Clinical factors associated with mortality within three months after radiosurgery of asymptomatic brain metastases from non-small cell lung cancer

Abstract

Purpose: Routine brain MRI surveillance frequently diagnoses small, asymptomatic brain metastases from non-small cell lung cancer (NSCLC) that are effectively treated with stereotactic radiosurgery (SRS). A subset of patients, however, may die prior to the onset of symptoms. This study identifies clinical features that distinguish neurologically-asymptomatic NSCLC brain metastases patients that die prior to routine 3 month follow-up after SRS.

Methods: Retrospective chart review from 2007 to 2017 identified 18 patients with neurologically-asymptomatic NSCLC brain metastases who died < 3 months after SRS. Twenty-eight additional patients meeting criteria and surviving > 6 months after SRS were identified. Clinical factors were examined to determine characteristics correlated with survival using cox proportional hazards and nominal logistic regression models. Logistic regression models using salient factors were trained with 10-fold cross-validation and compared to the graded prognostic assessment (GPA) and score index of radiosurgery (SIR) using the AUC from receiver operant characteristic curves.

Results: The median survival following SRS was 1.4 and 9.2 months for the < 3 months and > 6 months groups, respectively. Age, number of brain metastases, and Karnofsky performance status were associated with overall survival while gender and interval between primary cancer and first brain metastasis diagnoses were associated with < 3 months and > 6 months survival, respectively. Models using GPA and SIR performed poorly compared to preliminary metrics generated in this study for prognosis of both < 3 months and > 6 months survival.

Conclusion: Physicians require data to provide high-value, cost-conscious health care. Clinical metrics can screen patients with asymptomatic NSCLC brain metastases likely to die prior to the standard screening interval and observation could be considered.

Keywords: Brain metastasis; Clinical metric; NSCLC; Neurologically asymptomatic; Prognostic; Stereotactic radiosurgery.

Fig. 1

Patient flow chart

Fig. 2

Kaplan–Meier survival for disease time-course of patients that died within 3 months of SRS and those that survived for greater than 6 months. a OS probability of the two groups from time of primary cancer to death or last follow-up, b from time of primary cancer to brain metastasis diagnosis, and c from time of last SRS to death or last follow-up. In (b), 2 and 5 patients in the < 3-months and > 6-months, respectively, were diagnosed with brain metastases at the time of primary cancer diagnosis. CI confidence interval, GPA graded prognostic assessment, SIR stereotactic index of radiosurgery, OS overall survival, SRS stereotactic radiosurgery

Fig. 3

ROC curves for prognostic models were developed by training logistic regression models with 10-fold cross-validation on the entire cohort. The < 3 months and > 6 months models were generated based on factors determined to be prognostic in univariate logistic regression (Table S1 and Table S2). These were compared to logistic regression models based on KPS, GPA and SIR to examine prognostic value for a death within 3 months of SRS and for b survival of at least 6 months after SRS. AUC area under the curve, CI confidence interval, GPA graded prognostic assessment, KPS Karnofsky performance status, SIR stereotactic index of radiosurgery, SRS stereotactic radiosurgery