Association between high serum total bilirubin and post-stroke depression


Correspondence: Wai Kwong Tang, MD, Department of Psychiatry, Shatin Hospital, 33 A Kung Kok Street, Ma On Shan, Shatin, NT, Hong Kong. Email:



High serum bilirubin predicts depression in non-stroke subjects, but it is unknown whether it also predicts post-stroke depression (PSD). This study examined the association between the risk of PSD and bilirubin level.


Six hundred and thirty-five patients with acute ischemic stroke in Hong Kong were recruited. Serum total bilirubin, alanine transaminase and alkaline phosphatase levels were measured in all patients during their hospital stay. A psychiatrist gave the Structured Clinical Interview for DSM-IV to all patients 3 months after the index stroke, with 61 patients diagnosed with PSD: 27 with major depression, 24 with minor depression and 10 with dysthymia.


In the full sample, the 25%, 50% and 75% percentile bilirubin levels were 7.0, 10.0 and 14.0 μmol/L, respectively. Significant differences were found between the PSD and non-PSD groups in terms of bilirubin level (P = 0.006). In post-hoc comparisons, the proportion of patients with bilirubin ≥14.1 μmol/L was significantly higher in the PSD group (37.7% vs 19.7%, P = 0.001). In the final regression model, bilirubin level (≥14.1 μmol/L) remained a significant independent predictor of PSD, with an odds ratio of 2.4.


High bilirubin level is associated with PSD. Further investigations are needed to clarify the underlying pathophysiological link between bilirubin level and PSD.

DEPRESSION IS THE most common and serious affective disorder following stroke,[1, 2] occurring in approximately 33% of stroke patients.[3] Post-stroke depression (PSD) has been associated with both poor functional outcome[4] and higher mortality.[5]

The pathogenesis of PSD remains unclear. In non-stroke subjects, depression is associated with an increase in the serum levels of some inflammatory cytokines, such as interleukins (IL)[6] and C-reactive protein (CRP).[7] Other molecular markers, including brain-derived neurotrophic factor (BDNF) and leptin, have been associated with the development of depression in clinical and experimental studies.[8, 9] Previous studies have examined the common biomarkers of depression in stroke, including inflammatory molecular markers (IL-1β, IL-6, IL-18, intracellular adhesion molecule 1, tumor necrosis factor-α, leptin and high-sensitivity CRP), neurotrophic factors (BDNF) and vitamin B-related metabolites (homocysteine). The results were negative except for IL-18, leptin and homocysteine.[10-12]

Bilirubin, an efficient antioxidant, is the end product of the heme cleavage pathway, which is catalyzed by heme oxygenase and biliverdin reductase.[13] Bilirubin level may be related to the risk and outcome of stroke. Serum bilirubin is reported to be negatively correlated with carotid atherosclerosis,[14] whereas a high bilirubin level has been associated with stroke severity[15] and poor prognosis for ischemic stroke.[16]

The role of serum bilirubin level in depression in non-stroke subjects has been explored, with low nocturnal bilirubin level found to be associated with winter seasonal depression.[17] Increased concentration of bilirubin oxidative metabolites (biopyrrins) was found in the urine of patients with psychiatric disorders, and biopyrrin level was found to be correlated with Hamilton Depression Rating Scale score.[18] To date, however, no study has examined the relationship between serum bilirubin level and PSD. The lack of data in this field provided the impetus for the study reported herein.



Patients with first-ever or recurrent acute ischemic stroke admitted to the Acute Stroke Unit of the Prince of Wales Hospital between June 2004 and November 2008 were screened for study entry. The Prince of Wales Hospital is a university-affiliated general hospital serving a population of 800 000 in Hong Kong. The study's inclusion criteria were: (i) Chinese ethnicity; (ii) Cantonese as the primary language; (iii) age ≥18 years; (iv) well-documented (clinical presentation and computed tomography of the brain) first or recurrent acute stroke occurring within the 7 days before admission; (v) magnetic resonance imaging (MRI); (vi) Mini-Mental State Examination (MMSE) score ≥20;[19] and (vii) the ability and willingness to give informed consent. The exclusion criteria included: (i) transient ischemic attack; (ii) history of any central nervous system disease such as dementia, tumor, trauma or hydrocephalus; (iii) history of depression or other psychiatric disorders before the index stroke; (iv) severe aphasia (best language score on the National Institute of Health Stroke Scale [NIHSS] on admission ≥2), severe visual or auditory impairment; (v) history of liver disease or physically unfit for interview; and (vi) recurrent stroke before 3-month interview.

The study protocol was approved by the Clinical Research Ethics Committee of the Chinese University of Hong Kong. All participants signed a consent form.

Collection of demographic, clinical and radiological data

A trained research nurse, who was blind to the psychiatrist's diagnoses, collected the demographic data (age and sex) and assessed stroke severity using the NIHSS[20] within 2 days of admission. A research assistant evaluated all subjects with the MMSE, Lubben Social Network Scale (LSNS),[21] Modified Life Event Scale (MLES)[22] and Barthel index (BI)[23] 3 months after the onset of the index stroke. Cronbach's α for the NIHSS, MMSE and BI were 0.92, 0.86 and 0.93, respectively.[19, 24, 25] The LSNS is a composite social network scale that was designed for use in the elderly. It measures the level of social support that patients receive and their social interactions with relatives and friends. The LSNS has been translated into Chinese and validated in the Hong Kong elderly, with a Cronbach's α of 0.86.[26] The MLES records the presence of 18 adverse and distressing life events in the past 6 months. The score indicates the total number of adverse events encountered by subjects. Data on statin treatment and living environment were also recorded.

MRI with diffusion-weighted imaging (DWI) and conventional sequences, including a gradient echo (blood product-sensitive) sequence, was performed on each participant with a 1.5-T system (Sonata, Siemens Medical, Erlangen, Germany) within 7 days of admission. The locations and volumes of acute brain infarcts were recorded. Their total area on DWI was measured using manual outlines with restricted water diffusion identified on DWI with b values of 1000. The total volume was calculated by multiplying the total area by the sum of the slice thickness and the gap.

Assessment of PSD

Three months after the onset of the index stroke, two psychiatrists who were blind to the subjects' laboratory results gave them the Chinese version of the Structured Clinical Interview for DSM-IV (SCID-DSM-IV).[27, 28] Inter-rater reliability was examined in 25 patients and the kappa was 0.84. The timing of the assessment was chosen to avoid the period of transient emotional adjustment to the disability caused by the stroke.[29]

Laboratory test

Blood samples were obtained within 24 h of the subjects' index admission. Serum total bilirubin was determined with the diazotization method using DP Modular Analytics (Roche Diagnostics, Indianapolis, IN, USA). Alanine transaminase and alkaline phosphatase were recorded in IU/L and bilirubin was recorded in mg/dL and divided into four quartiles (≤8.0 μmol/L, 8.1–10.0 μmol/L, 10.1–14.0 μmol/L and ≥14.1 μmol/L), as the raw bilirubin data were skewed. The number of patients in each quartile was recorded; because of ties, the number of patients in each quartile was not even.

Statistical methods

The distribution of the data was examined using the Kolgomorov–Smirnov test. The demographic and clinical variables (age, sex, previous stroke, and NIHSS, MMSE, LSNS and MLES scores) and bilirubin level quartiles of patients with (PSD group) and without PSD (non-PSD group) were compared using the χ2 test, t-test and Mann–Whitney U test, as appropriate. Risk factors with P < 0.10 were then analyzed via multivariate logistic regression. In this analysis, the odds ratio of quartiles of bilirubin levels[15] and other independent risk factors was interpreted as the risk of subsequent PSD when all other risk factors were held constant. All statistical tests were performed using SPSS for Windows (Release 14.0; SPSS, Chicago, IL, USA). The level of significance was set at 0.05.


A total of 3976 consecutively admitted patients with ischemic stroke were screened, and 635 met the entry criteria (Fig. 1). Patients excluded from the study had a higher mean (±SD) age (68.4 ± 12.3 years vs 65.8 ± 11.3 years; t = 4.314, d.f. = 1418, P < 0.001), were more likely to be female (47.6% vs 37.5%; P < 0.001) and had a higher NIHSS mean score (6.2 ± 5.1 vs 4.3 ± 3.3; t = 8.314, d.f. = 1419, P < 0.001).

Figure 1.

Study recruitment profile. Not assessed at 3-month post-stroke follow up (n = 680, 17.1%): loss to follow up (n = 203, 5.1%), refused participation (n = 142, 3.6%), physical frailty (n = 78, 2.0%), history of central nervous system diseases (n = 61, 1.5%), aphasia (n = 41, 1.0%), history of dementia (n = 40, 1.0%), recurrent stroke (n = 36, 0.9%), severe auditory or visual impairment (n = 27, 0.7%), deceased (n = 24, 0.6%), non-Chinese ethnicity or non-Cantonese speaking (n = 22, 0.6%), transient ischemic attack (n = 6, 0.2%).. MMSE, Mini-Mental State Examination; MRI, magnetic resonance imaging.

Of the 635 patients who formed the study sample, 61 (32 men, 29 women) were diagnosed with PSD: 27 with major depression, 24 with minor depression and 10 with dysthymia. None of them received antidepressant treatment. Patient demographic and stroke-related data stratified by PSD status are listed in Table 1. The PSD and non-PSD groups did not differ in terms of age or previous stroke, but the PSD group had a lower level of social support, poorer cognitive function, more adverse life events, more severe stroke and lower level of functioning. The PSD group also had more female subjects. A significant difference in bilirubin level was found between the two groups (P = 0.006). In post-hoc comparisons, the proportion of patients in the highest quartile (≥14.1 μmol/L) was significantly higher in the PSD group (37.7% vs 19.7%, χ2 = 15.049, d.f. = 1, P = 0.001; Table 1).

Table 1. Subject characteristics vs PSD status

Yes (n = 61)

mean ± SD or n (%)

No (n = 574)

mean ± SD or n (%)

  1. t-test.; χ2 test; §Fisher's exact test; Mann–Whitney U-test. ††Post-hoc comparison; level of significance adjusted to 0.0125 (χ2, d.f.).
  2. ALP, alkaline phosphatase; ALT, alanine transaminase; BI, Barthel index; LSNS, Lubben Social Network Scale; MLES, Modified Life Events Scale; MMSE, Mini-Mental State Examination; NIHSS, National Institute of Health Stroke Scale; PSD, post-stroke depression.
Age (years)66.9 ± 11.065.6 ± 11.3t = 0.8450.40
Female sex29 (47.5)209 (36.4)χ2 = 2.9150.09
Institutionalized§3 (4.9)18 (3.1) 0.44
Hypertension40 (65.6)405 (70.8)χ2 = 0.7220.40
Diabetes mellitus21 (34.4)186 (32.5)χ2 = 0.0970.76
Hyperlipidemia35 (57.4)359 (62.7)χ2 = 0.6520.42
Ischemic heart disease§4 (6.6)41 (7.2) 0.86
Previous stroke13 (21.3)114 (19.9)χ2 = 0.0690.80
On statins6 (9.8)68 (11.8)χ2 = 0.2210.64
MMSE score25.6 ± 3.126.6 ± 2.8t = −2.6930.01
NIHSS total score5.3 ± 3.64.2 ± 3.3t = 2.3340.01
BI score18.7 ± 2.319.2 ± 1.8t = −2.2360.02
LSNS score26.3 ± 8.130.6 ± 8.3t = −3.919<0.001
MLES score2.3 ± 1.31.8 ± 0.9t = 3.7520.006
Volume of acute infarcts (mL)23.5 ± 5.923.2 ± 16.2z = −0.1040.92
Location of acute infarcts    
Frontal lobe§6 (9.8)42 (7.3) 0.44
Parietal lobe§1 (1.3)41 (7.1) 0.17
Temporal lobe§1 (1.3)23 (4.0) 0.72
Occipital lobe§3 (4.9)10 (1.7) 0.12
Subcortical white matter14 (23.0)166 (28.9)χ2 = 0.8310.45
Basal ganglia8 (13.1)55 (9.6)χ2 = 0.8610.36
Infratentorial11 (18.0)106 (18.5)χ2 = 0.0000.98
ALT (IU/L)23.5 ± 16.523.2 ± 16.3t = 0.1490.88
ALP (IU/L)78.3 ± 24.278.0 ± 27.8t = 0.0050.94
Bilirubin  χ2 = 12.5830.006
Quartile 112 (19.7)152 (26.5)χ2 = 1.3340.28††
Quartile 218 (29.5)166 (28.9)χ2 = 0.0090.92††
Quartile 38 (13.1)143 (24.9)χ2 = 4.2350.04††
Quartile 423 (37.7)113 (19.7)χ2 = 10.6370.001††

The following variables were entered into the regression model: bilirubin level, female sex, and LSNS, MLES, MMSE, NIHSS and BI scores. Bilirubin level (≥14.1 μmol/L) remained a significant independent predictor of PSD, with an odds ratio of 2.4 (Table 2).

Table 2. Multivariate logistic model of the clinical determinants of PSD
VariableOdds ratio (95%CI)Pa
  1. aLogistic regression. BI, Barthel index; CI, confidence interval; LSNS, Lubben Social Network Scale; MLES, Modified Life Events Scale; MMSE, Mini-Mental State Examination; NIHSS, National Institute of Health Stroke Scale; PSD, post-stroke depression.
Bilirubin (quartile 4)2.383 (1.303–4.357)0.005
MLES score1.649 (1.274–2.135)<0.001
LSNS score0.948 (0.917–0.980)<0.002
Female sex 0.06
NIHSS score 0.12
MMSE score 0.29
BI score 0.58
R square 0.15


To the best of our knowledge, this is the first paper to report an association between bilirubin level and PSD. The results suggest that the bilirubin level is an important biological marker for the risk of depression in patients with ischemic stroke.

Bilirubin, a powerful antioxidant, increases after hemorrhagic stroke.[30] Higher bilirubin level was associated with more severe stroke on univariate analysis, possibly reflecting the intensity of initial oxidative stress, even though the level of bilirubin may not be an independent prognostic factor.[15] Severe stroke is also a known risk factor for PSD.[31] Hence, stroke severity could be a partial explanation for the link between bilirubin level and PSD. In this study, however, the association between bilirubin level and PSD was independent of stroke severity. Accordingly, there should be other explanations for the association.

High levels of psychological stress have been reported in stroke survivors.[32] Acute psychological stress increases bilirubin metabolites in human urine,[33] the level of which is correlated with the level of perceived stress. Bilirubin metabolites can be useful markers of stress.[33] Although the level of bilirubin was not reported in the aforementioned study, it is possible that an acute stroke, which is a life-threatening condition, acts as a major psychological stressor, which in turn increases the level of bilirubin production. A higher level of bilirubin, like a higher level of cortisol,[34] may indicate a higher level of perceived stress, which increases the risk of depression.[35] Future studies should measure the level of perceived stress in acute stroke to test this hypothesis.

In this study, the volume and location of acute infarcts were not associated with PSD. Despite numerous attempts to establish a relationship between localization of stroke and PSD, this issue remains contentious,[36] possibly due to differences in study design, measurement of depression and neuroimaging methods.[37] Doubts have been raised about the classical single-stroke-based neuroanatomical model of PSD, suggesting that the vascular burden due to the chronic accumulation of small macro- and microvascular lesions may be the crucial determinant of PSD.[38]

The main limitation of this study was the fairly low recruitment rate and frequency of PSD, which resulted in a relatively small number of affected patients and, in turn, reduced the statistical power of the study. The low PSD frequency may be because subjects with a history of depression were excluded and the severity of stroke in the sample was mild. Another limitation was that PSD assessment was made only once, at the 3-month follow up, whereas the NIHSS was used only at the acute stage. Patients who had more severe stroke died before the 3-month follow up or became depressed later were not included. Those who were unable to give informed consent due to dementia or aphasia were also excluded. It should also be emphasized that the observed association should not be treated as a measure of depression to inform clinical practice. A further limitation was that MRI was not performed on the same day after the onset of the stroke. The literature suggests that serum bilirubin decreases during the acute phase of ischemic stroke,[16] which may have affected the present results. Finally, all subjects were of Chinese ethnicity. These selection biases limit the generalizability of the findings.


High bilirubin level is associated with the risk of PSD. Further investigations are needed to clarify the underlying pathophysiological link between bilirubin level and PSD.


There is no conflict of interest.