Caffeine is frequently used to treat apnea of prematurity in preterm infants. Because caffeine has a narrow therapeutic window, plasma concentrations are generally monitored weekly. It would be advantageous to monitor this therapy without blood sampling; saliva might offer this possibility. Paired plasma–saliva and saliva–saliva observations were made in preterm infants (n = 140, gestational ages between 24 and 34 weeks) who received caffeine for the treatment of apnea of prematurity. Three methods were used to collect saliva: no stimulation, dilute citric acid on collection gauze, and dilute citric acid in the cheek pouch before collection. Plasma and saliva caffeine concentrations were determined using high-performance liquid chromatography (HPLC). For all collection methods, the plots of the plasma/saliva outcomes showed linear relationships. The correlation between caffeine concentration in plasma and saliva and the reproducibility of saliva sampling was better with stimulation of saliva production using citric acid in the cheek pouch (r = 0.89) than with no stimulation (r = 0.68) or with stimulation using citric acid on the collection swab (r = 0.79). Monitoring of caffeine therapy in saliva can be applied reliably for routine use in clinical practice, but its reliability and reproducibility depend on the saliva sampling method used; saliva stimulation with citric acid in the cheek pouch is the best method studied.