Background GAMT deficiency is an autosomal recessive disorder of creatine biosynthesis causing developmental delays or intellectual disability in untreated patients as a result of irreversible brain damage occurring prior to diagnosis. Normal neurodevelopmental outcome has been reported in patients treated from neonatal period highlighting the importance of early treatment. Methods Five hundred anonymized newborns from the National Newborn Screening Program of The Netherlands were included into this pilot study. Direct sequencing of the coding region of the GAMT gene was applied following DNA extraction. The disease causing nature of novel missense variants in the GAMT gene was studied by overexpression studies. GAA and creatine was measured in blood dot spots. Results We detected two carriers, one with a known common (c.327G>A) and one with a novel mutation (c.297_309dup (p.Arg105Glyfs*) in the GAMT gene. The estimated incidence of GAMT deficiency was 1:250,000. We also detected five novel missense variants. Overexpression of these variants in GAMT deficient fibroblasts did restore GAMT activity and thus all were considered rare, but not disease causing variants including the c.131G>T (p.Arg44Leu) variant. Interestingly, this variant was predicted to be pathogenic by in silico analysis. The variants were included in the Leiden Open Variation Database (LOVD) database (www.LOVD.nl/GAMT). The average GAA level was 1.14 μmol/L ± 0.45 standard deviations. The average creatine level was 408 μmol/L ± 106. The average GAA/creatine ratio was 2.94 ± 0.136. Conclusion The estimated incidence of GAMT deficiency is 1:250,000 newborns based on our pilot study. The newborn screening for GAMT deficiency should be implemented to identify patients at the asymptomatic stage to achieve normal neurodevelopmental outcome for this treatable neurometabolic disease. Biochemical investigations including GAA, creatine and GAMT enzyme activity measurements are essential to confirm the diagnosis of GAMT deficiency. According to availability, all missense variants can be assessed functionally, as in silico prediction analysis of missense variants is not sufficient to confirm the pathogenicity of missense variants.