The aim of this study was to investigate the population pharmacokinetics of theophylline in premature Korean infants and to assess the influence of clinical covariates.Blood samples were first obtained after 1 week of maintenance dosing and then acquired approximately 4 weeks after continuous dosing. The time points were just before dosing and 2, 4, or 6 hours (at randomly assigned time points) after dosing. Two single-nucleotide polymorphism markers, -3860 G>A (CYP1A2*1C) and -163C>A (CYP1A2*1F), were genotyped. Gestational age (GA), postnatal age (PNA), postconceptional age (PCA = GA + PNA), body weight (BW), height, serum AST, serum ALT, serum BUN, serum creatinine, oxygen support, sex, delivery mode, and CYP1A2 genotypes were used for covariate model building. External validation was analyzed using data from an additional 27 patients.A total of 334 serum concentration measurements were made in 100 patients. A one-compartment absorptive model with first-order elimination was fitted to the data in NONMEM (version 7.1.2). The final model included the following parameters: Clearance (L/h) = 0.00492 × (BW)(3.53) + 0.00646 × (PNA), and volume of distribution (L) = 1.53 × (BW). The addition of the CYP1A2*1C or CYP1A2*1F genotypes to the model did not improve the model. The external validation results confirmed the predictive performance without bias in the final model.The selected covariates were generally consistent with previous studies. However, the mean volume of distribution was higher than the values reported in other population pharmacokinetic studies, which may have been due to the use of 2 sampling time points. The predictive performance was reasonably acceptable. Therefore, the present model may permit more accurate selection of doses to achieve target theophylline concentrations in premature infants.