Good afternoon. So I would like to thank the organizing committee as well as the selection committee for this tremendous opportunity. My name is Annick Nader. I'm a Ph.D. in Neurosurgery at the University of Toronto. That's in Canada. And I'm very pleased to present this afternoon. So I have no personal disclosure. I'm going to be referring to metastatic epidural spinal cord compression as MESCC. So, so far, most common existing clinical prognostic models, CPMs, of survival for patients with MESCC are based on study mostly using retrospective data, small sample size, and often an unclear methodology. So to date, there's no external validation that has been conducted in accordance to the tripod guidelines. So in order to enhance the quality of studies related to the development and validation of clinical prediction models, the tripod, which stands for the Transparent Reporting of Multivariable Prediction Model for Individual Prognosis or Diagnosis, has been published in 2015, and along with those 22 items on the checklist. Using the database provided by the AOSPI North America MESCC study, which was a prospective multicenter observational cohort study that included 142 surgical patients, we aimed to identify the preoperative predictive factors of survival in patients with MESCC who underwent surgery, to examine how these predictors relate to eight existing CPMs, and to perform the first external validation in accordance to the tripod guidelines. For univariable analyses, we used a Kaplan-Meyer, Derlag-Rank test, and the Cox regression, and any non-collinear predictors with less than 10% missing data, more than 10 even per stratum, and a p-value inferior to 0.05 were included into our multivariable Cox model using a backward elimination. For the first full external validation, eight models were assessed, the original revised Tokurashi, Tomita, modified Bauer, the van der Linden, the Bartels, oswestry, the spinal risk index, and the Bolin model. They were all assessed in terms of calibration and discrimination, as well survival was stratified by risk groups. So we identified three independent predictors of survival in our multivariable Cox model, the type of primary tumor, the presence of organ metastases, and the SF36 physical component score. For discrimination, which describes the ability of a model to distinguish between low and high risk patient, the Bartels model had the highest C-statistic at 0.68. As a reminder, a 0.5 C-statistic describes no discrimination, whereas a C-statistic of 0.60 to 0.75 describes a possibly helpful discrimination, and anything of a model with a C-statistic higher than 0.75 describes a clearly useful discrimination. For calibration, calibration is the degree to which the predicted probabilities match the observed values. So in a calibration plot where you have the 45 degree line that we present perfect calibration, anything under the line represent an overestimate of survival. So we can see that overall calibration was relatively poor. For risk group classification, the van der Linden was the only model that distinguished between our risk groups. So in conclusion, preoperative predictors for longer survival in patient with EMSCC who underwent surgery, we identified three independent predictors, the type of primary tumor as being breast, prostate, or thyroid cancer, the absence of visceral metastases or organ metastases, and a higher score on the SF36 physical component. For the full external validation in accordance with the tripod guidelines, calibration was overall poor and discrimination was possibly helpful. So therefore, physicians should be careful in using those clinical prognostic models given that the predictive ability is or was suboptimal in this research. I would like to thank my co-authors as well as AO Spine and CIHR for their funding and my supervisor, Dr. Michael Fillings. Thank you.

Annick Nader

metastatic epidural spinal cord compression

clinical prognostic models

Transparent Reporting of Multivariable Prediction Model for Individual Prognosis or Diagnosis

AOSPI North America MESCC study