identify women at risk is low, owing to the modest strength of the association, the high prevalence of BV in some populations, and the high degrees of variation in the accuracies of the tests used to detect BV (Honest et al., 2004).

Testing of the cervicovaginal secretions of asymptomatic women have been screened for markers of preterm birth, including fetal fibronectin (Goldenberg et al., 1996b,c, 1998), interleukins 6 and 8 (Goepfert et al., 2001; Kurkinen-Raty et al., 2001), and tumor necrosis factor alpha and matrix metalloproteinases (Vogel et al., 2005). The biochemical test for fetal fibronectin is the one about which the most data are available and is the only one marketed in the United States for that purpose.

Fetal fibronectin is a glycoprotein of fetal origin that normally resides at the decidual-chorionic interface within the uterus but is present in cervicovaginal secretions in 3 to 4 percent of pregnant women at between 21 and 37 weeks of gestation (Goldenberg et al., 1996c; Lockwood et al., 1991). Evaluation of fibronection levels in asymptomatic women at 24 weeks of gestation has a sensitivity of 20 to 30 percent for the prediction of spontaneous preterm birth at before 35 weeks of gestation (Goldenberg et al., 1996c; Iams et al., 2002). The sensitivity of determination of fibronectin levels for prediction of early preterm births before 28 weeks of gestation was found to be 63 percent (Goldenberg et al., 1996c). Determination of fibronectin levels is a better test for screening for the risk of delivery within 2 weeks of sampling than for delivery before a specific gestational week of pregnancy (Goldenberg et al., 1997).

Given the pathophysiological heterogeneity of the causes of spontaneous preterm birth, the clinical utility of any individual biomarker for predicting preterm birth is limited. Combinations of markers can increase the sensitivity of the prediction by combining risk predictors that address diverse causes of spontaneous and indicated preterm birth (Goldenberg et al., 2001). The positive predictive value can also be increased (at the sacrifice of sensitivity) by combining selected markers such as the woman’s obstetric history and cervical length; for example, a sonographic cervical length less than 25 mm predicted recurrent preterm birth in 100 percent of African American women with a prior preterm birth (Yost et al., 2004).

Goldenberg and colleagues (1998) sought to develop a multiple-marker test for preterm birth with data collected in the Preterm Prediction Study of