Associations of C-reactive protein and psychological distress are modified by antidepressants, supporting an inflammatory depression subtype: findings from UKHLS

Publication type

Journal Article

Published in

Brain, Behavior, and Immunity


Amanda Hughes and Meena Kumari

Publication date


Background: Clinical evidence increasingly suggests inflammation may be important specifically for an etiologically distinct depression subtype, characterised by resistance to antidepressant medication. However, epidemiological investigations of the relationship of inflammation with depression and psychological distress have failed to acknowledge these developments, which may have resulted in bias or masking of associations driven by the subtype. This may have contributed to inconsistent results in epidemiological studies, and equivocal support for an inflammation-depression link. Methods: An antidepressant-resistant, inflammatory depression subtype would result in stronger associations of depressive symptomatology with inflammation among antidepressant users than non-users, due to over-representation of subtype individuals among antidepressant users experiencing severe or persistent symptoms. We investigate, in a sample of 10,363 UK adults aged 16–98, modification by antidepressants of cross-sectional and longitudinal associations between C-reactive protein and psychological distress (General Health Questionnaire score, GHQ). We account for confounding by age, gender, income, inflammatory somatic illness, body mass index and, in longitudinal models, baseline psychological distress. Sensitivity analyses consider smoking, ethnicity, and other medications. Results: Robust associations of log-CRP and GHQ were seen for antidepressant users but not for non-users in both cross-sectional (coeff: 0.54, p = 0.01 vs 0.06, p = 0.28) and longitudinal models (coeff: 0.57, p = 0.006 vs 0.04, p = 0.39 two waves post-baseline). Cross-sectional associations were strongest for tricyclic users, and longitudinal associations strongest for SSRI users. In multilevel, repeated-measures longitudinal models, associations for antidepressant users peaked two waves after baseline before declining. Conclusions: Results suggest evidence for existence of an inflammatory depression subtype. Previous studies’ exclusion of antidepressant users and failure to consider interactive effects may have obscured associations driven by the subgroup. Follow-up work is now needed in community samples with clinical depression measures and prescription histories, to further elucidate the mechanisms involved.

Volume and page numbers

66, 89-93





Psychology, Medicine, Well Being, Health and Biology


Open Access; Open Access funded by Economic and Social Research Council; Under a Creative Commons license