MAJOR ARTICLE
HIV/AIDS
The Impact of a 9-Valent Pneumococcal Conjugate
Vaccine on the Public Health Burden of Pneumonia
in HIV-Infected and -Uninfected Children
Shabir A. Madhi,1,2 Locadiah Kuwanda,1 Clare Cutland,1,2 and Keith P. Klugman1,3
1
National Health Laboratory Service/University of the Witwatersrand/Medical Research Council Respiratory and Meningeal Pathogens Research
Unit and 2Paediatric Infectious Diseases Research Unit, Wits Health Consortium, University of the Witwatersrand, Johannesburg, South Africa;
and 3Department of Global Health, Rollins School of Public Health and Division of Infectious Diseases, School of Medicine, Emory University,
Atlanta, Georgia
(See the editorial commentary by Greenwood on pages 1519–20)
Introduction. Pneumococcal conjugate vaccine (PnCV) may be used as a probe to define the burden of
pneumococcal disease and better characterize the clinical presentation of pneumococcal pneumonia.
Methods. This study used a 9-valent PnCV to define different end points of vaccine efficacy and the preventable
burden of pneumococcal pneumonia in 39,836 children who were randomized in a double-blind, placebo-controlled
trial in South Africa.
Results. Whereas the point-estimate of vaccine efficacy was greatest when measured against the outcome of
vaccine-serotype specific pneumococcal bacteremic pneumonia (61%; P p .01 ), the sensitivity of blood culture to
measure the burden of pneumococcal pneumonia prevented by vaccination was only 2.6% in human immunodeficiency virus (HIV)–uninfected children and 18.8% in HIV-infected children. Only 37.8% of cases of pneumococcal pneumonia prevented by PnCV were detected by means of chest radiographs showing alveolar consolidation. A clinical diagnosis of pneumonia provided the best estimate of the burden of pneumococcal pneumonia
prevented through vaccination in HIV-uninfected children (267 cases prevented per 100,000 child-years) and HIVinfected children (2573 cases prevented per 100,000 child-years).
Conclusion. Although outcome measures with high specificity, such as bacteremic pneumococcal pneumonia,
provide a better estimate as to vaccine efficacy, the burden of disease prevented by vaccination is best evaluated
using outcome measures with high sensitivity, such as a clinical diagnosis of pneumonia.
The lack of a gold standard tool that is high in sensitivity
and specificity for diagnosing the etiology of pneumonia is a major hurdle in defining the efficacy and
potential public health benefit of vaccines against pneumonia in children. Although the World Health Organization (WHO) clinical criteria for the diagnosis of
severe pneumonia are highly sensitive (∼80%), they
have a low positive predictive value (25%) [1, 2]. In
contrast, blood cultures in individuals with pneumonia
Received 3 November 2004; accepted 4 January 2005; electronically published
7 April 2005.
Reprints or correspondence: Dr. Shabir A. Madhi, PO Bertsham, Chris HaniBaragwanath Hospital, Old Nurses Home, 1st Flr. West Wing, Bertsham, Gauteng,
2013, South Africa ([email protected]).
Clinical Infectious Diseases 2005; 40:1511–8
2005 by the Infectious Diseases Society of America. All rights reserved.
1058-4838/2005/4010-0019$15.00
lack sensitivity (10%–15%) but have good specificity
for defining the bacterial etiology of pneumonia [3].
Although lung puncture taps are an alternate strategy
for determining an etiological diagnosis of pneumonia,
they are rarely used and are mainly performed in cases
in which there is a distinctly accessible area of consolidation [4]. On the basis of the postulate that the most
likely chest radiographic manifestation of pneumococcal pneumonia is alveolar consolidation [5], a working
group of the WHO recently published guidelines for
the interpretation and reporting of chest radiographs
[5]. Chest radiographs, however, have been found to
be unreliable for distinguishing between bacterial and
viral pneumonia [6–8]. This study determined the efficacy of pneumococcal conjugate vaccine (PnCV) and
measured the relative burden of disease prevented
through vaccination by comparing different outcome
measures that could be used as surrogate markers of
pneumococcal pneumonia.
HIV/AIDS • CID 2005:40 (15 May) • 1511
METHODS
We have previously reported on the efficacy of a 9-valent
PnCV in reducing the rate of chest radiograph–confirmed
pneumonia [9] and against clinically and/or radiographicly
confirmed pneumonia [10]. The purpose of the current hypothesis-generating analysis was to compare the relative sensitivity of the above-mentioned outcomes with those of blood
culture–confirmed pneumococcal pneumonia and various
modifications of WHO-defined clinically diagnosed pneumonia [1] to determine which of the outcome measures provides the best estimate of the burden of pneumococcal pneumonia prevented by vaccination.
Study population. The demographic information for the
children participating in this study has been detailed elsewhere
[9]. In brief, the study was a double-blind, placebo-controlled
trial that included 39,836 children who were randomized to
receive either a 9-valent PnCV (Wyeth Vaccines and Pediatrics)
or placebo at 6, 10, and 14 weeks of age.
Passive surveillance for all-cause hospitalization was conducted by study staff at the local hospital (Chris Hani-Baragwanath Hospital; Soweto, South Africa). Indications for hospitalization were at the discretion of the hospital physicians.
All children were examined by one of the study staff when
hospitalized for any illness. A standard examination form was
completed that captured clinical signs and symptoms of lower
respiratory tract infection (LRTI). Outcome cases included in
this report were those that occurred through 15 November
2001.
HIV-1 prevalence and testing. On the basis of the measured prevalence of HIV infection among women attending
antenatal clinics during the course of the study period (22.5%–
27.8%), it was estimated that 24.87% of the children enrolled
into the study were born to HIV-infected mothers [11]. Furthermore, on the basis of the estimated vertical transmission
rate of HIV infection from mother to child in the absence of
any antiretroviral intervention (26%) [12], it was estimated that
6.47% of the children recruited into either arm of the study
were infected with HIV. The denominators for calculating the
incidence rate and vaccine-attributable reduction (VAR) in disease in HIV-infected and HIV-uninfected children were based
on the estimates above. Testing for HIV infection status of
children hospitalized with LRTIs was performed as described
elsewhere [9].
Bacterial cultures. Blood samples were cultured for bacterial
growth at admission to the hospital for all children with suspected
LRTI and were processed using the BacT/Alert microbial detection system (Organon Teknika), and isolates of Streptococcus
pneumoniae were serotyped as described elsewhere [9].
Chest radiographs. Chest radiographs were obtained for
all children with a suspected LRTI and were interpreted for the
1512 • CID 2005:40 (15 May) • HIV/AIDS
presence of alveolar infiltrate using WHO recommendations
for reporting [5, 9].
Definitions. The efficacy of the vaccine against possible
pneumococcal pneumonia was analyzed using the following
outcome measures. WHO-confirmed radiological pneumonia
(WHO-AC) was defined as pneumonia associated with chest
radiograph–confirmed alveolar consolidation (i.e., the presence
of a dense opacity that could be a fluffy consolidation of a
portion of a lobe, a whole lobe, or the entire lung, often containing air bronchograms and sometimes associated with pleural effusion, or a pleural effusion in the lateral pleural space
associated with a pulmonary infiltrate or an effusion large
enough to obscure such an opacity), in accordance with WHO
recommendations [5]. Clinical LRTI (C-LRTI) was defined to
include cases in all children hospitalized with a study physician
diagnosis of pneumonia or bronchiolitis, irrespective of their
clinical or radiographic features. Clinical pneumonia was defined as the presence of WHO-AC or was considered to have
occurred if the child fulfilled the clinical diagnosis of LRTI
without wheeze on chest auscultation but had crackles and/or
bronchial breathing on chest wall auscultation [10]. Bronchiolitis was defined as the presence of wheezing on chest auscultation performed by one of the study doctors in the absence
of documented alveolar consolidation (i.e., WHO-AC) on chest
radiography or bronchial breathing on chest wall auscultation.
WHO-defined mild pneumonia was defined as cough of !14
days duration in a child with tachypnea (defined as 150 breaths/
min in children !12 months of age and 140 breaths/min in
children ⭓12 months of age) in the absence of lower chest wall
in-drawing or other signs and symptoms of WHO-defined severe pneumonia [1]. WHO-defined severe pneumonia was defined as a cough of !14 days duration in a child with lower
chest wall in-drawing and/or any of the following signs and
symptoms of severe pneumonia: feeding difficulties, convulsions, central cyanosis, or encephalopathy [1]. Bacteremic
pneumococcal pneumonia was defined as pneumonia associated with a blood culture positive for S. pneumoniae, irrespective of the serotype. Vaccine serotype–specific bacteremic pneumococcal pneumonia was defined as pneumonia associated
with a blood culture positive for S. pneumoniae belonging to
1 of the 9 vaccine serotypes. VAR was defined as the burden
of disease prevented expressed in number of cases per 100,000
child-years of observation.
Statistical methods. Data were analyzed using Stata software, version 8.0 (StataCorp), and EpiInfo software, version
6.04d (Centers for Disease Control and Prevention). Vaccine
efficacy (VE) was calculated using the VE calculation function
in EpiInfo software for cohort studies. This is based on the
formula VE (expressed as a percentage) p {(attack rate in the
unvaccinated⫺attack rate in the vaccinated)/attack rate in the
unvaccinated} ⫻ 100. VAR was calculated on the basis of the
formula VAR p (incidence rate of disease in the unvaccinated
cohort⫺incidence rate of disease in the vaccinated cohort)/
median follow-up time (which was 847.5 days in the PnCV
group and 847 days in the placebo group) and expressed as
number of cases per 100,000 child-years of observation. Categorical data were analyzed using the Pearson x2 test or the
Yates x2 test if an expected cell value was !5. A P value of ⭐.05
was considered statistically significant.
All analyses performed on an intent-to-treat (ITT) basis included any child that had received at least a single dose of study
vaccine from the date of receiving the first dose of study vaccine;
per-protocol analysis included only children who had received
3 doses of the same study vaccine within protocol-defined time
periods [9]. Only the first episode of an event per individual
was considered for all analyses.
Ethical issues. This study was approved by the Ethics Committee for research on Human Subjects of the University of
the Witwatersrand (Johannesburg, South Africa), and signed
informed consent was obtained from the parent or legal guardian of the children at the time of enrollment into the study.
RESULTS
VE in entire study cohort irrespective of the HIV status of the
children. Among children with C-LRTI who were included
in the ITT analysis, chest radiographs were unavailable for 115
(7.7%) of 1493 vaccinees and 119 (7.2%) of 1646 placebo recipients (P p .61); blood cultures were not performed for 55
(3.7%) of the vaccinees and 72 (4.4%) of the placebo recipients
with C-LRTI (P p .33).
Although the point estimate of VE was greatest for vaccine serotype–specific bacteremic pneumococcal pneumonia
(60.9%), the outcome measure of clinical pneumonia was the
most sensitive in measuring the burden of pneumococcal pneumonia prevented by vaccination (VAR, 410 cases per 100,000
child-years; table 1). Using the outcome of clinical pneumonia
as the benchmark of the burden of disease prevented by vaccination, the sensitivity of other outcome measures to detect
the burden of pneumococcal pneumonia prevented was determined by comparing the measured VAR with the VAR of clinical
pneumonia (VAR, 410 cases per 100,000 child-years). The sensitivities of WHO-defined severe pneumonia (VAR, 279 cases
per 100,000 child-years), WHO-AC (VAR, 155 cases per
100,000 child-years), and pneumococcal bacteremic pneumonia (VAR, 37 cases per 100,000 child-years) in detecting the
burden of disease prevented through vaccination were 68.0%
(95% CI, 63.3%–72.5%), 37.8% (95% CI, 33.1%–42.7%), and
9.0% (95% CI, 6.4%–1.2%), respectively. However, vaccinees
were more likely than other subjects to be hospitalized for
WHO-defined mild pneumonia (relative risk [RR], 1.2; 95%
CI, 1–1.46), although this increase was not statistically significant when children with concurrent wheezing were excluded
(table 1). The HIV infection status was unknown or indeterminate for 48 (3.2%) of 1493 vaccinees and 53 (3.2%) of 1646
control children who were hospitalized for LRTI (P p .99).
VE in HIV-uninfected children. Table 2 shows the VE
against the different outcome measures of possible pneumococcal pneumonia in HIV-uninfected children. Among children
with C-LRTI who were included in the ITT analysis, chest
radiographs were unavailable for 70 (6.8%) of 1033 vaccinees
and 65 (5.9%) of 1106 placebo recipients (P p .39); blood
cultures were not performed for 39 (3.8%) of the vaccinees
and 45 (4.1%) of the placebo recipients (P p .73).
The outcome measure of clinical pneumonia yielded the
highest point estimate of the burden of hospitalized pneumococcal pneumonia prevented through vaccination (VAR, 267
cases per 100,000 child-years; P p .001) (table 1). The sensitivities of WHO-defined severe pneumonia (VAR, 164 cases per
100,000 child-years), WHO-AC (VAR, 100 cases per 100,000
child-years), and pneumococcal bacteremic pneumonia (VAR,
7 cases per 100,000 child-years), relative to the clinical pneumonia outcome in measuring the burden of pneumonia prevented by the PnCV, were 61.4% (95% CI, 55.3%–66.3%),
37.5% (95% CI, 31.6%–43.6%), and 2.6% (95% CI, 1.0%–
5.3%), respectively.
The lower sensitivity (64.4%; 95% CI, 58.4%–70.2%) of CLRTI (VAR, 172 cases per 100,000 child-years) for detection of
the burden of pneumonia prevented through vaccination was
most likely because of an excess (RR, 1.3; 95% CI, 1.1–1.6) of
WHO-defined mild pneumonia observed among vaccinees (table 2). As observed for the entire cohort of children, the burden
of WHO-defined mild pneumonia was greater for vaccinees
than for placebo recipients in the ITT (P p .02) and per-protocol analyses (P p .02); however, this was not statistically significant when children with wheezing were excluded (table 2).
There was a nonsignificant decrease in the overall (ITT) incidence of mechanical ventilation and a trend of fewer deaths
from C-LRTI in vaccinees (table 2).
VE against pneumonia in HIV-infected children. Chest
radiographs were unavailable for 35 (8.5%) of 412 vaccinees
and 47 (9.7%) of 487 placebo recipients for their first episodes
of C-LRTI (P p .59); blood culture results were unavailable for
8 (1.9%) of the vaccinees and 13 (2.7%) of the placebo recipients (P p .47).
Despite the lack of a statistically significant difference between vaccinees and placebo recipients for WHO-AC (13%;
P p .19), vaccinees had a significantly lower incidence of hospitalization when evaluating the outcomes of C-LRTI (15%;
P p .002), clinical pneumonia (15%; P p .004), and WHOdefined severe pneumonia (17%; P p .006 (table 3). The outcome measure with the greatest sensitivity for detecting the
burden of pneumococcal pneumonia prevented by vaccination
in HIV-infected children was that of C-LRTI (VAR, 2573 cases
HIV/AIDS • CID 2005:40 (15 May) • 1513
Table 1. Estimated efficacy of a 9-valent pneumococcal conjugate vaccine in preventing pneumococcal pneumonia in hospitalized HIV-infected and HIV-uninfected
children, by analysis.
Per-protocol analysis
Intent-to-treat analysis
Outcome measure
Clinical LRTI
Clinical pneumonia
Bronchiolitis
No. of
No. of
vaccine
placebo
recipients
recipients
(n p 19,922) (n p 19,914)
1514
1493
975
650
1646
1162
643
WHO-ACf
356
428
WHO-defined mild pneumonia
WHO-defined mild pneumonia
without wheezing
WHO-defined severe pneumonia
WHO-defined severe pneumonia
without wheezing
Bacteremic pneumocooccal pneumonia
All
Associated with a vaccine serotype
IPPV for LRTI
Death due to LRTI
232
192
100
916
105
1043
598
22
9
18
161
Vaccine
efficacy,a
% (95% CI)
Pb
9 (3–15)
.004
16 (9–23)
.00003
⫺1 (⫺11 to 10) .85
Incidence in
placebo
Power,e
c
d
group
VAR
%
858
544
970
679
11 (3–19)
20 (10–28)
155
!5
72
431
251
419
303
⫺3 (⫺15 to 11)
17 (2–30)
416
⫺86
48
155
117
⫺25 (⫺41 to ⫺4)
5 (⫺25 to 28) .72
12 (4–20)
.003
227
2259
11
279
5
83
56
511
64
618
13 (⫺25 to 39)
17 (7–26)
697
14 (5–23)
.02
1509
215
80
310
383
19 (6–30)
39
23
28
160
44
61
36
⫺1
(5–67)
.03
(16–82)
.01
(⫺16 to 64) .14
(⫺20 to 24) .96
85
50
61
346
37
30
22
⫺2
58
60
26
16
8
8
29
16
9
45 (⫺2 to 70)
50 (⫺17 to 79)
11 (⫺131 to 66)
!5
67
68
1 (⫺38 to 30)
⫺17 (⫺32 to 0)
336
410
⫺13
.01
927
.05
Vaccine
efficacy,a
% (95% CI)
81
99
17 (4–28)
3565
2517
1392
No. of
No. of
vaccine
placebo
recipients
recipients
(n p 18,245) (n p 18,268)
NOTE. For definitions of outcome measures, see Methods. IPPV, intermittent positive-pressure mechanical ventilation; LRTI, lower respiratory tract infection; VAR, vaccine-attributable reduction;
WHO, World Health Organization.
a
b
c
d
e
f
For a definition of vaccine efficacy, see Methods.
P is for no. of vaccine recipients versus no. of placebo recipients.
Incidence rate in the placebo group, expressed as no. of cases per 100,000 child-years of observation.
Expressed as no. of cases per 100,000 child-years of observation.
Power to detect a difference with 95% confidence.
Confirmed by chest radiograph findings.
Table 2.
analysis.
Estimated efficacy of a 9-valent pneumococcal conjugate vaccine in preventing pneumococcal pneumonia in hospitalized HIV-uninfected children, by
Per-protocol analysis
Intent-to-treat analysis
Outcome measure
Clinical LRTI
Clinical pneumonia
Bronchiolitis
WHO-ACf
1515
WHO-defined mild pneumonia
WHO-defined mild pneumonia
without wheezing
WHO-defined severe pneumonia
WHO-defined severe pneumonia
without wheezing
Bacteremic pneumocooccal
pneumonia
Overall
Associated with a vaccine serotype
IPPV for LRTI
Death due to LRTI
No. of
No. of
vaccine
placebo
recipients
recipients
(n p 18,633) (n p 18,626)
1033
566
558
1106
681
539
169
177
212
135
Vaccine
efficacy,a
% (95% CI)
Pb
7 (⫺1 to 14)
17 ( 7–26)
⫺3 (⫺14 to 9)
.10
.001
.56
Incidence in
placebo
Power,e
c
d
group
VAR
%
20 ( 3–35)
.03
⫺24 (⫺39 to ⫺5) .02
3 (⫺33 to 39)
No. of
No. of
vaccine
placebo
recipients
recipients
(n p 17,356) (n p 17,350)
Vaccine
efficacy,a
% (95% CI)
2,566
1,573
1,246
172
267
⫺43
37
91
8
650
348
377
717
452
358
9 (⫺1 to 18)
23 (11–33)
⫺5 (⫺18 to 9)
491
313
100
⫺97
58
66
119
124
158
89
25 (4 to 40)
⫺28 (⫺45 to ⫺6)
57
59
.85
137
5
591
662
11 (1–20)
.04
1530
164
!5
52
32
367
41
440
22 (⫺24 to 51)
17 (4–27)
312
359
13 (⫺1 to 25)
.07
832
112
46
181
229
21 (4–35)
!5
17
26
6
3
1
7
8
5
4
8
9
5
8
38 (⫺91 to 80)
.42
19
7
2
15
18
6
25
22
67 (⫺65 to 93)
40 (⫺14 to 68)
18 (⫺52 to 56)
.18
.11
.52
14
58
51
9
23
9
40
75
13
11
(⫺151
(⫺124
(⫺141
(⫺130
to
to
to
to
86)
97)
68)
66)
NOTE. For definitions of outcome measures, see Methods. IPPV, intermittent positive-pressure mechanical ventilation; LRTI, lower respiratory tract infection; VAR, vaccine-attributable reduction;
WHO, World Health Organization.
a
b
c
d
e
f
For a definition of vaccine efficacy, see Methods.
P is for no. of vaccine recipients versus no. of placebo recipients.
Incidence rate in the placebo group, expressed as no. of cases per 100,000 child-years of observation.
Expressed as no. of cases per 100,000 child-years of observation.
Power to detect a difference with 95% confidence.
Confirmed by chest radiograph findings.
per 100,000 child-years), and this was used as the benchmark
in evaluating the sensitivity of the other outcomes for determining the burden of pneumonia prevented by vaccination.
Although the VE estimate was greatest for vaccine serotype–
specific bacteremic pneumococcal pneumonia (VE, 59%;
P p .04), the sensitivity of this outcome (VAR, 344 cases per
100,000 child-years) in measuring the burden of pneumonia
prevented through vaccination was 13.4% (95% CI, 12.1%–
14.7%). Relative to the outcome for C-LRTI, the sensitivities
of clinical pneumonia (VAR, 2302 cases per 100,000 childyears), WHO-defined severe pneumonia (VAR, 2052 cases per
100,000 child-years), WHO-AC (VAR, 909 cases per 100,000
child-years), and all bacteremic pneumococcal pneumonia
(VAR, 483 cases per 100,000 child-years) in detecting the burden of pneumococcal pneumonia prevented by vaccination
were 89.5% (95% CI, 88.2%–90.6%), 79.8% (95% CI, 78.1%–
81.3%), 35.3% (95% CI, 33.5%–37.2%), and 18.8% (95% CI,
17.3%–20.3%), respectively (table 3). An estimated 10.9% of
the HIV-infected children (281 of 2577) died from C-LRTI,
with no difference in rates observed between vaccinees and
placebo recipients.
DISCUSSION
This report provides insight into the clinical presentation of
pneumococcal pneumonia in HIV-infected and -uninfected
children. We show that an outcome measure with high specificity, such as bacteremic pneumococcal pneumonia, yields the
highest point efficacy estimate for the PnCV but has a very
limited role when evaluating the potential public health benefit
of the PnCV.
The outcome measures that most accurately evaluated the
impact of the PnCV on the public health burden of pneumonia
were C-LRTI (in HIV-infected children) and clinical pneumonia (in HIV-uninfected children). Although the primary objective of the study was to measure the efficacy of the vaccine
against chest radiograph–confirmed alveolar consolidation [9],
the current hypothesis-generating study suggests that only
37.8% of patients with pneumococcal pneumonia presented
with chest radiograph–confirmed alveolar consolidation. Although a radiological lag in the chest radiograph changes in
relation to the clinical presentation may explain, in part, the
low frequency of chest radiograph–confirmed alveolar consolidation, it would appear that most cases of pneumococcal pneumonia do not manifest themselves as alveolar consolidation on
chest radiographs. This observation, in addition to the high
frequency of concurrent pneumococcal infection in children
with viral pneumonia (31%) [10], may also explain why chest
radiograph findings are unhelpful in discriminating between
bacterial and viral causes of pneumonia [6–8].
Clinical pneumonia end points themselves are subject to
variation between individuals and within individuals. This mis1516 • CID 2005:40 (15 May) • HIV/AIDS
classification bias has likely decreased the observed differences
between vaccinees and placebo recipients and, therefore, the
vaccine impact is likely to be an underestimation. Although all
of the study physicians had previous pediatric experience, clinical review by an expert panel may be useful to standardize a
clinical end point for a vaccine trial.
A paradoxical finding of our study was the excess in cases
of WHO-defined mild pneumonia observed in vaccinees, particularly in HIV-uninfected children. The reason for hospitalization of these children, whom the WHO recommends be
treated on an outpatient basis, was because the hospital provided a 24-h clinical service, and children with mild pneumonia
sometimes slept at the hospital for logistical reasons. The incidence of WHO-defined mild pneumonia in this study is,
however, an underestimation of the true burden of disease,
because the majority of children with mild pneumonia would
have been treated outside of the hospital, where surveillance
for study outcomes was not undertaken.
The increase in WHO-defined mild pneumonia was evident
only when all children were included in analysis, irrespective
of the presence of wheezing on clinical examination. The excess
of pneumonia was abolished when children with wheeze were
excluded from the category of mild pneumonia. There are a
number of possible, speculative reasons for why vaccinees were
at increased risk of developing WHO-defined mild pneumonia
with wheeze. Immunity induced through vaccination may have
attenuated the clinical course of illness, with vaccinees less likely
than placebo recipients to progress to severe disease. Although
vaccination was successful in reducing concurrent pneumococcal superinfection in patients with viral-associated pneumonia [10], the clinical course of the viral-associated LRTI may
have been altered in vaccinees. This may result in vaccinees
with viral infections experiencing progression to WHO-defined
mild pneumonia with wheezing rather than to WHO-defined
severe pneumonia, which would have resulted had the vaccine
not protected against superimposed pneumococcal infection.
Importantly, the excess of cases of WHO-defined mild pneumonia was not significant when children with wheezing were
excluded from the analyses, and furthermore, the vaccine had
no measurable effect on bronchiolitis. These data support the
low likelihood of a role for the pneumococcus in children who
present with these clinical signs and symptoms (i.e., mild pneumonia with wheeze and bronchiolitis), although our study
lacked the power to detect a very small impact of vaccination
on these conditions. Our data suggest that the criteria for administration of antibiotics may be refined to exclude children
who have wheezing as their only clinical sign, although a role
for other bacteria in this presentation cannot be excluded by
our study. The use of auscultation would, however, require the
additional training of nonphysician health workers in the evaluation of chest auscultation as part of clinical assessment.
Table 3.
analysis.
Estimated efficacy of a 9-valent pneumococcal conjugate vaccine in preventing pneumococcal pneumonia in hospitalized HIV-infected children, by
Per-protocol analysis
Intent-to-treat analysis
Outcome measure
No. of
No. of
vaccine
placebo
recipients recipients
(n p 1289) (n p 1288)
487
446
85
209
45
Vaccine
efficacy,a
% (95% CI)
Pb
15 (6–24)
15 (5–24)
14 (⫺16 to 37)
13 (⫺7 to 28)
⫺8 (⫺38 to 37)
.002
.004
.32
.19
.76
Clinical LRTI
Clinical pneumonia
Bronchiolitis
f
WHO-AC
WHO-defined mild pneumonia
WHO-defined mild pneumonia
without wheezing
412
379
73
182
49
39
37
WHO-defined severe pneumonia
WHO-defined severe pneumonia
without wheezing
Bacteremic pneumocooccal pneumonia
All
Associated with a vaccine serotype
IPPV for LRTI
300
361
17 (5–27)
.006
271
327
17 (5–28)
.009
17
7
3
31
17
3
Death due to LRTI
143
138
No. of
No. of
Incidence in
vaccine
placebo
placebo
Power,e recipients recipients
c
d
group
VAR
%
(n p 1201) (n p 1200)
2573
2302
409
909
⫺134
87
80
15
30
5
186
181
46
128
30
228
210
52
140
23
18 (3–31)
14 (⫺4 to 28)
11 (⫺31 to 40)
8 (⫺15 to 27)
⫺23 (⫺55 to 31)
1237
⫺69
12,082
2052
!5
78
23
131
20
169
⫺13 (⫺52 to 58)
23 (4 –38)
10,944
1884
73
120
150
20 (0–36)
45 (1–70)
.04
59 (1–83)
.04
0 (⫺394 to 80) .68
1065
584
100
483
344
0
48
46
13
7
24
12
46 (⫺6 to 72)
42 (⫺48 to 77)
!5
1
1
0 (⫺1496 to 94)
⫺3 (⫺23 to 20) .76
4616
⫺164
!5
59
59
0 (⫺42 to 30)
⫺5 (⫺39 to 48) .82
16,724
15,319
2918
6996
1504
Vaccine
efficacy,a
% (95% CI)
NOTE. For definitions of outcome measures, see Methods. IPPV, intermittent positive-pressure mechanical ventilation; LRTI, lower respiratory tract infection; VAR, vaccine-attributable reduction;
WHO, World Health Organization.
a
b
c
d
e
f
For a definition of vaccine efficacy, see Methods.
P is for no. of vaccine recipients versus no. of placebo recipients.
Incidence rate in the placebo group, expressed as no. of cases per 100,000 child-years of observation.
Expressed as no. of cases per 100,000 child-years of observation.
Power to detect a difference with 95% confidence.
Confirmed by chest radiograph findings.
Although we only observed trends toward fewer episodes
requiring mechanical ventilation support and fewer deaths in
HIV-uninfected children, these results need to be interpreted
in the context of the lack of power of our study to evaluate
such outcomes. Furthermore, unlike in most other developing
countries and many rural areas of South Africa, health care
services, including oxygen and appropriate antibiotic therapy,
were freely available and accessible to all children in the study
area. This is evident in the low overall case-fatality rate associated with LRTI in HIV-uninfected children (40 [1.9%] of
2139 episodes).
The vaccine was not efficacious in reducing mortality in HIVinfected children, in whom the overall case-fatality rate due to
LRTI (281 [31.3%] of 899 children) was 16.7-fold greater (95%
CI, 14.5–19.3-fold greater) than that observed in HIV uninfected children (1.9%). This may be due to the complexity of
pneumonia in these children [13]. HIV-infected children may
have been disadvantaged because of the hospital’s policy of not
providing mechanical ventilation to HIV-infected children.
However, this is unlikely to have been the dominant reason for
the high mortality rate among these children, because a fatality
rate of 190% was previously documented for HIV-infected children receiving mechanical ventilation for pneumonia in the
absence of treatment with antiretroviral drugs at the same hospital [14].
Although the determination of HIV status was made directly
for patients with the outcomes reported in this study, the denominator of HIV-infected children at risk was based on indirect estimation. Direct determination of the HIV status of all
children at study entry would have increased the precision of
our vaccine impact analyses but would not have affected our
estimates of the relative sensitivity of the clinical end points
measured. Furthermore, the estimated burden of disease is so
high that variance from these estimates is unlikely to greatly
alter our assessment of vaccine impact.
Thus, although the PnCV did not reduce mortality among
HIV-infected children, its relevance in improving the quality
of life for these children was made evident by the greater
burden (OR, 9.9; 95% CI, 8.7–11.2) of hospitalization for
pneumococcal pneumonia prevented in HIV-infected children (VAR, 2573 cases per 100,000 child-years), compared
with the burden prevented in HIV-uninfected vaccine recipients (VAR, 267 cases per 100,000 child-years). This is particularly important given that, among placebo recipients, the
overall incidence of C-LRTI hospitalization for HIV-infected
children was 6.4-fold greater (95% CI, 5.8–7.0-fold greater)
than for HIV-uninfected children.
1518 • CID 2005:40 (15 May) • HIV/AIDS
Acknowledgments
We acknowledge the essential contribution of members of the Vaccine
Trialist Group [10] to the original study.
Financial support. Wyeth-Lederle Vaccines and Pediatrics and the
World Health Organization.
Potential conflicts of interest. S.A.M. was the recipient of research
grant awards and salary support from Wyeth Vaccines and Pediatrics during
the conduct of the study. C.C. received salary support from Wyeth Vaccines
and Pediatrics during the conduct of the study. K.P.K. has received research
grant awards from Wyeth Vaccines and Pediatrics, is a member of the
Wyeth speakers’ bureau, and is a consultant for Wyeth Vaccines and Pediatrics. L.K.: no conflicts.
References
1. World Health Organization (WHO). Acute respiratory tract infections
in children: case management in small hospitals in developing countries. WHO/ARI/90.5. Geneva: WHO, 1994.
2. Mulholland EK, Simoes EA, Costales MO, McGrath EJ, Manalac EM,
Gove S. Standardized diagnosis of pneumonia in developing countries.
Pediatr Infect Dis J 1992; 11:77–81.
3. Pneumonia in childhood. Lancet 1988; 1:741–3.
4. Vuori-Holopainen E, Peltola H. Reappraisal of lung tap: review of an
old method for better etiologic diagnosis of childhood pneumonia.
Clin Infect Dis 2001; 32:715–26.
5. Standardization of interpretation of chest radiographs for the diagnosis
of pneumonia in children. WHO/V&B/01.35. World Health Organization (WHO) Pneumonia Vaccine Trial Investigators Group. Geneva:
WHO Department of Vaccines and Biologicals, 2001. Available at:
http://www.who.int/vaccine_research/documents/en/. Accessed 16 December 2004.
6. Korppi M, Kiekara O, Heiskanen-Kosma T, Soimakallio S. Comparison
of radiological findings and microbial aetiology of childhood pneumonia. Acta Paediatr 1993; 82:360–3.
7. Virkki R, Juven T, Rikalainen H, Svedstrom E, Mertsola J, Ruuskanen
O. Differentiation of bacterial and viral pneumonia in children. Thorax
2002; 57:438–41.
8. Friis B, Eiken M, Hornsleth A, Jensen A. Chest x-ray appearances in
pneumonia and bronchiolitis: correlation to virological diagnosis and
secretory bacterial findings. Acta Paediatr Scand 1990; 79:219–25.
9. Klugman KP, Madhi SA, Huebner RE, Kohberger R, Mbelle N, Pierce
N. A trial of a 9-valent pneumococcal conjugate vaccine in children
with and those without HIV infection. N Engl J Med 2003; 349:1341–8.
10. Madhi SA, Klugman KP. A role for Streptococcus pneumoniae in virusassociated pneumonia. The Vaccine Trialist Group. Nat Med 2004; 10:
811–3.
11. Health Systems Research and Epidemiology. Summary report: national
HIV seroprevalence survey of women attending antenatal clinics in
South Africa. Pretoria, South Africa: V & R Printing Works, 2001.
12. Gray GE, McIntyre JA, Lyons SF. Effect of breastfeeding on vertical
transmission of HIV-1 in Soweto, South Africa [abstract Th.C.413].
In: Program and abstracts of the 11th International Conference on
AIDS, Vancouver. 1997.
13. Madhi SA, Cutland C, Ismail K, O’Reilly C, Mancha A, Klugman KP.
Ineffectiveness of trimethoprim-sulfamethoxazole prophylaxis and the
importance of bacterial and viral coinfections in African children with
Pneumocystis carinii pneumonia. Clin Infect Dis 2002; 35:1120–6.
14. Mathivha R, Luyt D, Hon H, Dance M, Lipmanovitch M. Outcome
of mechanically ventilation in children infected with human immunodeficiency virus. S Afr Med J 1998; 88:1447–51.