Search

Introduction: The goal of this study was to generate reference data for a wide range of sensory and motor nerve conduction studies (NCSs) that are commonly used in routine electrodiagnostic practice. Methods: NCS data were collected from control subjects with a normal neurological examination and no known risk factors for or symptoms to suggest disease of the peripheral nervous system. Thirteen sensory and motor nerves were studied in the upper and lower limbs using standard recording techniques. Quantile regression was used to obtain estimates of the lower (e.g., 5th) percentiles for sensory and motor amplitudes and conduction velocities, as well as estimates of the upper (e.g., 95th) percentiles for distal motor latency and F-wave latencies. Results: Reference data for sensory and motor amplitudes required adjustment for age; F-wave latencies required adjustment for gender. Covariate adjustments were not required for other reference data. Data from N = 100 and N = 190 subjects permitted estimation of as low as the 4th and 2nd percentile, respectively, but further increases to N = 296 did not permit estimation of a more extreme percentile. Discussion: Efforts to develop reference data require careful selection of study subjects; sufficient sample size; adjustment for relevant covariates; and appropriate statistical methods for analysis. Reference data that cannot be estimated reliably due to absent potentials or data variability must be interpreted with caution. We present data for a range of clinically applicable reference percentiles, but judgment regarding which is most appropriate for diagnostic purpose awaits empiric evaluation in a disease population. Muscle Nerve 40: 772–794, 2009

Introduction: The interpretation of nerve conduction studies, as with any diagnostic test, requires an ability to differentiate normal from abnormal. To this end there have been many efforts to establish reference or normative data, but they have been hampered by a variety of methodological shortcomings. The goal of this article is to introduce a statistical method known as quantile regression, which we contend is better suited than existing methods to generate reference data, especially when there is a need to adjust for covariates. Methods: Statistical methods previously used for generation of reference data are reviewed. Quantile regression is presented and used to estimate the lower percentiles for response amplitudes of the radial sensory and tibial motor nerves. Results: Using data from 190 subjects, it is possible to estimate as low as the 2nd percentile for the radial nerve. Using data from 99 subjects it is possible to estimate as low as the 4th percentile for the tibial nerve. Percentile estimation for both nerves required adjustment for age, but no other covariates. Discussion: Quantile regression is well suited to the estimation of extreme percentiles, the very percentiles that are most relevant to reference data. It is also less dependent on data distribution and permits covariate adjustment, even for continuous variables such as age, which are clinically important determinants of reference data for nerve conduction studies. We recommend the use of quantile regression for future studies of reference data. Muscle Nerve 40: 763–771, 2009