External validity of two nomograms for predicting distant brain failure after radiosurgery for brain metastases in a bi-institutional independent patient cohort.

J Neurooncol 2018 Mar 7;137(1):147-154. Epub 2017 Dec 7.

Southeast Radiation Oncology Group, Charlotte, NC, USA.

Patients treated with stereotactic radiosurgery (SRS) for brain metastases (BM) are at increased risk of distant brain failure (DBF). Two nomograms have been recently published to predict individualized risk of DBF after SRS. The goal of this study was to assess the external validity of these nomograms in an independent patient cohort. The records of consecutive patients with BM treated with SRS at Levine Cancer Institute and Emory University between 2005 and 2013 were reviewed. Three validation cohorts were generated based on the specific nomogram or recursive partitioning analysis (RPA) entry criteria: Wake Forest nomogram (n = 281), Canadian nomogram (n = 282), and Canadian RPA (n = 303) validation cohorts. Freedom from DBF at 1-year in the Wake Forest study was 30% compared with 50% in the validation cohort. The validation c-index for both the 6-month and 9-month freedom from DBF Wake Forest nomograms was 0.55, indicating poor discrimination ability, and the goodness-of-fit test for both nomograms was highly significant (p < 0.001), indicating poor calibration. The 1-year actuarial DBF in the Canadian nomogram study was 43.9% compared with 50.9% in the validation cohort. The validation c-index for the Canadian 1-year DBF nomogram was 0.56, and the goodness-of-fit test was also highly significant (p < 0.001). The validation accuracy and c-index of the Canadian RPA classification was 53% and 0.61, respectively. The Wake Forest and Canadian nomograms for predicting risk of DBF after SRS were found to have limited predictive ability in an independent bi-institutional validation cohort. These results reinforce the importance of validating predictive models in independent patient cohorts.

Download full-text PDF

Source
http://dx.doi.org/10.1007/s11060-017-2707-2DOI Listing
March 2018
24 Reads

Publication Analysis

Top Keywords

wake forest
12
patient cohort
8
validation cohorts
8
brain metastases
8
independent patient
8
patients treated
8
distant brain
8
brain failure
8
validity nomograms
8
external validity
8
freedom dbf
8
nomograms
5
entry criteria
4
partitioning analysis
4
analysis rpa
4
rpa entry
4
n = 282 canadian
4
canadian nomogram
4
nomogram n = 281
4
forest nomogram
4

References

(Supplied by CrossRef)

H Aoyama et al.
JAMA 2006

EL Chang et al.
Lancet Oncol 2009

M Kocher et al.
J Clin Oncol 2011

PD Brown et al.
JAMA 2016

R Soffietti et al.
J Clin Oncol 2013

DN Ayala-Peacock et al.
Neuro-oncology 2014

G Rodrigues et al.
Radiother Oncol 2014

DG Altman et al.
Stat Med 2000

DG Altman et al.
BMJ 2009

EW Steyerberg et al.
PLoS Med 2013

RH Press et al.
Cancer 2015

Similar Publications