Active Tuberculosis in HIV-Exposed Tanzanian Children up to 2 years of Age: Early-Life Nutrition, Multivitamin Supplementation and Other Potential Risk Factors

Abstract

Background: Over half a million children worldwide develop active tuberculosis (TB) each year. Early-life nutritional exposures have rarely been examined in relation to pediatric TB among HIV-exposed children. We therefore investigated independent associations of early-life nutritional exposures with active TB among HIV-exposed children up to 2 years of age.

Methods: Participants were children from a randomized controlled multivitamin supplementation trial conducted in Dar es Salaam, Tanzania, from August 2004 to May 2008, who received daily multivitamin supplements or placebo for 24 months.

Results: Lower mean corpuscular volumes [relative risks (RR): 0.48, 95% confidence interval (CI): 0.27, 0.87] and higher birth weights (RR: 0.61, 95% CI: 0.37, 0.99) were protective against active TB, whereas multivitamin supplementation was not associated with TB risk (RR: 0.87, 95% CI: 0.65, 1.16).

Conclusions: Knowledge of nutrition-related risk and protective factors for TB in HIV-exposed children could enhance preventive and case-finding activities in this population, contributing to efforts to reduce the global TB burden.

INTRODUCTION

Tuberculosis (TB) is the second most important single infectious cause of mortality worldwide [1]. According to the World Health Organization (WHO), over half a million children develop active TB each year [1], though the estimate may be closer to 1 million [2] because TB infection may be underdiagnosed [3], especially in children [4]. Some risk factors for TB have been identified by different studies; however, few studies have examined early-life nutritional factors as risk factors for symptomatic TB disease, particularly in young HIV-exposed children.

We therefore sought first to examine if risk factors suggested by previous studies were important in HIV-exposed children. We also aimed to examine relationships of early-life nutritional factors, such as birth weight, height/length-for-age and iron status, with active TB in HIV-exposed children up to 2 years old, after accounting for previously identified risk factors. Third, multivitamin supplementation led to increased hemoglobin concentrations of the HIV-exposed children in our study [5], and given the potential of the Mycobacterium tuberculosis organisms to use the heme component of hemoglobin as an iron source for their growth [6], we aimed to understand the effect of multivitamin supplementation on the incidence of active TB among HIV-exposed children.

MATERIALS AND METHODS

The study involved children who participated in a randomized controlled trial conducted in Dar es Salaam, Tanzania, from August 2004 to May 2008. In the trial (described elsewhere [7]), 2387 HIV-exposed infants were randomly assigned at age 6 weeks to receive daily oral placebo or multivitamin supplements (vitamin C, vitamin E, thiamine, riboflavin, niacin, vitamin B-6, folate and vitamin B-12) for 24 months.

At monthly study visits, nursing staff recorded infant feeding information, took medical history, performed symptoms checks and assessed compliance with assigned regimens. Children also received routine immunizations and growth monitoring. Study physicians diagnosed and treated illnesses whenever they occurred, in addition to providing routine examinations every 3 months to all study children. All children received standard care according to Tanzanian Ministry of Health guidelines. At enrollment and every 6 months thereafter, children provided blood samples for complete blood count estimation and T-lymphocyte subset counts. Children were tested for HIV infection at 6 weeks and 18 months of age using an HIV-1 DNA polymerase chain reaction test, and infected children were treated according to national guidelines. Starting in July 2005, the standard first-line antiretroviral (ARV) regimen administered to eligible adults comprised stavudine, lamivudine and nevirapine, while children received zidovudine, lamivudine and nevirapine. Before this, standard care consisted of nevirapine prophylaxis for preventing mother-to-child transmission of HIV—mothers received one nevirapine dose at the start of labor, while the newborn received one dose within 72 h of birth.

Statistical analysis

We defined incident active TB as a new presumptive diagnosis of active TB made by study physicians, possible active TB identified using clinical criteria modified from the Edwards TB score [8] (Supplementary Table S1) or both. We evaluated the performance of the case definition among 130 HIV-infected study children who were seen by the Management and Development for Health program, Dar es Salaam, no more than 18 months after identification as a TB case and within 18 months of the last study visit for non-cases. Of these children, 19 had pulmonary TB, which was diagnosed by physicians based on clinical and chest X-ray findings (prevalence 14.6%), and the case definition had a specificity of 77.5%, sensitivity of 47.4%, positive predictive value (PPV) of 26.5% and negative predictive value (NPV) of 89.6% among these HIV-infected children.

One infant with possible active TB at baseline was excluded from analyses. To identify independent risk factors for active TB in this population, we examined independent associations of maternal and family factors with active TB in children: family size, daily amount spent on food for each family member, mother’s years of education and time-varying factors: mother’s age, CD4 T-cell count, antiretroviral therapy (ART) initiation status, mean corpuscular volume (MCV) and hemoglobin concentration. Multivariate models included the listed variables, as well as the year of study enrollment.

Next, we examined the independent associations of child-specific factors with active pediatric TB, including risk factors related to nutritional status, such as birth weight, whether colostrum was given after birth, baseline length-for-age, number of months of breastfeeding and time-varying MCV as a surrogate for iron status. Other child factors were sex, Bacille-Calmette-Guerin (BCG) vaccination status, whether the child was born at term, size for gestational age, assigned treatment regimen and time-varying factors: child’s HIV-infection status, CD4 T-lymphocyte percentage and cumulative adherence to study visits and regimens. Multivariate models included the maternal and family factors, study enrollment year, child’s age at baseline, as well as the child factors listed above.

Finally, we examined the effect of multivitamin supplementation on active TB—analyses of multivitamin effects were by intention-to-treat—and explored any effect modification by the child’s baseline HIV-infection status, hemoglobin concentration and MCV status. Likelihood ratio tests were used to test for interactions between multivitamin supplementation and potential effect modifiers. For all analyses, hazard ratios [subsequently referred to as relative risks (RR)] were estimated using the Andersen-Gill formulation of Cox proportional hazards models [9]. All analyses were performed using SAS software version 9.4 (SAS Institute, Cary, NC, USA).

RESULTS

Of the 2387 infants enrolled in the randomized trial, 2358 qualified for the current study (Fig. 1). Most children had mothers with at least 7 years of schooling, received BCG vaccination at birth and were being breastfed at study enrollment; 11% were HIV-infected at baseline (over 60% of HIV-infected infants had started ART at baseline). Baseline characteristics were similar for children in the multivitamin and placebo groups (Table 1).

Fig. 1.

Study flow diagram.

Table 1.

Baseline characteristics of children by randomized treatment assignment^a

Baseline characteristics	Placebo group	Multivitamin group
	N = 1179	N = 1179
	n (%)a	n (%)a
Maternal and family factors
Family size	3 [2,4]	3 [2,4]
Amount spent on food per person (TSh)b	508 (442)	497 (417)
Mother’s education (years)	7.1 (2.8)	7.3 (2.7)
Mother’s age (years)	28.1 (4.9)	28.1 (5.0)
Mother’s pregnancy CD4 T-cell count (/μl) and ARV status
CD4 <350 and ARV in pregnancy	106 (10.0)	115 (11.1)
CD4 <350 but no ARV in pregnancy	183 (17.3)	214 (20.6)
CD4 ≥350	771 (72.7)	711 (68.4)
Mother’s hemoglobin (g/l) and MCV (fl)
Hemoglobin <110 and MCV <80	185 (17.7)	185 (17.9)
Hemoglobin <110 and MCV ≥80	136 (13.0)	125 (12.1)
Hemoglobin ≥110 and MCV <80	195 (18.6)	184 (17.8)
Hemoglobin ≥110 and MCV ≥80	530 (50.7)	539 (52.2)
Child factors
Age (months)	1.3 (0.1)	1.3 (0.1)
Sex (female)	551 (46.7)	531 (45.0)
Hemoglobin (g/dl)	10.3 (1.6)	10.4 (1.5)
MCV (fl)	89.9 (7.4)	90.0 (7.3)
Stunting (HAZc <−2)	109 (9.4)	90 (7.8)
Wasting (WHZc <−2)	89 (7.7)	88 (7.6)
HIV positive	135 (11.6)	123 (10.6)
CD4 T-cell percentage
<15	27 (2.9)	27 (2.9)
15–<25	92 (9.7)	98 (10.5)
≥25	827 (87.4)	812 (86.7)
Gestational age (GA) and size for GA
GA ≤37 weeks and SGAd	11 (1.0)	5 (0.4)
GA ≤37 weeks and not SGAd	207 (18.8)	232 (20.6)
GA >37 weeks and SGAd	142 (12.9)	129 (11.5)
GA >37 weeks and not SGAd	743 (67.4)	761 (67.5)
Birth weight (kg)	3.1 (0.5)	3.1 (0.5)
Neonatal BCG vaccine administered	906 (90.4)	929 (90.0)
Received colostrum after delivery	900 (89.8)	936 (90.7)
Exclusively breastfed	999 (84.7)	1025 (86.9)

^aMean (SD), median [Q1, Q3] for continuous variables and n (% of those with non-missing information) for categorical variables (any percentage totals exceeding 100% were a result of rounding). No significant differences between placebo and multivitamin groups.

^bTanzanian Shillings (TSh) spent by the child’s family on food per person per day. During the study, TSh 1250 ∼ 1 US Dollar.

^cHAZ is height/length-for-age Z score estimated using the WHO child growth standards of 2006 as reference; WHZ is weight-for-height/length Z score.

^dSGA, i.e. birth weight below the 10th percentile for the gestational age.

The incidence rate of active TB among children was 5.5 per 100 child years, with a median time to first diagnosis of 6.6 months (interquartile range 2.8−11.5 months) for the 183 TB cases. Maternal and family risk factors for active TB in children are presented in Table 2. After adjusting for other risk factors, having a mother with >7 years of schooling significantly lowered a child’s risk of TB by 40% [95% confidence interval (CI): 5–61% lower, p = 0.03].

Table 2.

Maternal and family risk factors for active TB in study children

Risk factor	Univariate	p	Multivariateb	p
	RR (95% CI)a		RR (95% CI)a
Family size		0.67		0.99
>3	Ref.		Ref.
≤3	1.07 (0.79, 1.44)		0.98 (0.68, 1.41)
Daily amount spent on food per person (TSh)		0.57		0.75
≥500 TSh	Ref.		Ref.
<500 TSh	1.09 (0.81, 1.48)		0.88 (0.59, 1.33)
Mother’s education (years)		0.22		0.09
None	0.95 (0.54, 1.67)		0.93 (0.50, 1.72)
1–7	Ref.		Ref.
≥8	0.61 (0.40, 0.93)		0.60 (0.39, 0.95)
Mother’s age (years)c	0.99 (0.97, 1.02)	0.70	0.99 (0.95, 1.02)	0.45
Mother’s CD4 T-cell count (/μl)c		0.02		0.21
≥500	Ref.		Ref.
350–<500	1.23 (0.83, 1.83)		1.27 (0.84, 1.91)
<350	1.53 (1.07, 2.18)		1.44 (0.97, 2.12)
Mother’s ARV initiation statusc		0.004		0.13
Did not initiate ARVs	Ref.		Ref.
Initiated ARVs	1.73 (1.19, 2.52)		1.43 (0.91, 2.25)
Mother’s hemoglobin (g/l)c		0.14		0.63
≥12.0	Ref.		Ref.
11.0–<12.0	0.96 (0.65, 1.42)		0.90 (0.61, 1.34)
8.0–<11.0	1.11 (0.75, 1.62)		1.01 (0.67, 1.53)
<8.0	1.75 (0.64, 4.84)		1.55 (0.54, 4.45)
Mother’s MCV (fl)c		0.56		0.78
Highest quartile: >87.7	Ref.		Ref.
Third quartile: 82.9–87.7	0.82 (0.53, 1.27)		0.86 (0.55, 1.36)
Second quartile: 77.1–82.8	0.81 (0.52, 1.26)		0.90 (0.57, 1.41)
Lowest quartile: <77.1	0.87 (0.57, 1.34)		0.90 (0.56, 1.45)

^aRR (95% CI).

^bAdjusted for year of study enrollment and the other presented factors, i.e. family size, daily amount spent on food per person, mother’s years of education and the time-varying factors: mother’s age, mother’s CD4 count, mother’s ARV initiation status, mother’s hemoglobin concentration and mother’s MCV.

^cTime-varying variables.

Table 3. shows univariate and multivariate estimates of the associations of child-specific factors with active TB. After adjusting for family, maternal and the other child-specific factors, boys were significantly more likely than girls to develop active TB (RR: 1.44, 95% CI: 1.03, 2.02). HIV-infected children had a TB risk 4.8 times (95% CI: 3.2–7.1 times) greater than the risk of HIV-uninfected children. Also associated with a higher risk of active TB were having CD4 T-cell percentages <15%, relative to percentages ≥25% (RR: 2.03, 95% CI: 1.04, 3.95), and being born preterm but not small-for-gestational-age (SGA) versus term and not SGA (RR: 1.89, 95% CI: 1.22, 2.90).

Table 3.

Child-specific risk factors for active TB in study children

Risk factor	Univariate	p	Multivariateb	p
	RR (95% CI)a		RR (95% CI)a
Sex		0.11		0.03
Female	Ref.		Ref.
Male	1.28 (0.95, 1.72)		1.44 (1.03, 2.02)
Neonatal BCG vaccination		0.04		0.49
Vaccinated	Ref.		Ref.
Not vaccinated	1.61 (1.02, 2.52)		1.46 (0.75, 2.87)
HIV statusc		<0.0001		<0.0001
HIV negative	Ref.		Ref.
HIV positive	5.76 (4.27, 7.77)		4.79 (3.24, 7.07)
CD4 T-cell percentagec		<0.0001		0.06
≥25	Ref.		Ref.
15–<25	2.72 (1.85, 3.98)		1.33 (0.84, 2.10)
<15	4.57 (2.62, 7.98)		2.03 (1.04, 3.95)
Gestational age (GA) and size for GA		<0.0001		0.06
GA >37 weeks and not SGAd	Ref.		Ref.
GA ≤37 weeks and SGAd	5.43 (1.99, 14.82)		2.28 (0.47, 11.08)
GA ≤37 weeks and not SGAd	2.49 (1.77, 3.50)		1.89 (1.22, 2.90)
GA >37 weeks and SGAd	2.17 (1.43, 3.30)		1.21 (0.67, 2.20)
Birth weight (kg)	0.39 (0.29, 0.54)	<0.0001	0.61 (0.37, 0.99)	0.05
Colostrum at delivery		0.33		0.30
Given	Ref.		Ref.
Not given	1.28 (0.78, 2.09)		0.66 (0.31, 1.44)
Stunting at baseline		0.0001		0.84
HAZe ≥−2	Ref.		Ref.
HAZe <−2	2.21 (1.48, 3.31)		0.85 (0.48, 1.50)
MCV (fl)c		0.04		0.02
Highest quartile: >83.4	Ref.		Ref.
Third quartile: 72.7–83.4	0.67 (0.41, 1.10)		0.64 (0.38, 1.08)
Second quartile: 66.1–72.6	0.38 (0.22, 0.68)		0.40 (0.22, 0.74)
Lowest quartile: <66.1	0.57 (0.33, 0.98)		0.48 (0.27, 0.87)
Duration of breastfeeding (months)c	1.02 (0.95, 1.09)	0.65	0.98 (0.91, 1.05)	0.54
Assigned regimen		0.42		0.44
Placebo	Ref.		Ref.
Multivitamins	0.87 (0.65, 1.16)		0.88 (0.63, 1.22)
Compliance with assigned regimenc		0.51		0.75
>Median compliancef	Ref.		Ref.
≤Median compliancef	1.10 (0.82, 1.48)		1.06 (0.76, 1.47)

^aRR (95% CI).

^bAdjusted for family size, daily amount spent on food per person, mother’s years of education, mother’s age, child’s age at baseline, child’s sex, study enrollment year, birth weight, neonatal BCG vaccination status, whether colostrum consumed at delivery, whether preterm and/or SGA, assigned treatment regimen, baseline stunting status and the time-varying factors: child’s cumulative compliance with assigned regimen, child’s duration of breastfeeding, child’s CD4 percentage, child’s MCV, child’s HIV status, mother’s CD4 count, mother’s ARV status, mother’s MCV, mother’s hemoglobin.

^cTime-varying variable.

^dSGA, i.e. birth weight below the 10th percentile for gestational age.

^eHeight/Length-for-age Z score estimated using the WHO child growth standards of 2006 as reference.

^fMedian cumulative compliance was 95.9%.

On the other hand, children with lower MCV and higher birth weights were significantly less likely to develop active TB. Each 1 kg increase in child’s birth weight was associated with 1−63% lower TB risk, and the risk of active TB was 60% lower and 52% lower for the lowest and second-lowest quartiles of MCV, respectively, compared with the highest MCV quartile.

Multivitamin supplementation did not significantly alter the risk of active TB (RR: 0.87, 95% CI: 0.65, 1.16) and there was no effect modification by child’s baseline HIV infection status, baseline MCV or hemoglobin concentration (p values for interactions with assigned regimen were 0.24, 0.74 and 0.88, respectively).

DISCUSSION

We found that among HIV-exposed children, lower MCV, higher birth weights and having a mother with more years of schooling were protective against developing active TB, and male sex, being born preterm, HIV infection and CD4 T-cell percentages <15% were independently associated with elevated risk of active TB. Daily use of multivitamin supplements did not affect children’s TB risk, nor did family size, mother’s age, mother’s CD4 T-cell counts, or child’s neonatal BCG vaccination, colostrum at delivery, baseline length-for-age or breastfeeding duration.

Nutritional exposures early in life may affect susceptibility to TB disease. It has been hypothesized that exposure to undernutrition in utero impairs development of CD4 T-cells responsible for immunity against TB [10]. A study involving monozygotic twins born in Sweden between 1920 and 1960 [11] also reported lower TB risks as birth weights increased.

Lower MCV, a surrogate for lower iron levels in this setting, was associated with lower risks of active TB among study children. Mycobacterium tuberculosis organisms need iron to function [12] and may fail to thrive in low-iron environments [13]. Intravenous iron administration significantly increased multiplication of M. tuberculosis in mice [14], while dietary iron intake has also been associated with increased TB risk in a human study [15]. Elevated plasma ferritin was associated with increased clinical TB incidence in one adult study [16], agreeing with our findings. However, another study reported higher TB recurrence in adults with low plasma ferritin concentrations [17], although the low levels of plasma iron may have resulted from iron redistribution following TB infection [16].

Half of our study participants received multivitamin supplements for 24 months. Multivitamin supplementation improved hemoglobin concentrations of study children [5] and of HIV-infected women [18]. Although the multivitamin supplement did not include iron, it included vitamin C, which has been shown to increase intestinal iron absorption [19], potentially increasing host iron available to pathogens. Additionally, M. tuberculosis organisms are capable of using heme as an iron source [6]. We therefore examined if multivitamin supplementation could have inadvertently raised children’s risk of active TB, but found no harmful effects of multivitamin supplementation on the incidence of active TB.

Consistent with other studies [20, 21], we observed that HIV-infected children with low CD4 T-cell percentages had elevated risks of developing active TB, found higher occurrence of TB in males [22] and protective associations of maternal education [23], possibly reflecting associated improvements in childcare and nutrition. Unlike in some prior studies, family size [23, 24] was not associated with active TB in this study, and we observed a tendency toward protective associations for neonatal BCG vaccination only in univariate analyses.

A limitation we encountered was that microbiological and radiological investigations useful for confirming TB were not routinely performed for study children, necessitating the use of clinical criteria to identify TB cases. This led to random outcome misclassification and underdiagnoses of TB in the children. Clinical scoring systems have been useful in resource-limited settings as TB screening tools, but they may have low specificity and sensitivity in HIV-infected individuals [25, 26]. We attempted to validate the case definition in a sample of those HIV-infected study children who had access to routine chest X-rays as part of their care, and found that while sensitivity of the clinical definition in detecting true cases was a little less than 50% and PPV about 27% (TB prevalence was 14.6%), specificity was nearly 80% and NPV about 90%, suggesting that our case definition may have tended to identify the more obvious TB cases, but usually did not wrongly identify non-cases as having TB disease. The performance of our TB diagnostic criteria among HIV-infected children is similar to reports from a South African study by Marais et al. [27] In the study, symptoms-based criteria comprising objective weight loss in the preceding 3 months reported fatigue, and persistent cough for >2 weeks had a sensitivity of 56.2% and specificity of 61.8% among HIV-infected children, 82.3% and 90.2%, respectively, among HIV-uninfected children ages ≥3 years and 51.8% and 92.5%, respectively, among HIV-uninfected children <3 years old. Although we did not have the information required for evaluating the performance of our TB diagnostic criteria in HIV-uninfected children, we expect that similar to other studies, our diagnostic criteria performed better in HIV-uninfected children than in HIV-infected children.

We examined surrogates of nutritional status as risk factors, for instance MCV to measure iron status, or birth weight and size for gestational age as markers of prenatal nutrition. While more direct measures of body iron are preferable, use of MCV as a surrogate may be appropriate in our setting, where 42% of Tanzanian children <24 months are iron deficient [28] and nearly half of the anemia is iron deficiency anemia [28].

Our study contributes to the literature by identifying nutrition-related factors associated with active TB in HIV-exposed children. Recent reports suggest that up to half of pediatric TB cases are undiagnosed and close to 1 million children may develop active TB each year [2]; yet, TB increases risk of mortality [1] if untreated. Given the high proportion of undiagnosed children, calls were recently made to encourage development of new methods for identifying individuals affected by TB [29]. More research will be required to determine if the early-life nutrition-associated factors we identified, such as MCV and birth weight, are related to TB disease in HIV-exposed and HIV-unexposed children elsewhere in the world. Knowledge of risk factors in young children could contribute to development of preventive and diagnostic strategies, and increase the effectiveness of case-finding activities in children. This would enhance the success of ongoing efforts aimed at reducing the burden of TB globally, as well as in vulnerable populations.

FUNDING

This work was supported by a grant from the National Institute of Child Health and Human Development [grant no. K24HD058795, to E. L., K. P. M., C. D., S. A. and W. W. F.]

SUPPLEMENTARY DATA

Supplementary data are available at Journal of Tropical Pediatrics online