Clinical Infectious Diseases Advance Access published June 9, 2015

Transcription

1 Clinical Infectious Diseases Advance Access published June 9, Development of BCG scar and subsequent morbidity and mortality in rural Guinea- Bissau Line Storgaard 1,2, Amabelia Rodrigues 1, Cesario Martins 1, Bibi Uhre Nielsen 1,2, Henrik Ravn 1,2,3, Christine Stabell Benn 1,2,3, Peter Aaby 1,2,3, Ane Bærent Fisker 1,2 1 Bandim Health Project, INDEPTH Network, Apartado 861, Bissau, Guinea-Bissau 2 Research Center for Vitamins and Vaccines (CVIVA), Bandim Health Project, Statens Serum Institut, Artillerivej 5, 2300 Copenhagen S, Denmark 3 OPEN, Institute of Clinical Research, University of Southern Denmark/Odense University Hospital, 5000 Odense C, Denmark Correspondence to: Line Storgaard, Research Center for Vitamins and Vaccines (CVIVA), Bandim Health Project, Statens Serum Institut, Artillerivej 5, 2300 Copenhagen S, Denmark; xks361@alumni.ku.dk Alternate author: Ane Bærent Fisker, Research Center for Vitamins and Vaccines (CVIVA), Bandim Health Project, Statens Serum Institut, Artillerivej 5, 2300 Copenhagen S, Denmark; a.fisker@bandim.org 40-words summary: Among BCG-vaccinated children having a BCG scar was associated with lower mortality: 52% (10-74%) during the first year of life and 26% (4-44%) during the first 5 years. Scar-positive children were also less likely to be admitted to hospital. The Author Published by Oxford University Press on behalf of the Infectious Diseases Society of America. All rights reserved. For Permissions, please journals.permissions@oup.com.

2 2 ABSTRACT Background: Previous studies have found that Bacillus Calmette-Guerin (BCG) vaccination has non-specific beneficial effects on child survival, especially among children who developed a BCG scar. These studies have mostly been done in settings with a high scar frequency. In rural Guinea-Bissau many children do not develop a scar; we tested the hypothesis that among BCG-vaccinated children, a vaccination scar was associated with lower mortality and fewer hospital admissions. Methods: During , children under 5 years of age in villages followed by Bandim Health Project s demographic surveillance system had their scar status assessed at 6-monthly visits. We compared mortality and hospital admission rates of scar-positive and scar-negative BCG-vaccinated children during 6 months of follow-up in Cox proportional hazards models. Results: Among 15,911 BCG-vaccinated children only 52% had a scar. There were 106 non-injury-related deaths among scar-positive children and 137 among scar-negative children. The mortality rate ratio (MRR) was 0.74 (95% CI= ) overall; 0.48 ( ) in infancy, 0.69 ( ) in the second year of life and 0.89 ( ) in the third-fifth year. The association between scar positivity and lower mortality differed significantly by cause of death and was strongest for respiratory infections (MRR=0.20 ( )). There were 99 hospital admissions among scarpositive children and 125 admissions among scar-negative children, resulting in an incidence rate ratio of 0.74 ( ). Conclusions: Among BCG-vaccinated children having a scar is associated with lower mortality and morbidity also in a setting with low scar prevalence. BCG scar prevalence may be an important marker of vaccination programme quality.

3 3 INTRODUCTION Bacillus Calmette-Guerin (BCG) vaccine is recommended at birth in low income countries to induce immunity against tuberculosis (TB). At present about 100 million children receive BCG vaccine each year[1]. Intradermal BCG vaccination is normally followed by development of a papule which ulcerates and heals spontaneously leaving a permanent scar within a few months[1]. Several observational studies from Guinea-Bissau in West Africa[2-5] and elsewhere[6-9] have suggested that the BCG vaccine has strong positive effects beyond protecting against TB. Randomised trials among low birth-weight children (<2500 g), for whom BCG is normally delayed, have shown that early BCG vaccination reduces neonatal mortality by 48%[10, 11]. In a randomised trial of BCG revaccination, child mortality was reduced by 64% when BCG was given to 19-month-old children who had received booster DTP[12]. TB does not cause 48-64% of early childhood deaths[13] and thus these strong beneficial effects cannot be ascribed to protection against TB. The reduction in neonatal mortality was due to fewer cases of sepsis and respiratory infections among the BCG-vaccinated neonates[10, 11]. Recent immunological studies have shown that BCG induces epigenetic modulations in monocytes and increased responsiveness against unrelated stimuli, thus providing a plausible immunological mechanism behind the beneficial non-specific effects[14, 15]. Prospective studies from urban Guinea-Bissau have demonstrated a survival advantage during the first years of life among those who developed a scar versus those who did not develop a scar after BCG vaccination[3, 5, 16]. This was unlikely to be due to resistance to TB since exclusion of children exposed to TB at home did not change the association[3, 5, 16]. Not all BCG-vaccinated children develop a BCG scar. In studies from urban Guinea-Bissau the prevalence was 72-97%[3, 5, 16, 17] which is similar to the scar prevalence found in other sub-saharan studies[18-20]. Correct vaccination technique is important for scar development[1, 21] and differences in vaccination technique may explain some of the observed variations. However, studies have also suggested that different BCG strains have different capacity to generate a scar[5, 22]. In the present study we examined the association between BCG scar and overall mortality and hospital admissions among BCG-vaccinated children in rural Guinea-Bissau. We hypothesised that those having a BCG scar had lower

4 4 mortality and fewer hospital admissions during the first 5 years of life. We also investigated whether the association depended on the size of the scar. METHODS Setting and study population The study was conducted in Guinea-Bissau, where children receive BCG at government health centres and hospitals throughout the country. The Bandim Health Project s Health and Demographic Surveillance System (HDSS) follows 182 randomly selected clusters of 100 women of fertile age and their children in rural Guinea-Bissau. Children are followed from first registration until migration, death or 5 years of age (Supplementary Material). At all visits vaccination status is assessed by inspection of the vaccination card and the child s upper-arm is inspected for a BCG scar. If a scar is found, two perpendicular diameters are measured. It is noted which fieldworker read the scar. The study population for the present study was extracted from the HDSS cohort. All BCG-vaccinated children less than 5 years of age visited between 01 July 2009 and 30 June 2011 were eligible for the study if they had been present for inspection of their arm for a scar. Children without a documented BCG vaccination or vaccinated less than 31 days prior to scar reading were excluded from the analysis (Supplementary Material). Statistical methods First, we compared background variables between scar-positive and scar-negative children using data from the first visit with a scar reading (Supplementary Material). Second, in a Cox proportional hazards model with age as underlying time we compared mortality and hospital admission rates among scar-positive and scar-negative children. The analyses followed children from the visit where scar status was assessed during the subsequent 6 months or to the next visit, migration, death, or hospital admission (Supplementary material). We adjusted the 95% confidence interval (CI) for village cluster using robust variance estimates. We report the estimates for the entire population and stratified by sex and age (Supplementary material).

5 5 To assess a possible impact of scar size on mortality, we defined scar size as the average of the horizontal and vertical diameter. Children were divided into three groups: no scar, below and above the median. Mortality rates in the three groups were compared in the Cox-model. Previous studies have found lower mortality among children vaccinated early[3, 23]. We investigated whether the association between scar positivity and mortality varied by age of BCG vaccination (1 st, 2 nd -4 th, >4 th week). All background variables were included one by one in the Cox proportional hazards model to see how much each variable changed the unadjusted mortality rate ratio (MRR). The estimate was compared with an unadjusted estimate for the same records. Any variable changing the MRR by 10% or more would be included in the Cox proportional hazard model to obtain adjusted MRRs for scar-positive versus scar-negative children. The robustness of the results was tested in sensitivity analyses (Supplementary material). Data were analysed using Stata 11.2 (StataCorp, College Station, TX). All tests were 2-sided. RESULTS Between 1 July 2009 and 30 June 2011 the mobile teams of Bandim Health Project surveyed 26,975 children. Among these, 16,014 (59%) children were eligible for analysis by being less than 5 years old, present for scar inspection and BCG-vaccinated more than 30 days earlier (Figure 1). Scar reading was conducted at 35,416 (98%) of the visits; at 18,535 (52%) visits the fieldworker registered and measured a scar while at 16,881 visits (48%) the fieldworker registered that there was no scar. With more than 35,000 visits over 5 years and 6 months of follow-up at any visit there were around 3,500 children under observation at any given time (Figure 2). Determinants for scar development Among BCG-vaccinated children, several background variables were associated with having a BCG scar: Scar prevalence was lower among girls, the relative risk (RR) for girls compared with boys being 0.96 ( ). Scar prevalence was higher in the rainy than the dry season (RR=1.05 ( )). Scar prevalence also varied by sequence of vaccinations, place of birth, health region, fieldworker who conducted the scar reading, and maternal ethnicity but did not depend on the nutritional status of the child or the mother (Table 1).

6 6 Mortality During the 6-month follow-up period 249 deaths were registered, the overall mortality rate being 16/1000 person years of observation (Figure 1). Six deaths were injury-related leaving 106 deaths among scar-positive children and 137 deaths among scar-negative children, the MRR being 0.74 ( ) (Table 2, Figure 2). The MRR estimates were 0.66 ( ) for boys and 0.82 ( ) for girls (p=0.57 for interaction between scar status and sex on mortality). The difference in mortality between scar-positive and scar-negative children declined with age: When survival was examined during first, second and third-fifth year of life, the estimates were 0.48 ( ), 0.69 ( ) and 0.89 ( ) (p=0.27 for interaction between scar status and year of life on mortality). Scar size had no marked impact on the mortality rates (Supplementary material). When the analysis was stratified by age at vaccination, the stratum specific estimates were not homogenous. A scar was associated with lower mortality among children vaccinated in the first week of life (MRR=0.40 ( )) and in the subsequent three weeks (MRR=0.47 ( )) but we observed no difference between scar-positive and scarnegative children vaccinated after the neonatal period (MRR=0.98 ( )) (Table 3). Hence, the association was significantly modified by age at vaccination (p=0.01 for interaction between scar status and age at vaccination on mortality). No sex-difference was seen in either of the vaccination age groups (p=0.88, p=0.97 and p=0.57 for interaction between scar status and sex on mortality in each age group, respectively). All background variables were tested for their ability to change the MRR; none of the variables changed the estimate by more than 1.2% and no adjusted analysis was made. Cause of death Based on interviews conducted following the deaths, fever without other reported symptoms was reported as the cause for 60% (145/243) of the deaths. Diarrhoea was the major symptom in 25% (61/243) and cough or difficulties breathing was the major symptom in 12% (28/243) (Supplementary Table 2). The association between scar positivity and mortality differed significantly by cause of death: The association was particularly strong with respect to respiratory symptoms (MRR=0.20 ( )) whereas associations were weaker for fever (MRR=0.81 ( )) and diarrhoea (MRR=0.88 ( )) (p=0.04 for same effect) (Supplementary Table 2).

7 7 Admission to hospital In this rural area hospitalisations were rare, the overall admission rate being 14/1000 person years of observation. There were 99 hospital admissions among scar-positive children and 125 among scar-negative children (Figure 1). Compared with children without a scar, children with a BCG scar had a lower overall hospital admission rate, the incidence rate ratio (IRR) being 0.74 ( ) (Table 4). For boys and girls the estimates were 0.90 ( ) and 0.55 ( ), respectively (p=0.19 for interaction between scar status and sex on hospital admission). The association between scar positivity and lower morbidity for girls was consistent throughout the first 5 years of life: IRR=0.43 ( ) during the first, 0.45 ( ) the second, and 0.41 ( ) the third to fifth year of life; no association was seen for boys at any time (Table 4). Sensitivity analyses Some fieldworkers classified fewer or more children as scar-positive (Table 1, Supplementary Table 3). Excluding children with readings done by fieldworkers >=15% from the median or stratifying by fieldworker had no effect on the results (Supplementary material). Some children may take longer to develop scars (Supplementary Figure 1) or smaller scars may be difficult to see over time. Excluding assessments made within 2, 3, 4 or 5 month after vaccination or keeping the first scar assessment fixed for up to 2 years did not change the conclusions (Supplementary material). DISCUSSION Main results Only approximately half of the children developed a BCG scar. Among BCG-vaccinated individuals, children with a BCG scar had a lower overall mortality rate compared with children without a scar. The effect was strongest for children vaccinated within the first month of life. Infant mortality for scar-positive children was half the mortality for scar-negative children. The mortality difference between scar-negative and scar-positive children diminished but was still present during the second and third-fifth year of life. Scar-positive children were also less likely to be admitted to hospital. Strengths and weaknesses This study was based on a country-representative population; all regions outside the capital, Bissau being represented. The surveyed households have been routinely visited for years and the fieldworkers were experienced and frequently

8 8 supervised during the study period. A source of bias is the accuracy with which scar status has been classified. Due to the prospective study design the outcome was unknown when scar status was assessed and the fieldworkers were not aware of the hypothesis. Exclusion of readings by the most extreme scar-reading fieldworkers, or stratifying the analysis by fieldworker, did not alter the conclusions. The fieldworkers are trained to use local expressions to explain symptoms to the respondent in the verbal autopsy interview. However, the symptom data are crude and should be interpreted cautiously; while the results did suggest that a BCG scar is particularly associated with fewer deaths from respiratory infections these deaths were only a small proportion of all deaths. Consistency with previous studies We found remarkably lower scar prevalence in the rural Guinean cohort (52%) than has been found previously in the urban Guinean cohorts (72-97%)[3, 5, 16, 17]. A possible explanation is that the vaccine strain used in rural Guinea- Bissau was the Russian strain produced by Serum Institute of India. The Russian strain has been associated with fewer scars compared with other BCG strains[22]. In contrast the urban cohorts were vaccinated with Merieux (Pasteur Merieux, France)[3, 5, 16], Connaught (Intervax Biologicals Ltd., Canada)[5], or the Danish strain (Statens Serum Institut, Denmark)[3, 5, 16, 17]. Three prospective studies in urban Guinea-Bissau have previously shown that among BCG-vaccinated children, scar positivity was associated with reduced childhood mortality[3, 5, 16]. This study corroborated these findings in a large cohort from a rural setting with much lower scar prevalence and notably also with less than half the infant mortality seen in the previous studies. Two of the previous studies compared mortality until the age of 1 year, the adjusted MRR being 0.45 ( )[16] and 0.44 ( )[5], very similar to what we found during first year of life (MRR=0.48 ( )[3, 5, 16]. Also consistent with the previous studies[3, 5, 16], the association between scar positivity and lower mortality weakened with age but was still present during the second year of life. In contrast to the beneficial nonspecific effects of a scar among BCG-vaccinated children, a large study from Malawi did not find that having a BCG scar protected against TB in the general population[24]. Only one of the previous studies has examined the influence of having a scar on different causes of death and found no difference for cause of death[16]. However, in case-control studies from Guinea-Bissau and Brazil, BCG vaccination

9 9 was associated with reduced risk of lower respiratory tract infections[25, 26]. Furthermore, a recent paper describes that being BCG-vaccinated was associated with a lower prevalence of respiratory symptoms in the past 14 days[6]. Interpretation Immunological studies have shown that BCG vaccination induces epigenetic modifications of monocytes, resulting in increased pro-inflammatory monocyte-derived cytokine response when subsequently exposed to unrelated bacterial and fungal pathogens[14, 27]. This phenomenon may provide an explanation for the non-specific effects of BCG. The present study suggests that the non-specific effects of BCG are more pronounced among children who develop a scar, and thus the scar can be seen as a sign of a generally enhanced resistance towards pathogens. We assessed several underlying factors related to the child and mother for their association with both scar and mortality to identify possible confounders, but none of these factors could explain the association between scar positivity and lower mortality. It could be speculated that the association between scar and mortality might reflect host immune characteristics prior to BCG vaccination, i.e. neonates with impaired or immature immune system may fail to form a scar after BCG vaccination and also fail to fight infection, a healthy reactor effect. HIV-infection is associated with both immune competence and mortality, and children of HIV-infected mothers were therefore excluded in another scar study[3], but it had no impact on the estimate. The HIV prevalence among pregnant women in Guinea-Bissau was 5.8 % in 2009[28], thus HIV alone cannot explain that almost half of the children in our study were scar-negative. Pre-term and low birth-weight infants have been less likely to develop a scar[21, 29, 30], but again this cannot explain the high rate of scar negativity in the present study. Also, we did not see lower scar prevalence among twins, who are often preterm and/or small for gestational age. We assessed whether children with poor nutritional status (measured by MUACfor-age) and whether children of mothers with poor nutritional status were less likely to develop a scar, but they were not. First and foremost, it would be unlikely that we found the scar prevalence so much lower than other studies from Guinea-Bissau if scar reaction is determined by an underlying biological factor. Hence, we do not believe our results reflect a healthy reactor effect. A study in urban Guinea-Bissau found vaccination technique and vaccine strain to affect the development of a vaccination scar[21]. Large post vaccination wheals, correct intradermal administration, and certain strains were associated with better scarring. An observational study in Uganda found also that scar development differed with two sequentially used vaccine strains[22]; the scar prevalence was markedly higher among children who received the Danish BCG strain, than among children who received Russian BCG[22]. The low scar prevalence in the

10 10 present study may therefore partially be due to the use of a BCG stain that causes fewer scars. Furthermore, poor vaccination technique may contribute to explain the low scar prevalence, as staff is less experienced in the rural areas of Guinea-Bissau where 1-2 nurses handle all tasks at a health centre, whereas at the larger health facilities in the capital BCG vaccination is delegated to specialized nurses. Implications Vaccines are considered one of the most successful and cost-effective health investments in history[31]. However, our study indicates that we may not obtain the maximum benefit of BCG vaccination. Improving and sustaining successful health interventions rely on continued monitoring, evaluation and redesigning of evaluation criteria to ensure that the most important parameters are assessed. At present the programme performance with regards to BCG vaccinations is assessed by measuring BCG coverage at 12 months of age. Currently programme strategy does not encourage early delivery of BCG and less than half of the children are vaccinated in the first month of life[32, 33]. The present analysis identified several potential problems in the current BCG vaccination programme which could lead to considerably lower mortality if resolved. First, BCG is more beneficial the earlier it is given, but this is not taken into consideration in the current program. Second, there is a need to monitor the quality of vaccination technique and to which extent a BCG scar is developed. Third, strain of BCG may be a determinant of BCG scarring. Since scarring is related to mortality risk, the strain of BCG used may have a major impact on the mortality level. There would seem to be a need to conduct randomized trials of different strains of BCG to identify those most effective in providing beneficial non-specific effects and to test whether revaccination with BCG will improve the survival of scar-negative children. Conclusion BCG scar was a marker of reduced mortality and reduced risk of hospital admission. Such effects should be taken into account in the planning of vaccination programmes. Much could be gained by using BCG scar prevalence and age of BCG vaccination as programme performance indicators rather than merely assessing BCG coverage by 12 month of age. Further studies are required to investigate whether revaccination of scar-negative children should be recommended.

11 11 Conflicts of interest None Funding This work was supported by Aase og Ejnar Danielsens Fond; Danish Council for Independent Research ( , ); Fonden til Lægevidenskabens Fremme, DANIDA (104.Dan.8-920), European Union FP7 support for OPTIMUNISE (Health-F ) and The Danish National Research Foundation (DNRF-108). Author contributions ABF, PA and CSB planned the study. LS, ABF, AR, CM and BN supervised data collection, data entry and data cleaning. LS analysed the data with assistance from ABF and HR. LS had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. LS wrote the first draft of the paper. ABF has primary responsibility for the final content. All authors contributed to and approved the final manuscript.

12 12 References 1. BCG vaccine. WHO position paper. Releve epidemiologique hebdomadaire / Section d'hygiene du Secretariat de la Societe des Nations = Weekly epidemiological record / Health Section of the Secretariat of the League of Nations 2004; 79(4): Kristensen I, Aaby P, Jensen H. Routine vaccinations and child survival: follow up study in Guinea- Bissau, West Africa. BMJ 2000; 321(7274): Garly ML, Martins CL, Bale C, et al. BCG scar and positive tuberculin reaction associated with reduced child mortality in West Africa. A non-specific beneficial effect of BCG? Vaccine 2003; 21(21-22): Aaby P, Jensen H, Gomes J, Fernandes M, Lisse IM. The introduction of diphtheria-tetanus-pertussis vaccine and child mortality in rural Guinea-Bissau: an observational study. International Journal of Epidemiology 2004; 33(2): Roth A, Sodemann M, Jensen H, et al. Tuberculin reaction, BCG scar, and lower female mortality. Epidemiology 2006; 17(5): Hollm-Delgado MG, Stuart EA, Black RE. Acute Lower Respiratory Infection Among Bacille Calmette- Guerin (BCG)-Vaccinated Children. Pediatrics 2014; 133(1): e Roth A, Garly ML, Jensen H, Nielsen J, Aaby P. Bacillus Calmette-Guerin vaccination and infant mortality. Expert Rev Vaccines 2006; 5(2): M El-Zein MP, A Benedetti, M Roussau. BCG vaccination protect against the development of childhood asthma? A systematic review and meta-analysis of epidemiological studies. International Journal of Epidemiology 2010; 39: Aaby P, Vessari H, Nielsen J, et al. Sex differential effects of routine immunizations and childhood survival in rural Malawi. Pediatric Infectious Disease Journal 2006; 25(8): Aaby P, Roth A, Ravn H, et al. Randomized trial of BCG vaccination at birth to low-birth-weight children: beneficial nonspecific effects in the neonatal period? J Infect Dis 2011; 204(2): Biering-Sorensen S, Aaby P, Napirna BM, et al. Small randomized trial among low-birth-weight children receiving bacillus Calmette-Guerin vaccination at first health center contact. Pediatr Infect Dis J 2012; 31(3): Roth AE, Benn CS, Ravn H, et al. Effect of revaccination with BCG in early childhood on mortality: randomised trial in Guinea-Bissau. BMJ 2010; 340: c Liu L, Johnson HL, Cousens S, et al. Global, regional, and national causes of child mortality: an updated systematic analysis for 2010 with time trends since Lancet 2012; 379(9832): Kleinnijenhuis J, Quintin J, Preijers F, et al. Bacille Calmette-Guerin induces NOD2-dependent nonspecific protection from reinfection via epigenetic reprogramming of monocytes. Proc Natl Acad Sci U S A 2012; 109(43): Aaby P, Benn CS. Saving lives by training innate immunity with bacille Calmette-Guerin vaccine. Proc Natl Acad Sci U S A 2012; 109(43): Roth A, Gustafson P, Nhaga A, et al. BCG vaccination scar associated with better childhood survival in Guinea-Bissau. International Journal of Epidemiology 2005; 34(3): Sartono E, Lisse IM, Terveer EM, et al. Oral polio vaccine influences the immune response to BCG vaccination. A natural experiment. PLoSOne 2010; 5(5): e Lockman S, Tappero JW, Kenyon TA, Rumisha D, Huebner RE, Binkin NJ. Tuberculin reactivity in a pediatric population with high BCG vaccination coverage. Int J Tuberc Lung Dis 1999; 3(1): Kheir AE, Alhaj AA, Ibrahim SA. The sensitivity of BCG scar as an indicator of previous vaccination among Sudanese infants. Vaccine 2011; 29(46):

13 Fine PE, Ponnighaus JM, Maine N. The distribution and implications of BCG scars in northern Malawi. Bull World Health Organ 1989; 67(1): Roth A, Sodemann M, Jensen H, et al. Vaccination technique, PPD reaction and BCG scarring in a cohort of children born in Guinea-Bissau Vaccine 2005; 23(30): Anderson EJ, Webb EL, Mawa PA, et al. The influence of BCG vaccine strain on mycobacteriaspecific and non-specific immune responses in a prospective cohort of infants in Uganda. Vaccine 2012; 30(12): Roth A, Jensen H, Garly M-L, et al. Low Birth Weight Infants and Calmette-Guerin Bacillus Vaccination at Birth: Community Study from Guinea-Bissau. Pediatric Infectious Disease Journal 2004; 23(6): Fine PE, Sterne JA, Ponnighaus JM, Rees RJ. Delayed-type hypersensitivity, mycobacterial vaccines and protective immunity. Lancet 1994; 344(8932): Stensballe LG, Nante E, Jensen IP, et al. Acute lower respiratory tract infections and respiratory syncytial virus in infants in Guinea-Bissau: a beneficial effect of BCG vaccination for girls - Community based case-control study. Vaccine 2005; 23(10): Niobey FM, Duchiade MP, Vasconcelos AG, de Carvalho ML, Leal Mdo C, Valente JG. [Risk factors for death caused by pneumonia in children younger than 1 year old in a metropolitan region of southeastern Brazil. A case- control study]. Revista de saude publica 1992; 26(4): Aaby P, Kollmann TR, Benn CS. Nonspecific effects of neonatal and infant vaccination: public-health, immunological and conceptual challenges. Nature immunology 2014; 15(10): SIDA SNdLcl. Rapport d Activité sur la Riposte Guinée-Bissau ce_gw_narrative_report%5b1%5dpdf. 29. Sedaghatian MR, Hashem F, Moshaddeque Hossain M. Bacille Calmette Guerin vaccination in preterm infants. Int J Tuberc Lung Dis 1998; 2(8): Sedaghatian MR, Kardouni K. Tuberculin response in preterm infants after BCG vaccination at birth. Arch Dis Child 1993; 69(3 Spec No): WHO U, World Bank. State of the world s vaccines and immunization, 3rd ed. Geneva, World Health Organization Clark A, Sanderson C. Timing of children's vaccinations in 45 low-income and middle-income countries: an analysis of survey data. Lancet 2009; 373(9674): Thysen SM, Byberg S, Pedersen M, et al. BCG coverage and barriers to BCG vaccination in Guinea- Bissau: an observational study. BMC Public Health 2014; 14(1): 1037.

14 14 Table 1: Prevalence of BCG scar by background characteristics. Only children with documented BCG vaccination Background variables Number of children Scar prevalence [P-value] a,b Factors related to the child Sex n. n. % scar Relative Risk (95% CI) [P=0.02] Boys 8,030 3, Girls 7,880 3, ( ) Twin [P=0.12] No 15,438 7, Yes ( ) Age at BCG vaccination [P=0.37] 0-7 days 2,120 1, days 4,016 1, ( ) >28 days 9,752 4, ( ) Co-administered Polio or Pentavalent/DTP with BCG [P=0.004] BCG alone 5,634 2, Co-administered vaccination 10,252 4, ( ) Co-administered Polio with BCG [P=0.01] No co-administered Polio 6,075 2, Co-administered Polio 9,810 4, ( ) Co-administered Pentavalent/DTP with BCG [P=0.15] No co-administered Pentavalent/DTP 10,024 4, Co-administered Pentavalent/DTP 5,859 2, ( ) Previous Pentavalent/DTP [P=0.38] No previous Pentavalent/DTP 9,223 4, Co-administered Pentavalent/DTP c 5,140 2, ( ) Previous Pentavalent/DTP 1, ( ) Season of vaccination [P=0.01] Dry 8,234 3, Rainy 7,657 3, ( ) Place of birth [P=<0.0001] Home 7,985 3, Health Center 1, ( ) Hospital 1, ( ) Other ( ) Missing 4,350 2, ( ) Assistant conducting the scar reading [P=<0.0001] A 1, B 2,397 1, ( ) C 2,256 1, ( ) D 1, ( ) E 1, ( ) F 1, ( ) G ( ) H 2,069 1, ( ) I ( ) J ( ) Other ( ) Region [P=<0.0001] Oio 2,073 1, Biombo 2,127 1, ( ) Gabu 2, ( ) Cacheu 1, ( ) Bafata 2,030 1, ( )

15 15 Quinera 1, ( ) Tombali 1, ( ) Bubaque ( ) Bolama ( ) Sao Domingos 1, ( ) Factors related to the mother Years of schooling [P=0.48] 0 10,230 4, ,965 1, ( ) >5 2,079 1, ( ) Ethnicity [P=<0.0001] Balanta 3,809 1, Bijogo ( ) Felupe ( ) Fula 3,149 1, ( ) Mancanha ( ) Mandinga 2,561 1, ( ) Manjaco 1, ( ) Beafada ( ) Papel 1, ( ) Other ( ) Mother has a scar [P=0.64] No 6,311 2, Yes 7,915 3, ( ) Age when child was born [P=0.47] <20 3,449 1, ,944 3, ( ) ,679 1, ( ) > ( ) Prenatal consultations [P=0.12] Yes 7,872 3, No 1, ( ) Unknown 6,730 3, ( ) MUAC at time of registration of pregnancy [P=0.34] 1st quartile (<245) 1, nd quartile ( ) 1, ( ) 3rd quartile ( ) 1, ( ) 4th quartile (> 278) 1, ( ) Unknown 9,716 4, ( ) Risk factor assessed at time of reading scar MUAC for age z-score d [P=0.23] 1st quartile ( <-1.02) 3,847 1, nd quartile (-1.01 ; -0.38) 3,796 1, ( ) 3rd quartile (-0.37 ; 0.26) 3,781 1, ( ) 4th quartile (0.27 ; 4.62) 3,798 1, ( ) a Test for no effect of background variable on scar prevalence b Excluding missing category c No previous Pentavalent/DTP d 389 children < 3 months excluded to use WHO growth for age standard references e Missing information when not specified: sex: 1, twin: 1, age at vaccination: 23, co-administered vaccine: 25, co-administered polio: 26, coadministered pentavalent/dtp: 28, previous Pentavalent/DTP: 23, season of vaccination: 20, MUAC of child: 300, years of schooling: 637, ethnicity: 199, scar: 1685, age: 280, prenatal consultations: 190

16 16 Table 2: Mortality of children with and without a BCG scar overall and by sex and age group. Only children with documented BCG vaccination All Sex a,b Boys Girls Year of life 1 st year of life 2 nd year of life 3 rd -5 th year of life Number of visits. Mortality rate/1000 person years (deaths/person years) Scar - Scar + n=16,881 n=18,535 Mortality rate ratio (95% CI) 18.5 (137 / 7,391) 13.0 (106 / 8,143) 0.74 ( ) n=8,371 n=9, (69 / 3,669) 11.9 (50 / 4,203) 0.66 ( ) n=8,509 n=8, (68 / 3,721) 14.2 (56 / 3,940) 0.82 ( ) n=3,250 n=2, (34 / 1,053) 15.8 (12 / 760) 0.48 ( ) n=5,666 n=5, (50 / 1,801) 19.2 (35 / 1,820) 0.69 ( ) n=11,437 n=13, (53 / 4,537) 10.6 (59 / 5,563) 0.89 ( ) a Excluding 1 child with missing information on sex b P=0.57 for interaction between scar status and sex on mortality

17 17 Table 3: Mortality and hospital admission of children with and without BCG scar stratified by age at BCG vaccination. Only children with documented BCG vaccination Mortality a,b 1 st week 2 nd -4 th week > 4 th week Hospital admission c,d 1 st week 2 nd -4 th week > 4 th week Number of visits. Mortality rate/1000 person years (deaths/person years) Scar - Scar + n=2,164 n=2, (26 / 949) 10.0 (11 / 1,096) n=4,236 n=4, (35 / 1,860) 8.7 (18 / 2,058) n=10,448 n=11, (76 / 4,567) 15.5 (77 / 4,984) Number of visits. Incidence rate/1000 person years (admissions/person years) Scar - Scar + n=2,164 n=2, (18 / 946) 21.1 (23 / 1,093) n=4,236 n=4, (35 / 1,854) 9.3 (19 / 2,054) n=10,448 n=11, (72 / 4,553) 11.5 (57 / 4,973) Mortality rate ratio (95% CI) 0.40 ( ) 0.47 ( ) 0.98 ( ) Incidence rate ratio (95% CI) 1.15 ( ) 0.49 ( ) 0.76 ( ) a Mortality: p=0.01 for interaction between scar status and age at vaccination on mortality b Excluding 23 children where the exact date of BCG was not known c Hospital admission: p=0.13 for interaction between scar status and age at vaccination on hospital admission d Excluding 23 children where the exact date of BCG was not known

18 18 Table 4: Hospital admission of children with and without a BCG scar overall and by sex and age group. Only children with documented BCG vaccination Number of visits. Incidence rate/1000 person years Incidence Rate Ratio (95% CI) (admissions/person years) Scar - Scar + All Sex a,b Boys Girls Year of life 1 st year of life 2 nd year of life 3 rd -5 th year of life Year of life by sex a Boys 1 st year of life 2 nd year of life 3 rd -5 th year of life Girls 1 st year of life 2 nd year of life 3 rd -5 th year of life n=16,881 n=18, (125 / 7,368) 12.2 (99 / 8,125) 0.74 ( ) n=8,371 n=9, (66 / 3,659) 15.8 (66 / 4,191) 0.90 ( ) n=8,509 n=8, (59 / 3,709) 8.4 (33 / 3,934) 0.55 ( ) n=3,250 n=2, (24 / 1,049) 19.8 (15 / 759) 0.82 ( ) n=5,656 n=5, (49 / 1,792) 18.2 (33 / 1,814) 0.66 ( ) n=11,424 n=13, (52 / 4,527) 9.2 (51 / 5,552) 0.78 ( ) n=1,569 n=1, (11 / 508) 28.5 (11 / 386) 1.24 ( ) n=2,795 n=2, (24 / 897) 22.7 (21 / 924) 0.83 ( ) n=5,686 n=7, (31 / 2,253) 11.8 (34 / 2,880) 0.84 ( ) n=1,681 n=1, (13 / 541) 10.7 (4 / 372) 0.43 ( ) n=2,860 n=2, (25 / 895) 13.5 (12 / 890) 0.45 ( ) n=5,737 n=6, (21 / 2,273) 6.4 (17 / 2,672) 0.41 ( ) a Excluding 1 child with missing information on sex b P=0.19 for interaction between scar status and sex on hospital admission

19 19 Legends for figures Figure 1: Flow chart for children who entered the analyses of the importance of having a BCG scar. Guinea- Bissau, Notes to Figure 1: a: Number of visits per child during study period: one visit (n=5279), two visits (n=4209), three visits (n=3827), four visits (n=2698), five visits (n=1) b: Doubt about scar status (n=559), missing information on scar (n=101) c: 6 injury-related deaths censored in the analysis (3 scar, 3 no scar) Figure 2: Kaplan-Meier curve comparing mortality of children aged 6 months to 5 years with and without BCG scar. Only children with documented BCG vaccination

20 20 Figure 1: Flow chart for children who entered the analyses of the importance of having a BCG scar. Guinea- Bissau, Children followed (n=26,975) Potential eligible children (n=16,014) a Total visits (n=35975) Study base: Children with arm inspected for scar (n=15,911) Visits with scar reading (n=35,416) (98%) Scar (n=18,535) (52 %) End of follow-up: Alive and living in study area (n=18,048) Moved (n=381) Dead (n=106 c ) Hospitalized (n=99) No scar (n= 16,881) 48 % End of follow-up: Alive and living in study area (n=16,402) Moved (n=342) Dead (n=137 c ) Hospitalized (n=125) Children excluded (n=10,961) Not present for scar reading at any visit / age > 5 years at visit: (n=3440) No documented BCG vaccine when scar inspected: (n=7243) Time from BCG vaccine to scar inspection < 31 days: (n=278) Visits excluded (n=559 b ) a Number of visits per child during study period: one visit (n=5279), two visits (n=4209), three visits (n=3827), four visits (n=2698), five visits (n=1) b Doubt about scar status (n=559), missing information on scar (n=101) c 6 injury-related deaths censored in the analysis (3 scar, 3 no scar)

21 21 Figure 2: Kaplan-Meier curve comparing the mortality of children aged 6 months to 5 years with and without BCG scar. Only children with documented BCG vaccination Age / years 4 5 Number at risk Scar No scar Scar No scar