Long Exercise Test in Periodic Paralysis: A Bayesian Analysis

ABSTRACT Introduction: The long exercise test (LET) is used to assess the diagnosis of periodic paralysis (PP), but LET methodology and normal “cutoff” values vary. Methods: To determine optimal LET methodology and cutoffs, we reviewed LET data (abductor digiti minimi motor response amplitude, area) from 55 patients with PP (32 genetically definite) and 125 controls. Receiver operating characteristic curves were constructed, and area under the curve (AUC) was calculated to compare (1) peak‐to‐nadir versus baseline‐to‐nadir methodologies and (2) amplitude versus area decrements. Using bayesian principles, we calculated optimal cutoff decrements that achieved 95% posttest probability of PP for various pretest probabilities (PreTPs). Results: AUC was highest for peak‐to‐nadir methodology and equal for amplitude and area decrements. For PreTP ≤ 50%, optimal decrement cutoffs (peak‐to‐nadir) were > 40% (amplitude) or > 50% (area). Discussion: For confirmation of PP, our data endorse the diagnostic utility of peak‐to‐nadir LET methodology using 40% amplitude or 50% area decrement cutoffs for PreTP ≤50%. Muscle Nerve 59:47–54, 2019 See editorial on pages 8–9 in this issue.