The iNPH Radscale May Better Predict Whether Ventriculoperitonal Shunting was Offered, Compared to Drain Trial Gait Variables
Xavier Soehnlen1, Claire Sonneborn2, Stephen Jones2, Sean Nagel2, Richard Rammo2, James Liao2
1OU-HCOM, 2Cleveland Clinic
Objective:
Idiopathic normal pressure hydrocephalus (iNPH), can be treated surgically with ventriculoperitoneal shunting (VPS); however, there is no standardized way to evaluate VPS candidacy. Though physicians may consider the iNPH Radscale when screening patients for drain trials, the decision to offer a VPS is based on physician judgement of gait improvement in pre/post drain trial gait tests. For this study, we determined which Radscale and gait variables best predicted whether patients were offered VPS.
Design/Methods:
iNPH Radscale scores from preoperative MRI/CT, and gait variables (walk speed and steps to complete standardized distance) from standard-of-care gait assessments pre and post drain trial, were collected from 221 patients who underwent iNPH workup (167 offered VPS and 54 not offered). From the imaging and gait variables, we used the Harrel step-down procedure to select a model predict to if VPS was offered. This procedure prioritizes the concordance (c) statistic for discriminatory ability and removes predictor variables sequentially in the order of their influence, from least to most important.
Results:
The final, best-fitting model contained the following predictors: 1) Radscale full score; Radscale subscores for 2) Narrow Sulci, 3) Sylvian Fissures, 4) Temporal Horns, and 5) Periventricular Hypodensities. This model correctly classified shunt status 92.5% of the time (C statistic). The gait variables were not part of this model, and were removed by the Harrel step-down procedure.
Conclusions:
Despite patients being offered VPS primarily based on gait improvements pre vs post drain trial, gait variables were not included in the best-fitting, most parsimonious model for prediction of whether VPS was offered. In other words, gait variables did not significantly improve the model’s C statistic of 0.925. This surprising finding needs confirmation with larger samples, and, should be extended to determine if iNPH Radscale predicts sustained VPS response, which is the most important clinical question.
Disclaimer: Abstracts were not reviewed by Neurology® and do not reflect the views of Neurology® editors or staff.