Background: With the possible exception of France, no country has reported a reduction in incidence of preterm birth. In fact, the United States and Canada have reported a small but steady increase since the early 1980s. Objective: To assess secular trends in preterm birth and its determinants in a Montreal hospital with a longstanding obstetric and neonatal database. Methods: Cohort study of 65,574 nonreferred births from 1978 to 1996. Gestational age was estimated using three alternative methods: last normal menstual period, early ultrasound, and an algorithm based on the menstrual and ultrasound estimates. Multiple logistic regression was used to assess yearly and interval changes in preterm birth after controlling for potential determinants and confounders that also showed secular trends over the study period. Results: A crude secular increase was seen for births <37, <34, and <32 completed weeks using all three of the gestational age estimation methods. Exclusion of births <500 g and those with induction or “cold” cesarean before each of the corresponding gestational age cut-offs eliminated the secular trends for births <34 and<32 weeks and attenuated the trend for births <37 weeks. Much of the remaining trend [OR=1.015 (95% CI = 1.008-1.021) per year] for births <37 weeks was accounted for by the increasing use of early ultrasound for gestational age dating. This residual trend [OR=1.009 (1.003-1.016) per year] was further attenuated and became nonsignificant [OR=1.003 (0.996-1.010) per year] after controlling for secular increases in unmarried status and the proportion of women ≥35 years of age; along with a decrease in alcohol consumption and increases in histologic chorioamnionitis and cocaine use, these changes appear to have prevented a significant reversed trend [OR=0.986(0.973-0.999) per year] that would otherwise have been observed since the mid-1980s. Conclusions: This Montreal hospital has experienced an increase in preterm birth since 1978 that parallels increases recently reported in population-based national studies from the United States and Canada. The increase appears largely attributable to preterm induction and cesarean section, the increasing use of early ultrasound dating, and changes in sociodemographic and behavioral factors.