How to Deal with the Latency of Unobtainable Responses in the Statistical Analysis

ABSTRACT

OBJECTIVE: To evaluate the usability of near-nerve needle recording techniques in cases of unrecordable sensory nerve action potentials (SNAPs) with a surface electrode and to determine a proper alternative value of the missing latencies. METHOD: Twenty six hands of 23 patients with a carpal tunnel syndrome (CTS) and an unobtainable median SNAP by surface electrode were evaluated by the near-nerve needle recording of median SNAPs. Using the nerve conduction data of 113 patients with CTS, we have established 3 alternative values maximal, 95 percentile and predictive latencies. The alternative values were compared with the mean onset latencies by the near-nerve needle recordings of median SNAPs. RESULTS: Median SNAPs were obtainable in the 22 out of 26 hands by the near-nerve recording technique. The mean onset latency was 5.51+/-0.36 ms. The alternative values from 113 patients with CTS were as follows maximum latency, 6.9 ms; 95 percentile latency, 5.6 ms; and predictive latency, 5.52 ms (Y = 0.123x X 5.52491; Y, onset latency; X, amplitude; r2=0.564; p=0.00). The Predictive latency was nearest to the mean onset latency. CONCLUSION: To minimize the selection bias and statistical errors, the near nerve recording techniques proved to be a valuable method in cases of unrecordable SNAPs with surface electrode. For compensation of missing data, a proper alternative value can be obtained by the predictive latency calculated from a linear regression.