local failure 60 40 20 0 1 6 11 16 21 Vertebra Number: Sup --> Inf With only 7 local failures, no correlation was seen with GTV volume 26 Conclusion • In both the 20 Gy 5-fraction and 24 Gy single fraction treatments, insufficient dose may be a factor in the observed local failures • For the 20 Gy 5-fraction treatments, the relatively high local failure rate, the large tumor recurrence region, and the lack of correlation with cold spots on the edge of the GTV may indicate the dose is too low • For the 24 Gy single fraction treatments, measures of tumor dose insufficiency that indicate the presence of a cold spot such as D95, D98, and Dmin may be important risk factors for local failure No local failures in any histology were observed when Dmin > 15 Gy suggesting this metric may be a predictor of local control • Confirmation of these results awaits the accrual and followup of more patients Additional Slides D98 1:renal 2:GI 3:melanoma 4:prostate 5:headandneck 6:sarcoma 7:breast 8:bladder 9:lung 10:leydig GTV D98 2800 2600 2400 2200 2000 1800 1600 1400 1200 0 2 4 6 8 10 12 D95 1:renal 2:GI 3:melanoma 4:prostate 5:headandneck 6:sarcoma 7:breast 8:bladder 9:lung 10:leydig GTV D95 2800 2600 2400 2200 2000 1800 1600 1400 0 2 4 6 8 10 12