Isolated decreased serum-immunoglobulin (Ig)M has been associated with severe and/or recurrent infections, atopy and autoimmunity. However, the reported high prevalence of clinical problems in IgM-deficient patients may reflect the skewed tertiary centre population studied so far. Also, many papers on IgM deficiency have included patients with more abnormalities than simply IgM-deficiency. We studied truly selective primary IgM deficiency according to the diagnostic criteria of the European Society for Immunodeficiencies (ESID) (true sIgMdef) by reviewing the literature (261 patients with primary decreased serum-IgM in 46 papers) and analysing retrospectively all patients with decreased serum-IgM in a large teaching hospital in 's-Hertogenbosch, the Netherlands [1 July 2005-23 March 2016; n = 8049 IgM < 0·4 g/l; n = 2064 solitary (IgG+IgA normal/IgM < age-matched reference)]. A total of 359 of 2064 (17%) cases from our cohort had primary isolated decreased serum-IgM, proven persistent in 45 of 359 (13%) cases; their medical charts were reviewed. Our main finding is that true sIgMdef is probably very rare. Only six of 261 (2%) literature cases and three of 45 (7%) cases from our cohort fulfilled the ESID criteria completely; 63 of 261 (24%) literature cases also had other immunological abnormalities and fulfilled the criteria for unclassified antibody deficiencies (unPAD) instead. The diagnosis was often uncertain (possible sIgMdef): data on IgG subclasses and/or vaccination responses were lacking in 192 of 261 (74%) literature cases and 42 of 45 (93%) cases from our cohort. Our results also illustrate the clinical challenge of determining the relevance of a serum sample with decreased IgM; a larger cohort of true sIgMdef patients is needed to explore fully its clinical consequences. The ESID online Registry would be a useful tool for this.