AIM: Gestational age estimation by ultrasonography is the gold standard for dating pregnancies. However, the availability of prenatal ultrasonography in low-to-middle-income countries is limited. This study aimed to assess the reliability and validity of last menstrual period (LMP) as a gestational age dating method among women in Johannesburg, South Africa. METHODS: A total of 741 pregnant women were enrolled into a longitudinal study (June 2013 to July 2016). Gestational age was determined by LMP and ultrasonography. Differences in ultrasound-based and LMP-based gestational age estimates were assessed according to the American College of Obstetrics and Gynecologists' guidelines and women were classified as having discrepant results or not. Multiple statistical analyses determined the level of agreement between the two methods and validity of LMP estimates. RESULTS: Compared to ultrasound, dating by LMP assessed gestational age as 0.2 days longer. Women with discrepant results were of significantly lower weight and household socioeconomic status than those without discrepancies. While there was a substantial agreement (k = 0.64; 95% confidence interval, CI: 0.54, 0.71, P < 0.001) between the two methods, LMP only had a 29.0% (95% CI: 14.2, 48.0) sensitivity in identifying late-term neonates and a 33.3% (95% CI: 4.33, 77.7) sensitivity in identifying post-term neonates. CONCLUSION: In the absence of ultrasound, LMP is a reliable alternative for gestational age dating during early pregnancy. However, it is not sensitive in identifying late- and post-term pregnancies and should not be relied upon to make clinical decisions regarding elective cesarean section or induction of labor for supposed prolonged pregnancies.