A Multi-institutional Validation Study of a New Prognostic Index (Graded Prognostic Assessment, GPA) for Patients with Brain Metastases

Validation of Nine Different Prognostic Grading Indexes for Radiosurgery of Brain Metastases in Breast Cancer Patients and Development of an All-Encompassing Prognostic Tool

Purpose: Several prognostic indexes for overall survival (OS) after radiotherapy of brain metastases in breast cancer patients exist but are mainly validated for whole-brain radiotherapy or not specifically for breast cancer patients. To date, no such index provides information beyond mere OS.

Methods: We retrospectively analyzed 95 breast cancer patients treated with stereotactic radiosurgery for 203 brain metastases. The Kaplan–Meier method with log-rank test was used to assess OS, local control (LC), distant cranial control (DCC), and extracranial control (EC). Cox regression was applied to detect prognostic outcome factors. A point scoring system was designed to stratify patients based on outcome. Nine established prognostic indexes were analyzed using the concordance index (c-index).

Results: Two out of nine analyzed prognostic indexes for OS showed a significant c-index, the breast graded prognostic assessment (bGPA; 0.631; 95% CI, 0.514–0.748; p = 0.037) and the modified bGPA (mod-bGPA; 0.662; 95% CI, 0.547–0.777; p = 0.010). Significant results from multivariate analysis (Karnofsky Performance Score, Her2/neu receptor status, extracranial control) were used to generate a new point system: the breast cancer stereotactic radiotherapy score (bSRS), which discriminated three significantly different prognostic groups, for LC, DCC, EC, and OS, respectively. However, the c-index was only significant for OS (0.689; 95% CI, 0.577–0.802; p = 0.003).

Conclusions: The new bSRS score was superior to the bGPA and mod-bGPA scores for prognosis of OS. The bSRS is easy to use and the first tool, which might also provide outcome assessment beyond mere OS. Future studies need to validate these findings.

Deep learning-based survival analysis for brain metastasis patients with the national cancer database

PURPOSE

Prognostic indices such as the Brain Metastasis Graded Prognostic Assessment have been used in clinical settings to aid physicians and patients in determining an appropriate treatment regimen. These indices are derivative of traditional survival analysis techniques such as Cox proportional hazards (CPH) and recursive partitioning analysis (RPA). Previous studies have shown that by evaluating CPH risk with a nonlinear deep neural network, DeepSurv, patient survival can be modeled more accurately. In this work, we apply DeepSurv to a test case: breast cancer patients with brain metastases who have received stereotactic radiosurgery.

METHODS

Survival times, censorship status, and 27 covariates including age, staging information, and hormone receptor status were provided for 1673 patients by the NCDB. Monte Carlo cross-validation with 50 samples of 1400 patients was used to train and validate the DeepSurv, CPH, and RPA models independently. DeepSurv was implemented with L2 regularization, batch normalization, dropout, Nesterov momentum, and learning rate decay. RPA was implemented as a random survival forest (RSF). Concordance indices of test sets of 140 patients were used for each sample to assess the generalizable predictive capacity of each model.

RESULTS

Following hyperparameter tuning, DeepSurv was trained at 32 min per sample on a 1.33 GHz quad-core CPU. Test set concordance indices of 0.7488 ± 0.0049, 0.6251 ± 0.0047, and 0.7368 ± 0.0047, were found for DeepSurv, CPH, and RSF, respectively. A Tukey HSD test demonstrates a statistically significant difference between the mean concordance indices of the three models.

CONCLUSION

Our results suggest that deep learning-based survival prediction can outperform traditional models, specifically in a case where an accurate prognosis is highly clinically relevant. We recommend that where appropriate data are available, deep learning-based prognostic indicators should be used to supplement classical statistics.

Stereotactic radiosurgery in the management of brain metastases from primary thyroid cancers

Patients with metastatic well-differentiated thyroid cancer have a generally favorable long-term outcome although multi-organ involvement is a known marker of poor prognosis. Brain metastases are rare, occurring in less than 1% of patients with thyroid cancer. Few patients have been managed with stereotactic radiosurgery (SRS). A retrospective database of 5,067 patients treated for brain metastases between 1985 and 2007 was generated from 11 institutions. Thyroid cancer patients were identified in this database and, when possible, additional information was obtained from further chart review. Patients were excluded if they had incomplete treatment or follow-up information. Two validated prognostic indices, Graded prognostic Assessment (GPA) and Recursive Partitioning Analysis (RPA), were calculated for each patient. The overall survival times were calculated by the Kaplan-Meier method. Twenty-three thyroid cancer patients were identified (51% male, 48% female). Median age was 63 years (range 20-81). Pathology of the primary thyroid disease was available for twelve patients; the majority were diagnosed with differentiated thyroid cancer (n = 9 papillary, n = 2 Hürthle cell; 92%) and one had medullary subtype (8%). Median time from diagnosis of primary disease to brain metastasis was 41.8 months (range 0-516). Fifteen (65%) patients underwent SRS as part of their initial treatment with a median number of lesions treated of 1.5 (range 1-9). The median follow-up time for living patients was 35.2 months. Overall median survival time was 20.8 months (40% alive at last follow-up) and 37.4 months for SRS-treated patients (P = NS). A poor Karnofsky performance status was predictive of worse outcome (P = 0.001). GPA and RPA did not provide additional prognostic information. In conclusion, patients treated with SRS for brain metastases from primary thyroid cancer have a favorable prognosis with an expected median survival greater than 3 years. It is unclear as to whether current prognostic indices are relevant to this patient population.

Diagnosis-specific prognostic factors, indexes, and treatment outcomes for patients with newly diagnosed brain metastases: a multi-institutional analysis of 4,259 patients

PURPOSE

Controversy endures regarding the optimal treatment of patients with brain metastases (BMs). Debate persists, despite many randomized trials, perhaps because BM patients are a heterogeneous population. The purpose of the present study was to identify significant diagnosis-specific prognostic factors and indexes (Diagnosis-Specific Graded Prognostic Assessment [DS-GPA]).

METHODS AND MATERIALS

A retrospective database of 5,067 patients treated for BMs between 1985 and 2007 was generated from 11 institutions. After exclusion of the patients with recurrent BMs or incomplete data, 4,259 patients with newly diagnosed BMs remained eligible for analysis. Univariate and multivariate analyses of the prognostic factors and outcomes by primary site and treatment were performed. The significant prognostic factors were determined and used to define the DS-GPA prognostic indexes. The DS-GPA scores were calculated and correlated with the outcomes, stratified by diagnosis and treatment.

RESULTS

The significant prognostic factors varied by diagnosis. For non-small-cell lung cancer and small-cell lung cancer, the significant prognostic factors were Karnofsky performance status, age, presence of extracranial metastases, and number of BMs, confirming the original GPA for these diagnoses. For melanoma and renal cell cancer, the significant prognostic factors were Karnofsky performance status and the number of BMs. For breast and gastrointestinal cancer, the only significant prognostic factor was the Karnofsky performance status. Two new DS-GPA indexes were thus designed for breast/gastrointestinal cancer and melanoma/renal cell carcinoma. The median survival by GPA score, diagnosis, and treatment were determined.

CONCLUSION

The prognostic factors for BM patients varied by diagnosis. The original GPA was confirmed for non-small-cell lung cancer and small-cell lung cancer. New DS-GPA indexes were determined for other histologic types and correlated with the outcome, and statistical separation between the groups was confirmed. These data should be considered in the design of future randomized trials and in clinical decision-making.