Abstract
To determine the accuracy with which transvaginal cervical ultrasonography predicts spontaneous preterm birth, studies were identified, without language restrictions, through nine different databases (1966–2002) and manual searching of bibliographies of known primary and review articles. Studies were selected if they undertook transvaginal ultrasonographic cervical length measurement among a population of pregnant women antenatally and had known gestation at delivery. Data were extracted for studies’ characteristics, quality and accuracy results for < 32, < 34 and < 37 weeks’ gestation outcome as the reference standards. Data were stratified according to singletons or twins, gestational age at testing, cervical length thresholds and the various reference standards. Data were pooled to produce summary estimates of likelihood ratios (LRs). Forty-six primary articles included a total of 31 577 women, consisting of 33 studies in asymptomatic and 13 studies in symptomatic women. There was a wide variation among studies with respect to gestational age at testing, definition of threshold for abnormality and definition of reference standard. The most commonly reported subgroup was testing of asymptomatic women at 20–24 weeks’ gestation using a threshold of 25 mm with spontaneous preterm birth before 34 weeks’ gestation as the reference standard. The summary LR+ for this group was 4.44 (95% confidence interval (CI) 3.23–6.08), with corresponding LR− of 0.68 (95% CI 0.49–0.94).