Investigating the Mechanisms of Reduced Blood Transfusions after Delayed Umbilical Cord Clamping: The TITANS Causal Mediation Analysis

Abstract

Introduction: Delaying clamping of the umbilical cord (deferred cord clamping [DCC]) in preterm infants reduces mortality and the need for blood transfusions. The mechanisms leading to these benefits are not well understood. The TITANS study investigates potential mediators of the reduction in blood transfusions in infants who received DCC.

Materials and methods: Additional patient data was sourced from Australian and New Zealand sites from the Australian Placental Transfusion Study (APTS). APTS randomized preterm infants <30 weeks' gestation to receive DCC (60 s) or immediate cord clamping. We examined whether placental transfusion or initial severity of illness mediated the reduced requirement for blood transfusions for infants randomized to DCC. Peak hematocrit in the first 7 days (Hct) was used as an indicator of placental transfusion quantity. Cumulative blood sampled, mechanical ventilation, and arterial sampling lines were used as indicators of severity of illness. We quantified the natural indirect effect of peak Hct and then for all mediators in a joint model with sequential mediation.

Results: Data from 1,260 (of 1,401) Australian and New Zealand APTS infants were obtained. The effect of DCC on subsequent blood transfusion was mediated through peak Hct (indirect effect OR = 0.85, 95% CI: 0.79-0.93; p < 0.001), which accounted for 37% of the total effect. Indicators of severity of illness did not mediate the effect independently of peak Hct.

Conclusion: Peak Hct mediated some, but not all, of the effect of DCC on blood transfusion, whereas markers of severity of illness were not independent mediators.

Keywords: Causal mediation; Cord clamping; Delayed cord clamping; Neonate; Newborn; Placental transfusion; Preterm.

Fig. 1.

Diagrams of the possible causal mechanisms for the outcome (Y) based on the cord clamping intervention (X). a, b depict single mediator models in which the intervention X (intervention – immediate cord clamping [ICC] or deferred cord clamping [DCC]) affects the outcome Y through the indirect path of the mediators M₁ (placental transfusion) and M₂ (severity of illness mediator) independently. In these models, M₁ and M₂ do not affect each other. c, d If M₁ is not independent of M₂, we employ sequential mediation depicted in models. First in (c), we examine a mediator model with only M₁. This captures the effect of M₁ and all the subsequent pathways it influences (i.e., x->M₁->y and x->M₁->M₂->y). Next in (d), we examine the joint model containing all mediators; here, the intervention (x) affects the outcome (y) through both mediators (M₁ and M₂).

Fig. 2.

Flow chart: the top gray box summarizes all infants randomized and included in APTS. The second gray box summarizes the TITANS follow-up.