Objective: We have carried out a systematic review of the association between elevated second trimester maternal serum alpha-fetoprotein (AFP) and singleton preterm birth in order to determine its accuracy and the best AFP cut-off level in clinical tests in the general population.
Study design: 24 studies published between January 1991 and October 2007 were included, comprising 207,135 women.
Results: An elevated AFP test (expressed as multiple of the median, MoM) had high specificity but low sensitivity to predict preterm birth: using a 2.5 MoM as the cut-off in the AFP test improved the accuracy compared with 2.0 MoM. However, the overall likelihood ratios for positive and negative tests were not improved. The likelihood ratios for positive tests were: 2.99 (95% CI: 2.45-3.66) and 3.18 (95% CI: 2.07-4.88) for 2.0 MoM and 2.5 MoM, respectively; and for negative tests were: 0.94 (95% CI: 0.91-0.97) and 0.97 (95% CI: 0.95-0.98) for 2.0 MoM and 2.5 MoM, respectively. The available data do not allow us to distinguish whether the association between elevated AFP and preterm birth occurs in spontaneous preterm labour, in elective preterm delivery, or in both. Moreover, in these studies AFP was measured together with other biomarkers (e.g. human chorionic gonadotrophin, oestriol) which often were also elevated. When we included only women in whom AFP was elevated in isolation, there was no association with preterm birth (OR=1.80, 95%CI: 0.92-2.68).
Conclusion: Our findings suggest that maternal AFP levels are strongly related to preterm birth, but only in the context of other abnormal pregnancy markers. The results question the potential usefulness of AFP screening as a primary preterm birth marker and highlight the need for further studies on the functional role of AFP in pregnancy.