Introduction
Worldwide, more children than before survive preterm birth. Preterm birth can affect long-term cognitive outcomes. The aim of our study was to investigate the association between preterm birth and academic performance and intelligence in youth.

Material and methods
This cohort study included all liveborn children in Denmark from 1978 to 2000. We used uni- and multivariable logistic and linear regressions to analyze associations between gestational age and school graduation, grade point average (GPA), attending conscription, and male intelligence scores at conscription. We adjusted for a priori defined potential confounders.

Results
The study included 1 450 681 children and found an association between preterm birth and lower academic performance, with children born extremely preterm having the lowest odds of graduating from lower- and upper secondary education (LSE and USE) and appearing before the conscription board (odds ratios of 0.45 [0.38–0.54], 0.52 [0.46–0.59], and 0.47 [0.39–0.56] for LSE, USE, and conscription, respectively, compared to the term group). Statistically significant differences were observed in LSE for total GPA and core subject GPA with higher GPAs in the term group, which were considered clinically relevant for mathematics with a 0.71 higher grade point for the term compared to the extremely preterm. Conversely, USE differences were less evident, and in linear regression models we found that preterm birth was associated with higher GPAs in the adjusted analyses; however, this was not statistically significant. We demonstrated statistically significant differences in intelligence scores at conscription with lower scores in the three preterm groups (−5.13, −2.73, and − 0.76, respectively) compared to the term group.

Conclusions
Low gestational age at birth was associated with not graduating from LSE and USE, achieving lower GPAs in LSE, not attending conscription, and lower intelligence scores in young adulthood. The findings remained significant after adjusting for potential confounders.