Toward a point of care approach for intra-amniotic infection or early delivery using minimally invasive prediction models in women with preterm labor.

Abstract

Among women with preterm labor and intact membranes, those with intra-amniotic infection or inflammation represent the group with the highest risk of spontaneous delivery and worse adverse outcomes. Identification of this group requires amniocentesis, which is perceived as too invasive by both patients and physicians. This study aimed to develop a minimally invasive prediction model for intra-amniotic infection or early delivery to better stratify patients (low or high risk) and rationalize the use of amniocentesis, limiting the indications to the highest-risk group of spontaneous preterm delivery. External validation of 4 prediction models was performed using data from 2022 to 2024 of women diagnosed with preterm labor at <34 weeks of gestation who were admitted to the Hospital Clinic, Hospital Sant Joan de Déu, Vall d'Hebron Barcelona University Hospital (Spain), and University Hospital Hradec Kralove (Kradec Kralove, Czech Republic) and underwent amniocentesis to rule in/out intra-amniotic infection or inflammation. Different prediction models, including transvaginal ultrasound cervical length, serum C-reactive protein, vaginal interleukin 6, vaginal pH, vaginal lactic acid, and vaginal Lactobacillus genus, were validated in these patients. Diagnostic performance was performed in 114 women with preterm labor at <34 weeks of gestation, of whom 42 (36.8%) had intra-amniotic infection or spontaneous delivery within 7 days. The areas under the curve of the different models ranged from 84.0% (95% confidence interval, 78.8%-89.2%) to 89.9% (95% confidence interval, 88.4%-91.4%), the sensitivities ranged from 78.6% (33/42) to 90.5% (38/42), and the specificities ranged from 70.8% (51/72) to 84.7% (61/72). The most feasible and efficient model was formed by combining ultrasound cervical length, serum C-reactive protein, and vaginal interleukin 6, showing an area under the curve of 84.0%, a sensitivity of 78.6% (33/42), a specificity of 84.7% (61/72), a positive predictive value of 75.0% (33/44), and a negative predictive value of 87.1% (61/70). Minimally invasive models were developed to screen women at high risk of intra-amniotic infection or early delivery and guide the selective use of amniocentesis, thereby improving both antenatal counseling and the clinical management of high-risk patients.