Revisiting the Extrapolated Reference Values (E‐Ref) Estimation
Sanjeev D. Nandedkar, Paul E. BarkhausABSTRACT
Introduction/Aims
Extrapolated reference values (E‐Ref) is a method to extract reference values (RVs) from a mixed data set (i.e., a data set containing normal and abnormal findings). This is an attractive way for any laboratory to develop their own RVs. Since its original description, we have made enhancements to the E‐Ref method. Due to growing interest in this methodology, we describe our current version of the algorithm and interpretation of its results.
Methods
The cumulative distribution function (CDF) is the foundation of E‐Ref methodology. Computer simulations were used to generate “normal” and mixed data sets of conduction velocity. The E‐Ref algorithm was revised based on the analysis of CDF in the simulation studies. Bootstrap analysis was performed to assess the variability of estimates and finalize the RV. Median motor nerve conduction parameters from patient studies were analyzed.
Results
The CDF shows a “plateau” that contains the majority of the normal data even in a mixed data set. The mean of data in the plateau approximated the mean of “normal” measurements. The difference between the maximum and minimum in the plateau was close to two standard deviations (SDs).
Discussion
The new algorithm establishes three ranges to define findings: definitely normal, definitely abnormal, and borderline. This is consistent with the customary clinical thought process where abnormalities are not “black and white.” Bootstrap analysis must be performed for reliable estimates. This may be useful to assess the adequacy of the sample size.