Wednesday, May 11, 2016

Predicting spontaneous preterm birth in twin pregnancies utilizing cervical length and gestational age: Individual patient data meta-analysis.


Multiple births are steadily climbing all around the world. Developed countries making a significantly higher contribution to this rising rate because of women delaying childbirth, elderly mothers and increased use of ARTs.

US twinning rate rose by 101% from 1980 – 2006. About 68,339 twins were born in 1980 that doubled to 137,085 in 2006. The US current twin birth rate is 33.9 per 1,000 live births.

 According to WHO the rate of singleton preterm birth ranges between 5% to 18% for singleton pregnancy worldwide, the average being 11%, while almost 60% of twins are delivered preterm. About 13% of twins are born before 34 weeks and 7% before 32 weeks.

A multitude of prophylactic therapies have been in use like to gain valuable gestational weeks by supplementing progesterone, vaginal pessaries and strict bed rest without substantially significant results.

The next step was to develop essential biomarkers that can predict the chances of preterm births. Cervical length(CL) has long   been used as a predictive indicator of preterm birth. An earlier review has shown that a CL < or=20 mm at 20-24 weeks' gestation was the most accurate in predicting preterm birth at <32 and <34 weeks respectively. Many other studies have combined fetal fibronectin with CL. 

Studies in singleton pregnancies have also shown that the relationship between CL and spontaneous preterm birth (sPTB) is dependent on the Gestational age (GA) at which the USG is done, a shorter CL early in pregnancy has greater significance than the same measurement at a later GA.

Such studies in twins are few with small sample size and are not comparable. Previous meta-analysis has shown a relationship between CL and sPTB in twins, but did not correlate the GA at screening with prediction of sPTB.

This recent study published in the May, 2016 issue of BJOG is a meta-analysis of independent patient data(IPD), and provides a new estimate in which CL and GA are treated as continuous variables to predict weeks at delivery.

Specific data collected for each patient from the original authors of the study included the exact GA at CL screening, the CL measurement in millimeters and the exact GA at birth in weeks and days.
23 studies met the inclusion criteria, resulting in a total of 6188 transvaginal scans, performed on 4409 twin pregnancies. 

In the first analysis, univariate regression was performed to see what other confounders like maternal age, ethnicity, smoking, BMI, chorionicity, parity and study location affects the GA at birth. 

As second analysis multinomial logistic regression model was derived predicting the probabilities of very early preterm, early preterm, late preterm, and term birth using GA at USG and CL as continuous variables.

Important study results were:

  • BMI was the only other variable that correlated significantly with GA at birth in the univariate analysis, but when it was incorporated into multinomial logistic regression model with CL and GA at ultrasound, prediction of GA at birth did not improve.
  • A short CL measured at ≤20+0 weeks by USG indicates a probability of birth significantly earlier than if the same CL was taken at a later GA.
  • When screening before 18+0 weeks, any cervical length <30 mm has a higher risk of sPTB at ≤28+0 weeks in twins than in singletons.Whereas the best prediction of birth between 28+1 and 36+0 weeks was provided by screening at ≥24+0 weeks.
  • A 100% probability of preterm birth not occurring before 28 weeks is achieved by CL of 65 mm and 43 mm at ultrasound GA at ≤18+0 weeks and at 22+1 to 24+0 weeks, respectively.


In the third analysis, the accuracy of the model to correctly predict term delivery as compared to preterm was assessed. The model has a 68.2% true negative rate, classifying correctly those who were predicted to deliver at ≥36+1 weeks, compared with 26.2, 13.3 and 36.2% correctly predicted to deliver at ≤28+0, 28+1 to 32+0 and 32+1 to 36+0 weeks, respectively (true positive rate).

Although effective intervention for sPTB in twins are limited, the study provides risks of very early, early and late preterm birth, so a personalized cost effective delivery plan, optimal timing of corticosteroids and referring to neonatal unit can be managed. It also justifies serial CL measurements, so that early and late sPTB could be predicted.

To conclude the authors, recommend to start the screening at ≤18+0 weeks with repeat screening at >22+0 weeks; this best identifies the patients that may deliver very early at ≤28+0 weeks as well as the more common later group of sPTB between 28+0 to 36+0 weeks. 


References:




No comments:

Post a Comment