OBJECTIVES: To integrate data on the performance of cervical length measurement for the prediction of preterm birth in symptomatic women.METHODS: MEDLINE, SCOPUS and manual searches for studies with transvaginal ultrasound measurement of the cervical length in symptomatic women were carried out. Random effects models were used for data integration, and pooled test estimates of sensitivity, specificity, and positive and negative likelihood ratios (LR+ and LR-) were calculated along with their 95% CIs.RESULTS: Twenty-eight studies fulfilled the selection criteria. For birth within 1 week from presentation, the pooled sensitivity, specificity, LR+ and LR- of cervical length < 15 mm were 59.9% (95% CI, 52.7-66.8%), 90.5% (95% CI, 89.0-91.9%), 5.71 (95% CI, 3.77-8.65) and 0.51 (95% CI, 0.33-0.80), respectively. The same estimates for studies with presentation at or before 34 + 0 weeks were 71.0% (95% CI, 60.6-79.9%), 89.8% (95% CI, 87.4-91.9%), 5.19 (95% CI, 2.29-11.74) and 0.38 (95% CI, 0.11-1.34), respectively. For prediction of birth before 34 weeks, the pooled sensitivity, specificity, LR+ and LR- of cervical length < 15 mm were 46.2% (95% CI, 34.8-57.8%), 93.7% (95% CI, 90.7-96.0%), 4.31 (95% CI, 2.73-6.82) and 0.63 (95% CI, 0.38-1.04), respectively. There was considerable heterogeneity across studies in most estimates.CONCLUSIONS: Measurement of cervical length in symptomatic women can detect a significant proportion of those who will deliver within 1 week and help to rationalize their management. The considerable heterogeneity across studies may be indicative of methodological flaws, which either were not reported at all or were under-reported.