12 Gy gamma knife radiosurgical volume is a predictor for radiation necrosis in non-AVM intracranial tumors

Discussion

Despite the extensive use of radiosurgery for the treatment of many intracranial tumors, relatively little data have been published regarding its primary late toxicity, radiation necrosis. Of the published toxicity data, the majority has been acquired from patients treated for AVMs with few studies investigating patients with non-AVM tumors. In addition, most have reported only risks of symptomatic postradiosurgical imaging changes with little mention of asymptomatic postradiosurgical imaging changes. Our study adds significantly to the previously published radiosurgical toxicity data as it describes factors associated with risk of asymptomatic and symptomatic postradiosurgical changes suggestive of radiation necrosis in the non-AVM population.

Radiosurgical dose and treatment volume have been demonstrated in multiple reports to correlate with the development of postradiosurgical imaging changes suggestive of radiation necrosis. Some of the earliest estimations of toxicity from radiosurgical treatment of AVMs came from proton beam work by Kjellberg et al., who established a 1% isoeffect curve for symptomatic radiation necrosis based on proton beam diameter and radiation dose . Steiner et al. at the Karolinska Hospital reported the incidence of postradiosurgery toxicity in 239 patients with AVMs ranging from 0.2 to 35 cc treated to 20–25 Gy at the 50% isodose curve using early gamma knife technology. They found a 3% incidence of “untoward effects” in these patients with a latency of 3–8 months and imaging changes showing a combination of increased edema, white matter changes, and contrast enhancement consistent with what is often now described as postradiosurgical imaging changes suggestive of radiation necrosis. In a later update, the Karolinska group found that AVM location was an important factor influencing postradiosurgery symptomatic necrosis with central lesions having a higher risk compared with peripheral lesions.

Flickinger et al. at the University of Pittsburgh reported initial toxicity data on 72 AVM patients receiving GKSRS . They observed that 31% of the patients had postradiosurgical imaging changes consistent with radiation necrosis at 2 years post-GKSRS with a median time of onset of 12 months. The observed radiation necrosis imaging changes were asymptomatic in 55% of patients, bringing the absolute actuarial risk of symptomatic radiation necrosis to 14%. In this early report, the only factor that significantly correlated with risk of necrotic imaging changes was radiosurgical volume. In a later update of their patient cohort, the Pittsburgh group performed risk modeling analysis of 307 patients with AVMs treated with GKSRS. They investigated several factors including 8, 10, and 12 Gy radiosurgical volume, AVM nidus volume, marginal radiosurgical volume (radiosurgical volume − AVM nidus volume), number of isocenters, minimum AVM nidus dose, maximum dose, and target dose inhomogeneity. In their multivariate analysis, only the 12-GyV correlated with the development of post-GKSRS necrosis imaging changes (symptomatic or asymptomatic). The group went on to attempt to correlate AVM location with risk of post-GKSRS necrosis imaging changes. They created a postradiosurgery injury expression score based on AVM location and 12-GyV. This score did not correlate with asymptomatic post-GKSRS necrosis imaging changes, but did correlate with the risk of symptomatic necrosis

In a toxicity study of patients with AVMs and skull base tumors treated with linear accelerator (linac)-based radiosurgery, Voges  found that the only variable that correlated significantly with risk of symptomatic radiation necrosis in patients treated with AVMs was the volume of AVM and surrounding brain receiving 10 Gy (10-GyV). The 10 Gy dose was tested based on animal models and personal experience cited by the authors. A 12-GyV was not tested. It is possible that the 10-GyV proved significant by the Voges group but not significant in the Pittsburgh group studies because of the much more conventional, smaller sized lesions treated by the Pittsburgh group (median volume, 4.3 cc vs. 12.4 cc).

Our results also support a correlation between the 12-GyV and the risk of S-NEC with a steep increase in risk with 12-GyV >10 cc of 55.3% vs. 22.5% for 12-GyV <10 cc (p = 0.007 in multivariate analysis). However, we found that the risk of A-NEC for all lesions was relatively constant and did not increase with 12-GyV, remaining at 19.1% for tumors with 12-GyV >10 cc compared with 18.5% for tumors with 12-GyV <10 cc. In multivariate analysis, we also found a differential S-NEC risk based on tumor location, with occipital and temporal locations having a higher risk of S-NEC than other locations.

Our observation of a higher risk of S-NEC in men treated for non-AVM tumors was not observed by the Pittsburgh group in patients treated for AVMs. This may indicate a difference in the risk of S-NEC of tumors compared with AVMs, but is more likely a result of sampling bias in this retrospective trial.

Dose inhomogeneity and treatment plan conformality have been thought by some groups to correlate with the development of symptomatic postradiosurgical imaging changes suggestive of radiation necrosis. One of the largest reports of late radiosurgical toxicity for patients treated with radiosurgery who had previously received fractionated radiotherapy is Radiation Therapy Oncology Group trial 90-05. In this trial, it was found in multivariate analysis that the two variables associated with symptomatic radiation necrosis were target volume >8.2 cc and dose inhomogeneity (maximum dose/prescribed dose) ≥2. The majority of the patients enrolled in this study were treated with linac radiosurgery where dose inhomogeneity is generally between 1.1 and 1.2 vs. gamma knife radiosurgery where it is typically 2. In another linac-based radiosurgery study, dose inhomogeneity (defined as maximum tumor dose − minimum tumor dose) >10 cc was associated with an increased risk of neurologic toxicity, but this was not correlated with imaging findings . A French study  also found symptomatic radiation necrosis to be correlated with conformality of the treatment planning for linac radiosurgery. In two gamma knife complication studies, from the Pittsburgh group and the University of California–San Francisco group, neither conformity nor dose inhomogeneity correlated with post-GKSRS symptomatic radiation necrosis. Our data agree with the Pittsburgh and University of California–San Francisco GKSRS studies, in that conformality did not correlate with either A-NEC or S-NEC. One possible explanation of this is improved conformality for larger volumes with gamma knife SRS compared with linac SRS as has been previously suggested.

Conclusion

We found that the risk of developing S-NEC, but not A-NEC after gamma knife radiosurgery for non-AVM tumors correlates with 12-GyV. The risk of S-NEC increased significantly for 12-GyV >10 cc, regardless of plan conformality. These observations add to the knowledge base of postradiosurgical toxicity and can be used when planning and optimizing gamma knife radiosurgical treatments.