The impact of CyberKnife's prescription isodose percentage on intracranial target planning

To the Editor: Recently, a detailed comparative study regarding intracranial Gamma Knife (GK) vs. CyberKnife (CK) intracranial dosimetry has been published by your Journal.(1) In a group of 15 patients with 26 brain metastases, we showed that CK produced more homogeneous and conformal plans, while the GK plans had sharper peripheral dose falloff in most cases. In the CK plans, by convention, the applied range of prescription isodose percentage (PIP) was 77%–92%, with a median value of 86%. Intrigued by the results, we hypothesized that lowering the PIP in CK planning would improve peripheral dose falloff, without compromising the excellent dosimetric conformality which had previously been achieved. Secondarily, it was expected that, as PIP decreased, the stereotactic radiosurgical (SRS) plans would become less homogeneous as maximum dose within the target increased. We thank you for the opportunity to share with you the additional results that were generated from this investigation. Parts of the methods and materials have been previously described.(1) We compared the relative dosimetric merit of various prescription isodose levels in CK’s MultiPlan (Accuray Inc, Sunnyvale, CA). The same 15-patient series was used for dosimetric planning. For each tumor, the PIP was varied at three levels averaging approximately 50, 65, and 85% (CK50, CK65 and CK85; Table 1). The homogeneity (HI) and gradient (GI) indices, modified conformity index (mCI, the ratio of the prescription isodose volume to the tumor volume receiving at least the prescription dose), and an MPS-defined quantity called “new CI” (nCI, the ratio of mCI to target coverage, also the inverse of van’t Riet’s Conformation Number) were computed. For peripheral dose falloff, GI50 was calculated as the ratio of the volume enclosed by the isodose at 50% of the prescription dose level to the volume enclosed by the original prescription isodose. GI25, GI40, GI60, and GI80 were calculated in a similar manner. Statistical analyses were performed using analysis of variance (ANOVA) and nonparametric Kruskal-Wallis tests. We found that the mean tumor volume was 4.4 cm3; a median dose of 18 Gy was prescribed. For CK50, CK65, and CK85 series, the coverage was maintained at 96%–100% in all cases. Optimized plans in each scenario across various PIPs were computed, and dosimetric constraints of critical organ structures were all met. Minimum, average, maximum doses, HI, nCI, GI25, GI50, and MU were reported (Table 2). Comparing across the CK50, CK65, and CK85 series, the median mCIs were: 1.48, 1.36, and 1.52, p =.086, respectively. The remaining gradient indices were: GI40 (5.8, 6.9, and 7.6, p = 0.0008); for GI60 (2.8, 3.4, and 3.8), GI80 (1.6, 1.9, and 2.2), and GI90 (1.3, 1.4, and 1.6), p < 0.00001 in all cases. In our study, as expected, the selection of a PIP had a statistically significant impact on HI, mean, and maximum doses. By both mCI and nCI, CK65 produced the most conformal plans which nearly reached statistical significance. Importantly, CK50 had significantly sharpest dose falloff at all gradient index levels, with the exception of GI25. However, dosimetric plans prescribing to CK50 required significantly longer treatment times by MU estimation. In current clinical practices, deciding a prescription isodose level in CK varies by individual plan and preference, and clearly no consensus exists. The CK is a relatively new modality for SRS, which is stereotactically capable for extracranial indications, as well. For GK and linacbased intracranial SRS, it is common to prescribe to 40%–60% and 80%–95% of PIP, respectively. It is then customarily believed that the PIPs of CK should fall between those of the GK and linac-based SRS plans, as CK shares features of both. For CK dosimetric planning, some JOURNAL OF APPLIED CLINICAL MEDICAL PHYSICS, VOLUME 15, NUMBER 5, 2014