Objective: To assess the repeatability and suitability for multicentre studies of MScanFit motor unit number estimation (MUNE), which involves modelling compound muscle action potential (CMAP) scans. Methods: Fifteen groups in 9 countries recorded CMAP scans twice, 1–2 weeks apart in healthy subjects from abductor pollicis brevis (APB), abductor digiti minimi (ADM) and tibialis anterior (TA) muscles. The original MScanFit program (MScanFit-1) was compared with a revised version (MScanFit-2), designed to accommodate different muscles and recording conditions by setting the minimal motor unit size as a function of maximum CMAP. Results: Complete sets of 6 recordings were obtained from 148 subjects. CMAP amplitudes differed significantly between centres for all muscles, and the same was true for MScanFit-1 MUNE. With MScanFit-2, MUNE differed less between centres but remained significantly different for APB. Coefficients of variation between repeats were 18.0% for ADM, 16.8% for APB, and 12.1% for TA. Conclusions: It is recommended for multicentre studies to use MScanFit-2 for analysis. TA provided the least variable MUNE values between subjects and the most repeatable within subjects. Significance: MScanFit was primarily devised to model the discontinuities in CMAP scans in patients and is less suitable for healthy subjects with smooth scans.