Validation of Recursive Partitioning Analysis Classification in Patients with Brain Metastases from Non-small Cell Lung Cancer Treated with Short-course Accelerated Radiotherapy

Abstract 

Aims

To study various prognostic factors affecting outcome and to validate Radiation Therapy Oncology Group recursive partitioning analysis (RPA) class in non-small cell lung cancer (NSCLC) with brain metastases treated with short-course accelerated radiotherapy (SCAR).

Materials and methods

The case records of 100 patients with NSCLC consecutively treated at Tata Memorial Hospital from August 2006 to August 2009 were studied for various patient, tumour and treatment-related prognostic factors. Patients received whole-brain radiotherapy to a dose of 20Gy/five fractions over 1 week (n=90) or 30Gy/10 fractions over 2 weeks (n=10). The Kaplan–Meier estimate was used for survival analysis in SPSS v15.

Results

The median overall survival was 4.0 months (range 0.5–30.0 months). The 6-, 12-, 18- and 24-month survival rates were 35.8, 18.0, 9.3 and 6.2%, respectively. Of the various prognostic factors, RPA class (II versus III, P value=0.023), Karnofsky performance score (<70 versus ≥70, P value=0.039) and the use of systemic therapy (yes versus no, P value=0.00) emerged as significant on univariate analysis. RPA classification effectively separated the patient population into prognostically distinct subgroups. The median overall survival for RPA class II and RPA class III was 6 and 4 months, respectively. The use of systemic therapy prolonged overall survival by 6 months (3 months versus 9 months).

Conclusion

The SCAR regimen is an effective and resource-sparing palliative strategy for brain metastases in NSCLC. The results validate the usefulness of RPA classification in this specific subset of patients treated with SCAR.

Key words: Brain metastases, non-small cell lung cancer, recursive partitioning analysis, short-course radiotherapy