Implications of Using a Fetuses-at-Risk Approach When Fetuses Are Not at Risk

Background: Gestational-age-specific rates of postnatal endpoints are sometimes estimated with denominators based on fetuses-at-risk (FAR), rather than live births. However, as infants can only be included in the numerator after they are born alive, interpretation of such rates is problematic. Methods: Using simple algebra it can be shown that, at each gestational week, FAR rates of postnatal endpoints are the product of the conventional risk of outcome among live births and the probability of live birth, which increases from near zero early in gestation to close to one in the final weeks. The consequences of such a pattern of live birth on FAR rates are further illustrated in hypothetical scenarios with known conditions. Results: FAR rates of postnatal endpoints will generally increase towards the end of pregnancy due to the rising probability of live birth, regardless of the 'true' effect of immaturity on risk. In the presence of an exposure that increases the probability of early birth, the same mechanism will cause FAR rates to be higher in the exposed group, even if the exposure has no effect. Conclusions: Gestational-age-specific FAR rates of postnatal outcomes strongly depend on the probability of live birth. Thus, they reflect neither the causal effect of gestational length, nor that of a given exposure. Indeed, if an exposure shortens gestation, FAR rates will be higher in exposed infants even when the exposure has no impact on the outcome under study. These intrinsic limitations should be taken into account when applying FAR analyses to postnatal endpoints.