ESTRO 2024 - Abstract Book

S1984

Clinical - Mixed sites, palliation

ESTRO 2024

In this work we investigate the possibility of establishing statistical models for local control (LC) for stereotactic body radiotherapy (SBRT) of spinal metastases. We evaluated the influence of the number of fractions, prescribed radiation dose, tumor histology and the evaluation time-point.

Material/Methods:

A systematic review and meta-analysis were performed using PRISMA methodology as part of the ESTRO spine SBRT guideline. Publications from January 2005 to September 2021 were included. All references were reviewed for data on LC (also referred to as local progression in some publications) at the 1- and 2-year time points. The current analysis included LC data from articles that used a reasonable method to account for competing risks (e.g. Kaplan Meier estimate or other methods to assess cumulative incidence, where death and lost to follow-up were treated as competing risks). Articles reporting crude rates and those with shorter follow-up that did not include data at the 1- and 2-year time point were excluded. The median (most common) prescribed dose and fractionation were recorded for each study. These were converted to single-fraction equivalent dose using alpha/beta=10. Some authors reported LC for a pooled data set with various prescription doses, in which case the range of the prescription doses was recorded. Absolute dose was used and differences in e.g. prescription isodose or normalized PTV coverage were not considered in this analysis. Whether or not the cohort included patients receiving SBRT as re-irradiation was recorded. Data on the histology of the primary tumor and whether or not the cohort was “radiation - resistant” only (e.g. renal cell carcinoma [RCC] and hepatocellular carcinoma), or “mixed” was evaluated qualitatively. Mixed histology may or may not include also tumors with radiation-resistant histology. In order to assess the impact of each data set, we collected the number of lesions and the median follow-up time for each article. If the number of patients at risk at the 1- and 2-year time points was not stated, it was estimated based on the number of lesions and median follow-up time.

Results:

A total of 69 studies were identified in the above timeframe, and two important recent studies were later included in the analysis (Ryu et al 2023, and Zeng et al 2023 which replaced Tseng et al 2018). From these 71 studies, a total of 33 finally contributed 39 LC data-points. Of the 39 LC data- points, 10 were labeled as “radiation - resistant” and the rest as “mixed”. In total, data from 3747 patients and 5093 lesions were extracted. The median number of treatment fractions were 2 (range: 1-10). Major contributions to the models came from the Zeng et al (Sunnybrook) and Yamada et al (MSKCC) data sets. Logistic regression showed that increased median prescription dose was associated with higher LC at both 1- and 2-years (p<0.001); see Table 1.

Table 1. Table with dose-response data for LC models. Histology TIME-POINT (years)

LC(90%) dose

D50 (Gy)

Gamma50

Mixed

1

18.1

12.22

1.14

Mixed

2

19.8

13.74

1.25

Radio-resistant

1

21.5

15.91

1.55

Whether or not parts of the cohort received SBRT as a re-irradiation treatment had no significant effect. Radiation resistant histology was associated with a markedly reduced LC as compared to the data presented on patients with