Preterm delivery affects approximately 10% of pregnancies worldwide and remains a major clinical challenge due to the lack of reliable early predictive tools. Existing strategies are often invasive, relying on blood or amniotic fluid samples and requiring complex processing. In this study, we describe a novel non-invasive approach based on the multiplex detection of inflammatory cytokines in small urine volumes from pregnant women. To account for clinical and temporal variability, we applied Generalized Additive Models for Location, Scale, and Shape (GAMLSS) to adjust for gestational age at sampling and obstetric factors. Correlation network analyses revealed cytokine interactions that distinguished preterm from term deliveries, with macrophage-derived cytokines—MIP-1α, MIP-1β, IL-15, and IL-22—emerging as central nodes. These findings highlight the involvement of the IL-1 pathway in the pathophysiology of preterm labor. Furthermore, urinary IL-5 and IL-31 levels correlated positively with pregnancy duration, whereas IL-1β and IL-1Ra in urine and TNFα in amniotic fluid showed inverse associations. Altogether, this non-invasive methodology provides insight into immune dynamics during pregnancy and offers a foundation for future studies focused on biomarker discovery and mechanistic understanding of preterm birth.