Recursive Partitioning Model to Identify Early Progression in Patients with Extracranial Oligometastatic Colorectal Cancer Treated with Stereotactic Body Radiotherapy

Purpose/Objective(s)

Metastases directed therapy of oligometastatic colorectal cancer (CRC) may confer long-term disease advantages. However, there is a paucity of data to predict optimal patient selection. Our hypothesis is that pretreatment factors can identify patients at higher risk of early progression after SBRT for CRC oligometastases.

Materials/Methods

An institutional cohort was reviewed to identify patients with extracranial oligometastatic (≤ 5 metastases) colorectal cancer who received SBRT to all metastases. Outcomes of interest were local failure (LF), overall survival (OS) and progression-free survival (PFS). Recursive partitioning analysis (RPA) was performed to identify "early progression", defined as those having a PFS event within 6 months of SBRT, from a training set consisting of 70% of the cohort. The remaining 30% were used in the validation analysis of the model.

Results

From January 2008 to May 2021, 254 patients were identified with 408 lesions irradiated. Median age was 75 years (IQR: 62-82), median follow-up was 30.5 months (IQR: 17.7-48.8), and median time from diagnosis of cancer to metastatic disease was 12.0 months (IQR: 0.2-26.1). The majority of patients had a single metastasis (n=167, 65.8%). The 12-, 24- and 36-month cumulative incidence of LF was 10.3%, 22.8% and 27.5%, respectively. On multivariable analysis (MVA), liver metastasis (vs. non-liver, HR 1.99, 95%CI: 1.32-3.00, p<0.01) and lower mean PTV dose BED₁₀ (HR: 1.01, 95%CI: 1.01-1.02, p<0.01) predicted for higher LF. Median OS was 51.1 months (95%CI: 46.2-56.3) and 12-, 24- and 36-month OS was 94.0%, 79.5% and 64.9%, respectively. On MVA, higher pre-SBRT CEA (HR: 1.21, 95%CI: 1.01-1.45, p=0.04) and larger PTV volume (HR: 1.61, 95%CI: 1.31-1.97, p<0.01) were significant predictors of worse OS. Median PFS was 11.8 months (95%CI: 9.9-13.7) and 6-, 12- and 24-month PFS was 73.2%, 49.2% and 30.4%, respectively. Of the 254 patients included, 68 (26.8%) had "early progression". The RPA model ranked the following variables in order of importance: presence or history of any liver metastases, pre-SBRT CEA and number of lesions treated at time of SBRT. Four terminal nodes (i.e., groups) were generated, classifying the probability of "early progression" with a training and validation receiver operating characteristic curve (AUC) of 0.72 and 0.71, respectively. The two groups at highest risk of early progression were: 1) history or presence of any liver metastases and with ≥2 lesions treated; and 2) history or presence of any liver metastases with 1 lesion treated at the time of SBRT, and pre-SBRT CEA ≥8 ng/mL.

Conclusion

History or presence of liver metastases, higher number of lesions, and higher CEA levels can aid in the identification of oligometastatic patients who progress soon after SBRT. Such patients can be considered for alternative treatment strategies.