Background: Childhood abuse is associated with adult suicidal behaviors, partially mediated by mental disorders. However, the direct effect of childhood abuse not mediated by mental disorders is uncertain because the same risk factors serve as mediators and confounders of mental disorders and suicidal behaviors. The aim of the study was to estimate the direct effect of childhood abuse not mediated by mental disorders on suicidal behaviors using marginal structural models. Methods: We used cross-sectional data of Japanese adults in Tokyo and neighboring prefectures (The Japanese Study on Stratification, Health, Income, and Neighborhood). We developed a causal diagram of childhood abuse and suicidal behaviors, and defined exposures (childhood abuse), confounders (age, childhood social characteristics), mediators (mental disorders), risk factors (coping, smoking, alcohol consumption, education, physical health), and outcome (suicidal behaviors). We estimated the direct effect of childhood abuse according to sex using inverse-probability weighting at 2 points and compared to total and direct effects estimated by conventional regression models. Results: Of 1776 male and 2016 female participants, childhood abuse showed significant total effects on suicidal behaviors. Physical abuse was associated with suicidal ideation (risk ratio: males 2.11 [95% confidence interval: 1.59–2.82], females 2.15 [1.69–2.72]), suicidal plan (risk ratio: males 2.18 [1.21–3.92], females 2.14 [1.24–3.74]), and suicidal attempt (risk ratio: males 2.11 [1.14–5.10], females 4.79 [2.93–7.83]) through direct effect. Limitations: Childhood abuse was measured according to self-report of the past experiences. Conclusions: Our study suggests that childhood abuse is associated with adult suicidal behaviors through direct effects not mediated by mental disorders.