This study shows that people with type 1 diabetes and type 2 diabetes differ in their associations between biomarkers of inflammation and changes in depressive symptoms. Higher baseline levels of multiple biomarkers were associated with smaller improvements in depressive symptoms in people with type 1 diabetes, but with larger improvements in people with type 2 diabetes. In people with type 1 diabetes, these associations were most pronounced for changes in somatic symptoms, whereas in people with type 2 diabetes, the associations appeared to be driven mainly by changes in cognitive-affective and anhedonia symptoms.

Differential associations between biomarkers of inflammation and improvement of depressive symptoms between diabetes types

Previous studies have suggested that higher levels of several proinflammatory biomarkers such as C-reactive protein, IL-6 and TNF-α may be related to non-response to antidepressant drugs in people with major depressive disorder [13, 14]. We are not aware of any studies that (1) analysed the associations of biomarkers of inflammation with improvement of depressive symptoms upon non-pharmacological treatment; (2) were based on a comprehensive biomarker panel to characterise subclinical inflammation; or (3) addressed this topic in people with diabetes irrespective of diabetes type. Thus, our data are novel and substantially extend the current knowledge in this field.

We found associations between higher biomarker levels and smaller improvements of depressive symptoms only in people with type 1 diabetes, which is characterised by autoimmune disease activity [30]. These biomarkers included chemokines (CCL4, CCL20, MIP-1α) and soluble forms of multiple cell-surface molecules that are involved in proinflammatory signalling and activation of cells from both innate and adaptive immune systems (CD5, CD6, CD244, CDCP1, IL-10RB, IL-15RA, LIF-R, PD-L1, SLAMF1). Of note, CDCP1 is a ligand for CD6, which is expressed on certain T cells and may play a role in cell migration and chemotaxis. Higher levels of many chemokines have been found to be increased in people with depression [17, 31], but their association with treatment response remains unexplored. People with depression also show alterations in several immune cell subsets that are involved in innate and adaptive immune responses [32], but potential links to autoimmune diseases such as type 1 diabetes have not been investigated in this context. Despite these gaps in our knowledge that limit data interpretation, our study identified novel candidate biomarkers for future studies to corroborate our findings.

In contrast, higher biomarker levels were associated with stronger improvements of depressive symptoms in people with type 2 diabetes, which is characterised by subclinical inflammation [33]. These biomarkers included secreted proteins with proinflammatory activity (CSF1, Flt3L, TNF-α), chemokines (CX3CL1, CXCL10, eotaxin, IL-8, MCP-1, MCP-3, MIP-1α) and soluble forms of transmembrane proteins with functions in cell–cell communication and activation of innate and adaptive immune cells (CDCP1, CD8A, CD40, IL-2RB, IL-15RA, IL-18R1, PD-L1). This direction of association was unexpected, and we are not aware of similar findings from other studies. However, it should be noted that previous studies focused on the association between subclinical inflammation and pharmacological treatment, whereas our analysis included only studies that primarily investigated non-pharmacological interventions such as education or cognitive behavioural therapy. It would also be interesting to study the trajectories of both depressive symptoms and subclinical inflammation longitudinally to better understand our findings, but a complete set of biomarker data from the 12-month follow-up was not available in our study.

There was only a small overlap in biomarkers that were associated with improvements in depressive symptoms in opposite directions in type 1 diabetes and type 2 diabetes. One of these proteins was CDCP1, the biomarker with the highest effect size in type 2 diabetes (ESM Table 8). CDCP1 has been found to be associated with a higher risk of all-cause dementia and Alzheimer’s disease [34], but associations with other neurological or psychiatric conditions have not been reported.

So far, it is unclear why the diabetes types show differences in the associations of multiple biomarkers with changes in depressive symptoms. It is possible that they are related to the distinct types of immune activation that characterise type 1 diabetes (autoimmunity) and type 2 diabetes (subclinical inflammation) [33]. As diabetes type had no impact on the protocols and procedures in our studies, methodological or experimental issues can be excluded. It will be of great interest to compare our findings to those from people without diabetes in future studies. It is also unclear why the associations described here differ from associations that we found in our previous cross-sectional analyses to which the baseline data from the DIAMOS, ECCE HOMO and DDCT studies contributed [22]. It will be important to conduct studies that involve assessment of depressive symptoms and biomarkers at multiple timepoints to better elucidate trajectories and cause–effect relationships in the bidirectional interplay between inflammation and depression. At this stage, our observations are hypothesis-generating and may be used to design future replication studies. Comparable studies that consider the association between biomarker levels and the response to antidepressant medication are also urgently needed.

Differential associations with diabetes symptoms clusters

Our findings indicate that the association between higher biomarker levels and lower reduction of depressive symptoms in people with type 1 diabetes was mainly driven by smaller improvements in somatic symptoms, with improvements in cognitive-affective and anhedonia symptoms being more independent of these biomarker levels at baseline. In contrast, biomarker associations with the reduction of depressive symptoms in people with type 2 diabetes were strongest for improvements in cognitive-affective and anhedonia symptoms; higher biomarker levels appeared less relevant for reduction of somatic symptoms.

Previous studies indicated that higher levels of biomarkers of inflammation were mainly associated with somatic symptoms, whereas associations were weaker for anhedonia symptoms and least pronounced for cognitive-affective symptoms [17, 18, 35,36,37,38]. However, these associations have not been compared between people with type 1 diabetes and those with type 2 diabetes.

For people with type 1 diabetes, we identified novel biomarkers that were related to a smaller improvement in depressive symptoms, which could potentially guide treatment decisions. It may be hypothesised that additional anti-inflammatory treatment may help to improve somatic symptoms in particular, in people with type 1 diabetes. The potential of anti-inflammatory drugs for the treatment of depression has been assessed in several studies [39]. It has been proposed that biomarkers of inflammation may identify endotypes of depression that would benefit from attenuating subclinical inflammation [39, 40], so studies that identify these subsets of patients are needed.

In contrast, our data for people with type 2 diabetes and high subclinical inflammation suggest that such patients are good candidates for non-pharmacological therapy approaches to reduce depressive symptoms, particularly to improve cognitive-affective and anhedonia symptoms, whereas those with lower biomarker levels may benefit more from treatment with antidepressant drugs. However, confirmation of our results in other studies is important to corroborate these hypotheses before initiating resource-intensive RCTs. Of note, the heterogeneity of diabetes is not sufficiently captured by the subdivision into type 1 diabetes and type 2 diabetes. Recent studies have suggested the existence of subtypes of type 2 diabetes, of which the severe insulin-resistant diabetes subtype is characterised by the highest inflammatory burden [15] and the highest level of depressive symptoms [41]. Thus, people with severe insulin-resistant diabetes may benefit more from psychotherapy interventions to reduce elevated depressive symptoms than those with other subtypes.

Collectively, our data suggest a role for consideration of biomarkers of inflammation in both precision diabetology and precision medicine for depression. Measurement of these and other biomarkers may be expected to lead to a better understanding of the heterogeneity of both diseases and its implications for more targeted therapies.

Strengths and limitations

Strengths of our study include the large sample size, the individual participant data analysis from three randomised controlled trials based on similar examinations and protocols, the comprehensive biomarker phenotyping, the availability of data for both diabetes types, the analysis of symptom clusters and the adjustment for multiple confounders.

Limitations of our study mainly relate to the generalisability of the results. We analysed the reduction of depressive symptoms in the context of non-pharmacological interventions, so our results cannot be extrapolated to the response to the use of antidepressant drugs, which is most likely determined by other mechanisms and predictors. Our study sample was characterised by elevated depressive symptoms and diabetes distress, but the results may not be generalisable to people with severe major depressive disorder. Detailed data on ethnicity or ancestry were not available, so these could not be considered as potential confounders. In addition, our study cohorts mainly consisted of people of European descent, so the findings may not be generalisable to people of different ethnicity or ancestry.

Conclusions

In our combined analysis of three intervention studies targeting depressive symptoms, higher baseline levels of multiple biomarkers of inflammation were associated with smaller improvements in depressive symptoms in people with type 1 diabetes. This finding appeared to be mainly driven by changes in somatic symptoms. In contrast, higher biomarker levels at baseline were linked with greater improvements in depressive symptoms in people with type 2 diabetes; these were related to greater reductions in cognitive-affective and anhedonia symptoms. Our findings indicate that immune activation may have an impact on recovery from depressive symptoms, and that these effects may not only differ between diabetes types but also be related to different symptom clusters of depression. If replicated by other studies, these results may help develop more targeted treatment approaches for precision medicine in diabetes and depression.