Bias in Retrospective Studies of Spontaneous Abortion Based on the Outcome of the Most Recent Pregnancy

In a retrospective studies of spontaneous abortion, bias can arise if the investigator restricts attention to the outcome of the most recent pregnancy. Bias happens because couples tend to replace spontaneous abortions with other pregnancies until they achieve a live birth, after which they may avoid conception for a long time. Consequently, the proportion of most recent conceptions ending in loss might be very low compared to what would be seen in a prospective study. Also, time since conception is negatively correlated with apparent risk, because most pregnancies which occurred relatively long ago and ended in loss have been replaced by subsequent pregnancies. If the exposure of interest may also have changed over time, then time since conception should be considered a potential confounder. In addition, the consistency with which couples use birth control varies. Some couples are planners who use effective contraception and plan the growth of their family with deliberation; others have most of their pregnancies unintentionally. As a result, the most recent pregnancy is less likely to be a spontaneous abortion for a planner than for a nonplanner. If planners and nonplanners also differ in their exposures, then bias will result. Other factors which affect the interpregnancy interval, such as fecundability, breastfeeding, and desired family size can distort the relative sampling fraction for spontaneous abortions compared to live births and, to the extent that such factors are associated with the exposure, can bias estimates of relative risks.