Objectives: This study aimed to explore the association between transitions in anxiety symptoms and the risk of diabetes in women, using longitudinal data. Study design: This longitudinal study measured diabetes, and transitions in anxiety symptoms, using validated instruments. Methods: Data obtained by the Mater-University of Queensland Study of Pregnancy were analysed. Anxiety was measured using the Delusion Symptoms States Inventory (DSSI). To examine possible transitions over different time periods, anxiety was grouped into four categories and assessed at different phases over a 21-year period. Results: Three hundred and one women reported diabetes 21 years after the index pregnancy. Almost half of the women who reported anxiety symptoms continued to report these at a subsequent follow-up (FU) phase. About 1 in 10 women who had not reported anxiety symptoms at 5-year FU did so at the subsequent 14-year FU. In prospective analyses, we did not find significant association of diabetes with negative transition (no anxiety to anxiety at subsequent phase) or with positive history of anxiety symptom, but an increasing risk was evident. Women with persistent symptoms had a 1.85-fold greater risk of diabetes (95% CI: 1.18-2.90). The cross-sectional analysis did not produce significant results. Conclusions: Despite some limitations, this study provides insight into the long-term association between events of anxiety and the risk of diabetes across the reproductive life of women. However, the evidence is not strong enough to support a direct effect of anxiety in causing diabetes.