A multivariable prediction model for intra-amniotic infection in patients with preterm labor and intact membranes including a point of care system that measures amniotic fluid MMP-8

Objectives: Among patients with preterm labor and intact membranes (PTL), those with intra-amniotic infection (IAI) present the highest risk of adverse perinatal outcomes. Current identification of IAI, based on microbiological cultures and/or polymerase chain reaction amplification of the 16S ribosomal RNA gene, delay diagnosis and, consequently, antenatal management. The aim to of the study was to assess the performance of a multivariable prediction model for diagnosing IAI in patients with PTL below 34.0 weeks using clinical, sonographic and biochemical biomarkers.

Methods: From 2019 to 2022, we prospectively included pregnant patients admitted below 34.0 weeks with diagnosis of PTL and had undergone amniocentesis to rule in/out IAI. The main outcome was IAI, defined by a positive culture and/or 16S ribosomal RNA gene in amniotic fluid. Based on the date of admission, the sample (n=98) was divided into a derivation (2019-2020, n=49) and validation cohort (2021-2022, n=49). Logistic regression models were developed for the outcomes evaluated. As predictive variables we explored ultrasound cervical length measurement at admission, maternal C-reactive protein, gestational age, and amniotic fluid glucose and matrix metalloproteinase-8 (MMP-8) levels. The model was developed in the derivation cohort and applied to the validation cohort and diagnostic performance was evaluated. Clinical management was blinded to the model results.

Results: During the study period, we included 98 patients admitted with a diagnosis of PTL. Of these, 10 % had IAI. The final model included MMP-8 and amniotic fluid glucose levels and showed an area under the receiver operating characteristic curve to predict the risk of IAI of 0.961 (95 % confidence interval: 0.860-0.995) with a sensitivity of 75 %, specificity of 93.3 %, positive likelihood ratio (LR) of 11.3 and negative LR of 0.27 in the validation cohort.

Conclusions: In patients with PTL, a multivariable prediction model including amniotic fluid MMP-8 and glucose levels might help in the clinical management of patients undergoing amniocentesis to rule in/out IAI, providing results within a few minutes.