C-reactive protein and post-stroke depressive symptoms

Our study aimed to explore the association between serum C-reactive protein (CRP) and post-stroke depressive symptoms. We prospectively recruited 572 patients with ischemic stroke or transient ischemic attack in whom serum CRP level was measured within 48 h after stroke onset. Depressive symptoms were assessed at day 8 and 3 months after stroke in 405 and 306 patients, respectively. Patients with greater depressive symptoms at day 8 and patients with greater depressive symptoms 3 months after stroke had higher CRP level (median: 7.9 vs 4.3 mg/L, P < 0.01 and 6.7 vs 3.4 mg/L, P = 0.01, respectively). In the univariate analysis, CRP > 9.2 mg/L was associated with depressive symptoms at day 8 (OR: 2.06, 95%CI: 1.30–3.28, P < 0.01) and CRP > 4.3 mg/L was associated with depressive symptoms 3 months after stroke (OR: 1.79, 95%CI: 1.06–3.02, P = 0.03). In the multivariate analysis, higher CRP level was related to depressive symptoms at day 8 (OR: 2.23, 95%CI: 1.28–3.90, P < 0.01), but not depressive symptoms 3 months after stroke (OR: 1.13, 95%CI: 0.59–2.17, P = 0.71). In conclusion, higher levels of CRP are associated with greater depressive symptoms at day 8 after stroke, but their effects on depressive symptoms 3 months after stroke are less significant.

used to compare continuous variables between groups. Logistic regression was used to determine the predictors of functional outcome. Variables with P < 0.05 in the univariate analysis were included in a multivariate analysis. The Box-Tidwell test was used to check the linearity of the logit for the continuous independent variables in logistic regression analysis. The receiver operating characteristic curves were used to identify an optimal cut-off level of CRP that differentiates patients with greater depressive symptoms from those with lower depressive symptoms. Since CRP level could raise within 48 h after stroke onset and its level measured within 12 h might be lower than its level measured between 24-48 h 12 , two sensitivity analyses were performed. In the first, the patients in whom the CRP level was measured within 12 h after stroke were excluded. In the second analysis, only patients with CRP measurement within 24 h after stroke were included. The calculations were performed using the program STATISTICA for Windows (version 12.5, Statsoft, Poland).

Results
Of 750 patients that participated in the PROPOLIS study, 572 patients had an ischemic stroke or TIA and the CRP measurement within 48 h after stroke onset. After the exclusion of patients who died or were not able to perform PHQ-9, depressive symptoms were assessed in 405 patients at day 8 and in 306 patients 3 months after stroke (Fig. 1).
Compared to patients with lower depressive symptoms, patients with greater depressive symptoms were more often women; suffered from hypertension, previous stroke, and pre-stroke cognitive decline; took antidepressants before index stroke; and had higher pre-stroke NPI score (Table 1). Serum CRP level was higher in patients with greater depressive symptoms.
Depressive symptoms 3 months after stroke onset. Of  Compared to patients with lower depressive symptoms, patients with greater depressive symptoms more often suffered from hypertension, diabetes mellitus, and post-stroke delirium; took antidepressants during 3 months after stroke onset; and had higher pre-stroke NPI score (Table 2). Serum CRP level was higher in patients with greater depressive symptoms.
In the univariate analysis, CRP level above 4.3 mg/L was associated with greater depressive symptoms (OR: 1.79, 95%CI: 1.06-3.02, P = 0.03). This association was nonsignificant in the multivariate analysis adjusted for hypertension, diabetes mellitus, delirium, the use of antidepressant and the NPI score www.nature.com/scientificreports www.nature.com/scientificreports/ When patients treated with anti-depressants were excluded from the analysis, CRP was still associated with depressive symptoms in the univariate (OR: 2.00, 95%CI: 1.10-3.62, P = 0.02), but not in the multivariate analysis (OR: 1.04, 95%CI: 0.50-2.16, P = 0.92).

Discussion
Our study revealed that higher levels of CRP are associated with greater depressive symptoms at day 8 after stroke, but their effects on depressive symptoms 3 months after stroke are less significant.
A few studies have examined the relationship between circulating CRP and risk of post-stroke depression. These studies yielded conflicting results. Jimenez et al. measured the serum CRP level in 134 patients with first-ever ischemic stroke 27 . Blood was collected at discharge (day 7 ± 2) and 1 month after stroke. About 19% of patients were diagnosed as having major depression at discharge according to DSM-IV criteria and 22% of patients had major depression 1 month after stroke. The authors did not find any association between CRP level and the risk of post-stroke depression. The study of Yang et al. included 226 ischemic stroke patients 28 . CRP level was measured within 24 h after stroke onset. Six months after stroke major depression was diagnosed in 30.5% of patients. In this study, serum CRP level above 0.85 mg/dL was associated with the increased risk of depression after adjusting for potential confounders. Similarly, Cheng et al. found that higher CRP level measured with 24 h after stroke onset predicts the increased risk of depression 1 year after stroke 29 .
The contribution of inflammatory factors to the pathogenesis of post-stroke depressive symptoms might be dependent on time after stroke. Depressive symptoms that occur very early after stroke onset could be a part of so-called sickness behaviour. Sickness behaviour is a set of behavioural and motivational changes triggered by acute infection or tissue injury and includes anhedonia, hyperalgesia, fever, anorexia, sleepiness, anxiety, and disinterest in social interactions 30,31 . Animal studies have demonstrated that systemic inflammation might induce both sickness behaviour and depressive-like symptoms 30 . Multiple symptoms, for example weight loss, anorexia, fatigue, hyperalgesia, anhedonia, anxiety, and neurocognitive symptoms are shared by both sickness behaviour and depression. Clinical studies involving cancer patients treated with interferon-alpha (INFα) shed light on cytokine-induced sickness behaviour and depressive symptoms. These studies have shown that somatic and Pre-stroke total NPI score, median (IQs)** 4.5 (0-14.5) 0 (0-6) <0.01 Pre-stroke NPI score for depression, median (IQs)** 0 (0-2) 0 (0-0) <0.01 Pre-stroke use of antidepressants 6 (5. www.nature.com/scientificreports www.nature.com/scientificreports/ vegetative symptoms appeared within 2 weeks of INFα therapy 32 . In contrast, mood and cognitive symptoms appeared later during INFα therapy and were more apparent in patients who developed major depression. The hypothesis that early-onset post-stroke depressive symptoms might be related to sickness behaviour is supported by the observation that stroke patients with early-onset (in-hospital) depression had higher frequency of vegetative (anxiety, loss of energy, morning depression, early awakening, weight loss) and melancholic (loss of interest, depressed mood, psychomotor retardation) symptoms compared to patients with late-onset depression 33 . Alternatively, early-onset post-stroke depression might represent a specific phenotype of depression with dominant somatic (vegetative) symptoms rather than sickness behaviour. Further studies with dimensional analyses of specific clusters of neuropsychiatric and somatic symptoms are needed to better characterize the relationship between sickness behaviour and post-stroke depressive symptoms. Differentiation of early depressive symptoms related to sickness behaviour from those predicting major depression could have therapeutic implications. In patients treated with INFα, mood and cognitive symptoms were more responsive, whereas vegetative symptoms, such as anorexia or fatigue, were less responsive to paroxetine treatment 32 .
In the univariate analysis, higher CRP was associated with greater depressive symptoms 3 months after stroke. This association was, however, non-significant after adjusting for potential confounders. Inflammatory markers decline gradually after stroke 13 . For this reason, the remote effect of inflammation on depressive symptoms could be weaker than in the acute phase of the stroke. Moreover, the persistence of depressive symptoms beyond the acute phase of stroke might require additional vulnerabilities, including unresolved inflammation, genetic predisposition, or alteration in neuronal networks responsible for mood regulation. We cannot exclude the possibility that our study had too low statistical power to detect an association between CRP and depressive symptoms occurring 3 months after stroke. It should also be noted that inflammatory markers might be selectively associated with only specific dimensions of depression. In MDD, circulating inflammatory markers are especially linked to atypical depression characterized by increased appetite and weight gain 34 .
There are several limitations to our study. First, the PROPOLIS was designed to determine frequency, predictors and clinical consequences of post-stroke delirium. Depressive symptoms were considered as a secondary end-point of the study and statistical power was not calculated a priori to check if the study was able to detect an association between CRP and depressive symptoms. Second, no formal psychiatric diagnosis of depression was made. Instead, we used a validated questionnaire to assess depressive symptoms. Third, the CRP level was measured only once. Repeated measurements of inflammatory parameters at different time points after stroke could give better insight into relationships between inflammation and depression. Fourth, about 27% of patients  www.nature.com/scientificreports www.nature.com/scientificreports/ who were examined in acute stroke did not attend a control visit. These patients were older and had a more severe neurological deficit on admission. Fifth, about 13% of our patients took anti-depressants 3 months after stroke. CRP can interact with anti-depressive medication. In MDD patients, the elevated CRP level was associated with treatment resistance 35 .
There are also some advantages to our work. These include the prospective design of the study and assessment of pre-stroke psychiatric symptoms and cognitive decline.
Observational studies are not able to demonstrate if an association between CRP and depressive symptoms is causative. To verify this association, we need an interventional study that will examine if the anti-inflammatory treatment will reduce post-stroke depressive symptoms. From a clinical perspective, depression that occurs early after stroke (e.g., during the first 2 weeks after stroke onset) is associated with long-term poor functional outcome 36,37 . Thus, anti-inflammatory strategies attenuating systemic inflammatory reaction could have a beneficial effect on stroke outcome.
In conclusion, higher levels of serum CRP are associated with early depressive symptoms after ischemic stroke.

Data availability
The datasets generated during and/or analysed during the current study are available from the corresponding author on reasonable request.