Identifying causal associations between women's reproductive traits and risk of schizophrenia: a multivariate validated two-sample Mendelian randomization analysis

Abstract

Background: A significant association between women's reproductive traits and the risk of schizophrenia (SCZ) has been discovered, but the causalities remain unclear. We designed a two-sample univariate Mendelian randomization (MR) study using female-specific SNPs collected from a large-scale genome-wide association study as a genetic tool to explore the causal effect of female reproductive traits on the risk of SCZ, and conducted a multivariate MR study to re-validate the above findings.

Methods: From extensive genome-wide association studies (GWAS) of people with European ancestry (n = 176,881 to 418,758 individuals), summary-level data on five female reproductive variables were extracted. Summary-level information on SCZ was taken from a GWAS meta-analysis involving 320,404 people with European ancestry. The inverse variance weighting estimations for both univariable MR (UVMR) and multivariable MR (MVMR) were presented as the primary results. MR-Egger, weighted median, simple mode, and weighted mode regression methods for UVMR, and MVMR-Egger, MVMR-Lasso, and MVMR-median methods for MVMR were used for sensitivity analyses.

Results: The UVMR produced compelling proof for a connection between genetically predicted later age at first sexual intercourse (AFS) (OR, 0.632; 95% CI, 0.512-0.777; P < 0.01) and decreased SCZ risk. Pleiotropy analysis of the AFS-SCZ association confirmed the robustness of the MR results (P > 0.05). Consistent, substantial causal effects of AFS (OR, 0.592; 95%CI, 0.407-0.862; P < 0.01) on the risk of SCZ were demonstrated after adjusting for body mass index, years of schooling, and smoking initiation using MVMR.

Conclusions: Our findings provide convincing evidence that early AFS is a risk factor for SCZ. SCZ risk may be decreased by raising awareness of reproductive healthcare for women.

Keywords: Causal relationship; Mendelian randomization; Schizophrenia; Women’s reproductive traits.

Fig. 1

Diagram for Mendelian randomization (MR). MR was developed on the premise of three assumptions. First, SNPs designated as instrumental variables (IVs) should be significantly associated with the exposure (Assumption 1). Second, SNPs selected as IVs are required to be independent of confounders (Assumption 2). Third, rather than being directly correlated, the relationship between IVs and the risk of SCZ (outcome) only occurs vis-a-vis reproductive variables (exposure) (Assumption 3). Fourth, the first assumption in the MVMR (Assumption 4) is that genetic variations are caused by one or more of the exposures. BMI, body mass index; YS, years of schooling; PA, physical activity; SI, smoking initiation; AD, alcoholic drinks per week; CI, coffee intake

Fig. 2

The causal effect of genetically predicted age at first sexual intercourse on the risk of SCZ. a Forrest plot. The MR findings of the MR-Egger test and the inverse variance weighted (IVW) method are the significance of the red lines. b MR leave-one-out sensitivity analysis. The leave-one-out sensitivity analysis found no single genetic variant-driven causality. c Scatter plot. The slope of each line in the SNP scatterplot corresponds to the MR effect estimated by each method, and the slopes differ in magnitude but are in the same direction for the five methods. d Funnel plot. The blue line in the SNP funnel plot indicates the IVW estimate and the dark blue line indicates the Mendelian randomization-Egger estimate. No evidence of asymmetry on either side of the blue line of the funnel plot