Problems and Paradoxes in Perinatal Epidemiology
The so-called birth weight paradox, in which risk factors appear to have protective effects on perinatal mortality at low birth weights, has perplexed the perinatal epidemiologic community for decades. This paradox appears consistently with several risk factors for mortality, including smoking and socioeconomic status, and appears when gestational age replaces birth weight. Risk factors can appear protective for preterm births even though they convey increased risk overall. In the last several years, several authors have proposed explanations for the paradox based on different causal frameworks, and methods for design and analysis of perinatal studies to avoid the paradox. Fundamental to these explanations is a precise research question and careful definition of the cohort under study. In this symposium, we first describe the paradox and provide a historical overview. We then describe methods for defining the problem and setting up the cohort, and methods for analysis, that avoid the paradox. Next, we give an example of the biases that arise with inappropriate designs and analyses in the study of seasonal effects on preterm birth. It can be shown that the use of live births, as opposed to foetuses at risk, can give rise to bias. We conclude with a summary and by showing the parallels between the perinatal paradox and other similar paradoxes in epidemiology.