Prediction Is Very Difficult, Especially if It Is About the Future

    loading  Checking for direct PDF access through Ovid


The devastating consequences of preterm birth have been well-established. It is a leading cause of neonatal mortality and both short-term and long-term morbidity in offspring.2 These adverse health outcomes result in billions of dollars in health care costs annually.2–5 These costs represent only those related to the preterm births that actually occur; billions more are spent on treating women who are thought to be in preterm labor but in fact are not.6 Indeed, the majority of women diagnosed with preterm labor will ultimately deliver at term.
Our inability to accurately diagnose preterm labor in symptomatic women by assessment of history and physical examination understandably has spawned efforts to discover adjunct tests that could provide enhanced and clinically valuable predictive capacity. Such tests theoretically could help us direct therapies to those truly in need and avoid overtreatment of those not truly at risk. Two such tests that have been investigated previously are fetal fibronectin and transvaginal ultrasonographic assessment of cervical length. Both have been shown to have high negative predictive values (NPV) but relatively low positive predictive values (PPV) for soon-to-occur preterm birth.
In this issue of the journal, Wing and colleagues (see page 1183) continue the quest to find a reliable predictor of spontaneous preterm birth that could be used to enhance patient care and control health care costs.7 In their study, they compare the test characteristics of placental alpha microglobulin-1 with those of fetal fibronectin for the prediction of imminent spontaneous preterm birth in women symptomatic for preterm labor. They found that placental alpha microglobulin-1 had an NPV similar to that of fetal fibronectin but a PPV that was statistically significantly higher than that of fetal fibronectin for spontaneous preterm birth within 7 days.
These findings raise the question of whether testing for placental alpha microglobulin-1 should be considered in the context of clinical care, either as a new diagnostic adjunct for those individuals who do not presently use either fetal fibronectin or cervical length, or, at the very least, as a replacement for fetal fibronectin (its comparator in the study) for those who already use fetal fibronectin. The trade name accorded with the bedside test, PartoSure, seems to suggest the confidence we should have in the test with regard to predictive capacity and the benefits its use could engender. At this point, though, it may be more appropriate to think of placental alpha microglobulin-1 as “PartoUnsure,” because there are several aspects of this test and its consequences that we remain unsure about.
We remain unsure about the actual PPV of the test given the very few spontaneous preterm births that occurred. Although the PPV of the placental alpha microglobulin-1 was calculated to be 23% for women with singleton pregnancies, because this point estimate is based on only three preterm births among 13 women with positive placental alpha microglobulin-1 test results, the 95% CI for the PPV descends to a value as low as 5%.
Even if we were more certain about the actual PPV, we would remain unsure about whether the statistical superiority of the PPV means that this test would be better in clinical use, even theoretically, than fetal fibronectin. Fetal fibronectin, given its high NPV, has been theorized to be beneficial in its ability to reduce overtreatment; that is, its use might avert interventions in women who would otherwise have an intervention based on history and physical examination alone.
    loading  Loading Related Articles