We performed a meta-analysis on the diagnostic value of IgM anti-GM1 antibodies. The reported frequencies of IgM anti-GM1 antibodies ranged from 0 to 100% for patients with multifocal motor neuropathy (MMN), from 0 to 33% in the Guillain-Barré syndrome, from 0 to 65% in amyotrophic lateral sclerosis (ALS), from 0 to 77% in chronic inflammatory demyelinating neuropathy, and from 0 to 81% in lower motor neuron disease (LMND). However, using funnel graphs and a chi-square test we determined that the method of ELISA was the most important factor explaining these differences. After allowing for two factors--the use of detergent and the duration and temperature of serum incubation-studies became homogeneous in all but the LMND group of method A (no detergent, duration of serum incubation 5 hours) and the ALS group of method B (no detergent, duration of serum incubation at least 12 hours [overnight]). Since the anti-GM1 antibody assay serves to confirm clinical suspicion of MMN rather than to exclude the disease, specificity is more important than sensitivity. ELISA methods that do not use detergent and that incubate serum overnight resulted in a specificity of 90% and sensitivity of 38% in the comparison of MMN and LMND. With these values we calculated incremental ruling-in and ruling-out gain curves. Prior probabilities between 20 and 60% for having MMN changed to post-test probabilities between 50 and 85%, which is of clinical importance. In conclusion, ELISA is a useful diagnostic test to demonstrate IgM anti-GM1 antibodies provided the methods do not use detergent and do incubate serum overnight.