Review For reprint orders, please contact
[email protected]
BCG vaccination and leprosy protection: review of current evidence and status of BCG in leprosy control Expert Rev. Vaccines 9(2), 209–222 (2010)
Corinne SC Merle†, Sergio S Cunha and Laura C Rodrigues Author for correspondence Department of Epidemiology and Population Health, Tropical Epidemiological Group, London School of Hygiene and Tropical Medicine, Keppel Street, London WC1E 7HT, UK Tel.: +44 207 927 2826 Fax: +44 207 636 8739
[email protected] †
The bacillus Calmette–Guérin (BCG) vaccine, initially developed to provide protection against TB, also protects against leprosy; and the magnitude of this effect varies. Previous meta-analyses did not provide a summary estimate of the efficacy due to the heterogeneity of the results. We conducted a meta-analysis of published data including recently published studies (up to June 2009) to determine the efficacy of BCG protection on leprosy and to investigate whether age at vaccination, clinical form, number of doses, type of study, the latitude of study area and year of publication influence the degree of efficacy and explain the variation. In the light of the results, we argue for more emphasis on the role of BCG vaccination in leprosy control and research. Keywords : bacillus Calmette–Guérin • BCG • efficacy • efficiency • leprosy • meta-analysis • review • vaccine protection
Leprosy is a chronic granulomatous disease caused by Mycobacterium leprae, primarily affecting the peripheral nerves and skin. Among infectious diseases, leprosy is a leading cause of permanent physical impairment. Three major endemic countries (India, Brazil and Indonesia) account for 81% of all new cases [1] . Globally, the annual detection of new cases continues to decline, from a peak of more than 763,000 in 2001 to 254,525 in 2007 [1] . The basic principles for leprosy control are based on early diagnosis and treatment with multidrug therapy (MDT). Although bacillus Calmette–Guérin (BCG) vaccination was originally developed as a vaccine against TB, it has also been demonstrated to offer protection against leprosy [2–4] . Its protective effect has been shown in various studies, ranging from 20 to 90%, and its role in reducing the incidence of leprosy is mentioned in various reports of the WHO, such as the last one on the strategy for further reducing the disease burden due to leprosy [101] . Since the level of protection is variable (as is BCG protection against pulmonary TB) none of the reviews aimed to produce estimates of the potential impact of BCG vaccination on leprosy burden and control [2–4] . The www.expert-reviews.com
10.1586/ERV.09.161
reasons for the variation in protection are not known. Several questions about BCG vaccination and leprosy remain unanswered, including the duration of protective immunity, whether the protective efficacy in multibacillary forms is different to paucibacillary ones, whether age at vaccination affects protection and whether latitude is related to the level of protection (as it is for TB) [5] . In this review, we aim to give an update on current evidence, point to a lack of evidence regarding these questions and, in light of the results, we discuss the current and future status of BCG vaccination in leprosy control. Methods Literature search
A systematic bibliographic search of the medical literature published from 1 January 1960 until 1 June 2009 was conducted through electronic databases including Medline and Latin America & Caribbean Health Sciences (LILACS). The search was performed using and combining the following index terms: ‘BCG efficacy’, ‘BCG effectiveness’, ‘leprosy protection’, ‘protective effect’ and ‘leprosy vaccine’. In addition, a systematic manual search of bibliographies
© 2010 Expert Reviews Ltd
ISSN 1476-0584
209
Review
Merle, Cunha & Rodrigues
of every relevant study retrieved from the electronic search and from review articles was performed. The contents of all potential articles identified through the literature search were reviewed independently by two investigators (Merle and Cunha). All human studies, experimental or observational, published in English, Portuguese, Spanish or French, and measuring the efficacy or effectiveness of BCG vaccination in preventing leprosy cases were considered for inclusion in the meta-analysis. Crosssectional studies, studies based on comparison of ‘before–after’ intervention, or studies where efficacy was assessed on serum conversion or skin test outcomes instead of leprosy disease were not included. For trials, we selected only controlled trials with a clearly defined placebo (or nonintervention) group. To be included in the analysis, case–control studies had to define the criteria for selecting cases and controls, and the method for determining their BCG vaccination status. If multiple reports of a single study were published, the most complete (in terms of duration of follow-up or study population) was included. Selection criteria were defined before the literature search was performed in order to avoid selection bias.
heterogeneity, and a random-effects model with their respective 95% confidence intervals if evidence of heterogeneity in results was found. The significance cutoff value of 0.1 was used to decide whether or not there was heterogeneity [8] . We developed a random-effects regression model to investigate the source of heterogeneity in the efficacy of the BCG vaccine reported in individual studies and for the two sub-meta -analyses mentioned earlier. Meta-regressions were weighted with the inverse of the standard error of the estimates, and the between-study variance estimated through restricted maximum likelihood [9] . Regression analyses were limited to five variables potentially associated with BCG efficacy based on the literature review as well as data availability and accuracy: latitude of study area (as a continuous variable corresponding to degree from equator line), study design (trials vs observational studies), dose of BCG received (one dose vs more than one dose), the target population of the study (general population vs contacts of leprosy patients) and the year of publication (as a continuous variable). Finally, we used a funnel plot approach, as well as Begg’s and Egger’s tests, to investigate potential publication bias or bias for other reasons in this meta-analysis. All the analyses were performed using Stata version 10.
Data collection
Results
The authors and year of publication, the study characteristics (i.e., study design, target population, study period, study area, type of vaccine used and number of BCG doses received), vaccine effect estimates (rate ratio or risk ratio [RR] or odds ratio [OR]) and their corresponding 95% confidence intervals provided by the reports were extracted. Adjusted (rather than crude) estimates were considered if available. When the RR and/or OR and their confidence intervals were not available, they were estimated from the raw data in the reports.
A total of 272 titles or abstracts were examined (242 citations in Medline and 30 in LILACS). We identified 28 suitable studies that were included in the analysis: five trials [10–14] , six cohort studies [15–20] and 17 case–control studies [18,21–36] . A study conducted in Venezuela by Convit et al. was excluded as it did not estimate vaccine protection [37] , but rather investigated killed Mycobacterium leprae plus BCG, compared with BCG alone. Another observational study conducted in India by Chaudhury et al. was excluded because it compared three groups vaccinated with different vaccines (one being BCG alone) but had no unvaccinated group [38] . A trial in Vietnam by Truoc et al. was excluded because the intervention allocation was not satisfactory [39] . Finally, two randomized control trials, one in Malawi conducted by the Karonga group [40] and a cluster-randomized trial conducted by Cunha et al. in Brazil [41] were not included in the main metaanalysis as they assessed the efficacy of BCG revaccination only and did not estimate the efficacy of one dose of the BCG vaccine. Tables 1 & 2 describe the characteristics of all the studies included in the meta-analysis, along with the point estimate of BCG vaccine efficacy against leprosy.
Inclusion & exclusion criteria
Data analysis
Characteristics of study design and study population, separately for trials, case–control studies and cohorts were described. Vaccine efficacy, defined as 100 × (1-RRs or ORs), were calculated. A meta-analysis was conducted, aiming: • To summarize BCG effect estimates (pooled estimate and separate ones for each type of study design); • To assess the presence of heterogeneity; • To try to identify possible sources of heterogeneity. Information on vaccine efficacy specific to clinical forms (paucibacillary and multibacillary forms) or for specific age groups (younger or older than 15 years at vaccination) were also summarized in two meta-analyses, including a subset of studies with relevant data. Since leprosy is a disease with low incidence, RRs or ORs are numerically similar and so estimates of protection, whether ORs or RRs (from different types of study), were analyzed together, as suggested by Greenland et al. [6] . We proposed to estimate pooled effects using a fixed-effects model (based on the inverse variance method [7]) when there was no significant 210
BCG protection against leprosy
shows the efficacy by type of study and overall, sorted by latitude in a forest plot obtained in the meta-analysis. Vaccine protection in experimental studies (n = 5) ranged from 20 [12] to 48% [10] for the general population and was approximately 80% for individuals who had contact with a leprosy case [13] . Combining data from the five trials using a random-effects model (instead of a fixed-effects model due to significant heterogeneity of the results) gave a RR for leprosy protection of 0.59 (95% CI: 0.34–0.84) equivalent to an overall vaccine Figure 1
Expert Rev. Vaccines 9(2), (2010)
BCG vaccination & leprosy protection
Review
Table 1. Characteristics of the trials and cohort studies used in the meta-analysis and estimates of vaccine protection of BCG against leprosy (n = 11). Study (year) Study area Follow- Source Individuals Leprosy cases Latitude Vaccine up years population (n) (n) used
Vaccine protection % (95% CI)
Ref.
Trials Stanley et al. (1981)
Uganda (Soroti)
8
Contacts
16,150
261
1° 43’
Glaxo
80 (72–86)
[13]
Bagshawe et al. (1989)
Papua New Guinea
16
General
5356
281
6° 05’
Japan
48 (34–59)
[10]
Lwin et al. (1985)
Myamar
14
General
25,978
1494
26° 0’
Glaxo
20 (12–28)
[12]
Tripathy et al. (1983)
India
12
General
191,696
10,948
12° 42’
French (2) Danish (1)
24 (21–28)
[14]
Gupte et al. (1998)
India
8
General
171,400
372
12° 42’
India
34 (13–50)
[11]
Cohort studies Fine et al. (1986)
Malawi
5
General
79,801
70
9° 56’
NA
57 (24–74)
[18]
Ponnighaus et al. (1992)
Malawi
4
General
91,809
80
9° 56’
NA
49 (33–61)
[20]
Matos et al. (1999)
Brazil
4
Contact
670
23
22° 53’
NA
62 (33–78)
[16]
Cunha et al. (2004)
Brazil
4
School children
112,744
128
3° 7’
NA
41 (12–60)*
[15]
Goulart et al. (2008)
Brazil
5
Contact
1396
28
18° 55’
NA
53 (41–87)
[19]
Duppre et al. (2008)
Brazil
5
Contact
1922
69
22° 53’
NA
59 (27–77) ‡
[17]
BCG protection for the historical cohort. BCG protection among scar-negative contacts prior BCG vaccination. BCG: Bacillus Calmette–Guérin; NA: Not available. * ‡
protection of 41%. All cohort studies showed significant protection of BCG against leprosy. The pooled estimate using a fixedeffects model was 0.42 (95% CI: 0.34–0.51), corresponding to an overall vaccine protection of 58%. Overall, 14 out of the 17 case–control studies showed a statistically significant vaccine protection varying from 20 [28] to 90% [27] . Among the case– control studies, the only study restricted to contacts showed an overall vaccine protection of 56% (95% CI: 27–74) [25] , but in this study the population received one or more doses of BCG. Owing to the heterogeneity of the results, we used a randomeffects model to pool the estimates, which produced a combined OR of 0.40 corresponding to an overall vaccine protection of 60% (95% CI: 51–70%) (Figure 1) .
51.15%. When combining all covariates in the random multi variable meta-regression model, there was no statistical evidence that the number of doses of BCG, the latitude or the years of publication were contributing to the heterogeneity of the BCG efficacy between the studies. There was still some evidence that the type of study design interfered with the results. The BCG efficacy was found to be larger in observational studies than in trials: 0.40 (95% CI: 0.33–0.48) versus 0.59 (95% CI: 0.34–0.84) (p = 0.09) (Table 3) . The only aspect that significantly explained the heterogeneity of the results after adjustment was the target population of the study, whether they were contacts of leprosy cases or the general population. BCG efficacy seemed to be significantly higher among contacts of leprosy patients than among the general population (p = 0.03).
Meta-regression
In univariate meta-regression, the proportion of variation in BCG protection explained by each of the variables was: the latitude of the study area: 1.2%; the number of doses of BCG: 6.8%; the target population (whether contacts or general population): 25.6%; the year of publication: 45.8%; and the type of study: www.expert-reviews.com
Vaccine protection & clinical forms
presents vaccine efficacy estimates by clinical forms (i.e., paucibacillary and multibacillary forms). This analysis only included studies that reported BCG protection separately by clinical forms. Table 4
211
Review
Merle, Cunha & Rodrigues
Table 2. Characteristics of the case–control studies used in the meta-analysis and estimates of vaccine protection of BCG against leprosy (n = 17). Study (year)
Study area
Fine et al. (1986)
Case
Controls
Ref.
n
Source
n
Malawi
260
Survey
All Community population
9° 56’
Glaxo
36 (16–51)
[18]
Abel et al. (1990)
Vietnam
50
Health service
50
Hospital
10° 45’
Pasteur
48(-15–77)
[21]
Muliyil et al. (1991)
India (Vellore) 397
Community
669
Community
13° 05’
Copenhagen
20 (-10–41)
[28]
Rodrigues et al. (1992)
Brazil
62
Health service
186
School
16° 40’
Moreau
81 (63–90)
[31]
Baker et al. (1993)
Malawi
145
Community
290
Community
14° 58’
Glaxo
64 (42–77)
[22]
Convit et al. (1993)
Venezuela
90
Survey contact
3641
Survey contact
7° 06’
NA
56 (27–74)
[25]
Orege et al. (1993)
Kenya
69
Health service
238
Community
0° 06’
Glaxo
81 (67–90)
[29]
Thuc et al. (1994)
Vietnam
177
Health service
354
Community
10° 47’
SaigonPasteur, Canadian and Japanese
29 (-10–55)
[32]
Boelens et al. (1995)
Indonesia
115
NA
326
Community
5° 07’
NA
76 (39–90)
[24]
Lombardi et al. (1996)
Brazil
97
Health service
385
Community
23° 33’
Moreau
90 (78–96)
[27]
Bertolli et al. Myanmar (1997)
245
Health service
245
Community
16° 47’
NA
66 (44–80)
[23]
Zodpey et al. India (1998)
314
Health service
314
Hospital
21° 09’
Danish
71 (59–79)
[36]
Zodpey et al. India (1999)
212
Survey
212
Survey
21° 09’
Danish
60 (31–76)
[34]
Zodpey et al. India (2005)
364
Survey
364
Survey
20° 24’
Danish
54 (39–66)
[33]
Kerr-pontes et al. (2006)
226
Health service
857
Community
7° 12’
NA
52 (30–67)
[26]
Rahete et al. India (Raipur) (2007)
201
Health service
201
Community
21° 14’
NA
35 (-04–60)
[30]
Zodpey et al. India (2007) (Nagpur)
292
Health service
292
Health service
21° 09’
NA
62 (45–74)
[35]
Brazil (NE)
Source
Latitude Vaccine used Vaccine protection % (95% CI)
BCG: Bacillus Calmette–Guérin; NA: Not available.
Taking all studies together, there was greater variability of the BCG vaccine effect against paucibacillary forms, whereas the estimates for multibacillary forms were more homogeneous. The pooled protections were 76% (95% CI: 69–83%) for multibacillary forms and 62% (95% CI: 51–73%) for paucibacillary forms. After adjustment for two covariates: ‘number of dose of BCG delivered’ and ‘type of study design’ in a random-effects meta-regression model, we failed to find a statistically significant difference in BCG protection against multibacillary and paucibacillary leprosy forms. 212
Vaccine protection & age at vaccination
In all trials, vaccine protection was reported to be larger for subjects vaccinated at younger age compared with those vaccinated after 15 years of age. In all the cohort studies, BCG vaccination was given to children under 15 years of age. There are, therefore, no estimates available for the subgroup of subjects over 15 years of age. For case–control studies, despite the fact that the study population of several studies had a broad age range, estimates by age at vaccination were not provided or were not clear enough Expert Rev. Vaccines 9(2), (2010)
BCG vaccination & leprosy protection
Study
Year of Latitude of publication trial area
Review
ES (95% CI) % Weight
Trial Stanley et al. 2 1981 Bagshawe et al. 1989 6 Tripathy et al. 1983 13 Gupte et al. 1998 13 Lwin et al. 1985 26 Subtotal (l-squared = 98.1%, p = 0.000)
0.20 (0.14, 0.28) 0.52 (0.41, 0.66) 0.76 (0.72, 0.79) 0.66 (0.50, 0.87) 0.80 (0.72, 0.88) 0.59 (0.34, 0.84)
4.01 3.87 4.06 3.64 3.99 19.57
Cohort Cunha et al. 2004 3 Fine et al. 1986 10 Ponnighaus et al. 1992 10 Boulart et al. 2007 19 Duppre et al. 23 2008 Matos et al. 1999 23 Subtotal (l-squared = 0.0%, p = 0.521)
0.59 (0.40, 0.88) 0.43 (0.26, 0.76) 0.51 (0.26, 0.76) 0.27 (0.13, 0.59) 0.41 (0.33, 0.63) 0.38 (0.22, 0.67) 0.42 (0.34, 0.51)
3.39 3.34 3.34 3.43 3.78 3.46 20.73
Case–control Orege et al. 1993 0 Boelens et al. 1995 5 Kerr-pontes et al. 2006 7 Convit et al. 1993 7 Fine et al. b 1986 10 1994 11 Thuc et al. Abel et al. 1990 11 Muliyil et al. 1991 13 Baker et al. 15 1993 Bertolli et al. 1997 17 Rodrigues et al. 1992 17 Zodpey et al. c 2005 20 Rahete et al. 21 2007 Zodpey et al. 1998 21 Zodpey et al. d 2007 21 Zodpey et al. b 1999 21 Lombardi et al. 1996 24 Subtotal (l-squared = 80.0%, p = 0.000)
0.19 (0.10, 0.33) 0.24 (0.10, 0.61) 0.48 (0.33, 0.70) 0.44 (0.26, 0.73) 0.66 (0.49, 0.84) 0.71 (0.45, 1.10) 0.52 (0.23, 1.15) 0.80 (0.59, 1.10) 0.36 (0.23, 0.58) 0.34 (0.20, 0.56) 0.19 (0.10, 0.37) 0.46 (0.34, 0.61) 0.65 (0.40, 1.04) 0.29 (0.21, 0.41) 0.38 (0.26, 0.55) 0.40 (0.24, 0.69) 0.10 (0.04, 0.22) 0.40 (0.30, 0.49)
3.90 3.31 3.64 3.41 3.68 2.96 2.33 3.31 3.68 3.66 3.83 3.83 2.99 3.94 3.80 3.46 3.97 59.70
Overall (l-squared = 94.6%, p = 0.000)
0.45 (0.34, 0.56)
100.00
Note: weights are from random-effects analysis
0 RR or OR
1 No protection
Figure 1. Forest plot displaying a fully Bayesien meta-analysis (random-effects model) of the effect of BCG vaccine on leprosy overall, stratified by type of study design and sorted by the latitude of the study area (n = 28). BCG: Bacillus Calmette–Guérin; ES: Point estimate (OR or RR); OR: Odds ratio; RR: Risk ratio.
to be taken into consideration in the meta-analysis. BCG vaccine effects (RRs and ORs) against leprosy by age at vaccination for the study included in the meta-analysis are shown in Table 3 with corresponding pooled-effect estimates by type of study design; the overall pooled efficacy for BCG given before the age of 15 years was 57%. There were not sufficient data for subjects vaccinated after 15 years of age to allow meta-regression in order to investigate the difference in BCG protection according to age at vaccination.
error. When stratifying according to the type of study design, evidence of bias of the estimates was essentially limited to case– control studies (Begg’s test and Egger’s test concordant with p = 0.03 and p = 0.02, respectively). Figure 2 gives a summary of the results of the Begg’s and Egger’s tests and presents the funnel plot for case–control studies. It shows a clear asymmetry in the funnel plot, suggesting that small case–control studies with a low efficacy were less likely to be published than those finding a large efficacy.
Publication bias
Discussion
Begg’s funnel plots were used to investigate publication bias. The plot shows evidence of publication bias with a p value of 0.007 for the correlation between the point estimate and its standard
There is consistent evidence that BCG protects against leprosy. However, the magnitude of such a protection varies greatly in different studies and is estimated to be approximately 41% (95% CI:
www.expert-reviews.com
213
Review
Merle, Cunha & Rodrigues
Table 3. Meta-regression model to explain heterogeneity in the efficacy of BCG vaccine against leprosy (n = 28). Study characteristics considered in the meta-regression model
BCG efficacy (95% CI) in subgroup analysis
p-values (metaregression model)
Dose of BCG
of a fixed-effects model to investigate the reasons behind the heterogeneity. Factors considered for the meta-regression and other aspects that could influence the findings are discussed in the following sections. Importance of the study design
A debate continues in the literature as to whether one should include observational 2 doses or more 59% (50–68) 0.43 studies in a meta-analysis [42,43] , if results Type of study from different study designs should be combined or analyzed separately [44] , and Experimental studies 41% (16–66) 0.09 whether if observational and interventional Observational studies 60% (52–67) 0.09 studies disagree, the trials are always right. Target population studied The existence of unknown and unmeasGeneral population 53% (41–64)* 0.03 ured confounding factors in observational studies is the main criticism. Furthermore, Contact subject 68% (56–80) ‡ 0.03 trials and observational studies aim to § Latitude of study area 0.29 answer different questions: except by prag§ Years of publication of the study 0.26 matic trials, most randomized controlled * 30% BCG efficacy (95% CI: 20–40) for experimental studies and 59% BCG efficacy (95% CI: 50–68) for trials are concerned with vaccine protection observational studies. ‡ under controlled conditions, close to ‘ideal 80% BCG efficacy (95% CI: 73–87) for experimental studies and 62% BCG efficacy (95% CI: 52–72) for observational studies. conditions’, and so with vaccine efficacy. § Continuous variable. Observational studies are concerned with BCG: Bacillus Calmette–Guérin. estimates of vaccine effectiveness and pro16–66%) for trials and 60% (95% CI: 51–70%) for observa- tection under routine conditions. We found in our meta-analysis tional studies. We investigated whether the heterogeneity of that 48% of the heterogeneity of the results could be explained by the results was influenced by the latitude of the study area, the type of study. Therefore, pooled estimates have to be considered the number of BCG doses delivered, the year of publication with caution. We decided to present them separately for trials and of the study, the study design and whether the population was observational studies and to adjust on this factor when conducting general or contacts of leprosy cases. Although there were differ- the meta-regression. ences, perhaps owing to a lack of power, we did not find statistiOther methodological aspects specific to some study designs may cally significant evidence that the heterogeneity of the results explain the difference in efficacy depending on the study design. across all studies was explained by latitude of the study area, Length of follow-up is an important factor to take into considthe number of BCG doses received or the year of publication eration. In various experimental or cohort studies [11,12,16,17,41] an of the study. There is some evidence of variation of the BCG increased risk of leprosy has been noticed among the population efficacy depending of the study design, with an estimated higher exposed to BCG during the first year of follow-up. The explanation protective effect in observational studies compared with trials, offered for this paradoxical finding is that BCG vaccination may and some indication of publication bias in case–control studies. induce the emergence of leprosy clinical manifestations among Finally, we found a statistically significant higher protective asymptomatic infected individuals during the initial period of effect of the BCG vaccination if studies were conducted among follow-up. It is worth distinguishing whether vaccine unmasks household contacts instead of the general population, even after clinical symptoms among people who would ultimately develop adjustment for the number of doses of BCG received, the type of the disease later without vaccination, or whether it triggers clinithe study design, the year of publication and the latitude of the cal symptoms among people who would not have developed the study area. We did not find statistical evidence of a difference disease if they were not vaccinated. There is currently no evidence in protection against multibacillary and paucibacillary clinical to support one view or the other. However, in practice, this implies forms. The available data did not permit a reliable analysis to the need for a sufficient follow-up period of the study population be carried out to assess the difference of vaccine protection con- in order to draw reliable conclusions about BCG protection. It ferred for children vaccinated before 15 years of age compared may explain the variability of efficacy estimates from one study to with later vaccination. another depending on the length of the follow-up period. The reliability of BCG scar reading to ascertain whether or not Heterogeneity in the protective efficacy of BCG BCG vaccination has been performed is a matter of concern in There is overwhelming evidence that BCG can offer protection observational studies, and it has been studied by various teams. against leprosy, but that the level of protection varies. Owing Rodrigues et al. found an overall good specificity (90%) and sento the heterogeneity, we used a random-effects model instead sitivity (98%) of scar reading in Brazil [45] . In Malawi, Floyd et al. 1 dose of BCG
214
55% (42–67)
0.43
Expert Rev. Vaccines 9(2), (2010)
BCG vaccination & leprosy protection
also found that BCG scar reading is overall a highly sensitive and repeatable indicator of BCG vaccination (93 and 94%, respectively) but when focusing on the population of infants vaccinated when younger than 1 month of age, they found that less than 80% had still recognizable scars 4 years later [46] . Given that most vaccinations in the world are administered soon after birth, this low sensitivity may lead to an underestimation of BCG protection in observational studies in which vaccination status is inferred from the presence or absence of a distinctive BCG scar. Paradoxically, in our meta-analysis we found a higher BCG vaccine effectiveness in observational studies compared with vaccine efficacy observed in trials (as was the case in other trials [2,4]), which should be the opposite if the measurement of vaccine protection was biased by misclassification of previous BCG exposure. Even if BCG scar reading is not a perfect indicator of previous BCG vaccination, it cannot explain the difference in terms of BCG protection noticed between observational studies and trials.
Review
protection to these individuals already exposed to EM. Therefore, the observed vaccine protection may depend on the prevalence of infection by EM in the study population, which varies in different regions throughout the world, and may explain in part the variation of the BCG protection from one study to another. There is no perfect variable that can embody the degree of exposure to EM. High humidity and temperature, as is found in tropical areas, favor the growth of such EM. Therefore, as investigated by Fine et al. for BCG protection against TB, we tried to adjust the results on vaccine protection on the latitude of the study area, considered as a proxy measure of mycobacterial exposition. We failed to find any relationship between the protection offered by BCG and the latitude of the study area, although all leprosy studies are carried out in tropical areas and thus there may not be enough variation in the case of leprosy studies; latitude contributes to only 4.3% of the heterogeneity. However, it is worth mentioning that humidity and temperature (likely to influence EM prevalence) do vary in different areas at the same latitude.
Number of doses of BCG vaccination & benefits of revaccination
We found no statistical difference in BCG protection between studies where patients are vaccinated once (55%; 95% CI: 42–67%) and studies where patients are vaccinated twice or more (59%; 95% CI: 50–68%), with a p value of 0.43. It is unlikely that the variability of the BCG efficacy between studies could be explained by the number of BCG doses received by the study population in one study compared with another. There is also no consistency on whether revaccination always provides additional protection. The two large trials conducted had very different results: a recent cluster randomized trial conducted by Cunha et al. among 99,770 school children 7–14 years of age and followed for 6 years in Brazil found an incidence rate ratio of leprosy of 0.99 in the revaccinated group compared with the control group (95% CI: 0.68–1.45), indicating no obvious gain of protection with revaccination [41] . By contrast, Fine et al. found in a randomized controlled trial conducted in northern Malawi that a second BCG vaccination affords 49% protection (95% CI: 0–75%) against leprosy [40] . The main difference between these two studies is the nature of the revaccinated population, which was restricted to school children in the study in Brazil and involved a broader age range of infants to adults in Malawi. Revaccination might give additional protection to adults for whom the efficacy of the first vaccination waned with time but, as shown by Cunha et al., there might be no benefit of revaccination when it is performed in school children. More studies need to be undertaken in order to draw a reliable conclusion; the level of vaccine protection from revaccination may vary from place to place. Exposure to environmental mycobacteria
Many species of bacteria of the genus Mycobacterium are found in the external environment, such as water and soil. There is growing evidence that the exposure of a population to other mycobacteria might interfere with the efficacy of the BCG vaccine against TB and leprosy [47] . Individuals exposed to environmental mycobacteria (EM) could have some degree of acquired ‘natural’ immunity against leprosy, and BCG may not offer additional www.expert-reviews.com
Study population & vaccination of high-risk populations
We found that BCG efficacy was significantly more important in studies when BCG vaccination was targeting household close contacts who are at higher risk of infection compared with the ones conducted in the general population (66% of protection for household close contacts versus 53% for the general population). It is noteworthy that leprosy has a long incubation period and it is difficult to define a clear date of onset. The main issue faced by the investigators conducting intervention among a contact population is the risk of misclassification of cases and controls. Duppre et al. found 28 new leprosy cases (21 cases in the BCG-vaccinated group) during the first 2–10 months of their study [17] . They made the assumption that these new cases were either within the incubation period or were mistakenly not diagnosed at the time of recruitment into the cohort study, and that BCG vaccination accelerated the natural history of Mycobacterium leprae infection in subjects infected prior to vaccination. Furthermore, it is not always easy to quantify the level of exposure of contacts and there might be variation in the definition of ‘contact’ from one study to another. Lastly, the summary estimates found in this meta-analysis were heterogeneous, which obliged us to consider random-effects model estimates that give greater weight to less-precise studies and might therefore have undesired effects on the summary risk estimates. In summary, although we are still unable to explain the variation in estimates, these findings clearly show the effect of BCG in preventing leprosy, and suggest a role for routine BCG vaccination of all leprosy close contacts in addition to treatment of index cases. There is a need for further studies measuring the impact of such a strategy in order to guide vaccination policy in endemic countries, for example, cost–effectiveness analyses and chemoprophylaxis followed by BCG vaccination. Publication bias & other factors
We found some evidence of publication bias, especially among case–control studies, which seems to show that small case–control studies demonstrating low protection might be missing in 215
Review
Merle, Cunha & Rodrigues
Table 4. Vaccine effect (rate ratio and odds ratio) and 95% confidence intervals by study and clinical forms or age at vaccination. Study (year)
Clinical form
Age at vaccination
Ref.
Paucibacillary
Multibacillary
