- Email: [email protected]
The risk of disseminated Bacille Calmette-Guerin (BCG) disease in HIV-infected children
Vaccine 25 (2007) 14–18
Short communication
The risk of disseminated Bacille Calmette-Guerin (BCG) disease in HIV-infected children Anneke C. Hesseling a,b,∗ , Ben J. Marais a , Robert P. Gie a , H. Simon Schaaf a , Paul E.M. Fine b , Peter Godfrey-Faussett b , Nulda Beyers a a
Desmond Tutu TB Centre, Department of Paediatrics and Child Health, Faculty of Health Sciences, Stellenbosch University, P.O. Box 19063, Tygerberg, 7505, South Africa b Department of Infectious and Tropical Diseases, London School of Hygiene and Tropical Medicine, UK Received 16 May 2006; received in revised form 15 July 2006; accepted 16 July 2006 Available online 1 August 2006
Abstract Objectives: Bacille Calmette-Guerin (BCG), a live attenuated Mycobacterium bovis vaccine, poses a risk to human immunodeficiency virus (HIV)-infected children; this risk has not been well quantified. We estimate the risk of disseminated BCG disease in HIV-infected children in a setting highly endemic for tuberculosis and HIV. Design and methods: We conducted a prospective hospital-based surveillance study in the Western Cape Province, South Africa. Clinical and laboratory-confirmed cases of disseminated BCG disease in children <1 year of age from January 2002 to December 2004 at a referral hospital were used as numerator data. Denominator data for calculations of disseminated BCG risk were obtained through estimating the total number of HIV-infected infants receiving BCG based on the known vaccination coverage in the study setting, combined with population data on the total number of children <1 year of age, the known HIV prevalence amongst women attending public antenatal care facilities and different scenarios (5–15%) for the rate of vertical HIV transmission. Results: Nine cases of disseminated BCG disease were identified over the study period, seven of these were in HIV-infected infants. The estimated risk for HIV-infected infants to develop disseminated BCG disease, given a 95% BCG coverage and an HIV prevalence of 12.4–15.4% amongst women, were as follows for different scenarios of vertical HIV transmission: 329–417/100,000 vaccinees (assuming 5% vertical HIV transmission), 164–208/100,000 vaccinees (assuming 10% vertical HIV transmission) and 110–139/100,000 vaccinees (assuming 15% vertical HIV transmission). Conclusions: The risk of disseminated BCG disease is increased several hundred fold in HIV-infected infants compared to the documented risk in HIV-uninfected infants. Data on the protective effect of BCG in HIV-exposed and infected children is lacking. Population- and hospital-based surveillance is vitally important to more accurately estimate the safety and benefits of BCG in HIV-exposed and infected infants. © 2006 Elsevier Ltd. All rights reserved. Keywords: BCG; HIV; Disseminated; Incidence
1. Introduction Bacille Calmette-Guerin (BCG), a live attenuated Mycobacterium bovis strain vaccine, is routinely given to infants at birth or shortly thereafter, in regions where tuber-
∗
Corresponding author. Tel.: +27 21 9389177; fax: +27 21 9389719. E-mail addresses: [email protected], [email protected] (A.C. Hesseling). 0264-410X/$ – see front matter © 2006 Elsevier Ltd. All rights reserved. doi:10.1016/j.vaccine.2006.07.020
culosis is endemic. The World Health Organization (WHO) currently recommends that BCG be given to all asymptomatic infants, irrespective of human immunodeficiency virus (HIV) exposure [1]. This policy has practical limitations, as most HIV-exposed infants are infected perinatally, and are therefore asymptomatic at birth. Vaccination with BCG has remained the standard of care for tuberculosis prevention in most developing countries because of its documented efficacy in preventing life-threatening forms of disease in HIV-uninfected infants and young children. It is the
A.C. Hesseling et al. / Vaccine 25 (2007) 14–18
only currently available vaccine, is inexpensive and requires only one encounter with an infant [2]. In the absence of HIV infection, BCG-associated adverse events in children include injection-site lesions, adenitis, suppurative adenitis and very rarely, disseminated disease. Adverse events vary with vaccine strain type, physical–chemical properties, bacillary load, administration method and host immune characteristics [3]. The reported frequency of disseminated BCG disease has traditionally been less than five per one million vaccinees and is mainly associated with congenital cellular immunodeficiency [4]. Despite several case reports of local and disseminated BCG disease in HIV-infected infants [5–9], the risk of disseminated BCG disease in children vertically infected with HIV remains poorly quantified, partly due to a lack of surveillance and adequate population-based data in HIV-endemic settings. Disseminated BCG disease in HIV-infected children is likely to be underreported, especially in settings highly endemic for tuberculosis and HIV, where diagnostic facilities are often limited [10]. Even where mycobacterial culture is available, disseminated BCG disease may be misdiagnosed as tuberculosis in HIV-infected children, in the absence of accurate Mycobacterium tuberculosis complex speciation [11]. This preliminary reports aims to estimate the risk of disseminated BCG disease in HIV-infected infants in a setting highly endemic for tuberculosis and HIV, using available paediatric population data and known BCG coverage rates and analyzing different hypothetical scenarios for vertical transmission of HIV combined with known maternal antenatal HIV prevalence in the public health sector.
2. Methods 2.1. Study setting This study was conducted from January 2002 to December 2004 at the Tygerberg Children’s Hospital, Western Cape Province, South Africa, which drains 30–50% of provincial tertiary paediatric referrals. The adult tuberculosis incidence in the province was 678/100,000 in 2003, the incidence of tuberculosis in children <13 years of age was 407/100,000 and the HIV prevalence among women attending public antenatal care facilities was 15.4% (95% CI: 12.5–18.2%) in 2004 [12–14]. Current South African BCG vaccination policy recommends universal intradermal vaccination at birth with Danish Strain BCG (1331, Statens Serum Institute, Copenhagen, Denmark) in the right deltoid region. 2.2. Risk calculation Disseminated BCG disease in HIV-infected children is predominantly seen in children <1 year of age with rapid progression to advanced HIV disease [5,8,9]. Therefore, we included all HIV-infected and uninfected children <1 year
15
of age diagnosed with disseminated BCG disease during the study period. 2.2.1. Numerator Cases of distant/disseminated BCG disease were detected through active prospective hospital laboratory-based and clinical surveillance. M. bovis BCG was detected through routine polymerase chain reaction (PCR) speciation of M. tuberculosis complex isolates [11] after mycobacterial culture by the automated Middlebrook 7H9 broth-based Mycobacterial Growth Indicator Tube method (MGIT; Becton Dickinson, Sparks, MD, USA), following rigorous measures to prevent mycobacterial cross-contamination [15]. Additionally, the Western Cape Provincial Expanded Programme on Immunization (EPI) register was reviewed to ensure that all routinely notified disseminated BCG disease cases over the study period were also included. Distant/disseminated BCG disease was confirmed by isolation of M. bovis BCG from a distant site beyond the ipsilateral axillary gland, e.g. from gastric aspiration, in the presence of systemic symptoms and local and/or regional evidence of BCG disease, based on a standard paediatric BCG disease classification [9]. BCG vaccination was documented both from neonatal vaccination records and the presence of a scar in the right deltoid region. HIV infection was diagnosed through enzyme-linked immunosorbent assay and confirmed by HIV DNA PCR. Informed consent was obtained for study participation and HIV testing from parents or legal guardians. The study was approved by the Ethics Committee, Faculty of Health Sciences, Stellenbosch University. 2.2.2. Denominator Mid-year projected population data on the total number of live children <1 year of age in the province were used to calculate denominator data [16]. This projected figure accounts for both the total number of live births and infant mortality. In the absence of reliable population-based routine surveillance data on the exact number of live HIV-infected children <1 year of age in the Western Cape Province over the study period, we used the projected mid-year population for the study period of children <1 year to estimate the total number of HIV-infected infants, given plausible scenarios of mother-to-child transmission of HIV for the study setting and the routinely documented HIV prevalence amongst women attending public antenatal clinics. Scenarios 1–3 assumed vertical HIV transmission rates of 5, 10 and 15%, respectively; this was applied to the maternal HIV prevalence data over the study period (Table 1) [14]. BCG vaccination coverage of 95% was assumed, based on a recent provincial vaccine coverage survey [17]. In the same manner as for scenarios 1–3, the total annual number of live children <1 year of age less the number of HIV-infected children was used as denominator to estimate the risk of disseminated BCG disease in HIV-uninfected children.
16
A.C. Hesseling et al. / Vaccine 25 (2007) 14–18
Table 1 Estimates of the total mid-year population of children <1 year, Western Cape Province, South Africa, along with the number of HIV-infected children vaccinated with BCG based on the known maternal antenatal HIV prevalence and assumed rates of vertical HIV transmission
Maternal HIV prevalence Routine provincial HIV surveillance data [14]a Total number of infants <1 year of age [16]b Scenario 1 Assuming 5% total HIV vertical infection Total number of HIV-infected infants vaccinated assuming 95% neonatal BCG coverage Scenario 2 Assuming 10% total HIV vertical infection Total number of HIV-infected infants vaccinated assuming 95% neonatal BCG coverage Scenario 3 Assuming 15% total HIV vertical infection Total number of HIV-infected infants vaccinated assuming 95% neonatal BCG coverage
2002
2003
2004
12.4% (95% CI: 8.8–15.9%)
13.1% (95% CI: 8.5–17.7%)
15.4% (95% CI: 12.5–18.2%)
96,946
97,778
98,339
601 571
640 608
757 719
1202 1142
1281 1217
1514 1439
1803 1713
1921 1825
2272 2158
Scenarios 1–3 used mid-year population estimates of total children <1 year of age, and assumed a vertical HIV transmission risk of 5, 10 and 15%, respectively, given the routine maternal HIV surveillance, to calculate the total number of HIV-infected infants. BCG coverage rates from the Western Cape Provincial vaccination survey, 2005 [17]. All numbers are rounded to the closest integer. a Routine provincial maternal antenatal surveillance data, as measured through annual sentinel surveillance. This only relates to the public health sector; no surveillance data are available for the private health sector. b ASSA 2003 model, Actuarial Society, South Africa; mid-year population estimates based taking into account total number of live births and infant mortality rate.
3. Results Nine children were diagnosed with disseminated BCG disease over the study period. All these children had immune deficiencies. Seven children were HIV-infected; all had advanced HIV disease and were <1 year of age; the mortality was 85.7% (6/7). The two HIV-uninfected children with disseminated BCG disease, both had primary immune deficiencies including one child with severe combined immunodeficiency and a second with an unidentified T-cell deficiency. Detailed data on the clinical presentation, HIV disease characteristics, BCG disease classification, management and outcome of all children with distant/disseminated BCG disease are reported elsewhere [9]. No additional cases of dissemi-
nated BCG disease were identified from the provincial notification register for the study period. Denominator data calculations resulted in a total number for 2002–2004 of 571–719 HIV-infected infants if 5% vertical HIV transmission was assumed, 1142–1439 HIVinfected infants if 10% vertical transmission was assumed and 1713–2158 HIV-infected infants if 15% vertical HIV transmission was assumed, respectively, for the study period. The total number of distant/disseminated BCG cases per study year and the calculated risks of distant/disseminated BCG disease in HIV-infected infants for 2002–2004 are indicated in Table 2. The calculated incidence of distant or disseminated disease in HIV-uninfected infants was 0.72–0.74 per 100,000 per year for the study period.
Table 2 Calculated risk of distant or disseminated BCG disease in HIV-infected children based on the known maternal antenatal HIV prevalence, assumed rates of vertical HIV transmission and estimates of total mid-year population of children <1 year, Western Cape Province, South Africa Risk scenarios of disseminated BCG disease
Cases/year 2002
Cases/year 2003
Cases/year 2004
Actual cases/year Risk of disseminated BCG disease Case scenario 1, assuming 5% total vertical HIV infection Case scenario 2, assuming 10% total vertical HIV infection Case scenario 3, assuming 15% total vertical HIV infection
2
2
3
2/571 = 350/100,000/year 2/1142 = 175/100,000/year 2/1713 = 117/100,000/year
2/608 = 329/100,000/year 2/1217 = 164/100,000/year 2/1825 = 110/100,000/year
3/719 = 417/100,000/year 3/1439 = 208/100,000/year 3/2158 = 139/100000/year
Routine provincial maternal antenatal surveillance data, as measured through annual sentinel surveillance [14]. This only relates to the public health sector; no surveillance data are available for the private health sector. ASSA 2003 model, Actuarial Society, South Africa [16] mid-year population estimates based taking into account total number of live births and infant mortality rate. Scenarios 1–3 used mid-year population estimates of total children <1 year of age, and assumed a vertical HIV transmission risk of 5, 10 and 15%, respectively, given the routine maternal HIV surveillance, to calculate the total number of HIV-infected infants. BCG coverage rates from the Western Cape Provincial vaccination survey, 2005 [17]. All numbers are rounded to the closest integer.
A.C. Hesseling et al. / Vaccine 25 (2007) 14–18
4. Discussion We have demonstrated that neonatal BCG vaccination may pose a considerable risk of distant or disseminated BCG disease to infants vertically infected with HIV, even when they are asymptomatic when vaccinated at birth. The estimated risk of disseminated BCG disease in HIV-infected children was increased several hundred fold compared to that historically described in HIV-negative children. Further surveillance of BCG safety is critically important to more accurately estimate the risk of BCG vaccination in HIVinfected infants, as these preliminary risk calculations have important limitations: we did not have reliable populationbased paediatric HIV data for the study period, and therefore used modeled data and assumptions to obtain denominator estimates for the total number of HIV-infected infants. We addressed some of these limitations by using different hypothetical scenarios for vertical HIV transmission. The rate of vertical HIV transmission depends on multiple factors including antiretroviral therapy in the mother as part of prevention of mother-to-child HIV transmission (PMTCT) regimens or as full (triple) antiretroviral therapy, breastfeeding practices and method of delivery. There is an active PMTCT program in the Western Cape Province, South Africa. By March 2003, the Western Cape Department of Health had achieved full roll-out of the programme in all districts in the province with uniform access for pregnant women attending public-sector maternity services [23]. The current PMTCT protocol recommends universal HIV testing of all pregnant women in the antenatal services and uses combination preventive drug regimens (AZT and NVP) during pregnancy, intrapartum and in the infant after birth. Women who have a CD4+ T lymphocyte count of ≤200 ul−1 at the first antenatal booking are offered triple-drug therapy. The vertical HIV transmission rate in the Cape Metropolitan area was approximately 4–5% in 2004 (personal communication, Dr. Pren Naidoo, Cape Town City Health), at a time when the PMTCT regimen had been widely implemented. We therefore consider the vertical HIV transmission rate scenario of 5% the most realistic for the study period. Vertical HIV transmission is however likely to be variable in other study settings, due to variable availability and uptake of PMTCT services. These risk calculations are likely to be a considerable under-estimate of the true disseminated BCG disease incidence in HIV-infected infants for the following reasons: we assumed that all tertiary referrals for disseminated BCG disease would have been seen at the study hospital, which is unlikely, given that only 30–50% of children in the province are eligible for referral. We only included numerator data from this hospital, as ongoing prospective surveillance is routinely conducted using standard criteria and case definitions. Further reasons for possible under-estimation are the stringent case definitions used and the possibility that HIVinfected children with disseminated BCG disease may have been misdiagnosed or died of HIV-or BCG-related causes
17
before referral. Despite these limitations, the validity of our estimations is supported by the risk calculation of disseminated BCG disease in HIV-uninfected children, which was in the same order of magnitude as that described in the literature. These preliminary risk calculations are therefore useful to quantify the direct risk posed by BCG to HIV-infected infants, given different scenarios of vertical HIV transmission. More extensive surveillance measures are likely to yield more exact estimates in the future. We are unable to comment on the potential non-specific effects BCG may have in HIVinfected infants. Apart from risk estimation, an assessment of the protective effect of BCG in HIV-infected children is important, given the extremely high risk of infection with M. tuberculosis in highly tuberculosis-endemic settings and the high risk of tuberculosis disease progression in HIV-infected children following M. tuberculosis infection [18,19]. In HIV-unexposed and uninfected children, evidence of the protective effect of BCG against disseminated forms of tuberculosis in the first 2 years of life is consistent, with reported efficacy ranges from 46 to 100% [3,20]. In a recent meta-analysis, BCG has also been shown to be cost-effective, preventing nearly one case of disseminated tuberculosis for roughly every 9300 inoculations, mostly in Southeast Asia (46%) and Africa (27%) [21]. However, these cost-effectiveness analyses did not quantify BCG benefits in HIV-infected children. Equal protective effect in HIV-infected children cannot be assumed, as the available epidemiological data on the protective effect of BCG in HIVexposed or infected children is extremely limited [1]. Most BCG efficacy and safety trials were conducted in the preHIV era or in settings where the burden of paediatric HIV was very low. As the mechanism of protection derived from BCG involves a reduction of the haematogenous spread of bacilli from the site of primary infection [22], it is plausible that when a specific T-cell mediated response is required for control of haematogenous spread of M. tuberculosis, the suppression of T-cell immunity as a result of HIV may render this defence inadequate in children. However, in settings highly endemic for tuberculosis and HIV, even partial protection due to BCG may confer considerable benefit to HIV-infected infants [10]. Based on the considerable risk of disseminated BCG disease in HIV-infected infants, possible vaccination policy implications include the more rapid identification of HIVinfected infants through PMTCT programs in HIV-endemic settings, and delaying BCG in HIV-exposed infants until their HIV status has been determined. As data on the protective effect of BCG in HIV-exposed and infected infants are extremely limited, more research is urgently needed in this area. Prospective surveillance incorporating hospital and population-based strategies in settings highly endemic for HIV and tuberculosis are vitally important to more accurately quantify the risks and benefits of BCG in specific sub-populations. New tuberculosis vaccine strategies should consider both the potential risks and benefits of live vaccines in HIV-infected children.
18
A.C. Hesseling et al. / Vaccine 25 (2007) 14–18
Acknowledgements The authors would like to thank Robin Warren, DST/NRF Centre of Excellence and Department of Medical Biochemistry, Faculty of Health Sciences, Stellenbosch University for PCR speciation, Wendy Brittle for mycobacterial culture, Cape Town City Health (Pren Naidoo and Ivan Toms), the Western Cape Province Department of Health (Keith Cloete and Fawzia Desai) for advice on immunization and HIV surveillance data and the South African Actuarial Society (Leigh Johnson and Keith Dorrington), for use and advice of the ASSA AIDS 2003 model. The authors would also like to thank the TB 21st Century Consortium, supported by the Norwegian Scientific Council and the Harry Crossly Foundation Trust for financial support. This manuscript is in partial fulfilment of a Ph.D. thesis. Conflict of interest: None declared.
References [1] World Health Organization. BCG vaccine. WHO position paper. Wkly Epidemiol Rec 2004;79:27–38 [accessed on 15 July 2006] http://www.who.int/wer. [2] Girard MP, Fruth U, Kieny M. Vaccine 2005;23:5725–31. [3] Fine PEM, Carneiro IAM, Milstien JB, Clements C. Issues relating to the use of BCG in immunization programmes: a discussion document WHO/V&B/99.23. Geneva: World Health Organization; 1999. [4] Casanova JL, Jouanguy E, Lamhamedi S, Blanche S, Fischer A. Immunological conditions of children with BCG disseminated infection. Lancet 1995;346:581. [5] Fallo A, Torrado L, Sanchez A, Cerquiero C, Shadrosky L, Lopez EL. Delayed complications of bacillus Calmette-Guerin vaccination in HIVinfected children. In: [Abstract] Program and abstracts of the Third IAS Conference on HIV Pathogenesis and Treatment. 2005 [WeOa0104]. [6] Lallemant-Le Coeur S, Lallemant M, Cheynier D, Nzingoula S, Drucker J, Larouze B. Bacillus Calmette-Guerin immunization in infants born to HIV-1-seropositive mothers. AIDS 1991;5:195– 9. [7] O’Brien KL, Ruff AJ, Louis MA, et al. Bacillus Calmette-Guerin complications in children born to HIV-1-infected women with a review of the literature. Pediatrics 1995;95:414–8.
[8] Hesseling AC, Schaaf HS, Hanekom WA, et al. Danish bacille Calmette-Guerin vaccine-induced disease in human immunodeficiency virus-infected children. Clin Infect Dis 2003;37:1226–33. [9] Hesseling AC, Rabie H, Marais BJ, Manders M, Lips M, Schaaf HS, et al. Bacille Calmette-Guerin vaccine-induced disease in HIV-infected and HIV-uninfected children. Clin Infect Dis 2006;42:548–58. [10] Talbot EA, Perkins MD, Silva SF, Frothingham R. Disseminated bacille Calmette-Guerin disease after vaccination: case report and review. Clin Infect Dis 1997;24:1139–46. [11] Talbot EA, Williams DL, Frothingham R. PCR identification of Mycobacterium bovis BCG. J Clin Microbiol 1997;35:566–9. [12] Cape Town TB Control. Progress report 1997–2003, available at: http://www.hst.org.za/publications/618 [accessed on 15 July 2006]. [13] Marais BJ, Hesseling AC, Gie RP, Schaaf HS, Beyers N. The burden of childhood tuberculosis and the accuracy of community-based surveillance data. Int J Tuberc Lung Dis 2006;10:259–63. [14] Department of Health. National HIV and Syphilis antenatal seroprevalence survey 2004, available at: www.doh.gov.za/docs/reports/ 2004/hiv-syphilis.pdf [accessed on 15 July 2006]. [15] Carroll NM, Richardson M, van Helden PD. Criteria for identification of cross-contamination of cultures of Mycobacterium tuberculosis in routine microbiology laboratories. J Clin Microbiol 2003;41:2269. [16] Actuarial Society of South Africa, AIDS model 2003, http://www.assa. org.za [accessed on 15 July 2006]. [17] Corrigal J. Vaccination coverage of the Western Cape Province: household survey 2005. Child Health and EPI sub-directorate of the Western Cape Province. [18] Mukadi YD, Wiktor SZ, Coulibaly, et al. Impact of HIV infection on the development, clinical presentation, and outcome of tuberculosis among children in Abidjan, Cote d’Ivoire. Aids 1997;11:1151–8. [19] Marais BJ, Gie RP, Schaaf HS, Hesseling AC, Obihara CC, Starke JJ, et al. The natural history of childhood intra-thoracic tuberculosis: a critical review of literature from the pre-chemotherapy era. Int J Tuberc Lung Dis 2004;4:392–402. [20] Rodrigues LC, Diwan VK, Wheeler JG. Protective effect of BCG against tuberculous meningitis and miliary tuberculosis: a metaanalysis. Int J Epidemiol 1993;6:1154–8. [21] Trunz BB, Fine P, Dye C. Effect of BCG vaccination on childhood tuberculous meningitis and miliary tuberculosis worldwide: a meta-analysis and assessment of cost-effectiveness. Lancet 2006;367:1173–80. [22] Arbelaez MP, Nelson KE, Munoz A. BCG vaccine effectiveness in preventing tuberculosis and its interaction with human immunodeficiency virus infection. Int J Epidemiol 2000;6:1085–91. [23] Western Cape Department of Health. Prevention of mother-to-child transmission of HIV (PMTCT), available at http://www.capegateway. gov.za/health [accessed on 15 July 2006].
Short communication
The risk of disseminated Bacille Calmette-Guerin (BCG) disease in HIV-infected children Anneke C. Hesseling a,b,∗ , Ben J. Marais a , Robert P. Gie a , H. Simon Schaaf a , Paul E.M. Fine b , Peter Godfrey-Faussett b , Nulda Beyers a a
Desmond Tutu TB Centre, Department of Paediatrics and Child Health, Faculty of Health Sciences, Stellenbosch University, P.O. Box 19063, Tygerberg, 7505, South Africa b Department of Infectious and Tropical Diseases, London School of Hygiene and Tropical Medicine, UK Received 16 May 2006; received in revised form 15 July 2006; accepted 16 July 2006 Available online 1 August 2006
Abstract Objectives: Bacille Calmette-Guerin (BCG), a live attenuated Mycobacterium bovis vaccine, poses a risk to human immunodeficiency virus (HIV)-infected children; this risk has not been well quantified. We estimate the risk of disseminated BCG disease in HIV-infected children in a setting highly endemic for tuberculosis and HIV. Design and methods: We conducted a prospective hospital-based surveillance study in the Western Cape Province, South Africa. Clinical and laboratory-confirmed cases of disseminated BCG disease in children <1 year of age from January 2002 to December 2004 at a referral hospital were used as numerator data. Denominator data for calculations of disseminated BCG risk were obtained through estimating the total number of HIV-infected infants receiving BCG based on the known vaccination coverage in the study setting, combined with population data on the total number of children <1 year of age, the known HIV prevalence amongst women attending public antenatal care facilities and different scenarios (5–15%) for the rate of vertical HIV transmission. Results: Nine cases of disseminated BCG disease were identified over the study period, seven of these were in HIV-infected infants. The estimated risk for HIV-infected infants to develop disseminated BCG disease, given a 95% BCG coverage and an HIV prevalence of 12.4–15.4% amongst women, were as follows for different scenarios of vertical HIV transmission: 329–417/100,000 vaccinees (assuming 5% vertical HIV transmission), 164–208/100,000 vaccinees (assuming 10% vertical HIV transmission) and 110–139/100,000 vaccinees (assuming 15% vertical HIV transmission). Conclusions: The risk of disseminated BCG disease is increased several hundred fold in HIV-infected infants compared to the documented risk in HIV-uninfected infants. Data on the protective effect of BCG in HIV-exposed and infected children is lacking. Population- and hospital-based surveillance is vitally important to more accurately estimate the safety and benefits of BCG in HIV-exposed and infected infants. © 2006 Elsevier Ltd. All rights reserved. Keywords: BCG; HIV; Disseminated; Incidence
1. Introduction Bacille Calmette-Guerin (BCG), a live attenuated Mycobacterium bovis strain vaccine, is routinely given to infants at birth or shortly thereafter, in regions where tuber-
∗
Corresponding author. Tel.: +27 21 9389177; fax: +27 21 9389719. E-mail addresses: [email protected], [email protected] (A.C. Hesseling). 0264-410X/$ – see front matter © 2006 Elsevier Ltd. All rights reserved. doi:10.1016/j.vaccine.2006.07.020
culosis is endemic. The World Health Organization (WHO) currently recommends that BCG be given to all asymptomatic infants, irrespective of human immunodeficiency virus (HIV) exposure [1]. This policy has practical limitations, as most HIV-exposed infants are infected perinatally, and are therefore asymptomatic at birth. Vaccination with BCG has remained the standard of care for tuberculosis prevention in most developing countries because of its documented efficacy in preventing life-threatening forms of disease in HIV-uninfected infants and young children. It is the
A.C. Hesseling et al. / Vaccine 25 (2007) 14–18
only currently available vaccine, is inexpensive and requires only one encounter with an infant [2]. In the absence of HIV infection, BCG-associated adverse events in children include injection-site lesions, adenitis, suppurative adenitis and very rarely, disseminated disease. Adverse events vary with vaccine strain type, physical–chemical properties, bacillary load, administration method and host immune characteristics [3]. The reported frequency of disseminated BCG disease has traditionally been less than five per one million vaccinees and is mainly associated with congenital cellular immunodeficiency [4]. Despite several case reports of local and disseminated BCG disease in HIV-infected infants [5–9], the risk of disseminated BCG disease in children vertically infected with HIV remains poorly quantified, partly due to a lack of surveillance and adequate population-based data in HIV-endemic settings. Disseminated BCG disease in HIV-infected children is likely to be underreported, especially in settings highly endemic for tuberculosis and HIV, where diagnostic facilities are often limited [10]. Even where mycobacterial culture is available, disseminated BCG disease may be misdiagnosed as tuberculosis in HIV-infected children, in the absence of accurate Mycobacterium tuberculosis complex speciation [11]. This preliminary reports aims to estimate the risk of disseminated BCG disease in HIV-infected infants in a setting highly endemic for tuberculosis and HIV, using available paediatric population data and known BCG coverage rates and analyzing different hypothetical scenarios for vertical transmission of HIV combined with known maternal antenatal HIV prevalence in the public health sector.
2. Methods 2.1. Study setting This study was conducted from January 2002 to December 2004 at the Tygerberg Children’s Hospital, Western Cape Province, South Africa, which drains 30–50% of provincial tertiary paediatric referrals. The adult tuberculosis incidence in the province was 678/100,000 in 2003, the incidence of tuberculosis in children <13 years of age was 407/100,000 and the HIV prevalence among women attending public antenatal care facilities was 15.4% (95% CI: 12.5–18.2%) in 2004 [12–14]. Current South African BCG vaccination policy recommends universal intradermal vaccination at birth with Danish Strain BCG (1331, Statens Serum Institute, Copenhagen, Denmark) in the right deltoid region. 2.2. Risk calculation Disseminated BCG disease in HIV-infected children is predominantly seen in children <1 year of age with rapid progression to advanced HIV disease [5,8,9]. Therefore, we included all HIV-infected and uninfected children <1 year
15
of age diagnosed with disseminated BCG disease during the study period. 2.2.1. Numerator Cases of distant/disseminated BCG disease were detected through active prospective hospital laboratory-based and clinical surveillance. M. bovis BCG was detected through routine polymerase chain reaction (PCR) speciation of M. tuberculosis complex isolates [11] after mycobacterial culture by the automated Middlebrook 7H9 broth-based Mycobacterial Growth Indicator Tube method (MGIT; Becton Dickinson, Sparks, MD, USA), following rigorous measures to prevent mycobacterial cross-contamination [15]. Additionally, the Western Cape Provincial Expanded Programme on Immunization (EPI) register was reviewed to ensure that all routinely notified disseminated BCG disease cases over the study period were also included. Distant/disseminated BCG disease was confirmed by isolation of M. bovis BCG from a distant site beyond the ipsilateral axillary gland, e.g. from gastric aspiration, in the presence of systemic symptoms and local and/or regional evidence of BCG disease, based on a standard paediatric BCG disease classification [9]. BCG vaccination was documented both from neonatal vaccination records and the presence of a scar in the right deltoid region. HIV infection was diagnosed through enzyme-linked immunosorbent assay and confirmed by HIV DNA PCR. Informed consent was obtained for study participation and HIV testing from parents or legal guardians. The study was approved by the Ethics Committee, Faculty of Health Sciences, Stellenbosch University. 2.2.2. Denominator Mid-year projected population data on the total number of live children <1 year of age in the province were used to calculate denominator data [16]. This projected figure accounts for both the total number of live births and infant mortality. In the absence of reliable population-based routine surveillance data on the exact number of live HIV-infected children <1 year of age in the Western Cape Province over the study period, we used the projected mid-year population for the study period of children <1 year to estimate the total number of HIV-infected infants, given plausible scenarios of mother-to-child transmission of HIV for the study setting and the routinely documented HIV prevalence amongst women attending public antenatal clinics. Scenarios 1–3 assumed vertical HIV transmission rates of 5, 10 and 15%, respectively; this was applied to the maternal HIV prevalence data over the study period (Table 1) [14]. BCG vaccination coverage of 95% was assumed, based on a recent provincial vaccine coverage survey [17]. In the same manner as for scenarios 1–3, the total annual number of live children <1 year of age less the number of HIV-infected children was used as denominator to estimate the risk of disseminated BCG disease in HIV-uninfected children.
16
A.C. Hesseling et al. / Vaccine 25 (2007) 14–18
Table 1 Estimates of the total mid-year population of children <1 year, Western Cape Province, South Africa, along with the number of HIV-infected children vaccinated with BCG based on the known maternal antenatal HIV prevalence and assumed rates of vertical HIV transmission
Maternal HIV prevalence Routine provincial HIV surveillance data [14]a Total number of infants <1 year of age [16]b Scenario 1 Assuming 5% total HIV vertical infection Total number of HIV-infected infants vaccinated assuming 95% neonatal BCG coverage Scenario 2 Assuming 10% total HIV vertical infection Total number of HIV-infected infants vaccinated assuming 95% neonatal BCG coverage Scenario 3 Assuming 15% total HIV vertical infection Total number of HIV-infected infants vaccinated assuming 95% neonatal BCG coverage
2002
2003
2004
12.4% (95% CI: 8.8–15.9%)
13.1% (95% CI: 8.5–17.7%)
15.4% (95% CI: 12.5–18.2%)
96,946
97,778
98,339
601 571
640 608
757 719
1202 1142
1281 1217
1514 1439
1803 1713
1921 1825
2272 2158
Scenarios 1–3 used mid-year population estimates of total children <1 year of age, and assumed a vertical HIV transmission risk of 5, 10 and 15%, respectively, given the routine maternal HIV surveillance, to calculate the total number of HIV-infected infants. BCG coverage rates from the Western Cape Provincial vaccination survey, 2005 [17]. All numbers are rounded to the closest integer. a Routine provincial maternal antenatal surveillance data, as measured through annual sentinel surveillance. This only relates to the public health sector; no surveillance data are available for the private health sector. b ASSA 2003 model, Actuarial Society, South Africa; mid-year population estimates based taking into account total number of live births and infant mortality rate.
3. Results Nine children were diagnosed with disseminated BCG disease over the study period. All these children had immune deficiencies. Seven children were HIV-infected; all had advanced HIV disease and were <1 year of age; the mortality was 85.7% (6/7). The two HIV-uninfected children with disseminated BCG disease, both had primary immune deficiencies including one child with severe combined immunodeficiency and a second with an unidentified T-cell deficiency. Detailed data on the clinical presentation, HIV disease characteristics, BCG disease classification, management and outcome of all children with distant/disseminated BCG disease are reported elsewhere [9]. No additional cases of dissemi-
nated BCG disease were identified from the provincial notification register for the study period. Denominator data calculations resulted in a total number for 2002–2004 of 571–719 HIV-infected infants if 5% vertical HIV transmission was assumed, 1142–1439 HIVinfected infants if 10% vertical transmission was assumed and 1713–2158 HIV-infected infants if 15% vertical HIV transmission was assumed, respectively, for the study period. The total number of distant/disseminated BCG cases per study year and the calculated risks of distant/disseminated BCG disease in HIV-infected infants for 2002–2004 are indicated in Table 2. The calculated incidence of distant or disseminated disease in HIV-uninfected infants was 0.72–0.74 per 100,000 per year for the study period.
Table 2 Calculated risk of distant or disseminated BCG disease in HIV-infected children based on the known maternal antenatal HIV prevalence, assumed rates of vertical HIV transmission and estimates of total mid-year population of children <1 year, Western Cape Province, South Africa Risk scenarios of disseminated BCG disease
Cases/year 2002
Cases/year 2003
Cases/year 2004
Actual cases/year Risk of disseminated BCG disease Case scenario 1, assuming 5% total vertical HIV infection Case scenario 2, assuming 10% total vertical HIV infection Case scenario 3, assuming 15% total vertical HIV infection
2
2
3
2/571 = 350/100,000/year 2/1142 = 175/100,000/year 2/1713 = 117/100,000/year
2/608 = 329/100,000/year 2/1217 = 164/100,000/year 2/1825 = 110/100,000/year
3/719 = 417/100,000/year 3/1439 = 208/100,000/year 3/2158 = 139/100000/year
Routine provincial maternal antenatal surveillance data, as measured through annual sentinel surveillance [14]. This only relates to the public health sector; no surveillance data are available for the private health sector. ASSA 2003 model, Actuarial Society, South Africa [16] mid-year population estimates based taking into account total number of live births and infant mortality rate. Scenarios 1–3 used mid-year population estimates of total children <1 year of age, and assumed a vertical HIV transmission risk of 5, 10 and 15%, respectively, given the routine maternal HIV surveillance, to calculate the total number of HIV-infected infants. BCG coverage rates from the Western Cape Provincial vaccination survey, 2005 [17]. All numbers are rounded to the closest integer.
A.C. Hesseling et al. / Vaccine 25 (2007) 14–18
4. Discussion We have demonstrated that neonatal BCG vaccination may pose a considerable risk of distant or disseminated BCG disease to infants vertically infected with HIV, even when they are asymptomatic when vaccinated at birth. The estimated risk of disseminated BCG disease in HIV-infected children was increased several hundred fold compared to that historically described in HIV-negative children. Further surveillance of BCG safety is critically important to more accurately estimate the risk of BCG vaccination in HIVinfected infants, as these preliminary risk calculations have important limitations: we did not have reliable populationbased paediatric HIV data for the study period, and therefore used modeled data and assumptions to obtain denominator estimates for the total number of HIV-infected infants. We addressed some of these limitations by using different hypothetical scenarios for vertical HIV transmission. The rate of vertical HIV transmission depends on multiple factors including antiretroviral therapy in the mother as part of prevention of mother-to-child HIV transmission (PMTCT) regimens or as full (triple) antiretroviral therapy, breastfeeding practices and method of delivery. There is an active PMTCT program in the Western Cape Province, South Africa. By March 2003, the Western Cape Department of Health had achieved full roll-out of the programme in all districts in the province with uniform access for pregnant women attending public-sector maternity services [23]. The current PMTCT protocol recommends universal HIV testing of all pregnant women in the antenatal services and uses combination preventive drug regimens (AZT and NVP) during pregnancy, intrapartum and in the infant after birth. Women who have a CD4+ T lymphocyte count of ≤200 ul−1 at the first antenatal booking are offered triple-drug therapy. The vertical HIV transmission rate in the Cape Metropolitan area was approximately 4–5% in 2004 (personal communication, Dr. Pren Naidoo, Cape Town City Health), at a time when the PMTCT regimen had been widely implemented. We therefore consider the vertical HIV transmission rate scenario of 5% the most realistic for the study period. Vertical HIV transmission is however likely to be variable in other study settings, due to variable availability and uptake of PMTCT services. These risk calculations are likely to be a considerable under-estimate of the true disseminated BCG disease incidence in HIV-infected infants for the following reasons: we assumed that all tertiary referrals for disseminated BCG disease would have been seen at the study hospital, which is unlikely, given that only 30–50% of children in the province are eligible for referral. We only included numerator data from this hospital, as ongoing prospective surveillance is routinely conducted using standard criteria and case definitions. Further reasons for possible under-estimation are the stringent case definitions used and the possibility that HIVinfected children with disseminated BCG disease may have been misdiagnosed or died of HIV-or BCG-related causes
17
before referral. Despite these limitations, the validity of our estimations is supported by the risk calculation of disseminated BCG disease in HIV-uninfected children, which was in the same order of magnitude as that described in the literature. These preliminary risk calculations are therefore useful to quantify the direct risk posed by BCG to HIV-infected infants, given different scenarios of vertical HIV transmission. More extensive surveillance measures are likely to yield more exact estimates in the future. We are unable to comment on the potential non-specific effects BCG may have in HIVinfected infants. Apart from risk estimation, an assessment of the protective effect of BCG in HIV-infected children is important, given the extremely high risk of infection with M. tuberculosis in highly tuberculosis-endemic settings and the high risk of tuberculosis disease progression in HIV-infected children following M. tuberculosis infection [18,19]. In HIV-unexposed and uninfected children, evidence of the protective effect of BCG against disseminated forms of tuberculosis in the first 2 years of life is consistent, with reported efficacy ranges from 46 to 100% [3,20]. In a recent meta-analysis, BCG has also been shown to be cost-effective, preventing nearly one case of disseminated tuberculosis for roughly every 9300 inoculations, mostly in Southeast Asia (46%) and Africa (27%) [21]. However, these cost-effectiveness analyses did not quantify BCG benefits in HIV-infected children. Equal protective effect in HIV-infected children cannot be assumed, as the available epidemiological data on the protective effect of BCG in HIVexposed or infected children is extremely limited [1]. Most BCG efficacy and safety trials were conducted in the preHIV era or in settings where the burden of paediatric HIV was very low. As the mechanism of protection derived from BCG involves a reduction of the haematogenous spread of bacilli from the site of primary infection [22], it is plausible that when a specific T-cell mediated response is required for control of haematogenous spread of M. tuberculosis, the suppression of T-cell immunity as a result of HIV may render this defence inadequate in children. However, in settings highly endemic for tuberculosis and HIV, even partial protection due to BCG may confer considerable benefit to HIV-infected infants [10]. Based on the considerable risk of disseminated BCG disease in HIV-infected infants, possible vaccination policy implications include the more rapid identification of HIVinfected infants through PMTCT programs in HIV-endemic settings, and delaying BCG in HIV-exposed infants until their HIV status has been determined. As data on the protective effect of BCG in HIV-exposed and infected infants are extremely limited, more research is urgently needed in this area. Prospective surveillance incorporating hospital and population-based strategies in settings highly endemic for HIV and tuberculosis are vitally important to more accurately quantify the risks and benefits of BCG in specific sub-populations. New tuberculosis vaccine strategies should consider both the potential risks and benefits of live vaccines in HIV-infected children.
18
A.C. Hesseling et al. / Vaccine 25 (2007) 14–18
Acknowledgements The authors would like to thank Robin Warren, DST/NRF Centre of Excellence and Department of Medical Biochemistry, Faculty of Health Sciences, Stellenbosch University for PCR speciation, Wendy Brittle for mycobacterial culture, Cape Town City Health (Pren Naidoo and Ivan Toms), the Western Cape Province Department of Health (Keith Cloete and Fawzia Desai) for advice on immunization and HIV surveillance data and the South African Actuarial Society (Leigh Johnson and Keith Dorrington), for use and advice of the ASSA AIDS 2003 model. The authors would also like to thank the TB 21st Century Consortium, supported by the Norwegian Scientific Council and the Harry Crossly Foundation Trust for financial support. This manuscript is in partial fulfilment of a Ph.D. thesis. Conflict of interest: None declared.
References [1] World Health Organization. BCG vaccine. WHO position paper. Wkly Epidemiol Rec 2004;79:27–38 [accessed on 15 July 2006] http://www.who.int/wer. [2] Girard MP, Fruth U, Kieny M. Vaccine 2005;23:5725–31. [3] Fine PEM, Carneiro IAM, Milstien JB, Clements C. Issues relating to the use of BCG in immunization programmes: a discussion document WHO/V&B/99.23. Geneva: World Health Organization; 1999. [4] Casanova JL, Jouanguy E, Lamhamedi S, Blanche S, Fischer A. Immunological conditions of children with BCG disseminated infection. Lancet 1995;346:581. [5] Fallo A, Torrado L, Sanchez A, Cerquiero C, Shadrosky L, Lopez EL. Delayed complications of bacillus Calmette-Guerin vaccination in HIVinfected children. In: [Abstract] Program and abstracts of the Third IAS Conference on HIV Pathogenesis and Treatment. 2005 [WeOa0104]. [6] Lallemant-Le Coeur S, Lallemant M, Cheynier D, Nzingoula S, Drucker J, Larouze B. Bacillus Calmette-Guerin immunization in infants born to HIV-1-seropositive mothers. AIDS 1991;5:195– 9. [7] O’Brien KL, Ruff AJ, Louis MA, et al. Bacillus Calmette-Guerin complications in children born to HIV-1-infected women with a review of the literature. Pediatrics 1995;95:414–8.
[8] Hesseling AC, Schaaf HS, Hanekom WA, et al. Danish bacille Calmette-Guerin vaccine-induced disease in human immunodeficiency virus-infected children. Clin Infect Dis 2003;37:1226–33. [9] Hesseling AC, Rabie H, Marais BJ, Manders M, Lips M, Schaaf HS, et al. Bacille Calmette-Guerin vaccine-induced disease in HIV-infected and HIV-uninfected children. Clin Infect Dis 2006;42:548–58. [10] Talbot EA, Perkins MD, Silva SF, Frothingham R. Disseminated bacille Calmette-Guerin disease after vaccination: case report and review. Clin Infect Dis 1997;24:1139–46. [11] Talbot EA, Williams DL, Frothingham R. PCR identification of Mycobacterium bovis BCG. J Clin Microbiol 1997;35:566–9. [12] Cape Town TB Control. Progress report 1997–2003, available at: http://www.hst.org.za/publications/618 [accessed on 15 July 2006]. [13] Marais BJ, Hesseling AC, Gie RP, Schaaf HS, Beyers N. The burden of childhood tuberculosis and the accuracy of community-based surveillance data. Int J Tuberc Lung Dis 2006;10:259–63. [14] Department of Health. National HIV and Syphilis antenatal seroprevalence survey 2004, available at: www.doh.gov.za/docs/reports/ 2004/hiv-syphilis.pdf [accessed on 15 July 2006]. [15] Carroll NM, Richardson M, van Helden PD. Criteria for identification of cross-contamination of cultures of Mycobacterium tuberculosis in routine microbiology laboratories. J Clin Microbiol 2003;41:2269. [16] Actuarial Society of South Africa, AIDS model 2003, http://www.assa. org.za [accessed on 15 July 2006]. [17] Corrigal J. Vaccination coverage of the Western Cape Province: household survey 2005. Child Health and EPI sub-directorate of the Western Cape Province. [18] Mukadi YD, Wiktor SZ, Coulibaly, et al. Impact of HIV infection on the development, clinical presentation, and outcome of tuberculosis among children in Abidjan, Cote d’Ivoire. Aids 1997;11:1151–8. [19] Marais BJ, Gie RP, Schaaf HS, Hesseling AC, Obihara CC, Starke JJ, et al. The natural history of childhood intra-thoracic tuberculosis: a critical review of literature from the pre-chemotherapy era. Int J Tuberc Lung Dis 2004;4:392–402. [20] Rodrigues LC, Diwan VK, Wheeler JG. Protective effect of BCG against tuberculous meningitis and miliary tuberculosis: a metaanalysis. Int J Epidemiol 1993;6:1154–8. [21] Trunz BB, Fine P, Dye C. Effect of BCG vaccination on childhood tuberculous meningitis and miliary tuberculosis worldwide: a meta-analysis and assessment of cost-effectiveness. Lancet 2006;367:1173–80. [22] Arbelaez MP, Nelson KE, Munoz A. BCG vaccine effectiveness in preventing tuberculosis and its interaction with human immunodeficiency virus infection. Int J Epidemiol 2000;6:1085–91. [23] Western Cape Department of Health. Prevention of mother-to-child transmission of HIV (PMTCT), available at http://www.capegateway. gov.za/health [accessed on 15 July 2006].