Background: Diagnosis of Cytomegalovirus (CMV) primary infection during pregnancy or in immunocompetent patients relies on serology with detection of specific CMV-IgG and IgM. In case of positive CMV-IgM in pregnant women, CMV-IgG avidity is now widely recommended, but in general population it is not currently performed. Objective: In this study, we aimed to determine CMV-IgM positive predictive values (PPV) in different clinical settings. Material and methods: We conducted a retrospective study on positive CMV-IgM in our virology laboratory from 2013 to 2019, in three clinical groups: screening in non-symptomatic pregnant women (group 1), pregnant women with ultrasound (US) abnormalities (group 2) and patients (general population) with clinical signs suggestive of CMV primary infection (group 3). CMV-IgG avidity had been performed in all cases allowing to evaluate PPV of positive CMV-IgM to diagnose CMV primary-infection in each group. Results: Between 2013 and 2019, 6859 serum samples were found positive for CMV-IgM and had been tested for CMV-IgG avidity, with 6560 sera for group 1, 30 for group 2 and 269 for group 3. Overall, low avidity confirming primary infection was observed respectively in 16.4 % for group 1, 36.7 % for group 2, and 35.3 % for group 3. CMV-IgM PPV was significantly lower in group 1 compared to groups 2 (p = 0.01) and 3 (p < 0.001). Discussion: Our observations highlight the major importance of including CMV-IgG avidity in the diagnostic algorithm, whatever the clinical situation (for immunocompetent patients), to confirm or exclude a recent CMV primary infection in case of positive CMV-IgM.