1541. Development and Validation of Clinical Prediction Models of Survival and Clinical Outcomes for Patients with Metastatic Epidural Spinal Disease: A Systematic Review
Authors: Anick Nater-Goulet, MD; Jetan Badhiwala; James Hong; So Kato; Melanie Anderson; David Choi; Michael Fehlings (Toronto, Canada)
In multivariable prognostic research, the development and external validation are the first phases typically involved towards the establishment of clinical prognostic models (CPMs) in practice. This systematic review aims to identify and assess CPMs created to predict clinical outcomes in patients with metastatic epidural spinal disease (MESD) and their subsequent validation studies.
Three electronic databases were searched (January 1, 1990 to December 31, 2017), without language restriction, to identify studies that addressed review question: What are the existing CPMs that have been developed and/or externally validated to predict survival or other clinical outcomes in patients with MESD? Data extraction, reporting and appraisal of the selected studies were conducted following recommended guidance: CHARMS, TRIPOD, and PROBAST (CRD42017072908).
Among 8,077 unique full-text articles, 117 were included. Among the 52 articles describing a CPM (CPM creation, n=44; update of an existing CPM, n=8), 44 did not include any assessment of model performance (calibration and/or discrimination) while 20 reported the number of outcome events and 7 discussed missing data. Among the 5 articles with the term “external validation” or “external validity” in the title or abstract, missing data, number of outcome events, and calibration along with discrimination were discussed in 4, 3 and 2 studies, respectively.
Since 1990, while over 50 CPMs predicting clinical outcomes in patients with MESD were developed, only 5 studies claimed performing an external validation of any of these tools. The majority of the studies included in this review did not report on key methodological and data analysis elements. Greater rigor in the development and validation of CPMs could promote the establishment of CPMs in clinical practice in this patient population.