The objective of this study was to compare cervicovaginal fetal fibronectin (FFN) versus transvaginal sonographic cervical length as predictors of preterm birth in twin pregnancy and to test whether 17-hydroxyprogesterone caproate (17OHPc) modifies the predictive value of FFN. Secondary analysis of a randomized trial of 17OHPc versus placebo in dichorionic-diamniotic twins, analyzing the subset with screening FFN (N = 198) and/or cervical length (N = 214) at 24 to 26 weeks of gestation. Positive FFN was found in 7%, cervical length ≤ 25 mm in 8%, and both positive FFN and cervical length ≤ 25 mm in 3%. Birth < 32, < 34, and < 37 weeks occurred in 8, 30, and 67%, respectively. In logistic regression analysis controlling for FFN, cervical length, prior preterm birth, and treatment group, positive FFN was significantly associated with birth < 30 and < 32 weeks (odds ratio 55.0 [95% confidence interval 5.2-582], 18.1 [3.3-99], respectively, p < 0.001 for both) but cervical length ≤ 25 mm was not (odds ratio 0.1 [0.002-1.6], 0.6 [0.1-4.3]). Positive FFN was stronger than cervical length ≤ 25 mm in predicting early preterm birth in twins, regardless of 17OHPc use. Treatment with 17OHPc did not appear to alter the predictive value of FFN.