OBJECTIVE: To assess the predictive capacity of cervical length (CL) measurement underwent during the second trimester ultrasound for prediction preterm birth <32, 34, and 37 weeks of gestation in an unselected risk population. METHODS: A retrospective cohort study was performed with 751 singleton pregnancies between 20 and 24+6 weeks of gestation. The CL measurement (mm) using the transvaginal route was obtained in a sagittal view and the calipers positioned to measure the linear distance between the triangular area of echodensity at the external os and the internal os. To compare the preterm (<37 weeks) and term births (≥37 weeks), we used unpaired t test. We assessed whether the CL measurement was dependent of gestational age by performing a linear regression and assessing the coefficient of determination (R2). We additionally assessed the accuracy of CL measurement to predict preterm birth by assessing the area under receiver operating characteristics curves with its respective confidence intervals (CIs) 95%. RESULTS: Preterm birth <37 weeks was found in 13.6% (102/751) of pregnant women. Short cervix (≤25 mm) was found in 2.7% (20/751) of pregnancies. Only 30% (6/20) of pregnant women with short cervix have used progesterone to prevent preterm birth. There was a weak correlation between CL measurement and gestational age at delivery (R2=0.01, P=0.002). Receiver operating characteristics curve analysis of the ability of CL measurement to predict preterm birth <32, 34, and 37 weeks, showed an area under the curve of 0.693 (95% CI, 0.512 to 0.874), 0.472 (95% CI, 0.353 to 0.591), 0.490 (95% CI, 0.426 to 0.555), respectively. CONCLUSION: There was a weak correlation between CL measurement and gestational age at delivery. In an unselected population, CL measurement screening at 20 to 24+6 weeks of gestation does not seem to be a good predictor of preterm birth.