ORIGINAL RESEARCH ARTICLE
BioDrugs 2000 Dec; 14 (6): 389-408 1173-8804/00/0012-0389/$20.00/0 © Adis International Limited. All rights reserved.
Meta-Analysis of Published Clinical Trials of a Ribosomal Vaccine (Ribomunyl®1) in Prevention of Respiratory Infections Peter Boyle,1 Joseph A. Bellanti2 and Chris Robertson1 1 Division of Epidemiology and Biostatistics, European Institute of Oncology, Milan, Italy 2 Department of Paediatrics, Georgetown University, Washington, D.C., USA
Abstract
Objective: To perform a meta-analysis using data from all clinical trials and studies of a ribosomal vaccine (Ribomunyl®) in order to estimate its overall effect on the number of infections and antibacterial courses used per person. Design and setting: Meta-analysis of studies performed between 1985 and 1999 in 7 European countries and also in Kazakhstan, Tunisia, Morocco and Argentina. Patients and participants: Information from 14 213 adults and children. Results: There were 9 randomised, double-blind, placebo-controlled studies, 3 randomised nonblind studies and 16 nonblind studies with no placebo arm in which the response to ribosomal vaccine was compared with historical information. The mean number of infections per person in a study period of 3 months using placebo was found to be 2.39 (standard error ± 0.50), and in a study period of 6 months was 3.35 (±0.41) infections. In both study periods, ribosomal vaccine use was associated with a reduction in the number of infections per person of 1.43 (±0.26). In the study period, patients on placebo reported 3.02 (±0.44) antibacterial courses, whereas ribosomal vaccine was associated with a reduction of 1.32 (±0.42) antibacterial courses. Conclusions: In spite of variability in data quality, and the small sample size in some of the studies, we conclude that in patients with recurrent respiratory infections ribosomal vaccine significantly reduces both the number of infections and the number of antibacterial courses compared with placebo. This study is a strong and objective demonstration of the efficacy of ribosomal vaccine in limiting the number of otorhinolaryngological infections in children and adults.
The mucous membrane lining the airways provides a natural barrier to the main source of environmental pathogens in humans.[1] This barrier is equipped with highly efficient biochemical, me1 Use of tradenames is for product identification only, and does not imply endorsement.
chanical and immunological defence systems. The immune defences of the nasal cavities also have an essential role, being part of the mucosae-associated lymphoid tissue.[2] If tissue defences become ineffective, infection and inflammation results, typified by rhinitis or rhinopharyngitis[3] and persistent infection can be complicated by sinusitis, otitis,[4]
390
laryngitis and bronchitis. Although otorhinolaryngological infections are usually mild and self-limited, at times they may be recurrent, chronic and disabling and lead to therapeutic difficulties.[5] Since the frequency of these recurrences may indicate inadequate immune protection, immunostimulant treatment has been demonstrated to be effective in reducing the frequency of these recurrences. Ribosomal vaccine (Ribomunyl®) is comprised of 2 parts: ribosomal fractions of Klebsiella pneumoniae, Streptococcus pneumoniae, Streptococcus pyogenes and Haemophilus influenzae, and membrane fractions of Klebsiella pneumoniae. This composition gives the preparation a dual mode of action. The immunogenic properties of ribosomes were initially demonstrated for Mycobacterium tuberculosis ribosomes[6] and have been confirmed subsequently.[7,8] It has been demonstrated that ribosomal vaccine induces the production of humoral and secretory specific antibodies[9-11] against the 4 bacterial strains included in the compound. This has also been associated with the nonspecific immunostimulant properties of the membrane fractions of K. pneumoniae: activation of natural killer cells,[12] polyclonal stimulation of B lymphocytes[13] and T lymphocytes,[14] and activation of polymorphonuclear cells[15] as well as macrophages by increasing phagocytosis and cytokine production (interleukin-1,[16] interleukin-6 and interleukin-8).[17,18] Ribosomal vaccine, through this double mechanism of action, provides protection against bacterial and viral infections by stimulating the mucosal immune system.[19] The prevalence of chronic obstructive pulmonary diseases and the increasing cost of treatment has justified a preventive approach, and has led to renewed interest in this class of drugs.[20] In order to clarify the effectiveness of ribosomal vaccine in protecting or reducing infections, we have assembled data from all published clinical trials and studies conducted on ribosomal vaccine and performed a meta-analysis. Data from randomised trials (both double-blind and nonblind) were as© Adis International Limited. All rights reserved.
Boyle et al.
sessed, and are reported independently from data derived from nonblind single arm studies. Material and Methods The principal primary publication of each trial was the main source of data for the meta-analysis. This information was obtained from the manufacturer of Ribomunyl®. Data from 28 studies are analysed; these represent all the peer-reviewed trials of ribosomal vaccine as well as studies published in other forms. For 16 trials there was a publication in a peer-reviewed journal. For 1 trial there was a statistical report available. In a few cases (5) there was only an abstract or a conference paper, whereas for 8 the full conference poster or slides were available. Half of the studies were published in French. Most of the remainder were published in English, apart from 2 in German and 1 in Czech. The studies were carried out between 1985 and 1999 in 7 European countries and also in Kazakhstan, Tunisia, Morocco, and Argentina. Fourteen of the studies were carried out in France where the compound was developed, although it is now used clinically in 50 countries for protection against upper and lower respiratory tract infections. There are 3 major groups of studies: • 9 randomised, double-blind, placebo-controlled trials • 3 randomised studies that were not blinded; one was placebo-controlled, another was a trial of ribosomal vaccine against a comparator drug [Klebsiella pneumoniae glycoprotein extract (Biostim®)], and in the third ribosomal vaccine was compared with antibronchitic bacterial vaccine (Broncho Vaxom®). • 16 nonblind studies of ribosomal vaccine alone with no comparative treatment or placebo arm, but which used historical information on the study patients. In the randomised comparative trials the response to ribosomal vaccine was compared with the response to placebo or the comparator drug in the same study period. In the historical comparison studies, the response to ribosomal vaccine in the study period was compared with a previous referBioDrugs 2000 Dec; 14 (6)
Ribosomal Immunostimulant and Respiratory Infection
391
Table I. Demographic data from each trial Study
Reference Location
Randomised double-blind studies Lacomme & Narcy (1985) 21 Haguenauer et al. (1987) Debas (1989) Garabedian et al. (1990) Guérin et al. (1990) Vautel et al. (1993) Hüls et al. (1995) Serrano et al. (1997) Fiocchi & Giovannini (1997)
22 23 24 25 26 27 28 29
Randomised nonblind studies Akoun et al. (1988) 30 Palma-Carlos et al. (1990)
31
France France Argentina France France France Germany
Centre
10 Single Single 3 4 Single Multi
France/Belgium Multi Italy
France Portugal
11
Multi Single
Group
Patients
Male (%) Age (y)a Population
RV
53
65
7.0
P
51
65
7.0
RV
19
78
3.9
P
19
45
3.9
RV
25
60
27.1
P
25
44
32.9
RV
78
46
3.8
P
76
63
3.8
RV
51
75
58.4
P
52
71
59.0
RV
32
53
3.1
P
32
63
2.8
RV
77
49
6.0
P
77
48
6.0
RV
168
42
39.5
P
159
42
42.7
RV
290
60
32.8
P
287
55
32.8
RV
66
66
56.0
KPGE
67
66
56.0
RV
15
40
33.0
ABV
15
47
35.0
C C A C A C C A A
A A
∨
Slapák (1993)
32
Czech
Single
RV
28
NS
7.1
P
24
NS
6.6
C
Nonblind single arm studies Cotin & Lesbros (1986) 33
France
Multi
RV
964
55
16.1
M
Castel (1986)
France
Multi
RV
137
58
8.3
C
34
Legros (1986)
35
France
Multi
RV
185
53
24.6
M
Traissac & Petit (1986)
36
France
Single
RV
77
53
11.2
C
Grimfeld (1988)
37
France
Single
RV
50
70
6.2
C
Perruchet & Vautel (1990)
38
France
Multi
RV
1989
58
5.0
C
Menardo & Perruchet (1990)
39
France
Multi
RV
2021
64
5.0
C
Sapène et al. (1990)
40
France
Multi
RV
1127
59
54.4
A
Huls et al. (1991)
41
Germany
Multi
RV
492
59
1-16
C
Gonçalves et al. (1993)
42
Portugal
Multi
RV
1117
53
46.7
A
Benchakroun et al. (1994)
43
Morocco
8
RV
191
54
29.0
M
Alfarro (1995)
44
Portugal
Multi
RV
1046
54
6.5
C
Belaeva (1997)
45
Belarus
Single
RV
124
NS
3-15
C
Shortanbaev (1997)
46
Kazakhstan
Single
RV
153
NS
7-14
C
Arfa et al. (1997)
47
Tunisia
Multi
RV
762
61
4.8
C
Nascimento (1998)
48
Portugal
Multi
RV
1992
53
5.6
C
a
Average age (or range) of patients in the study arm.
A = adults; ABV = antibronchitic bacterial vaccine (Broncho Vaxom®); C = children; KPGE = Klebsiella pneumoniae glycoprotein extract (Biostim®); M = mixed adults and children; NS = not stated; P = placebo; RV = ribosomal vaccine (Ribomunyl®).
© Adis International Limited. All rights reserved.
BioDrugs 2000 Dec; 14 (6)
392
Boyle et al.
Table II. Baseline data from each trial Study
Randomised double-blind studies Lacomme & Narcy (1985) Haguenauer et al. (1987) Debas (1989) Garabedian et al. (1990) Guérin et al. (1990) Vautel et al. (1993) Hüls et al. (1995) Serrano et al. (1997) Fiocchi & Giovannini (1997)
Randomised nonblind studies Akoun et al. (1988) Palma-Carlos et al. (1990)
Reference
21 22 23 24 25 26 27 28 29
30 31
Group
Infectious episodes
Antibacterial courses
mean
mean
SD
SD
RV
6.5
NA
5
NA
P
6.5
NA
5
NA
R
6.9
NA
6
NA
P
6.5
NA
5.2
NA
RV
7.4
1.7
NA
NA
P
7.4
1.7
NA
NA
RV
NA
NA
3.6
NA
P
NA
NA
3.4
NA
RV
NA
NA
NA
NA
P
NA
NA
NA
NA
RV
5.9
1.3
NA
NA
P
6.1
1.5
NA
NA
RV
6.3
3.1
NA
NA
P
6
3.5
NA
NA
RV
4.4
1.9
NA
NA
P
4.3
1.7
NA
NA
RV
3
1.3
NA
NA
P
3
1.3
NA
NA
NA
RV
5.9
NA
5.3
KPGE
5.9
NA
5.5
NA
RV
NA
NA
NA
NA
ABV
NA
NA
NA
NA
RV
3
NA
NA
NA
P
1.3
NA
NA
NA
∨
Slapák (1993)
32
Nonblind single arm studies Cotin & Lesbros (1986)
33
RV
6.6
3.7
5
3.2
Castel (1986)
34
RV
6.2
2.4
4.8
2.5
Legros (1986)
35
RV
6.1
2.7
3.9
2.5
Traissac & Petit (1986)
36
RV
6.4
2.8
4.1
1.9
Grimfeld (1988)
37
RV
7.6
NA
6.8
NA
Perruchet & Vautel (1990)
38
RV
5.3
NA
4.3
NA
Menardo & Perruchet (1990)
39
RV
5
NA
5
NA NA
Sapène et al. (1990)
40
RV
5
NA
4.9
Huls et al. (1991)
41
RV
8.3
NA
NA
NA
Gonçalves et al. (1993)
42
RV
4.5
NA
4
NA
Benchakroun et al. (1994)
43
RV
6
NA
5.1
NA
Alfarro (1995)
44
RV
5.5
NA
5.5
NA
Belaeva (1997)
45
RV
NA
NA
NA
NA
Shortanbaev (1997)
46
RV
NA
NA
NA
NA
Arfa et al. (1997)
47
RV
6.9
NA
5.9
NA
Nascimento (1998)
48
RV
5.6
NA
4.6
NA
ABV = antibronchitic bacterial vaccine (Broncho Vaxom®); KPGE = Klebsiella pneumoniae glycoprotein extract (Biostim®); NA = data not available; P = placebo; RV = antibronchitic bacterial vaccine (Ribomunyl®); SD = standard deviation.
© Adis International Limited. All rights reserved.
BioDrugs 2000 Dec; 14 (6)
Ribosomal Immunostimulant and Respiratory Infection
ence period. For example, if the study was carried out in the winter of 1997 then the reference period would be the winter of 1996. The papers were all read in detail and the baseline data, including reference period data and end of study information, were extracted where available. Although the randomised studies were conducted according to Good Clinical Practice guidelines and standards, these have varied greatly over the period of time when these studies were conducted. Consequently, the data quality in the papers and reports is not uniform. Many studies only publish mean values without any measure of standard deviation. Also many studies do not quote the baseline means. The demographic data for each trial are presented in table I.[21-48] It is immediately apparent that there is a wide variation in the size of the studies. The gender of the study patients does not vary a great deal, but some studies were carried out in adults whereas the majority were only carried out in children. There is information from 14 213 participants in total, of whom 57% are male. Most of the studies (16) included only children, with 9332 children, whereas 6 studies included adults only, with 2837 adults. Both adults and children were recruited in 6 studies, which included a total of 2044 patients. The data in table II give information on the endpoint variables during the period before the study was initiated. Usually this period was exactly the same period in the preceding winter. If the study period was October to March in one year then the reference period was October to March of the previous year. In 3 cases the reference period was loosely defined as the previous winter or previous year and in 5 cases was not stated. There were 2 common study periods, months 0 to 3 or months 0 to 6, depending on the protocol for the study. From the 28 studies, no data were used from 3 studies. The nonblind study of Belaeva[45] did not report any numerical information and only talked about a halving of the number of infections. Since no further information was available for this study it was excluded from the analysis. The nonblind © Adis International Limited. All rights reserved.
393
study of Shortanbaev[46] reported only data on the duration of infections and duration of antibacterial use but not on the number of infections or the number of antibacterial courses. The randomised study in Portugal of Palma-Carlos et al.,[31] in which 15 patients were treated with ribosomal vaccine and 15 with antibronchitic bacterial vaccine, did not report sufficient data. Only percentage reductions were reported and no information on the number of infections or antibacterial courses. The analysis only considered the ribosomal vaccine or placebo arms of the trials and studies. The comparator arms of the 3 comparative trials were not used. The analysis uses information from the 2 types of studies – randomised, both double blind and nonblind, and nonblind single arm studies. Although the randomised studies had a greater homogeneity of patients as a result of the more stringent entry criteria, they were generally smaller, whereas the nonblind single arm studies tended to be larger but more heterogeneous in terms of the types of infections of the patients. The results are presented separately for the randomised studies only. The analysis is based only upon the response in the study period and changes from the reference period are not used. The reference period information is only used as an explanatory variable in the metaanalysis regression. The primary end-points for this meta-analysis are (i) the number of infections per person in the study period; and (ii) the number of courses of antibacterials prescribed per person during the study period. Data were also available on the duration of the infections, the number of days in bed, the duration of the antibacterial courses, and absenteeism (from school or from work). These responses were not analysed in the meta-analysis as the information was not recorded in a substantial number of studies. Duration of antibacterial use was recorded in 7 studies, absenteeism in 9 studies and the number of days in bed in 8 studies. The complete response data are presented in table III for the number of infections per person and in table IV for the number of antibacterial courses per person. In some studies there is information on BioDrugs 2000 Dec; 14 (6)
394
Boyle et al.
Table III. Number of infections per person (by study period) in each trial Study
Reference
Group
0-6 months mean
Randomised double-blind studies Lacomme & Narcy (1985) 21 Haguenauer et al. (1987)
22
Debas (1989)
23
Garabedian et al. (1990) Guérin et al. (1990)
24 25
Vautel et al. (1993)
26
Hüls et al. (1995)
27
Serrano et al. (1997)
28
Fiocchi & Giovannini (1997)
Randomised nonblind studies Akoun et al. (1988) Palma-Carlos et al. (1990)
29
30 31
0-3 months
3-6 months
SD
mean
SD
mean
SD
NA
RV
3.7
2.5
1.9
1.2
NA
P
5.1
2.8
2.4
1.1
NA
NA
RV
4
2.1
2.2
1.2
1.8
1.7
P
7.8
2.2
4.8
1.4
3.33
1.8
RV
NA
NA
0.4
0.7
NA
NA
P
NA
NA
1.8
0.9
NA
NA
RV
NA
NA
0.7
1.1
NA
NA
P
NA
NA
1.5
1.2
NA
NA 0.9
RV
0.8
1.4
0.4
1.2
0.4
P
1.6
1.4
0.9
1.2
0.8
0.9
RV
3.4
2.1
1.8
1.4
1.6
1.3
P
5.6
2.2
2.9
1.4
2.6
1.3
RV
1.7
1.4
NA
NA
NA
NA
P
2.5
1.8
NA
NA
NA
NA
RV
1
1.1
0.5
0.7
0.5
0.8
P
1.5
1.4
0.8
0.8
0.7
1.1
RV
NA
NA
1.8
1.6
NA
NA
P
NA
NA
2.6
1.5
NA
NA
RV
1.2
1.1
0.9
NA
0.5
NA
KPGE
1.6
1.2
1.1
NA
0.8
NA
RV
58% reduction
ABV
62% reduction
∨
Slapák (1993)
32
RV
0.6
NA
0.3
NA
NA
NA
P
2.3
NA
1.1
NA
NA
NA
Nonblind single arm studies Cotin & Lesbros (1986)
33
RV
2.3
2.6
1.4
1.8
0.9
1.3
Castel (1986)
34
RV
2.3
1.7
1.4
1.1
0.9
1
Legros (1986)
35
RV
2.5
2.3
1.4
1.4
1
1.2
Traissac & Petit (1986)
36
RV
1.9
1.7
NA
NA
NA
NA
Grimfeld (1988)
37
RV
3.3
NA
1.9
NA
1.4
NA
Perruchet & Vautel (1990)
38
RV
2.6
NA
NA
NA
NA
NA
Menardo & Perruchet (1990)
39
RV
2.7
NA
1.7
NA
NA
NA
Sapène et al. (1990)
40
RV
1.8
NA
NA
NA
NA
NA
Huls et al. (1991)
41
RV
5.8
NA
NA
NA
NA
NA
Gonçalves et al. (1993)
42
RV
NA
NA
1.2
NA
NA
NA
Benchakroun et al. (1994)
43
RV
2
1.3
1.2
0.9
0.8
0.8
Alfarro (1995)
44
RV
NA
NA
1.4
NA
1.1
NA
Belaeva (1997)
45
RV
NA
NA
NA
NA
NA
NA
Shortanbaev (1997)
46
RV
NA
NA
NA
NA
NA
NA
Arfa et al. (1997)
47
RV
2.3
NA
1.7
NA
NA
NA
Nascimento (1998)
48
RV
NA
NA
1.8
NA
NA
NA
ABV = antibronchitic bacterial vaccine (Broncho Vaxom®); KPGE = Klebsiella pneumoniae glycoprotein extract (Biostim®); NA = data not available; P = placebo; RV = ribosomal vaccine (Ribomunyl®); SD = standard deviation.
© Adis International Limited. All rights reserved.
BioDrugs 2000 Dec; 14 (6)
Ribosomal Immunostimulant and Respiratory Infection
395
Table IV. Number of antibacterial courses per person (by study period) in each trial Study
Reference
Randomised double-blind studies Lacomme & Narcy (1985) 21 Haguenauer et al. (1987)
22
Debas (1989)
23
Garabedian et al. (1990)
24
Guérin et al. (1990) Vautel et al. (1993)
25 26
Hüls et al. (1995)
27
Serrano et al. (1997)
28
Fiocchi & Giovannini (1997)
Randomised nonblind studies Akoun et al. (1988) Palma-Carlos et al. (1990)
29
30 31
Group
0-6 months
0-3 months
3-6 months
mean
mean
mean
SD
SD
SD
RV
2.3
NA
NA
NA
NA
NA
P
3.5
NA
NA
NA
NA
NA NA
RV
2
NA
1.3
NA
0.7
P
5.6
NA
3.5
NA
2.2
NA
RV
NA
NA
NA
NA
NA
NA
P
NA
NA
NA
NA
NA
NA
RV
NA
NA
0.7
1.1
NA
NA
P
NA
NA
1.5
1.2
NA
NA
RV
0.7
1.2
0.3
1
0.4
0.8
P
1.3
1.2
0.7
1
0.6
0.8
RV
1.3
2
NA
NA
NA
NA
P
3.3
2
NA
NA
NA
NA
RV
NA
NA
NA
NA
NA
NA
P
NA
NA
NA
NA
NA
NA
RV
0.8
1.3
NA
NA
NA
NA
P
1.3
1.6
NA
NA
NA
NA
RV
NA
NA
0.4
0.8
NA
NA
P
NA
NA
0.6
1
NA
NA
RV
NA
NA
NA
NA
0.4
NA
KPGE
NA
NA
NA
NA
0.7
NA
RV
NA
NA
NA
NA
NA
NA
ABV
NA
NA
NA
NA
NA
NA
RV
NA
NA
NA
NA
NA
NA
P
NA
NA
NA
NA
NA
NA
∨
Slapák (1993)
32
Nonblind single arm studies Cotin & Lesbros (1986)
33
RV
1.6
1.9
0.9
1.1
0.6
1.1
Castel (1986)
34
RV
1.2
1.3
0.8
0.9
0.5
0.8
1.6
0.7
1
0.5
0.9
Legros (1986)
35
RV
1.2
Traissac & Petit (1986)
36
RV
very much diminished
Grimfeld (1988)
37
RV
3
NA
1.7
NA
1.3
NA
Perruchet & Vautel (1990)
38
RV
1.7
NA
NA
NA
NA
NA
Menardo & Perruchet (1990)
39
RV
2.4
NA
1.4
NA
NA
NA
Sapène et al. (1990)
40
RV
1.7
NA
NA
NA
NA
NA
Huls et al. (1991)
41
RV
NA
NA
NA
NA
NA
NA
Gonçalves et al. (1993)
42
RV
NA
NA
1.1
NA
NA
NA
Benchakroun et al. (1994)
43
RV
1.5
NA
0.9
NA
0.6
NA
Alfarro (1995)
44
RV
NA
NA
1.7
NA
1.3
NA
Belaeva (1997)
45
RV
NA
NA
NA
NA
NA
NA
Shortanbaev (1997)
46
RV
NA
NA
NA
NA
NA
NA
Arfa et al. (1997)
47
RV
1.8
NA
1.4
NA
NA
NA
Nascimento (1998)
48
RV
NA
NA
1.4
NA
NA
NA
ABV = antibronchitic bacterial vaccine (Broncho Vaxom®); KPGE = Klebsiella pneumoniae glycoprotein extract (Biostim®); NA = data not available; P = placebo; RV = ribosomal vaccine (Ribomunyl®); SD = standard deviation.
© Adis International Limited. All rights reserved.
BioDrugs 2000 Dec; 14 (6)
396
both response periods. In others, data were only presented for one of these periods. In some studies data were presented for months 0 to 3 and months 3 to 6, and here the month 0 to 6 response was obtained by addition. In order to carry out the meta-analysis it is necessary to have an estimated effect for each study and an estimate of the standard error of this effect, for example the mean number of infections in the study period of 0 to 6 months and the standard deviation of the number of infections in this period. Full information was not available for all studies, although most had some information. The general strategy for dealing with this missing information was the use of all the available information and imputing values from the other similar studies when information was not available. Although this strategy is sometimes used, it is one of the unsatisfactory aspects of meta-analyses of published data. It was not possible to access the original data, which for many of these studies was not available. We stress that there was never any imputation on the response variables, i.e. the mean number of infections or antibacterial courses in the study period. There was only some imputation of the mean response in the reference period and in the response period standard deviations. The effect of these imputations were assessed by a sensitivity analysis. The estimation of the effects from the metaanalysis used a general methodology.[49] This is a multilevel or hierarchical model,[50] including random effects for the studies. The parameter estimates for this model were estimated using the Mln software for hierarchical modelling.[51] Full details are given in the Statistical Appendix. Results Findings will be presented separately for the number of infections and the number of antibacterial courses prescribed. Number of Infections
The data used in the meta-analysis of the number of infections per person are presented in table © Adis International Limited. All rights reserved.
Boyle et al.
V. The analysis was based upon data from 25 studies, amounting to 36 treatment arms. The mean number of infectious episodes per person in the study period are plotted in figure 1. There is considerable variability over the trials, and individual trials do not have precise estimates as the confidence intervals for the smaller trials are very wide. From the summary measures for the fixed effects meta-analysis (figure 2), it can be seen that there are fewer infections (i) when ribosomal vaccine is used and (ii) when the study period is only 3 months. Within the ribosomal vaccine arms there are fewer infections in the randomised studies compared with the nonblind single arm ones. The difference between a 3- and 6-month response period is greater among the nonblind single arm studies. The parameter estimates are presented in table VI. Results for 3 analyses are presented, one using all the studies and the second using only those studies for which there was a value recorded for the mean number of infections in the reference period. The third used only the results of the randomised studies. The estimated coefficients from the first 2 analyses are virtually the same, so the imputation of the mean number of infections in the reference period over all studies into studies for which this is not recorded appears to be reasonable. This was only necessary in 2 studies with 255 patients in total, so no major influence is likely. From the model based upon all the studies there was no significant difference between the randomised studies and the nonblind single arm studies, nor was the comparison of ribosomal vaccine with placebo different in 3-month compared with 6month studies, and these terms were omitted from the full model. The mean number of infections in the study period of 3 months using placebo is 2.39 (standard error ± 0.50) infections. If ribosomal vaccine is used then one would expect 0.97 (±0.37) infections. If the study period was 6 months rather than 3 months, then one would expect 3.35 (±0.41) infections with placebo and only 1.92 (±0.25) infections with ribosomal vaccine. Thus, the use of ribosomal vaccine is associated with a reduction of 1.43 (±0.26) [p < 0.001] infections compared with BioDrugs 2000 Dec; 14 (6)
Ribosomal Immunostimulant and Respiratory Infection
397
Table V. Data used in the meta-analysis for the number of infections per person (values in italics denote imputed values) Reference
Response period (mo)
Randomised double-blind studies Lacomme & Narcy (1985) 21 Haguenauer et al. (1987) Debas (1989) Garabedian et al. (1990) Guérin et al. (1990) Vautel et al. (1993) Hüls et al. (1995) Serrano et al. (1997) Fiocchi & Giovannini (1997)
22 23 24 25 26 27 28 29
Randomised nonblind studies Akoun et al. (1988) 30
Group
Baseline mean
Response mean
6
RV
6.5
3.7
2.5
6
P
6.5
5.1
2.8
6
RV
6.9
4.0
2.1
6
P
6.5
7.8
2.2
3
RV
7.4
0.4
0.7
3
P
7.4
1.8
0.9
3
RV
5.5
0.7
1.1
3
P
5.5
1.5
1.2
6
RV
5.5
0.8
1.4
6
P
5.5
1.6
1.4
6
RV
5.9
3.4
2.1
6
P
6.1
5.6
2.2
6
RV
6.3
1.7
1.4
6
P
6.0
2.5
1.8
6
RV
4.4
1.0
1.1
6
P
4.3
1.5
1.4
3
RV
3.0
1.8
1.6
3
P
3.0
2.6
1.5
6
RV
5.9
1.2
1.1
32
6
RV
3.0
0.6
2.1
32
6
P
1.3
2.3
2.1
∨
Slapák (1993) ∨
Slapák (1993)
Response SD
Nonblind single arm studies Cotin & Lesbros (1986) 33
6
RV
6.6
2.3
2.6
Castel (1986)
6
RV
6.2
2.3
1.7
34
Legros (1986)
35
6
RV
6.1
2.5
2.3
Traissac & Petit (1986)
36
6
RV
6.4
1.9
1.7
Grimfeld (1988)
37
6
RV
7.6
3.3
2.1
Perruchet & Vautel (1990)
38
6
RV
5.3
2.6
2.1
Menardo & Perruchet (1990)
39
6
RV
5.0
2.7
2.1
Sapène et al. (1990)
40
6
RV
5.0
1.8
2.1
Huls et al. (1991)
41
6
RV
8.3
5.8
2.1
Gonçalves et al. (1993)
42
3
RV
4.5
1.2
1.5
Benchakroun et al. (1994)
43
6
RV
6.0
2.0
1.3
Alfarro (1995)
44
3
RV
5.5
1.4
1.5
Arfa et al. (1997)
47
6
RV
6.9
2.3
2.1
Nascimento (1998)
48
3
RV
5.6
1.8
1.5
P = placebo; RV = ribosomal vaccine (Ribomunyl®); SD = standard deviation.
placebo. A separate analysis which permitted different effects of ribosomal vaccine in the 3-month response period compared with a 6-month response period was carried out. In the 3-month response period, ribosomal vaccine was associated with © Adis International Limited. All rights reserved.
1.23 (±0.45) [p < 0.001] fewer infections, whereas in the 6-month response period the reduction is 1.51 (±0.31) [p = 0.006] infections. All of these values relate to a reference period with 5 infections. For every infection over 5 in the reference BioDrugs 2000 Dec; 14 (6)
398
Boyle et al.
NB RV 3mo 42 44 48 NB RV 6mo 33 34 35 36 37 38 39 40 41 43 47 RCT RV 3mo 23 24 29 RCT P 3mo 23 24 29 RCT RV 6mo 21 22 25 26 27 28 30 32 RCT P 6mo 21 22 25 26 27 28 32 0
2
4
6
8
Mean number of infections
Fig. 1. Mean number of infections and 95% confidence intervals for all studies. The data for the individual studies are plotted on the same level as the reference number. The summary estimates for the 6 subgroups are plotted on the same level as the subgroup name. NB = nonblind single arm study; P = placebo; RCT = randomised controlled trial; RV = ribosomal vaccine (Ribomunyl®).
period, one would expect an increase in the number of infections in the study period. This figure is only applicable provided there are 3 or more infections in the reference period as this was the minimum inclusion criterion for the studies. There was no evidence of any interaction between the study period © Adis International Limited. All rights reserved.
or placebo increment and the mean number of infections in the reference period. There was no evidence of any influence of age, gender or year of the study on the number of infections. For clinicians, the diagnosis is important in the assessment. Although this was recorded in all BioDrugs 2000 Dec; 14 (6)
Ribosomal Immunostimulant and Respiratory Infection
studies, frequently a single patient had multiple diagnoses and it was not possible to assess the effects of the underlying conditions of the patients. The results from only the randomised trials are a little different in that there was no significant effect of response period or baseline number of infections. The mean number of infections with ribosomal vaccine or placebo was only slightly changed from the values reported in all studies and the estimated reduction in the number of infections using ribosomal vaccine compared to placebo is 1.35 (±0.27) over both 0-3 months and 0-6 months. The variance over the studies in the effects of ribosomal vaccine and placebo was greater among only the randomised studies, and this is in accord with the greater variation among the randomised studies compared with the nonblind single arm studies seen in figure 1. This model is a random effects model, and the variances of the ribosomal vaccine and placebo effects are both large relative to their standard errors. There is certainly more variation over studies in the responses to placebo than the responses to ribo-
399
somal vaccine. This implies that there is a great deal of variation in the mean numbers of infectious episodes in the response periods over the studies that cannot be explained by known differences in the studies which have already been taken into account. Number of Antibacterial Courses
The data used in the meta-analysis of the number of courses of antibacterials per person are presented in table VII. The analysis was based upon data from 19 studies, amounting to 26 observations. There were 15 studies with the mean number of antibacterial courses in the reference period recorded. There were 461 patients in the 4 studies for which the mean number of antibacterial courses in the reference period was imputed. These were all randomised trials in which the main outcome was a comparison of placebo with ribosomal vaccine in the study period. A plot showing the mean number of antibacterial courses in the study period is shown in figure
NB RV 3mo
NB RV 6mo
RCT RV 3mo
RCT P 3mo
RCT RV 6mo
RCT P 6mo
1
1.5
2
2.5
3
Mean number of infections
Fig. 2. Summary of meta-analysis of the number of infections. NB = nonblind single arm study; P = placebo; RCT = randomised controlled trial; RV = ribosomal vaccine (Ribomunyl®).
© Adis International Limited. All rights reserved.
BioDrugs 2000 Dec; 14 (6)
400
Boyle et al.
Table VI. Meta-analysis results for infectious episodes per person Effect
All studies estimate
SE
Reference infections recordeda
Randomised studies only
estimate
estimate
SE
SE
Fixed effects Ribosomal vaccine (Ribomunyl®)
0.972
0.365
1.032
0.408
1.312
0.568
Placebo
2.393
0.503
2.559
0.568
2.660
0.732
Response periodb
0.949
0.403
0.957
0.447
0.640
0.602
Reference period infectiousc
0.155
0.151
0.197
0.155
NA
NA
Reference period infections squared
0.145
0.058
0.132
0.062
NA
NA
Random effect variances Ribosomal vaccine
0.805
0.229
0.780
0.232
1.439
0.622
Covariance
1.233
0.406
1.167
0.433
2.146
0.956
Placebo
2.401
0.865
2.427
1.001
3.580
1.563
a
Only studies with the mean number of infections in the reference period recorded.
b
The coefficient gives the effect of a 6-month response period compared to a 3-month response period.
c
The coefficient gives the effect of the number of infections in the reference period.
NA = not available; SE = standard error.
3. There is more variation among the randomised trials than there is among the nonblind single arm trials and this is partially, but not completely, due to the size of the trials. There is a general trend for the number of antibacterial courses in the study period to increase as the number of antibacterial courses used in the reference period increases (data not shown), and a linear term for the number of antibacterial courses used is also included in the model. The fixed effect meta-analysis estimates are shown in figure 4. Within the nonblind single arm studies, there is greater use of antibacterial courses in 6 months compared with 3 months. The same tendency is visible in the randomised trials, but as these studies tend to be smaller there is considerable uncertainty in the estimated effects. These show that antibacterial use in the randomised trials is lower than in the nonblind single arm trials for those patients receiving ribosomal vaccine. First, the results for all studies are discussed separately for the two response periods, 3 or 6 months. An average of 2.09 (±1.48) antibacterial courses were used among individuals using a placebo for 3 months and 1.45 (±0.40) courses among individuals using ribosomal vaccine for 3 months. In a 6month trial period, 3.37 (±0.93) antibacterial courses were used by individuals on placebo compared with 1.78 (±0.25) on ribosomal vaccine. The © Adis International Limited. All rights reserved.
reduction in the number of antibacterial courses used on ribosomal vaccine compared with placebo was greater at 6 months relative to 3 months, but this was not a statistically significant difference (p = 0.25). There was a slight increase in the number of antibacterial courses used in a 6-month period compared with a 3-month period, but again this difference was not significant. The estimated effect was 0.29 (±0.22) [p = 0.18], and consequently study duration was not included in the model presented in table VIII. There was evidence, however, of a reduction of 0.44 (±0.20) in the number of antibacterial courses used per person in randomised trials compared with nonblind single arm trials over both 0-3 months and 0-6 months. With both response periods combined, individuals receiving placebo had an average of 3.02 (±0.44) antibacterial courses per person, whereas use of ribosomal vaccine was associated with an average of 1.70 (±0.12) antibacterial courses used per person (table VIII). This corresponds to a reduction of 1.32 (±0.42) [p = 0.002] antibacterial courses per person using ribosomal vaccine compared with placebo. All of these values are relevant to a reference period in which there were 5 antibacterial courses prescribed. An additional antibacterial course in the reference period is associated with an increase of 0.47 (±0.13) in the mean numBioDrugs 2000 Dec; 14 (6)
Ribosomal Immunostimulant and Respiratory Infection
ber of antibacterial courses in the study period. No study reported a mean number of antibacterial courses lower than 3 and the rate of increase is only applicable provided there are 3 or more antibacterial courses in the reference period. When the 4 studies with no information on the mean number of antibacterial courses used in the reference period are omitted, then the estimated difference between the randomised and nonblind studies changes considerably (table VIII). These 4 randomised studies all have a very low recorded use of antibacterial courses in the study period and their omission means that there is no longer any significant difference between the randomised and nonblind single arm studies. In this analysis the esti-
401
mated reduction in the number of antibacterial courses associated with ribosomal vaccine is 2.04 (±0.73) [p = 0.005]. This is slightly greater than when all studies are included, but the standard error is much larger as much information is lost when these 4 randomised studies are excluded. There is a slight reduction in the variance of the ribosomal vaccine and placebo effects over the studies compared with the analysis with all studies, as the 4 omitted studies are at one extreme of the distribution of antibacterial use and the studies remaining are more homogeneous. When only the randomised trials are included in the analysis (table VIII), there was no influence of the mean number of antibacterial courses used in
Table VII. Data used in the meta-analysis for the number of antibacterial courses per person (values in italics denote imputed values) Reference
Response period (mo) Group
Randomised double-blind studies Lacomme & Narcy (1985) 21 Haguenauer et al. (1987) Garabedian et al. (1990) Guérin et al. (1990) Vautel et al. (1993) Serrano et al. (1997) Fiocchi & Giovannini (1997)
22 24 25 26 28 29
Baseline mean
Response mean Response SD
6
RV
5.0
2.3
1.7
6
P
5.0
3.5
1.7
6
RV
6.0
2.0
1.7
6
P
5.2
5.6
1.7
3
RV
3.6
0.7
1.1
3
P
3.4
1.5
1.2
6
RV
4.8
0.7
1.2
6
P
4.8
1.3
1.2
6
RV
4.8
1.3
2.0
6
P
4.8
3.3
2.0
6
RV
4.8
0.8
1.3
6
P
4.8
1.3
1.6
3
RV
4.8
0.4
0.8
3
P
4.8
0.6
1.0
Nonblind single arm studies Cotin & Lesbros (1986)
33
6
RV
5.0
1.6
1.9
Castel (1986)
34
6
RV
4.8
1.2
1.3
Legros (1986)
35
6
RV
3.9
1.2
1.6
Grimfeld (1988)
37
6
RV
6.8
3.0
1.7
Perruchet & Vautel (1990)
38
6
RV
4.3
1.7
1.7
Menardo & Perruchet (1990)
39
6
RV
5.0
2.4
1.7
Sapène et al. (1990)
40
6
RV
4.9
1.7
1.7
Gonçalves et al. (1993)
42
3
RV
4.0
1.1
1.0
Benchakroun et al. (1994)
43
6
RV
5.1
1.5
1.7
Alfarro (1995)
44
3
RV
5.5
1.7
1.0
Arfa et al. (1997)
47
6
RV
5.9
1.8
1.7
Nascimento (1998)
48
3
RV
4.6
1.4
1.0
P = placebo; RV = ribosomal vaccine (Ribomunyl®); SD = standard deviation.
© Adis International Limited. All rights reserved.
BioDrugs 2000 Dec; 14 (6)
402
Boyle et al.
NB RV 3mo 42 44 48 NB RV 6mo 33 34 35 37 38 39 40 43 47 RCT RV 3mo 24 29 RCT P 3mo 24 29 RCT RV 6mo 21 22 25 26 28 RCT P 6mo 21 22 25 26 28
1
2
3
4
5
6
Mean number of antibacterial courses Fig. 3. Mean number of antibacterial courses and 95% confidence intervals for all studies. The data for the individual studies are plotted on the same level as the reference number. The summary estimates for the 6 subgroups are plotted on the same level as the subgroup name. NB = nonblind single arm study; P = placebo; RCT = randomised controlled trial; RV = ribosomal vaccine (Ribomunyl®).
the reference period or of the length of the study period. The mean number of antibacterial courses used by individuals on placebo is 2.45 (±0.62), whereas those on ribosomal vaccine used 1.19 (±0.26) antibacterial courses. The estimated number of antibacterial courses was smaller in both the placebo and ribosomal vaccine arms compared with the analysis on all studies. However, the reduction in the mean number of antibacterial courses associated with ribosomal vaccine use was virtually the same at 1.26 (±0.40) [p = 0.002] antibacterial courses per person as it was in all studies, © Adis International Limited. All rights reserved.
where the reduction was 1.32 (±0.40) antibacterial courses per person. As in the case of the number of infections, there was greater variation over the studies in the placebo arm compared with the ribosomal vaccine arm. There was also greater variation among the randomised studies than there was among the nonblind single arm studies. Discussion Interpretation of these study results must be preceded by discussions of the data and the data qualBioDrugs 2000 Dec; 14 (6)
Ribosomal Immunostimulant and Respiratory Infection
ity. First of all, it is essential that the definitions employed in each of the studies are comparable across all studies. Even when this was done, there were still outstanding issues regarding the quality of the data. These are not technical issues relating to the statistical methods but are related to the availability of the data. There were many instances where individual study estimates of variability were unavailable, and this is a serious weakness of the work. The estimates rest upon an assumption that the imputed standard deviations are reasonable for each study. In table III and table IV the published individual standard deviations are all comparable, and as the imputed values are based upon a weighted average the imputed value is comparable with the observed values. The imputed standard deviations are 2.1 and 1.5 for the number of infections and 1.7 and 1.0 for the number of antibacterial courses, for 6and 3-month responses, respectively. The responses are the study means and the precision of the mean depends on the sample size as well as on the standard deviation. The sample sizes are all known
403
and as they are mostly large there is not much influence of different imputed standard deviations on the estimated effects. Our conclusions are robust to variation in the imputed values for the standard deviations. This has also been demonstrated by a multiple imputation analysis. The results were not presented as they were similar to those in the tables but are available from the authors. There was a wide variation in the size of the individual studies, ranging from 30 to 2021 participants. As stated above, the data quality in the studies was not uniform and many studies published mean values without any measures of standard deviation. Some studies did not quote the baseline mean values in the publications, although more information was available from the statistical reports of the studies. Bearing this in mind, it is a considerable strength of this meta-analysis that data from all available clinical trials of ribosomal vaccine have been brought together for this analysis. In the statistical analysis, there was no evidence of any age-related, gender-related or temporal events that could affect the study. The between-
NB RV 3mo
NB RV 6mo
RCT RV 3mo
RCT P 3mo
RCT RV 6mo
RCT P 6mo
0.5
1
1.5
2
2.5
Mean number of antibacterial courses
Fig. 4. Summary of meta-analysis of the number of antibacterial courses. NB = nonblind single arm study; P = placebo; RCT = randomised controlled trial; RV = ribosomal vaccine (Ribomunyl®).
© Adis International Limited. All rights reserved.
BioDrugs 2000 Dec; 14 (6)
404
Boyle et al.
Table VIII. Meta-analysis results for antibacterial courses per person Effect
All studies estimate
SE
Reference courses recordeda
Randomised studies only
estimate
SE
estimate
SE
Fixed effects Ribosomal vaccine (Ribomunyl®)
1.698
0.122
1.701
0.097
1.188
0.261
Placebo
3.018
0.439
3.745
0.793
2.448
0.618
Randomised studyb
–0.441
0.203
0.053
0.224
NA
NA
Reference period antibacterialsc
0.473
0.129
0.583
0.105
NA
NA
Random effect variances Ribosomal vaccine
0.176
0.057
0.113
0.042
0.471
0.254
Covariance
0.360
0.201
–0.081
0.248
0.999
0.597
Placebo
1.783
0.902
1.564
1.286
2.653
1.422
a
Only studies with the mean number of antibacterial courses in the reference period recorded.
b
The coefficient gives the effect of a randomised study compared with a nonblind study.
c
The coefficient gives the effect of the number of antibacterial courses in the reference period.
NA = not applicable; SE = standard error.
study variances of the effects of ribosomal vaccine and placebo are both large relative to their standard errors. This implies that there is variation in the mean numbers of infectious episodes and the mean numbers of antibacterial courses used over the studies that could not be explained in this analysis. Some known factors contributed to this variation, such as the length of the study, in the case of infections, and the study type, in the case of antibacterial courses, but even allowing for these effects there is still some heterogeneity across the studies. One possible explanation for this excess variation could be related to the lack of similarity of the diagnostic groups enrolled in the studies – otorhinolaryngological or bronchial infections such as rhinitis are relatively easy to treat and bronchitis relatively difficult. Also, the randomised studies tended to focus on different but more homogeneous patient groups, whereas the nonblind single arm studies were larger but used more heterogeneous patients within each study. Different studies also had more or less severe inclusion criteria such as greater than 3 or 5 infections in the previous winter. Some studies only used patients with otorhinolaryngological infections, whereas others, especially the nonblind single arm studies, used a wider mixture of patients. There may also be differences among the countries in the prescribing of antibacterials.[52,53] © Adis International Limited. All rights reserved.
The strengths of this analysis are that all available data have been employed in the analysis and that the differences found are strong and statistically significant. The use of the nonblind single arm studies contributed to the extrapolation of results to a situation that is more in line with clinical practice than a randomised trial. We have investigated if there were systematic differences between the randomised and nonblind studies and have documented the cases where such differences existed. Similar results were obtained whether the data are analysed comparing ribosomal vaccine with placebo using all studies or whether only the data obtained from randomised studies of ribosomal vaccine versus placebo were used. This is so because the randomised studies were the only studies that provided information on the comparison of ribosomal vaccine with placebo. The nonblind single arm studies provided information on the effect of ribosomal vaccine and contributed to the estimation of the effects of the response period and the baseline number of infections. As there were more patients in the nonblind studies, there was more precision in the estimated effects when all studies were included. At no stage were contradictory estimates obtained from the randomised studies compared with the nonblind single arm studies. BioDrugs 2000 Dec; 14 (6)
Ribosomal Immunostimulant and Respiratory Infection
Conclusions In patients with recurrent respiratory infections, ribosomal vaccine was found to be associated not only with a reduction in the mean number of antibacterial courses used but also a reduction in the mean number of infections in the study period. In a 6-month period, we estimate a frequency of 3.35 (±0.41) infections and 3.02 (±0.44) antibacterial courses with placebo but only 1.92 (±0.25) infections and 1.70 (±0.12) antibacterial courses with ribosomal vaccine. Given all of the above considerations, it is possible to conclude from this analysis that ribosomal vaccine significantly reduces the number of infections compared with placebo. Ribosomal vaccine also significantly reduces the number of courses of antibacterials prescribed. Thus, ribosomal vaccine use appears to have significant activity in reducing the risk of infections and the necessity for antibacterials. APPENDIX: Statistical Analysis It is necessary to have an estimated effect for each study and an estimate of the standard error of this effect. For those studies publishing the mean and standard deviation of the number of infections during months 0 to 6, these quantities were used. For studies publishing the mean number of infections during months 0 to 6 but not the standard deviation, then a value of the standard deviation was imputed from a weighted average of the standard deviations in the studies reporting standard deviations. The weights were the sample sizes so that large studies contributed more weight than small studies. There are then still a number of studies for which there is no response information for months 0 to 6 but there is response information for months 0 to 3. This was used in the analysis with an indicator to note the use of a shorter time period. In this way there is a response for all studies. Imputation of the standard deviation was necessary in 10 of the 25 studies contributing to the analysis of the number of infections, and in 11 of the 19 studies contributing to the analysis of the number of antibacterial courses. © Adis International Limited. All rights reserved.
405
Similar procedures were used for the baseline mean number of infections in the previous winter. If this was missing then a value was imputed from the weighted average of all studies reporting baseline information. Although pragmatic, this is not an ideal solution. This imputation was used in 2 studies when analysing the number of infections, and in 4 studies for antibacterial use. We stress that there was never any imputation on the response variables, the mean number of infections or antibacterial courses in the study period. There was only some imputation of the mean response in the reference period and in the response period standard deviations. The effect of these imputations were assessed by a sensitivity analysis. The estimation of the effects from the metaanalysis followed the model proposed by Stram.[49] If ybijk denotes the baseline observation for individual i in study j in treatment group k and yeijk is the corresponding value at the end of the study, then _ _ yejk denotes the mean end-of-study value and ybjk is the corresponding baseline mean. If these data were always available, then we would be able to calculate the mean reduction in the number of infections compared with the previous reference period. This cannot be calculated for many of the studies here, as no baseline information was _ collected. Instead, the end-of-study response yejk is used as the response from the arms of the trials. This was the response variable used in all the double-blind and comparative studies. It was also the variable reported in the nonblind single arm studies. The standard error of this effect is given by: sjk =
⎯⎯ √
2
sejk nejk
where s2 ejk is the variance of the values in the study period in treatment group k in study j, and nejk is the corresponding number of individuals. There are corresponding values for the beginning of the study for some studies but these were not used. BioDrugs 2000 Dec; 14 (6)
406
Each arm of the studies is summarised by the effect, and its standard error and the simplest random effects meta-analysis model is written as: _ yejk = μ + τR + uj + εjk
where μ represents the overall mean effect, R is an indicator variable taking the value 1 for ribosomal vaccine (Ribomunyl®) and 0 for placebo, τ represents the effect of ribosomal vaccine compared to placebo, uj represents the random effect of study j and εjk represents the sampling variability of the effect. The standard deviation of the latter term is assumed to be known and is equal to sjk. This model is a multilevel or hierarchical model,[50] and the parameter estimates for this model were estimated using the Mln software for hierarchical modelling.[52] The multilevel model can be written as a 2-level model, where the first level is: _ yejk = μj + τR + εjk
and the second is: μ j = μ + uj
Within this framework it can be seen that the second level model can be extended to take into account study variables that may influence the average study effect in a systematic way. Such variables include the length of the study, the average age of the patients, the percentage of the patients who were male and also the baseline response variable in the previous reference period. Also included are the systematic differences between the trials, such as nonblind single arm versus randomised and a 3- or 6-month response period. This leads to a second level model, which is: μj = μ + βcj +uj
where β measures the effect of the study level covariate cj on the study effect, μj. The full meta-analysis model used is: _ yejk
= μj = + τjR + β1S + β2S + β3SR + β4LR + β5B + β6B2 + εjk
μj = μ + u0j © Adis International Limited. All rights reserved.
Boyle et al.
τj = τ + u1j
where R is a dummy variable taking the value 0 for placebo and 1 for ribosomal vaccine, S is a dummy variable taking the value 0 for a nonblind single arm study and 1 for a randomised study, L is a dummy variable taking the value 0 for a study of 3 months duration and 1 for a study of 6 months duration. The 2 product terms SR and LR give dummy variables also taking the value 1 for ribosomal vaccine in randomised studies and ribosomal vaccine in 6-month studies, respectively. B is the baseline mean and B2 is the baseline mean squared. These 2 terms allow the response to depend upon the background level of infections in the patients. The quadratic term is required because the relationship is slightly curved. The treatment effect varies randomly over the studies. This means that the estimated effects of placebo and ribosomal vaccine can vary over studies. The variability in these effects may be different. Acknowledgements This work was conducted within the framework of support from the Associazione Italiana per la Ricerca sul Cancro (Italian Association for Cancer Research).
References 1. Mestecky J, Abraham R, Ogra PL. Common mucosal immune system and strategies for the development of vaccines. In: Ogra PL, Metecky H. Handbook of mucosal immunology. London: Academic Press 1995; 357-72 2.. McGhee JR, Mestecky J, Dertzbaugh MT, et al. The mucosal immune system: from fundamental concepts to vaccine development. Vaccine 1992; 10: 75-80 3. Turner RB. The epidemiology, pathogenesis, and treatment of the common cold. Semin Pediatr Infect Dis 1995; 6: 57-61 4. Paradise JL, Rockette HE, Colburn DK, et al. Otitis media in 2253 Pittsburgh-area infants: prevalence and risk factors during the first two years of life. Pediatrics 1997; 99: 318-33 5. International Rhinitis Management Working Group. International consensus report on the diagnosis and management of rhinitis. Allergy 1994; Suppl. 19: 1-34 6. Youmans AS, Youmans GP. Immunogenic activity of a ribosomal fraction obtained from Mycobacterium tuberculosis. J Bacteriol 1965; 89: 1291-8 7. Nikolaeva LV, Savlev EP. Molecular, biological and immunological properties of ribosomal vaccines. Biomed Sci 1991; 2: 1-10 8. Levenson VJ, Mallett CP, Hale TL. Protection against local Shigella sonnei infection in mice by parenteral immunisation
BioDrugs 2000 Dec; 14 (6)
Ribosomal Immunostimulant and Respiratory Infection
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19. 20. 21.
22.
23.
24.
with a nucleoprotein subcellular vaccine. Infect Immun 1995; 63: 2762-5 Faure GC, Béné MC, Simon C, et al. Increase in specific antibody-forming cells in human tonsils after oral stimulation with D-53, a ribosomal vaccine. Int J Immunopharmacol 1990; 12: 315-20 Zanin C, Perrin P, Perruchet AM, et al. Antibody-producing cells in peripheral blood and tonsils after oral treatment of children with bacterial ribosomes. Int J Immunopharmacol 1994; 16: 497-505 Kolopp-Sarda MN, Béné MC, Allaire JM, et al. Kinetics of specific IgA secretory response in human saliva after oral challenge by a ribosomal immunostimulant. Int J Immunopharmacol 1997; 19 (3): 181-6 Allavena P, Erroi A, Pirelli A, et al. Stimulation of cytotoxic and non-cytotoxic functions of Natural Killer cells by bacterial membrane proteoglycans and ribosomes. Int J Immunopharmacol 1989; 11 (1): 29-34 Lafont S, Millet I, Kouassi E, et al. Induction of murine B cell proliferation and immunoglobulin synthesis by some bacterial ribosomes. Microbiol Immunol 1988; 32: 1043-58 Millet I, Lafont S, Jeannin M, et al. Proliferative response of human T lymphocytes to a vaccinal preparation of ribosomes from Streptococcus pyogenes. Int Arch Allergy Appl Immunol 1988; 86: 432-5 Roques C, Frayret MN, Luc J, et al. Effect of an in vivo immunostimulant treatment on PMN functions: interactions with antibiotics in vitro. Int J Immunopharmacol 1991; 13 (8): 1051-57 Pujol JL, Klein B, Godard P, et al. Bacterial ribosomal immunostimulants prime alveolar macrophages in vivo to produce interleukin 1 in vitro. Chest 1991; 100: 644-5 Sironi M, Sica A, Riganti F, et al. Interleukine 6 gene expression and production induced in human monocytes by membrane proteoglycans from klebsiella pneumoniae. Int J Immunopharmacol 1990; 12 4: 397-402 Luini W, De Rossi M, Licciardello L, et al. Chemotactic cytokine gene expression and production induced in human monocytes by membrane proteoglycans from Klebsiella pneumoniae. Int J Immunopharmacol 1991; 13 (6): 631-7 Clot J. Pharmacology of ribosomal immunotherapy. Drugs 1997; 54 Suppl. 1: 33-6 Anthonisen NR. OM-85 BV for COPD [editorial]. Am J Respir Crit Care Med 1997; 156: 1713-4 Lacomme Y, Narcy P. Prévention par immunothérapie ribosomale des épisodes de surinfection récidivantes de la sphère ORL chez l’enfant. Résultats cliniques d’une étude multicentrique. Immunologie Med 1985; 11: 73-5 Haguenauer JP. Prévention des épisodes infectieux récidivants de la sphère ORL par D53 comprimés chez l’enfant de moins de 5 ans. Immunologie Médicale 1987; 18: 36-9 Debas J. Relapsing tonsillitis prevention with a ribosomal vaccine. Conference Proceedings, 1989 Sep 10-15; Madrid, Spain Garabedian EN, Dubreuil C, Triglia JC. Effectiveness and tolerance of Ribomunyl tablets in preventing middle ear infections in children affected by S.O.M. [poster and abstract] International Congress on Prevention of Infection, 1990 June; Nice, France
© Adis International Limited. All rights reserved.
407
25. Guérin JC, Léophonte P, Muir JF, et al. Prevention of bronchial superinfections in chronic bronchitic patients with Ribomunyl tablets [poster and abstract]. Congrès International Prévention Infection Nice 1990 [poster and abstract] 26. Vautel JM, Cauquil J, Perruchet AM, et al. Prevention of recurrent ear, nose and throat infections in young children with Ribomunyl: a double blind, placebo-controlled study. Curr Ther Res 1993; 53 (6): 722-9 27. Hüls P, Welbers P, Lindemann H. Immuntherapie bei kindern mit hilfe eines multibakteriellen ribosomenpräparates. Der Kinderarzt 1995; 26 (8): 1018-24 28. Serrano E, Morgon A, Demanez JP, et al. Effectiveness of ribosomal fractions of Klebsiella pneumoniae, Streptococcus pneumoniae, Streptococcus pyogenes, Haemophilus influenzae and the membrane fraction of Kp (Ribomunyl) in prevention of clinical recurrences of infectious rhinitis. Results of a multicenter double blind placebo-controlled study. Int Arch Otorhinolaryngol 1997; 254: 372-5 29. Fiocchi A, Giovannini M. Clinical evaluation of ribosomal immunotherapy in recurrent respiratory infections: results of a meta-analysis. Drugs 1997; 54: 45 30. Akoun G, Arnaud A, Blaive B. Efficacité comparée de 2 immunostimulants dans la prévention des épisode infectieux bronchiques chez le bronchopathe chronique. Immunologie Médicale 1988; 5, 3: 135-8 31. Palma-Carlos G, Trindade MF, Azevedo-Conde T, et al. Oral non specific immunotherapy with bacterial extract versus ribosomal extracts adjuved with membrane proteoglycans clinical efficacy. Clin Exp Allergy 1990; 20 Suppl. 1: 126 32. Šlapák I. Klinicke zkusenosti pouziti Ribomunylu u recidivujicich ORL onemocneni. Bulletin Saad-Sdruzeni Pro Alergické a Astmatické Deti c 1993; 2: 1-14 33. Cotin G, Lesbros D. Prévention des récidives infectieuses des voies respiratoires par Ribomunyl comprimés. Immunologie Médicale 1986; 14: 28-33 34. Castel Y. Prévention des récidives infectieuses des voies respiratoires chez l’enfant par Ribomunyl comprimés. Immunologie Médicale 1986; 14: 34-40 35. Legros M. Prévention des récidives infectieuses de la sphère O.R.L. par Ribomunyl comprimés. Immunologie Médicale 1986; 14: 41-6 36. Traissac L, Petit J. Traitement préventif des infections O.R.L. récidivantes par Ribomunyl comprimé. Immunologie Médicale 1986; 16: 52-4 37. Grimfeld A. Infections respiratoires récidivantes de l’enfant: Intérêt du Ribomunyl. Tempo Médical 1988; 299: 57-61 38. Perruchet AM, Vautel JM. Epidemiologic and therapeutic survey : secretory otitis media [poster and abstract]. Congrès International Prévention Infection; 1990 June: Nice, France 39. Menardo JL, Perruchet AM. Epidemiologic and therapeutic survey : prevention of respiratory infections among children with respiratory allergy. Eur Respir J 1990; 3 Suppl. 10: 385s-386s 40. Sapène M, Grignet JP, Perruchet AM. Epidemiologic and therapeutic survey : chronic bronchitis and asthma [poster]. Congrès SEP/SEPCR. 1990 Sep; London, UK 41. Huls G, Hirche H, Lindemann H. Klinische effizienz eines neuen multibakteriellen immuntherapeutikums (Ribomunyl) bei kindern und jugendlichen mit rezidivierenden respiratorischen infekten. Jatros Paediatrie 1991; 7: 1-14
BioDrugs 2000 Dec; 14 (6)
408
42. Gonçalves JR, Sotto Mayor R, Thomas AM. Prevention of recurrent infections of the respiratory tract: concerning 1.117 patients [abstract]. Congrès International de Langue Française, Marrakech: 1993 Jun 20-23 43. Benchakroun Y, Bouayad Z, Chekkouri I, et al. Prévention des recidives infectieuses des voies respiratoires par Ribomunyl comprimés : Résultats d’une étude clinique multicentrique au Maroc. L’objectif Médical Magazine 1994; 10: 35-8 44. Alfarro C. Prévention des récidives infectieuses O.R.L. et bronchopulmonaires chez l’enfant par RIBOMUNYL sachet - résultat d’un essai clinique multicentrique au Portugal. Médecine Digest 1995; XXI 5: 1-8 45. Belaeva L. Clinical experience with ribosomal immunotherapy in paediatrics in Belarus. Drugs 1997; 54 Suppl. 1: 40 46. Shortanbaev, A. Prevention of recurrent infections: clinical evaluation in Kazakhstan. Drugs 1997; 54 Suppl. 1: 39 47. Arfa F. Therapeutic survey with ribosomal immunotherapy in the prevention of recurrent episodes of ENT and bronchopulmonary infections in Tunisia. Drugs 1997; 54 Suppl. 1: 41 48. Nascimento J. Prevention of ENT and BP infectious episodes in children with Ribomunyl sachet in Portugal. [abstract] Congress ‘From Immunology to Clinical Efficacy’, St Julien en Genevois, 1998; 09
© Adis International Limited. All rights reserved.
Boyle et al.
49. Stram DO. Meta analysis of published data using a linear mixed effects model. Biometrics 1996: 52: 536-44 50. Bryk AS, Raudenbush SW. Hierarchical linear models. London: Sage Publications, 1992 51. Woodhouse G, Rasbash J, Goldstein H, et al. MLn command reference. London: Multilevel Models Project, Institute of Education, 1995 52. Van Cauwenberge P, Berdeaux G, Morineau A, et al. Use of diagnosis clusters to assess the economic consequences of rhinopharyngitis in children in Italy and France during the winter. Clin Ther 1999; 21 (2): 404-21 53. Banz K, Thomas AM, Olivieri D. Economic evaluation of ribosomal immunotherapy in patients with chronic ear, nose and throat and respiratory tract infections – results for Italy. BioDrugs 1998; 10 (5): 385-96
Correspondence and offprints: Dr Peter Boyle, Division of Epidemiology and Biostatistics, European Institute of Oncology, Via Ripamonti 435, 20141 Milan, Italy. E-mail:
[email protected]
BioDrugs 2000 Dec; 14 (6)
