Confidential Draft

Confidential second draft –September 2007

Report for Immunization, Vaccines and Biologicals, World Health Organization

Effectiveness of pneumococcal polysaccharide vaccine: systematic review and meta-analysis of randomised controlled trials

Pippa Scott, Matthias Egger, Anke Huss

Institute of Social and Preventive Medicine (ISPM)

University of Bern, Switzerland

www.ispm.ch

Contract ID: OD/AP-07-02439

Corresponding author:

Professor Matthias Egger MD MSc FFPH DTM&H

Institute of Social and Preventive Medicine (ISPM)

University of Bern

Finkenhubelweg 11

CH-3012 Bern, Switzerland

Tel: +41 31 631 35 01

Fax: +41 31 631 35 20

egger@ispm.unibe.ch

Abstract 323 words, main text 4845 words, 3 tables, 39 figures, 1 appendix

1

Confidential Draft

Abstract

Background: Infection with Streptococcus pneumoniae causes substantial morbidity and mortality

worldwide. Polysaccharide pneumococcal vaccines were first developed over 50 years ago, but results

of meta-analyses have given conflicting results with respect to the efficacy of the vaccine.

Objectives: We aimed to conduct a systematic review and meta-analysis of randomised controlled

trials (RCTs) examining the effect of pneumococcal polysaccharide vaccination on clinical outcomes,

taking into account trial quality.

Search strategy: We searched Medline, Embase, the Cochrane CENTRAL database, LILACS (Latin

American and Caribbean Health Sciences Literature), IndMed (Indian Medlars Centre) and AIM

(African Index Medicus) using relevant search terms. Searches were conducted in May 2007

Selection criteria: Studies were scrutinised by two reviewers to identify RCTs which compared

polysaccharide pneumococcal vaccine recipients to a control group (receiving placebo, another

vaccine or no intervention) and which reported clinically relevant outcomes. No language restrictions

were applied.

Data extraction and analysis: Data on interventions, outcomes and quality (concealment of

allocation, blinding) were extracted in duplicate. Results were combined in random-effects meta-

analyses. Stratified analyses and meta-regression was used to examine sources of between-trial

heterogeneity, including trial quality, valency of the vaccine and other variables.

Results: We identified 31 eligible trials, which included a total of 110,676 individuals. Nineteen trials

were performed in developed countries, nine in developing countries and three in Russia. Study

populations included patients with respiratory illness, HIV-infected patients, children, miners or

soldiers, the elderly and patients with chronic illnesses. The 23-valent vaccine was used in eight trials

and the 14-valent vaccine in 13. The combined relative risk (RR) of all-cause pneumonia was 0.74

(95% CI 0.57- 0.94). There was substantial between-trial heterogeneity (test of heterogeneity

p<0.001). In meta-regression analysis, trial quality had an important effect on efficacy estimates, with

trials of higher quality showing less benefit. Amongst seven controlled double-blind trials, there was

little evidence of a protective effect of the vaccine against all-cause pneumonia: RR 1.19 (95% CI

0.97-1.47, heterogeneity p=0.126). Combined RRs were similar for trials of 23-valent vaccine, 14-

valent vaccine and trials of vaccines with lower valencies. There was some evidence, based on seven

double-blind trials, of a protective effect on mortality due to pneumonia, with a combined RR of 0.64

(95% CI 0.39-1.04), but little evidence for an effect on all-cause mortality (RR 0.94; 95% CI 0.80 -

1.10).

Conclusions: The evidence from randomised controlled trials on the efficacy of polysaccharide

pneumococcal vaccines indicates that the vaccine may be less efficacious than previously assumed.

2

Confidential Draft

Background

Infection with Streptococcus pneumoniae causes substantial morbidity and mortality

worldwide. The burden of disease due to S. pneumoniae falls mainly on children and elderly and

extends to persons with underlying conditions such as HIV infection or chronic respiratory diseases.

Patients who have undergone splenectomy or transplantation are also considered to be at increased

risk of pneumococcal disease.1 Concerningly, antibiotic resistance among pneumococcal strains is

increasing.2,3 Polysaccharide pneumococcal vaccines were first developed over 50 years ago and

have progressed from 2-valent vaccines to today’s 23-valent polysaccharide vaccine registered in the

US in 1983.4 There are more than 80 known serotypes of S. pneumoniae and the majority of those

causing disease are included the 23-valent vaccine. It has been shown that between 72% 5 and over

95% 6 of serotypes causing invasive pneumococcal disease are covered by the vaccine, with some

variation by geographical area.

Results of meta-analyses 7-22 of randomised clinical trials (RCTs) have produced conflicting

results. They differ with respect to study populations, for example children, or patients with chronic

obstructive pulmonary disease (COPD) as well as outcomes considered, for example definitive or

presumptive pneumococcal pneumonia, pneumonia from all causes or mortality. Although many of the

meta-analyses examined quality items of the trials, these were rarely considered in their statistical

analyses. The conflicting results are mirrored in variation in vaccination recommendations between

countries. The US Centres for Disease Control and Prevention (CDC), for example, recommend

vaccination for numerous high-risk groups (for example persons over 65 years) and high-risk settings

(for example nursing homes).23 In European countries recommendations generally cover fewer risk

groups. Even among the countries with wide-reaching recommendations there is some disagreement

on the groups included in the recommendations.

We conducted a systematic review and meta-analysis of RCTs examining the effect of

pneumococcal polysaccharide vaccination on clinical outcomes, taking into account the quality of the

RCTs.

3

Confidential Draft

Methods Literature search

Our literature search covered the Medline and Embase databases as well as the Cochrane

CENTRAL database of randomised controlled trials. Key words were “pneumococcal vaccine” or

“pneumococcal immunisation” in combination with “polysaccharide”. In Medline and Embase searches,

study design terms were also included (“randomized controlled trial.pt.”, “clinical trial.pt.”, “controlled

clinical trial.pt.”). We conducted an additional search in the LILACS (South American), IndMed (Indian

Medlars Centre) and AIM (African Index Medicus) databases with the search words “pneumococc*”,

“vac*” as well as “polysac*”, In addition, we screened reference lists for relevant studies, reviews and

meta-analyses. Searches were conducted in May 2007. We also searched the Cochrane library for

meta-analyses and systematic reviews, using the search terms “pneumococc*” and “polysac*”.

Eligibility criteria for studies We included all randomised controlled trials where polysaccharide pneumococcal vaccine was

compared with placebo, other vaccines or no intervention and which reported on clinical outcomes or

death. No language restrictions were applied. Studies examining antibody responses only were

excluded. We did not consider uncontrolled studies, observational intervention studies or animal and

laboratory studies. We also excluded studies where polysaccharide pneumococcal vaccine was used

as a booster after the conjugate pneumococcal vaccine.

Selection of studies and data extraction Eligibility of studies was determined independently by two reviewers (AH and PS). We

extracted data on the study population, intervention types (valency of vaccine and the control

intervention/s), sample size, sample size calculations, loss to follow up, outcomes and source of

funding. We also extracted methodological quality data, including the reporting of randomisation

processes, the method of allocation sequence generation and allocation (concealed or not), whether a

double-blind design was reported and who was blinded. All data were extracted in duplicate, and

discrepancies resolved by consensus, with the senior epidemiologist (ME) acting as the arbiter.

Authors were not contacted for further clarification.

Outcomes We analysed data for the following outcomes:

• Definitive pneumococcal pneumonia (typical clinical/radiological findings and S. pneumoniae

isolated from normally sterile body fluid such as blood),

• Presumptive pneumococcal pneumonia (typical clinical/radiological findings and either S

pneumoniae. isolated from respiratory tract samples, or seroconversion against S.

pneumoniae),

• Invasive pneumococcal disease (S. pneumoniae isolated from a usually sterile body fluid such

as blood),

4

Confidential Draft

• Pneumonia (all causes),

• Bronchitis (all cause),

• Mortality (all cause; due to pneumonia; due to pneumococcal infection),

• Meningitis,

• Bacteraemia.

Data analysis

We used a version of the “metan” command for Stata statistical software (version 9, Stata

Corporation, College Station, TX) adapted for vaccine trials to conduct DerSimonian and Laird

random-effects meta-analyses.25 Results were expressed as relative risks (RRs) with 95% confidence

intervals. In forest plots both RRs and vaccine efficacy estimates are shown. We quantified between-

trial heterogeneity using the I2 statistic, which can be interpreted as the proportion of the total variation

in estimated risk ratios that is due to between-trial heterogeneity,26 and did the standard test of

heterogeneity. Analyses are based on episodes when available, otherwise the number of cases

developing the event were used. We compared results from episode- and case-based analyses for

trials reporting both. The saline placebo group was used in analyses of trials with more than one

control group. In some instances investigators reported data from combined control groups only and

these were then included in analyses. Because some of the publications analyzed data that were

already up to ten years old, we extracted the start year of the trial. If this information was not given by

the authors, we approximated the starting year by subtracting the maximum follow-up time in years of

the publication year.

When the meta-analysis included 10 or more trials, we stratified analyses by trial quality, the

valency of the vaccine, the study setting and study population. For all-cause pneumonia, the extent to

which one or several of these study level variable explained heterogeneity was further explored using

random-effects meta-regression models (using the metareg command in Stata), adjusted and not

adjusted for trial quality. In these models we examined to what extent between-study variance, tau2,

was reduced in models including different sets of variables. We did not use meta-regression for other

outcomes because the number of trials was considered too small. In stratified and meta-regression

analyses, blinding was analysed in three groups: described as double-blind with placebo or other

vaccine used in controls (described as double-blind, controlled), not described as double-blind but

placebo or other vaccine used in controls (not described as double blind, controlled) and no

intervention in controls (unblinded). Concealment of allocation was grouped as adequate or unclear,

the valency of the vaccine as 23-valent versus 14-valent versus other, and the study population as

COPD or other lung disease, HIV-infected patients, children, miners or soldiers, the chronically ill or

elderly, and other. The trial setting was defined following common United Nations practice as

developing country, industrialized country and other (the latter including the countries of Eastern

Europe). Differences in the results between small and large trials were assessed by visual inspection

of funnel plots and by means of a statistical test for small study effects (funnel plot asymmetry), for

outcomes with 10 or more trials.

5

Confidential Draft

Results

Our literature search identified 516 references, 91 were potentially eligible. Thirty-one trials

reported in 27 articles met our inclusion criteria and were included in our meta-analysis (Figure 1).

Trial characteristics An overview of study characteristics is given in Table 1 and Appendix 1. Nineteen trials (61%)

were performed in developed countries, nine (29%) in developing countries and three (10%) in Russia.

There were five studies (16%) in patients with respiratory illness (4 COPD and 1 bronchogenic

carcinoma), one study in HIV-infected patients and nine (29%) in children. Seven studies (23%) were

in miners or soldiers, eight (26%) in the elderly or patients with chronic illnesses and one study in

people over the age of 10 years in Papua New Guinea. The 14-valent vaccine was used in 13 (42%)

trials and the 23-valent vaccine in eight (26%) trials. Eight (26%) studies reported use of another

valency of the vaccine (2-valent to 17-valent) and in two studies the vaccine was changed from 14-

valent to 23-valent vaccine mid-trial. Valencies of vaccines increased over time (Figure 2). Seven

trials (23%) reported intra-muscular administration of the vaccine and 11 (37%) subcutaneous

administration. Thirteen (42%) did not report the administration route.

Control interventions

Eight trials used no intervention in the control group. Nineteen used saline placebo. Six trials

used a vaccine in a control group. Honkanen 1999 and Koivula 1997 used influenza vaccines in both

arms. Austrian 1976 and Smit 1977 (a and b) used meningococcal vaccines in a second control group

(with saline placebo in the first). Austrian 1976 reported results for control groups separately

throughout. Only combined results for control groups were available for all-cause pneumonia and

bronchitis in Smit. Mäkelä 1981 and John 1984 used Haemophilus influenzae B vaccination in

controls.

Trial quality

Eleven of the 31 (35%) trials described the way allocation sequence was generated and in six

we judged this to be adequate. Fourteen (45%) of the trials described the way they concealed

allocation and in nine (29%) this was done in an adequate way. Thirteen trials were described as

double-blind and used placebo or other vaccine used in controls, ten were not described as double-

blind but used placebo or another vaccine in controls and 8 trials were unblinded. Eight of the 31

(26%) trials reported a sample size calculation, with one study reaching the calculated sample size.

Seven trials (23%) reported mean follow-up times. These ranged from 0.5 to 2.9 years with a mean,

over trials, of 2.1 years. Nineteen trials (61%) reported a maximum follow-up time of between 0.5 and

5 years with a mean of 2.5 years. Seven studies (23%) reported no time frame for the follow-up of trial

participants. Ten trials reported values for loss to follow-up, but only 5 (16%) of all 31 trials report loss

6

Confidential Draft

to follow-up by intervention group. Eighteen (58%) trials reported deaths and all of these trials reported

deaths by intervention group.

One trial reported the loss of records for some trial participants and this may affect results for this trial

(Riley 1977, people over 10yo in Papua New Guinea)

Diagnostic criteria for pneumonia

Of the 22 trials reporting on either all-cause, presumptive or definitive pneumococcal

pneumonia, the majority (14) reported only radiographically confirmed cases. Five did not report the

manner of pneumonia case assessment and 2 used radiographic assessment for an unspecified

percentage of cases and clinical assessment only in the remainder. One trial (MacLeod) diagnosed

pneumonia by clinical assessment only. Of the 11 trials reporting presumptive pneumococcal

pneumonia, 7 based this on culture of S. pneumoniae in addition to clinical/radiological diagnosis. Two

of the remaining studies based it on serological findings and 1 on either culture or serology (Appendix

1).

Meta-analyses

Depending on the outcome, 2 to 20 trials and 794 to 83,517 trial participants could be included

in each analysis (Table 2).

Definitive pneumococcal pneumonia

Only two trials assessed outcomes which met the criteria for definitive pneumococcal

pneumonia (i.e. typical clinical/radiographic signs and the culture of S. pneumoniae from a normally

sterile body fluid). One of these used the 23-valent vaccine and the other the 14 –valent vaccine; both

studies were reported to be double-blind. The combined RR was 0.62 (95% CI 0.05-8.61); the I2

statistic was 48.6%, with a P value for heterogeneity of 0.163 (Figure 3).

Presumptive pneumococcal pneumonia

Eleven studies assessed presumptive pneumococcal pneumonia (typical clinical/radiological

findings and either S. pneumoniae isolated from sputum/other respiratory tract samples, or sero-

conversion against S. pneumoniae). The combined RR was 0.64 (95% CI 0.43-0.96), however there

was substantial heterogeneity with an I2 value of 74.4% (heterogeneity p<0.001, Figure 4).

Three trials were described as double-blind. In these trials the combined RR was 1.20 (95% CI

0.75-1.92), with little evidence of between-trial heterogeneity (I2 =0%, heterogeneity p=0.59). Studies

that were not described as double-blind but were placebo-controlled, were heterogeneous (I2=80%,

p<0.001) and produced a combined RR of 0.54 (95% CI 0.33-0.86). One unblinded and non-placebo-

controlled trial had an RR of 0.09 (0.01-1.64) (Figure 5). Similarly, meta-analysis of the three trials with

adequate allocation concealment resulted in a combined RR of 1.06 (95% CI 0.67-1.67, I2 =0%,

heterogeneity p=0.45). Conversely, for trials with unclear allocation concealment the combined RR

was 0.55 (95% CI 0.33-0.90) with marked heterogeneity (I2=79%, p<0.001) (Figure 6). There was little

7

Confidential Draft

evidence for a protective effect in three trials of the 23-valent vaccine: the combined RR was 1.12

(95% CI 0.66-1.90, I2 =36%, p=0.21) and similar results were obtained for 14-valent vaccines (three

trials): combined RR 0.95 (95% CI 0.63-1.43), I2 =0%, p=0.46). The remaining studies were

heterogeneous (I2 =70%, p=0.010), the combined RR was 0.39 (95% CI 0.23-0.66) (Figure 7). A forest

plot ordered by the incidence of presumptive pneumococcal pneumonia in controls showed no clear

pattern (Figure 8).

The eight trials performed in developed countries gave a combined RR of 0.92 (95% CI 0.69-

1.22, I2 =31%, p=0.18) compared to an RR of 0.28 (95% CI 0.14-0.56) from three trials from

developing countries, I2 =59%, p=0.089 (Figure 9). Trials of patients with respiratory illness showed

some heterogeneity (I2 =47%, p=0.153) with a combined RR of 0.52 (95% CI 0.08-3.58). Studies in

miners or soldiers were heterogeneous (I2 =78%, p=0.004); the combined RR was 0.38 (95% CI 0.21-

0.68). Conversely, trials in the elderly or people with chronic illness were homogeneous (I2 =0%,

p=0.74) and the combined RR was 1.08 (95% CI 0.88-1.41) (Figure 10).

All cause pneumonia

Twenty trials reported on all cause pneumonia. In eight (40%) of the trials the 23-valent

vaccine was used. As shown in Figure 11, trial results were very heterogeneous (I2 =89.1%, p<0.001),

with a combined RR of 0.74 (95% CI 0.57 – 0.94). Results from stratified analyses were similar to

presumptive pneumococcal pneumonia: the double-blind and placebo-controlled trials showed only

moderate heterogeneity (I2 =40%, p=0.126), with little evidence of a protective effect of the vaccine

(RR 1.19; 95% CI 0.97-1.47). Trials that were not described as double-blind but used placebo or

another vaccine in controls and unblinded trials were highly heterogeneous (I2 =88%, p<0.001 and I2

=83%, p<0.001) with combined RRs of 0.81 (95% CI 0.58-1.12) and 0.43 (95% CI 0.27-0.67),

respectively (Figure 12). The same picture emerged for concealment of allocation: the six trials with

adequate concealment of allocation showed little heterogeneity (I2 =9%, p=0.36) with an RR close to

one (RR 1.02; 95% CI 0.88-1.18) whereas the 14 trials with unclear concealment of allocation were

very heterogeneous (I2 =92%, p<0.001) with a combined RR of 0.64 (95% CI 0.46-0.90) (Figure 13).

There were eight trials of 23-valent vaccine, six trials of 14-valent vaccine and six trials of

vaccines with valencies ranging from 3-valent to 13-valent. Between-trial heterogeneity was

substantial in all three groups (I2 between 76% and 92%). Combined RRs were similar: 0.73 (95% CI

0.44-1.24) for 23-valent vaccines, 0.80 (95% CI 0.50-1.29) for 14-valent vaccines and 0.69 (95% CI

0.47-1.02) for vaccines of other valencies (Figure 14). A forest plot ordered by the incidence of

presumptive pneumococcal pneumonia in controls showed no clear pattern (Figure 15). There was

much heterogeneity both among the twelve trials in developed countries and the five trials from

industrialized countries (I2 =83, p<0.001 and I2 =86%, p<0.001), with combined RRs of 0.89 (95% CI

0.70-1.14) and 0.76 (95% CI 0.47-1.25). The combined RR from the three Russian trials was 0.29

(95% CI 0.18 – 0.47) (Figure 16).

8

Confidential Draft

Six population groups were represented in this analysis but for three groups there was only

one trial each (HIV-infected patients, children ,and people older than 10yo in Papua New Guinea).

Trials in patients with respiratory illness showed little heterogeneity (I2 =0%, p=0.67), with a combined

RR of 0.88 (95% CI 0.63-1.23). Studies in miners or soldiers were heterogeneous (I2 =74%, p=0.002);

the combined RR was 0.42 (95% CI 0.29-0.61). The studies in the elderly or patients with chronic

illnesses were also heterogeneous (I2 =89%, p<0.001), with a combined RR of 0.90 (95% CI 0.63-

1.20) (Figure 17). Finally, six trials reported pneumonia cases as well as episodes. A comparison of

these results produced combined RRs of 1.22 (95% CI 1.03-1.46) for episodes and 1.16 (95% CI 0.97-

1.40) for cases (Figure 18).

Bronchitis

Four studies reported on bronchitis from any cause (1 used the 14-valent vaccine, 2 a 12-

valent and 1 a 6-valent). There was only weak evidence of a protective effect of the vaccine (RR

0.92;95% CI 0.76-1.12). Between-trial heterogeneity was moderate (I2 =54%, p=0.090). Only one trial

was reported to be double-blind (Simberkoff 1986), and both this trial and Austrian b 1980 were

considered to have adequate allocation concealment (Figure 19).

All cause mortality

Fourteen studies reported all-cause mortality. The proportion of study participants dying in

control groups ranged from 0% to 28%. The meta-analysis showed little evidence of a protective effect

of the vaccine with a combined RR of 0.94 (95% CI 0.84-1.05) and some between-trial heterogeneity

(I2 =46%, p=0.031) (Figure 20). In contrast to presumptive pneumococcal pneumonia and all-cause

pneumonia, blinding status had little influence on results: The combined RR was 0.94 (95% CI 0.80 -

1.10) for nine double-blind studies using placebo or another vaccine in controls, 0.98 (95% CI 0.82-

1.18) for two trials that were not described as double blind but used a placebo in controls and 0.86

(95% CI 0.60-1.23) for the three unblinded trials (Figure 21). Similarly, there was little difference

between the five adequately concealed trials and the nine trials with unclear concealment of allocation.

Combined RRs were 0.97 (95% CI 0.75-1.24) and 0.93 (95% CI 0.81-1.05), respectively (Figure 22).

Combined RRs in groups defined by the valency of the vaccine used in trials were also fairly similar:

1.00 (95% CI 0.87-1.16) for three trials of 23-valent vaccine, 0.95 (95% CI 0.81-1.11) for eight trials of

14-valent vaccine and 0.79 (95% CI 0.54-1.14) for the three trials of other vaccine valencies (Figure

23). When sorting the trials by mortality in controls there was some evidence to indicate a weak

protective effect of the vaccine in trials with lower mortality (Figure 24).

In developing countries (four trials) there was some evidence of a protective effect of the

vaccine with an RR of 0.84 (95% CI 0.70-1.02) while in developed countries (10 trials) the combined

RR was 1.00 (95% CI 0.87-1.14) (Figure 25). The trials from Russia did not report on mortality. Trials

in patients with respiratory illness showed little heterogeneity (I2 =0%, p=0.49), with a combined RR of

0.96 (95% CI 0.73-1.28). The two studies in children (both from Papua New Guinea) were also

homogeneous (I2 =0%, p=0.84), indicating a reduction in mortality (combined RR 0.70; 95% CI 0.51-

9

Confidential Draft

0.96). Conversely, the trials in the elderly or people with chronic illness were more heterogeneous (I2

=54%, p=0.042), with a combined RR of 0.99 (95% CI 0.85-1.17) (Figure 26).

10

Confidential Draft

Mortality due to pneumonia

Ten trials reported on mortality due to pneumonia. In two trials the 23-valent vaccine was

used, six trials used the 14-valent vaccine and two a vaccine of lower valency. The combined RR over

all trials was 0.77 (95% CI 0.53-0.1.11), with moderate heterogeneity (I2 of 37%, p=0.12) (Figure 27).

There were seven trials which were described as double-blind and used placebo or another vaccine in

controls. Meta-analysis of these trials showed some evidence of a protective effect of the vaccine with

an RR of 0.64 (95% CI 0.39-1.04) (Figure 28). The five trials with adequate concealment of allocation

produced a combined RR of 0.89 (95% CI 0.50-1.61), compared to a combined RR of 0.59 (95% CI

0.37-0.94) in the five trials with unclear concealment (Figure 29). The combined RR was 0.90 (95% CI

0.35-2.33) for the two trials of 23-valent vaccine, 0.69 (95% CI 0.39-1.20) for the six trials of 14-valent

vaccines and 0.88 (95% CI for the two trials of other valencies (Figure 30). No obvious pattern

emerged when ordering trials according to mortality from pneumonia in controls (Figure 31).

Stratifying trials by setting gave a combined RR of 0.50 (95% CI 0.33-0.76) for seven trials

from developing countries, compared to an RR of 1.06 (95% CI 0.76-1.47) for three trials from

industrialized countries (Figure 32). The trials from Russia did not report on mortality from pneumonia.

Four different types of study populations were included in this analysis. Two trials in patients with

respiratory illness gave a combined RR of 0.82 (95% CI 0.32-2.07), two studies of children in Papua

New Guinea an RR of 0.39 (95% CI 0.19-0.80), and the five studies in the elderly or patients with

chronic illness an RR of 1.08 (95% CI 0.72-1.62). The RR from a study in people over the age of 10yo

in Papua New Guinea was 0.57 (95% CI 0.34-0.94) although the authors of this trial report a loss of

records which may have affected results (Figure 33).

Other outcomes

Four trials reported mortality due to pneumococcal infection: Two used the 14-valent vaccine,

1 a 12-valent and 1 a 17 valent. There was little heterogeneity (I2 =14%, p=0.31). The combined RR

was 0.93 (95% CI 0.29-3.05) (Figure 34). Eight trials reported on bacteraemia, septicaemia or invasive

pneumococcal disease. Three used the 23-valent vaccine, 3 used the 14-valent vaccine, 1 a 12-valent

and 1 a 17-valent vaccine. Again, there was little heterogeneity in the results (I2 =0%, p=0.50). The

combined RR was for this outcome was 0.98 (95% CI 0.55-1.75) (Figure 35).

Meta-regression analyses The results from meta-regression analyses of the 20 trials that examined pneumonia of all

causes are shown in Table 3. Results are presented as ratios of relative risks (relative risk with

characteristic divided by relative risk without characteristic). Ratios above 1.0 correspond to a larger

relative risk for trials with characteristic and hence a smaller apparent benefit of the vaccine. In

univariate analyses, the year of publication, the year the study started and vaccine valency had little

influence on the between-trial variance (tau2). In other words, these variables contributed little to

heterogeneity. The degree of blinding was strongly associated with trial results, accounting for about

half of between-trial heterogeneity. The association with concealment of allocation was weak. The type

11

Confidential Draft

of study population and the trial setting also accounted for some heterogeneity. In multivariate

analyses adjusted for trial quality, ratios of relative risks generally moved closer to 1, indicating that

some of the effects observed in univariate analyses were due to differences in trial quality. Trial quality

and the type of study population explained virtually all between-trial heterogeneity. In this analysis, a

beneficial effect of the vaccine continued to be evident in the studies of miners and soldiers and the

study in individuals older than 10 years in Papua New Guinea.

Funnel plots Funnel plots for outcomes with ten or more trials (presumptive pneumococcal pneumonia,

pneumonia from all causes, mortality from all causes, mortality from pneumonia) are shown in Figures

36 to 39. There was some evidence of funnel plot asymmetry for trials of all-cause mortality, but not in

trials of the other outcomes.

12

Confidential Draft

Discussion

Our meta-analysis of the effectiveness of polysaccharide pneumococcal vaccine (PPV) on a

range of clinically relevant outcomes revealed a high degree of heterogeneity between trials, some of

which could be explained by the methodological quality of trials. Trials with higher quality, that is, those

with a double-blind design and adequate concealment of allocation, generally showed little evidence of

a protective effect of the vaccine in preventing presumptive pneumococcal pneumonia, pneumonia

from any cause, and all-cause mortality. There was some evidence, based on seven trials of high

quality, of a protective effect on mortality due to pneumonia, but confidence intervals were wide and

included 1. It nevertheless is possible that PPV has little effect on the risk of pneumonia but improves

outcome. Analyses of definitive pneumococcal pneumonia, bronchitis, mortality due to pneumococcal

infection, and bacteraemia were based on few trials or few events, and the results provide can

therefore neither confirm nor exclude a protective effect of the vaccine.

Strengths and limitations

This is the most comprehensive analysis of trials of PPV to date: our literature search was

thorough (with dedicated database searches covering South America, India and Africa as well as the

industrialized world) and we examined a range of clinically relevant health outcomes in numerous

population groups. Unlike many previous meta-analyses, we assessed sources of heterogeneity

between studies. A detailed analysis by age of the participants was not possible due to missing data in

a number of trials, however, we were able to approximate this in the analyses by population groups

where three broad age groups exist in our sub-groupings (children, young adults (the miner/soldier

group) and older/elderly adults)

Although it seems likely, based on empirical evidence of bias and theoretical considerations,

that trials of higher methodological quality will provide results that are closer to the truth than trials of

low quality, our study cannot prove that this is the case for the trials examined in the present review.

Our stratified analyses and meta-regressions are observational in nature, and other factors associated

both with the quality of the trial and the estimates of vaccine efficacy could have confounded our

results. When restricting analyses to trials of higher quality, between-trial heterogeneity tended to

decrease, which speaks against the presence of important confounding. Increasing the number of

strains included in the vaccine from 2 up to 23 was not associated with improved vaccine efficacy in

any of the outcomes examined. This is an unexpected but important observation, which casts doubt on

the vaccine’s efficacy in general. In fact, there was a trend in the opposite direction, with the older

trials using vaccines of lower valencies showing greater effects of the vaccine. However, the older

trials also tended to be of lower methodological quality.

Our review and meta-analysis included several disease outcomes but did not systematically

assess adverse events. Our protocol stipulated that adverse events would be examined, however, we

found that these were poorly reported: they were often listed only for a sub-group of study participants

13

Confidential Draft

or not reported separately for vaccines and controls. It is nevertheless important to note that no

serious side effects have been reported in any trials of a commercial PPV, which supports the safety

of the vaccine. There is some unavoidable inaccuracy in our results due to poor reporting in some of

the included trials. This was particularly an issue in the assessment of trial quality and of follow-up

time. Our results may therefore underestimate the importance of trial quality, due to imprecision in

measurement resulting in (non-differential) misclassification and residual confounding. For example, in

multivariable meta-regression analysis a beneficial effect of the vaccine on the risk of all-cause

pneumonia continued to be evident in the studies of miners and soldiers. Some of these studies were

done in the 1940s and 1970s, with poor reporting of methodology, and it is possible that their quality

was rated too highly in our study. In particular, assessment of outcomes in these trials may have been

unblinded, despite the use of an intervention in controls.. It is also likely that, due to the fundamental

difficulty in obtaining a definite aetiological cause of pneumonia, only a fraction of pneumonias caused

by S. pneumoniae are included in our definitive pneumococcal pneumonia analysis.

The trials we have included span many decades and there have been changes in the vaccine

and trial quality and reporting over this time. It could also be speculated that changes in the pathogen

may have occurred over this time. However, there is no obvious pattern in the effect of the vaccine

over time. Although widely used all over the world, there are relatively few trials of the 23-valent

vaccine which limits the extent of analyses which can be undertaken within this group. Finally, there

are no randomised controlled trials in splenectomised patients, a group in which the use of PPV is

frequently recommended.

Context

Our findings are at odds with those of several previous meta-analyses. We find that, after

controlling for methodological quality, there is little evidence that vaccination with PPV provides the

beneficial effect seen in previous analyses, which did not consider trial quality. An additional difference

relates to the three trials from Zhogolev 2003, which we included in our analysis. Since 2003, eight

meta-analyses have been done. However, six of these focused on restricted populations (for example

patients with sickle-cell anaemia) and therefore would not include these trials (in military recruits). Of

the remaining two meta-analyses, one only searched Medline, which does not index the Zhogolev

study, and the other lists its most recent amendment at a date before this paper could have appeared

in the Cochrane listings.

Implications and future research

The lack of clear evidence for a protective effect of the pneumococcal polysaccharide vaccine

should inform discussions on its place developing and industrialized countries, and different target

groups. Further high quality trials with appropriately blinded assessment of outcomes would be useful

to confirm or refute the hypothesis that the benefits observed in previous trials were associated with

lower methodological quality. Alternatively, emphasis could be placed on conducting high quality trials

of the conjugate vaccine in adults and increasing availability of this vaccine if found to be effective.

14

Confidential Draft

A systematic review of case-control studies would also be helpful. Such a review should again pay

attention to the quality of studies, and particularly to blinding of interviewers and patients to the study

objective.

15

Confidential Draft

References 1. Obaro, S.K., M.A. Monteil, and D.C. Henderson, The pneumococcal problem. Bmj, 1996.

312(7045): p. 1521-5. 2. Marrie, T.J., Therapeutic implications of macrolide resistance in pneumococcal community-

acquired lower respiratory tract infections. Int J Clin Pract, 2004. 58(8): p. 769-76. 3. Bruinsma, N., K.G. Kristinsson, S. Bronzwaer, P. Schrijnemakers, J. Degener, E. Tiemersma,

W. Hryniewicz, J. Monen, and H. Grundmann, Trends of penicillin and erythromycin resistance among invasive Streptococcus pneumoniae in Europe. J Antimicrob Chemother, 2004. 54(6): p. 1045-50.

4. Lipsky, B.A. and J.V. Hirschmann, Pneumococcal polysaccharide vaccines do not protect the elderly from pneumococcal infections. Neth J Med, 2004. 62(2): p. 33-5.

5. Rohani, M.Y., A. Raudzah, A.J. Ng, P.P. Ng, A.A. Zaidatul, I. Asmah, M. Murtaza, N. Parasakthy, M.Y. Mohd Yasmin, and Y.M. Cheong, Epidemiology of Streptococcus pneumoniae infection in Malaysia. Epidemiol Infect, 1999. 122(1): p. 77-82.

6. Kyaw, M.H., S. Clarke, G.F. Edwards, I.G. Jones, and H. Campbell, Serotypes/groups distribution and antimicrobial resistance of invasive pneumococcal isolates: implications for vaccine strategies. Epidemiol Infect, 2000. 125(3): p. 561-72.

7. Chang, C.C., R.J. Singleton, P.S. Morris, and A.B. Chang, Pneumococcal vaccines for children and adults with bronchiectasis. Cochrane Database Syst Rev, 2007(2): p. CD006316.

8. Granger, R., J. Walters, P.J. Poole, T.J. Lasserson, P. Mangtani, C.J. Cates, and R. Wood-Baker, Injectable vaccines for preventing pneumococcal infection in patients with chronic obstructive pulmonary disease. Cochrane Database Syst Rev, 2006(4): p. CD001390.

9. Chaithongwongwatthana, S., W. Yamasmit, S. Limpongsanurak, P. Lumbiganon, J.A. Desimone, J. Baxter, and J.E. Tolosa, Pneumococcal vaccination during pregnancy for preventing infant infection. Cochrane Database Syst Rev, 2006(1): p. CD004903.

10. Melegaro, A. and W.J. Edmunds, The 23-valent pneumococcal polysaccharide vaccine. Part I. Efficacy of PPV in the elderly: a comparison of meta-analyses. Eur J Epidemiol, 2004. 19(4): p. 353-63.

11. Conaty, S., L. Watson, J. Dinnes, and N. Waugh, The effectiveness of pneumococcal polysaccharide vaccines in adults: a systematic review of observational studies and comparison with results from randomised controlled trials. Vaccine, 2004. 22(23-24): p. 3214-24.

12. Davies, E.G., C. Riddington, R. Lottenberg, and N. Dower, Pneumococcal vaccines for sickle cell disease. Cochrane Database Syst Rev, 2004(1): p. CD003885.

13. Straetemans, M., E.A. Sanders, R.H. Veenhoven, A.G. Schilder, R.A. Damoiseaux, and G.A. Zielhuis, Pneumococcal vaccines for preventing otitis media. Cochrane Database Syst Rev, 2004(1): p. CD001480.

14. Dear, K., J. Holden, R. Andrews, and D. Tatham, Vaccines for preventing pneumococcal infection in adults. Cochrane Database Syst Rev, 2003(4): p. CD000422.

15. Puig-Barbera, J., A. Belenguer Varea, M. Goterris Pinto, and M.J. Brines Benlliure, [Pneumococcal vaccine effectiveness in the elderly. Systematic review and meta-analysis]. Aten Primaria, 2002. 30(5): p. 269-81; discussion 281-3.

16. Watson, L., B.J. Wilson, and N. Waugh, Pneumococcal polysaccharide vaccine: a systematic review of clinical effectiveness in adults. Vaccine, 2002. 20(17-18): p. 2166-73.

17. Cornu, C., D. Yzebe, P. Leophonte, J. Gaillat, J.P. Boissel, and M. Cucherat, Efficacy of pneumococcal polysaccharide vaccine in immunocompetent adults: a meta-analysis of randomized trials. Vaccine, 2001. 19(32): p. 4780-90.

18. Sheikh, A., B. Alves, and S. Dhami, Pneumococcal vaccine for asthma. Cochrane Database Syst Rev, 2002(1): p. CD002165.

19. Moore, R.A., P.J. Wiffen, and B.A. Lipsky, Are the pneumococcal polysaccharide vaccines effective? Meta-analysis of the prospective trials. BMC Fam Pract, 2000. 1: p. 1.

16

Confidential Draft

20. Hutchison, B.G., A.D. Oxman, H.S. Shannon, S. Lloyd, C.A. Altmayer, and K. Thomas, Clinical effectiveness of pneumococcal vaccine. Meta-analysis. Can Fam Physician, 1999. 45: p. 2381-93.

21. Go, E.S. and Z.K. Ballas, Anti-pneumococcal antibody response in normal subjects: a meta-analysis. J Allergy Clin Immunol, 1996. 98(1): p. 205-15.

22. Fine, M.J., M.A. Smith, C.A. Carson, F. Meffe, S.S. Sankey, L.A. Weissfeld, A.S. Detsky, and W.N. Kapoor, Efficacy of pneumococcal vaccination in adults. A meta-analysis of randomized controlled trials. Arch Intern Med, 1994. 154(23): p. 2666-77.

23. CDC, Recommended Adult Immunization Schedule - United States, October 2006-September 2007. MMWR, 2006. 55(40): p. Q1-Q4.

24. Pebody, R.G., T. Leino, H. Nohynek, W. Hellenbrand, S. Salmaso, and P. Ruutu, Pneumococcal vaccination policy in Europe. Euro Surveill, 2005. 10(9): p. 174-8.

25. DerSimonian, R. and N. Laird, Meta-analysis in clinical trials. Control Clin Trials, 1986. 7(3): p. 177-88.

26. Higgins, J.P. and S.G. Thompson, Quantifying heterogeneity in a meta-analysis. Stat Med, 2002. 21(11): p. 1539-58.

27. Alfageme, I., R. Vazquez, N. Reyes, J. Munoz, A. Fernandez, M. Hernandez, M. Merino, J. Perez, and J. Lima, Clinical efficacy of anti-pneumococcal vaccination in patients with COPD. Thorax, 2006. 61(3): p. 189-95.

28. Austrian, R., R.M. Douglas, G. Schiffman, A.M. Coetzee, H.J. Koornhof, S. Hayden-Smith, and R.D. Reid, Prevention of pneumococcal pneumonia by vaccination. Trans Assoc Am Physicians, 1976. 89: p. 184-94.

29. Austrian, R., Surveillance of pneumococcal infection for field trials of polyvalent pneumococcal vaccines, M.N.I.o.H. Bethesda, National Institute of Allergy and Infectious Diseases, Editor. 1980.

30. Davis, A.L., C.P. Aranda, G. Schiffman, and L.C. Christianson, Pneumococcal infection and immunologic response to pneumococcal vaccine in chronic obstructive pulmonary disease. A pilot study. Chest, 1987. 92(2): p. 204-12.

31. Douglas, R.M. and H.B. Miles, Vaccination against Streptococcus pneumoniae in childhood: lack of demonstrable benefit in young Australian children. J Infect Dis, 1984. 149(6): p. 861-9.

32. Douglas, R.M., D. Hansman, B. McDonald, J. Paton, and K. Kirke, Pneumococcal vaccine in aboriginal children--a randomized controlled trial involving 60 children. Community Health Stud, 1986. 10(2): p. 189-96.

33. French, N., C.F. Gilks, A. Mujugira, C. Fasching, J. O'Brien, and E.N. Janoff, Pneumococcal vaccination in HIV-1-infected adults in Uganda: humoral response and two vaccine failures. Aids, 1998. 12(13): p. 1683-9.

34. Gaillat, J., D. Zmirou, M.R. Mallaret, D. Rouhan, J.P. Bru, J.P. Stahl, P. Delormas, and M. Micoud, [Clinical trial of an antipneumococcal vaccine in elderly subjects living in institutions]. Rev Epidemiol Sante Publique, 1985. 33(6): p. 437-44.

35. Honkanen, P.O., T. Keistinen, L. Miettinen, E. Herva, U. Sankilampi, E. Laara, M. Leinonen, S.L. Kivela, and P.H. Makela, Incremental effectiveness of pneumococcal vaccine on simultaneously administered influenza vaccine in preventing pneumonia and pneumococcal pneumonia among persons aged 65 years or older. Vaccine, 1999. 17(20-21): p. 2493-500.

36. John, A.B., A. Ramlal, H. Jackson, G.H. Maude, A.W. Sharma, and G.R. Serjeant, Prevention of pneumococcal infection in children with homozygous sickle cell disease. Br Med J (Clin Res Ed), 1984. 288(6430): p. 1567-70.

37. Karsh, J., N. Pavlidis, G. Schiffman, and H.M. Moutsopoulos, Immunization of patients with Sjogren's syndrome with pneumococcal polysaccharide vaccine: a randomized trial. Arthritis Rheum, 1980. 23(11): p. 1294-8.

38. Kaufman, P., Pneumonia in old age: active immunization against pneumonia with pneumococcus polysaccharide; results of a six year study. Arch Intern Med, 1947. 79: p. 518.

39. Kaufman, P., Studies on old age pneumonia. II. Prophylactic effect of pneumococcus polysaccharide against pneumonia. Archives of Internal Medicine, 1941. 61: p. 304-319.

40. Klastersky, J., P. Mommen, F. Cantraine, and A. Safary, Placebo controlled pneumococcal immunization in patients with bronchogenic carcinoma. Eur J Cancer Clin Oncol, 1986. 22(7): p. 807-13.

41. Koivula, I., M. Sten, M. Leinonen, and P.H. Makela, Clinical efficacy of pneumococcal vaccine in the elderly: a randomized, single-blind population-based trial. Am J Med, 1997. 103(4): p. 281-90.

17

Confidential Draft

18

42. Leech, J.A., A. Gervais, and F.L. Ruben, Efficacy of pneumococcal vaccine in severe chronic obstructive pulmonary disease. Cmaj, 1987. 136(4): p. 361-5.

43. Lehmann, D., T.F. Marshall, I.D. Riley, and M.P. Alpers, Effect of pneumococcal vaccine on morbidity from acute lower respiratory tract infections in Papua New Guinean children. Ann Trop Paediatr, 1991. 11(3): p. 247-57.

44. MacLeod, C.M., R.D. Hodges, M. Heidelberger, and W.G. Bernhard, Prevention of pneumococcal pneumonia by immunization with specific capsular serotypes. J Exp Med, 1945. 82: p. 445-465.

45. Mäkelä, P.H., M. Sibakov, E. Herva, J. Henrichsen, J. Luotonen, M. Timonen, M. Leinonen, M. Koskela, J. Pukander, S. Pontynen, P. Gronroos, and P. Karma, Pneumococcal vaccine and otitis media. Lancet, 1980. 2(8194): p. 547-51.

46. Örtqvist, A., J. Hedlund, L.A. Burman, E. Elbel, M. Hofer, M. Leinonen, I. Lindblad, B. Sundelof, and M. Kalin, Randomised trial of 23-valent pneumococcal capsular polysaccharide vaccine in prevention of pneumonia in middle-aged and elderly people. Swedish Pneumococcal Vaccination Study Group. Lancet, 1998. 351(9100): p. 399-403.

47. Riley, I.D., P.I. Tarr, M. Andrews, M. Pfeiffer, R. Howard, P. Challands, and G. Jennison, Immunisation with a polyvalent pneumococcal vaccine. Reduction of adult respiratory mortality in a New Guinea Highlands community. Lancet, 1977. 1(8026): p. 1338-41.

48. Riley, I.D., F.A. Everingham, D.E. Smith, and R.M. Douglas, Immunisation with a polyvalent pneumococcal vaccine. Effect of respiratory mortality in children living in the New Guinea highlands. Arch Dis Child, 1981. 56(5): p. 354-7.

49. Riley, I.D., D. Lehmann, M.P. Alpers, T.F. Marshall, H. Gratten, and D. Smith, Pneumococcal vaccine prevents death from acute lower-respiratory-tract infections in Papua New Guinean children. Lancet, 1986. 2(8512): p. 877-81.

50. Rosen, C., P. Christensen, J. Henrichsen, B. Hovelius, and K. Prellner, Beneficial effect of pneumococcal vaccination on otitis media in children over two years old. Int J Pediatr Otorhinolaryngol, 1984. 7(3): p. 239-46.

51. Schuller, D.E., Prophylaxis of otitis media in asthmatic children. Pediatr Infect Dis, 1983. 2(4): p. 280-3.

52. Steentoft, J., H.B. Konradsen, J. Hilskov, G. Gislason, and J.R. Andersen, Response to pneumococcal vaccine in chronic obstructive lung disease--the effect of ongoing, systemic steroid treatment. Vaccine, 2006. 24(9): p. 1408-12.

53. Shahid, N.S., M.C. Steinhoff, S.S. Hoque, T. Begum, C. Thompson, and G.R. Siber, Serum, breast milk, and infant antibody after maternal immunisation with pneumococcal vaccine. Lancet, 1995. 346(8985): p. 1252-7.

54. Simberkoff, M.S., A.P. Cross, M. Al-Ibrahim, A.L. Baltch, P.J. Geiseler, J. Nadler, A.S. Richmond, R.P. Smith, G. Schiffman, D.S. Shepard, and et al., Efficacy of pneumococcal vaccine in high-risk patients. Results of a Veterans Administration Cooperative Study. N Engl J Med, 1986. 315(21): p. 1318-27.

55. Smit, P., D. Oberholzer, S. Hayden-Smith, H.J. Koornhof, and M.R. Hilleman, Protective efficacy of pneumococcal polysaccharide vaccines. Jama, 1977. 238(24): p. 2613-6.

56. Zhogolev, S.D., V.D. Mosiagin, V.U. Demidovich, P.I. Mel'nichenko, and P.I. Ogarkov, [Efficacy of pneumococcal vaccine in military units]. Zh Mikrobiol Epidemiol Immunobiol, 2003(2): p. 36-42.

Confidential Draft

Table 1: Characteristics of included studies. Author, year of publication

Trial N

Contributed to outcomes1

Blinding2 Allocation conceal- ment adequate

Study population Country Vaccine valency

% male 3

Age mean (range) 3

Incid. of pneu- monia in controls (%)

Mortality in controls (%)

Max.length follow up (years)

Alfageme (2006) 600 2, 3, 5, 6 3 Unclear COPD patients Spain 23 95 68.5 (61 - 73) 15 19 2.7 Austrian (1976) 4500 2, 3 2 Unclear Miners South Africa 13 100 nr 12 Nr Austrian a (1980) 1300 3, 5, 6, 8, 12 1 Unclear Hospitalised/home USA 12 NR NR (NR - NR) 21 6.2 3.0 Austrian b (1980) 13600 3, 4, 5, 6, 7 2 Adequate Elderly/chronic illness USA 12 NR NR (45 - NR) 4 0.69 2.8 Davis (1987) 103 1, 2, 3, 5, 6 1 Unclear COPD patients USA 14 NR 62.5 (NR - NR) 13 25 Nr Douglas (1986) 39 10 1 Adequate Children Australia 14 NR 3.3 (2 - NR) 1.5 Douglas (1984) 853 3, 10 1 Adequate Children Australia 14 55 NR (2 - 4.5) 0.7 2.0 French (2000) 1323 3, 5, 8 1 Unclear HiV+ Uganda 23 29 31 (15 - NR) 4 28 2.7 Gaillat (1985) 1827 3, 5 3 Unclear Hospitalised/home France 14 34 74 (NR - NR) 4.1 23 2.0 Honkanen (1999) 26925 2, 3, 8 2 Unclear Elderly/chronic illness Finland 23 38 73.5 (65 - NR) 0.9 3.2 John (1984) 242 7, 8 2 Unclear Children Jamaica 14 53 1.2 (0.5 - 2.9) 5.0 Kaufman (1947) 8783 3, 5 3 Unclear Hospitalised/home USA 3 NR 67 (NR - NR) 4.1 1.3 1.5 Klastersky (1986) 50 2, 7, 8 2 Adequate Bronchogenic carcin. Belgium 17 96 61 (42 - 78) Nr Koivula (1997) 2837 2, 3, 5, 6 2 Unclear Elderly/chronic illness Finland 14 37 NR (60 - NR) 4.7 11 3.0 Leech (1987) 189 5, 8, 12 1 Unclear COPD patients Canada 14 71 67 (40 - 89) 11 2.2 Lehmann (1991) 1371 12 1 Unclear Children Papua New Guinea 14/23 NR NR (0.3 - 4.9) Nr MacLeod (1945) 17035 2 2 Uncear Soldiers USA 4 100 23.3 (18 - NR) Nr Mäkelä (1981) 446 10 2 Unclear Children Finland 14 NR NR (2 - 6.9) 0.5 Örtqvist (1998) 691 1, 2, 3, 5, 6, 8 1 Adequate Elderly/chronic illness Sweden 23 48 69.2 (50 - 85) 16 8 Nr Riley (1986) 4862 5, 6 1 Unclear Children Papua New Guinea 14/23 NR NR (0.3 - 5) 2.9 4.0 Riley (1977) 11958 3, 5, 6, 12 1 Adequate People over 10yo Papua New Guinea 14 NR NR (10 - NR) 0.8 2.8 Nr Riley (1981) 871 5, 6, 12 1 Adequate Children Papua New Guinea 14 NR NR (0.5 - 4.9) 3.8 3.0 Rosen (1984) 405 10 1 Unclear Children Sweden 14 54 3.3 (0.5 - 5) 2.0 Schuller (1983) 80 10 3 Unclear Children USA 14 61 3.5 (2 - 6) 1.0 Simberkoff (1986) 2295 2, 3, 4, 5, 6, 7, 8 1 Adequate Elderly/chronic illness USA 14 NR 61.2 (55 - NR) 3.6 15 Nr Smit a (1977) 3019 2, 3, 4 2 Unclear Miners South Africa 6 100 NR (NR - NR) 12 2.3 Smit b (1977) 1675 2, 3, 4 2 Unclear Miners South Africa 12 100 NR (NR - NR) 5.1 1.6 Steentoft (2006) 49 3, 12 3 Adequate COPD patients Denmark 23 55 NR (47 - 86) 42 Nr Zhogolev a (2003) 144 3 3 Unclear Soldiers Russia 23 100 NR (NR - NR) 5.5 Nr Zhogolev b (2003) 827 3 3 Unclear Soldiers Russia 23 100 NR (NR - NR) 21 Nr Zhogolev c (2003) 1777 3 3 Unclear Soldiers Russia 23 100 NR (NR - NR) 11 Nr 1list of outcomes: 1) Definitive pneumococcal pneumonia, 2) Presumptive pneumococcal pneumonia, 3) All-cause pneumonia, 4) Bronchitis, 5) All-cause mortality, 6) Mortality due to pneumonia, 7) Mortality due to pneumococcal infection, 8) Bacteraemia, septicaemia or invasive pneumococcal disease, 10) Otitis media, 12) Other, such as acute lower respiratory tract infection. 2 Trials were grouped into 1) Reported to be double-blind if the wording “double-blind” was used in the publication and the trial was placebo-controlled or another vaccine was used in controls, 2) Not reported to be double-blind, but trial was placebo-controlled or another vaccine was used in controls, or 3) Unblinded if not reported to be double-blind and no intervention used in controls. 3 NR = not reported

19

20

Confidential Draft

Outcome No of trials reporting this outcome

Combined n of study population for this outcome

Combined number of cases in both vaccinees and controls

Definitive pneumococcal pneumonia 2 794 7 Presumptive p. pneumonia 11 56564 589 Pneumonia, all causes 20 83517 2729 Bronchitis 4 20589 1689 Mortality all cause 14 51098 2396 Mortality due to pneumonia 10 39117 251 Mortality due to pneumoc. infection 4 16,041 18 Bacteraemia 8 32,869 50 Meningitis None of the studies reported meningitis as an outcome. Otitis media 5 1823 170 Adverse events Many of the studies reported only a comment, e.g. “no serious

side effects were reported”.

Table 2: Reported outcomes of RCTs using pneumococcal polysaccharide vaccine

Confidential Draft

Table 3: Univariable and multivariable meta-regression analysis of the effect of pneumococcal polysaccharide vaccine on the risk of pneumonia from all causes in 20 trials.

Univariable analysis Multivariable analysisa

Variable Ratio of relative risks*

(95%CI)

Tau2 Ratio of relative risks*

(95%CI)

Tau2

Nil 0.3116 0.3116

Publication year 1.01 (0.99-1.03) 0.3243 1.01 (0.99-1.02) 0.1638

Year study started 1.00 (0.98-1.02) 0.3312 1.00 (0.99-1.02) 0.1731

Vaccine valencyb

- 14 valent

1.07 (0.49-2.33)

0.93 (0.45-1.92)

0.3510

0.70 (0.36-1.34)

0.71(0.38-1.31)

0.1758

- Other valencies

Concealment of allocationc

1.56 (0.83-2.93)

0.2919

1.11 (0.63-1.95)

0.1659

- Adequate

Blindingd

- Not describd as double- blind, controlled

1.91 (1.12-3.26)

2.76 (1.60-4.78)

0.1520

1.88 (1.08-3.30)

2.62 (1.41-4.90)

0.1659

- Described as double blind, controlled Study population e

- Respiratory

- HIV patients

- Children

- Miners/soldiers

0.89 (0.39-2.04)

2.12 (0.67-6.72)

1.49 (0.21-10.68)

0.48 (0.27- 0.87)

0.85 (0.27- 2.65)

0.1929

1.71 (0.97-3.02)

1.56 (0.75-3.25)

1.14 (0.15-0.45)

0.52 (0.38-0.73)

0.64 (0 .32-1.31)

0.0054

- >10 years old in PNG

Trial setting f

- Industrialized countries

1.16 (0.66-2.05)

- Other

0.40 (0.18-0.92)

0.1929

1.46 (0.87-2.45)

0.80 (0.33-1.95)

0.1027

a Adjusted for blinding and concealment of allocation. For these two variables, adjusted results reported are those from the model containing these two variables alone b Compared to 23 valent PPV c Adequate compared to unclear/inadequate group d Compared to unblinded e Compared to elderly/chronically ill f Compared to trials in developing countries * Relative risk with characteristic divided by relative risk without characteristic. Ratios above 1.0 correspond to a larger relative risk for trials with characteristic and hence a smaller apparent benefit of the vaccine. Trials described as double blind and using placebo or another vaccine in the control group show a less beneficial effect than unblended trials, for example.

21

Confidential Draft

Figure 1: Identification of eligible trials

516 studies identified: 269 Embase and Medline 147 Cochrane 63 from references lists 22 LILACS 1 AIM 14 IndMed

excluded: 73 duplicates 73 conjugate vaccine only 202 were not RCT’s 74a did not report clinically relevant outcomes (usually antibody response only) 1 PPV used as booster only 1 conference proceedings 1 only children up to 9 month of age

91 potentially eligible trials identified after screening of title and abstract

Included 27 publications, reporting 31 trials (2 with 2 trials, 1 with 3 trials)

65 Excluded after screening the full text: 40 were not RCT’s 8 were only conference proceedings, editorials or short letters 6 did not report clinically relevant outcomes 5 reported a follow-up of a subpopulation only, or of an earlier trial phase, or an additional data analysis 2 were in children under 2 yrs of age 2 used PPV only as booster vaccination 1 reported only pilot phase of trial

22

Confidential Draft

Figure 2: Year of trial start and valency of polysaccharide pneumococcal vaccine.

26

1423

Val

ency

of v

acci

ne

1935 1945 1955 1965 1975 1985 1995 2005year trial started

23

Confidential Draft

Definitive pneumococcal pneumonia Figure 3: Meta-analysis of clinical trials of the effect of pneumococcal polysaccharide vaccines on the risk of definitive pneumococcal pneumonia.

NOTE: Weights are from random effects analysis

Overall (I-squared = 48.6%, p = 0.163)

ID

Study

Davis 1987

Örtqvist 1998

0.62 (0.05, 8.61)

RR (95% CI)

3.18 (0.13, 76.20)

0.21 (0.02, 1.77)

2/389

vaccinees

Events,

1/50

1/339

5/405

controls

Events,

0/53

5/352

38 (-761, 95)

efficacy (%)

Vaccine

-218 (-7520, 87)

79 (-77, 98)

valency

Vaccine

14

23

0.62 (0.05, 8.61)

RR (95% CI)

3.18 (0.13, 76.20)

0.21 (0.02, 1.77)

2/389

vaccinees

Events,

1/50

1/339

Reduced risk Increased risk 1.125.25 .5 1 2 4 8

Definitive pneumococcal pneumonia

24

Confidential Draft

Presumptive pneumococcal pneumonia Figure 4: Meta-analysis of clinical trials of the effect of pneumococcal polysaccharide vaccines on the risk of presumptive pneumococcal pneumonia. Trials are sorted by start of study.

NOTE: Weights are from random effects analysis

Overall (I-squared = 74.4%, p = 0.000)

Alfageme 2006

Study

Klastersky 1986

Honkanen 1999

Simberkoff 1986

Smit a 1977

Davis 1987

Smit b 1977

ID

Koivula 1997

Austrian 1976

MacLeod 1945

Örtqvist 1998

0.64 (0.43, 0.96)

0.09 (0.01, 1.64)

0.40 (0.08, 1.99)

1.20 (0.80, 1.82)

1.07 (0.53, 2.16)

0.24 (0.12, 0.49)

5.29 (0.26, 107.63)

0.06 (0.01, 0.48)

RR (95% CI)

0.85 (0.51, 1.41)

0.41 (0.29, 0.58)

0.69 (0.50, 0.97)

1.23 (0.64, 2.36)

231/28806

0/300

Events,

2/26

52/13980

16/1145

9/983

2/50

1/540

vaccinees

26/1364

44/1493

60/8586

19/339

358/27758

5/300

Events,

4/21

40/12945

15/1150

38/985

0/53

16/550

controls

33/1473

106/1480

85/8449

16/352

36 (4, 57)

91 (-64, 99)

Vaccine

60 (-99, 92)

-20 (-82, 20)

-7 (-116, 47)

76 (51, 88)

-429 (-10663, 74)

94 (52, 99)

efficacy (%)

15 (-41, 49)

59 (42, 71)

31 (3, 50)

-23 (-136, 36)

23

Vaccine

17

23

14

6

14

12

valency

14

13

4

23

0.64 (0.43, 0.96)

0.09 (0.01, 1.64)

0.40 (0.08, 1.99)

1.20 (0.80, 1.82)

1.07 (0.53, 2.16)

0.24 (0.12, 0.49)

5.29 (0.26, 107.63)

0.06 (0.01, 0.48)

RR (95% CI)

0.85 (0.51, 1.41)

0.41 (0.29, 0.58)

0.69 (0.50, 0.97)

1.23 (0.64, 2.36)

231/28806

0/300

Events,

2/26

52/13980

16/1145

9/983

2/50

1/540

vaccinees

26/1364

44/1493

60/8586

19/339

Reduced risk Increased risk

1.125 .25 .5 1 2 4 8

Sorted by start of studyPresumptive pneumococcal pneumonia

25

Confidential Draft

Figure 5: Meta-analysis of clinical trials of the effect of pneumococcal polysaccharide vaccines on the risk of presumptive pneumococcal pneumonia. Trials are sorted by blinding status.

NOTE: Weights are from random effects analysis

.

.

.

Overall (I-squared = 74.4%, p = 0.000)

Not reported to be double blind, placebo-controlled

Reported to be double blind, placebo-controlled

MacLeod 1945

Smit b 1977

Simberkoff 1986

Örtqvist 1998

Davis 1987

Subtotal (I-squared = 79.8%, p = 0.000)

Klastersky 1986

Koivula 1997

ID

Subtotal (I-squared = .%, p = .)

Alfageme 2006

Smit a 1977

Honkanen 1999

Austrian 1976

Subtotal (I-squared = 0.0%, p = 0.593)

Unblinded, not placebo-controlled

Study

0.64 (0.43, 0.96)

0.69 (0.50, 0.97)

0.06 (0.01, 0.48)

1.07 (0.53, 2.16)

1.23 (0.64, 2.36)

5.29 (0.26, 107.63)

0.54 (0.33, 0.86)

0.40 (0.08, 1.99)

0.85 (0.51, 1.41)

RR (95% CI)

0.09 (0.01, 1.64)

0.09 (0.01, 1.64)

0.24 (0.12, 0.49)

1.20 (0.80, 1.82)

0.41 (0.29, 0.58)

1.20 (0.75, 1.92)

231/28806

60/8586

1/540

16/1145

19/339

2/50

194/26972

2/26

26/1364

vaccinees

0/300

0/300

9/983

52/13980

44/1493

37/1534

Events,

358/27758

85/8449

16/550

15/1150

16/352

0/53

322/25903

4/21

33/1473

controls

5/300

5/300

38/985

40/12945

106/1480

31/1555

Events,

36 (4, 57)

31 (3, 50)

94 (52, 99)

-7 (-116, 47)

-23 (-136, 36)

-429 (-10663, 74)

46 (14, 67)

60 (-99, 92)

15 (-41, 49)

efficacy (%)

91 (-64, 99)

91 (-64, 99)

76 (51, 88)

-20 (-82, 20)

59 (42, 71)

-20 (-92, 25)

Vaccine

4

12

14

23

14

17

14

valency

23

6

23

13

Vaccine

0.64 (0.43, 0.96)

0.69 (0.50, 0.97)

0.06 (0.01, 0.48)

1.07 (0.53, 2.16)

1.23 (0.64, 2.36)

5.29 (0.26, 107.63)

0.54 (0.33, 0.86)

0.40 (0.08, 1.99)

0.85 (0.51, 1.41)

RR (95% CI)

0.09 (0.01, 1.64)

0.09 (0.01, 1.64)

0.24 (0.12, 0.49)

1.20 (0.80, 1.82)

0.41 (0.29, 0.58)

1.20 (0.75, 1.92)

231/28806

60/8586

1/540

16/1145

19/339

2/50

194/26972

2/26

26/1364

vaccinees

0/300

0/300

9/983

52/13980

44/1493

37/1534

Events,

Reduced risk Increased risk

1.125 .25 .5 1 2 4 8

By blindingPresumptive pneumococcal pneumonia

26

Confidential Draft

Figure 6: Meta-analysis of clinical trials of the effect of pneumococcal polysaccharide vaccines on the risk of presumptive pneumococcal pneumonia. Trials are sorted by concealment of allocation.

NOTE: Weights are from random effects analysis

.

.

Overall (I-squared = 74.4%, p = 0.000)

ID

Austrian 1976

Simberkoff 1986

Davis 1987

Smit a 1977

Subtotal (I-squared = 0.0%, p = 0.447)

Honkanen 1999

Unclear

Örtqvist 1998

Adequate

Subtotal (I-squared = 78.8%, p = 0.000)

Smit b 1977

Koivula 1997

Klastersky 1986

Alfageme 2006

MacLeod 1945

Study

0.64 (0.43, 0.96)

RR (95% CI)

0.41 (0.29, 0.58)

1.07 (0.53, 2.16)

5.29 (0.26, 107.63)

0.24 (0.12, 0.49)

1.06 (0.67, 1.67)

1.20 (0.80, 1.82)

1.23 (0.64, 2.36)

0.55 (0.33, 0.90)

0.06 (0.01, 0.48)

0.85 (0.51, 1.41)

0.40 (0.08, 1.99)

0.09 (0.01, 1.64)

0.69 (0.50, 0.97)

231/28806

vaccinees

44/1493

16/1145

2/50

9/983

37/1510

52/13980

19/339

194/27296

1/540

26/1364

2/26

0/300

60/8586

Events,

358/27758

controls

106/1480

15/1150

0/53

38/985

35/1523

40/12945

16/352

323/26235

16/550

33/1473

4/21

5/300

85/8449

Events,

36 (4, 57)

efficacy (%)

59 (42, 71)

-7 (-116, 47)

-429 (-10663, 74)

76 (51, 88)

-6 (-67, 33)

-20 (-82, 20)

-23 (-136, 36)

45 (10, 67)

94 (52, 99)

15 (-41, 49)

60 (-99, 92)

91 (-64, 99)

31 (3, 50)

Vaccine

valency

13

14

14

6

23

23

12

14

17

23

4

Vaccine

0.64 (0.43, 0.96)

RR (95% CI)

0.41 (0.29, 0.58)

1.07 (0.53, 2.16)

5.29 (0.26, 107.63)

0.24 (0.12, 0.49)

1.06 (0.67, 1.67)

1.20 (0.80, 1.82)

1.23 (0.64, 2.36)

0.55 (0.33, 0.90)

0.06 (0.01, 0.48)

0.85 (0.51, 1.41)

0.40 (0.08, 1.99)

0.09 (0.01, 1.64)

0.69 (0.50, 0.97)

231/28806

vaccinees

44/1493

16/1145

2/50

9/983

37/1510

52/13980

19/339

194/27296

1/540

26/1364

2/26

0/300

60/8586

Events,

Reduced risk Increased risk

1.125 .25 .5 1 2 4 8

By concealment of allocationPresumptive pneumococcal pneumonia

27

Confidential Draft

Figure 7: Meta-analysis of clinical trials of the effect of pneumococcal polysaccharide vaccines on the risk of presumptive pneumococcal pneumonia. Trials are sorted by vaccine valency.

NOTE: Weights are from random effects analysis

.

.

.

Overall (I-squared = 74.4%, p = 0.000)

Alfageme 2006

Koivula 1997

Subtotal (I-squared = 35.9%, p = 0.210)

Klastersky 1986

Study

Smit b 1977

23-valent

MacLeod 1945

Subtotal (I-squared = 0.0%, p = 0.461)

14-valent

Örtqvist 1998

other valencies

Honkanen 1999

Davis 1987

Subtotal (I-squared = 70.1%, p = 0.010)

Simberkoff 1986

ID

Smit a 1977

Austrian 1976

0.64 (0.43, 0.96)

0.09 (0.01, 1.64)

0.85 (0.51, 1.41)

1.12 (0.66, 1.90)

0.40 (0.08, 1.99)

0.06 (0.01, 0.48)

0.69 (0.50, 0.97)

0.95 (0.63, 1.43)

1.23 (0.64, 2.36)

1.20 (0.80, 1.82)

5.29 (0.26, 107.63)

0.39 (0.23, 0.66)

1.07 (0.53, 2.16)

RR (95% CI)

0.24 (0.12, 0.49)

0.41 (0.29, 0.58)

231/28806

0/300

26/1364

71/14619

2/26

Events,

1/540

60/8586

44/2559

19/339

52/13980

2/50

116/11628

16/1145

vaccinees

9/983

44/1493

358/27758

5/300

33/1473

61/13597

4/21

Events,

16/550

85/8449

48/2676

16/352

40/12945

0/53

249/11485

15/1150

controls

38/985

106/1480

36 (4, 57)

91 (-64, 99)

15 (-41, 49)

-12 (-90, 34)

60 (-99, 92)

Vaccine

94 (52, 99)

31 (3, 50)

5 (-43, 37)

-23 (-136, 36)

-20 (-82, 20)

-429 (-10663, 74)

61 (34, 77)

-7 (-116, 47)

efficacy (%)

76 (51, 88)

59 (42, 71)

23

14

17

Vaccine

12

4

23

23

14

14

valency

6

13

0.64 (0.43, 0.96)

0.09 (0.01, 1.64)

0.85 (0.51, 1.41)

1.12 (0.66, 1.90)

0.40 (0.08, 1.99)

0.06 (0.01, 0.48)

0.69 (0.50, 0.97)

0.95 (0.63, 1.43)

1.23 (0.64, 2.36)

1.20 (0.80, 1.82)

5.29 (0.26, 107.63)

0.39 (0.23, 0.66)

1.07 (0.53, 2.16)

RR (95% CI)

0.24 (0.12, 0.49)

0.41 (0.29, 0.58)

231/28806

0/300

26/1364

71/14619

2/26

Events,

1/540

60/8586

44/2559

19/339

52/13980

2/50

116/11628

16/1145

vaccinees

9/983

44/1493

Reduced risk Increased risk

1.125 .25 .5 1 2 4 8

By vaccine valencyPresumptive pneumococcal pneumonia

28

Confidential Draft

Figure 8: Meta-analysis of clinical trials of the effect of pneumococcal polysaccharide vaccines on the risk of presumptive pneumococcal pneumonia. Trials are ordered by the incidence of pneumonia in controls.

NOTE: Weights are from random effects analysis

Overall (I-squared = 74.4%, p = 0.000)

ID

Austrian 1976

Klastersky 1986

Smit a 1977

Koivula 1997

MacLeod 1945

Simberkoff 1986

Örtqvist 1998

Davis 1987

Smit b 1977

Alfageme 2006

Honkanen 1999

Study

0.64 (0.43, 0.96)

RR (95% CI)

0.41 (0.29, 0.58)

0.40 (0.08, 1.99)

0.24 (0.12, 0.49)

0.85 (0.51, 1.41)

0.69 (0.50, 0.97)

1.07 (0.53, 2.16)

1.23 (0.64, 2.36)

5.29 (0.26, 107.63)

0.06 (0.01, 0.48)

0.09 (0.01, 1.64)

1.20 (0.80, 1.82)

231/28806

vaccinees

44/1493

2/26

9/983

26/1364

60/8586

16/1145

19/339

2/50

1/540

0/300

52/13980

Events,

358/27758

controls

106/1480

4/21

38/985

33/1473

85/8449

15/1150

16/352

0/53

16/550

5/300

40/12945

Events,

36 (4, 57)

efficacy (%)

59 (42, 71)

60 (-99, 92)

76 (51, 88)

15 (-41, 49)

31 (3, 50)

-7 (-116, 47)

-23 (-136, 36)

-429 (-10663, 74)

94 (52, 99)

91 (-64, 99)

-20 (-82, 20)

Vaccine

valency

13

17

6

14

4

14

23

14

12

23

23

Vaccine

0.64 (0.43, 0.96)

RR (95% CI)

0.41 (0.29, 0.58)

0.40 (0.08, 1.99)

0.24 (0.12, 0.49)

0.85 (0.51, 1.41)

0.69 (0.50, 0.97)

1.07 (0.53, 2.16)

1.23 (0.64, 2.36)

5.29 (0.26, 107.63)

0.06 (0.01, 0.48)

0.09 (0.01, 1.64)

1.20 (0.80, 1.82)

231/28806

vaccinees

44/1493

2/26

9/983

26/1364

60/8586

16/1145

19/339

2/50

1/540

0/300

52/13980

Events,

Reduced risk Increased risk

1.125 .25 .5 1 2 4 8

By incidence of presumptive pneumococcal pneumonia in controlsPresumptive pneumococcal pneumonia

29

Confidential Draft

Figure 9: Meta-analysis of clinical trials of the effect of pneumococcal polysaccharide vaccines on the risk of presumptive pneumococcal pneumonia. Trials are ordered trial setting.

NOTE: Weights are from random effects analysis

.

.

Overall (I-squared = 74.4%, p = 0.000)

Davis 1987

Smit b 1977

Honkanen 1999

Koivula 1997

Subtotal (I-squared = 58.7%, p = 0.089)

ID

MacLeod 1945

Örtqvist 1998

Developing

Simberkoff 1986

Alfageme 2006

Smit a 1977

Klastersky 1986

Subtotal (I-squared = 31.4%, p = 0.177)

Austrian 1976

Developed

Study

0.64 (0.43, 0.96)

5.29 (0.26, 107.63)

0.06 (0.01, 0.48)

1.20 (0.80, 1.82)

0.85 (0.51, 1.41)

0.28 (0.14, 0.56)

RR (95% CI)

0.69 (0.50, 0.97)

1.23 (0.64, 2.36)

1.07 (0.53, 2.16)

0.09 (0.01, 1.64)

0.24 (0.12, 0.49)

0.40 (0.08, 1.99)

0.92 (0.69, 1.22)

0.41 (0.29, 0.58)

231/28806

2/50

1/540

52/13980

26/1364

54/3016

vaccinees

60/8586

19/339

16/1145

0/300

9/983

2/26

177/25790

44/1493

Events,

358/27758

0/53

16/550

40/12945

33/1473

160/3015

controls

85/8449

16/352

15/1150

5/300

38/985

4/21

198/24743

106/1480

Events,

36 (4, 57)

-429 (-10663, 74)

94 (52, 99)

-20 (-82, 20)

15 (-41, 49)

72 (44, 86)

efficacy (%)

31 (3, 50)

-23 (-136, 36)

-7 (-116, 47)

91 (-64, 99)

76 (51, 88)

60 (-99, 92)

8 (-22, 31)

59 (42, 71)

Vaccine

14

12

23

14

valency

4

23

14

23

6

17

13

Vaccine

0.64 (0.43, 0.96)

5.29 (0.26, 107.63)

0.06 (0.01, 0.48)

1.20 (0.80, 1.82)

0.85 (0.51, 1.41)

0.28 (0.14, 0.56)

RR (95% CI)

0.69 (0.50, 0.97)

1.23 (0.64, 2.36)

1.07 (0.53, 2.16)

0.09 (0.01, 1.64)

0.24 (0.12, 0.49)

0.40 (0.08, 1.99)

0.92 (0.69, 1.22)

0.41 (0.29, 0.58)

231/28806

2/50

1/540

52/13980

26/1364

54/3016

vaccinees

60/8586

19/339

16/1145

0/300

9/983

2/26

177/25790

44/1493

Events,

Reduced risk Increased risk

1.125 .25 .5 1 2 4 8

By trial settingPresumptive pneumococcal pneumonia

30

Confidential Draft

Figure 10: Meta-analysis of clinical trials of the effect of pneumococcal polysaccharide vaccines on the risk of presumptive pneumococcal pneumonia. Trials are ordered by type of study population.

NOTE: Weights are from random effects analysis

.

.

.

Overall (I-squared = 74.4%, p = 0.000)

Chronic illness or elderly

Klastersky 1986

Austrian 1976

Subtotal (I-squared = 46.8%, p = 0.153)

Smit b 1977

Simberkoff 1986

ID

Davis 1987

Alfageme 2006

Örtqvist 1998

COPD or brochogenic carcinoma

Subtotal (I-squared = 77.5%, p = 0.004)

Koivula 1997

MacLeod 1945

Miners or soldiers

Honkanen 1999

Smit a 1977

Study

Subtotal (I-squared = 0.0%, p = 0.736)

0.64 (0.43, 0.96)

0.40 (0.08, 1.99)

0.41 (0.29, 0.58)

0.52 (0.08, 3.58)

0.06 (0.01, 0.48)

1.07 (0.53, 2.16)

RR (95% CI)

5.29 (0.26, 107.63)

0.09 (0.01, 1.64)

1.23 (0.64, 2.36)

0.38 (0.21, 0.68)

0.85 (0.51, 1.41)

0.69 (0.50, 0.97)

1.20 (0.80, 1.82)

0.24 (0.12, 0.49)

1.08 (0.83, 1.41)

231/28806

2/26

44/1493

4/376

1/540

16/1145

vaccinees

2/50

0/300

19/339

114/11602

26/1364

60/8586

52/13980

9/983

Events,

113/16828

358/27758

4/21

106/1480

9/374

16/550

15/1150

controls

0/53

5/300

16/352

245/11464

33/1473

85/8449

40/12945

38/985

Events,

104/15920

36 (4, 57)

60 (-99, 92)

59 (42, 71)

48 (-258, 92)

94 (52, 99)

-7 (-116, 47)

efficacy (%)

-429 (-10663, 74)

91 (-64, 99)

-23 (-136, 36)

62 (32, 79)

15 (-41, 49)

31 (3, 50)

-20 (-82, 20)

76 (51, 88)

Vaccine

-8 (-41, 17)

17

13

12

14

valency

14

23

23

14

4

23

6

Vaccine

0.64 (0.43, 0.96)

0.40 (0.08, 1.99)

0.41 (0.29, 0.58)

0.52 (0.08, 3.58)

0.06 (0.01, 0.48)

1.07 (0.53, 2.16)

RR (95% CI)

5.29 (0.26, 107.63)

0.09 (0.01, 1.64)

1.23 (0.64, 2.36)

0.38 (0.21, 0.68)

0.85 (0.51, 1.41)

0.69 (0.50, 0.97)

1.20 (0.80, 1.82)

0.24 (0.12, 0.49)

1.08 (0.83, 1.41)

231/28806

2/26

44/1493

4/376

1/540

16/1145

vaccinees

2/50

0/300

19/339

114/11602

26/1364

60/8586

52/13980

9/983

Events,

113/16828

Reduced risk Increased risk

1.125 .25 .5 1 2 4 8

By study populationPresumptive pneumococcal pneumonia

31

Confidential Draft

All cause pneumonia Figure 11: Meta-analysis of clinical trials of the effect of pneumococcal polysaccharide vaccines on the risk of pneumonia from all causes. Trials are sorted by start of study.

NOTE: Weights are from random effects analysis

Overall (I-squared = 89.1%, p = 0.000)

Honkanen 1999

Kaufman 1947

Zhogolev a 2003

Koivula 1997

Alfageme 2006

Örtqvist 1998

Davis 1987

Douglas 1984

Smit b 1977

Zhogolev c 2003

Austrian 1976

Simberkoff 1986

Austrian a 1980

Steentoft 2006

Austrian b 1980

ID

Riley 1977

Smit a 1977

Zhogolev b 2003

Gaillat 1985

French 2000

Study

0.74 (0.57, 0.94)

1.16 (0.91, 1.48)

0.44 (0.34, 0.57)

0.73 (0.23, 2.37)

1.14 (0.83, 1.57)

0.96 (0.65, 1.41)

1.15 (0.83, 1.59)

0.61 (0.19, 1.94)

1.33 (0.30, 5.92)

0.68 (0.32, 1.42)

0.21 (0.14, 0.33)

0.45 (0.35, 0.58)

1.37 (0.92, 2.04)

1.22 (1.00, 1.49)

0.71 (0.31, 1.64)

0.98 (0.83, 1.16)

RR (95% CI)

0.76 (0.49, 1.17)

0.63 (0.44, 0.91)

0.30 (0.20, 0.46)

0.23 (0.11, 0.48)

1.89 (1.19, 3.00)

1185/41857

145/13980

85/4750

4/99

73/1364

43/300

63/339

4/50

4/426

9/540

24/990

83/1493

56/1145

154/607

11/37

268/6782

vaccinees

36/5946

37/983

27/422

9/937

50/667

Events,

1544/41660

116/12945

164/4033

8/145

69/1473

45/300

57/352

7/53

3/426

28/1135

90/787

182/1480

41/1150

144/693

5/12

274/6818

controls

48/6012

121/2036

85/405

31/749

26/656

Events,

26 (6, 43)

-16 (-48, 9)

56 (43, 66)

27 (-137, 77)

-14 (-57, 17)

4 (-41, 35)

-15 (-59, 17)

39 (-94, 81)

-33 (-492, 70)

32 (-42, 68)

79 (67, 86)

55 (42, 65)

-37 (-104, 8)

-22 (-49, -0)

29 (-64, 69)

2 (-16, 17)

efficacy (%)

24 (-17, 51)

37 (9, 56)

70 (54, 80)

77 (52, 89)

-89 (-200, -19)

Vaccine

23

3

23

14

23

23

14

14

12

23

13

14

12

23

12

valency

14

6

23

14

23

Vaccine

0.74 (0.57, 0.94)

1.16 (0.91, 1.48)

0.44 (0.34, 0.57)

0.73 (0.23, 2.37)

1.14 (0.83, 1.57)

0.96 (0.65, 1.41)

1.15 (0.83, 1.59)

0.61 (0.19, 1.94)

1.33 (0.30, 5.92)

0.68 (0.32, 1.42)

0.21 (0.14, 0.33)

0.45 (0.35, 0.58)

1.37 (0.92, 2.04)

1.22 (1.00, 1.49)

0.71 (0.31, 1.64)

0.98 (0.83, 1.16)

RR (95% CI)

0.76 (0.49, 1.17)

0.63 (0.44, 0.91)

0.30 (0.20, 0.46)

0.23 (0.11, 0.48)

1.89 (1.19, 3.00)

1185/41857

145/13980

85/4750

4/99

73/1364

43/300

63/339

4/50

4/426

9/540

24/990

83/1493

56/1145

154/607

11/37

268/6782

vaccinees

36/5946

37/983

27/422

9/937

50/667

Events,

Reduced risk Increased risk

1.125 .25 .5 1 2 4 8

Sorted by start of studyAll cause pneumonia

Note - Smit a and b show combined results placebo and meningococcal control groups

32

Confidential Draft

Figure 12: Meta-analysis of clinical trials of the effect of pneumococcal polysaccharide vaccines on the risk of pneumonia from all causes. Trials are ordered by blinding status.

NOTE: Weights are from random effects analysis

.

.

.

Overall (I-squared = 89.1%, p = 0.000)

Zhogolev b 2003

French 2000

Gaillat 1985

Honkanen 1999

Subtotal (I-squared = 87.5%, p = 0.000)

Subtotal (I-squared = 82.5%, p = 0.000)

Koivula 1997

Austrian 1976

Zhogolev a 2003

Steentoft 2006

Simberkoff 1986

Kaufman 1947

Riley 1977

Smit b 1977

Örtqvist 1998

Alfageme 2006

Unblinded

Davis 1987

Austrian a 1980

Austrian b 1980

Subtotal (I-squared = 39.9%, p = 0.126)

Zhogolev c 2003

ID

Smit a 1977

Described as double blind;also controlled

Douglas 1984

Not described as double blind;placebo/vaccine controlled

Study

0.74 (0.57, 0.94)

0.30 (0.20, 0.46)

1.89 (1.19, 3.00)

0.23 (0.11, 0.48)

1.16 (0.91, 1.48)

0.81 (0.58, 1.12)

0.43 (0.27, 0.67)

1.14 (0.83, 1.57)

0.45 (0.35, 0.58)

0.73 (0.23, 2.37)

0.71 (0.31, 1.64)

1.37 (0.92, 2.04)

0.44 (0.34, 0.57)

0.76 (0.49, 1.17)

0.68 (0.32, 1.42)

1.15 (0.83, 1.59)

0.96 (0.65, 1.41)

0.61 (0.19, 1.94)

1.22 (1.00, 1.49)

0.98 (0.83, 1.16)

1.19 (0.97, 1.47)

0.21 (0.14, 0.33)

RR (95% CI)

0.63 (0.44, 0.91)

1.33 (0.30, 5.92)

1185/41857

27/422

50/667

9/937

145/13980

615/25142

203/7535

73/1364

83/1493

4/99

11/37

56/1145

85/4750

36/5946

9/540

63/339

43/300

4/50

154/607

268/6782

367/9180

24/990

vaccinees

37/983

4/426

Events,

1544/41660

85/405

26/656

31/749

116/12945

790/25887

428/6431

69/1473

182/1480

8/145

5/12

41/1150

164/4033

48/6012

28/1135

57/352

45/300

7/53

144/693

274/6818

326/9342

90/787

controls

121/2036

3/426

Events,

26 (6, 43)

70 (54, 80)

-89 (-200, -19)

77 (52, 89)

-16 (-48, 9)

19 (-12, 42)

57 (33, 73)

-14 (-57, 17)

55 (42, 65)

27 (-137, 77)

29 (-64, 69)

-37 (-104, 8)

56 (43, 66)

24 (-17, 51)

32 (-42, 68)

-15 (-59, 17)

4 (-41, 35)

39 (-94, 81)

-22 (-49, -0)

2 (-16, 17)

-19 (-47, 3)

79 (67, 86)

efficacy (%)

37 (9, 56)

-33 (-492, 70)

Vaccine

23

23

14

23

14

13

23

23

14

3

14

12

23

23

14

12

12

23

valency

6

14

Vaccine

0.74 (0.57, 0.94)

0.30 (0.20, 0.46)

1.89 (1.19, 3.00)

0.23 (0.11, 0.48)

1.16 (0.91, 1.48)

0.81 (0.58, 1.12)

0.43 (0.27, 0.67)

1.14 (0.83, 1.57)

0.45 (0.35, 0.58)

0.73 (0.23, 2.37)

0.71 (0.31, 1.64)

1.37 (0.92, 2.04)

0.44 (0.34, 0.57)

0.76 (0.49, 1.17)

0.68 (0.32, 1.42)

1.15 (0.83, 1.59)

0.96 (0.65, 1.41)

0.61 (0.19, 1.94)

1.22 (1.00, 1.49)

0.98 (0.83, 1.16)

1.19 (0.97, 1.47)

0.21 (0.14, 0.33)

RR (95% CI)

0.63 (0.44, 0.91)

1.33 (0.30, 5.92)

1185/41857

27/422

50/667

9/937

145/13980

615/25142

203/7535

73/1364

83/1493

4/99

11/37

56/1145

85/4750

36/5946

9/540

63/339

43/300

4/50

154/607

268/6782

367/9180

24/990

vaccinees

37/983

4/426

Events,

Reduced risk Increased risk

1.125 .25 .5 1 2 4 8

By blindingAll cause pneumonia

33

Confidential Draft

Figure 13: Meta-analysis of clinical trials of the effect of pneumococcal polysaccharide vaccines on the risk of pneumonia from all causes. Trials are ordered by by concealment of allocation.

NOTE: Weights are from random effects analysis

.

.

Overall (I-squared = 89.1%, p = 0.000)

Honkanen 1999

Subtotal (I-squared = 9.2%, p = 0.357)

Örtqvist 1998

Smit a 1977

Kaufman 1947

Austrian b 1980

Study

ID

Simberkoff 1986

Austrian a 1980

Adequate

Subtotal (I-squared = 91.5%, p = 0.000)

Davis 1987

Gaillat 1985

Zhogolev c 2003

Austrian 1976

Riley 1977

Koivula 1997

Alfageme 2006

Smit b 1977

Zhogolev a 2003

Douglas 1984

French 2000

Zhogolev b 2003

Unclear

Steentoft 2006

0.74 (0.57, 0.94)

1.16 (0.91, 1.48)

1.02 (0.88, 1.18)

1.15 (0.83, 1.59)

0.63 (0.44, 0.91)

0.44 (0.34, 0.57)

0.98 (0.83, 1.16)

RR (95% CI)

1.37 (0.92, 2.04)

1.22 (1.00, 1.49)

0.64 (0.46, 0.90)

0.61 (0.19, 1.94)

0.23 (0.11, 0.48)

0.21 (0.14, 0.33)

0.45 (0.35, 0.58)

0.76 (0.49, 1.17)

1.14 (0.83, 1.57)

0.96 (0.65, 1.41)

0.68 (0.32, 1.42)

0.73 (0.23, 2.37)

1.33 (0.30, 5.92)

1.89 (1.19, 3.00)

0.30 (0.20, 0.46)

0.71 (0.31, 1.64)

1185/41857

145/13980

438/14675

63/339

37/983

85/4750

268/6782

Events,

vaccinees

56/1145

154/607

747/27182

4/50

9/937

24/990

83/1493

36/5946

73/1364

43/300

9/540

4/99

4/426

50/667

27/422

11/37

1544/41660

116/12945

428/14770

57/352

121/2036

164/4033

274/6818

Events,

controls

41/1150

144/693

1116/26890

7/53

31/749

90/787

182/1480

48/6012

69/1473

45/300

28/1135

8/145

3/426

26/656

85/405

5/12

26 (6, 43)

-16 (-48, 9)

-2 (-18, 12)

-15 (-59, 17)

37 (9, 56)

56 (43, 66)

2 (-16, 17)

Vaccine

efficacy (%)

-37 (-104, 8)

-22 (-49, -0)

36 (10, 54)

39 (-94, 81)

77 (52, 89)

79 (67, 86)

55 (42, 65)

24 (-17, 51)

-14 (-57, 17)

4 (-41, 35)

32 (-42, 68)

27 (-137, 77)

-33 (-492, 70)

-89 (-200, -19)

70 (54, 80)

29 (-64, 69)

23

23

6

3

12

Vaccine

valency

14

12

14

14

23

13

14

14

23

12

23

14

23

23

23

0.74 (0.57, 0.94)

1.16 (0.91, 1.48)

1.02 (0.88, 1.18)

1.15 (0.83, 1.59)

0.63 (0.44, 0.91)

0.44 (0.34, 0.57)

0.98 (0.83, 1.16)

RR (95% CI)

1.37 (0.92, 2.04)

1.22 (1.00, 1.49)

0.64 (0.46, 0.90)

0.61 (0.19, 1.94)

0.23 (0.11, 0.48)

0.21 (0.14, 0.33)

0.45 (0.35, 0.58)

0.76 (0.49, 1.17)

1.14 (0.83, 1.57)

0.96 (0.65, 1.41)

0.68 (0.32, 1.42)

0.73 (0.23, 2.37)

1.33 (0.30, 5.92)

1.89 (1.19, 3.00)

0.30 (0.20, 0.46)

0.71 (0.31, 1.64)

1185/41857

145/13980

438/14675

63/339

37/983

85/4750

268/6782

Events,

vaccinees

56/1145

154/607

747/27182

4/50

9/937

24/990

83/1493

36/5946

73/1364

43/300

9/540

4/99

4/426

50/667

27/422

11/37

Reduced risk Increased risk

1.125 .25 .5 1 2 4 8

By concealment of allocationAll cause pneumonia

34

Confidential Draft

Figure 14: Meta-analysis of clinical trials of the effect of pneumococcal polysaccharide vaccines on the risk of pneumonia from all causes. Trials are ordered by vaccine valency.

NOTE: Weights are from random effects analysis

.

.

.

Overall (I-squared = 89.1%, p = 0.000)

Study

Smit b 1977

Honkanen 1999

ID

Douglas 1984

Riley 1977

Smit a 1977

French 2000

Zhogolev c 2003

Austrian b 1980

Simberkoff 1986

other valencies

Davis 1987

Örtqvist 1998

Koivula 1997

Steentoft 2006

Zhogolev b 2003

14-valent

Austrian 1976

Alfageme 2006

23-valent

Subtotal (I-squared = 92.4%, p = 0.000)

Austrian a 1980

Subtotal (I-squared = 91.7%, p = 0.000)

Zhogolev a 2003

Gaillat 1985

Kaufman 1947

Subtotal (I-squared = 75.8%, p = 0.001)

0.74 (0.57, 0.94)

0.68 (0.32, 1.42)

1.16 (0.91, 1.48)

RR (95% CI)

1.33 (0.30, 5.92)

0.76 (0.49, 1.17)

0.63 (0.44, 0.91)

1.89 (1.19, 3.00)

0.21 (0.14, 0.33)

0.98 (0.83, 1.16)

1.37 (0.92, 2.04)

0.61 (0.19, 1.94)

1.15 (0.83, 1.59)

1.14 (0.83, 1.57)

0.71 (0.31, 1.64)

0.30 (0.20, 0.46)

0.45 (0.35, 0.58)

0.96 (0.65, 1.41)

0.69 (0.47, 1.02)

1.22 (1.00, 1.49)

0.73 (0.44, 1.24)

0.73 (0.23, 2.37)

0.23 (0.11, 0.48)

0.44 (0.34, 0.57)

0.80 (0.50, 1.29)

1185/41857

Events,

9/540

145/13980

vaccinees

4/426

36/5946

37/983

50/667

24/990

268/6782

56/1145

4/50

63/339

73/1364

11/37

27/422

83/1493

43/300

636/15155

154/607

367/16834

4/99

9/937

85/4750

182/9868

1544/41660

Events,

28/1135

116/12945

controls

3/426

48/6012

121/2036

26/656

90/787

274/6818

41/1150

7/53

57/352

69/1473

5/12

85/405

182/1480

45/300

913/16195

144/693

432/15602

8/145

31/749

164/4033

199/9863

26 (6, 43)

Vaccine

32 (-42, 68)

-16 (-48, 9)

efficacy (%)

-33 (-492, 70)

24 (-17, 51)

37 (9, 56)

-89 (-200, -19)

79 (67, 86)

2 (-16, 17)

-37 (-104, 8)

39 (-94, 81)

-15 (-59, 17)

-14 (-57, 17)

29 (-64, 69)

70 (54, 80)

55 (42, 65)

4 (-41, 35)

31 (-2, 53)

-22 (-49, -0)

27 (-24, 56)

27 (-137, 77)

77 (52, 89)

56 (43, 66)

20 (-29, 50)

Vaccine

12

23

valency

14

14

6

23

23

12

14

14

23

14

23

23

13

23

12

23

14

3

0.74 (0.57, 0.94)

0.68 (0.32, 1.42)

1.16 (0.91, 1.48)

RR (95% CI)

1.33 (0.30, 5.92)

0.76 (0.49, 1.17)

0.63 (0.44, 0.91)

1.89 (1.19, 3.00)

0.21 (0.14, 0.33)

0.98 (0.83, 1.16)

1.37 (0.92, 2.04)

0.61 (0.19, 1.94)

1.15 (0.83, 1.59)

1.14 (0.83, 1.57)

0.71 (0.31, 1.64)

0.30 (0.20, 0.46)

0.45 (0.35, 0.58)

0.96 (0.65, 1.41)

0.69 (0.47, 1.02)

1.22 (1.00, 1.49)

0.73 (0.44, 1.24)

0.73 (0.23, 2.37)

0.23 (0.11, 0.48)

0.44 (0.34, 0.57)

0.80 (0.50, 1.29)

1185/41857

Events,

9/540

145/13980

vaccinees

4/426

36/5946

37/983

50/667

24/990

268/6782

56/1145

4/50

63/339

73/1364

11/37

27/422

83/1493

43/300

636/15155

154/607

367/16834

4/99

9/937

85/4750

182/9868

Reduced risk Increased risk

1.125 .25 .5 1 2 4 8

By vaccine valencyAll cause pneumonia

35

Confidential Draft

Figure 15: Meta-analysis of clinical trials of the effect of pneumococcal polysaccharide vaccines on the risk of pneumonia from all causes. Trials are ordered by incidence of pneumonia in controls.

NOTE: Weights are from random effects analysis

Overall (I-squared = 89.1%, p = 0.000)

Riley 1977

Kaufman 1947

Honkanen 1999

Steentoft 2006

Austrian a 1980

ID

Austrian 1976

Zhogolev a 2003

Zhogolev b 2003

Davis 1987

Zhogolev c 2003

Smit a 1977

Örtqvist 1998

Alfageme 2006

Smit b 1977

Koivula 1997

Study

French 2000

Simberkoff 1986

Douglas 1984

Gaillat 1985

Austrian b 1980

0.74 (0.57, 0.94)

0.76 (0.49, 1.17)

0.44 (0.34, 0.57)

1.16 (0.91, 1.48)

0.71 (0.31, 1.64)

1.22 (1.00, 1.49)

RR (95% CI)

0.45 (0.35, 0.58)

0.73 (0.23, 2.37)

0.30 (0.20, 0.46)

0.61 (0.19, 1.94)

0.21 (0.14, 0.33)

0.63 (0.44, 0.91)

1.15 (0.83, 1.59)

0.96 (0.65, 1.41)

0.68 (0.32, 1.42)

1.14 (0.83, 1.57)

1.89 (1.19, 3.00)

1.37 (0.92, 2.04)

1.33 (0.30, 5.92)

0.23 (0.11, 0.48)

0.98 (0.83, 1.16)

1185/41857

36/5946

85/4750

145/13980

11/37

154/607

vaccinees

83/1493

4/99

27/422

4/50

24/990

37/983

63/339

43/300

9/540

73/1364

Events,

50/667

56/1145

4/426

9/937

268/6782

1544/41660

48/6012

164/4033

116/12945

5/12

144/693

controls

182/1480

8/145

85/405

7/53

90/787

121/2036

57/352

45/300

28/1135

69/1473

Events,

26/656

41/1150

3/426

31/749

274/6818

26 (6, 43)

24 (-17, 51)

56 (43, 66)

-16 (-48, 9)

29 (-64, 69)

-22 (-49, -0)

efficacy (%)

55 (42, 65)

27 (-137, 77)

70 (54, 80)

39 (-94, 81)

79 (67, 86)

37 (9, 56)

-15 (-59, 17)

4 (-41, 35)

32 (-42, 68)

-14 (-57, 17)

Vaccine

-89 (-200, -19)

-37 (-104, 8)

-33 (-492, 70)

77 (52, 89)

2 (-16, 17)

14

3

23

23

12

valency

13

23

23

14

23

6

23

23

12

14

Vaccine

23

14

14

14

12

0.74 (0.57, 0.94)

0.76 (0.49, 1.17)

0.44 (0.34, 0.57)

1.16 (0.91, 1.48)

0.71 (0.31, 1.64)

1.22 (1.00, 1.49)

RR (95% CI)

0.45 (0.35, 0.58)

0.73 (0.23, 2.37)

0.30 (0.20, 0.46)

0.61 (0.19, 1.94)

0.21 (0.14, 0.33)

0.63 (0.44, 0.91)

1.15 (0.83, 1.59)

0.96 (0.65, 1.41)

0.68 (0.32, 1.42)

1.14 (0.83, 1.57)

1.89 (1.19, 3.00)

1.37 (0.92, 2.04)

1.33 (0.30, 5.92)

0.23 (0.11, 0.48)

0.98 (0.83, 1.16)

1185/41857

36/5946

85/4750

145/13980

11/37

154/607

vaccinees

83/1493

4/99

27/422

4/50

24/990

37/983

63/339

43/300

9/540

73/1364

Events,

50/667

56/1145

4/426

9/937

268/6782

Reduced risk Increased risk

1.125 .25 .5 1 2 4 8

Sorted by incidence of pneumonia in controls over study periodAll cause pneumonia

36

Confidential Draft

Figure 16: Meta-analysis of clinical trials of the effect of pneumococcal polysaccharide vaccines on the risk of pneumonia from all causes. Trials are ordered by trial setting.

NOTE: Weights are from random effects analysis

.

.

.

Overall (I-squared = 89.1%, p = 0.000)

Subtotal (I-squared = 86.3%, p = 0.000)

Örtqvist 1998

Subtotal (I-squared = 82.9%, p = 0.000)

Kaufman 1947

Smit a 1977

Zhogolev b 2003

Developing

Zhogolev c 2003

French 2000

Douglas 1984

Simberkoff 1986

Undefined

Zhogolev a 2003

Subtotal (I-squared = 53.2%, p = 0.118)

Smit b 1977

ID

Steentoft 2006

Study

Honkanen 1999

Gaillat 1985

Koivula 1997

Austrian 1976

Austrian a 1980

Riley 1977

Austrian b 1980

Davis 1987

Developed

Alfageme 2006

0.74 (0.57, 0.94)

0.76 (0.47, 1.25)

1.15 (0.83, 1.59)

0.89 (0.70, 1.14)

0.44 (0.34, 0.57)

0.63 (0.44, 0.91)

0.30 (0.20, 0.46)

0.21 (0.14, 0.33)

1.89 (1.19, 3.00)

1.33 (0.30, 5.92)

1.37 (0.92, 2.04)

0.73 (0.23, 2.37)

0.29 (0.18, 0.47)

0.68 (0.32, 1.42)

RR (95% CI)

0.71 (0.31, 1.64)

1.16 (0.91, 1.48)

0.23 (0.11, 0.48)

1.14 (0.83, 1.57)

0.45 (0.35, 0.58)

1.22 (1.00, 1.49)

0.76 (0.49, 1.17)

0.98 (0.83, 1.16)

0.61 (0.19, 1.94)

0.96 (0.65, 1.41)

1185/41857

215/9629

63/339

915/30717

85/4750

37/983

27/422

24/990

50/667

4/426

56/1145

4/99

55/1511

9/540

vaccinees

11/37

Events,

145/13980

9/937

73/1364

83/1493

154/607

36/5946

268/6782

4/50

43/300

1544/41660

405/11319

57/352

956/29004

164/4033

121/2036

85/405

90/787

26/656

3/426

41/1150

8/145

183/1337

28/1135

controls

5/12

Events,

116/12945

31/749

69/1473

182/1480

144/693

48/6012

274/6818

7/53

45/300

26 (6, 43)

24 (-25, 53)

-15 (-59, 17)

11 (-14, 30)

56 (43, 66)

37 (9, 56)

70 (54, 80)

79 (67, 86)

-89 (-200, -19)

-33 (-492, 70)

-37 (-104, 8)

27 (-137, 77)

71 (53, 82)

32 (-42, 68)

efficacy (%)

29 (-64, 69)

Vaccine

-16 (-48, 9)

77 (52, 89)

-14 (-57, 17)

55 (42, 65)

-22 (-49, -0)

24 (-17, 51)

2 (-16, 17)

39 (-94, 81)

4 (-41, 35)

23

3

6

23

23

23

14

14

23

12

valency

23

Vaccine

23

14

14

13

12

14

12

14

23

0.74 (0.57, 0.94)

0.76 (0.47, 1.25)

1.15 (0.83, 1.59)

0.89 (0.70, 1.14)

0.44 (0.34, 0.57)

0.63 (0.44, 0.91)

0.30 (0.20, 0.46)

0.21 (0.14, 0.33)

1.89 (1.19, 3.00)

1.33 (0.30, 5.92)

1.37 (0.92, 2.04)

0.73 (0.23, 2.37)

0.29 (0.18, 0.47)

0.68 (0.32, 1.42)

RR (95% CI)

0.71 (0.31, 1.64)

1.16 (0.91, 1.48)

0.23 (0.11, 0.48)

1.14 (0.83, 1.57)

0.45 (0.35, 0.58)

1.22 (1.00, 1.49)

0.76 (0.49, 1.17)

0.98 (0.83, 1.16)

0.61 (0.19, 1.94)

0.96 (0.65, 1.41)

1185/41857

215/9629

63/339

915/30717

85/4750

37/983

27/422

24/990

50/667

4/426

56/1145

4/99

55/1511

9/540

vaccinees

11/37

Events,

145/13980

9/937

73/1364

83/1493

154/607

36/5946

268/6782

4/50

43/300

Reduced risk Increased risk

1.125 .25 .5 1 2 4 8

By trial settingAll cause pneumonia

37

Confidential Draft

Figure 17: Meta-analysis of clinical trials of the effect of pneumococcal polysaccharide vaccines on the risk of pneumonia from all causes. Trials are ordered by type of study population.

NOTE: Weights are from random effects analysis

.

.

.

.

.

.

Overall (I-squared = 89.1%, p = 0.000)

Honkanen 1999

Miners or soldiers

Simberkoff 1986

Zhogolev c 2003

>10yo in Papua New Guinea

HIV positive

Kaufman 1947

COPD or brochogenic carcinoma

Subtotal (I-squared = 88.9%, p = 0.000)

French 2000

Austrian b 1980

Davis 1987

Children

Austrian 1976

Subtotal (I-squared = .%, p = .)

Gaillat 1985

Zhogolev a 2003

Douglas 1984

Smit a 1977

Austrian a 1980

Örtqvist 1998

Riley 1977

Subtotal (I-squared = .%, p = .)

Alfageme 2006

Chronic illness or elderly

Subtotal (I-squared = 0.0%, p = 0.665)

Steentoft 2006

Subtotal (I-squared = 74.2%, p = 0.002)

Zhogolev b 2003

ID

Smit b 1977

Study

Subtotal (I-squared = .%, p = .)

Koivula 1997

0.74 (0.57, 0.94)

1.16 (0.91, 1.48)

1.37 (0.92, 2.04)

0.21 (0.14, 0.33)

0.44 (0.34, 0.57)

0.90 (0.67, 1.20)

1.89 (1.19, 3.00)

0.98 (0.83, 1.16)

0.61 (0.19, 1.94)

0.45 (0.35, 0.58)

0.76 (0.49, 1.17)

0.23 (0.11, 0.48)

0.73 (0.23, 2.37)

1.33 (0.30, 5.92)

0.63 (0.44, 0.91)

1.22 (1.00, 1.49)

1.15 (0.83, 1.59)

0.76 (0.49, 1.17)

1.33 (0.30, 5.92)

0.96 (0.65, 1.41)0.88 (0.63, 1.23)

0.71 (0.31, 1.64)

0.42 (0.29, 0.61)

0.30 (0.20, 0.46)

RR (95% CI)

0.68 (0.32, 1.42)

1.89 (1.19, 3.00)

1.14 (0.83, 1.57)

1185/41857

145/13980

56/1145

24/990

85/4750

853/29904

50/667

268/6782

4/50

83/1493

36/5946

9/937

4/99

4/426

37/983

154/607

63/339

36/5946

4/426

43/30058/387

11/37

184/4527

27/422

vaccinees

9/540

Events,

50/667

73/1364

1544/41660

116/12945

41/1150

90/787

164/4033

896/28213

26/656

274/6818

7/53

182/1480

48/6012

31/749

8/145

3/426

121/2036

144/693

57/352

48/6012

3/426

45/30057/365

5/12

514/5988

85/405

controls

28/1135

Events,

26/656

69/1473

26 (6, 43)

-16 (-48, 9)

-37 (-104, 8)

79 (67, 86)

56 (43, 66)

10 (-20, 33)

-89 (-200, -19)

2 (-16, 17)

39 (-94, 81)

55 (42, 65)

24 (-17, 51)

77 (52, 89)

27 (-137, 77)

-33 (-492, 70)

37 (9, 56)

-22 (-49, -0)

-15 (-59, 17)

24 (-17, 51)

-33 (-492, 70)

4 (-41, 35)12 (-23, 37)

29 (-64, 69)

58 (39, 71)

70 (54, 80)

efficacy (%)

32 (-42, 68)

Vaccine

-89 (-200, -19)

-14 (-57, 17)

23

14

23

3

23

12

14

13

14

23

14

6

12

23

14

23

23

23

valency

12

Vaccine

14

0.74 (0.57, 0.94)

1.16 (0.91, 1.48)

1.37 (0.92, 2.04)

0.21 (0.14, 0.33)

0.44 (0.34, 0.57)

0.90 (0.67, 1.20)

1.89 (1.19, 3.00)

0.98 (0.83, 1.16)

0.61 (0.19, 1.94)

0.45 (0.35, 0.58)

0.76 (0.49, 1.17)

0.23 (0.11, 0.48)

0.73 (0.23, 2.37)

1.33 (0.30, 5.92)

0.63 (0.44, 0.91)

1.22 (1.00, 1.49)

1.15 (0.83, 1.59)

0.76 (0.49, 1.17)

1.33 (0.30, 5.92)

0.96 (0.65, 1.41)0.88 (0.63, 1.23)

0.71 (0.31, 1.64)

0.42 (0.29, 0.61)

0.30 (0.20, 0.46)

RR (95% CI)

0.68 (0.32, 1.42)

1.89 (1.19, 3.00)

1.14 (0.83, 1.57)

1185/41857

145/13980

56/1145

24/990

85/4750

853/29904

50/667

268/6782

4/50

83/1493

36/5946

9/937

4/99

4/426

37/983

154/607

63/339

36/5946

4/426

43/30058/387

11/37

184/4527

27/422

vaccinees

9/540

Events,

50/667

73/1364

Reduced risk Increased risk

1.125 .25 .5 1 2 4 8

By study populationAll cause pneumonia

38

Confidential Draft

Figure 18: Meta-analysis of clinical trials of the effect of pneumococcal polysaccharide vaccines on the risk of pneumonia from all causes. Comparison of results from analyses based on cases (first pneumonias only) and pneumonia episodes in six trials reporting both outcomes.

NOTE: Weights are from random effects analysis

Overall (I-squared = 27.5%, p = 0.228)

Austrian a 1980

Koivula 1997

Davis 1987

French 2000

Study

ID

Simberkoff 1986

Alfageme 2006

1.22 (1.03, 1.46)

1.22 (1.00, 1.49)

1.14 (0.83, 1.57)

0.61 (0.19, 1.94)

1.89 (1.19, 3.00)

RR (95% CI)

1.37 (0.92, 2.04)

0.96 (0.65, 1.41)

380/4133

154/607

73/1364

4/50

50/667

Events,

vaccinees

56/1145

43/300

332/4325

144/693

69/1473

7/53

26/656

Events,

controls

41/1150

45/300

-22 (-46, -3)

-22 (-49, -0)

-14 (-57, 17)

39 (-94, 81)

-89 (-200, -19)

Vaccine

efficacy (%)

-37 (-104, 8)

4 (-41, 35)

12

14

14

23

Vaccine

valency

14

23

1.22 (1.03, 1.46)

1.22 (1.00, 1.49)

1.14 (0.83, 1.57)

0.61 (0.19, 1.94)

1.89 (1.19, 3.00)

RR (95% CI)

1.37 (0.92, 2.04)

0.96 (0.65, 1.41)

380/4133

154/607

73/1364

4/50

50/667

Events,

vaccinees

56/1145

43/300

Reduced risk Increased risk

1.25 .5 1 2 4

Sorted by publication yearAll cause pneumonia episodes

NOTE: Weights are from random effects analysis

Overall (I-squared = 17.9%, p = 0.297)

Davis 1987

Simberkoff 1986

Alfageme 2006

Koivula 1997

ID

French 2000

Austrian a 1980

Study

1.16 (0.97, 1.40)

0.45 (0.12, 1.66)

1.27 (0.84, 1.93)

1.03 (0.67, 1.57)

1.13 (0.81, 1.58)

RR (95% CI)

1.87 (1.12, 3.14)

1.08 (0.84, 1.41)

290/4133

3/50

48/1145

38/300

67/1364

vaccinees

40/667

94/607

Events,

266/4325

7/53

38/1150

37/300

64/1473

controls

21/656

99/693

Events,

-16 (-40, 3)

55 (-66, 88)

-27 (-93, 16)

-3 (-57, 33)

-13 (-58, 19)

efficacy (%)

-87 (-214, -12)

-8 (-41, 16)

Vaccine

14

14

23

14

valency

23

12

Vaccine

1.16 (0.97, 1.40)

0.45 (0.12, 1.66)

1.27 (0.84, 1.93)

1.03 (0.67, 1.57)

1.13 (0.81, 1.58)

RR (95% CI)

1.87 (1.12, 3.14)

1.08 (0.84, 1.41)

290/4133

3/50

48/1145

38/300

67/1364

vaccinees

40/667

94/607

Events,

Reduced risk Increased risk

1.25 .5 1 2 4

Sorted by publication yearAll cause pneumonia cases

39

Confidential Draft

Bronchitis Figure 19: Meta-analysis of clinical trials of the effect of pneumococcal polysaccharide vaccines on the risk of bronchitis from any cause.

NOTE: Weights are from random effects analysis

Overall (I-squared = 53.7%, p = 0.090)

Smit a 1977

Simberkoff 1986

ID

Austrian b 1980

Smit b 1977

Study

0.92 (0.76, 1.12)

0.91 (0.71, 1.16)

0.82 (0.58, 1.15)

RR (95% CI)

1.07 (0.96, 1.19)

0.58 (0.32, 1.07)

791/9450

84/983

56/1145

vaccinees

638/6782

13/540

Events,

907/11139

191/2036

69/1150

controls

600/6818

47/1135

Events,

8 (-12, 24)

9 (-16, 29)

18 (-15, 42)

efficacy (%)

-7 (-19, 4)

42 (-7, 68)

Vaccine

6

14

valency

12

12

Vaccine

0.92 (0.76, 1.12)

0.91 (0.71, 1.16)

0.82 (0.58, 1.15)

RR (95% CI)

1.07 (0.96, 1.19)

0.58 (0.32, 1.07)

791/9450

84/983

56/1145

vaccinees

638/6782

13/540

Events,

Reduced risk Increased risk 1.5 .67 1 1.5 2

Bronchitis

Note - Smit a and b show reported results placebo and meningococcal control groups

40

Confidential Draft

Mortality from all causes Figure 20: Meta-analysis of clinical trials of the effect of pneumococcal polysaccharide vaccines on the risk of death from all causes. Trials are sorted by start of study.

NOTE: Weights are from random effects analysis

Overall (I-squared = 45.8%, p = 0.031)

Study

Davis 1987

Riley 1977

Simberkoff 1986

Örtqvist 1998

Austrian b 1980

Leech 1987

French 2000

ID

Riley 1986

Alfageme 2006

Riley 1981

Kaufman 1947

Austrian a 1980

Gaillat 1985

Koivula 1997

0.94 (0.84, 1.05)

1.14 (0.60, 2.18)

0.79 (0.63, 0.99)

1.24 (1.03, 1.49)

1.08 (0.65, 1.77)

0.96 (0.64, 1.45)

0.58 (0.22, 1.49)

1.00 (0.84, 1.19)

RR (95% CI)

0.71 (0.50, 1.01)

0.98 (0.71, 1.37)

0.65 (0.30, 1.39)

0.53 (0.34, 0.83)

0.93 (0.60, 1.43)

1.06 (0.89, 1.26)

0.99 (0.80, 1.22)

1192/25825

Events,

14/50

133/5946

211/1145

29/339

45/6782

6/92

185/667

vaccinees

51/2445

57/300

10/401

32/4750

35/607

232/937

152/1364

1204/25273

Events,

13/53

170/6012

171/1150

28/352

47/6818

11/97

182/656

controls

71/2417

58/300

18/470

51/4033

43/693

175/749

166/1473

6 (-5, 16)

Vaccine

-14 (-118, 40)

21 (1, 37)

-24 (-49, -3)

-8 (-77, 35)

4 (-45, 36)

42 (-49, 78)

0 (-19, 16)

efficacy (%)

29 (-1, 50)

2 (-37, 29)

35 (-39, 70)

47 (17, 66)

7 (-43, 40)

-6 (-26, 11)

1 (-22, 20)

Vaccine

14

14

14

23

12

14

23

valency

23

14

3

12

14

14

0.94 (0.84, 1.05)

1.14 (0.60, 2.18)

0.79 (0.63, 0.99)

1.24 (1.03, 1.49)

1.08 (0.65, 1.77)

0.96 (0.64, 1.45)

0.58 (0.22, 1.49)

1.00 (0.84, 1.19)

RR (95% CI)

0.71 (0.50, 1.01)

0.98 (0.71, 1.37)

0.65 (0.30, 1.39)

0.53 (0.34, 0.83)

0.93 (0.60, 1.43)

1.06 (0.89, 1.26)

0.99 (0.80, 1.22)

1192/25825

Events,

14/50

133/5946

211/1145

29/339

45/6782

6/92

185/667

vaccinees

51/2445

57/300

10/401

32/4750

35/607

232/937

152/1364

Reduced risk Increased risk

1.125 .25 .5 1 2 4 8

Sorted by start of studyAll cause mortality

41

Confidential Draft

Figure 21: Meta-analysis of clinical trials of the effect of pneumococcal polysaccharide vaccines on the risk of death from all causes. Trials are ordered by blinding status.

NOTE: Weights are from random effects analysis

.

.

.

Overall (I-squared = 45.8%, p = 0.031)

Leech 1987

Gaillat 1985

French 2000

Reported to be double blind, placebo-controlled

Not reported to be double blind, placebo-controlled

Koivula 1997

Subtotal (I-squared = 0.0%, p = 0.908)

Austrian b 1980

Riley 1986

Subtotal (I-squared = 49.2%, p = 0.046)

Subtotal (I-squared = 75.7%, p = 0.016)

Örtqvist 1998

Riley 1977

Riley 1981

Simberkoff 1986

Unblinded, not placebo-controlled

Davis 1987

Alfageme 2006

ID

Study

Kaufman 1947

Austrian a 1980

0.94 (0.84, 1.05)

0.58 (0.22, 1.49)

1.06 (0.89, 1.26)

1.00 (0.84, 1.19)

0.99 (0.80, 1.22)

0.98 (0.82, 1.18)

0.96 (0.64, 1.45)

0.71 (0.50, 1.01)

0.94 (0.80, 1.10)

0.86 (0.60, 1.23)

1.08 (0.65, 1.77)

0.79 (0.63, 0.99)

0.65 (0.30, 1.39)

1.24 (1.03, 1.49)

1.14 (0.60, 2.18)

0.98 (0.71, 1.37)

RR (95% CI)

0.53 (0.34, 0.83)

0.93 (0.60, 1.43)

1192/25825

6/92

232/937

185/667

152/1364

197/8146

45/6782

51/2445

674/11692

321/5987

29/339

133/5946

10/401

211/1145

14/50

57/300

vaccinees

Events,

32/4750

35/607

1204/25273

11/97

175/749

182/656

166/1473

213/8291

47/6818

71/2417

707/11900

284/5082

28/352

170/6012

18/470

171/1150

13/53

58/300

controls

Events,

51/4033

43/693

6 (-5, 16)

42 (-49, 78)

-6 (-26, 11)

0 (-19, 16)

1 (-22, 20)

2 (-18, 18)

4 (-45, 36)

29 (-1, 50)

6 (-10, 20)

14 (-23, 40)

-8 (-77, 35)

21 (1, 37)

35 (-39, 70)

-24 (-49, -3)

-14 (-118, 40)

2 (-37, 29)

efficacy (%)

Vaccine

47 (17, 66)

7 (-43, 40)

14

14

23

14

12

23

14

14

14

14

23

valency

Vaccine

3

12

0.94 (0.84, 1.05)

0.58 (0.22, 1.49)

1.06 (0.89, 1.26)

1.00 (0.84, 1.19)

0.99 (0.80, 1.22)

0.98 (0.82, 1.18)

0.96 (0.64, 1.45)

0.71 (0.50, 1.01)

0.94 (0.80, 1.10)

0.86 (0.60, 1.23)

1.08 (0.65, 1.77)

0.79 (0.63, 0.99)

0.65 (0.30, 1.39)

1.24 (1.03, 1.49)

1.14 (0.60, 2.18)

0.98 (0.71, 1.37)

RR (95% CI)

0.53 (0.34, 0.83)

0.93 (0.60, 1.43)

1192/25825

6/92

232/937

185/667

152/1364

197/8146

45/6782

51/2445

674/11692

321/5987

29/339

133/5946

10/401

211/1145

14/50

57/300

vaccinees

Events,

32/4750

35/607

Reduced risk Increased risk

1.125 .25 .5 1 2 4 8

By blindingAll cause mortality

42

Confidential Draft

Figure 22: Meta-analysis of clinical trials of the effect of pneumococcal polysaccharide vaccines on the risk of death from all causes. Trials are ordered by concealment of allocation.

NOTE: Weights are from random effects analysis

.

.

Overall (I-squared = 45.8%, p = 0.031)

Subtotal (I-squared = 37.2%, p = 0.121)

Leech 1987

Adequate

Subtotal (I-squared = 62.8%, p = 0.030)

Koivula 1997

Riley 1981

Austrian b 1980

French 2000

Riley 1986

Austrian a 1980

Alfageme 2006

Gaillat 1985

Davis 1987

ID

Riley 1977

Örtqvist 1998

Study

Kaufman 1947

Unclear

Simberkoff 1986

0.94 (0.84, 1.05)

0.93 (0.81, 1.05)

0.58 (0.22, 1.49)

0.97 (0.75, 1.24)

0.99 (0.80, 1.22)

0.65 (0.30, 1.39)

0.96 (0.64, 1.45)

1.00 (0.84, 1.19)

0.71 (0.50, 1.01)

0.93 (0.60, 1.43)

0.98 (0.71, 1.37)

1.06 (0.89, 1.26)

1.14 (0.60, 2.18)

RR (95% CI)

0.79 (0.63, 0.99)

1.08 (0.65, 1.77)

0.53 (0.34, 0.83)

1.24 (1.03, 1.49)

1192/25825

764/11212

6/92

428/14613

152/1364

10/401

45/6782

185/667

51/2445

35/607

57/300

232/937

14/50

vaccinees

133/5946

29/339

Events,

32/4750

211/1145

1204/25273

770/10471

11/97

434/14802

166/1473

18/470

47/6818

182/656

71/2417

43/693

58/300

175/749

13/53

controls

170/6012

28/352

Events,

51/4033

171/1150

6 (-5, 16)

7 (-5, 19)

42 (-49, 78)

3 (-24, 25)

1 (-22, 20)

35 (-39, 70)

4 (-45, 36)

0 (-19, 16)

29 (-1, 50)

7 (-43, 40)

2 (-37, 29)

-6 (-26, 11)

-14 (-118, 40)

efficacy (%)

21 (1, 37)

-8 (-77, 35)

Vaccine

47 (17, 66)

-24 (-49, -3)

14

14

14

12

23

12

23

14

14

valency

14

23

Vaccine

3

14

0.94 (0.84, 1.05)

0.93 (0.81, 1.05)

0.58 (0.22, 1.49)

0.97 (0.75, 1.24)

0.99 (0.80, 1.22)

0.65 (0.30, 1.39)

0.96 (0.64, 1.45)

1.00 (0.84, 1.19)

0.71 (0.50, 1.01)

0.93 (0.60, 1.43)

0.98 (0.71, 1.37)

1.06 (0.89, 1.26)

1.14 (0.60, 2.18)

RR (95% CI)

0.79 (0.63, 0.99)

1.08 (0.65, 1.77)

0.53 (0.34, 0.83)

1.24 (1.03, 1.49)

1192/25825

764/11212

6/92

428/14613

152/1364

10/401

45/6782

185/667

51/2445

35/607

57/300

232/937

14/50

vaccinees

133/5946

29/339

Events,

32/4750

211/1145

Reduced risk Increased risk

1.125 .25 .5 1 2 4 8

By concealment of allocationAll cause mortality

43

Confidential Draft

Figure 23: Meta-analysis of clinical trials of the effect of pneumococcal polysaccharide vaccines on the risk of death from all causes. Trials are ordered by the valency of the vaccine used.

NOTE: Weights are from random effects analysis

.

.

.

Overall (I-squared = 45.8%, p = 0.031)

Austrian b 1980

ID

Austrian a 1980

Subtotal (I-squared = 55.8%, p = 0.104)

Subtotal (I-squared = 56.8%, p = 0.023)

other valencies

14-valent

Kaufman 1947

Riley 1986

Koivula 1997

French 2000

23-valent

Simberkoff 1986

Riley 1981

Örtqvist 1998

Leech 1987

Davis 1987

Riley 1977

Subtotal (I-squared = 0.0%, p = 0.955)

Gaillat 1985

Study

Alfageme 2006

0.94 (0.84, 1.05)

0.96 (0.64, 1.45)

RR (95% CI)

0.93 (0.60, 1.43)

0.79 (0.54, 1.14)

0.95 (0.81, 1.11)

0.53 (0.34, 0.83)

0.71 (0.50, 1.01)

0.99 (0.80, 1.22)

1.00 (0.84, 1.19)

1.24 (1.03, 1.49)

0.65 (0.30, 1.39)

1.08 (0.65, 1.77)

0.58 (0.22, 1.49)

1.14 (0.60, 2.18)

0.79 (0.63, 0.99)

1.00 (0.87, 1.16)

1.06 (0.89, 1.26)

0.98 (0.71, 1.37)

1192/25825

45/6782

vaccinees

35/607

112/12139

809/12380

32/4750

51/2445

152/1364

185/667

211/1145

10/401

29/339

6/92

14/50

133/5946

271/1306

232/937

Events,

57/300

1204/25273

47/6818

controls

43/693

141/11544

795/12421

51/4033

71/2417

166/1473

182/656

171/1150

18/470

28/352

11/97

13/53

170/6012

268/1308

175/749

Events,

58/300

6 (-5, 16)

4 (-45, 36)

efficacy (%)

7 (-43, 40)

21 (-14, 46)

5 (-11, 19)

47 (17, 66)

29 (-1, 50)

1 (-22, 20)

0 (-19, 16)

-24 (-49, -3)

35 (-39, 70)

-8 (-77, 35)

42 (-49, 78)

-14 (-118, 40)

21 (1, 37)

-0 (-16, 13)

-6 (-26, 11)

Vaccine

2 (-37, 29)

12

valency

12

3

14

23

14

14

23

14

14

14

14

Vaccine

23

0.94 (0.84, 1.05)

0.96 (0.64, 1.45)

RR (95% CI)

0.93 (0.60, 1.43)

0.79 (0.54, 1.14)

0.95 (0.81, 1.11)

0.53 (0.34, 0.83)

0.71 (0.50, 1.01)

0.99 (0.80, 1.22)

1.00 (0.84, 1.19)

1.24 (1.03, 1.49)

0.65 (0.30, 1.39)

1.08 (0.65, 1.77)

0.58 (0.22, 1.49)

1.14 (0.60, 2.18)

0.79 (0.63, 0.99)

1.00 (0.87, 1.16)

1.06 (0.89, 1.26)

0.98 (0.71, 1.37)

1192/25825

45/6782

vaccinees

35/607

112/12139

809/12380

32/4750

51/2445

152/1364

185/667

211/1145

10/401

29/339

6/92

14/50

133/5946

271/1306

232/937

Events,

57/300

Reduced risk Increased risk

1.125 .25 .5 1 2 4 8

By vaccine valencyAll cause mortality

44

Confidential Draft

Figure 24: Meta-analysis of clinical trials of the effect of pneumococcal polysaccharide vaccines on the risk of death from all causes. Trials are sorted by mortality in controls.

NOTE: Weights are from random effects analysis

Overall (I-squared = 45.8%, p = 0.031)

Koivula 1997

Riley 1986

Austrian b 1980

Simberkoff 1986

French 2000

ID

Riley 1981

Leech 1987

Gaillat 1985

Örtqvist 1998

Austrian a 1980

Study

Davis 1987

Riley 1977

Kaufman 1947

Alfageme 2006

0.94 (0.84, 1.05)

0.99 (0.80, 1.22)

0.71 (0.50, 1.01)

0.96 (0.64, 1.45)

1.24 (1.03, 1.49)

1.00 (0.84, 1.19)

RR (95% CI)

0.65 (0.30, 1.39)

0.58 (0.22, 1.49)

1.06 (0.89, 1.26)

1.08 (0.65, 1.77)

0.93 (0.60, 1.43)

1.14 (0.60, 2.18)

0.79 (0.63, 0.99)

0.53 (0.34, 0.83)

0.98 (0.71, 1.37)

1192/25825

152/1364

51/2445

45/6782

211/1145

185/667

vaccinees

10/401

6/92

232/937

29/339

35/607

Events,

14/50

133/5946

32/4750

57/300

1204/25273

166/1473

71/2417

47/6818

171/1150

182/656

controls

18/470

11/97

175/749

28/352

43/693

Events,

13/53

170/6012

51/4033

58/300

6 (-5, 16)

1 (-22, 20)

29 (-1, 50)

4 (-45, 36)

-24 (-49, -3)

0 (-19, 16)

efficacy (%)

35 (-39, 70)

42 (-49, 78)

-6 (-26, 11)

-8 (-77, 35)

7 (-43, 40)

Vaccine

-14 (-118, 40)

21 (1, 37)

47 (17, 66)

2 (-37, 29)

14

12

14

23

valency

14

14

14

23

12

Vaccine

14

14

3

23

0.94 (0.84, 1.05)

0.99 (0.80, 1.22)

0.71 (0.50, 1.01)

0.96 (0.64, 1.45)

1.24 (1.03, 1.49)

1.00 (0.84, 1.19)

RR (95% CI)

0.65 (0.30, 1.39)

0.58 (0.22, 1.49)

1.06 (0.89, 1.26)

1.08 (0.65, 1.77)

0.93 (0.60, 1.43)

1.14 (0.60, 2.18)

0.79 (0.63, 0.99)

0.53 (0.34, 0.83)

0.98 (0.71, 1.37)

1192/25825

152/1364

51/2445

45/6782

211/1145

185/667

vaccinees

10/401

6/92

232/937

29/339

35/607

Events,

14/50

133/5946

32/4750

57/300

Reduced risk Increased risk

1.125 .25 .5 1 2 4 8

Sorted by mortality in controlsAll cause mortality

45

Confidential Draft

Figure 25: Meta-analysis of clinical trials of the effect of pneumococcal polysaccharide vaccines on the risk of death from all causes. Trials are ordered by trial setting.

NOTE: Weights are from random effects analysis

.

.

Overall (I-squared = 45.8%, p = 0.031)

Davis 1987

Subtotal (I-squared = 39.0%, p = 0.098)

Riley 1986

Austrian a 1980

French 2000

Subtotal (I-squared = 40.5%, p = 0.168)

Koivula 1997

Developing

Leech 1987

Developed

Riley 1977

Alfageme 2006

Austrian b 1980

Riley 1981

Kaufman 1947

Örtqvist 1998

ID

Simberkoff 1986

Gaillat 1985

Study

0.94 (0.84, 1.05)

1.14 (0.60, 2.18)

1.00 (0.87, 1.14)

0.71 (0.50, 1.01)

0.93 (0.60, 1.43)

1.00 (0.84, 1.19)

0.84 (0.70, 1.02)

0.99 (0.80, 1.22)

0.58 (0.22, 1.49)

0.79 (0.63, 0.99)

0.98 (0.71, 1.37)

0.96 (0.64, 1.45)

0.65 (0.30, 1.39)

0.53 (0.34, 0.83)

1.08 (0.65, 1.77)

RR (95% CI)

1.24 (1.03, 1.49)

1.06 (0.89, 1.26)

1192/25825

14/50

813/16366

51/2445

35/607

185/667

379/9459

152/1364

6/92

133/5946

57/300

45/6782

10/401

32/4750

29/339

vaccinees

211/1145

232/937

Events,

1204/25273

13/53

763/15718

71/2417

43/693

182/656

441/9555

166/1473

11/97

170/6012

58/300

47/6818

18/470

51/4033

28/352

controls

171/1150

175/749

Events,

6 (-5, 16)

-14 (-118, 40)

0 (-14, 13)

29 (-1, 50)

7 (-43, 40)

0 (-19, 16)

16 (-2, 30)

1 (-22, 20)

42 (-49, 78)

21 (1, 37)

2 (-37, 29)

4 (-45, 36)

35 (-39, 70)

47 (17, 66)

-8 (-77, 35)

efficacy (%)

-24 (-49, -3)

-6 (-26, 11)

Vaccine

14

12

23

14

14

14

23

12

14

3

23

valency

14

14

Vaccine

0.94 (0.84, 1.05)

1.14 (0.60, 2.18)

1.00 (0.87, 1.14)

0.71 (0.50, 1.01)

0.93 (0.60, 1.43)

1.00 (0.84, 1.19)

0.84 (0.70, 1.02)

0.99 (0.80, 1.22)

0.58 (0.22, 1.49)

0.79 (0.63, 0.99)

0.98 (0.71, 1.37)

0.96 (0.64, 1.45)

0.65 (0.30, 1.39)

0.53 (0.34, 0.83)

1.08 (0.65, 1.77)

RR (95% CI)

1.24 (1.03, 1.49)

1.06 (0.89, 1.26)

1192/25825

14/50

813/16366

51/2445

35/607

185/667

379/9459

152/1364

6/92

133/5946

57/300

45/6782

10/401

32/4750

29/339

vaccinees

211/1145

232/937

Events,

Reduced risk Increased risk

1.125 .25 .5 1 2 4 8

By trial settingAll cause mortality

46

Confidential Draft

Figure 26: Meta-analysis of clinical trials of the effect of pneumococcal polysaccharide vaccines on the risk of death from all causes. Trials are ordered by the type of study population.

NOTE: Weights are from random effects analysis

.

.

.

.

.

Overall (I-squared = 45.8%, p = 0.031)

ID

COPD or brochogenic carcinoma

Chronic illness or elderly

Riley 1977

Study

Leech 1987

Örtqvist 1998

Koivula 1997

Subtotal (I-squared = 0.0%, p = 0.840)

Gaillat 1985

Kaufman 1947

Austrian b 1980

Riley 1981

Subtotal (I-squared = 54.1%, p = 0.042)

Riley 1986

French 2000

HIV positive

Subtotal (I-squared = .%, p = .)

Austrian a 1980

>10yo in Papua New Guinea

Davis 1987

Alfageme 2006

Subtotal (I-squared = 0.0%, p = 0.494)

Simberkoff 1986

Subtotal (I-squared = .%, p = .)

Children

0.94 (0.84, 1.05)

RR (95% CI)

0.79 (0.63, 0.99)

0.58 (0.22, 1.49)

1.08 (0.65, 1.77)

0.99 (0.80, 1.22)

0.70 (0.51, 0.96)

1.06 (0.89, 1.26)

0.53 (0.34, 0.83)

0.96 (0.64, 1.45)

0.65 (0.30, 1.39)

0.99 (0.85, 1.17)

0.71 (0.50, 1.01)

1.00 (0.84, 1.19)

1.00 (0.84, 1.19)

0.93 (0.60, 1.43)

1.14 (0.60, 2.18)

0.98 (0.71, 1.37)

0.96 (0.73, 1.28)

1.24 (1.03, 1.49)

0.79 (0.63, 0.99)

1192/25825

vaccinees

133/5946

Events,

6/92

29/339

152/1364

61/2846

232/937

32/4750

45/6782

10/401

736/15924

51/2445

185/667

185/667

35/607

14/50

57/300

77/442

211/1145

133/5946

1204/25273

controls

170/6012

Events,

11/97

28/352

166/1473

89/2887

175/749

51/4033

47/6818

18/470

681/15268

71/2417

182/656

182/656

43/693

13/53

58/300

82/450

171/1150

170/6012

6 (-5, 16)

efficacy (%)

21 (1, 37)

Vaccine

42 (-49, 78)

-8 (-77, 35)

1 (-22, 20)

30 (4, 49)

-6 (-26, 11)

47 (17, 66)

4 (-45, 36)

35 (-39, 70)

1 (-17, 15)

29 (-1, 50)

0 (-19, 16)

0 (-19, 16)

7 (-43, 40)

-14 (-118, 40)

2 (-37, 29)

4 (-28, 27)

-24 (-49, -3)

21 (1, 37)

valency

14

Vaccine

14

23

14

14

3

12

14

23

12

14

23

14

0.94 (0.84, 1.05)

RR (95% CI)

0.79 (0.63, 0.99)

0.58 (0.22, 1.49)

1.08 (0.65, 1.77)

0.99 (0.80, 1.22)

0.70 (0.51, 0.96)

1.06 (0.89, 1.26)

0.53 (0.34, 0.83)

0.96 (0.64, 1.45)

0.65 (0.30, 1.39)

0.99 (0.85, 1.17)

0.71 (0.50, 1.01)

1.00 (0.84, 1.19)

1.00 (0.84, 1.19)

0.93 (0.60, 1.43)

1.14 (0.60, 2.18)

0.98 (0.71, 1.37)

0.96 (0.73, 1.28)

1.24 (1.03, 1.49)

0.79 (0.63, 0.99)

1192/25825

vaccinees

133/5946

Events,

6/92

29/339

152/1364

61/2846

232/937

32/4750

45/6782

10/401

736/15924

51/2445

185/667

185/667

35/607

14/50

57/300

77/442

211/1145

133/5946

Reduced risk Increased risk

1.125 .25 .5 1 2 4 8

By study populationAll cause mortality

47

Confidential Draft

Mortality due to pneumonia Figure 27: Meta-analysis of clinical trials of the effect of pneumococcal polysaccharide vaccines on the risk of death from pneumonia. Trials are ordered by the year the trial started.

NOTE: Weights are from random effects analysis

Overall (I-squared = 36.7%, p = 0.115)

Austrian b 1980

Riley 1986

Koivula 1997

Study

Riley 1981

Simberkoff 1986

Riley 1977

Davis 1987

Alfageme 2006

Örtqvist 1998

Austrian a 1980

ID

0.77 (0.53, 1.11)

1.15 (0.73, 1.82)

0.42 (0.19, 0.92)

0.90 (0.28, 2.94)

0.20 (0.02, 1.62)

2.01 (0.86, 4.68)

0.57 (0.34, 0.94)

0.53 (0.10, 2.77)

1.00 (0.33, 3.07)

0.69 (0.12, 4.12)

0.53 (0.20, 1.38)

RR (95% CI)

109/19379

39/6782

9/2445

5/1364

Events,

1/401

16/1145

23/5946

2/50

6/300

2/339

6/607

vaccinees

142/19738

34/6818

21/2417

6/1473

Events,

6/470

8/1150

41/6012

4/53

6/300

3/352

13/693

controls

23 (-11, 47)

-15 (-82, 27)

58 (8, 81)

10 (-194, 72)

Vaccine

80 (-62, 98)

-101 (-368, 14)

43 (6, 66)

47 (-177, 90)

0 (-207, 67)

31 (-312, 88)

47 (-38, 80)

efficacy (%)

12

14

Vaccine

14

14

14

14

23

23

12

valency

0.77 (0.53, 1.11)

1.15 (0.73, 1.82)

0.42 (0.19, 0.92)

0.90 (0.28, 2.94)

0.20 (0.02, 1.62)

2.01 (0.86, 4.68)

0.57 (0.34, 0.94)

0.53 (0.10, 2.77)

1.00 (0.33, 3.07)

0.69 (0.12, 4.12)

0.53 (0.20, 1.38)

RR (95% CI)

109/19379

39/6782

9/2445

5/1364

Events,

1/401

16/1145

23/5946

2/50

6/300

2/339

6/607

vaccinees

Reduced risk Increased risk

1.125 .25 .5 1 2 4 8

Sorted by start of studyMortality due to pneumonia

48

Confidential Draft

Figure 28: Meta-analysis of clinical trials of the effect of pneumococcal polysaccharide vaccines on the risk of death from pneumonia. Trials are ordered by blinding status.

NOTE: Weights are from random effects analysis

.

.

.

Overall (I-squared = 36.7%, p = 0.115)

Subtotal (I-squared = .%, p = .)

Subtotal (I-squared = 38.5%, p = 0.135)

ID

Simberkoff 1986

Subtotal (I-squared = 0.0%, p = 0.702)

Riley 1981

Not reported to be double blind, placebo-controlled

Austrian b 1980

Örtqvist 1998

Riley 1986

Koivula 1997

Reported to be double blind, placebo-controlled

Riley 1977

Davis 1987

Austrian a 1980

Alfageme 2006

Unblinded, not placebo-controlled

Study

0.77 (0.53, 1.11)

1.00 (0.33, 3.07)

0.64 (0.39, 1.04)

RR (95% CI)

2.01 (0.86, 4.68)

1.12 (0.73, 1.71)

0.20 (0.02, 1.62)

1.15 (0.73, 1.82)

0.69 (0.12, 4.12)

0.42 (0.19, 0.92)

0.90 (0.28, 2.94)

0.57 (0.34, 0.94)

0.53 (0.10, 2.77)

0.53 (0.20, 1.38)

1.00 (0.33, 3.07)

109/19379

6/300

59/10933

vaccinees

16/1145

44/8146

1/401

39/6782

2/339

9/2445

5/1364

23/5946

2/50

6/607

6/300

Events,

142/19738

6/300

96/11147

controls

8/1150

40/8291

6/470

34/6818

3/352

21/2417

6/1473

41/6012

4/53

13/693

6/300

Events,

23 (-11, 47)

0 (-207, 67)

36 (-4, 61)

efficacy (%)

-101 (-368, 14)

-12 (-71, 27)

80 (-62, 98)

-15 (-82, 27)

31 (-312, 88)

58 (8, 81)

10 (-194, 72)

43 (6, 66)

47 (-177, 90)

47 (-38, 80)

0 (-207, 67)

Vaccine

valency

14

14

12

23

14

14

14

12

23

Vaccine

0.77 (0.53, 1.11)

1.00 (0.33, 3.07)

0.64 (0.39, 1.04)

RR (95% CI)

2.01 (0.86, 4.68)

1.12 (0.73, 1.71)

0.20 (0.02, 1.62)

1.15 (0.73, 1.82)

0.69 (0.12, 4.12)

0.42 (0.19, 0.92)

0.90 (0.28, 2.94)

0.57 (0.34, 0.94)

0.53 (0.10, 2.77)

0.53 (0.20, 1.38)

1.00 (0.33, 3.07)

109/19379

6/300

59/10933

vaccinees

16/1145

44/8146

1/401

39/6782

2/339

9/2445

5/1364

23/5946

2/50

6/607

6/300

Events,

Reduced risk Increased risk

1.125 .25 .5 1 2 4 8

By blindingMortality due to pneumonia

49

Confidential Draft

Figure 29: Meta-analysis of clinical trials of the effect of pneumococcal polysaccharide vaccines on the risk of death from pneumonia. Trials are ordered by concealment of allocation.

NOTE: Weights are from random effects analysis

.

.

Overall (I-squared = 36.7%, p = 0.115)

Alfageme 2006

Subtotal (I-squared = 0.0%, p = 0.716)

Riley 1986

Austrian b 1980

Austrian a 1980

Subtotal (I-squared = 59.4%, p = 0.043)

Örtqvist 1998

Adequate

Study

ID

Riley 1977

Simberkoff 1986

Davis 1987

Koivula 1997

Unclear

Riley 1981

0.77 (0.53, 1.11)

1.00 (0.33, 3.07)

0.59 (0.37, 0.94)

0.42 (0.19, 0.92)

1.15 (0.73, 1.82)

0.53 (0.20, 1.38)

0.89 (0.50, 1.61)

0.69 (0.12, 4.12)

RR (95% CI)

0.57 (0.34, 0.94)

2.01 (0.86, 4.68)

0.53 (0.10, 2.77)

0.90 (0.28, 2.94)

0.20 (0.02, 1.62)

109/19379

6/300

28/4766

9/2445

39/6782

6/607

81/14613

2/339

Events,

vaccinees

23/5946

16/1145

2/50

5/1364

1/401

142/19738

6/300

50/4936

21/2417

34/6818

13/693

92/14802

3/352

Events,

controls

41/6012

8/1150

4/53

6/1473

6/470

23 (-11, 47)

0 (-207, 67)

41 (6, 63)

58 (8, 81)

-15 (-82, 27)

47 (-38, 80)

11 (-61, 50)

31 (-312, 88)

Vaccine

efficacy (%)

43 (6, 66)

-101 (-368, 14)

47 (-177, 90)

10 (-194, 72)

80 (-62, 98)

23

12

12

23

Vaccine

valency

14

14

14

14

14

0.77 (0.53, 1.11)

1.00 (0.33, 3.07)

0.59 (0.37, 0.94)

0.42 (0.19, 0.92)

1.15 (0.73, 1.82)

0.53 (0.20, 1.38)

0.89 (0.50, 1.61)

0.69 (0.12, 4.12)

RR (95% CI)

0.57 (0.34, 0.94)

2.01 (0.86, 4.68)

0.53 (0.10, 2.77)

0.90 (0.28, 2.94)

0.20 (0.02, 1.62)

109/19379

6/300

28/4766

9/2445

39/6782

6/607

81/14613

2/339

Events,

vaccinees

23/5946

16/1145

2/50

5/1364

1/401

Reduced risk Increased risk

1.125 .25 .5 1 2 4 8

By concealment of allocationMortality due to pneumonia

50

Confidential Draft

Figure 30: Meta-analysis of clinical trials of the effect of pneumococcal polysaccharide vaccines on the risk of death from pneumonia. Trials are ordered by vaccine valency.

NOTE: Weights are from random effects analysis

.

.

.

Overall (I-squared = 36.7%, p = 0.115)

Simberkoff 1986

Subtotal (I-squared = 51.9%, p = 0.149)

Riley 1986

Austrian b 1980

Subtotal (I-squared = 0.0%, p = 0.732)

Austrian a 1980

other valencies

Riley 1977

Davis 1987

Koivula 1997

Örtqvist 1998

23-valent

ID

Subtotal (I-squared = 49.3%, p = 0.079)

Alfageme 2006

14-valent

Riley 1981

Study

0.77 (0.53, 1.11)

2.01 (0.86, 4.68)

0.88 (0.42, 1.82)

0.42 (0.19, 0.92)

1.15 (0.73, 1.82)

0.90 (0.35, 2.33)

0.53 (0.20, 1.38)

0.57 (0.34, 0.94)

0.53 (0.10, 2.77)

0.90 (0.28, 2.94)

0.69 (0.12, 4.12)

RR (95% CI)

0.69 (0.39, 1.20)

1.00 (0.33, 3.07)

0.20 (0.02, 1.62)

109/19379

16/1145

45/7389

9/2445

39/6782

8/639

6/607

23/5946

2/50

5/1364

2/339

vaccinees

56/11351

6/300

1/401

Events,

142/19738

8/1150

47/7511

21/2417

34/6818

9/652

13/693

41/6012

4/53

6/1473

3/352

controls

86/11575

6/300

6/470

Events,

23 (-11, 47)

-101 (-368, 14)

12 (-82, 58)

58 (8, 81)

-15 (-82, 27)

10 (-133, 65)

47 (-38, 80)

43 (6, 66)

47 (-177, 90)

10 (-194, 72)

31 (-312, 88)

efficacy (%)

31 (-20, 61)

0 (-207, 67)

80 (-62, 98)

Vaccine

14

12

12

14

14

14

23

valency

23

14

Vaccine

0.77 (0.53, 1.11)

2.01 (0.86, 4.68)

0.88 (0.42, 1.82)

0.42 (0.19, 0.92)

1.15 (0.73, 1.82)

0.90 (0.35, 2.33)

0.53 (0.20, 1.38)

0.57 (0.34, 0.94)

0.53 (0.10, 2.77)

0.90 (0.28, 2.94)

0.69 (0.12, 4.12)

RR (95% CI)

0.69 (0.39, 1.20)

1.00 (0.33, 3.07)

0.20 (0.02, 1.62)

109/19379

16/1145

45/7389

9/2445

39/6782

8/639

6/607

23/5946

2/50

5/1364

2/339

vaccinees

56/11351

6/300

1/401

Events,

Reduced risk Increased risk

1.125 .25 .5 1 2 4 8

By vaccine valencyMortality due to pneumonia

51

Confidential Draft

Figure 31: Meta-analysis of clinical trials of the effect of pneumococcal polysaccharide vaccines on the risk of death from pneumonia. Trials are ordered by mortality from all-cause pneumonia in controls.

NOTE: Weights are from random effects analysis

Overall (I-squared = 36.7%, p = 0.115)

Koivula 1997

Alfageme 2006

Riley 1981

ID

Riley 1986

Simberkoff 1986

Austrian b 1980

Riley 1977

Austrian a 1980

Örtqvist 1998

Davis 1987

Study

0.77 (0.53, 1.11)

0.90 (0.28, 2.94)

1.00 (0.33, 3.07)

0.20 (0.02, 1.62)

RR (95% CI)

0.42 (0.19, 0.92)

2.01 (0.86, 4.68)

1.15 (0.73, 1.82)

0.57 (0.34, 0.94)

0.53 (0.20, 1.38)

0.69 (0.12, 4.12)

0.53 (0.10, 2.77)

109/19379

5/1364

6/300

1/401

vaccinees

9/2445

16/1145

39/6782

23/5946

6/607

2/339

2/50

Events,

142/19738

6/1473

6/300

6/470

controls

21/2417

8/1150

34/6818

41/6012

13/693

3/352

4/53

Events,

23 (-11, 47)

10 (-194, 72)

0 (-207, 67)

80 (-62, 98)

efficacy (%)

58 (8, 81)

-101 (-368, 14)

-15 (-82, 27)

43 (6, 66)

47 (-38, 80)

31 (-312, 88)

47 (-177, 90)

Vaccine

14

23

14

valency

14

12

14

12

23

14

Vaccine

0.77 (0.53, 1.11)

0.90 (0.28, 2.94)

1.00 (0.33, 3.07)

0.20 (0.02, 1.62)

RR (95% CI)

0.42 (0.19, 0.92)

2.01 (0.86, 4.68)

1.15 (0.73, 1.82)

0.57 (0.34, 0.94)

0.53 (0.20, 1.38)

0.69 (0.12, 4.12)

0.53 (0.10, 2.77)

109/19379

5/1364

6/300

1/401

vaccinees

9/2445

16/1145

39/6782

23/5946

6/607

2/339

2/50

Events,

Reduced risk Increased risk

1.125 .25 .5 1 2 4 8

By mortality in controlsMortality due to pneumonia

52

Confidential Draft

Figure 32: Meta-analysis of clinical trials of the effect of pneumococcal polysaccharide vaccines on the risk of death from pneumonia. Trials are ordered by trial setting.

NOTE: Weights are from random effects analysis

.

.

Overall (I-squared = 36.7%, p = 0.115)

Developing

Simberkoff 1986

Davis 1987

Austrian a 1980

Koivula 1997

Örtqvist 1998

Riley 1981

Riley 1977

Austrian b 1980

Subtotal (I-squared = 0.0%, p = 0.555)

ID

Developed

Riley 1986

Study

Alfageme 2006

Subtotal (I-squared = 0.0%, p = 0.501)

0.77 (0.53, 1.11)

2.01 (0.86, 4.68)

0.53 (0.10, 2.77)

0.53 (0.20, 1.38)

0.90 (0.28, 2.94)

0.69 (0.12, 4.12)

0.20 (0.02, 1.62)

0.57 (0.34, 0.94)

1.15 (0.73, 1.82)

0.50 (0.33, 0.76)

RR (95% CI)

0.42 (0.19, 0.92)

1.00 (0.33, 3.07)

1.06 (0.76, 1.47)

109/19379

16/1145

2/50

6/607

5/1364

2/339

1/401

23/5946

39/6782

33/8792

vaccinees

9/2445

Events,

6/300

76/10587

142/19738

8/1150

4/53

13/693

6/1473

3/352

6/470

41/6012

34/6818

68/8899

controls

21/2417

Events,

6/300

74/10839

23 (-11, 47)

-101 (-368, 14)

47 (-177, 90)

47 (-38, 80)

10 (-194, 72)

31 (-312, 88)

80 (-62, 98)

43 (6, 66)

-15 (-82, 27)

50 (24, 67)

efficacy (%)

58 (8, 81)

Vaccine

0 (-207, 67)

-6 (-47, 24)

14

14

12

14

23

14

14

12

valency

Vaccine

23

0.77 (0.53, 1.11)

2.01 (0.86, 4.68)

0.53 (0.10, 2.77)

0.53 (0.20, 1.38)

0.90 (0.28, 2.94)

0.69 (0.12, 4.12)

0.20 (0.02, 1.62)

0.57 (0.34, 0.94)

1.15 (0.73, 1.82)

0.50 (0.33, 0.76)

RR (95% CI)

0.42 (0.19, 0.92)

1.00 (0.33, 3.07)

1.06 (0.76, 1.47)

109/19379

16/1145

2/50

6/607

5/1364

2/339

1/401

23/5946

39/6782

33/8792

vaccinees

9/2445

Events,

6/300

76/10587

Reduced risk Increased risk

1.125 .25 .5 1 2 4 8

By trial settingMortality due to pneumonia

53

Confidential Draft

Figure 33: Meta-analysis of clinical trials of the effect of pneumococcal polysaccharide vaccines on the risk of death from pneumonia. Trials are ordered by type of study population.

NOTE: Weights are from random effects analysis

.

.

.

.

Overall (I-squared = 36.7%, p = 0.115)

Riley 1977

Koivula 1997

Subtotal (I-squared = 0.0%, p = 0.533)

Subtotal (I-squared = 0.0%, p = 0.498)

Chronic illness or elderly

Austrian b 1980

>10yo in Papua New Guinea

Riley 1981

Riley 1986

Subtotal (I-squared = 13.3%, p = 0.329)

Örtqvist 1998

ID

Subtotal (I-squared = .%, p = .)

Children

Davis 1987

COPD or brochogenic carcinoma

Alfageme 2006

Austrian a 1980

Simberkoff 1986

Study

0.77 (0.53, 1.11)

0.57 (0.34, 0.94)

0.90 (0.28, 2.94)

0.82 (0.32, 2.07)

0.39 (0.19, 0.80)

1.15 (0.73, 1.82)

0.20 (0.02, 1.62)

0.42 (0.19, 0.92)

1.08 (0.72, 1.62)

0.69 (0.12, 4.12)

RR (95% CI)

0.57 (0.34, 0.94)

0.53 (0.10, 2.77)

1.00 (0.33, 3.07)

0.53 (0.20, 1.38)

2.01 (0.86, 4.68)

109/19379

23/5946

5/1364

8/350

10/2846

39/6782

1/401

9/2445

68/10237

2/339

vaccinees

23/5946

2/50

6/300

6/607

16/1145

Events,

142/19738

41/6012

6/1473

10/353

27/2887

34/6818

6/470

21/2417

64/10486

3/352

controls

41/6012

4/53

6/300

13/693

8/1150

Events,

23 (-11, 47)

43 (6, 66)

10 (-194, 72)

18 (-107, 68)

61 (20, 81)

-15 (-82, 27)

80 (-62, 98)

58 (8, 81)

-8 (-62, 28)

31 (-312, 88)

efficacy (%)

43 (6, 66)

47 (-177, 90)

0 (-207, 67)

47 (-38, 80)

-101 (-368, 14)

Vaccine

14

14

12

14

23

valency

14

23

12

14

Vaccine

0.77 (0.53, 1.11)

0.57 (0.34, 0.94)

0.90 (0.28, 2.94)

0.82 (0.32, 2.07)

0.39 (0.19, 0.80)

1.15 (0.73, 1.82)

0.20 (0.02, 1.62)

0.42 (0.19, 0.92)

1.08 (0.72, 1.62)

0.69 (0.12, 4.12)

RR (95% CI)

0.57 (0.34, 0.94)

0.53 (0.10, 2.77)

1.00 (0.33, 3.07)

0.53 (0.20, 1.38)

2.01 (0.86, 4.68)

109/19379

23/5946

5/1364

8/350

10/2846

39/6782

1/401

9/2445

68/10237

2/339

vaccinees

23/5946

2/50

6/300

6/607

16/1145

Events,

Reduced risk Increased risk

1.125 .25 .5 1 2 4 8

By study populationMortality due to pneumonia

54

Confidential Draft

Other outcomes Figure 34: Meta-analysis of clinical trials of the effect of pneumococcal polysaccharide vaccines on the risk of death from pneumococcal infection.

NOTE: Weights are from random effects analysis

Overall (I-squared = 14.2%, p = 0.312)

Klastersky 1986

John 1984

ID

Study

Austrian b 1980

Simberkoff 1986

0.93 (0.29, 3.05)

0.81 (0.05, 12.16)

(Excluded)

RR (95% CI)

0.63 (0.21, 1.92)

7.03 (0.36, 135.95)

9/8015

1/26

0/62

vaccinees

Events,

5/6782

3/1145

9/8026

1/21

0/37

controls

Events,

8/6818

0/1150

7 (-205, 71)

19 (-1116, 95)

efficacy (%)

Vaccine

37 (-92, 79)

-603 (-13495, 64)

17

14

valency

Vaccine

12

14

0.93 (0.29, 3.05)

0.81 (0.05, 12.16)

(Excluded)

RR (95% CI)

0.63 (0.21, 1.92)

7.03 (0.36, 135.95)

9/8015

1/26

0/62

vaccinees

Events,

5/6782

3/1145

Reduced risk Increased risk 1.125 .25 .5 1 2 4 8

Mortality due to pneumococcal infection

55

Confidential Draft

Figure 35: Meta-analysis of clinical trials of the effect of pneumococcal polysaccharide vaccines on the risk of bacteraemia, septicaemia or invasive pneumococcal disease, sorted by start of study.

NOTE: Weights are from random effects analysis

Overall (I-squared = 0.0%, p = 0.503)

Klastersky 1986

Leech 1987

Simberkoff 1986

ID

Örtqvist 1998

Austrian a 1980

Honkanen 1999

Study

French 2000

John 1984

0.98 (0.55, 1.75)

0.81 (0.05, 12.16)

3.16 (0.13, 76.63)

0.50 (0.05, 5.53)

RR (95% CI)

0.21 (0.02, 1.77)

(Excluded)

0.37 (0.07, 1.91)

1.48 (0.67, 3.26)

1.19 (0.23, 6.20)

25/16918

1/26

1/92

1/1145

vaccinees

1/339

0/607

2/13980

Events,

15/667

4/62

25/15951

1/21

0/97

2/1150

controls

5/352

0/693

5/12945

Events,

10/656

2/37

2 (-75, 45)

19 (-1116, 95)

-216 (-7563, 87)

50 (-453, 95)

efficacy (%)

79 (-77, 98)

63 (-91, 93)

Vaccine

-48 (-226, 33)

-19 (-520, 77)

17

14

14

valency

23

12

23

Vaccine

23

14

0.98 (0.55, 1.75)

0.81 (0.05, 12.16)

3.16 (0.13, 76.63)

0.50 (0.05, 5.53)

RR (95% CI)

0.21 (0.02, 1.77)

(Excluded)

0.37 (0.07, 1.91)

1.48 (0.67, 3.26)

1.19 (0.23, 6.20)

25/16918

1/26

1/92

1/1145

vaccinees

1/339

0/607

2/13980

Events,

15/667

4/62

Reduced risk Increased risk 1.125 .25 .5 1 2 4 8

Sorted by start of studyBacteraemia, septicaemia or invasive pneumococcal disease

NB “john 1984” uses HiB vaccine as a control

56

Confidential Draft

Funnel plots Figure 36: Funnel plot for outcome presumptive pneumococcal pneumonia.

se(lo

gRR

)

Funnel plot presumptive pneumococcal pneumoniaRR (log scale)

.125 .25 .5 1 2 4 8

0

1

2

Asymmetry p= 0.66 Figure 37: Funnel plot for outcome all-cause pneumonia.

se(lo

gRR

)

Funnel plot all cause pneumonia, cases/episodes combinedRR (log scale)

.125 .25 .5 1 2 4 8

0

.5

1

Asymmetry p=0.33 Figure 38: Funnel plot for outcome all-cause mortality.

se(lo

gRR

)

Funnel plot all cause mortalityRR (log scale)

.125 .25 .5 1 2 4 8

0

.25

.5

.75

Asymmetry p= 0.072

57

Confidential Draft

Figure 39: Funnel plot for outcome mortality due to pneumonia.

se(lo

gRR

)

Funnel plot mortality due to pneumoniaRR (log scale)

.125 .25 .5 1 2 4 8

0

.5

1

1.5

Asymmetry p= 0.47

58

Confidential Draft

Appendix 1: Study characteristics

Author Alfageme 200627 Approximate year of trial start

2003

Study population Spanish COPD patients, excluding those with specified co-morbidities. 95% male. Mean 68.5 years old, range 61-73 years

Interventions 23-valent PPV. No intervention in control group. Follow up checks only, as in vaccine group

Quality Described as randomised. Generation of allocation sequence described and adequate but concealment of allocation not described Not described as double-blind but outcome assessors are blinded

Outcomes Presumptive pneumococcal pneumonia (radiographic and bacteriologic) All-cause pneumonia (radiographic) All-cause mortality Mortality due to pneumonia

Author Austrian 197628 Approximate year of trial start

1972

Study population Young gold-miners in South Africa 100% male. Age not well defined

Interventions 13 valent PPV. 2 control groups : saline placebo and Meningococcal A vaccine Quality Described as randomised. Generation of allocation sequence described and

adequate but concealment of allocation not described Not reported as double-blind and no description of who, if anyone, was blinded

Outcomes Presumptive pneumococcal pneumonia (radiographic and is suggestive that culture performed but not explicitly stated) All-cause pneumonia (radiographic) Bacteraemia/septicaemia

Author Austrian 198029 Approximate year of trial start

1973

Study population a) US psychiatric hospital patients (resident for over 3m) b) clients (>45yo) of a health insurance plan Sex and age distribution not well defined for either trial

Interventions 12v PPV with saline placebo in controls Quality a) Not described as randomised. Neither generation of allocation sequence nor

concealment of allocation described. Described as double-blind but not possible to determine who is truly blinded b) Described as randomised. Generation of allocation sequence and concealment of allocation described and adequate Not described as double-blind. Some trial staff blinded but unclear regarding outcome assessors

Outcomes a) All-cause pneumonia (radiographic) All-cause mortality Mortality due to pneumonia Bacteraemia/septicaemia b) All-cause pneumonia (radiographic) All-cause bronchitis All-cause mortality Mortality due to pneumonia Mortality due to pneumococcal infection

59

Confidential Draft

Author Davis 198730 Approximate year of trial start

1978

Study population US COPD patients Mean age 62.5 years. Sex distribution not defined

Interventions 14-valent PPV. Control group given saline placebo Quality Described as randomised. Generation of allocation sequence described and

adequate but concealment of allocation not described Described as double-blind. Participants are blinded. Outcome assessors are blinded.

Outcomes Definitive pneumococcal pneumonia (radiographic and bacteriologic) Presumptive pneumococcal pneumonia (radiographic and bacteriologic) All-cause pneumonia (radiographic) All-cause mortality Mortality due to pneumonia

Author Douglas 198431 Approximate year of trial start

1980

Study population Australian children recruited from GP lists, seen for any reason in previous 12m 55% male, age range reported 2-4.5 years

Interventions 14-valent PPV. Saline placebo in controls Quality Described as randomised. Generation of allocation sequence described but not

clear if adequate. Concealment of allocation described and adequate Described as double-blind. Participants are blinded. Outcome assessors are blinded.

Outcomes All-cause pneumonia (unclear diagnostic procedure, report those hospitalised for pneumonia) Otitis media

Author Douglas 198632 Approximate year of trial start

1982

Study population Australian Aboriginal children Mean age 3.3 years. Minimum age 2 years. Sex distribution not reported.

Interventions 14-valent PPV .Placebo in control group (not stated which substance but prepared by Merck, Sharp and Dohme – likely to be saline)

Quality Described as randomised. Generation of allocation sequence not described. Concealment of allocation described and adequate Described as double-blind. Participants are blinded. Outcome assessors are blinded.

Outcomes Otitis media Notes Data only extracted for children over 2yo Author French 199833 Approximate year of trial start

1995

Study population Ugandan HIV+ 15-55yo, not pregnant, not on rifampicin, not stage 4, no acute febrile illnesses 29% male. Mean age 31 years. Minimum age 15 years.

Interventions 23-valent PPV. Controls received sodium phosphate carrier (placebo) Quality Described as randomised but neither generation of allocation sequence nor

allocation concealment described. Described as double-blind. Participants are blinded. Some trial staff blinded but unclear if outcome assessors are blinded.

Outcomes All-cause pneumonia (radiographic) All-cause mortality Invasive pneumococcal disease

60

Confidential Draft

Author Gaillat 198534 Approximate year of trial start

1980

Study population French elderly in hospitals or nursing homes 34% male. Mean age 74 years

Interventions 14-valent PPV. No intervention in control group Quality Described as randomised. Generation of allocation sequence not described.

Concealment of randomisation described but not adequate. Not apparently blinded

Outcomes All-cause pneumonia (appears not all cases had radiographic confirmation but percentage unclear) All-cause mortality

Author Honkanen 199935 Approximate year of trial start

1992

Study population Finnish older than 65yo Mean age 73.5 years

Interventions 23-valent PPV and influenza vaccine in pneumococcal vaccine group. Influenza vaccine in control group.

Quality Not described as randomised. Method of allocation described (by odd or even year of birth) but not adequate Not described as double-blind. Unclear if outcome assessors blind

Outcomes Presumptive pneumococcal pneumonia (radiographic and presence of circulating pneumolysin specific immune complexes) All-cause pneumonia (radiographic) Bacteraemia/septicaemia

Author John 198436 Approximate year of trial start

1978

Study population Jamaican children with sickle cell disease 53% male. Mean age 1.2 years, range 0.5-2.9 years

Interventions 14 valent PPV. Haemophilus influenzae type B vaccine in control group Quality Described as randomised but neither generation of allocation sequence nor

allocation concealment described. Not reported as double-blind and blinding not discussed

Outcomes Mortality due to pneumococcal infection Bacteraemia/septicaemia

Notes Factorial trial including penicillin use in one level. Result extracted only for those not receiving penicillin. Protocol revised by authors moving some from penicillin groups to vaccine-only groups maintaining vaccine-type allocation. Not possible to extract outcomes on original allocations

Author Kaufman 194738 Approximate year of trial start

1937

Study population Elderly in New York City Home Mean 67 years. Sex distribution not defined

Interventions 3v PPV. No intervention in control group Quality Described as randomised. Randomisation not adequate.

Not described as double-blind and no further description of any blinding in paper Outcomes All-cause pneumonia (unclear diagnostic criteria)

All-cause mortality Notes Results from Kaufman 194139 are summarised in Kaufman 1947 – only these data

were extracted. A randomisation process is reported starting in the trial’s second year (alternation; volunteers taken in first year), results were extracted from second year on.

61

Confidential Draft

Author Klastersky 198640 Approximate year of trial start

1987

Study population Belgian bronchogenic carcinoma patients, most with no radiotherapy or chemotherapy prior to vaccination 96% male. Mean age 61 years, range 42-78 years

Interventions 17v PPV. Saline placebo in control group Quality Described as randomised. No description of generation of allocation sequence but

concealment described and adequate Not reported as double-blind but participants blind and at least some trial staff. Unclear if outcome assessors blind

Outcomes Presumptive pneumococcal pneumonia (radiographic and bacteriologic) Mortality due to pneumococcal infection Bacteraemia/septicaemia

Author Koivula 199741 Approximate year of trial start

1982

Study population Finnish elderly (over 60yo) Interventions 14-valent PPV and influenza vaccine in intervention group. Influenza vaccine in

control group 37% male.

Quality Described as randomised. Generation of allocation sequence described and adequate. Concealment described and but difficult to determine if adequate. Not reported as double-blind but outcome assessors blinded.

Outcomes Presumptive pneumococcal pneumonia (radiographic and two-fold rise of pneumolysin antibodies ) All-cause pneumonia (radiographic) All-cause mortality Mortality due to pneumonia

Notes Database of elderly residents randomised prior to recruitment. Outcomes also reported for non-responders

Author Leech 198742 Approximate year of trial start

1981

Study population Canadian COPD patients 71% male. Mean age 67 years, range 40-89 years

Interventions 14-valent PPV. Saline placebo in control group. Both groups given influenza vaccine at 0, 1 and 2 years

Quality Described as randomised but neither generation of allocation sequence nor allocation concealment described. Described as double-blind. At least some trial staff blinded but unclear if outcome assessors blind

Outcomes All-cause mortality Bacteraemia/septicaemia Emergency department visits for pneumonia

Author Lehmann 199143 Approximate year of trial start

1981

Study population Children under 5yo in Papua New Guinea Interventions 14-valent and 23-valent used in trial. Unclear which used when. Control group

received saline placebo 0.3-4.9 years old. Sex distribution not defined

Quality Described as randomised but neither generation of allocation sequence nor allocation concealment described. Described as double-blind but unclear who is blinded

Outcomes Acute lower respiratory tract infections Notes Same trial as Riley 1986 but reports different outcomes

62

Confidential Draft

Author MacLeod 194544 Approximate year of trial start

1944

Study population US trainees at the Army Airforce Technical School Interventions 4V PPV. Saline placebo in control group

100% male. Mean age 23.3 years. Minimum age 18 years Quality Described as randomised. Both generation of allocation sequence and allocation

concealment described but neither are adequate. Not described as double-blind and no further description of any blinding in paper

Outcomes Presumptive pneumococcal pneumonia (clinical and bacteriologic) Author Mäkelä 198045 Approximate year of trial start

1977

Study population Finnish children with prior otitis media Age range 2-6.9 years.

Interventions 14-valent PPV. Haemophilus influenzae B vaccination in control group Quality Described as randomised. Both generation of allocation sequence and allocation

concealment described but neither are adequate. Not reported as double-blind but outcome assessors blinded.

Outcomes Otitis media Author Örtqvist 199846 Approximate year of trial start

1991

Study population Swedish non-immunocompromised middle aged and elderly (50yo and over) who had previously been hospitalised for community aquired pneumonia 48% male. Mean age 69.2 years, range 50-85 years

Interventions 23-valent PPV. Control group received saline placebo Quality Described as randomised and both generation of allocation sequence and

allocation concealment are adequate. Described as double-blind and both participants and outcome assessors are blinded

Outcomes Definitive pneumococcal pneumonia Presumptive pneumococcal pneumonia (radiographic and bacteriologic/two-fold rise of pneumolysin antibodies) All-cause pneumonia (radiographic) All-cause mortality Mortality due to pneumonia Bacteraemia/septicaemia

Author Riley 197747 Approximate year of trial start

1974

Study population Highlanders over 10yo in Papua New Guinea Sex distribution not defined

Interventions 14-valent PPV. Control received placebo Quality Described as randomised. Generation of allocation sequence not described.

Concealment of allocation described and adequate Described as double-blind. Participants are blinded. Outcome assessors are blinded.

Outcomes All-cause pneumonia (radiographic where possible, percentage unclear) All-cause mortality Mortality due to pneumonia Acute lower respiratory tract infections

63

Confidential Draft

Author Riley 198148 Approximate year of trial start

1974

Study population 6m-5yo highland children in Papua New Guinea Sex distribution not defined

Interventions 14-valent PPV. Control group received saline placebo Quality Described as randomised. Generation of allocation sequence not described.

Concealment of allocation described and adequate. Described as double-blind. Participants are blinded. Outcome assessors are blinded.

Outcomes All-cause mortality Mortality due to pneumonia Acute lower respiratory tract infections

Author Riley 198649 Approximate year of trial start

1981

Study population 4m-5yo highland children in Papua New Guinea Sex distribution not defined

Interventions 14-valent and 23-valent used in trial. Unclear which used when. Control group received saline placebo

Quality Described as randomised but neither generation of allocation sequence nor allocation concealment described. Described as double-blind but unclear who is blinded

Outcomes All-cause mortality Mortality due to pneumonia

Notes Same trial as Lehmann but reports different outcomes Author Rosen 198450 Approximate year of trial start

1982

Study population Swedish children under 5yo 54% male. Mean age 3.3 years, range 0.5-5 years

Interventions 14-valent PPV. Saline placebo in control group Quality Not described as randomised and no clear description of randomisation method.

Reported to be double-blind by no description of who is blinded Outcomes Otitis media Notes 2 publications of this study. 1984 publication divides under 2yo and over 2yo Author Schuller 198351 Approximate year of trial start

1981

Study population US asthmatic children with a minimum of 4 prior episodes of otitis media 61% male. Mean age 3.5 years, range 2-6 years

Interventions 14-valent PPV. Control group received only “normal treatment” for otitis media (i.e. antibiotics when otitis media occurs)

Quality Described as randomised but neither generation of allocation sequence nor allocation concealment described. Not described as double-blind and no description of any blinding

Outcomes Otitis media Notes Multiple interventions. Data only extracted for vaccine only and baseline control

group (no active interventions)

64

Confidential Draft

Author Simberkoff 198654 Study population US “high-risk”patients i.e. those 55yo or older, or with chronic disease Interventions 14-valent PPV. Saline placebo in control group Quality Described as randomised. No description of generation of allocation sequence but

concealment described and adequate Described as double-blind but no description of who is blinded

Outcomes Presumptive pneumococcal pneumonia (radiographic and bacteriologic) All-cause pneumonia (radiographic) All-cause bronchitis All-cause mortality Mortality due to pneumonia Mortality due to pneumococcal infection Bacteraemia/septicaemia

Author Smit 197755 Approximate year of trial start

a)1973 b) 1974

Study population Young gold-miners in South Africa 100% male. Age distribution not well defined

Interventions a)6v PPV. Two control groups: saline placebo and Meningococcal A and C vaccine b)12v PPV.Two control groups: saline placebo and Meningococcal A and C vaccine

Quality a and b) Described as randomised but neither generation of allocation sequence nor allocation concealment described. Not described as double-blind but outcome assessors appear to be blinded

Outcomes a and b) Presumptive pneumococcal pneumonia (radiographic and bacteriologic) All-cause pneumonia (radiographic), results only available for both control groups combined All-cause bronchitis), results only available for both control groups combined

Author Steentoft 200652 Approximate year of trial start

2005

Study population Danish COPD patients 55% male. Age range 47-86 years

Interventions 23-valent PPV and combinations of steroid treatments in vaccine groups (3). Control group only has the steroid combinations

Quality Described as randomised. No description of generation of allocation sequence but concealment described and adequate No blinding reported

Outcomes All-cause pneumonia (radiographic) Exacerbation of COPD

Notes Three vaccine groups combined as this reflects the steroid treatment patterns in the control group

Author Zhogolev 200356 Approximate year of trial start

2001

Study population Russian soldiers in a) North-west Russia, b) Central Russia, c) East Russia, d) also in East Russia, but the last trial compared 23PPV plus influenza vaccination to no intervention. 100% male. Age range not defined

Interventions 23-valent polysaccharide, control group received no intervention Quality No description of generation of allocation sequence, concealment of allocation or

blinding Outcomes All cause pneumonia (diagnostic criteria unclear) Notes 4 trials are reported in three Russian regions with differing risk settings. Incidence

of pneumonia in controls was much higher in the “central” region, compared to the others. One trial (d) excluded as compares PPV and influenza vaccination to no

65

Confidential Draft

intervention

66