INTRODUCTION: Saliva is a promising option for therapeutic drug monitoring, with studies since the 1970s indicating a good correlation between plasma and saliva levels for early anti-seizure medications, although limited data exist for newer generation drugs. OBJECTIVES: To evaluate the reliability and predictive power of saliva as a minimally invasive surrogate marker of plasma concentration for the routine therapeutic drug monitoring (TDM) of newer anti-seizure medications (ASM). METHODS: We collected blood samples at steady state in patients at least 6 h post-dose, paired with unstimulated saliva samples. We evaluated the correlation between plasma and saliva drug levels and the positive and negative predictive value for plasma values extrapolation from saliva levels. A very low saliva level was defined as below half the plasma lower reference limit. RESULTS: 294 adult patients (53 % male) with a mean age of 40 (SD: 16) were enrolled and 589 paired saliva-plasma samples were quantified. The highest significant correlations between saliva and plasma were observed for zonisamide (R CONCLUSIONS: This large newer ASM paired plasma-saliva collection allows to precise the potential use of saliva in the management of epilepsy, especially for commonly used ASM such as lamotrigine and levetiracetam. Although they correlate well, extrapolating plasma levels from saliva samples is still an imprecise approximation, making it inadequate for fine dosage adjustments. Yet, a very low saliva level has an appreciable discriminative ability for low plasma level. Unstimulated saliva represents a convenient non-invasive alternative to plasma, to readily identify compliance issues or major drug-drug interactions.