Background: Threatened preterm labor is the major cause of hospital admission during the second half of pregnancy. An early diagnosis is crucial for adopting pharmacologic measures to reduce perinatal mortality and morbidity. Current diagnostic criteria are based on symptoms and short cervical length. However, there is a high false-positive rate when using these criteria, which implies overtreatment, causing unnecessary side effects and an avoidable economic burden.
Objective: This study aimed to compare the use of placental alpha microglobulin-1 and interleukin-6 as vaginal biomarkers combined with cervical length and other maternal characteristics to improve the prediction of preterm delivery in symptomatic women.
Study design: A prospective observational study was conducted in women with singleton pregnancies complicated by threatened preterm labor with intact membranes at 24+0 to 34+6 weeks of gestation. A total of 136 women were included in this study. Vaginal fluid was collected with a swab for placental alpha microglobulin-1 determination using the PartoSure test, interleukin-6 was assessed by electrochemiluminescence immunoassay, cervical length was measured by transvaginal ultrasound, and obstetrical variables and newborn details were retrieved from clinical records. These characteristics were used to fit univariate binary logistic regression models to predict time to delivery <7 days, time to delivery <14 days, gestational age at delivery ≤34 weeks, and gestational age at delivery ≤37 weeks, and multivariate binary logistic regression models were fitted with imbalanced and balanced data. Performance of models was assessed by their F2-scores and other metrics, and the association of their variables with a risk or a protective factor was studied.
Results: A total of 136 women were recruited, of whom 8 were lost to follow-up and 7 were excluded. Of the remaining 121 patients, 22 had a time to delivery <7 days and 31 had a time to delivery <14 days, and 30 deliveries occurred with a gestational age at delivery ≤34 weeks and 55 with a gestational age at delivery ≤37 weeks. Univariate binary logistic regression models fitted with the log transformation of interleukin-6 showed the greatest F2-scores in most studies, which outperformed those of models fitted with placental alpha microglobulin-1 (log[interleukin-6] vs placental alpha microglobulin-1 in time to delivery <7 days: 0.38 vs 0.30; time to delivery <14 days: 0.58 vs 0.29; gestational age at delivery ≤34 weeks: 0.56 vs 0.29; gestational age at delivery ≤37 weeks: 0.61 vs 0.16). Multivariate logistic regression models fitted with imbalanced data sets outperformed most univariate models (F2-score in time to delivery <7 days: 0.63; time to delivery <14 days: 0.54; gestational age at delivery ≤34 weeks: 0.62; gestational age at delivery ≤37 weeks: 0.73). The performance of prediction of multivariate models was drastically improved when data sets were balanced, and was maximum for time to delivery <7 days (F2-score: 0.88±0.2; positive predictive value: 0.86±0.02; negative predictive value: 0.89±0.03).
Conclusion: A multivariate assessment including interleukin-6 may lead to more targeted treatment, thus reducing unnecessary hospitalization and avoiding unnecessary maternal-fetal treatment.
Keywords: multivariate assessment; preterm birth prediction; threatened preterm labor.
Copyright © 2023 Elsevier Inc. All rights reserved.