Intra- and interobserver agreement and statistical clustering of placental histopathologic features relevant to preterm birth.

Abstract

Many previous studies of agreement in identifying placental histopathologic lesions have been based on small sample sizes, and none has examined whether individual histologic features cluster robustly together within and between observers. We studied 767 placental specimens from case-control studies of preterm birth and preeclampsia nested within a prospective cohort of pregnant women recruited from 4 large Montreal maternity hospitals. The specimens were fixed, embedded, stained, and examined using a standardized protocol; a 10% random sample (n = 81) was then blindly reexamined at least 6 months later by the same pathologist and a second pathologist. Intra- and interobserver agreement were high (kappa > or = 0.50) for membrane inflammation, funisitis, and umbilical cord vasculitis, and these 3 features were robustly clustered statistically, consistent with an underlying mechanism of ascending infection. Agreement and clustering were also high or moderate for features of placental underperfusion: infarction, decidual vasculopathy, and syncytial knotting. Our results should help researchers to interpret future findings relating placental histopathology to preterm birth, preeclampsia, and other adverse pregnancy outcomes, and to their etiologic determinants and causal pathways.