Page et al propose a simple measure for determining the optimal timing of delivery among twins. The index proposed, namely, the composite rate of fetal and infant death at any gestational week, involves a summation of the stillbirth rate over the previous gestational week and the infant death rate at the gestational week in question.

Significant problems with the proposed index include a lack of clarity regarding the denominator for the composite fetal and infant mortality rate. Although both stillbirths and deaths after birth are relevant in the context of obstetric intervention, summing the stillbirth rate (an actuarial incidence density rate expressed as week ^–1 ) and the infant mortality rate (a dimensionless cumulative incidence) does not appear to be conceptually or mathematically correct. Also, the summation of stillbirth and infant mortality rates (without consideration of the relative frequency of stillbirth and infant deaths) results in a heterogeneous composite that is more heavily weighted in favor of the infant death rate at early gestation and increasingly weighted in favor of the stillbirth rate at late gestation. A logical problem with the index is highlighted by the extreme example of immediate delivery of all fetuses at 21 weeks’ gestation. With all fetuses delivered at 21 weeks dying in infancy following such hypothetical intervention (ie, 1000 deaths per 1000 live births), the addition of the fetal deaths from the previous week will lead to the numerator exceeding the denominator.

Numerous studies in the recent obstetric literature have attempted to identify the optimal timing of delivery in specific high-risk and low-risk pregnancies. Randomized trials on this topic have consistently estimated the cumulative incidence of the primary outcome among all trial subjects in the immediate delivery vs the expectant management arm. Nonexperimental studies, on the other hand, have contrasted cumulative incidence or incidence density rates of the primary outcome in various subsets of the study population.

The diversity of indices used to assess optimal timing of delivery in nonexperimental studies suggests that the correct epidemiologic index for assessing the primary outcome is not intuitively obvious. In our considered opinion, nonexperimental studies on timing of delivery should attempt to replicate the clinical situation that confronts patients and care providers, emulate traditions within modern obstetrics so that all relevant severe morbidity and mortality is contrasted, and draw on the paradigm of the randomized trial to correctly design the form of the end result (ie, the epidemiologic measure expressing outcome frequencies in the immediate delivery and expectant management groups).