Early BCG Vaccination, Hospitalizations, and Hospital Deaths: Analysis of a Secondary Outcome in 3 Randomized Trials from Guinea-Bissau

Abstract

Background

This study was performed to examine the effects of early BCG vaccination on the risk, cause, and severity of infant hospitalizations. The

analysis included 3 trials randomizing low-weight neonates to early BCG vaccination (intervention) versus no BCG vaccination (usual practice in low-weight neonates, control), with hospitalizations as secondary outcome.

Methods

Hospitalization data were collected at the pediatric ward of the National Hospital. Effects of BCG vaccination on hospitalization risk were assessed in Cox models providing overall and major disease-group incidence rate ratios (IRRs). Severity was assessed by means of in-hospital case-fatality rates and compared by group as cohort study risk ratios (RRs).

Results

Among 6583 infants (3297 in BCG group, 3286 controls), there were 908 infant hospitalizations (450 BCG, 458 controls) and 135 in-hospital deaths (56 BCG, 79 controls). The neonatal (28 days), 6-week, and infant (1-year) BCG versus control hospitalization IRRs were 0.97 (95% confidence interval [CI], .72–1.31), 0.95 (.73–1.24), and 0.96 (.84–1.10). Corresponding BCG versus control case-fatality rate RRs were 0.58 (95% CI, .35–.94), 0.56 (.35–.90), and 0.72 (.53–.99). BCG vaccination tended to reduce neonatal and infant sepsis hospitalization rates (IRR, 0.75 [95% CI, .50–1.13] and 0.78 [.55–1.11], respectively), and it reduced the neonatal in-hospital sepsis mortality rate (RR, 0.46; 95% CI, .22–.98). There were no confirmed hospitalizations for tuberculosis.

Conclusions

BCG vaccination did not affect hospitalization rates but reduced in-hospital mortality rates significantly, primarily by preventing fatal cases of sepsis. The observed beneficial effects of BCG on the in-hospital mortality rate were entirely nonspecific.

Clinical Trials Registration

NCT00146302, NCT00168610, and NCT00625482.

BCG vaccine, infant mortality, infant morbidity, in-hospital case-fatality, neonatal mortality, neonatal morbidity, non-specific (heterologous) effects of vaccines, sepsis, tuberculosis

Observational studies and randomized controlled trials (RCTs) from low-income countries indicate that BCG vaccine against tuberculosis provides notable protection against nontuberculosis mortality in the first months of life; in other words, the vaccine has beneficial nonspecific effects [1]. BCG vaccine may alter the susceptibility to other pathogens by inducing heterologous immunity and/or epigenetic reprogramming of innate immune cells, so-called trained immunity [2, 3]. The World Health Organization recommends BCG vaccination as soon as possible after birth as a key component in its “End TB” strategy [4], but it is often delayed because infants are low weight (<2500 g) or for logistic reasons, including vaccine wastage policies restricting the opening of BCG vials to situations with many eligible children present [5, 6].

We tested the overall effect on mortality of providing BCG vaccination shortly after birth (BCG intervention group) against the usual policy of providing delayed BCG vaccination (control group) to low-weight infants in 1 small RCT [7] and 2 large RCTs [8, 9] conducted in Guinea-Bissau from 2002 to 2013. All trials showed a beneficial effect of BCG vaccine with effects most pronounced in the neonatal period (first 28 days of life), when only few infants in the control group had yet received BCG vaccine and other infant vaccines had not been administered. In the combined 2002–2013 estimate, the mortality rate ratio for the BCG versus the control group was 0.62 (95% CI, .46–.83) in the neonatal period and 0.84 (.71–1.00) in the infant period (first year of life) (Table 1) [9].

Table 1.

Overview of Neonatal and Infant Mortality Rates, Hospitalization Incidence and in-Hospital Mortality Rates Across the 3 RCTs of BCG Versus No BCG, by Trial and Combined

Trial	Neonatal Period (Up to 28 d of life)a			Infant Period (Up to 1 y after Birth)b
	Mortality MRRc (95% CI) [Deaths/ Person-Years]	Hospitalization IRRd (95% CI) [Admissions/ Person-Years]	Case-Fatality Rate RRe (95% CI) [Deaths/ Person-Years]	Mortality MRR (95% CI) [Deaths/ Person-Years]	Hospitalization IRR (95% CI) [Admissions/ Person-Years]	Case-Fatality Rate RR (95% CI) [Deaths/ Person-Years]
Trial I	0.28 (.06–1.37)[8/6]	0.84 (.21–3.40) [8/6]	0.33 (.06–1.99)[4/6]	0.41 (.14–1.18)[16/87]	1.20 (.41–3.47) [14/87]	0.50 (.12–2.12)[5/87]
Trial II	0.55f (.34–.89)[75/150]	0.89 (.56–1.41)[72/149]	0.56 (.26–1.22)[21/149]	0.83 (.63–1.08)[229/2038]	1.05 (.84–1.30)[321/2031]	0.73 (.45–1.17)[57/2031]
Trial III	0.70 (.47–1.04) [106/272]	1.04 (.68–1.57)[90/266]	0.64 (.32–1.27)[25/266]	0.88 (.70–1.10) [322/3776]	0.91 (.77–1.07) [573/3770]	0.71 (.46–1.10)[73/3770]
Combined	0.62f (.46–.83)[189/428]	0.97 (.72–1.31)[170/421]	0.58g (.35–.94) [50/421]	0.84f (.71–1.00) [567/5901]	0.96 (.84–1.10) [908/5888]	0.72g (.53–.99) [135/5888]

Abbreviations: CI, confidence interval; IRR, incidence rate ratio; MRR, mortality rate ratio; RCTs, randomized controlled trials; RR, cohort risk ratio.

^aWithin 170 neonatal admissions, 166 infants (80 in the BCG group, 86 controls) were admitted once and 2 (2 BCG, 0 controls) were admitted twice.

^bWithin 908 infant admissions, 735 infants (367 BCG, 368 controls) were admitted once, 64 (29 BCG, 35 controls) were admitted twice, 11 (7 BCG, 4 controls) were admitted 3 times, and 3 (1 BCG, 2 controls) were admitted 4 times.

^cMRRs for main trial.

^dIRRs for admission at the pediatric ward of the Simão Mendes National Hospital

^eRRs for death while admitted at the Pediatric Ward of the Simão Mendes National Hospital

^fP < .05 (main trial Cox analysis).

^gP < .05 (2-sided Fisher exact test).

Based on the causes of death obtained by standard verbal autopsy (VA) in the 2 large RCTs, BCG vaccine provided protection against deaths from infectious diseases, particularly sepsis, whereas there was no protection against deaths from noninfectious causes, such as sudden infant death syndrome, prematurity, and respiratory distress syndrome [8, 9]. Little is known, however, as to whether BCG vaccine works by reducing susceptibility to infection, thus lowering overall incidence of disease, by reducing the severity of infection, or by a combination of both. In the present study, we examined the effect of BCG vaccination on hospitalizations and in-hospital case-fatality rates within the 3 randomized trials.

METHODS

Hospitalization was a secondary outcome in 1 small and 2 large RCTs with enrollment between 2002–2004 [7] (trial I), 2004–2008 [8] (trial II), and 2008–2013 [9] (trial III) (Figure 1). All trials were conducted by the Bandim Health Project (BHP; www.bandim.org) in Guinea-Bissau.

Figure 1.

Timeline of the 3 randomized trials of BCG vaccination versus no BCG vaccination conducted between 2002 and 2013.

Setting

The BHP maintains a Health Demographic Surveillance System (HDSS) in the capital Bissau. At the maternity ward of the National Hospital Simão Mendes (HNSM), a BHP team documents all deliveries and vaccinations. Furthermore, all vaccinations are monitored daily by BHP assistants at 3 health centers in the HDSS area. The vaccination schedule at the time of the trials in Guinea-Bissau was BCG plus oral polio vaccine (OPV) at birth, 3 doses of diphtheria-tetanus-pertussis (DTP) vaccine at age 6, 10, and 14 weeks, and measles vaccine (MV) at age 9 months; BCG vaccination was normally deferred for low-weight infants.

The Simão Mendes Hospital hosts a pediatric ward with uptake from the capital, which serves as the pediatric referral hospital for the country. Hospitalizations at HNSM and their outcomes have been documented by a BHP team since the 1990s [10]. Guinea-Bissau is among the poorest countries in the world and has one of the world’s highest infant mortality rates [11]. At the time the RCTs were conducted, national physicians had obtained medical training in different medical schools across Europe, Russia, and Cuba. The pediatric ward of HNSM has scarce resources, and diagnostic procedures are based mostly on clinical presentation, occasional blood sampling, and microscopy of blood slides for malaria parasites; blood cultures are not routinely performed. The in-hospital mortality for admitted infants was 25% in 2002, when enrollment into trial I began, and 10% in 2013 when enrollment into trial III ended (unpublished data).

Study Design

The RCT design has been described in detail elsewhere [7–9]. Briefly, the primary objective was to investigate the effect of early BCG vaccination on infant mortality among low-weight infants. Provided informed oral and written consent, healthy neonates ready for discharge from the maternity ward and healthy infants brought to a HDSS area health center for the first vaccinations were invited to participate in the study. The inclusion criterion was a weight <2500 g on the day of inclusion. Exclusion criteria were prior BCG vaccination or severe malformations. Most Infants were enrolled within the first days of life, but a few participants were brought only for their first vaccines after the neonatal period. Mothers or guardians were provided written and oral information during the informed consent procedure and the opportunity to ask questions. Informed oral and written consent was obtained in all cases. In total, 6583 infants (104 in trial I, 2320 in trial II, and 4159 in trial III) were enrolled and randomized 1:1 to receive BCG vaccine (BCG intervention group, 3297 infants) or to be treated according to standard policy (control group, 3286 infants); that is, they were informed that their child could be vaccinated with BCG when reaching a weight of 2500 g, typically when the 6-week childhood vaccines were due.

Randomization

The mother drew a sealed envelope containing the randomization allocation. In all trials, the intervention group received trivalent (OPV) and a single 0.05-mL dose of live attenuated BCG (Copenhagen strain 1331, Statens Serum Institut) by intradermal injection in the left deltoid region. Control group infants received only trivalent OPV. The trials were not blinded.

Follow-up for Deaths and Hospitalizations

Home Visit Follow-up

During the first year of life, data on anthropometrics, mortality, and morbidity were collected at 4 scheduled visits to the child’s home 3 days after inclusion and by 2, 6, and 12 months of age. For children known to have died, a standard VA [12] was performed at a home visit 3 months after death.

Hospital Registration and Follow-up

Severe pediatric cases are typically sent to the HNSM pediatric ward for treatment. There, BHP staff documents all hospitalizations and collects data on admission weight, temperature, unique BHP study and/or HDSS identification numbers (if applicable), admission and discharge dates, vaccinations, information on diagnosis, and the outcome of the hospitalization. At the ward, all infants are assigned a preliminary entry diagnosis in the triage room. A final diagnosis is assigned to the admission chart at discharge, by the physician responsible for the treatment, for most but not all infants By hospital routine, admission charts are mostly but not always delivered to the BHP staff after a hospitalization, ensuring that the discharge diagnosis is registered in most cases. If a final diagnosis was not available, the entry diagnosis was used for data analysis (see footnote in Table 2 for further details). A standardized data linkage protocol was applied to link study subjects with the pediatric ward hospitalization database (Supplementary Material). Only hospitalization data from the Simão Mendes Hospital Pediatric ward were included in the present study.

Table 2.

Neonatal and Infant Admission Incidence and Case-Fatality Rate by Disease

Diagnosis	Neonatal Period (Up to 28 d of Life)				Infant Period (Up to 1 y After Birth)
	BCG, Admissions (Deaths), No. (n = 3262)	Controls, Admissions (Deaths), No. (n = 3259)	Admission Risk for BCG vs Controls, IRR (95% CI)	Case-Fatality Rate Ratio for BCG vs Controls, RRa (95% CI)	BCG, Admissions (Deaths), No. (n = 3298)	Controls, Admissions (Deaths), No. (n = 3285)	Admission Risk for BCG vs Controls, IRR (95% CI)	Case-Fatality Rate Ratio for BCG vs Controls, RRa (95% CI)
Infectious disease
Bronchitis	1 (0)	0 (0)	…	…	5 (0)	15 (2)	0.33 (.12–.90)b	…
Fever	2 (0)	4 (1)	0.50 (.091–2.70)	…	4 (0)	7 (2)	0.42 (.11–1.63)	…
Gastroenteritis	4 (0)	1 (0)	3.96 (.44–35.4)	…	73 (3)	77 (5)	0.93 (.67–1.28)	0.63 (.16–2.55)
Malaria	12 (3)	8 (2)	1.49 (.61–3.63)	1.00 (.21–4.71)	130 (12)	130 (13)	0.98 (.77–1.25)	0.92 (.44–1.95)
Meningitis	0 (0)	0 (0)	…	…	4 (1)	0 (0)	…	…
Pneumonia	4 (2)	8 (2)	0.49 (.15–1.64)	2.00 (.42–9.42)	116 (18)	98 (16)	1.16 (.88–1.51)	0.95 (.51–1.76)
Sepsis	41 (7)	54 (20)	0.75 (.50–1.13)	0.46 (.22–.98)b	57 (10)	72 (23)	0.78 (.55–1.11)	0.55 (.28–1.06)
Tetanus	1 (0)	1 (1)	0.99 (.06–15.8)	…	2 (0)	2 (2)	0.99 (.14–7.00)	…
Tuberculosis	0 (0)	0 (0)	…	…	0 (0)	0 (0)	…	…
Any infectious diseases	65 (12)	76 (26)	0.85 (.61–1.18)	0.54 (.30–.98)b	391 (44)	401 (63)	0.95 (.83–1.10)	0.72 (.50–1.03)
Noninfectious condition
Anemia	10 (3)	4 (3)	2.47 (.78–7.89)	0.40 (.13–1.21)	26 (6)	22 (6)	1.16 (.66–2.05)	0.85 (.32–2.25)
Congenital heart disease	0 (0)	1 (0)	…	…	2 (2)	3 (1)	0.66 (.11–3.92)	3.00 (.61–14.9)
Hypoglycemia	2 (0)	0 (0)	…	…	5 (0)	3 (1)	0.64 (.39–6.87)	…
Malnutrition	0 (0)	0 (0)	…	…	4 (0)	6 (2)	0.65 (.18–2.31)	…
Otherc	3 (1)	3 (1)	0.99 (.20–4.91)	1.00 (.10–9.61)	15 (2)	18 (3)	0.87 (.44–1.71)	0.75 (.14–3.94)
Prematurity	4 (2)	2 (2)	1.98 (.36–10.8)	0.50 (.19–1.33)	5 (2)	3 (2)	1.65 (.40–6.92)	0.60 (.16–2.29)
Unknown	0 (0)	0 (0)	…	…	2 (0)	2 (1)	0.98 (.14–6.96)	…
Any noninfectious condition	19 (6)	10 (6)	1.88 (.88–4.05)	0.53 (.23–1.21)	59 (12)	57 (16)	1.02 (.71–1.46)	0.72 (.38–1.39)
Total admissionsd	84 (18)	86 (32)	0.97 (.72–1.31)	0.58 (.35–.94)b	450 (56)	458 (79)	0.96 (.84–1.10)	0.72 (.52–.99)b

Abbreviations: CI, confidence interval; IRR, incidence rate ratio; RR, cohort risk ratio.

^aRRs for death while admitted at the Pediatric Ward of the Simão Mendes National Hospital

^bP < .05 (2-sided Fisher exact test).

^c”Other” includes other somatic disease (1 infant), dermatologic condition (1 infant), intoxication (2 infants), and dehydration (4 infants).

^dInfants are assigned an entry diagnosis on entry to the ward. In addition, those with final discharge paperwork available (81%; 739 of 908) had a discharge diagnosis assigned. The diagnoses in this table represent discharge diagnoses when available. Note that some infants (36 in the BCG group, 26 controls) were enrolled at an age of >4 weeks; they are included in estimates to 1 year but not in the neonatal estimate.

Evaluation of VA Data in Study Infants

For infants registered at the routine home visits to have died within less than a week after discharge from the pediatric ward and with an unclear hospital exit status (hospital records misplaced or no final status recorded), the VA history was reviewed to determine whether the child had in fact died at the pediatric ward or not. For a total of 30 infants (15 in BCG group, 15 controls) the VA data were reviewed; 15 infants (10 BCG, 5 controls) were registered to have died at the hospital, 7 (2 BCG, 5 controls) had died at home, 1 (control) died en route from the hospital, 3 (2 BCG, 1 control) died when healthcare was being sought elsewhere, and 1 (control) who died at the pediatric ward according to the VA could not be matched with a hospitalization. In infants (1 BCG, 2 controls), no VA was performed because the family had moved.

Statistical Analyses

Cumulative hospitalization curves were computed using the Kaplan-Meier estimate based on the date of hospitalization. Incidence rate ratios (IRRs) of hospitalization events comparing randomization groups were estimated in multiple-event Cox proportional hazards models, with age as the underlying time variable; therefore, age was inherently controlled for in all analyses. Hospitalization estimates are reported as IRRs with 95% confidence intervals (CIs). In-hospital case-fatality rates are reported as cohort study risk ratios (RRs) with 95% CIs; P values were determined with the Fisher 2-sided exact test.

The primary outcome in trials I and II was the infant mortality rate (up to 365 days), and for trial III it was the neonatal mortality rate (up to 28 days) [7–9]. Many control group children received BCG vaccine at 6 weeks of age along with the first dose of DTP. In subsequent RCTs designed to examine the nonspecific effects of either early BCG or effects of different strains of BCG to newborns, the investigators have focused on evaluating effects of BCG up to either 4 or 6 weeks [13–15]—before other vaccines are administered and before a substantial part of the control group has received BCG vaccine—as well as up to 1 year. In the present study, results are therefore presented up to 28 days (neonatal), 42 days (6 weeks), and 365 days (infant) after birth.

Person-years at risk were calculated from enrollment (day of randomization). Children were censored in the hospitalization analysis if they moved or died. While admitted, children were not considered at risk of a new hospitalization and the time hospitalized thus did not contribute to the person-days at risk. Tests for proportionality of hazard rates were computed using Schoenfeld’s partial residuals and by visual inspection of Nelson-Aalen estimators.

Season of enrollment and sex were defined as potential effect modifiers and an analysis of BCG vaccine effects by sex and season of enrollment (rainy season, June to October; dry season, November to May) is thus presented. In the Supplementary Material, we include a case-fatality rate analysis by date of death rather than the date of hospitalization, to account for lengthier fatal hospitalizations crossing the neonatal, 6-week, or infant time periods. All analyses were made using Stata MP 12 software.

RESULTS

Baseline characteristics were evenly distributed between the BCG and control group within the 3 studies and combined [7–9]. The overall incidence of hospitalizations was 2.6% in the neonatal period and 13.8% in the infant period, and the overall in-hospital case-fatality rates were 29.4% and 14.9%, respectively. Readmissions accounted for 2 of 170 hospitalizations (1.2%) in the neonatal and 95 of 908 (10.5%) in the infant period (footnote in Table 1).

Overall Effects of BCG Vaccination on Hospitalizations and Their Severity During the Infant Period

Figure 2 presents the cumulative Kaplan-Meier curve of hospitalizations up to 12 months of age in the intervention and control groups, and Supplementary Table 1 provides a detailed overview, by trial and combined, of inclusions, main trial deaths and in-hospital deaths up to 1 year of age.

Figure 2.

Kaplan-Meier cumulative hazards curve for hospitalization at National Hospital Simão Mendes, combined for the 3 randomized controlled trials.

Early BCG vaccination had no effect on the incidence of hospitalizations in the neonatal, 6-week, or infant periods (Figures 2 and 3). BCG vaccination was associated with a reduced all-cause case-fatality rate RR at the hospital in all 3 trials; the combined RR was 0.58 (95% CI, .35–.94) in the neonatal period, 0.56 (.35–.90) by 6 weeks, and 0.72 (.53–.99) in the infant period (Table 1). Although the difference in case-fatality rate was still significant by 12 months of age, most of the effect occurred in the first 6 weeks of life; between 6 weeks and 12 months, the case-fatality rate RR was 0.87 (95% CI, .57–1.31).

Figure 3.

Main trial mortality [9], hospitalization, and in-hospital mortality rates across the 3 randomized controlled trials of BCG versus no BCG vaccination; data include combined BCG versus control mortality rate ratio (MRR), incidence rate ratio (IRR), and cohort risk ratio(RR) estimates for the neonatal and infant periods. *P < .05 (main trial Cox analysis); †P < .05 (2-sided Fisher exact test, Pediatric Ward data).

Effects of BCG Vaccination by Cause of Hospitalization

BCG vaccination was not associated with hospitalization incidence or in-hospital deaths either from noninfectious diseases or from the major infectious disease groups—gastroenteritis, malaria, and pneumonia (Table 2). It tended to reduce the risk of neonatal sepsis hospitalization (IRR, 0.75; 95% CI, .50–1.13) (Table 2). The trends were the same in the first 6 weeks of life (Supplementary Table 2) and in the infant period (IRR, 0.78; 95 CI, .55–1.11) (Table 2). In the infant period, BCG vaccination was associated with reduced risk of hospitalization due to bronchitis, with an IRR of 0.33 (95% CI, .12–.90) (Table 2).

There were 50 neonatal deaths at the hospital (BCG 18, controls 32) and of these, 38 deaths (BCG 12, controls 26) were due to infections (Table 2). BCG vaccination was associated with a reduced risk of fatal neonatal hospitalizations caused by infections, with an RR of 0.54 (95% CI, .30–.98) (Table 2). The absolute BCG-versus-control difference of 14 neonatal deaths from infectious diseases was attributable to fewer cases of fatal neonatal sepsis (BCG 7, controls 20); RR 0.46 (95% CI, .22–.98) (Figure 4). For the infant period, the RR for fatal sepsis was 0.55 (95% CI, .28–1.06) (Table 2).

Figure 4.

BCG versus control in-hospital deaths by disease and period of follow-up. *P < .05 (2-sided Fisher exact test). Abbreviation: RR, cohort risk ratio (pediatric ward data).

Effect of BCG Vaccination by Season of Enrollment and Sex

The overall hospitalization incidence tended to be higher both among males and among infants enrolled in the rainy season (Table 3). Both neonatal and infant BCG-versus-control IRR and RR estimates were comparable by season of enrollment and by sex (Table 3).

Table 3.

Hospitalization Incidence and Case-Fatality Rate by Sex and Season of Enrollment

	Admission Rate per 100 Person-Years (Admissions/Person-Years) [Infants Enrolled, No.]
	Neonatal (28 d)					Infant (1 y)
	BCG	Controls	IRR (95% CI) for BCG vs Controls	Case-Fatality Rate RR	P Value for Same Effect on RRa	BCG	Controls	IRR (95% CI) for BCG vs Controls	Case-Fatality Rate RR	P Value for Same Effect on RRa
Sex
Male	41 (31/75) [1174]	49 (36/74) [1168]	0.85 (.53–1.38)	0.48 (.23–1.00)b	.47	19 (201/1061) [1185]	18 (187/1031) [1180]	1.05 (.86–1.28)	0.64 (.39–1.06)	.53
Female	39 (53/136) [2092]	37 (50/135) [2091]	1.05 (.71–1.55)	0.69 (.35–1.36)	.47	13 (249/1914) [2113]	14 (271/1878) [2105]	0.90 (.76–1.07)	0.79 (.52–1.18)	.53
Season of enrollmentc
Dry	37 (50/134) [2078]	37 (48/131) [2053]	1.01 (.68–1.50)	0.58 (.32–1.05)	.95	14 (259/1907) [2097]	15 (269/1832) [2069]	0.93 (.78–1.10)	0.64 (.42–.97)b	.36
Rainy	44 (34/77) [1188]	49 (38/78) [1206]	0.91 (.57–1.45)	0.56 (.24–1.33)	.95	18 (191/1068) [1201]	18 (189/1077) [1216]	1.02 (.83–1.25)	0.86 (.53–1.39)	.36

Abbreviations: CI, confidence interval; IRR, incidence rate ratio; RR, cohort risk ratio.

^aP values comparing male vs female infants and dry vs rainy seasons.

^bP < .05 (2-sided Fisher’s exact test).

^cAs defined by the Köppen climate classification for tropical climates (rainy season: average monthly precipitation >60 mm). The dry season is from November through May, and the rainy season from June through October

Sensitivity Analysis

An analysis of case-fatality rates by timing of death rather than date of hospitalization is presented in Supplementary Table 3. Case-fatality rate RRs (95% CIs) for the neonatal, 6-week, and infant periods were 0.56 (.34–.93), 0.58 (.36–.93), and 0.71 (.51–.97). In Supplementary Table 4, an overview of control group BCG vaccination catch-up is presented by study and period of life.

Effect of BCG Vaccination on the Risk of Tuberculosis Hospitalization

One infant in the BCG group was admitted before 1 year of age with an entry diagnosis of tuberculosis, but this infant was discharged alive after 26 days with a final diagnosis of pneumonia. There were no confirmed cases of tuberculosis at the hospital.

DISCUSSION

BCG vaccination had no effect on the incidence of hospitalizations, but among hospitalized infants, it significantly reduced the case-fatality rate. This effect was due mainly to a reduction in fatal neonatal sepsis cases.

Strengths and Weaknesses

The 3 studies were RCTs with balanced randomization between BCG and control groups. Some families of control group neonates brought their children for BCG vaccination during the first weeks of life, and the results presented here are therefore likely to provide a conservative estimate of the effects of early BCG vaccination. Given that BCG vaccination was associated with markedly reduced overall mortality, it may be that early vaccination saved frail children in the BCG group and that these relatively vulnerable infants could subsequently have been subject to a higher incidence of hospitalization, which could explain the lack of an effect on admission incidence throughout the infant period. By the time of the standard 6-week vaccinations, coadministration of BCG plus DTP (controls) versus just DTP (BCG group) could also have disadvantaged the latter group [16].

The HNSM admissions occurring during the night or on weekends might not always have been registered by the BHP hospital team, or the outcome (discharged alive or died at the ward and final diagnosis) might not have been registered; infants admitted at odd hours (10 pm to 6 am) that died shortly after admission could likewise have remained undetected. There were several infants with hospital discharge codes such as “fled” or “hospital documents misplaced,” and it was possible to establish in the VA data that some of them had died during hospitalization. After the review of VA information, only 3 infants had unknown discharge status, and 1 could not be matched with a hospitalization.

Diagnostic procedures at HNSM are insufficient owing to lack of funds and chronic nationwide political and economic instability. Thus, the diagnoses presented are based largely on clinical presentation, and a proportion of the infants (19%) had only an entry diagnosis assigned. Using the entry diagnosis as an approximation represents the least loss of information when a final diagnosis is not available.

It should be noted that the BCG effect estimates were comparable across the 3 RCTs. The proportion of the total study deaths occurring at the pediatric ward, approximately 25%, was rather low, whereas the in-hospital mortality rate was high. This suggests that when infants were admitted, it was at a late stage of disease. Some children may not have been admitted, owing to lack of resources, or may have been admitted to other clinics or nongovernment hospitals, but we found no indication that the proportion admitted to HNSM differed by intervention group, so the comparison by group is therefore unlikely to be biased. However, in settings with earlier presentation to the hospital and broader diagnostic and treatment capacities, vaccine-induced effects on case-fatality rates could be diluted by early effective treatment. Our results may thus not apply to all settings.

Interpretation

It has been proposed that BCG-induced epigenetic reprogramming leading to innate immune training may explain how BCG vaccination could provide protection against unrelated pathogens [17]. In a randomized, placebo-controlled human challenge study published in 2018, BCG vaccine was shown to significantly reduce viremia after challenge with yellow fever vaccine virus [18]. Because BCG vaccine induces nonspecific protection against unrelated pathogens both in vitro and in vivo, it is biologically plausible that our results reflect similar preventive effects.

Our analysis interestingly indicates that BCG vaccination does not necessarily affect the susceptibility to infection, but rather that it reduces infection severity, particularly for neonatal sepsis. Because any major infection might progress to sepsis in the absence of sufficient immune resistance and medical treatment, it is possible that BCG vaccine induces improved resistance against a wider range of infections, ultimately preventing the progression to fatal sepsis. Our data document that the in-hospital case-fatality rate is correlated with the total effects of BCG vaccination on mortality, whereas the incidence of hospitalization was not correlated with main study mortality estimates.

Owing to declining child mortality rates in low-income countries, effects on hospitalization rates have been considered as a surrogate for effects on all-cause mortality rates when evaluating health interventions; for example, we have used the incidence of hospitalizations to assess the possible nonspecific effects of MV in Guinea-Bissau and of the measles-mumps-rubella vaccine in Denmark [19, 20]. The present study covers a period of 12 years in a scenario of declining mortality rates, and in none of the trials were effects on the incidence of hospitalizations a surrogate for effects on all-cause mortality rate.

An intervention may thus have a substantial effect on mortality rate but little or no effect on the risk of hospitalization. This has implications for the design of future BCG vaccine trials. It is unclear why hospitalization incidence was affected in the MV trials but not in the BCG vaccine trials. It could reflect the comparison between early MV recipients and control group children that had the third dose of killed DTP vaccine as the most recent vaccination, whereas controls in theBCG vaccine trials had received the live-vaccine OPV at enrollment. It could also reflect the fact that the BCG vaccine trials were conducted in the first month of life, when parents may be more inclined to take their sick neonate to the hospital; the MV was tested in infants 4.5–9 months of age, who may be taken to the hospital only for more severe infections. Finally, the immunological mechanisms of action through which the vaccines exert their effects on mortality could differ.

Implications and Conclusions

Childhood mortality rates remain substantial in low-income countries and are especially high in the neonatal period, when almost half of all deaths in children <5 years old occur; 75% of these neonatal deaths occur during the first week of life [21]. Although perinatal deaths occurring within the first hours and days after birth are mostly caused by prematurity, birth defects, and birth complications, severe infections swiftly gain importance. Thus, any intervention that can alter the severity of infection in the host during the first most vulnerable weeks of life is highly warranted.

Furthermore, improving the host response to infections will become increasingly important, because resistance to antibiotic interventions is rising in all world regions [22]. If BCG vaccine stimulates, trains, and epigenetically reprograms the immune system to improve handling of unrelated pathogens, thereby providing protection against fatal infectious diseases such as sepsis, newborns should be vaccinated as early as possible, without delay. Because BCG vaccination is cheap, safe, and feasible on a mass scale, the implications are immense, and barriers to vaccination such as restrictive vial wastage policies should be eliminated [5, 6].

Supplementary Data

Supplementary materials are available at The Journal of Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author.

Notes

Author contributions. P. A. and C. S. B. were the chief investigators and guarantors, and they designed and initiated the RCTs. F. S. B., S. B. S., N. L., I. M, P. U., A. B. F., and A. R. were responsible for the recruitment and follow-up of participants. A. A. was responsible for the statistical analysis, and F. S. B. wrote the first draft of the manuscript. All authors contributed to and approved the final version of the manuscript.

Acknowledgments. We thank Anders Rehfeld, MD, PhD, Mike Berendsen, MSc, and Anton Pottegård, cand. pharm, PhD, for assistance with a presubmission peer review.

Disclaimer.  The sponsors had no role in study design, data collection, data analysis, data interpretation, or the writing of the report.

Financial support. This work was supported by the European Union (grant ICA4-CT-2002–10053), the March of Dimes, the Danish International Development Agency and the Danish National Research Foundation (grant DNRF108 to the BHP), the Novo Nordisk Foundation (research professorship grant to P. A.), the Lundbeck Foundation (research year grant to F. S. B.), and the Danish National Research Foundation and the University of Southern Denmark (PhD grants to F. S. B.).

Potential conflicts of interest. Several of the authors (F. S .B, A. B. F., P. A., C. S. B) are affiliated with the Statens Serum Institute, which administers their grants and is a producer of BCG but did not fund the vaccines, the study, or the researchers. All other authors report no potential conflicts. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.