Category: Maternal Fetal Medicine
Introduction: Studies of interpregnancy interval and pregnancy outcomes should only consider characteristics measured at the time of the first (index) delivery as potential confounders, as this time point precedes the interpregnancy interval. However, studies often only have information on these variables measured at the subsequent pregnancy. Because these factors are on the causal pathway between exposure and outcome, adjustment may induce mediator over-adjustment bias and collider stratification bias. The objective of this study was to quantify the magnitude and direction of bias induced by controlling for time-varying confounders measured at the subsequent pregnancy in between-woman analyses of interpregnancy intervals and adverse pregnancy outcomes.
Methods: We analyzed all pregnancies in British Columbia from 2004-2014 to women with ≥2 singleton pregnancies (n=121,143 women). We used log binomial regression to estimate risk ratios (RRs) at short (<6, 6-11, 12-17-month) interpregnancy intervals vs the reference 18-23-month interval for 5 outcomes: adverse fetal-neonatal composite outcome; small-for-gestational age (SGA), preterm delivery, and spontaneous and indicated preterm delivery, separately. We estimated RRs from two model sets adjusted for confounders (maternal age, low neighborhood income, body mass index, and smoking) measured at the 1) index delivery and 2) subsequent pregnancy. We calculated relative percent change in RR between models, and considered changes <5% minimal, 5-<10% modest, and ≥10% substantial.
Results: For intervals <6 months compared with the 18-23-month reference interval, RRs for all outcomes were attenuated after adjustment for confounders measured at the index delivery. Adjustment for factors measured at the subsequent pregnancy minimally biased RRs for the fetal-neonatal composite and indicated preterm delivery toward the null, and minimally biased RRs for SGA, preterm delivery, and spontaneous preterm delivery away from the null.
Conclusion: We found that adjusting for time-varying confounders ascertained at the time of subsequent delivery introduced minimal bias in analyses of interpregnancy intervals and adverse pregnancy outcomes.
Jennifer Hutcheon– Associate Professor, University Of British Columbia
Wendy Norman– Associate Professor, University of British Columbia
Jessica Liauw– Maternal-Fetal Medicine Fellow, University of British Columbia
Talshyn Bolatova– Research Assistant, University of British Columbia
Katherine Ahrens– Assistant Research Professor, University of Southern Maine