Abstract
Adjunctive treatment to improve outcome from bacterial meningitis has centered on dexamethasone. Among Vietnamese patients with bacterial meningitis, cerebrospinal fluid (CSF) opening pressure and CSF:plasma glucose ratios were significantly improved and levels of CSF cytokines interleukin (IL)–6, IL-8, and IL-10 and were all statistically significantly lower after treatment in patients who were randomized to dexamethasone, compared with levels in patients who received placebo.
The mortality rate associated with bacterial meningitis and the frequency of neurologic sequelae among those who survive the diseases remains high [1, 2]. Adjunctive treatment approaches to bacterial meningitis have focused on mitigating the inflammatory response in the subarachnoid space with agents such as corticosteroids. In animal models of bacterial meningitis, adjunctive dexamethasone with antimicrobial treatment ameliorated markers of subarachnoid space inflammation [3-5]. In randomized, controlled clinical trials, dexamethasone has been associated with improved cerebrospinal fluid (CSF) inflammatory parameters, but only in children with Haemophilus influenzae type b infection [6, 7]. In a second study involving children, dexamethasone was associated with improved CSF concentrations of tumor necrosis factor (TNF) α and platelet activating factor but was not associated with a reduction in mortality [8]. Finally, a third study failed to demonstrate any modulation of CSF inflammatory indices in children or adults with bacterial meningitis, despite an improvement in associated mortality among patients who received dexamethasone [9]. More-recent randomized controlled trials have provided conflicting results regarding the clinical value of dexamethasone as adjunctive treatment in bacterial meningitis for both adults and children [10-12] In this study, we examined biochemical and immunological markers of the subarachnoid space inflammatory response among adult Vietnamese patients with bacterial meningitis who were randomized to receive either dexamethasone or placebo.
Patients and methods
The study participants were from a randomized, double-blind, placebo-controlled trial of dexamethasone involving 435 patients >14 years of age who had suspected bacterial meningitis. The study was conducted from November 1996 through May 2005 at the Hospital for Tropical Diseases, Ho Chi Minh City, Vietnam [13]. A total of 217 patients were assigned to the dexamethasone group, and 218 were assigned to the placebo group. Lumbar puncture was performed and CSF opening pressure was measured at hospital admission and then repeated after initiation of treatment as the standard of care. Aliquots of CSF were sent for routine investigations, and when there was sufficient CSF collected, an aliquot was stored at −70°C for cytokine measurements. Bacterial meningitis was confirmed in 300 patients (69.0%) using conventional diagnostic approaches [13], and an additional 41 patients (11.8%) had a diagnosis that was confirmed by molecular methods (polymerase chain reaction [PCR]). Probable meningitis was diagnosed in 82 patients (19%), and an alternative diagnosis was made in 12 patients (3%).
A diagnosis of bacteriologically confirmed meningitis was made if bacteria were detected in CSF by (1) Gram or acridine orange stain, (2) latex agglutination test, (3) PCR, or (4) culture from CSF or blood samples. Details on the routine methods of pathogen detection have been described elsewhere [13]. Real-time PCR for detection of H. influenzae type B, Streptococcus pneumoniae, Neisseria meningitidis and Streptococcus suis was performed on stored samples as described elsewhere [14]. CSF cytokine levels were measured only in patients with confirmed bacterial meningitis for whom there were paired CSF samples obtained at the time of randomization (baseline) and again 1–4 days after initiation of treatment.
CSF concentrations of interleukin (IL)–6, IL-8, IL-10, IL-12p70, TNF-α, and IL-1β were determined using a multiplex bead array assay (CBA assay; BD Biosciences), and analysis was performed according to the manufacturer’s instructions. The limit of detection was 10 pg/mL for each cytokine. Analysis was performed by a technician blinded to the treatment assignment of each patient, and paired CSF samples were analyzed simultaneously.
Laboratory measurements, except for the CSF opening pressure and the ratio of CSF glucose to plasma glucose, were log-transformed before analysis. Several of the cytokine measurements were below the lower limit of detection, and therefore, methods for left-censored data were used [15]. We compared follow-up laboratory measurements between the 2 arms using the Wilcoxon rank-sum test for uncensored data or the Peto and Peto modification of the Gehan-Wilcoxon test for cytokines. We also performed an adjusted analysis that modeled the follow-up measurement as depending linearly on the baseline value of the respective laboratory measurement and the sampling day in addition to the study arm. In a sensitivity analysis, the analysis was adjusted by the addition of sex plus the variables most strongly associated with survival [13]: age, Glascow coma score (GCS), the presence of paresis, and meningitis caused by a bacterium other than Streptococcus suis. For the adjusted analyses, we used standard linear regression for uncensored data; for cytokines, we used maximum likelihood estimation accounting for left-censored data and assuming a normal distribution of the log-transformed measurements. No imputation of missing laboratory markers and no adjustment for multiplicity was performed, and all reported confidence intervals (CIs) are 95% CIs. Analyses were performed with the statistical software R, version 2.8.1 [16], and the contributed R package Nondetects and Data Analysis.
Results
From November 1996 through June 2005, 435 patients with suspected bacterial meningitis were randomly assigned to receive either a placebo or dexamethasone. Overall, 341 (78%) of 435 patients had confirmed bacterial meningitis. The number of patients with confirmed cases is higher than previously reported (300 patients) [13], because PCR identified an additional 41 confirmed patients. Paired CSF cytology or biochemistry data, collected at enrolment and 1–6 days later, was available from almost all of the confirmed patients (dexamethasone group, 154 [94%] of 164; placebo group, 164 [93%] of 177). In 146 (42%) of 341 patients, there was insufficient CSF to analyze cytokine concentrations, or the second CSF sample was collected >4 days after randomization and was therefore not analyzed. Thus, cytokine levels in paired CSF samples were measured for a total of 195 (58%) of 341 patients. Baseline characteristics of all patients with confirmed bacterial meningitis are shown in Table 1. The baseline features of patients for whom CSF cytokine levels were measured were similar to those for the overall group of patients with confirmed bacterial meningitis (Table 1).
Table 1. Baseline Characteristics and Outcome for Patients with Confirmed Bacterial Meningitis.
| Patients with confirmed bacterial meningitis | Patients with CSF cytokine measurements | |||
|---|---|---|---|---|
| Variable | Dexamethasone group (n = 164) | Placebo group (n = 177) | Dexamethasone group (n = 88) | Placebo group (n = 107) |
| Age, median years (IQR) | 43 (32—58) | 40 (28—51) | 45.5 (33.5—59) | 41 (29—52) |
| Male sex | 115 (70.1) | 136 (76.8) | 64 (72.7) | 81 (75.7) |
| Duration of illness, median day (IQR) | 3 (2–6) | 3 (2–5) | 3 (2–5) | 3 (2–5) |
| Findings at hospital admission | ||||
| Glascow coma score, median value (IQR) | 13 (9–15) | 13 (10–15) | 12 (9–15) | 13 (10–15) |
| Papilloedema | 8 (4.9) | 4 (2.3) | 1 (1.1) | 2 (1.9) |
| Cranial nerve palsy | 16 (9.8) | 19 (10.7) | 7 (8.0) | 9 (8.4) |
| Paresia | 29 (17.7) | 25 (14.1) | 13 (14.8) | 11 (10.3) |
| Positive test results | ||||
| CSF gram stain | 106 (64.6) | 116 (65.5) | 59 (67.0) | 71 (66.4) |
| CSF acridine orange stain stain | 11 (6.7) | 13 (7.3) | 4 (4.5) | 7 (6.5) |
| CSF culture | 108 (65.9) | 113 (63.8) | 59 (67.0) | 74 (69.2) |
| Blood culture | 65 (39.6) | 62 (35.0) | 36 (40.9) | 37 (34.6) |
| Latex antigen | 23 (14.0) | 27 (15.3) | 13 (14.8) | 15 (14.0) |
| PCR | 128 (78.0) | 134 (75.7) | 74 (84.1) | 81 (75.7) |
| Index of CSF inflammation, median value (IQR) | ||||
| White blood cell count, ×1000 cells/μL | 3785 (1055–8150) | 2808 (1125–7605) | 3290 (1155–8580) | 3285 (1510–8600) |
| Glucose level, mg/dL | 20 (10–36) | 23 (10–38) | 19 (10–39) | 20 (10–35) |
| Lactate level, mmol/L | 9.1 (6.1–14.7) | 10.1 (6.1–14.3) | 11.6 (7.6–16.7) | 11.0 (6.3–16.0) |
| Protein level, mg/L | 253 (153–420) | 244 (159–421) | 290 (142–434) | 249 (170–445) |
| Causative organism | ||||
| Streptococcus suis | 72 (43.9) | 65 (36.7) | 42 (47.7) | 41 (38.3) |
| Streptococcus pneumoniae | 36 (22.0) | 42 (23.7) | 22 (25.0) | 25 (23.4) |
| Neiseria meningitidis | 14 (8.5) | 14 (7.9) | 8 (9.1) | 7 (6.5) |
| Klebsiella pneumoniae | 7 (4.3) | 3 (1.7) | 2 (2.3) | 1 (0.9) |
| Escherichia.coli | 6 (3.7) | 3 (1.7) | 4 (4.6) | 1 (0.9) |
| Staphylococcus aureus | 3 (1.8) | 6 (3.4) | 1 (1.1) | 3 (2.8) |
| Hemophilus influenzea | 2 (1.2) | 6 (3.4) | 0 | 4 (3.7) |
| Death | 10 (6.1) | 22 (12.5) | 3 (3.4) | 9 (8.4) |
NOTE. Data are no. (%) of patients, unless otherwise indicated. CSF, cerebrospinal fluid; IQR, interquartile range; PCR, polymerase chain reaction.
The distribution of sampling times for the follow-up samples that were used for cytological or biochemical analysis was not statistically significantly different between the placebo and dexamethasone treatment arms (P = .61, by Fishers exact test). CSF glucose level (P < .001) and the ratio of CSF glucose to plasma glucose (P = .02) at follow-up were statistically significantly higher for patients who received dexamethasone than they were for patients who received placebo (Table 2), and the findings were consistent when the analysis was adjusted for additional baseline covariates that had been previously identified as being associated with outcome in this clinical trial (GCS, paresis, age, sex, and Streptococcus suis infection) (P < .001 for CSF glucose; P = .003 for ratio CSF glucose to plasma glucose. The CSF opening pressure was lower at follow-up among patients who received dexamethasone (P = .04) and was borderline significant after adjustment for additional covariates (P = .07). There were no statistically significant differences in follow-up measurements between the 2 arms with respect to CSF protein levels, CSF lactate levels, or CSF leukocyte count.
Table 2. Cerebrospinal Fluid (CSF) Opening Pressure, Leukocyte Count, and Biochemical Analysis Results for 341 Patients with Confirmed Bacterial Meningitis.
| Dexamethasone group (n = 164) | Placebo group (n = 177) | P for unadjusted comparison of follow-up values | Adjusted comparison of follow-up values | ||||
|---|---|---|---|---|---|---|---|
| CSF parameter | At baseline | At follow-upaa | At baseline | At follow-upb | Estimate of dexamethasone effect (95% CI) | P | |
| Opening pressure | |||||||
| No. of patients | 128 | 135 | 144 | 143 | |||
| Median cm (IQR) | 20 (14–34) | 13.5 (9–18) | 20 (15–28) | 14 (11–20) | .04 | −1.97 (−3.84 to −0.09) | .04 |
| Leukocyte count | |||||||
| No. of patients | 164 | 153 | 176 | 164 | |||
| Median leukocytes per mm3 (IQR) | 3785 (1065–8135) | 825 (260–2630) | 2808 (1138–7433) | 865 (400–1943) | .40 | 0.95 (0.70–1,28)b | .74 |
| Glucose level | |||||||
| No. of patients | 164 | 154 | 177 | 164 | |||
| Median mg/dL (IQR) | 20 (10–36) | 63 (45–80) | 23 (10–38) | 44 (27–55) | <.001 | 1.61 (1.41–1,84)b | <.001 |
| Ratio of CSF glucose to plasma glucose | |||||||
| No. of patients | 162 | 154 | 177 | 162 | |||
| Median % (IQR) | 15 (8–31) | 40 (32–48) | 17 (9–30) | 37 (24–49) | .02 | 4.36 (0.84–7.89) | .02 |
| Lactate level | |||||||
| No. of patients | 145 | 134 | 156 | 148 | |||
| Median mmol/L (IQR) | 11.60 (7.07–17.10) | 4.20 (3.33–6.28) | 10.80 (6.82–15.72) | 4.65 (3.23–6.13) | .73 | 0.99 (0.88–1.10)b | .79 |
| Protein level | |||||||
| No. of patients | 162 | 148 | 172 | 159 | |||
| Median mg/L (IQR) | 253 (155–419) | 108 (69–176) | 244 (159–421) | 108 (66–165) | .84 | 1.03 (0.88–1,20)b | .70 |
NOTE. CI, confidence interval; IQR, interquartile range.
The 318 follow-up samples were obtained on days 1–6 after randomization. Sampling day was day 1 for 19 (6%) of patients, day 2 for 218 (69%), day 3 for 69 (22%), day 4 for 5 (2%), day 5 for 3 (1%), and day 6 for 4 (1%).
Because data was log-transformed before analysis, this corresponds to an (antilog-transformed) multiplicative effect (eg, follow-up CSF glucose level for patients who received dexamethasone is estimated to be higher than that for patients who received placebo by a factor of 1.61).
Baseline and follow-up CSF cytokine concentrations are shown in Table 3. The distribution of sampling times for followup samples used in cytokine analysis was not statistically significantly different between the placebo and dexamethasone treatment arms (P = .21, by Fisher’s exact test). CSF concentrations of IL-6, IL-8, and IL-10 were all statistically significantly lower in follow-up samples from patients who received dexamethasone than they were in samples from patients who received placebo (Table 3). For all 3 cytokines, the results remained statistically significant (P < .01 for all) after adjustment for additional covariates (GCS, paresis, age, sex, and Streptococcus suis infection). CSF concentrations of IL-12p70, IL-1β, and TNF-α were either not measurable or were low at baseline. At follow-up, these cytokines were seldom detected in CSF, and there was no statistically significant difference in the distribution of follow-up measurements between the 2 groups or in the rate of patients with detectable values (Table 3).
Table 3. Cerebrospinal Fluid (CSF) Cytokine Concentrations at Baseline and Follow-Up for 195 Patients with Confirmed Bacterial Meningitis.
| Dexamethasone group (n = 88) | Placebo group (n = 107) | P for unadjusted comparison of follow-up values | Adjusted comparison of follow-up values | ||||
|---|---|---|---|---|---|---|---|
| CSF parameter | Baseline | Follow-upa | Baseline | Follow-upa | Estimate of dexamethasone effect (95% CI) | P | |
| IL-6 | |||||||
| No. (%) of patients with detectable values | 83 (94) | 83 (94) | 103 (96) | 105 (98) | |||
| Median log10 pg/mL (IQR) for patients with detectable values | 4.97 (4.38–5.37 | 3.23 (2.43–4.19) | 4.89 (4.5–5.5) | 3.65 (2.8–4.33) | .01 | −0.43 (−0.73 to −0.12) | .006 |
| IL-8 | |||||||
| No. (%) of patients with detectable values | 86 (98) | 86 (97) | 106 (99) | 106 (99) | |||
| Median log10 pg/mL (IQR) of patients with detectable values | 4.33 (3.81–4.68) | 3.24 (2.66–3.69) | 4.3 (3.82–4.68) | 3.45 (2.94–3.89) | .03 | −0.21 (−0.41 to −0.008) | .04 |
| IL-10 | |||||||
| No. (%) of patients with detectable values | 83 (94) | 43 (49) | 106 (99) | 74 (69) | |||
| Median log10 pg/mL (IQR) of patients with detectable values | 2.58 (2.06–3.09) | 1.57 (1.19–1.94) | 2.53 (2.04–3.06) | 1.52 (1.25–1.87) | .02 | −0.24 (−0.42 to −0.06) | .01 |
| IL-12 | |||||||
| No. (%) of patients with detectable values | 14 (16) | 5 (6) | 17 (16) | 5 (5) | |||
| Median log10 pg/mL (IQR) of patients with detectable values | 1.33 (1.18–1.64) | 1.27 (1.26–1.46) | 1.3 (1.22–1.69) | 1.20 (1.15–1.22) | .71b | 0.08 (–0.26 to 0.42) | .64 |
| IL-1β | |||||||
| No. (%) of patients with detectable values | 72 (82) | 38 (43) | 90 (84) | 51 (48) | |||
| Median log10 pg/mL (IQR) of patients with detectable values | 2.63 (2.13–3.3) | 1.77 (1.42–1.92) | 2.44 (2.17–3.24) | 1.87 (1.51–2.02) | .27b | −0.17 (−0.45 to 0.10) | .22 |
| TNF-α | |||||||
| No. (%) of patients with detectable values | 64 (72) | 12 (14) | 73 (68) | 11 (10) | |||
| Median log10 pg/mL (IQR) of patients with detectable values | 2.28 (1.49–3.16) | 1.23 (1.16–1.39) | 2.07 (1.57–3.35) | 1.31 (1.17–1.51) | .50b | 0.08 (−0.23 to 0.38) | .62 |
NOTE. IL, interleukin; TNF tumor necrosis factor.
Samples were obtained on days 1–4 after randomization. Sampling day was day 1 for 6 (3%) of the patients, day 2 for 144 (74%), day 3 for 44 (23%), and day 4 for 1 (1%).
Comparisons of the rates of detectable values by Fisher’s exact test were also nonsignificant. P values were .76 (IL-12), .57 (IL-1β), and .51 (TNF-α).
Discussion
A role for adjunctive dexamethasone therapy in bacterial meningitis has been considered for >30 years, but there is still no consensus on the value of this intervention [17, 18]. We recently demonstrated that dexamethasone was associated with a significant reduction in mortality among human immunodeficiency virus (HIV)–negative Vietnamese adult patients with confirmed bacterial meningitis [13]. The key finding from the present study was that dexamethasone was associated with a statistically significant (albeit small) reduction, in absolute terms, in CSF biochemical and immune markers of inflammation in Vietnamese adult patients with confirmed bacterial meningitis
Receipt of dexamethasone, compared with receipt of placebo, was associated with a statistically significantly greater improvement in the CSF glucose to plasma glucose ratio. This may reflect rapid antimicrobial-mediated killing of bacteria in the central nervous system and a dexamethasone-augmented reduction in inflammatory metabolic activity, but it could also reflect increased transport of glucose into the CSF. At the time of study enrolment, the concentration of IL-6, IL-8, and IL-10 in CSF samples from patients with bacterial meningitis was very high and decreased with therapy. CSF concentrations of TNF-α, IL-1β, and IL-12 were lower or were not measurable at presentation, and this may reflect a different quantitative balance of production, consumption, degradation, and binding to inhibitory receptors in the CSF for these molecules, compared with IL-6, IL-8, and IL-10. IL-6 and IL-8 are classical proinflammatory cytokines that are produced by diverse cell types, with IL-8 in a potent neutrophil chemoattractant and likely a mediator of the neutrophil infiltrate in the CSF of patients with bacterial meningitis [19].
The strength of the current study was the large number of patients observed and the breadth of pathogens studied and inflammatory parameters measured. A weakness of the study was the possible bias associated with not measuring cytokine responses in all patients who were randomized to receive placebo or dexamethasone. Nevertheless, the results provide a rational, although not conclusive, explanation of the clinical benefit provided by adjunctive dexamethasone in HIV-negative Vietnamese adolescent and adult patients with confirmed bacterial meningitis [13].
Acknowledgments
Financial support. The Wellcome Trust.
Footnotes
Potential conflicts of interest. All authors: no conflicts.
References
- 1.Durand ML, Calderwood SB, Weber DJ, et al. Acute bacterial meningitis in adults: a review of 493 episodes. N Engl J Med. 1993;328:21–8. doi: 10.1056/NEJM199301073280104. [DOI] [PubMed] [Google Scholar]
- 2.Bohr V, Hansen B, Jessen O, et al. Eight hundred and seventy-five cases of bacterial meningitis. Part I of a three-part series: clinical data, prognosis, and the role of specialised hospital departments. J Infect. 1983;7:21–30. doi: 10.1016/s0163-4453(83)90894-0. [DOI] [PubMed] [Google Scholar]
- 3.Mustafa MM, Ramilo O, Mertsola J, et al. Modulation of inflammation and cachectin activity in relation to treatment of experimental Hemophilus influenzae type b meningitis. J Infect Dis. 1989;160:818–25. doi: 10.1093/infdis/160.5.818. [DOI] [PubMed] [Google Scholar]
- 4.Syrogiannopoulos GA, Olsen KD, Reisch JS, McCracken GH., Jr. Dexamethasone in the treatment of experimental Haemophilus influenzae type b meningitis. J Infect Dis. 1987;155:213–9. doi: 10.1093/infdis/155.2.213. [DOI] [PubMed] [Google Scholar]
- 5.Tauber MG, Khayam-Bashi H, Sande MA. Effects of ampicillin and corticosteroids on brain water content, cerebrospinal fluid pressure, and cerebrospinal fluid lactate levels in experimental pneumococcal meningitis. J Infect Dis. 1985;151:528–34. doi: 10.1093/infdis/151.3.528. [DOI] [PubMed] [Google Scholar]
- 6.Lebel MH, Freij BJ, Syrogiannopoulos GA, et al. Dexamethasone therapy for bacterial meningitis: results of two double-blind, placebo-controlled trials. N Engl J Med. 1988;319:964–71. doi: 10.1056/NEJM198810133191502. [DOI] [PubMed] [Google Scholar]
- 7.Mustafa MM, Lebel MH, Ramilo O, et al. Correlation of interleukin-1 β and cachectin concentrations in cerebrospinal fluid and outcome from bacterial meningitis. J Pediatr. 1989;115:208–13. doi: 10.1016/s0022-3476(89)80067-8. [DOI] [PubMed] [Google Scholar]
- 8.Odio CM, Faingezicht I, Paris M, et al. The beneficial effects of early dexamethasone administration in infants and children with bacterial meningitis. N Engl J Med. 1991;324:1525–31. doi: 10.1056/NEJM199105303242201. [DOI] [PubMed] [Google Scholar]
- 9.Girgis NI, Farid Z, Mikhail IA, Farrag I, Sultan Y, Kilpatrick ME. Dexamethasone treatment for bacterial meningitis in children and adults. Pediatr Infect Dis J. 1989;8:848–51. doi: 10.1097/00006454-198912000-00004. [DOI] [PubMed] [Google Scholar]
- 10.Scarborough M, Gordon SB, Whitty CJ, et al. Corticosteroids for bacterial meningitis in adults in sub-Saharan Africa. N Engl J Med. 2007;357:2441–50. doi: 10.1056/NEJMoa065711. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Molyneux EM, Walsh AL, Forsyth H, et al. Dexamethasone treatment in childhood bacterial meningitis in Malawi: a randomised controlled trial. Lancet. 2002;360:211–8. doi: 10.1016/s0140-6736(02)09458-8. [DOI] [PubMed] [Google Scholar]
- 12.Peltola H, Roine I, Fernandez J, et al. Adjuvant glycerol and/or dexamethasone to improve the outcomes of childhood bacterial meningitis: a prospective, randomized, double-blind, placebo-controlled trial. Clin Infect Dis. 2007;45:1277–86. doi: 10.1086/522534. [DOI] [PubMed] [Google Scholar]
- 13.Nguyen TH, Tran TH, Thwaites G, et al. Dexamethasone in Vietnamese adolescents and adults with bacterial meningitis. N Engl J Med. 2007;357:2431–40. doi: 10.1056/NEJMoa070852. [DOI] [PubMed] [Google Scholar]
- 14.Mai NT, Nga TV, Linh le D, et al. Streptococcus suis meningitis in adults in Vietnam. Clin Infect Dis. 2008;46:659–67. doi: 10.1086/527385. [DOI] [PubMed] [Google Scholar]
- 15.Helsel DR. More than obvious: better methods for interpreting non-detect data. Environ Sci Technol. 2005;39:419A–23A. doi: 10.1021/es053368a. [DOI] [PubMed] [Google Scholar]
- 16.R Development Core Team: a language and environment for statistical computing. R Foundation for Statistical Computing; [Accessed 13 January 2009]. Available at: http://www.R-project.org. [Google Scholar]
- 17.Coyle PK. Glucocorticoids in central nervous system bacterial infection. Arch Neurol. 1999;56:796–801. doi: 10.1001/archneur.56.7.796. [DOI] [PubMed] [Google Scholar]
- 18.Greenwood BM. Corticosteroids for acute bacterial meningitis. N Engl J Med. 2007;357:2507–9. doi: 10.1056/NEJMe0707474. [DOI] [PubMed] [Google Scholar]
- 19.Kobayashi Y. Neutrophil infiltration and chemokines. Crit Rev Immunol. 2006;26:307–16. doi: 10.1615/critrevimmunol.v26.i4.20. [DOI] [PubMed] [Google Scholar]