Associations of androgens with depressive symptoms and cognitive status in the general population

Hanna Kische, Stefan Gross, Henri Wallaschofski, Hans Jörgen Grabe, Henry Völzke, Matthias Nauck, Robin Haring

Research output: Contribution to journalArticleResearchpeer-review

19 Citations (Scopus)


Objectives: Associations between androgens and depressive symptoms were mostly reported from cross-sectional and patient-based studies. Study design/main outcome measures: Longitudinal data from 4,110 participants of the Study of Health in Pomerania were used to assess sex-specific associations of baseline total and free testosterone, androstenedione and sex hormone-binding globulin with incident depressive symptoms and cognitive status at 5- and 10-year follow-up. Results: Despite sex-specific differences in depressive symptoms prevalence at baseline (women: 17.4%, men: 8.1%), cross-sectional analyses showed no associations between sex hormones and depressive symptoms. In age-adjusted longitudinal analyses, total testosterone was associated with incident depressive symptoms (relative risk at 5-year follow-up: 0.73, 95% confidence interval: 0.58-0.92). Similarly, age-adjusted analyses showed a positive association between sex hormone-binding globulin and cognitive status in men (β-coefficient per standard deviation: 0.44, 95% confidence interval: 0.13-0.74). In women, ageadjusted associations of androstenedione with baseline depressive symptoms (relative risk: 0.88, 95% confidence interval: 0.77-0.99) were found. None of the observed associations remained after multivariable adjustment. Conclusions: The present population-based, longitudinal study revealed inverse associations between sex hormones and depressive symptoms. However, the null finding after multivariable adjustment suggests, that the observed associations were not independent of relevant confounders including body mass index, smoking and physical inactivity. Furthermore, the low number of incident endpoints in our non-clinical population-based sample limited the statistical power and reduced the chance to detect a statistically significant effect.

Original languageEnglish
Article numbere0177272
Number of pages13
JournalPLoS ONE
Issue number5
Publication statusPublished - 1 May 2017

Cite this