The validity of three methods (last menstrual period [LPM], Ballard and Dubowitz scores) for assessment of gestational age for premature infants in a low-resource setting was assessed, using antenatal ultrasound as the gold standard. It was hypothesized that LMP and other methods would perform similarly in determining postnatal gestational age. Concordance analysis was applied to data on 355 neonates of <33 weeks gestational age enrolled in a topical skin-therapy trial in a tertiary-care children's hospital in Bangladesh. The concordance coefficient for LMP, Ballard, and Dubowitz was 0.878, 0.914, and 0.886 respectively. LMP and Ballard underestimated gestational age by one day (+/-11) and 2.9 days (+/-7.8) respectively while Dubowitz overestimated gestational age by 3.9 days (+/-7.1) compared to ultrasound finding. LMP in a low-resource setting was a more reliable measure of gestational age than previously thought for estimation of postnatal gestational age of preterm infants. Ballard and Dubowitz scores are slightly more reliable but require more technical skills to perform. Additional prospective trials are warranted to examine LMP against antenatal ultrasound for primary assessment of neonatal gestational age in other low-resource settings.