Background An educational gradient in preterm birth has been observed in many countries. The distribution of education and other characteristics of women giving birth are the end result of complex processes that play out over the life course. If these processes play out differently in different contexts, cross-cohort comparisons may be helpful in understanding the educational gradient in preterm birth. We performed a cross-cohort comparison of educational disparities in preterm birth based on European birth cohort data.

Methods The study included individual-level data from 12 European cohorts from Denmark, England, France, Lithuania, the Netherlands, Norway, Italy, Portugal and Spain. The studies covered the period 1990 to 2012 and included between 2434 and 99,655 pregnancies each. The association between maternal education and preterm birth for singleton live births was examined separately in each cohort by linear and logistic regression and reported as risk ratios, risk differences and slope indexes of inequality.

Results The percentage of singleton preterm live births varied between 3.7% and 7.5%. There were large variations between the cohorts in the distribution of education and other maternal characteristics. In spite of this, there were similar educational differences in risk of preterm birth in 8 of the 12 cohorts. In these cohorts, the slope index of inequality varied between 2.2 and 3.2 per 100 singleton live births, and relative risk between the lowest and highest education group varied from 1.4 to 1.9. In the remaining 4 cohorts from England, France, Italy and Portugal, one showed a weaker gradient (Slope Index of Inequality =1.5 (-1.0, 3.9)), two showed no educational differences (SII=-0.3 (-2.0, 1.5) and SII=0.9 (-1.3, 3.1)), and the gradient was reversed in the last cohort (SII=-1.2 (-4.6, 2.3)).

Conclusion Educational disparities in risk of preterm birth were found in birth cohorts from all over Europe. Despite differences in the distribution of education and level of preterm birth, the results were remarkably similar across the cohorts. For those few cohorts that did not follow the pattern, study and country characteristics were unable to explain the differences.