Acute type 1 diabetes (AD1) is a rare but definitive immune-related adverse event associated with anti-PD1. Most of the reported cases are close to what has been described as "fulminant type 1 diabetes." We sought to determine whether anti-PD1 could impair glycoregulation and whether occurrence of AD1 could be anticipated by prior glycemic changes. Fasting glycemia collected before, under, and after treatment in melanoma patients treated with anti-PD1 over a period of 36 months were retrospectively analyzed. Glycemic trend analyses were performed using linear regression analysis. In total, 1470 glucose values were monitored in 163 patients treated for a mean duration of 5.96 months. Three patients developed an AD1 (1, 84%). Two other cases were observed in the same period in a still-blinded trial of anti-PD1 versus ipilimumab. All cases of AD1 occurred in patients with a normal pretreatment glycemia, and there was no detectable drift of glycemia before ketoacidosis onset. In 4 of 5 cases of AD1, the HLA subgroups were DRB01* 03 or 04, known to increase type 1 diabetes risk in the general population. In the 28 patients with preexisting type 2 diabetes, there was a slight trend for glycemia increase with anti-PD1 infusions (0.05 mmol/L/infusion P=0.004). In the 132 patients with normal pretreatment glycemia, there was a slight trend for a decrease of glycemia with anti-PD1 infusions (-0.012/mmol/L/infusion P=0.026). These data suggest that the monitoring of glycemia under anti-PD1 cannot help to anticipate AD1, and there is no general tendency to glycemic disorder. HLA genotyping before treatment may help to focus surveillance in patients with the HLA DRB1*03/04 group.