Tuberculin reaction and BCG scar: association with infant mortality

Transcription

1 Tropical Medicine and International Health doi: /tmi volume 20 no 12 pp december 2015 Tuberculin reaction and BCG scar: association with infant mortality Clara Amalie Gade Timmermann 1, Sofie Biering-Sørensen 2,3, Peter Aaby 2,3, Ane Bærent Fisker 2,3, Ivan Monteiro 3, Amabelia Rodrigues 2,3, Christine Stabell Benn 2,4 and Henrik Ravn 2,4 1 Department of Environmental Medicine, Institute of Public Health, University of Southern Denmark, Odense, Denmark 2 Research Centre for Vitamins and Vaccines, Bandim Health Project, Bissau, Guinea-Bissau 3 Bandim Health Project, INDEPTH Network, Bissau, Guinea-Bissau 4 OPEN, Institute of Clinical Research, University of Southern Denmark/Odense University Hospital, Odense, Denmark Abstract objective To test the hypothesis that having a scar and a positive tuberculin skin test (TST) response after vaccination with Bacille Calmette Guerin (BCG) is associated with reduced infant mortality. methods We studied cohorts of 2709 normal-birthweight (NBW) and 1102 low-birthweight (LBW) infants in Guinea-Bissau. Children were enrolled in randomised trials between year 2002 and 2008 and received BCG vaccination at birth. BCG scars and TST responses were assessed at 2 and 6 months of age. The infants were followed for mortality to 12 months of age, and survival was analysed using Cox regression. results At age 2 months, 88% of NBW children and 91% of LBW children had a BCG scar, and 36% and 17% had a TST response, respectively. The LBW infants had nearly twofold higher mortality (4.5%) than the NBW infants (2.8%) between 2 and 12 months of age. In the LBW cohort, the adjusted mortality rate ratio (MRR) comparing children with a BCG scar with those without was 0.42 (95% CI = 0.19; 0.93). There was a similar tendency for TST positivity: MRR = 0.47 (95% CI = 0.14; 1.54). For LBW children who had both a positive TST reaction and a scar, the MRR was 0.22 (95% CI = 0.05; 0.87). For NBW children, a scar and a positive TST were associated with 20% reductions in mortality, which did not reach statistical significance. conclusion We confirmed previous observations that having a scar and a TST response after BCG vaccination is associated with lower mortality risk. The possibility of revaccinating scar-negative children should be considered. keywords tuberculin skin test, BCG scar, infant mortality, BCG vaccine, Guinea-Bissau Introduction The response to Bacille Calmette Guerin (BCG) vaccine has been evaluated by assessing whether the recipient develops a BCG scar [1] or by the tuberculin skin test (TST) [2] measuring the reaction to purified protein derivate of M. tuberculosis (PPD) [3]. Numerous observational studies and randomised trials have shown that BCG vaccination is associated with strong survival benefits, which cannot be explained by prevention of TB [4]. A decade ago, several studies were published, showing that among BCG-vaccinated children, having a BCG scar and a TST response are associated with reduced all-cause mortality [5 7]. This supports the notion that BCG has important beneficial non-specific effects and also points to the possibility of using especially BCG scar, which is easy to assess, as a vaccination programme performance indicator. However, little has happened, and BCG vaccination is still only being evaluated by means of BCG vaccination coverage by 12 months of age. The aim of this study was to verify the association between BCG scars and TST responses, on the one hand, with survival, on the other hand, and to identify determinants of TST responses and BCG scars within two cohorts of normalbirthweight (NBW) and low-birthweight (LBW) infants in urban Guinea-Bissau. Methods The study was conducted at the Bandim Health Project (BHP) in Guinea-Bissau, West Africa. BHP runs a Health and Demographic Surveillance System (HDSS) in six 2015 John Wiley & Sons Ltd 1733

2 districts of Bissau, the capital of Guinea-Bissau. All houses in the study area are visited monthly to register new pregnancies and births, and all children are followed with home visits every third month from birth until 3 years of age. Cohorts We used data from two randomised trials conducted among NBW children and LBW children, respectively. The infants were enrolled in the trials at the maternity ward of the national hospital as well as three health centres in the HDSS study area when they were due to receive their first vaccinations. They were only enrolled if they had no overt disease or malformation. As a part of the trials, visits were planned for all children in their home at 2 and 6 months of age, where BCG scars and TST responses were to be assessed. The children were included in the present analysis if they were visited at least once for assessments of BCG scars and TST responses at 2 or 6 months of age. The NBW cohort consisted of 2709 infants with a birthweight of at least 2500 g enrolled between November 2002 and November At enrolment, all children received BCG and were randomised to receive vitamin A or placebo [8]. The LBW cohort consisted of 1102 infants with a birthweight of less than 2500 g enrolled between November 2004 and March In addition to including children from the BHP study area, LBW children living outside the study area were also included if born at the maternity ward at the national hospital [9]. In Guinea- Bissau, as in many other low-income countries, the policy is to postpone BCG vaccination to LBW newborns. During this period, the LBW children were randomised to receive BCG at enrolment or the usual practice of postponed BCG [9]. Only the children randomised to BCG at enrolment were included in the present analysis. A subgroup of the LBW children born at the national hospital was furthermore randomised to vitamin A or placebo at enrolment [9, 10]. The BCG vaccine (Statens Serum Institut, Copenhagen, Denmark) was given intradermal in a dose of 0.05 ml in the upper left deltoid area. diameters using a ruler. The child was revisited 2 3 days later by a trained field worker who measured the perpendicular diameters of the TST response using a ruler. The size of both scar and TST response was defined as the mean of the two diameters [11]. If the TST response was larger than 10 mm at 2 or 6 months of age, the child was considered tuberculosis (TB) suspect [12] and referred for further examination at a TB clinic. Assessments of mortality All children in the two trials were followed for mortality to 12 months of age. The LBW children living outside the study area were visited at home at 2, 6 and 12 months of age. For the NBW and LBW children living within the study area, deaths were ascertained through the routine HDSS household visits every third month. When a child had died, an HDSS field worker performed a short verbal autopsy and a physician classified the likely cause of death. This information was used to exclude deaths from accidents; however, none of the children in the present analyses died from accidents [8, 9]. Background factors A large number of background factors were collected as part of the trials (Table 2). At inclusion, the children in the two cohorts were weighed; the children born at the maternity ward at the national hospital were also weighed at birth. Among children born at home or at a local health centre, there was no information on birthweight; for these children, the inclusion weight was used as an indicator of birthweight if the child was included within 10 days of the birth. The sequence in which BCG and diphtheria tetanus pertussis (DTP) vaccines are given might be important for child survival [13, 14]. Information about which vaccines the child had received most recently was therefore included in the analyses. The BHP follows TB patients in the study area, and children in this study were considered exposed to TB before BCG vaccination if born less than 6 months before a person in the same multifamily house was diagnosed with TB [8]. Assessments of TST responses and BCG scars At the scheduled visits at 2 and 6 months of age, a trained nurse applied the TST (Tuberculin PPD RT23; Statens Serum Institut) in a dose of 0.1 ml ventrally on the child s forearm. At the same visit, a trained field worker assessed the child s BCG scar by measuring the perpendicular Ethics The two trials were approved by the ethical committee of the Ministry of Health in Guinea-Bissau and the Danish Central Ethical Committee, and the LBW study was additionally approved by the Gambia/MRC Scientific and ethics committees [8, 9] John Wiley & Sons Ltd

3 Statistical methods Logistic regression analyses were performed to identify determinants for having a positive TST response and BCG scar. The two cohorts were combined. Information about cohort (NBW/LBW), birthweight (continuous), sex, place of inclusion, age at BCG vaccination (continuous), season at BCG vaccination, if the child received OPV at birth, if the child was randomised to receive vitamin A, most recent vaccine received (BCG/DTP), maternal ethnic group, maternal parity, maternal education, cohabitation with the father, place of residence, electricity in the house, type of roof on the house, exposure to TB before BCG vaccination, timing of the 2/6 months visit and field worker assessing the TST response (TST response analyses only) were included in logistic regression models. Timing of the visit and maternal parity were tested using a trend test. Potential determinants, for which information was missing for more than 1% of the children, were analysed in separate models in which the other potential determinants of TST responses/bcg scars were included. Sex-differential effects were tested one at a time by including two-way interaction terms between sex and each of the factors: cohort (NBW/LBW), age at BCG vaccination, oral polio vaccine (OPV) at birth, randomisation to vitamin A and most recent vaccine received. By including all factors in joint models, only direct associations between each factor and TST response/bcg scar can be seen. To see the full effect without adjusting for potential confounders and mediators, we therefore also performed the analyses without adjusting for any of the other factors. TST response and BCG scar as predictors of mortality were analysed using Cox regression models, and the proportional hazards assumption was tested using Schoenfeld residuals. The date of entry was the first day a valid TST application/bcg scar assessment had been made; the underlying time in the model was time since the TST application/bcg scar assessment. The children were followed until death or censoring, or until 12 months of age, whichever came first. Combined effects of TST response and BCG scar were also analysed by categorising in three groups: both TST response and BCG scar, TST response or BCG scar and none of the two. Only children with a valid assessment of both the TST response and the BCG scar were included in these analyses. The Cox regression analyses were adjusted for all factors significantly associated with either having a positive TST response or BCG scar at 2 months of age, as these factors are all assumed to be either directly or indirectly associated with survival. They were, however, only included if information was available for at least 99% of the children. Age at the visit was measured in days. Factors that were significantly associated with either having a positive TST response or BCG scar at 2 months of age but were missing for >1% of the children were subsequently tentatively included in the models. To show potential differences between LBW and NBW children, an interaction term between cohort and TST response or BCG scar was included and results are shown for each cohort. Furthermore, interactions between sex and TST response or BCG scar were tested. Infants who had received BCG vaccination less than 30 days before the visit and infants who were suspect of TB were excluded from all analyses. Furthermore, infants were excluded from the TST analyses if they had a history of chloroquine in the last 7 days before the visit, because chloroquine has been shown to suppress TST reactions [15]. Children were also excluded from the TST analyses if the response had not been assessed between 2 and 3 days after the application of the TST (manufacturer s guidelines). All analyses were adjusted for possible dependence within twin pairs using clustered (twin pairs) robust standard errors. The analyses were conducted using Stata 13. Results Figure 1 shows the number of children with valid assessments of TST responses and BCG scars at 2 and 6 months of age. At 2 months of age, 3007 children had a valid TST response or BCG scar assessment. At 6 months of age, 2606 children had a valid TST response or BCG scar assessment. Among those with a valid assessment, a third of the NBW children had a positive TST response, and a fifth and a fourth of the LBW children had a TST response at 2 and 6 months of age, respectively. Around 90% of both NBW and LBW children had a BCG scar (Table 1). Determinants for TST responses and BCG scars Information on most background factors was available for >99% of the children. However, among the children with a valid TST response or BCG scar assessment at 2 months, information about birthweight was missing for 345 children (11%), information about exposure to TB was missing for 52 children (2%), and information about maternal education was missing for 185 children (6%). 2 months of age. Inclusion at a health centre rather than at the national hospital and receiving OPV at birth was associated with reduced odds of having a TST response 2015 John Wiley & Sons Ltd 1735

4 Eligible NBW: 2709 LBW: months: 2575 Valid TST response assessment: 1663 TST not applied: 267 TST applied but not read: 51 Invalid TST response assessment: 594 Valid BCG scar assessment: 2068 No BCG scar assessment: 295 Invalid BCG scar assessment: 212 Valid TST response and BCG scar assessment: 1651 Valid TST response or BCG scar assessment: months: 1082 Valid TST response assessment: 807 TST not applied: 143 TST applied but not read: 48 Invalid TST response assessment: 84 Valid BCG scar assessment: 926 No BCG scar assessment: 142 Invalid BCG scar assessment: 14 Valid TST response and BCG scar assessment: 806 Valid TST response or BCG scar assessment: 927 Died: 26 Not seen at 2 Died: 31 months of age 24 of these had a valid TST response or BCG scar assessment at 2 months. NBW: 134 LBW: of these had a valid TST response or BCG scar assessment at 2 months. Lost to follow-up: 20 Lost to follow-up: 42 6 months: months: 1043 Valid TST response assessment: 1350 TST not applied: 476 TST applied but not read: 60 Invalid TST response assessment: 625 Valid TST response assessment: 621 TST not applied: 230 TST applied but not read: 48 Invalid TST response assessment: 144 Not seen at 6 months of age: 125 Valid BCG scar assessment: 1808 No BCG scar assessment: 508 Invalid BCG scar assessment: 195 Valid BCG scar assessment: 775 No BCG scar assessment: 224 Invalid BCG scar assessment: 44 Not seen at 6 months of age: 13 Valid TST response and BCG scar assessment: 1327 Valid TST response and BCG scar assessment: 621 Valid TST response or BCG scar assessment: 1831 Valid TST response or BCG scar assessment: 775 Died: 6 3 of these had a valid TST response or BCG scar assessment at 2 months. Lost to follow-up: 12 Died: of these had a valid TST response or BCG scar assessment at 2 months. 32 of these had a valid TST response or BCG scar assessment at 6 months. Lost to follow-up: 137 Died: of these had a valid TST response or BCG scar assessment at 2 months. 16 of these had a valid TST response or BCG scar assessment at 6 months. Lost to follow-up: 181 Died: 7 1 of these had a valid TST response or BCG scar assessment at 2 months. Lost to follow-up: 1 Followed until 12 months of age: 2434 Followed until 12 months of age: 843 Figure 1 Flowchart of TST response and BCG scar assessments in the two cohorts John Wiley & Sons Ltd

5 Table 1 TST responses and BCG scars among those with valid assessments at 2 and 6 months of age, by cohort 2 months 6 months TST response/valid TST response assessment NBW 36% (598/1663) 35% (477/1350) LBW 17% (134/807) 28% (173/621) BCG scar/valid BCG scar assessment NBW 88% (1811/2068) 93% (1689/1808) LBW 91% (845/926) 93% (718/775) and BCG scar at 2 months of age, and a late 2-month visit was associated with increased odds of having a TST response and BCG scar. Being female, being LBW, older age at BCG vaccination and cohabitation with the father were associated with reduced odds of having a TST response. High maternal parity and higher birthweight were associated with increased odds of having a TST response. Furthermore, the TST response was dependent on the field worker assessing the response, and the odds of having a BCG scar were dependent on the place of residence (Table 2). In addition, a significant interaction was found between sex and cohort (P < 0.01); LBW girls had 2.12 times higher odds (95% CI: 1.13; 3.99) of having a BCG scar compared to NBW girls, whereas no significant difference was seen between NBW boys and LBW boys, and LBW girls had 93% higher odds (95% CI: 1.19; 3.11) of having a BCG scar compared to LBW boys, whereas no difference was seen between NBW girls and boys. In the unadjusted analyses, the odds of having a TST response also depended on not being randomised to vitamin A or placebo at birth, most recent vaccine received, maternal ethnic group, place of residence, electricity in the house and maternal education, and the odds of having a BCG scar also depended on age at BCG vaccination, electricity in the house and birthweight (Table 2), but these associations were explained by confounding or mediation by other factors. 6 months of age. Only a few associations were significant at 6 month of age (data not shown). The TST response was still dependent on the field worker assessing the response, and inclusion at a health centre was associated with 26% (95% CI: 4; 44) reduced odds of having a TST response. Furthermore, children visited timely or late had 40% (95% CI: 21; 247) and 85% (95% CI: 0; 341), respectively, higher odds of having a BCG scar than those visited early. In the unadjusted analyses, being in the LBW cohort, older age a BCG vaccination, not receiving OPV at birth, not being randomised to vitamin A or placebo at birth, field worker assessing the response and having low birthweight were associated with decreased odds of having a TST response. OPV at birth and place of residence were associated with decreased odds of having a BCG scar (results not shown). Mortality analyses Among the 3007 children with at a valid TST response or BCG scar assessment at 2 months of age, 100 died before 12 months of age, 58 of 2080 (2.8%) NBW children and 42 of 927 (4.5%) LBW children. The unadjusted mortality rates for NBW and LBW children with and without a TST response and BCG scar are shown in Figure 2. No significant interactions were found between cohort and TST response and BCG scar, but the associations between TST responses/bcg scars and survival were stronger in the LBW cohort than in the NBW cohort. In the LBW cohort, children with a BCG scar at 2 months of age had significantly lower mortality than children with no BCG scar, the MRR being 0.42 (0.19; 0.93) when adjusting for sex, place of inclusion, age at BCG vaccination, OPV at birth, maternal parity, cohabitation with father, place of residence, age at the visit and field worker assessing the TST response. There was a similar tendency for children having a positive TST response, the MRR being 0.47 (0.14; 1.54). For LBW children who had both a positive TST reaction and BCG scar, the MRR was 0.22 (0.05; 0.87). The same tendencies were seen in the NBW cohort and when combing the two cohorts, although these associations were not significant (Table 3). Information about birthweight was missing for 3% of LBW children and 15% of NBW children. Including birthweight in the Cox models had little impact on the associations among the LBW children, whereas it diluted the associations among the NBW children (results not shown). No significant interactions with sex were found. Among the 2606 children with a valid TST response or BCG scar assessment at 6 months of age, 48 (32 NBW and 16 LBW) died before 12 months of age, and significant associations were not found between TST responses and BCG scars at 6 months of age and subsequent mortality (results not shown) John Wiley & Sons Ltd 1737

6 Table 2 Odds ratios (ORs) for having a TST response/bcg scar at 2 months of age TST response BCG scar Crude Adjusted Crude Adjusted % TST (n TST/n) OR (95% CI) P % TST (n TST/n) OR (95% CI) P % scar (n scar/n) OR (95% CI) P % scar (n scar/n) Cohort NBW 36.0 (598/1663) LBW 16.6 (134/807) Sex Boys 34.2 (418/1224) Place of inclusion Girls 25.2 (314/1246) National hospital 30.1 (472/1569) Health centres 28.9 (260/901) Age at BCG vaccination (weeks) 29.6 (732/2470) Season at BCG vaccination Dry (Dec Jun) 29.2 (333/1142) Rainy (July November) 30.1 (399/1328) OPV at birth No 38.4 (140/365) Randomisation to vitamin A Most recent vaccine received Yes 28.1 (592/2105) Placebo 32.8 (378/1152) Vitamin A 29.5 (329/1115) Not randomised 12.3 (25/203) BCG 25.1 (121/482) DTP 30.7 (611/1988) 0.35 (0.29; 0.44) 0.65 (0.55; 0.77) 0.94 (0.79; 1.13) 0.81 (0.72; 0.91) 1.04 (0.88; 1.24) 0.63 (0.50; 0.79) 0.86 (0.72; 1.03) 0.29 (0.19; 0.45) 1.32 (1.05; 1.66) 36.0 (592/1645) < (131/800) 34.1 (413/1212) < (310/1233) 30.0 (466/1555) (257/890) < (723/2445) 29.2 (330/1130) (393/1315) 37.9 (137/362) < (586/2083) 32.9 (375/1141) (324/1102) < (24/202) 24.7 (117/473) (606/1972) 0.35 (0.17; 0.75) 0.69 (0.57; 0.83) 0.71 (0.56; 0.91) 0.86 (0.76; 0.98) 0.91 (0.75; 1.11) 0.74 (0.57; 0.96) 0.87 (0.72; 1.05) 0.66 (0.40; 1.09) 0.91 (0.71; 1.18) 87.6 (1811/2068) (845/926) 88.6 (1318/1488) < (1338/1506) 91.7 (1728/1884) (928/110) (2656/2994) 89.7 (1232/1374) (1424/1620) 92.2 (425/461) (2231/2533) 88.7 (1237/1394) (1219/1367) (200/233) 87.0 (496/570) (2160/2424) Interaction with sex Interaction with cohort 0.46 (0.37; 0.58) 0.82 (0.74; 0.91) 0.84 (0.66; 1.06) 0.63 (0.44; 0.90) 1.05 (0.82; 1.33) 0.77 (0.50; 1.17) 1.22 (0.92; 1.62) 87.5 (1789/2045) 91.3 (838/918) 88.5 (1304/1474) 88.9 (1323/1489) 91.8 (1714/1868) < (913/1095) < (2627/2963) 89.7 (1221/1362) (1406/1601) 92.1 (420/456) (2207/2507) 88.8 (1226/1381) (1202/1350) (199/232) 86.9 (485/558) (2142/2405) OR (95% CI) P Interaction with sex Interaction with cohort 0.57 (0.42; 0.77) 0.90 (0.80; 1.02) 0.83 (0.65; 1.06) 0.56 (0.38; 0.82) 1.06 (0.83; 1.35) 0.72 (0.44; 1.20) 1.16 (0.85; 1.57) < < John Wiley & Sons Ltd

7 Table 2 (Continued) TST response BCG scar Crude Adjusted Crude Adjusted % TST (n TST/n) OR (95% CI) P % TST (n TST/n) OR (95% CI) P % scar (n scar/n) OR (95% CI) P % scar (n scar/n) Maternal ethnic group Maternal parity Cohabitation with father Place of residence Electricity in the house Type of roof on the house Fula 23.4 (104/444) Balanta 25.4 (73/288) Mancanha 34.5 (70/203) Mandinga 29.7 (58/195) Manjaco 29.6 (91/308) Pepel 32.0 (209/653) Other 33.4 (124/371) (210/789) (176/571) (344/1105) No 31.6 (306/968) Yes 28.4 (422/1487) Bandim 33.4 (429/1284) Belem 33.5 (114/340) Mindera 38.8 (73/188) Cuntum 17.5 (18/103) Other 17.5 (97/554) No 27.9 (464/1664) Yes 33.0 (263/796) Corrugated iron/tiles/ cement 29.5 (686/2323) Thatched 29.9 (41/137) 1.11 (0.78; 1.57) 1.72 (1.19; 2.48) 1.38 (0.95; 2.03) 1.37 (0.98; 1.91) 1.54 (1.17; 2.03) 1.64 (1.20; 2.24) 1.23 (0.97; 1.56) 1.25 (1.02; 1.53) 0.86 (0.72; (0.78; 1.30) 1.27 (0.92; 1.74) 0.42 (0.25; 0.71) 0.42 (0.33; 0.54) 1.28 (1.06; 1.53) 1.02 (0.70; 1.49) 23.2 (102/439) (73/286) 34.5 (70/203) 29.7 (58/195) 29.7 (91/306) 31.7 (206/649) 33.5 (123/367) 26.5 (207/782) 0.04* 30.9 (175/567) 31.1 (341/1096) 31.5 (304/964) (419/1481) 33.6 (426/1269) < (113/338) 38.5 (72/187) 17.5 (18/103) 17.2 (94/548) 27.9 (461/1651) (262/794) 29.5 (682/2310) (41/135) 1.16 (0.79; 1.68) 1.47 (1.00; 2.18) 1.51 (1.00; 2.29) 1.06 (0.74; 1.52) 1.20 (0.88; 1.62) 1.47 (1.05; 2.06) 1.24 (0.95; 1.61) 1.32 (1.03; 1.69) 0.79 (0.64; 0.98) 0.84 (0.63; 1.12) 1.08 (0.76; 1.53) 1.27 (0.63; 2.57) 1.47 (0.83; 2.61) 1.17 (0.96; 1.44) 1.09 (0.73; 1.63) 88.1 (458/520) (330/361) 87.9 (225/256) 90.5 (200/221) 88.8 (325/366) 87.8 (716/816) 88.5 (391/442) 88.4 (842/953) 0.03* 87.5 (597/682) 89.5 (1208/1350) 87.9 (1034/1177) (1602/1797) 86.5 (1362/1574) (387/426) 89.7 (210/234) 82.9 (102/123) 93.4 (594/636) 87.9 (1752/1994) (890/985) 88.8 (2501/2817) (141/163) 1.44 (0.91; 2.29) 0.98 (0.62; 1.56) 1.29 (0.75; 2.21) 1.07 (0.70; 1.64) 0.97 (0.69; 1.36) 1.04 (0.70; 1.55) 0.93 (0.68; 1.26) 1.12 (0.86; 1.46) 1.14 (0.90; 1.43) 1.54 (1 08; 2.21) 1.36 (0.87; 2.13) 0.76 (0.44; 1.29) 2.20 (1.54; 3.15) 1.29 (1.00; 1.67) 0.81 (0.51; 1.29) 88.0 (453/515) (328/359) 87.9 (225/256) 90.5 (199/220) 89.0 (323/363) 87.7 (712/812) 88.4 (387/438) 88.4 (837/947) 0.35* 87.4 (592/677) 89.5 (1198/1339) 87.9 (1030/1172) (1597/1791) 86.4 (1344/1556) < (384/423) 90.1 (209/232) 82.9 (102/123) 93.5 (588/629) 87.8 (1741/1982) (886/981) 88.8 (2488/2802) (139/161) OR (95% CI) P 1.36 (0.84; 2.19) 0.93 (0.57; 1.52) 1.08 (0.63; 1.87) 1.19 (0.76; 1.85) 1.18 (0.82; 1.71) 0.99 (0.65; 1.51) 1.06 (0.76; 1.47) 1.33 (0.98; 1.81) 1.06 (0.81; 1.38) 1.25 (0.85; 1.83) 1.08 (0.67; 1.74) 0.53 (0.26; 1.08) 1.24 (0.68; 2.26) 1.07 (0.82; 1.40) 0.97 (0.59; 1.57) * John Wiley & Sons Ltd 1739

8 Table 2 (Continued) TST response BCG scar Crude Adjusted Crude Adjusted % TST (n TST/n) OR (95% CI) P % TST (n TST/n) OR (95% CI) P % scar (n scar/n) OR (95% CI) P % scar (n scar/n) OR (95% CI) P Timing of the visit Early (ref.) 30.9 (59/191) Timely (2 months +/ 7 days) Field worker assessing TST (6 different) 28.5 (445/1559) Late 31.7 (228/720) Birthweight (kg) 30.2 (663/2195) Exposure to TB Maternal education No 30.2 (716/2369) Yes 20.3 (12/59) No 25.9 (177/683) Yes 31.4 (515/1642) 0.89 (0.64; 1.24) 1.04 (0.73; 1.47) 2.22 (1.91; 2.58) 0.59 (0.31; 1.11) 1.31 (1.07; 1.60) 30.5 (58/190) 0.35* 28.6 (441/1542) 31.4 (224/713) 1.33 (0.94; 1.89) 1.62 (1.11; 2.36) 79.0 (192/243) <0.01* 88.5 (1672/1890) 92.0 (792/861) 2.04 (1.45; 2.86) 3.05 (2.05; 4.54) 78.8 (190/241) <0.01* 88.4 (1656/1873) 92.0 (781/849) 1.77 (1.23; 2.54) 2.53 (1.66; 3.85) <0.01 <0.01 < (655/2174) 30.2 (707/2344) (12/59) 25.6 (171/667) < (513/1635) 1.38 (1.07; 1.78) 0.60 (0.31; 1.19) 1.04 (0.81; 1.33) / (2552/2871) (64/71) 88.2 (720/816) (1788/1994) 0.82 (0.67; 1.00) 1.14 (0.52; 2.52) 1.16 (0.89; 1.50) (2352/2625) 88.8 (2523/2840) (64/71) 88.2 (704/798) (1780/1986) 1.13 (0.79; 1.63) 1.29 (0.55; 3.03) 1.22 (0.90; 1.67) <0.01* *P-value for trend test. ORs are adjusted for all factors in the table except birthweight, exposure to TB, and maternal education due to a high number of missing values on these three variables. OR indicates the change in odds with each increase in 1 unit of the continuous variable John Wiley & Sons Ltd

9 Kaplan-Meier failure estimates NBW children Accumulated mortality (%) Accumulated mortality (%) Time since 2 months visit (months) LBW children Time since 2 months visit (months) Figure 2 Cumulated mortality among NBW and LBW children with and without a TST response and BCG scar at 2 months of age. Children with neither TST response nor BCG scar Children with either TST response or BCG scar Children with both TST response and BCG scar Discussion We confirmed previous studies; having a BCG scar or a positive TST reaction at 2 months of age was associated with better survival during infancy. Observational studies on the association between vaccination status and mortality are often criticised because the association might be explained by socio-economic differences between vaccinated and unvaccinated children. Our study eliminates this source of confounding by only looking at the association between BCG scar, TST responses and mortality among BCG-vaccinated children. To be able to adjust for potential determinants of reaction, we tested the impact of many background factors on TST responses and BCG scarring and included them in the mortality analysis. LBW children had fewer TST responses than NBW children, and increasing birthweight was associated with the TST response. However, controlling for birthweight, there was still a significant effect on survival of having a BCG scar among LBW children. HIV status was not taken into account in the present study. This is, however, unlikely to have affected the results as the HIV prevalence among infants in Guinea- Bissau is expected to be less than 1% [8]. Furthermore, a previous study from Guinea-Bissau of TST responses and BCG scars found that exclusion of dead children with own or maternal HIV antibodies did not modify the estimates [5]. In our analyses, the strongest findings regarding survival were seen in the LBW cohort, which has the highest mortality rate, but most analyses showed tendencies for TST response and BCG scar being positively associated with better survival. Hence, results are consistent with previous findings from Guinea-Bissau, showing that a 2015 John Wiley & Sons Ltd 1741

10 Table 3 Adjusted mortality rate ratios (MRR) for children with/without TST responses and BCG scars Combined NBW and LBW cohorts NBW cohort* LBW cohort* Adjusted MRR (95% CI) Rates (deaths/person years) [n ] P-value for cohort interaction Adjusted MRR (95% CI) Rates (deaths/person years) [n ] Adjusted MRR (95% CI) Rates (deaths/person years) [n ] 2 months No TST response 3.7 (31/833.9) [1056] 6.8 (35/514.8) [671] 4.89 (66/1348) [1727] TST response 3.0 (14/470.2) [595] 0.75 (0.40; 1.42) 3.0 (3/100.7) [132] 0.47 (0.14; 1.54) (17/571) [727] 0.67 (0.39; 1.14) No BCG scar 4.4 (9/206.2) [257] 13.2 (8/60.6) [81] 6.37 (17/267) [338] BCG scar 3.5 (49/1412.9) [1795] 0.80 (0.39; 1.68) 5.3 (34/645.6) [840] 0.42 (0.19; 0.93) (83/2059) [2635] 0.62 (0.36; 1.07) 4.0 (6/151.3) [187] 14.6 (7/47.8) [64] 6.53 (13/199) [251] Neither BCG scar nor TST response 3.6 (25/694.2) [883] 0.98 (0.40; 2.42) 6.0 (28/467.8) [608] 0.40 (0.17; 0.94) 4.56 (53/1162) [1491] 0.66 (0.35; 1.26) BCG scar or TST response 3.1 (14/449.5) [570] 0.79 (0.30; 2.10) 3.0 (3/99.1) [130] 0.22 (0.05; 0.87) (17/549) [700] 0.50 (0.24; 1.05) BCG scar and TST response *The two cohorts were combined, and an interaction term between cohort and TST response/bcg scar was included in the models. The figures do not correspond exactly to the figures in the flow chart as a few children had missing information on one or more variables. The results are adjusted for sex, place of inclusion, age at BCG vaccination, OPV at birth, maternal parity, cohabitation with father, place of residence, age at the visit and field worker assessing the TST response (TST response and combined TST response and scar analyses only). TST response [5, 7] and a BCG scar [5 7] were associated with lower mortality. The infant mortality rates in Guinea-Bissau have decreased as these previous three studies were conducted in [5], [6] and [7], respectively. The declining mortality rate among NBW children may be an important reason that we found only a limited effect among NBW children. Conclusions Two randomised trials among LBW children have shown that BCG at birth compared with delayed BCG reduces neonatal mortality [9, 16]. As this is a non-tb effect, BCG probably has a beneficial immune training effect. Immunological studies in humans and animals have shown that this training effect may be linked to an epigenetic reprogramming of innate immunity providing stronger anti-inflammatory responses [17, 18]. Although this has not been tested, it seems likely that a positive TST response and a BCG scar are associated with the epigenetic reprogramming of monocytes. All children in the present study were BCG vaccinated, and a positive TST response and BCG scar may therefore reflect that they had better immune training. It is important to consider whether those who did not react to TST or did not have a scar were somehow immunologically deficient. We found TST responses and BCG scars to be associated with reduced mortality rates, and it could be argued that this is due to children with positive reactions having stronger immune systems prior to vaccination. However, a study from Guinea-Bissau found that improved survival was only seen with TST and not with a positive delayed hypersensitivity response to diphtheria and tetanus, which makes it unlikely that the association between TST response and survival is merely a reflection of some children having a stronger immune system [5]. Also, a recent study in the rural areas of Guinea-Bissau has shown a strong correlation between having a BCG scar and reduced mortality in a population where only 52% developed a scar, which was probably due to the use of another vaccine strain [19]. When BCG scars are associated with reduced mortality irrespective of the proportion of children who develop a scar, it strongly supports the notion that this association is not merely a reflection of underlying biological/genetic differences between those who develop and those who do not develop a scar. It has previously been shown that the quality of the BCG vaccination is a major determinant of the subsequent reactions [20]. The same can be seen in the present study; the TST response and BCG scarring were of John Wiley & Sons Ltd

11 poorer quality among children vaccinated at health centres than among children enrolled and BCG vaccinated at the maternity ward of the national hospital (Table 2). One nurse and one midwife gave all BCG vaccinations at the maternity ward, so they have had far more supervision in applying BCG vaccination correctly [20]. Surprisingly, BCG scarring was better among LBW girls than among NBW girls and no difference was found among the boys; if the absence of BCG scarring reflects some degree of immune deficiency, there should have been less scarring among the LBW children. However, the LBW children were predominantly enrolled at the maternity ward (90%) and they were included more recently than the NBW children. The maternity ward staff therefore had had better training in setting BCG vaccinations when the LBW children were included. Hence, it seems likely that a large part of the non-response is due to poor BCG vaccination technique rather than an underlying poor immune system. There are two implications. First, BCG scarring, which is cheap and easy to assess, can be used to monitor the quality of BCG vaccinations; when the odds of developing a scar when vaccinated at the health centres is only around half of the odds of developing a scar when vaccinated at the hospital, there is clearly room for improving the technique. Second, given the strong survival benefit from having a BCG scar, it should be worthwhile to examine whether an effective BCG revaccination of children with no BCG scar can subsequently provide them with the same benefit as if they had had an effective vaccination at birth. In conclusion, we confirmed that children who do not develop TST responses and BCG scars have higher mortality rates than children with TST responses and BCG scars. There may be a good reason to use BCG reactions to assess the quality of vaccinations at different institutions and possibly also reason to BCG-revaccinate children who have not responded with a scar. References 1. Santiago EM, Lawson E, Gillenwater K et al. A prospective study of bacillus Calmette-Guerin scar formation and tuberculin skin test reactivity in infants in Lima, Peru. Pediatrics 2003: 112: e Wang L, Turner MO, Elwood RK, Schulzer M, Fitzgerald JM. A meta-analysis of the effect of Bacille Calmette Guerin vaccination on tuberculin skin test measurements. Thorax 2002: 57: Smith KC, Starke JR, Orme IM, Kaper J. Tuberculosis vaccines. In: Plotkin S, Orenstein W, Offit P (eds). Vaccines (5th edn). Saunders: Philadelphia, Aaby P, Kollmann TR, Benn CS. Nonspecific effects of neonatal and infant vaccination: public-health, immunological and conceptual challenges. Nat Immunol 2014: 15: 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: Roth A, Gustafson P, Nhaga A et al. BCG vaccination scar associated with better childhood survival in Guinea-Bissau. Int J Epidemiol 2005a: 34: Roth A, Sodemann M, Jensen H et al. Tuberculin reaction, BCG scar, and lower female mortality. Epidemiology 2006: 17: Benn CS, Diness BR, Roth A et al. Effect of 50,000 IU vitamin A given with BCG vaccine on mortality in infants in Guinea-Bissau: randomised placebo controlled trial. BMJ 2008: 336: 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: Benn CS, Fisker AB, Napirna BM et al. Vitamin A supplementation and BCG vaccination at birth in low birthweight neonates: two by two factorial randomised controlled trial. BMJ 2010: 340: c Diness BR, Fisker AB, Roth A et al. Effect of high-dose vitamin A supplementation on the immune response to Bacille Calmette-Guerin vaccine. Am J Clin Nutr 2007: 86: WHO. Guidance for National Tuberculosis Programmes on the Management of Tuberculosis in Children. World Health Organization: Geneva, Switzerland, 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. Int J Epidemiol 2004: 33: Kristensen I, Aaby P, Jensen H. Routine vaccinations and child survival: follow up study in Guinea-Bissau, West Africa. BMJ 2000: 321: Garly ML, Bale C, Martins CL et al. BCG vaccination among West African infants is associated with less energy to tuberculin and diphtheria-tetanus antigens. Vaccine 2001: 20: 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: 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 USA 2012: 109: Netea MG, Quintin J, van der Meer JW. Trained immunity: a memory for innate host defense. Cell Host Microbe 2011: 9: John Wiley & Sons Ltd 1743

12 19. Storgaard L, Rodrigues A, Martins C et al. Development of BCG scar and subsequent morbidity and mortality in rural Guinea-Bissau. Clin Infect Dis 2015: 61: 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 2005b: 23: Corresponding Author Clara Amalie Gade Timmermann, Department of Environmental Medicine, Institute of Public Health, University of Southern Denmark, J.B. Winsløws Vej 17, 2., DK-5000 Odense C, Denmark. atimmermann@health.sdu.dk John Wiley & Sons Ltd