Optimization of the cut-offs in acetylcholine receptor antibodies and diagnostic performance in myasthenia gravis patients

Objective: This study aims to establish an optimization procedure to define the cut-offs of quantitative assays for acetylcholine receptor antibody (AChRAb), evaluate their diagnostic performance in myasthenia gravis (MG), and explore the association with clinical features. Methods: Samples from a representative cohort of 77 MG patients, 80 healthy controls (HC) and 80 other autoimmune diseases (OAD) patients were tested using competitive inhibition ELISA and RIA. Raw values (OD and cpm) and processed values (inhibition rate, binding rate and concentration) were used to define the cut-offs with statistical methods, a rough method, and receiver operating characteristic (ROC) curve. Optimal cut-offs were selected by comparing false positive rates in HC and OAD individuals. The diagnostic performance was evaluated in whole MG cohort and subgroups. Agreement between ELISA and RIA for AChRAb positivity were examined with Kappa test and McNemar test. Clinical association with AChRAb was explored by comparison among subgroups and with Spearman rank correlation. Results: The optimal cut-offs for AChRAb positivity were determined as OD <= 1.79 for ELISA and cpm > 1234.12 for RIA, which derived from statistical method and performed better than those derived from ROC curves. The sensitivity and specificity were 74.03%, 100% for ELISA, and 74.03%, 99.37% for RIA. There was good agreement between ELISA and RIA for AChRAb positivity in whole cohort and subgroups (weighted k > 0.71, p < 0.01; McNemar test, p > 0.05). Levels of AChRAb were different in MG subgroups (p < 0.01). Correlation between Quantitative Myasthenia Gravis scores and AChRAb levels was moderate for ELISA and RIA (rs = -0.60 and 0.57, p < 0.01). Conclusion: The raw testing values of ELISA and RIA were found as optimal quantitative measures of AChRAb levels. There are good agreements on diagnostic performance between two assays. Quantitative values are more informative than positivity in association with clinical features.