Glycerol vs. Dexamethasone in Meningitis • CID 2007:45 (15 November) • 1277

M A J O R A R T I C L E

Adjuvant Glycerol and/or Dexamethasone toImprove the Outcomes of Childhood BacterialMeningitis: A Prospective, Randomized, Double-Blind, Placebo-Controlled Trial

Heikki Peltola,1 Irmeli Roine,3 Josefina Fernandez,4 Ines Zavala,5 Silvia Gonzalez Ayala,7 Antonio Gonzalez Mata,10

Antonio Arbo,12 Rosa Bologna,8 Greta Mino,6 Jose Goyo,11 Eduardo Lopez,9 Solange Dourado de Andrade,13

and Seppo Sarna2

1Helsinki University Central Hospital, Hospital for Children and Adolescents, and 2University of Helsinki, Department of Public Health, Helsinki,Finland; 3University Diego Portales, Faculty of Health Sciences, Santiago, Chile; 4Clinica Infantil Dr. Robert Reid Cabral, Santo Domingo,Dominican Republic; 5Hospital de Ninos Dr. Roberto Gilbert, and 6Hospital del Nino Dr. Francisco de Icaza Bustamante, Guayaquil, Ecuador;7Hospital de Ninos Sor Marıa Ludovica, La Plata, and 8Hospital de Pediatrıa Dr. Juan P. Garrahan and 9Hospital de Ninos Dr. Ricardo Gutierrez,Buenos Aires, Argentina; 10Hospital Pediatrico Dr. Agustin Zubillaga, Barquisimeto, and 11Hospital Universitario de los Andes, Merida, Venezuela;12Instituto de Medicina Tropical, Universidad Nacional de Asuncion, Asuncion, Paraguay; and 13Fundacao de Medicina Tropical do Amazonas,Institute for Tropical Diseases, Manaus, Brazil

(See the editorial commentary by Saez-Llorens and McCracken on pages 1287–9)

Background. Despite favorable meta-analyses, no study involving third-generation cephalosporins for the treat-ment of childhood bacterial meningitis has documented a benefit of adjuvant dexamethasone therapy if theoutcomes are examined individually.

Methods. We conducted a prospective, randomized, double-blind trial comparing adjuvant dexamethasone orglycerol with placebo in children aged from 2 months through 16 years in Latin America. Ceftriaxone wasadministered to all children; children were randomized to also receive dexamethasone intravenously, glycerol orally,both agents, or neither agent. Primary end points were death, severe neurological sequelae, or deafness, with thefirst 2 end points forming a composite end point. A subgroup analysis for Haemophilus influenzae type b meningitiswas undertaken. Intention-to-treat analysis was performed using binary logistic regression models.

Results. H. influenzae type b, pneumococci, and meningococci were the main agents found among 654 patients;dexamethasone was given to 166, dexamethasone and glycerol were given to 159, glycerol was given to 166, andplacebo was given to 163. No adjuvant therapy significantly affected death or deafness. In contrast, glycerol anddexamethasone plus glycerol reduced severe neurological sequelae, compared with placebo; the odds ratios were0.31 (95% confidence interval [95% CI], 0.13–0.76; ) and 0.39 (95% CI, 0.17–0.93; ), respectively.P p .010 P p .033For neurological sequelae and death, the odds ratios were 0.44 (95% CI, 0.25–0.76; ) and 0.55 (95% CI,P p .0030.32–0.93; ), respectively. Dexamethasone therapy prevented deafness in patients with H. influenzae typeP p .027b meningitis only if patients were divided grossly into dexamethasone recipients and nonrecipients and if timingbetween dexamethasone and ceftriaxone administration was not taken into account (odds ratio, 0.27; 95% CI,0.09–0.77; ).P p .014

Conclusion. Oral glycerol therapy prevents severe neurological sequelae in patients with childhood meningitis.Safety, availability, low cost, and oral administration also add to its usefulness, especially in resource-limited settings.

Despite successful Haemophilus influenzae type b (Hib)

and Streptococcus pneumoniae vaccinations in many

Received 2 March 2007; accepted 24 July 2007; electronically published 15October 2007.

Reprints or correspondence: Prof. H. Peltola, HUCH, Hospital for Children andAdolescents, P.O. Box 281 (11 Stenback St.), 00029 HUS Helsinki, Finland(heikki.peltola@hus.fi).

Clinical Infectious Diseases 2007; 45:1277–86� 2007 by the Infectious Diseases Society of America. All rights reserved.1058-4838/2007/4510-0004$15.00DOI: 10.1086/522534

parts of the world [1], childhood bacterial meningitis

remains a challenge [2–4]. Even when seasonal menin-

gococcal epidemics in Africa are excluded, 11 million

people are affected annually; ∼350,000 die, and at least

30% of survivors experience sequelae. Death often fol-

lows neurological damage, especially in regions where

rehabilitation facilities are virtually nonexistent [4, 5].

Globally, Hib, S. pneumoniae, and Neisseria meningitidis

cause ∼90% of nonneonatal, nontuberculous cases of

bacterial meningitis.

by guest on May 14, 2011

cid.oxfordjournals.orgD

ownloaded from

1278 • CID 2007:45 (15 November) • Peltola et al.

Except in cases of drug-resistant infection, the use of anti-

microbials that are newer than third-generation cephalosporins

has not improved outcomes [6]. As has been documented by

biochemical parameters, dexamethasone dampens the inflam-

matory response [7–10], but no study of optimal antimicrobial

therapy for childhood meningitis has shown a significant re-

duction in deafness, neurological sequelae, or mortality when

these outcomes were examined separately. In Malawi, the first

trial that was large enough to allow an examination of the

individual outcomes failed to find any benefit [11]. Unfortu-

nately, cephalosporins could not be used routinely in that piv-

otal study. Revised Cochrane analysis [12] supports the use of

corticosteroids in high-income countries; however, as in meta-

analysis in general, very dissimilar populations were directly

compared, and the presenting status of the children was not

taken into account.

Glycerol (glycerine, 1, 2, 3-propanetriol), which is a naturally

occurring trivalent alcohol, an essential compound of the hu-

man cell membrane, a hyperosmolar agent, and an osmotic

diuretic, was long used in neurosurgery, neurology, and oph-

thalmology to reduce raised tissue pressure [13–17]. Glycerol

was given experimentally to treat bacterial meningitis in a few

children in the United States in the 1970s [18], but the first

systematic trial was performed in Finland during the period

1987–1991 [19]. Glycerol appeared to reduce profound hearing

loss and persistent neurological abnormalities as efficaciously

as dexamethasone, but the series was too small for definitive

conclusions. To validate that finding, we launched a much larger

study in Latin America, in which the potentials of glycerol and

dexamethasone were examined in terms of different outcomes,

and the results were compared with those of a placebo group.

PATIENTS AND METHODS

Setting and patients. This prospective multicenter, random-

ized, double-blind clinical trial examined the potential of in-

travenous dexamethasone, oral glycerol, or their combination

as adjuvant medications to improve different outcomes of

childhood bacterial meningitis. The series comprised children

with bacterial meningitis who were aged 2 months through 16

years at 10 institutions in Argentina, Brazil, Dominican Re-

public, Ecuador, Paraguay, and Venezuela, during the period

1996–2003 (Santo Domingo and Manaus joined the trial in

2001).

Meningitis was defined by (1) CSF culture positive for a

bacterial agent known to cause meningitis, (2) characteristic

CSF findings and positive blood culture results, (3) character-

istic CSF findings and a CSF sample with a positive latex ag-

glutination test result, and (4) symptoms and signs that were

compatible with bacterial meningitis and at least 3 of the fol-

lowing criteria: CSF pleocytosis (WBC count, �1000 cells/

mm3), decreased CSF glucose level (!40 mg/dL), increased CSF

protein concentration (�40 mg/dL), increased serum C-reac-

tive protein level (�40 mg/L) [20, 21], or occasionally, when

data regarding C-reactive protein level was not available, blood

leukocyte count 115,000 cells/mm3. Bacteriological analysis was

performed at each institute’s laboratory and was standardized

by training where necessary.

The exclusion criteria were a history of recent head injury,

previous neurosurgical precedure (e.g., intracranial shunt place-

ment), previous neurological disease (e.g., cerebral palsy and

Down syndrome), immunosuppression, and known hearing

impairment. Pretreatment antimicrobial therapy was registered

in detail but did not prevent study enrollment if oral therapy

or �1 parenteral dose had been administered.

Study drugs. All children received intravenous ceftriaxone

at a dosage of 80–100 mg per kg of body weight once daily for

7–10 days. Ceftriaxone was bought locally, although for the

centers in Guayaquil, Ecuador, and Asuncion, Paraguay, cef-

triaxone was purchased from Chile. Antipyretics were given as

needed, and convulsions were treated according to local prac-

tice. No fluid restriction was used [22, 23], but in patients with

hypovolemia, deficits were restored before changing to main-

tenance fluids with isotonic crystalloids.

The patients were randomized to 1 of the following adjuvant

medication groups: intravenous dexamethasone and oral pla-

cebo, intravenous dexamethasone and oral glycerol, oral glyc-

erol and intravenous placebo, or intravenous placebo and oral

placebo. The dosing of dexamethasone was 0.15 mg/kg ad-

ministered every 6 h for 48 h [24], the first dose being ad-

ministered 15 min prior to administration of ceftriaxone (if

possible).

Oral 85% glycerol (1 mL of which contained 1 g of glycerol)

was given for 48 h at a dosage of 1.5 g (1.5 mL) per kg every

6 h; the maximum volume was 25 mL per dose. The first dose

was given 15 min prior to ceftriaxone administration. At most

study centers, a nasogastric tube was inserted routinely. If the

child vomited within 30 min, the dose was repeated

immediately.

Randomization and blinding. Stratified block randomi-

zation took place in blocks of 20, except at 2 hospitals in Buenos

Aires, Argentina, in which the placebo-placebo group was not

allowed; at these hospitals, the block size was 24. All treatment

kits were packaged according to the randomization lists in San-

tiago, Chile. Saline and carboxymethylcellulose were the pla-

cebo preparations for dexamethasone and glycerol, respectively.

The agents were provided in identical ampoules or bottles and

were labeled only with a study code. Because all patients had

an intravenous line and received a test agent orally, the blinding

was complete. Each treatment kit, marked only with the study

number, contained the medication or placebo and a sealed

by guest on May 14, 2011

cid.oxfordjournals.orgD

ownloaded from

Glycerol vs. Dexamethasone in Meningitis • CID 2007:45 (15 November) • 1279

Table 1. Patient characteristics at presentation to the hospital.

Variables

DXM andplacebo group

(n p 166)

DXM andglycerol group

(n p 159)

Glycerol andplacebo group

(n p 166)

Placebo andplacebo group

(n p 163)

Age, median months (range) 13.0 (2–178) 12.0 (2–184) 10.0 (2–152) 10.0 (2–168)Male sex 93 (56) 94 (59) 92 (55) 98 (60)First symptoms 24–47 h before presentationa 42/146 (29) 34/145 (23) 41/144 (28) 47/144 (33)First symptoms �48 h before presentation 19/146 (13) 20/145 (14) 18/144 (13) 19/144 (13)Convulsions prior to or at hospital admission 55/154 (36) 52/150 (35) 50/153 (33) 54/153 (36)Prior use of antimicrobialsb 49/154 (32) 52/144 (36) 52/147 (35) 61/144 (42)Causative agent (no. of patients)

Haemophilus influenzae type b (221) 54 54 53 60Streptococcus pneumoniae (132) 35 31 30 36Meningococcus (110) 26 25 33 26Other (21) 9 2 6 4Unknown (170) 42 47 44 37

Glasgow Coma ScaleMean score � SD 12.2 � 2.8 12.4 � 2.8 12.5 � 2.5 12.1 � 3.0Score !13 66/155 (43) 60/149 (40) 61/155 (39) 65/153 (42)

CSF test resultsLeucocyte count, median cells/mm3 (IQR value) 2000 (6040) 1800 (7506) 1890 (3892) 1822 (6220)Glucose level, median mg/dL (IQR value) 19 (33) 16 (43) 16 (32) 13 (28)Protein level, median g/dL (IQR value) 151 (140) 144 (168) 155 (162) 171 (182)

Serological test resultsLeucocyte count, median 103 cells/mm3 (IQR value) 15.0 (10.7) 15.6 (11.7) 14.9 (10.4) 15.1 (12.5)Glucose level, median g/dL (IQR value) 106 (58) 105 (48) 100 (49) 106 (52)Hemoglobin level, median g/dL (IQR value) 9.4 (3.2) 9.3 (3.0) 9.1 (3.0) 9.0 (2.6)Sodium level, median mmoL/L (IQR value) 137 (7) 137 (6) 137 (7) 137 (8)

NOTE. Data are no. (%) of patients, unless otherwise indicated. Data is given for the intention-to-treat population. DXM, dexamethasone; IQR, interquartilerange.

a Irritability, vomiting, absent look, neck rigidity, or convulsions observed by mother.b During present illness, before bacterial meningitis was diagnosed.

envelope. The envelope disclosed the medication and was to

be opened in an emergency (no envelopes were opened during

the course of the study). Persons treating the patients, the study

monitor (I.R.), and the scientific advisor (H.P.) were not aware

of the specific treatments until the code was broken. This was

done after the study was completed.

Sample size. The sample size was calculated assuming that

a given adjuvant medication would decrease the rate of sequelae

from 20% to 5%. Accepting a 5% error in a 2-tailed test and

a power of 80%, at least 88 patients in each arm were required.

To be able to adjust for possible confounding factors, the sample

size was doubled. However, enrollment was to be stopped on

31 December 2003, whatever the number of patients.

Follow-up. All findings at presentation were recorded on

specially designed forms by the physician in charge. Records

included exact information on the nature and route of all an-

timicrobials and of the test agents, the age-adjusted Glasgow

Coma Scale [25], and the Denver Developmental Score [26].

A child was assessed for neurological, developmental, and hear-

ing sequelae on hospital discharge. If any deficits were found,

he or she was scheduled for a follow-up visit 1–2 months later.

After 200 patients had been enrolled and again after 400

patients had been enrolled, an ethicist and a statistician who

were not involved in the study reviewed the data to ensure that

there was no statistical difference between groups with respect

to case-fatality rate. The study protocol was approved by all

local ethical committees. Because not all mothers were literate,

an oral consent was accepted, after full information had been

given. The study was designed, conducted, and analyzed in-

dependently of any funding source.

End points and statistical analysis. The 3 primary end

points were death, severe neurological sequelae, and profound

hearing loss (inability to detect sounds with the better ear at

80 dB, determined with brain stem evoked response audiometry

or traditional audiometry). Severe neurological sequelae were

defined as blindness, quadriparesis or quadriplegia, hydro-

cephalus requiring a shunt, or severe psychomotor retardation

(in which the patient does not sit or walk, does not speak or

by guest on May 14, 2011

cid.oxfordjournals.orgD

ownloaded from

1280 • CID 2007:45 (15 November) • Peltola et al.

Figure 1. Study profile. DXM, dexamethasone; GLY, glycerol.

establish contact, or requires institutionalization). Because se-

vere neurological sequelae and death may form a continuum

[4, 5], these 2 outcomes formed a composite end point. Because

it was unlikely that all patients would report for follow-up,

assessments were performed primarily at hospital discharge.

Dexamethasone has been reported to be especially beneficial

in treating Hib meningitis [8–10]. Therefore, we planned a

subgroup analysis for Hib versus non-Hib meningitis, taking

into account the receipt or nonreceipt of pretreatment anti-

microbial drugs. The 2 centers in Buenos Aires that did not

include the placebo-only arm were included in the analysis,

provided that their results did not change the results of the 4-

arm study.

The x2 test was used to test the heterogeneity of proportions

between groups. To compare the main outcome measures, a

multivariable binary logistic model with and without covariates

was used. The treatment effects were taken into account by

applying a reference coding system using the (0.1)-indicator

variables, with the placebo recipients serving as the reference

group. All analyses were performed on an intention-to-treat

basis and were checked by per-protocol analysis.

Receipt of potential pretreatment antimicrobial drugs and

the timing of their administration with respect to the initiation

of adjuvant therapy were included as covariates in the post hoc

analysis for Hib meningitis versus other meningitides. We also

checked whether the findings remained the same when the

etiologically unconfirmed cases were excluded from analysis.

Finally, the entire series was divided into dexamethasone or

glycerol recipients and nonrecipients; this rough division was

intended to identify differences that were so small that they

would remain undetected in direct comparisons with the pla-

cebo-placebo group.

The results are expressed as ORs, 95% CIs, and P values. An

OR !1.0 indicated a beneficial effect, and it was also considered

to be statistically significant if the upper value of the 95% CI

was !1.0. P values of !.05 were considered to be statistically

significant.

RESULTS

General. As shown by the patient characteristics at hospital

admission (table 1), there was no significant difference between

the 4 groups, and there was no major differences in enrollment

of patients or outcomes during the study period. Of the 763

patients who were assessed for eligibility, 109 did not fulfill the

criteria of bacterial meningitis. Thus, 654 children (figure 1)

had data analyzed; of these, 166 received dexamethasone and

placebo, 159 received dexamethasone and glycerol, 166 received

glycerol and placebo, and 163 received placebo-placebo adju-

vant treatment. Per-protocol analysis, which did not change

the results, used data from 640 children; the reasons for ex-

cluding data from 14 children from this analysis are given in

figure 1. Eighty-six patients (13%) died. Of the remaining 568

children, 556 (98%) underwent a full neurological evaluation,

whereas 534 (94%) were tested for deafness (bilateral hearing

threshold, �80 dB).

In all, 37% of patients had received prior antimicrobial

treatment. Because inclusion of the 86 children from Buenos

Aires did not change the results, our data represent all 654

patients. Their characteristics during the hospital stay are

shown in table 2.

The causative agent was identified for 484 patients (74%).

Hib was the most common pathogen (found in 221 patients),

followed by S. pneumoniae (132 patients) and N. meningitidis

(110 patients); 21 patients had cases that were caused by other

bacteria (mostly Escherichia coli or Salmonella enteritidis). For

by guest on May 14, 2011

cid.oxfordjournals.orgD

ownloaded from

Glycerol vs. Dexamethasone in Meningitis • CID 2007:45 (15 November) • 1281

Table 2. Patient characteristics during hospital stay for the intention-to-treat population.

Variables

DXM andplacebo group

(n p 166)

DXM andglycerol group

(n p 159)

Glycerol andplacebo group

(n p 166)

Placebo andplacebo group

(n p 163)

Duration of fever, mean days 1.9 2.4 2.6 2.4

Secondary fevera 65/143 (45) 65/140 (46) 69/143 (48) 66/144 (46)

Convulsions 48/155 (31) 43/148 (29) 42/148 (28) 50/150 (33)

Duration of Glasgow Coma Scale !15, mean days 3.2 3.3 3.3 3.5

Duration of neck stiffness, mean days 3.1 3.1 2.8 3.4

Duration of irritability, mean days 2.0 2.5 2.3 2.3

Duration of poor feeding, mean days 2.4 2.2 2.3 2.8

Vomiting 43/156 (28) 40/149 (27) 51/148 (34) 36/148 (24)

Diarrhea 25/156 (16) 16/144 (11) 29/141 (21) 17/145 (12)

Visible blood in stool 6/111 (5) 6/103 (6) 1/115 (1) 2/99 (2)

Mortality, by etiological agent (no. of patients who died)

Haemophilus influenzae type b (32) 7/54 (13) 8/54 (15) 7/53 (13) 10/60 (17)

Streptococcus pneumoniae (30) 8/35 (23) 6/31 (19) 6/30 (20) 10/36 (28)

Neisseria meningitidis (1) 0/26 (0) 0/25 (0) 0/33 (0) 1/26 (4)

Other (9) 5/9 (56) 1/2 (50) 2/6 (33) 1/4 (25)

Unknown (14) 3/42 (7) 5/47 (11) 2/44 (5) 4/37 (11)

Severe neurological sequelae, by etiological agent(no. of patients with severe neurological sequelae)b

H. influenzae type b (13) 4/45 (9) 0/43 (0) 1/46 (2) 8/49 (16)

S. pneumoniae (13) 3/27 (11) 3/25 (12) 3/23 (13) 4/26 (15)

N. meningitidis (1) 0/25 (0) 0/24 (0) 0/33 (0) 1/25 (4)

Other (2) 0/4 (0) 1/1 (100) 1/4 (25) 0/3 (0)

Unknown (17) 3/38 (8) 4/41 (10) 2/41 (5) 8/34 (24)

Profound hearing loss, by etiological agent(no. of patients with profound hearing loss)c

H. influenzae type b (21) 3/46 (7) 2/45 (4) 8/43 (19) 8/47 (17)

S. pneumoniae (9) 2/26 (7) 2/24 (8) 3/23 (13) 2/25 (8)

N. meningitidis (3) 2/24 (8) 1/23 (4) 0/30 (0) 0/24 (0)

Other (1) 0/3 (0) 1/1 (100) 0/3 (0) 0/3 (0)

Unknown (9) 3/36 (8) 3/39 (8) 1/37 (3) 2/32 (5)

NOTE. Data are no. (%) of patients, unless otherwise indicated. DXM, dexamethasone.a Any fever after defeverscence of at least 24-h duration.b Blindness, quadriparesis or quadriplegia, hydrocephalus requiring a shunt, or severe psychmotor retardation.c Hearing threshold �80 dB in both ears. Audiological testing will be discussed in extenso in a separate article.

170 patients, 80 (47%) of whom had received pretreatment

antimicrobials, no causative agent was identified. In total, 174

patients were enrolled in Argentina, 143 in Ecuador, 122 in

Venezuela, 120 in the Dominican Republic, 80 in Paraguay, and

15 in Brazil.

Death, severe neurological sequelae, and hearing loss.

Table 3 shows the number of deaths (86 [13%] of 654), severe

neurological sequelae (44 [8%] of 556), deaths and severe neu-

rological sequelae combined (130 [20%] of 642), and profound

hearing loss (43 [8%] of 534) in the 4 groups. Poor outcomes

were most common in the placebo-placebo group, but statis-

tical significance was reached only with respect to severe neu-

rological sequelae and the category of severe neurological dam-

age or death. Profound hearing loss occurred with similar

frequency in all 4 groups (detailed audiological analysis will be

presented separately). For any end point, no significant inter-

action was observed between glycerol and dexamethasone.

The results with respect to outcome in the 3 adjuvant groups

tested against the placebo group are shown in table 4; this table

also gives data for children with information regarding the

timing of ceftriaxone therapy compared with that of adjuvant

therapy and data regarding etiologically confirmed cases. The

incidence of severe neurological sequelae was significantly re-

duced among patients who received glycerol alone (OR, 0.31

[95% CI, 0.13–0.76]; ) and patients who received theP p .01

dexamethasone-glycerol combination (OR, 0.39 [95% CI, 0.17–

0.93]; ), whereas among those who received dexa-P p .033

methasone alone, no more than a tendency towards a reduction

in severe neurological sequelae was observed. On the other

hand, a tendency towards lowered mortality rate was observed

in the group that received only glycerol. Receipt or nonreceipt

of pretreatment antimicrobials and the timing of their admin-

istration with respect to the initiation of an adjuvant medication

left the results essentially unchanged.

by guest on May 14, 2011

cid.oxfordjournals.orgD

ownloaded from

1282 • CID 2007:45 (15 November) • Peltola et al.

Table 3. Death, severe neurological sequelae, the composite end point of death or severe neurological sequelae, and profoundhearing loss, by treatment group.

Outcome

No. ofpatients

evaluated

DXM andplacebo group

(n p 166)

DXM andglycerol group

(n p 159)

Glycerol andplacebo group

(n p 166)

Placebo andplacebo group

(n p 163) Pa

Death

All patients 654 23/166 (14) 20/159 (13) 17/166 (10) 26/163 (16) .383

Patients with confirmed etiology 484 20/123 (16) 14/112 (13) 15/122 (12) 23/127 (18) .498

Severe neurological sequelaeb

All patients 556 10/139 (7) 8/134 (6) 7/147 (5) 19/136 (14) .022

Patients with confirmed etiology 403 7/100 (7) 4/94 (4) 5/106 (5) 13/103 (13) .081

Severe neurological sequelae or death

All patients 642 33/162 (20) 27/153 (18) 24/164 (15) 46/163 (28) .016

Patients with confirmed etiology 475 27/120 (23) 18/108 (17) 20/121 (17) 36/126 (29) .068

Profound hearing lossc

All patients 534 10/135 (7) 9/132 (7) 12/136 (9) 12/131 (9) .879

Patients with confirmed etiology 390 7/98 (7) 6/94 (6) 11/99 (11) 10/99 (10) .593

NOTE. Data are no. (%) of patients, unless otherwise indicated. DXM, dexamethasone.a Determined by x2 test between 4 groups (for heterogeneity).b Severe neurological sequelae included blindness (7 patients), quadriplegia (8), hydrocephalus requiring a shunt (4), or severe psychomotor retardation (26),

with each child counted only once.c Hearing threshold of �80 dB for the better ear.

For the composite end point group (those who experienced

severe neurological sequelae or death), the results were similar.

The OR for the glycerol-only group was 0.44 (95% CI, 0.25–

0.76; ), and for the dexamethasone-glycerol group, theP p .003

OR was 0.55 (95% CI, 0.32–0.93; ). Including theP p .027

timing of antimicrobial administration in the analysis or ex-

cluding the patients for whom an etiological agent was not

identified did not change the observation that the glycerol-only

group had the best outcomes, followed by the dexamethasone-

glycerol group; the group that received dexamethasone alone

had the worst outcomes.

Hib versus non-Hib meningitis. Effects of the adjuvant

therapies on Hib versus non-Hib meningitis are presented in

table 5. All statistically significant differences were found in the

glycerol-only group. The incidence of severe neurological se-

quelae was reduced in patients with Hib meningitis, regardless

of whether all cases were examined (OR, 0.11 [95% CI, 0.01–

0.95]; ) or only those cases with information on pre-P p .045

treatment antimicrobials and the timing of adjuvant therapy

(OR 0.11 [95% CI, 0.01–0.96]; ).P p .046

Examining severe neurological sequelae and death together,

the effect of treatment with glycerol remained statistically sig-

nificant among patients with Hib meningitis with information

on prior receipt of antimicrobials and the timing of adjuvant

therapy (OR, 0.33 [95% CI, 0.12–0.92]; ) and amongP p .035

children with non-Hib meningitis (OR, 0.46 [95% CI, 0.23–

0.91]; ). Among children with non-Hib meningitis,P p .025

even the mortality rate was somewhat reduced by receipt of

glycerol (OR, 0.49 [95% CI, 0.21–1.13]; ).P p .094

When the data were analyzed taking into consideration only

whether glycerol or dexamethasone was or was not given, the

only new finding was with respect to Hib meningitis. Analyzing

all 181 cases together and neglecting the timing between dex-

amethasone and ceftriaxone administration, treatment with

dexamethasone prevented profound hearing loss (OR, 0.27

[95% CI, 0.09–0.77]; ). Statistical significance was lostP p .014

if data from patients with cases of Hib meningitis who had not

received pretreatment antimicrobials (123 patients) were ex-

amined separately.

Safety. Few adverse effects attributable to either adjuvant

medication were observed. Visible blood in the stool was noted

in 6 (5%) of 111 patients in the dexamethasone group and in

5 (5%) of 101 patients in the dexamethasone-glycerol group

but in only 1 (1%) of 113 patients in the glycerol-only group

and 2 (2%) of 99 patients in the placebo group (table 2). The

difference between the dexamethasone recipients and those who

did not receive dexamethasone was statistically significant

( ). Vomiting or diarrhea were not more common inP p .032

the glycerol group, and massive gastrointestinal hemorrhage or

other severe adverse events were not found in any child.

DISCUSSION

This prospective, double-blind, randomized trial from 6 Latin

American countries that involved 654 patients is, to our knowl-

edge, the largest clinical study of childhood bacterial meningitis

to date. Oral glycerol improved outcomes especially by reducing

severe neurological sequelae. The study was performed in con-

ditions similar to those in which the great majority of children

with meningitis worldwide are treated [3, 4, 27–29]. Many of

the children in our study presented late in the course of disease,

many were anemic, and many had previously been given oral

by guest on May 14, 2011

cid.oxfordjournals.orgD

ownloaded from

Glycerol vs. Dexamethasone in Meningitis • CID 2007:45 (15 November) • 1283

Table 4. Risk of death, severe neurological sequelae, the composite end point of death or severe neurological sequelae, and profoundhearing loss in the 3 adjuvant therapy groups, compared with the placebo only group.

Outcome

No. ofpatients

evaluated

DXM andplacebo group

(n p 166)

DXM andglycerol group

(n p 159)

Glycerol andplacebo group

(n p 166)

OR (95% CI) P OR (95% CI) P OR (95% CI) P

Death

Overall 654 0.82 (0.45–1.49) .509 0.69 (0.37–1.31) .257 0.58 (0.30–1.11) .099

Including receipt of prior antimicrobial and timing of DXMadministrationa

All cases 578 0.92 (0.48–1.76) .796 0.79 (0.40–1.56) .502 0.60 (0.29–1.22) .155

Cases with confirmed etiology 430 0.94 (0.46–1.90) .853 0.75 (0.35–1.62) .466 0.63 (0.29–1.37) .239

Severe neurological sequelaeb

Overall 556 0.48 (0.21–1.07) .072 0.39 (0.17–0.93) .033 0.31 (0.13–0.76) .010

Including receipt of prior antimicrobial and timing of DXMadministrationa

All cases 497 0.50 (0.21–1.19) .119 0.46 (0.19–1.11) .084 0.37 (0.14–0.93) .034

Cases with confirmed etiology 363 0.52 (0.19–1.39) .192 0.32 (0.10–1.03) .057 0.36 (0.12–1.07) .065

Combined outcome of severe neurological sequelae or death

Overall 642 0.65 (0.39–1.09) .100 0.55 (0.32–0.93) .027 0.44 (0.25–0.76) .003

Including receipt of prior antimicrobial and timing of DXMadministrationa

All cases 572 0.72 (0.42–1.25) .241 0.63 (0.36–1.11) .108 0.47 (0.26–0.85) .013

Cases with confirmed etiology 425 0.75 (0.41–1.38) .357 0.56 (0.28–1.09) .088 0.49 (0.25–0.96) .037

Profound hearing loss

Overall 534 0.79 (0.33–1.91) .604 0.73 (0.30–1.79) .485 0.96 (0.42–2.22) .923

Including receipt of prior antimicrobial and timing of DXMadministrationa

All cases 476 0.70 (0.27–1.81) .455 0.78 (0.31–1.98) .603 1.04 (0.44–2.47) .932

Cases with confirmed etiology 351 0.56 (0.19–1.61) .279 0.62 (0.21–1.80) .378 1.11 (0.44–2.78) .825

NOTE. P values were determined by Wald’s test. DXM, dexamethasone.a Receipt versus nonreceipt of prior antimicrobial treatment and timing of adjuvant medication administration versus initiation of ceftriaxone therapy (before

vs. at the same time or later) included as covariates.b Severe neurological sequelae included blindness (7 patients), quadriplegia (8), hydrocephalus requiring a shunt (4), or severe psychomotor retardation (26),

with each child counted only once.

antimicrobials. A few of these children may have been infected

with HIV, although the prevalence of HIV infection in Latin

America was low during the study period [30].

Because most children could not be expected to return for

a control visit, we were forced to use the outcome data that

were available at hospital discharge. Some subtle sequelae may,

therefore, have remained undetected, but this shortcoming

probably did not distort the results, because most of the se-

quelae described in our study were very unlikely to resolve with

time. Children with these sequelae probably died or remained

permanently disabled. Other limitations of the study were the

inclusion of patients who had received pretreatment antimi-

crobial drugs (which are administered indiscriminately in Latin

America) and patients with etiologically unidentified cases;

however, the study design distributed such patients evenly to

each arm. Because the details of antimicrobial administration

were recorded meticulously, we could also examine children

who did not receive a single dose of any antimicrobial before

hospitalization.

The comprehensiveness of the series allowed us to analyze

the adjuvant effects separately for each outcome. Except for 2

earlier studies that were sufficiently powered—one from Africa,

involving children [11], and the other from Europe, involving

adults who received dexamethasone for 4 days [31]—this has

not been possible before. By keeping the outcomes separated,

potential benefits between adjuvants could be examined spe-

cifically. Glycerol not only reduced severe neurological sequelae

but was also beneficial when severe neurological sequelae and

mortality were examined together [4, 5]. Nevertheless, preven-

tion of severe neurological sequelae alone was a major achieve-

ment; many clinicians classify survival with severe sequelae and

death as poor outcomes of equal importance [4]. Reducing the

impact of meningitis on the family and on society clearly re-

quires more than merely increasing the survival rate [32].

How does glycerol work? One-third of the children with

bacterial meningitis experience significantly impaired cerebral

blood flow [33, 34] and intracranial edema. Oral glycerol in-

creases plasma osmolality, which, among other effects, lowers

intraocular pressure within 30 min [17]. Oral glycerol also in-

creases plasma osmolality in children with bacterial meningitis

by guest on May 14, 2011

cid.oxfordjournals.orgD

ownloaded from

1284 • CID 2007:45 (15 November) • Peltola et al.

Table 5. Risk of death, severe neurological sequelae, the composite end point of death or severe neurological sequelae, and profoundhearing loss among the 3 adjuvant therapy groups, compared with the placebo group, in patients with Haemophilus influenzae typeb (Hib) meningitis and patients with non-Hib meningitis.

Outcome

No. ofpatients

evaluated

DXM andplacebo group

(n p 166)

DXM andglycerol group

(n p 159)

Glycerol andplacebo group

(n p 166)

OR (95% CI) P OR (95% CI) P OR (95% CI) P

Death

Hib meningitis

Overall 221 0.75 (0.26–2.12) .508 0.87 (0.32–2.39) .787 0.76 (0.27–2.17) .608

Including prior antimicrobial and DXMadministration timinga 204 0.82 (0.28–2.42) .724 0.79 (0.27–2.30) .669 0.62 (0.19–2.00) .422

Optimal DXM therapyb 114 1.71 (0.39–7.60) .478 0.73 (0.11–4.77) .740 0.92 (0.17–5.02) .926

Non-Hib meningitis, overall 433 0.84 (0.40–1.77) .652 0.65 (0.30–1.45) .293 0.49 (0.21–1.13) .094

Severe neurological sequelaec

Hib meningitis

Overall 183 0.50 (0.14–1.79) .287 … NE 0.11 (0.01–0.95) .045

Including prior antimicrobial and DXMadministration timinga 172 0.46 (0.13–1.68) .240 … NE 0.11 (0.01–0.96) .046

Optimal DXM therapyb 100 0.29 (0.05–1.67) .167 … NE 0.15 (0.02–1.41) .098

Non-Hib meningitis, overall 374 0.43 (0.16–1.21) .109 0.61 (0.24–1.57) .307 0.40 (0.14–1.12) .080

Severe neurological sequelaec or death

Hib meningitis

Overall 215 0.61 (0.26–1.45) .265 0.42 (0.17–1.08) .072 0.41 (0.16–1.03) .058

Including prior antimicrobial and DXMadministration timinga 201 0.64 (0.26–1.58) .319 0.39 (0.15–1.06) .064 0.33 (0.12–0.92) .035

Optimal DXM therapyb 114 0.73 (0.23–2.30) .591 0.21 (0.04–1.09) .063 0.40 (0.10–1.52) .176

Non-Hib meningitis, overall 427 0.68 (0.36–1.28) .233 0.62 (0.32–1.20) .158 0.46 (0.23–0.91) .025

Profound hearing loss

Hib meningitis

Overall 181 0.34 (0.08–1.37) .130 0.23 (0.05–1.13) .071 1.11 (0.38–3.29) .845

Including prior antimicrobial and DXMadministration timinga 168 0.37 (0.09–1.52) .167 0.26 (0.05–1.33) .106 1.15 (0.38–3.47) .800

Optimal DXM therapyb 97 0.19 (0.02–1.86) .154 0.23 (0.02–2.21) .202 0.48 (0.08–2.89) .420

Non-Hib meningitis, overall 353 1.71 (0.48–6.06) .408 1.75 (0.49–6.21) .387 0.90 (0.22–3.71) .883

NOTE. Intention-to-treat subgroup analysis. For outcomes, see table 3. P values were determined by Wald’s test. DXM, dexamethasone; NE, not estimable.a Receipt versus nonreceipt of prior antimicrobial treatment and timing of adjuvant medication administration versus initiation of ceftriaxone therapy (before

vs. at the same time or later) included as covariates.b Patients with no prior antimicrobial treatment and ceftriaxone administered 15 min after dexamethasone.c Severe neurological sequelae included blindness (7 patients), quadriplegia (8), hydrocephalus requiring a shunt (4), or severe psychomotor retardation (26),

with each child counted only once.

(S. Singhi, personal communication), as it does in adult healthy

volunteers [35]. This change in osmolality reduces edema and

enhances cerebral circulation by reducing the excretion of CSF

[36]. As water moves by osmosis back into the plasma, extra-

vascularization of water and subsequent occult hypovolemia

are reduced. Decrease of the intracranial pressure by glycerol-

induced osmotic diuresis seems to be less important [37], be-

cause a gradient between the body compartments requires an

intact or almost intact blood-brain barrier, and this is not the

case in patients with bacterial meningitis. Glycerol is also a

scavenger of free oxygen radicals, and this action may further

alleviate meningeal inflammation.

Our glycerol dosage (6 g per kg of body weight per day

divided into 4 doses) was not evidence-based but was derived

from dosages recommended earlier [38] in neurology and neu-

rosurgery. Our practice of keeping to a single maximum dose

of 25 mL was based on our previous experience [19], which

showed that some children vomited if they received a larger

total dose. In our study, vomiting was not more common

among glycerol recipients. No data suggest the substitution of

another osmotic diuretic, such as mannitol, for glycerol; in fact,

intravenous mannitol may be harmful in patients with bacterial

meningitis [39].

Further studies on adjuvant glycerol to treat bacterial men-

ingitis are clearly warranted. Interestingly, although the Malawi

trial [11] found no advantage associated with dexamethasone

treatment, some of our patients benefited. This was best shown

among patients with Hib meningitis when the dexamethasone

recipients were compared with the non-recipients and the tim-

ing of antimicrobial administration was not taken into account.

by guest on May 14, 2011

cid.oxfordjournals.orgD

ownloaded from

Glycerol vs. Dexamethasone in Meningitis • CID 2007:45 (15 November) • 1285

Unfortunately, we cannot easily identify patients who likely

would benefit from adjuvant dexamethasone. Evidently, they

are not simply those patients with nonpretreated Hib meningitis

for whom cephalosporin therapy was initiated after giving

dexamethasone.

In conclusion, oral glycerol reduced the incidence of severe

neurological sequelae associated with childhood bacterial men-

ingitis. Five properties of glycerol make its widespread use pos-

sible and desirable: it can be taken orally, is inexpensive, is

easily available, has no special storage requirements, and is safe

[18, 19, 37, 40]. Glycerol is a common ingredient of ingested

(e.g., chewing gum) and topical (e.g., shampoo) substances.

Some patients have received intravenous glycerol for 1 week

[14]; we gave 8 oral doses over a 48-h period.

Glycerol is a novel tool for the treatment of bacterial men-

ingitis, which is a life-threatening disease that ∼5000 children

per day contract worldwide. Oral glycerol is also an agent that

is cheap enough and simple enough for treatment of the poorest

patients. Since the advent of chloramphenicol and ampicillin

∼40 years ago, no other medication has improved the prognosis

of childhood meningitis, especially Hib meningitis, as much as

glycerol.

STUDY CONTRIBUTORS

Jesus Feris-Iglesias and Chabela Pena (Santo Domingo, Do-

minican Republic); Mariella Chang and Ruth Flor (Guayaquil,

Ecuador); Marıa Rosa Agosti (La Plata, Argentina); Miriam

Maitin and Lesbia Colina (Barquisimeto, Venezuela); Dolores

Lovera (Asuncion, Paraguay); Marıa Teresa Rosanova, Ilse Vil-

laroel, and Mari Carmen Cifro (Buenos Aires, Argentina); Mag-

dalena Correa (Merida, Venezuela); and Marcos Fernandes and

Vania Prazeres (Manaus, Brazil).

Bacteriological and other laboratory investigations were di-

rected by Jacqueline Sanchez (Santo Domingo, Dominican Re-

public); Rafael Roas (Barquisimeto, Venezuela); Cecilia Vescina,

Marta Altschuler, and Patricie Lazarte (La Plata, Argentina);

Wilma Basualdo (Asuncion, Paraguay); Maria del Carmen Cei-

nos (Buenos Aires, Argentina); and Rossicleia Monte (Manaus,

Brazil).

Audiological examinations were performed by Clemente Ter-

rero (Santo Domingo, Dominican Republic); Beila Pire (Bar-

quisimeto, Venezuela); Pedro Toledo (Guayaquil, Ecuador);

Luis Pedersoli, Alicia Calcaterra, and Silvia Jury (La Plata, Ar-

gentina); Arturo Campos (Asuncion, Paraguay); and Marıa E.

Prieto (Buenos Aires, Argentina).

The formula for the placebo of glycerol was developed by

Pedro Valora (Buenos Aires, Argentina).

Acknowledgments

We thank Dr. Ralf Clemens, who organized the initial grant for thisnonprofit study, and Dr. Elizabeth Molyneux, who revised the text. Ossi

Hiltunen (Orion Diagnostica) kindly provided the equipment for quan-titative C-reactive protein measurements.

Financial support. GlaxoSmithKline, Alfred Kordelin, Paivikki andSakari Sohlberg, and Sigfrid Juselius Funds. Farmacia Ahumada donatedglycerol and both placebo preparations. Laboratorio de Chile partly donatedceftriaxone.

Potential conflicts of interest. H.P. is currently a scientific consultantof Serum Institute of India. All other authors: no conflicts.

References

1. Peltola H. Worldwide Haemophilus influenzae type b disease at thebeginning of the 21st century; global analysis of the disease burden 25years after the use of polysaccharide vaccine and a decade after theadvent of conjugates. Clin Microbiol Rev 2000; 13:302–17.

2. Daoud AS, Al-Sheyyab M, Batchoun RG, et al. Bacterial meningitis:still a cause of high mortality and severe neurological morbidity inchildhood. J Trop Med Paed 1995; 41:308–10.

3. Goetghebuer T, West TE, Wermenbol V, et al. Outcome of meningitiscaused by Streptococcus pneumoniae and Haemophilus influenzae typeb in children in The Gambia. Trop Med Int Health 2000; 5:207–13.

4. Duke T, Mokela D, Frank D, et al. Management of meningitis inchildren with oral fluid restriction or intravenous fluid at maintenancevolumes: a randomized trial. Ann Trop Paediatr 2002; 22:145–57.

5. Principi N, Esposito S. Dexamethasone in acute bacterial meningitis.Lancet 2002; 360:1610.

6. Peltola H, Anttila M, Renkonen O-V; The Finnish Study Group. Ran-domised comparison of chloramphenicol, ampicillin, cefotaxime, andceftriaxone for childhood bacterial meningitis. Lancet 1989; 1:1281–7.

7. Mustafa MM, Ramilo O, Mertsola J, et al. Modulation of inflammationand cachectin activity in relation to treatment of experimental Hae-mophilus influenzae type b meningitis. J Infect Dis 1989; 160:818–25.

8. Lebel MH, Freij BJ, Syrogiannopoulos GA, et al. Dexamethasone ther-apy for bacterial meningitis: results of two double-blind, placebo-con-trolled trials. N Engl J Med 1988; 319:964–71.

9. Odio CM, Faingezicht I, Paris M, et al. The beneficial effects of earlydexamethasone administration in infants and children with bacterialmeningitis. N Engl J Med 1991; 324:1525–31.

10. Schaad UB, Lips U, Gnehm HE, Blumberg A, Wedgwood J. Dexa-methasone therapy for bacterial meningitis in children. The Swiss Men-ingitis Study Group. Lancet 1993; 342:457–61.

11. Molyneux EM, Walsh AL, Forsyth H, et al. Dexamethasone treatmentin childhood bacterial meningitis in Malawi: a randomised controlledtrial. Lancet 2002; 360:211–8.

12. van de Beek D, de Gans J, McIntyre P, Prasad K. Corticosteroids inacute bacterial meningitis. Cochrane Database Syst Rev 2007:CD004405.

13. Buckell M, Walsh L. Effect of glycerol by mouth on raised intracranialpressure in man. Lancet 1964; 2:1151–2.

14. Gilsanz V, Rebollar JL, Buencuerpo J, Chantres MT. Controlled trialof glycerol versus dexamethasone in the treatment of cerebral oedemain acute cerebral infarction. Lancet 1975; 1:1049–51.

15. Rottenberg DA, Hurwitz BJ, Posner JB. The effect of oral glycerol onintraventricular pressure in man. Neurology 1977; 27:600–8.

16. Bayer AJ, Pathy MSJ, Newcombe R. Double-blind randomised trial ofintravenous glycerol in acute stroke. Lancet 1987; 1:405–8.

17. McCurdy DK, Schneider B, Scheie HG. Oral glycerol: the mechanismon intraocular hypotension. Am J Ophthalmol 1966; 61:1244–9.

18. Herson VC, Todd JK. Prediction of morbidity in Haemophilus influen-zae meningitis. Pediatrics 1977; 59:35–9.

19. Kilpi T, Peltola H, Kallio MK, et al. Oral glycerol versus intravenousdexamethasone in preventing hearing impairment due to childhoodbacterial meningitis. Pediatr Infect Dis J 1995; 14:270–8.

20. Peltola HO. C-reactive protein in rapid monitoring of infections ofthe central nervous system. Lancet 1982; 1:980–2.

21. Roine I, Banfi A, Bosch P, Ledermann W, Contreras C, Peltola H.Serum C-reactive protein in childhood meningitis in countries with

by guest on May 14, 2011

cid.oxfordjournals.orgD

ownloaded from

1286 • CID 2007:45 (15 November) • Peltola et al.

limited laboratory resources: a Chilean experience. Pediatr Infect DisJ 1991; 10:923–8.

22. Tureen JM, Tauber MG, Sande MA. Effect of hydration status oncerebral blood flow and cerebrospinal fluid lactic acidosis in rabbitswith experimental meningitis. J Clin Invest 1992; 89:947–53.

23. Singhi SC, Singhi PD, Srinivas B, et al. Fluid restriction does notimprove the outcome of acute meningitis. Pediatr Infect Dis J 1995;14:495–503.

24. Syrogiannopoulos GA, Lourida AN, Theodoridou MC, et al. Dexa-methasone therapy for bacterial meningitis in children: 2- versus 4-day regimen. J Infect Dis 1994; 169:853–8.

25. Teasdale G, Jennett B. Assessment of coma and impaired consciousness:a practical scale. Lancet 1974; 2:81–4.

26. Glascoe FP, Byrne KE, Ashford LK, Johnson KL, Chang B, StricklandB. Accuracy of the Denver-II in developmental screening. Pediatrics1992; 89:1221–5.

27. Peltola H. Haemophilus influenzae type b disease and vaccination inLatin America and the Caribbean. Pediatr Infect Dis J 1997; 16:780–7.

28. Basualdo W, Arbo A. Invasive Haemophilus influenzae type b infectionsin children in Paraguay. Arch Med Res 2004; 35:126–33.

29. Peltola H. Burden of meningitis and other severe bacterial infectionsof children in Africa: implications for prevention. Clin Infect Dis2001; 32:64–75.

30. Salooje H, Violari A. HIV infection in children. BMJ 2001; 323:670–4.31. de Gans J, van de Beek D; European Dexamethasone in Adulthood

Bacterial Meningitis Study Investigators. Dexamethasone in adults withbacterial meningitis. N Engl J Med 2002; 347:1549–56.

32. Dexamethasone therapy for bacterial meningitis. New Engl J Med1989; 320:463–5.

33. Ashwal S, Stringer W, Tornasi L, Schneider S, Thompson J, Perkin R.Cerebral blood flow and carbon dioxide reactivity in children withbacterial meningitis. J Pediatr 1990; 117:523–30.

34. Møller K, Larsen FS, Qvist J, et al. Dependency of cerebral blood flowon mean arterial pressure in patients with acute meningitis. Crit CareMed 2000; 28:1027–32.

35. Sommer S, Nau R, Wieland E, Prange HW. Pharmacokinetics of glyc-erol administered orally in healthy volunteers. Arzneimittelforschung1993; 43:744–7.

36. Zoghbi HV, Okumura S, Laurent JP, Fishman MA. Acute effect ofglycerol on net cerebrospinal fluid production in dogs. J Neurosurg1985; 63:759–62.

37. Nau R, Prins F-J, Kolenda H, Prange HW. Temporary reversal of serumto cerebrospinal fluid glycerol concentration gradient after intravenousinfusion of glycerol. Eur J Clin Pharmacol 1992; 42:181–5.

38. Frank MSB, Nahata MC, Hilty MD. Glycerol: a review of its phar-macology, phamacokinetics, adverse reactions, and clinical use. Phar-macotherapy 1981; 1:147–60.

39. Syrogiannopoulos GA, Olsen KD, McCracken GH. Mannitol treatmentin experimental Haemophilus influenzae type b meningitis. Pediatr Res1987; 22:118–22.

40. Tourtelotte WW, Reinglass JL, Newkirk TA. Cerebral dehydration ac-tion of glycerol. I. Historical aspects with emphasis on the toxicity andintravenous administration. Clin Pharmacol Ther 1972; 13:159–71.

by guest on May 14, 2011

cid.oxfordjournals.orgD

ownloaded from