Aim: DBS might be an appropriate microsampling technique for therapeutic drug monitoring of caffeine in infants. Nevertheless, its application presents several issues that still limit its use. This paper describes a validated DBS–LC–MS/MS method for caffeine. Results: The results of the method validation showed an hematocrit dependence. In the analysis of 96 paired plasma and DBS clinical samples, caffeine levels measured in DBS were statistically significantly lower than in plasma but the observed differences were independent from hematocrit. Conclusion: These results clearly showed the need for extensive validation with real-life samples for DBS-based methods. DBS–LC–MS/MS can be considered to be a good alternative to traditional methods for therapeutic drug monitoring or PK studies in preterm infants.