European Journal of Neurology 2009, 16: 498–505 doi:10.1111/j.1468-1331.2008.02512.

A decrease of human leucocyte antigen-DR expression on

monocytes in peripheral blood predicts stroke-associated infection
in critically-ill patients with acute stroke
D. P. Zhanga,b, F. L. Yanc, H. Q. Xuc, Y. X. Zhuc, Y. Yind and H. Q. Lud
a
Department of Neurology, Zhong-da Hospital, Southeast University, Nanjing, China; bDepartment of Neurology, PeopleÕs Hospital of
Zhengzhou, Zhengzhou, China; cDepartment of Neurology, Zhong-da Hospital, Southeast University, Nanjing, China; and dDetermination
Center of Flow Cytometry, Southeast University; Nanjing, China

Keywords:
human leucocyte antigen-
DR, neurological intensive
care unit, predictive value,
stroke, stroke-associated
infection
Received 6 August 2008
Accepted 28 November 2008
Background and purpose: To investigate changes in human leucocyte antigen (HLA)-DR
expression on peripheral monocytes, determine the value of predicting the development of
stroke-associated infection (SAI), and determine the correlation with other conditions in crit-
ically-ill patients in the neurological intensive care unit (NICU) who suffered an acute stroke.
Methods: All patients were enrolled consecutively and admitted to NICU within 24 h
after the onset of symptoms. Patients were followed in order to identify whether
infection developed and determine survival status within 2 weeks after the stroke.
Patients were divided into stroke or control group by study design, infection or non-
infection group by whether or not they had an infection, survival or death group by
prognosis and cerebral infarction or cerebral haemorrhage group by stroke type. Pa-
tients in which acute stroke was excluded by head CT or MRI were admitted to general
ward and were used as a control group. Blood samples were collected serially on days 1,
2, 4, 6 and 14 after stroke, then monocyte human leucocyte antigen-DR (HLA-DR)
expression was determined by flow cytometry. The National Institute of Health Stroke
Scale (NIHSS), Acute Physiology and Chronic Health Evaluation II (APACHEII) and
Glasgow Coma Scale (GCS) scores were recorded over the course of observation.
Results: Fifty-three subjects and 39 controls were enrolled in the study. On days 1, 2, 4, 6
and 14, there was a significant difference in monocyte HLA-DR expression between stroke
group and control group (all P < 0.001), but no difference was found between ischaemic
stroke group and haemorrhagic stroke group (all P > 0.05). The infection group compared
with non-infection group did not exhibit a significant difference in HLA-DR expression on
days 1 and 2 (all P > 0.05), but significant differences emerged on days 4, 6 and 14 (all
P < 0.01). On days 1 and 2 the HLA-DR expression in the survival group compared with
death group, was not significantly different (all P > 0.05), but differences became signifi-
cant on days 4 and 6 (P < 0.01). On day 1, HLA-DR expression <62.80% had the
predictive value to SAI (sensitivity 83.3%, specificity 55.2%, AUC = 0.661, P = 0.031)
and on day 2, HLA-DR expression <57.83% had the predictive value to SAI (sensitivity
95.8%, specificity 79.3%, AUC = 0.907, P = 0.000) in acute stroke patients. A statisti-
cally significant inverse correlation was found between NIHSS and HLA-DR on days 2 and
4 during the observation period (all P < 0.01), but there was no statistically significant
negative correlation on days 1, 6 or 14 (all P > 0.05). HLA-DR expression did not correlate
with APACHEII (all P > 0.05) or GCS (all P > 0.05) during the measurement period.
Conclusions: Human leucocyte antigen-DR (HLA-DR) expression decreases and sustains
a dynamic change and it also relates to the severity of patientÕs condition in the critically-ill
patients with stroke. Progressively persistent low monocyte HLA-DR expression is asso-
ciated with a poor prognosis. The decline in HLA-DR expression contributes to infection in
critically-acute stroke patients. Monitoring of monocyte HLA-DR expression may be
useful for identifying patients suffering from acute stroke who are at high risk for infection.

Correspondence: Yan Fu-ling, Department of Neurology, Southeast University, Nanjing 210009, P.R. China (tel.: +86 25 83272357; fax: +86 25
83272011; e-mail: yanfuling218@163.com).

 2009 The Author(s)

498 Journal compilation  2009 EFNS
 A decrease of human leucocyte antigen-DR expression predicts stroke-associated infection 499

better understanding of the complex interaction be-

Introduction
Stroke-associated infection (SAI), predominantly chest
and urinary-tract infections occur in 23–65% of all
stroke patients within the first few days after the stroke
[1–5]. Although early mortality is due to direct com-
plications from large strokes, pneumonia is the leading
cause of death in the post-acute phase of stroke,
regardless of hospitalization [6–8]. The high incidence
of infections in these patients is probably the result of
an impaired immune function. Recently, stroke-induced
immunodeficiency in a mouse model of cerebral
ischaemia has been described, which results in sponta-
neous bacterial infections resulting from impaired
cell-mediated immune responses [9]. Infection after
experimental focal ischaemia may result from brain-
induced immunodepression, but it is undetermined
whether a similar syndrome occurs in human stroke
[10]. Recent studies have analysed immunological
changes caused by stroke in humans and identified a
persistent loss of CD4+ T-lymphocytes in patients
developing an infection, suggesting that defects of T
helper-cell functions contribute to immunosuppression
in stroke patients [11,12]. These results give additional
support to the notion that stroke induces immunosup-
pression in humans. However, the clinical significance
of monocytes has not been determined in stroke-in-
duced immunosuppression in humans [13,14]. Mea-
surement of human leucocyte antigen-DR (HLA-DR)
expression on peripheral monocytes is also a good
marker of global immune function since HLA-DR
expression reflects the net effect of multiple mediators
acting in concert [15]. The existence of a stroke-induced
immunodepression syndrome might be an adaptive
mechanism to brain ischaemia, although further re-
search is required to unravel the clinical consequences
of these immunological changes [16]. In clinical prac-
tice, the National Institute of Health Stroke Scale
(NIHSS), Acute Physiology and Chronic Health Eval-
uation II (APACHEII) and the Glasgow Coma Scale
(GCS) are normally applied to evaluate the patientÕs
condition, especially in patients with severe stroke. A
tween the central nervous system and the immune sys-
tem might lead to more effective stroke therapies.
Therefore, the present study analyses immunological
changes caused by stroke in humans and identifies the
relationship between HLA-DR expression on periph-
eral monocytes and NIHSS, APACHEII and GCS.
Materials and methods
Patients and control subjects
We prospectively screened patients admitted to the
Zhong-da Hospital [affiliated with Southeast University
neurological intensive care unit (NICU)] between
October 2006 and January 2007. The patients not only
met one of the criteria for NICU admission, including
haemodynamic instability, acute respiratory failure or
intubation for airway protection, unstable neurological
status and/or the GCS <10, but also had at least one
NIHSS ‡6 after stroke onset during the observation
period or stroke symptoms onset <24 h before
admission to the NICU. Patients with clinical signs of
infection or with compromised immune function upon
admission were not recruited.
Sixty-seven acute stroke patients were admitted to
the NICU from October 2006 to January 2007, 62 pa-
tients were recruited into the study and 53 patients were
included in the study. These patients were compared
with 39 age-matched non-stroke control subjects in the
general neurological ward (Tables 1 and 2). Stroke
neurologists, infectious disease specialists and research
nurses worked in close collaboration to warrant a
harmonized application of diagnostic and therapeutic
measures. Patients had a brain CT scan or brain MRI
before admission, as well as the diagnostic workup
aimed to identify the cause of ischaemic or haemor-
rhagic strokes. Blood was obtained from the patients
immediately at admission and between 5:30 AM and
7:00 AM on days 1, 2, 4, 6 and 14 thereafter. The blood
samples were sent to the determination centre for flow
cytometry analysis within 3 h. The NIHSS scores were

Table 1 Clinical data for stroke subjects

compared with controls Stroke Control
Variables (n = 53) (n = 39) t/x2 P-value

Age (years, mean ± SD) 67.8 ± 11.6 69.6 ± 8.8 1.295 0.399
Sex (male, n, %) 26 (49.1) 18 (46.2) 0.076 0.783
Hypertension (n, %) 34 (64.2) 24 (61.5) 0.066 0.798
Diabetes mellitus (n, %) 8 (15.1) 3 (17.9) 0.134 0.714
Heart disease (n, %) 19 (35.8) 16 (41.0) 0.146 0.613
COPD (n, %) 3 (5.7) 4 (10.3) 0.675 0.411
Smoking (n, %) 19 (35.8) 11 (28.2) 0.597 0.440
Drinking (n, %) 17 (32.1) 11 (28.2) 0.159 0.690

COPD, chronic obstructive pulmonary disease.

 2009 The Author(s)

Journal compilation  2009 EFNS European Journal of Neurology 16, 498–505
500 D. P. Zhang et al.

Table 2 Patient characteristics

Total no. Age (years) Male Ischaemic Haemorrhage Death (mean ± SD)
Non-stroke control subjects 39 69.6 ± 8.8 18 0 0 0
Total patients included 62 68.1 ± 7.9 30 34 28 8
Cohort included 53 67.8 ± 11.6 26 29 24 7
Infected cohort 24 72.4 ± 10.8 12 13 11 6
Non-infected cohort 29 65.4 ± 11.3 14 16 13 1

recorded on days 1, 2, 4, 6 and 14, whereas the The study protocol was approved by the Research
APACHEII and GCS scores were recorded on days 1, 2 Ethics Committee of Affiliated ZhongDa Hospital,
and 3 (Table 3). Southeast University. All patients gave fully informed
For the purpose of this study, we defined the fol- consent directly or through a surrogate when appropri-
lowing criteria of infection: (i) presence of clinical signs ate. Routine cerebral CT images were acquired on a
of infection. Pneumonia was described as pulmonary 64-row multislice CT scanner (Somatom 64; SIEMENS
infiltrates in a chest radiograph, fever, respiratory Medical Systems, Erlangen, Germany). Cerebral MR
symptoms (cough, dyspnea, or pleuritic pain) and blood images were acquired on Eclipse 1.5T MR scanner
leucocyte count >11 000 cell/ml or <4000 cell/ml. (SIEMENS Medical Systems). Cell types were defined by
Urinary-tract infection was defined as low urinary tract surface marker expression determined in a fluorescence-
symptoms with a positive urine culture for an activated cell sorter analysis performed on a FACScali-
uropathogen (>105 colony-forming units [CFU]/mm3) bur (Beckton Dickinson, Heidelberg, Germany) using
or fever with a positive urine culture for an uropatho- Tru-Count tubes, Simulset IMK Plus-reagents (Becton
gen in absence of other infectious source. Fever of Dickinson, Germany). Monocytes were identified by
unknown origin was described as none of the above direct immunofluorescence staining by a combination of
criteria identified or idiopathic fever; (ii) serum the fluorescein isothiocyanate (FITC)-labelled mono-
concentrations of C-reactive protein >50 mg/ml; and clonal anti-CD14 IgG1. The expression of HLA-DR was
(iii) procalcitonin serum concentrations >0.5 ng/ml. determined with use of phycoerythrin (PE)-labelled
To compare patients who developed infection after monoclonal anti-HLA-DR IgG1 (Becton Dickinson).
stroke with those who did not, two cohorts were FITC-labelled mouse IgG1, PE-labelled mouse IgG1
formed. In the infected cohort, all three criteria for (Becton Dickinson) were used as controls.
infection had to be fulfilled on days 2, 4, 6 or 14. In the In brief, 2-ml blood samples were collected in EDTA-
non-infected cohort, none of the criteria were attained coated tubes. The anticoagulated peripheral blood
throughout the study period. Patients who had some samples were processed using the whole lysis procedure
criteria but not all were not assigned to either cohort, according to the manufacturerÕs protocol. An appro-
involved in 9 patients. priate volume of whole blood was added to the mixture
of a selected panel of monoclonal antibodies. The
Table 3 Scores of infection compared with non-infection
samples were incubated for 30 min at 4C, washed in
Infected or non-infected phosphate-buffered saline and then lyzed with a com-
mercial lyzing reagent (Becton Dickinson). Thereafter,
Infected Non-infected
Scales (n = 24) (n = 29) t/u P-value the samples were analysed on a FACSCallibur flow
cytometer equipped with a 15-mW air-cooled 488-nm
NIHSS(mean ± SD)
argon–ion laser for the excitation of FITC and PE. The
day1 18.8 ± 6.0 7.6 ± 3.9 5.942 0.012
day2 20.0 ± 6.4 8.8 ± 4.2 7.032 0.000
fluorescence of FITC was measured by a 530/30-nm
day4 20.4 ± 7.4 8.5 ± 4.3 7.599 0.000 band pass filter. The orange emission from PE was
day6 19.5 ± 7.9 7.1 ± 4.1 6.254 0.000 measured by a 582/42-nm band pass filter. The pro-
day14 14.6 ± 5.4 5.7 ± 3.6 4.424 0.023 portion of HLA-DR positive cells in the monocyte
APACHEII (mean ± SD)
population was determined. Final analysis was per-
day1 16.6 ± 5.4 10.3 ± 3.9 4.752 0.001
day2 17.6 ± 7.1 10.2 ± 3.7 4.650 0.000
formed using Quanticalc software (all from Becton
day3 18.8 ± 7.4 10.1 ± 3.7 5.226 0.000 Dickinson) to obtain the molecules HLA-DR per cell
GCS (M, IQR) from the measured geometric means.
day1 9.5 (8–13) 15 (13.5–15) )4.271 0.000 Procalcitonin serum concentration was measured on
day2 10.0 (7.3–13.8) 15 (14–15) )4.171 0.000
a Liaison analyzer (DiaSorin, Dietzenbach, Germany;
day3 10.0 (8.3–13.8) 15 (14–15) )4.275 0.000
reagents: Brahms, Hennigsdorf, Germany); C-reactive
M, median; IQR, interquartile range. protein on a Dimension RxL (Dade Behring, Eschborn,

 2009 The Author(s)

Journal compilation  2009 EFNS European Journal of Neurology 16, 498–505
 A decrease of human leucocyte antigen-DR expression predicts stroke-associated infection
Germany). It was first determined whether the studied
variables had normality. The differences in proportions
were then compared using the Pearson x2 test and
normally distributed continuous variables were com-
pared using StudentÕs two-tailed t-test or one-way
ANOVA. Statistical analyses were performed using SPSS
13.0 for Windows (SPSS Inc, Chicago, IL, USA).
Correlations between two variances were applied to
GraphPad Prism 4.0 (GraphPad Software, San Diego,
CA, USA), Pearson correlation analysis was used for
normally distributed variables and Spearman non-
parametric correlation analysis was employed for non-
normally distributed variables. The Receiver Operating
Characteristic (ROC) curve was performed using MED-
CALC 9.2 (MedCalc Software, Broekstraat, Mariakerke,
Belgium) in order to evaluate the predictive value of
HLA-DR on stroke-associated infection. All the prob-
ability values were two-tailed and the level of statistical
significance was set at P < 0.05.
Results
Stroke induces rapid immunological changes in
peripheral blood
Stroke and control
Monocyte HLA-DR expression in individuals with
stroke was below the control individuals and further
decreased reaching its nadir on days 2 and 4, then
recovered gradually over the follow-up period. On days
1, 2, 4, 6 and 14, the monocyte HLA-DR expression in
the stroke group was 61.66 ± 16.81%, 54.89 ±
16.61%, 55.54 ± 15.39%, 62.21 ± 18.04% and
71.12 ± 16.67%, respectively. In the control group, the
values were 78.14 ± 7.87%, 79.92 ± 10.48%, 82.67 ±
8.69%, 80.09 ± 9.45% and 80.54 ± 8.23%, respec-
tively. There were statistically significant differences at
these time points (all P = 0.000; Fig. 1).
Control Stroke
100
90
80
70
HLA-DR (%)
60
50
40
30
20
10
0
Day 1 Day 2 Day 4 Day 6
Figure 1 Expression of HLA-DR in control and stroke. Control:
n = 39; Stroke: n (day1) = 53, n (day2) = 53, n (day4) = 52, n
(day6) = 49, n (day14) = 46. *P < 0.05; Data: mean ± SD.
 2009 The Author(s)
Journal compilation  2009 EFNS European Journal of Neurology 16, 498–505
501
Ischaemia and haemorrhage
Monocyte HLA-DR expression was slightly degraded
in cerebral haemorrhage patients compared with the
cerebral infarction patients. On days 1, 2, 4, 6 and 14,
monocyte HLA-DR expression in the ischaemic stroke
group was 64.73 ± 15.32%, 57.56 ± 17.84%,
57.43 ± 13.53%, 62.81 ± 15.75% and 69.89 ±
19.23%, respectively. In the haemorrhagic stroke
group, these values were 57.95 ± 18.08%, 51.66 ±
14.73%, 53.25 ± 17.40%, 61.48 ± 20.84% and
72.88 ± 12.45%, respectively. There were no signifi-
cant differences between groups (P = 0.146, 0.201,
0.329, 0.797 and 0.555, respectively; Fig. 2).
Infection and non-infection
Monocyte HLA-DR expression decreased more rapidly
and recovered more slowly in patients who were
assigned to the infected cohort in comparison with the
patients who remained uninfected. On days 1, 2, 4, 6
and 14, the monocyte HLA-DR expression in the
infection group was 59.74 ± 16.70%, 51.44 ± 16.34%,
48.86 ± 16.10%, 51.78 ± 21.15% and 60.26 ±
20.89%, respectively. In the non-infection group,
these values were 63.25 ± 17.03%, 57.74 ± 16.58%,
61.07 ± 12.54%, 70.12 ± 9.77% and 77.49 ± 9.14%,
respectively. In the infection group compared with non-
infection group, there were no significant differences in
HLA-DR expression on days 1 and 2 (P = 0.445 and
0.172), but significant differences emerged on days 4, 6
and 14 (P = 0.003, 0.001 and 0.004, respectively;
Fig. 3).
Survival and death
Monocyte HLA-DR expression decreased over time in
the death patients, whilst it gradually normalized over
the follow-up period in the survival patients. On days 1,
100 Ischaemic stroke Hemorrhagic stroke
80
HLA-DR (%)
60
40
20
0
Day 1 Day 2 Day 4 Day 6 Day 14
Time
Day 14 Figure 2 Expression of HLA-DR in ischaemic and haemorrhagic
stroke. Ischaemic stroke: n (day1) = 29, n (day2) = 29, n
(day4) = 28, n (day6) = 26, n (14) = 26. Haemorrhage stroke: n
(day1) = 24, n (day2) = 24, n (day3) = 24, n (day6) = 23, n
(day14) = 20. Data: mean ± SD.
502 D. P. Zhang et al.

100 Non-infection Infection Correlations between NIHSS, APACHEII, GCS and HLA-
90 DR
80
A statistically significant inverse correlation was found
70
HLA-DR (%)

60
between the NIHSS and HLA-DR expression on days 2
50 and 4 (day 2: r = )0.447, P = 0.001; day 4: r =
40 )0.626, P = 0.000), but there was no statistically sig-
30 nificant correlation on days 1, 6 or 14(day 1:
20 r = )0.105, P = 0.571; day 6: r = )0.293, P = 0.104;
10 day 14: r = )0.167, P = 0.268). HLA-DR expression
0 did not statistically correlate with the APACHEII score
Day 1 Day 2 Day 4 Day 6 Day 14
(day 1: r = )0.112, P = 0.424; day 2: r = )0.159,
Time
P = 0.255) or GCS (day 1: r = 0.135, P = 0.337; day
Figure 3 Expression of HLA-DR in infection and non-infection 2: r = 0.105, P = 0.454; Fig. 5).
groups. Non-infection: n (day1) = 29, n (day2) = 29, n
(day4) = 28, n (day6) = 28, n (day14) = 27; Infection: n
(day1) = 24, n (day2) = 24, n (day4) = 24, n (day6) = 21, Predictive value to SAI
n (day14) = 19. *P < 0.05; Data: mean ± SD.
Receiver operating characteristic curve indicated that a
HLA-DR expression <62.80% had predictive value for
SAI (sensitivity 83.3%, specificity 55.2%, AUC =
2, 4, 6 and 14, monocyte HLA-DR expression in the
0.661, P = 0.031, 95% CI 0.518–0.785, positive likeli-
survival group was 62.59 ± 15.98%, 55.43 ± 15.82%,
hood ratio 1.73, negative likelihood ratio 0.32, positive
57.95 ± 13.63%, 65.82 ± 14.41% and 71.12 ±
predictive value 60.6%, negative predictive value
16.67%, respectively, whereas on days 1, 2, 4 and 6 the
80.8%) on day 1. On day 2, HLA-DR expression
monocyte HLA-DR expression in the death group was
55.51 ± 21.96%, 51.32 ± 22.33%, 39.66 ± 17.91%
and 28.93 ± 14.06%, respectively (all patients died
within 8 days after the stroke). There were no signifi- Day 2
(a) 40
cant differences between groups on days 1 and 2
(P = 0.303 and 0.548), but differences became signifi-
cant on days 4 and 6 (P = 0.003 and 0.000; Fig. 4). In 30
logistic regression (admission NIHSS and age
NIHSS2

adjusted), there was a statistical significance in the 20

association between poor outcome and decreased
expression of HLA-DR at day 4 (OR 5.72, 95% CI
1.63–45.18, P = 0.020), and at day 6 (OR 7.44, 95% CI 10
2.85–84.71, P = 0.009).
0
0 10 20 30 40 50 60 70 80 90 100
Survival Death HLA-DR2
100
90
80 (b) 40 Day 4
70
HLA-DR (%)

60
50 30
40
NIHSS4

30
20 20
10
0
Day 1 Day 2 Day 4 Day 6 Day 14 10
Time

Figure 4 Expression of HLA-DR in survival and death groups. 0

Survival: n (day1) = 46, n (day2) = 46, n (day4) = 46, n 0 10 20 30 40 50 60 70 80 90 100
(day6) = 46, n (day14) = 46. Death: n (day1) = 7, n (day2) = 7, HLA-DR4
n (day4) = 5, n (day6) = 3, n (day14) = 0. *P < 0.05; Data:
mean ± SD. Figure 5 Correlation between NIHSS and HLA-DR (%).

 2009 The Author(s)

Journal compilation  2009 EFNS European Journal of Neurology 16, 498–505
 A decrease of human leucocyte antigen-DR expression predicts stroke-associated infection
100
80
60
Sensitivity
40 Day 1
Day 2
20
0 monocyte deactivation was more serious in the infec-
0 20 40 60 80 100
100-Specificity
Figure 6 Value of HLA-DR (%) predicted SAI.
<57.83% had predictive value for SAI (sensitivity
95.8%, specificity 79.3%, AUC = 0.907, P = 0.000,
95% CI 0.794–0.969, positive likelihood ratio 4.63,
negative likelihood ratio 0.053, positive predictive value
79.3%, negative predictive value 95.8%) in acute stroke
patients (Fig. 6).
Discussion
Observations in mouse models suggest that stroke-in-
duced immunodepression is a cause of infection fol-
lowing cerebral stroke [9,17]. However, attempts to
transfer these findings into clinical practice have yielded
contradictory results [18–20]. Most recently, there have
been two clinical studies supporting the notion that
stroke induces immunosuppression in humans. One
study [11] analysed lymphocyte subsets after stroke and
demonstrated that low CD4+ T cell counts as well as a
low percentage of lymphocytes on day 1 may pre-dis-
pose patients to infection after cerebral ischaemia. The
other study [12] also found cellular immunodepression
preceding infectious complications after acute ischae-
mic stroke in humans. A diminished monocytic HLA-
DR expression early after stroke onset has also been
described simply by the above mentioned two studies
[11,12]. However, these outcomes mainly observed the
lymphocyte subsets after stroke and monocytes proba-
bly played more important role in defending microor-
ganism which close linked to SAI. Measurement of
HLA-DR expression on peripheral monocytes also
seems to be a good marker of global immune function
since HLA-DR expression reflects the net effect of
 2009 The Author(s)
Journal compilation  2009 EFNS European Journal of Neurology 16, 498–505
503
multiple mediators acting in concert [15]. In the present
study, earlier differences of HLA-DR expression had
been detected, remarkably indicating the studied infec-
tions occurring in early time-points was focused on,
however, HLA-DR was not different in the infection
group in the first days, different only occurring by days
4, 6 or 14 in the infection group compared to non-
infection group.
Our results suggest that stroke induces long-lasting
dramatic depression of monocyte immunity function
(HLA-DR expression decrease) associated with SAI
and monocyte immunity function showed a remarkably
dynamic change. The expression of HLA-DR in the
infection group was lower than in the non-infection
group and the recovery in infection group was slower
than in the non-infection group, suggesting that
tion group. Stroke type has no influence on the
expression of HLA-DR. ROC indicated that on day 2,
HLA-DR expression <57.83% had predictive value for
SAI (sensitivity 95.8%, specificity 79.3%, AUC =
0.907, P = 0.000, 95% CI 0.794–0.969, positive likeli-
hood ratio 4.63, negative likelihood ratio 0.053, positive
predictive value 79.3%, negative predictive value
95.8%) in acute stroke patients. A reduced HLA-DR
expression predicted subsequent stroke associated
infections was also observed recently by Harms et al.
[21]. However, our study included different patients
(ischaemic and haemorrhagic stroke) and used different
analysis methods (ROC) confirmed the predictive value
of HLA-DR for post-stroke infections. These facts
further reveal that decreased HLA-DR expression was
close linked to the SAI.
Expression of HLA-DR <30% predicted infection in
trauma patients [22], which implies that stroke patients
are more liable to suffer from infection due to immun-
odepression because stroke patients have more risk
factors for infection than the trauma patients. For
example, stroke patients are more prone to incur
impaired consciousness and aspiration for various rea-
sons. These results support to the notion that stroke
induces monocyte immunosuppression in humans,
which contributes to developing an infection after
stroke.
On days 2 and 4, HLA-DR expression correlated
inversely with NIHSS, but this effect was not seen on
days 1, 6 and 14. The fact that HLA-DR expression is
inversely correlated with NIHSS suggests that the brain
lesion promotes the down-regulation of HLA-DR.
These findings also indicate the recovery of HLA-DR
and NIHSS are partly asynchronous; the former
recovered more rapidly than the latter. To observe
stroke-induced immunodepression, the level of HLA-
DR after stroke should be determined. HLA-DR did
504 D. P. Zhang et al.
not correlate with APACHEII or GCS at the observa-
tional points in the present study. This may be because
APACHEII is mainly associated with peripheral
inflammatory conditions (i.e., a chronic health state).
Stroke-induced immunodepression results from a local
increase in proinflammatory cytokines in damaged
brain tissue, which triggers a systemic immunoinhibi-
tory response via sympathoadrenal activation [9].
Whether or not this is involved in the peripheral
inflammatory reaction is debated based on differing
results in humans and laboratory animals [12,23]. The
GCS normally evaluates the severity of brain insult, but
stroke location, lateralization and mental status
together affected stroke-induced immunodepression
[16]. The relationship between stroke scales and stroke-
induced immunodepression is a very interesting phe-
nomenon and future studies should investigate their
relationship.
In the present study, we demonstrated that the
expression of HLA-DR progressively decreased in
deceased stroke patients. Logistic regression (admission
NIHSS and age adjusted) indicated that decreased
expression of HLA-DR (at day 4 and 6) was indepen-
dently associated with the subsequent death patients. In
contrast, the level of HLA-DR temporarily decreased
and then progressively recovered in the survival
patients. These results suggest that only ÔsustainedÕ or
Ôlong-termÕ immunodepression as a result of stroke,
results in a poor prognosis.
There are some limitations to our study. First, for the
sake of saving time and study funding, there were a
small number of patients included in this study. We
plan to enlarge our sample size in order to further study
this interesting problem. Secondly, we chose a control
group that was somewhat similar to the stroke patients
based on the fact that they had neurological abnor-
malities, however, the control group did not exhibit
significant differences compared with the stroke group
on many of the measures we tested. Despite these lim-
itations, we have characterized stroke-induced mono-
cyte deactivation, analysed the expression of HLA-DR
predicting SAI after acute stroke, investigated the
relationship between stroke scales and stroke-induced
immunodepression and observed the prognosis of
stroke-induced immunodepressed patients. If the pre-
dictive value of the level HLA-DR could be validated in
a larger prospective clinical trial, this would provide an
early and easily obtainable parameter to identify pa-
tients at high risk for subsequent infection. In addition
to antimicrobial approaches, immunotherapeutic
approaches may help to prevent or even reverse
inappropriate immunodepression and decrease the risk
of infectious complications after a stroke. Further
research is needed to understand the signals and
Journal compilation  2009 EFNS European Journal of Neurology 16, 498–505
mechanisms by which the adaptive and innate immune
systems react to brain tissue damage and to unravel the
consequences that this has for the patient.
Acknowledgements
We wish to express our gratitude to the healthcare
workers at the Neurological Intensive Care Unit at
Zhong-da Hospital affiliated with Southeast University
for their helpful cooperation. This study was supported
by the Natural Science Foundation of Jiang-su Prov-
ince (BK2008299).
References
1. Davenport RJ, Dennis MS, Wellwood I, Warlow CP.
Complications after acute stroke. Stroke 1996; 27: 415–
420.
2. Johnston KC, Li JY, Lyden PD, et al. Medical and neu-
rological complications of ischemic stroke: experience
from the RANTTAS trial. RANTTAS investigators.
Stroke 1998; 29: 447–453.
3. Georgilis K, Plomaritoglou A, Dafni U, Bassiakos Y,
Vemmos K. Aetiology of fever in patients with acute
stroke. Journal of Internal Medicine 1999; 246: 203–209.
4. Grau AJ, Buggle F, Schnitzler P, Spiel M, Lichy C, Hacke
W. Fever and infection early after ischemic stroke. Journal
of the Neurological Sciences 1999; 171: 115–120.
5. Langhorne P, Stott DJ, Robertson L, et al. Medical
complications after stroke: a multicenter study. Stroke
2000; 31: 1223–1229.
6. Kalra L, Evans A, Perez I, Knapp M, Donaldson N, Swift
CG. Alternative strategies for stroke care: a prospective
randomised controlled study of stroke unit, stroke team,
and domiciliary management of stroke. Lancet 2000; 356:
894–899.
7. Henon H, Godefroy O, Leys D, et al. Early predictors of
death and disability after acute cerebral ischemic event.
Stroke 1995; 26: 392–398.
8. Kalra L, Yu G, Wilson K, Roots P. Medical complica-
tions during stroke rehabilitation. Stroke 1995; 26: 990–
994.
9. Prass K, Meisel C, Höflich C, et al. Stroke-induced
immunodeficiency promotes spontaneous bacterial infec-
tions and is mediated by sympathetic activation reversal
by poststroke T helper cell type 1-like immunostimula-
tion. Journal of Experimental Medicine 2003; 198: 725–
736.
10. Chamorro A, Urra X, Planas AM. Infection after acute
ischemic stroke: a manifestation of brain-induced
immunodepression. Stroke 2007; 38: 1097–1103.
11. Vogelgesang A, Grunwald U, Langner S, et al. Analysis
of lymphocyte subsets in patients with stroke and their
influence on infection after stroke. Stroke 2008; 39: 237–
241.
12. Haeusler KG, Schmidt WU, Föhring F, et al. Cellular
immunodepression preceding infectious complications
after acute ischemic stroke in humans. Cerebrovascular
Diseases 2008; 25: 50–58.
13. Chamorro A, Amaro S, Vargas M, et al. Interleukin-10,
monocytes, and increased risk of early infection in
 2009 The Author(s)
 A decrease of human leucocyte antigen-DR expression predicts stroke-associated infection
ischemic stroke. Journal of Neurology, Neurosurgery, and
Psychiatry 2006; 77: 1279–1281.
14. Chamorro A, Amaro S, Vargas M, et al. Catecholamines,
infection, and death in acute ischemic stroke. Journal of
the Neurological Sciences 2007; 252: 29–35.
15. Wunderink RG. Nosocomial pneumonia, including ven-
tilator-associated pneumonia[J]. Proceedings of the
American Thoracic Society 2005; 2: 440–444.
16. Meisel C, Schwab JM, Prass K, Meisel A, Dirnagl U.
Central nervous system injury-induced immune deficiency
syndrome. Nature Reviews. Neuroscience 2005; 6: 775–786.
17. Prass K, Braun JS, Dirnagl U, Meisel C, Meisel A. Stroke
propagates bacterial aspiration to pneumonia in a model
of cerebral ischemia. Stroke 2006; 37: 2607–2612.
18. Chamorro A, Horcajada JP, Obach V, et al. The early
systemic prophylaxis of infection after stroke study: a
randomized clinical trial. Stroke 2005; 36: 1495–1500.
 2009 The Author(s)
Journal compilation  2009 EFNS European Journal of Neurology 16, 498–505
505
19. Vargas M, Horcajada JP, Obach V, et al. Clinical conse-
quences of infection in patients with acute stroke: is it
prime time for further antibiotic trials? Stroke 2006; 37:
461–465.
20. Dirnagl U, Klehmet J, Braun JS, et al. Stroke-induced
immunodepression: experimental evidence and clinical
relevance. Stroke 2007; 38: 770–773.
21. Harms H, Prass K, Meisel C, et al. Preventive antibacte-
rial therapy in acute ischemic stroke: a randomized con-
trolled trial. PLoS ONE 2008; 3: e2158.
22. Volk HD, Reinke P, Docke WD. Clinical aspects: from
systemic inflammation to ÔimmunoparalysisÕ. Chemical
Immunology 2000; 74: 162–177.
23. Woiciechowsky C, Schöning B, Lanksch WR, Volk HD,
Döcke WD. Mechanisms of brain-mediated systemic anti-
inflammatory syndrome causing immunodepression.
Journal of Molecular Medicine 1999; 77: 769–780.