Objective Meta-analyses of high-quality studies have not revealed significant risk of preterm birth in women working night shifts during pregnancy. However, inaccurate self-reports of timing and intensity of night work might result in bias towards null, whereas recall bias may point in the opposite direction. The aim of this study was to investigate if night work (at least three hours between 12pm–5am) in 1st and 2nd trimesters was associated with increased risk of preterm birth and to examine if this was modified by change of work schedule during pregnancy.

Methods We obtained individual day-to-day information on working hours from the Danish Working Hour Database (DWHD), a payroll database including all public service employees in Danish hospitals and public institutions from 2007–2012. Linking DWHD with the Medical Birth Registry, at the individual level, we identified 26,666 births in 19,676 women and retrieved data on among others gestational length. The risk of preterm birth (delivery before 38 weeks/gestation) among night workers compared to day workers (at least three hours between 6am-8pm) was analysed by logistic regression adjusted for age, parity and smoking.

Results The prevalence of night work in the 1st trimester was 42% and in the 2nd trimester 35%. The adjusted odds ratio (OR, 95% CI) of preterm birth among night workers during 1st trimester was 1.04 (0.90; 1.20), and during the 2nd trimester 0.83 (0.71; 0.96). When comparing women changing from night work in 1st trimester to only day work in 2nd with women working at night in both trimesters and exclusively day workers, we found ORs of 1.42 (1.14; 1.77) and 1.29 (1.05; 1.57), respectively. No significant association was observed between women with night work in both trimesters and dayworkers.

Conclusion The study indicates that healthy worker selection during pregnancy may cause bias and underestimation of risk. Previous studies should be reevaluated taking this into account.