A simulation study of antimicrobial resistance carriage in neonatal intensive care units: implications for cluster-randomised trials

Background: 

Premature neonates in the neonatal intensive care unit (NICU) frequently experience hospital-acquired infection. Antimicrobial resistance (AMR) among hospital-acquired pathogens limits antibiotic treatment options. Infection prevention and control (IPC) interventions can reduce resistant bacterial carriage among infants in the NICU, but there is no high-level evidence to quantify this effect. In this data-driven simulation study, we explored the power of a parallel cluster-randomised trial evaluating the impact of an IPC intervention on prevalence of AMR genes carriage. 

Methods: 

Neonates admitted to 24 NICUs from 8 European countries participating in the EU-funded NeoIPC feasibility study[1] were screened during four point-prevalence surveys (PPSs) at variable intervals of 4, 7, and 10 days. Neonatal stool samples were centrally tested by PCR for presence of ESBL (CTX-M group 1 and 9), CPE (blaIMP, blaKPC, blaNDM, blaOXA-48, blaVIM), and VRE genes (vanA, vanB). Data from neonates born <32 weeks gestational age were modelled using a mixed effects logistic regression. Based on this, we simulated a multi-country, parallel cluster-randomised trial with a duration of 12 months, while varying the number of clusters, microbiological sampling frequency, and countries. We assumed 25% missing stool samples. Length of stay was based on published data [2]. The main outcome was the power of each set of design characteristics to detect a reduction in AMR genes carriage.  

Results: 

Stool samples were available in 631 (67%) of 947 infant-PPS combinations. AMR genes were detected in 119/631 (19%) of samples, varying from 0% to 46% between countries (Figure 1). The best model had a fixed effect for admission duration and random intercepts for NICU and infant (Figure 2). Adequate power could only be attained with a high number of NICUs or when restricting to NICUs from high-prevalence countries (Figure 3). Power was minimally affected by the sampling frequency.  

Conclusions: 

To evaluate the impact of IPC interventions on prevalence of AMR gene carriage in a cluster-randomised trial in the NICU, a very large number of clusters or restriction to high-prevalence settings is required. Frequent sampling is not needed. 

References: 

[1] NeoIPC feasibility phase: Colonisation Surveillance. NeoIPC project website: https://neoipc.org/feasibility/  

[2] S.E. Seaton et al. Estimating neonatal length of stay for babies born very preterm. Arch Dis Child Fetal Neonatal Ed. 2019 Mar;104(2):F182-F186. doi: 10.1136/archdischild-2017-314405. 

Acknowledgements: 

This work has been funded by the European Union’s Horizon 2020 research and innovation programme under grant agreement No. 965328.