Pitfalls of the E-Ref Procedure: Tie Values and the Proportion of the Abnormal Data.

Abstract

Introduction: Extrapolated reference values (E-Ref) procedure is a new method for determining the cutoff value without collecting the control data. We tried to apply this method to determine the cutoff value for the distal motor latency of the median nerve (median DML). During this process, we found two pitfalls of the E-Ref method. First, the E-Ref procedure did not correctly work when the DML values measured with 0.1 ms accuracy frequently took on tie values. Second, the result was influenced by the proportion of abnormal values. This study investigated these issues.

Methods: Data of the median DML were extracted from our laboratory database. To solve the problem of tie values, we tried a wider post-smoothing window in the original E-Ref method. We also devised a modified method conducting pre-smoothing. To see the effect of the proportion of abnormal data, we simulated many datasets having different proportion of abnormal data.

Results: In total, 1016 DML values were identified. False deflections due to tie values were often identified as the E-Ref point using the original methods even using a wider window, resulting in unrealistically low values. Modified method was free from this drawback. For all methods, the E-Ref value increased as the proportion of abnormal values increased.

Discussion: The problem of tie values, a pitfall of the E-Ref method, might be solved by pre-smoothing the data. The E-Ref value is influenced by the proportion of the normal data, and datasets containing less than 20% abnormal data may achieve appropriate results.