Perinatal & Pediatric
A Statewide Sibling Study of Cerebral Palsy in California Haoran Zhuo* Haoran Zhuo Tormod Rogne Zeyan Liew
Background and Aim Sibling design has gained popularity in epidemiologic research, but no study has been performed for cerebral palsy (CP). We developed the first sibling study on CP to evaluate the robustness of several identified and suspected risk factors of CP against uncontrolled confounding.
Methods Within a cohort of singleton births of 2007-2015 in California and CP diagnosis ascertained from a statewide registry, we identified discordant siblings (1213 CP and 1544 non-CP siblings) born to the same biological mothers. We performed logistic regression on the full cohort and conditional logistic regression on the sibling data to evaluate the associations between CP and a priori selected maternal factors (e.g. pre-pregnancy BMI, cigarette smoking, gestational weight gain), pregnancy/obstetric complications (e.g. preeclampsia, gestational diabetes, infections), and neonatal adverse birth outcomes (e.g. preterm birth, low birth weight, low Apgar score). Sensitivity tests were performed to evaluate the impacts of the carryover effect and generalizability when using sibling design.
Results The association between a strong risk factor, preterm birth, and CP was robust in the full cohort (OR=4.6, 95% CI 4.3-4.9) and the sibling analysis (OR=3.3, 95% CI 2.6-4.2). Similar findings were also on other investigated neonatal factors and maternal obstetric complications. On the contrary, the observed associations between maternal pre-pregnancy BMI, gestational preeclampsia, and gestational diabetes in the cohort (ORs range 1.2-1.4) were attenuated towards the null among siblings. Although maternal insufficient gestational weight gain and cigarette smoking were consistently associated with CP in the sibling data, our sensitivity analysis indicated that these lifestyle factors may be susceptible to the carryover effect. Other assumptions, such as the generalizability of the sibling analysis to the full cohort, did not impact our findings.
Conclusions Several investigated maternal and neonatal factors of CP remain robust in the sibling design, while uncontrolled confounding needs to be considered for pre-pregnancy BMI and some pregnancy complications. We demonstrated that if appropriately applied, a sibling design is useful to enhance causal inference of the etiological research findings of CP.