Nutritional status and complementary feeding among HIV‐exposed infants: a prospective cohort study

Abstract

Complementary feeding is crucial for improving child survival and promoting growth and development, particularly among HIV‐exposed children who have higher risk of morbidity and mortality than their un‐exposed peers. This prospective study employed an infant and child feeding index (ICFI) to measure complementary feeding and determine its association with nutritional status among 2092 HIV‐exposed infants followed from 6 to 24 months of age in Dar es Salaam, Tanzania. The ICFI measured both quality and quantity of complementary feeding, including current breastfeeding status, food consistency, dietary diversity scores (DDS), food group frequency score, and meal frequency. The ICFI score ranged from 0 to 9; the median score was 6 (Inter‐Quartile Range, IQR= 4–7). After adjusting for potential confounders, high ICFI scores were associated with reduced risk of stunting (high vs. low tertile hazard ratio, HR: 0.72; 95% confidence interval, CI: 0.57, 0.91; P< 0.01) and underweight (high vs. low tertile HR: 0.79; 95% CI: 0.61, 1.02; P= 0.07). Low DDS were associated with higher risk of stunting (low vs. high tertile HR: 1.59; 95% CI: 1.23, 2.07; P< 0.01) and underweight (low vs. high tertile HR: 1.48; 95% CI: 1.12, 1.96; P= 0.01). In this setting, high DDS and ICFI scores were protective of stunting and underweight. We recommend for nutrition programs in low‐income countries to emphasize educating HIV‐exposed children's caregivers on the importance of dietary diversity and optimal complementary feeding to improve nutritional status in this important subpopulation.

Introduction

While global rates of undernutrition have been on the decline since 1990 (UNICEF, WHO, World Bank 2015), an estimated 45% of global child deaths (3.1 million annual child deaths) are caused by undernutrition (World Health Organization 2015). Low income countries bear a disproportionate share of this burden with 96% of all wasted and 94% of all stunted children under‐five living in Asia and Africa in 2014 (UNICEF, WHO, World Bank 2015).

Infant and young child feeding is a key area for improving child survival and promoting healthy growth and development particularly in the first 2 years of a child's life (World Health Organization 2015). WHO and UNICEF recommend exclusive breastfeeding for the first 6 months followed by introduction of nutritionally adequate and safe complementary foods with continued breastfeeding up to 2 years of age or beyond (World Health Organization 2015). Over one million HIV‐exposed children are born worldwide every year (Sugandhi et al. 2013) and evidence suggests that these children are at higher risk of morbidity and mortality compared with their un‐exposed peers (Brahmbhatt et al. 2006; Marinda et al. 2007; Shapiro et al. 2007; Filteau 2009; Landes et al. 2012). It is recommended that HIV‐infected mothers take antiretroviral (ARV) drugs and exclusively breastfeed their babies for 6 months, then introduce appropriate complementary foods and continue breastfeeding at least up to the child's first birthday. Even in the absence of antiretroviral therapy (ART), breastfeeding should continue until a nutritionally adequate and safe diet without breast milk can be provided (WHO 2010; World Health Organization 2015).

Efforts to quantify complementary feeding practices and assess their associations with children's health outcomes have been limited as these practices encompass a series of inter‐related behaviours that must be considered simultaneously and are, therefore, difficult to summarize into one or a few variables that accurately reflect these practices (Ruel & Menon 2002). To examine the effect of child feeding practices on child stunting, Ruel and Menon developed the infant and child feeding index (ICFI) based on an age‐specific scoring system and employed it to demographic and health surveys (DHS) data from five Latin American countries. Several investigators have since adapted the score to measure infant and young child feeding practices in various countries. However, previous studies that used the ICFI were for the most part cross sectional and conducted among HIV un‐exposed children. Furthermore, in the sub‐Saharan African context, most of the literature on feeding practices among HIV‐exposed children has focused on specific feeding behaviours such as exclusive breastfeeding and age at introduction of complementary foods (Kiarie et al. 2004; Young et al. 2010; Mwiru et al. 2011; Natchu et al. 2012; Okanda et al. 2014) and has not captured the multidimensionality of feeding practices.

We examined the relation of complementary feeding as measured by the ICFI with nutritional status among HIV‐exposed infants who participated in a randomized controlled trial of micronutrients in Dar es Salaam, Tanzania.

Key messages.

• High dietary diversity was associated with reduced risk of stunting and underweight.

• Optimal complementary feeding as measured by the infant and child feeding index (ICFI) was associated with reduced risk of stunting and underweight.

• Nutrition programmes in resource limited settings should emphasize educating HIV‐exposed children's caregivers on improving quality of complementary food, example, in terms of increasing dietary diversity, to facilitate optimal feeding and therefore improve nutritional status in this important subpopulation.

Materials and methods

Study design and population

HIV‐exposed infants between 5 and 7 weeks age whose mothers were residents of Dar es Salaam, Tanzania, were enrolled in a trial of multiple micronutrients and followed for 24 months. The trial aimed to evaluate whether direct supplementation of these micronutrients to HIV‐exposed infants reduces morbidity and mortality. Between 2004 and 2007, clinical and other health‐care services for pregnant women and children were provided during monthly visits at a study clinic according to the Tanzania Ministry of Health guidelines.

The study was a randomized, double‐blind, placebo‐controlled trial. Women aged at least 18 years presenting for prenatal care at the 32^nd week of gestation or earlier in 1 of 8 antenatal clinics in Dar es Salaam were offered HIV screening with pre‐ and posttest counselling. Women who tested HIV positive were further screened for eligibility, including the intention to reside in Dar es Salaam for the duration of follow‐up. Written informed consent was obtained from women for participation in the trial while still pregnant. Eligibility for infant participation in the trial included singleton birth and age between 5 and 7 weeks at randomization. Excluded were infants born of multiple gestation or those with serious congenital anomalies or other conditions that would interfere with study procedures, including the ability to consume a daily micronutrient supplement. The primary outcome of the trial was all‐cause mortality while the secondary outcomes were hospitalization, unscheduled clinic visits and common infectious morbidities including diarrheal disease and lower respiratory tract infections. Detailed descriptions of the trial design and follow‐up procedures have been published elsewhere (Duggan et al. 2012; Liu et al. 2013).

Data collection

Mothers and their children were asked to return to the clinic every month for research visits and standard clinical care. Mothers were counselled on the risks and benefits of exclusive breastfeeding in keeping with WHO recommendations in place during the study period. At the time of the study, WHO recommended that HIV‐exposed children be exclusively breastfed ‘for the first months of life’ followed by replacement feeding as soon as it was acceptable, feasible, affordable, sustainable and safe, taking into account local circumstances and the individual woman's situation (World Health Organization 2001). At each clinic visit, women were asked about infant feeding practices in the past seven days such as breastfeeding status and frequency, introduction of other liquid or foods including water, tea, juice, cow's milk, infant formula, porridge, mashed vegetable and meat. We defined exclusive breastfeeding as feeding a child with breast milk only without additional foods or water. The duration of exclusive breastfeeding was calculated as the mean of infant ages at which the last time the mother reported that the child was still exclusively breastfeeding and the first time the mother reported that the child was given other foods or water in addition to breast milk.

At the time of the study, ARV medication was limited to nevirapine prophylaxis for maternal to child transmission; one dose given to the mother at the onset of labour and one dose given to the infant within 72 h of birth (Guay et al. 1999). All children were tested for HIV infection at 6 weeks of age by using the Amplicor HIV‐1 DNA assay version 1.5 (Roche Molecular Systems Inc). Tests at 18 months of age were performed by using HIV ELISAs followed by Enzygnost anti‐HIV‐1/2 Plus (Dade Behring); discordant results were resolved by using a Western blot test. Samples from children who tested positive at 18 months were then back tested via polymerase chain reaction to estimate time of transmission. Children who were HIV‐negative at 18 months and still breastfed were tested again before they were discharged from the study after 24 months of follow‐up. Blood specimens were requested from each mother at baseline to measure complete blood counts including haemoglobin concentrations and T‐cell subset counts. Complete blood counts, haemoglobin concentrations and T‐cell subset counts were also measured for children at baseline and every six months thereafter, until the end of the follow‐up. haemoglobin concentrations were measured using AcT5 Diff AL haematology analyser (Beckman Coulter, Jersey City, NJ, USA) and T‐cell subset, for example absolute CD4 cell count and percentage were performed with the FACSCalibur system (Becton‐Dickinson, San Jose, CA, USA) (Liu et al. 2013).

At monthly visits, research nurses measured maternal and child anthropometry including weight, height or length and mid upper‐arm circumference (MUAC) using standardized procedures and calibrated instruments. Child's weight was measured to the nearest 10 g with a digital infant balance (TANITA), and length was measured to the nearest 1 mm with a rigid length board with a movable foot piece. Anthropometric indices were computed and compared with reference data from the 2006 World Health Organization growth chart (WHO Multicentre Growth Reference Study Group 2006). Children below −2 SD of the WHO median weight‐for‐age, length‐for‐age and weight‐for‐length were considered as underweight, stunted and wasted, respectively.

Infant and child feeding index

Complementary feeding was measured using a composite index combining multiple dimensions of child feeding practices based on both current infant and young child feeding recommendations (World Health Organization 2005; World Health Organization 2008) and previous work on the subject. We defined the ICFI score for three age groups: 6–8 months, 9–11 months and 12–24 months. The index consisted of five components: current breastfeeding status, consistency of complementary food, dietary diversity score (DDS), food group frequency score (FGFS) and meal frequency score (MFS), which were attributed in a manner reflecting the age change over the study period.

Current breastfeeding

A score of +1 was given to breastfeeding children while a score of 0 was given to non‐breastfeeding children at all ages.

Complementary food consistency

Food consistency replaced bottle feeding because data on bottle use was not collected for this cohort of children. It was defined as the thickest consistency of complementary food fed to the child in the previous week. For 6 to 8‐month‐old infants, liquid food was given a score of 0, gruel‐like food such as porridge +1 and semi‐solid food such as mashed vegetables +2. For older children, gruel‐like food, semi‐solid food and solid food were scored as 0, +1, and +2, respectively. A similar study (Ma et al. 2012) replaced the bottle feeding component with food consistency. For regression analysis purposes, we grouped together children in the low and average categories of this score because in majority of child visits it was reported that the children were fed solid foods and thus fell in the high food consistency category.

DDS

DDS was calculated on the basis of the number of different food groups infants consumed over the prior 24 h. We categorized food items into four food groups: (1) cereals; (2) dairy products (infant formula, cow's milk and milk powder; excludes breast milk); (3) flesh foods (meat, fish); and (4) fruits and vegetables. Each food group consumed in the previous 24 h was scored as 1. These scores were summed to give a possible range of 0–4 then categorized based on tertiles. Consuming 0 to 2 food groups in the previous 24 h was categorized as a DDS score of 0, 3 food groups a score of +1 and 4 food groups a score of +2.

FGFS

FGFS was assessed based on the frequency of consumption of the four food groups described above by infants over the previous 7 days. Each food group scored 0 if not consumed during the previous week, +1 if consumed on 1 to 3 days and +2 if consumed on 4 days or more (Moursi et al. 2008; Moursi et al. 2009; Ma et al. 2012; Lohia & Udipi 2014). These scores were summed to give a possible range of 0–8 then new scores were assigned based on tertiles.

MFS

MFS was based on current feeding recommendations, according to which 6 to 8‐month‐old breastfed children should receive solid, semi‐solid or soft foods (excluding milk feeds) at least twice a day and 9 to 23‐month‐old breastfed children at least 3 times a day (WHO 2008), while non breastfed children aged 6–23 months should be fed solid, semi‐solid or soft foods (including milk feeds) four to five times a day (World Health Organization 2005; WHO 2008).

The final ICFI was a summation of the scores obtained for each variable described and ranged from 0 to 9 for the three age groups. For each age group, the feeding index was grouped into tertiles to form three categories: low, medium and high. Table 1 summarizes the variables and scoring system used to create the ICFI for 6 to 24‐month‐old children by age group.

Table 1.

Description of the scoring system used to construct the infant and child feeding index

Variable	6–8 months	9–11 months	12–24 months
Currently breast feeding	No = 0	No = 0	No = 0
	Yes = +1	Yes = +1	Yes = +1
Food consistency*	Liquid = 0	Gruel‐like = 0	Gruel‐like = 0
	Gruel‐like = +1	Semisolid = +1	Semisolid = +1
	Semi‐solid = +2	Solid = +2	Solid = +2
Dietary diversity score (DDS) past 24 h †	0–2 FG = 0	0–2 FG = 0	0–2 FG = 0
	3 FG = +1	3 FG = +1	3 FG = +1
	4 FG = +2	4 FG = +2	4FG = +2
Food group frequency score (FGFS) past 7 days ‡	0–4 FGFS = 0	0–5 FGFS = 0	0–5 FGFS = 0
	5 or 6 FGFS = +1	6 or 7 FGFS = +1	6 or 7 FGFS = +1
	>6 FGFS = +2	8 FGFS = +2	8 FGFS = +2
Meal frequency score (MFS) past 24 h §	0 meals for BF, 0–1 for non‐BF = 0	0–1 meals for BF, 0–1 for non‐BF =0	0–1 meals for BF, 0–2 meals for non‐BF = 0
	1 meal for breastfed, 2–3 for non‐BF = +1	2 meals for BF, 2–3 for non‐BF = +1	2 meals for BF, 3–4 for non‐BF = +1
	≥2 meals for BF, ≥4 for non‐BF = +2	≥3 meals for BF, ≥4for non‐BF = +2	≥3 meals for BF, ≥5 for non‐BF = +2

FG, Food Group; BF, breastfed.

^* Food consistency was the thickest consistency of complementary food taken by the child in the prior week. For 6 to 8‐month‐old infants, liquid food was given a score of 0, gruel‐like food such as porridge +1 and semi‐solid food such as mashed vegetables +2. For older children, gruel‐like food, semi‐solid food and solid food were scored as 0, +1 and +2, respectively.

^† DDS was calculated on based on four food groups: (1) cereals, (2) dairy products; (3) flesh foods and (4) fruits and vegetables. Each food group consumed in the previous 24 h was scored as 1. These scores were summed to give a possible range of 0–4 then categorized based on tertiles. Consuming 0 to 2 food groups in the previous 24 h was categorized as a DDS score of 0, 3 food groups a score of +1 and 4 food groups a score of +2.

^‡ FGFS was assessed based on the frequency of consumption of different food groups over the previous 7 days. Each food group scored 0 if not consumed during the previous week, +1 if consumed on 1 to 3 days and +2 if consumed on 4 days or more. These scores were summed to give a possible range of 0–8 then new scores were assigned based on tertiles.

^§ Meal frequency scoring was based on current feeding recommendations, according to which 6 to 8‐month‐old breastfed children should receive solid, semi‐solid or soft complementary foods (excluding milk feeds) at least twice a day and 9 to 23‐month‐old breastfed children at least 3 times a day, while non breastfed children aged 6–23 months should receive solid, semi‐solid or soft complementary feeds (plus milk feeds) four to five times a day.

Statistical analysis

A total of 2387 singleton children were born alive to HIV‐infected women in this study. Of these, 273 children did not have data in the relevant age period of 6 to 24 months and 22 children lacked data on infant feeding, leaving 2092 children available for analysis.

Examination of associations of the ICFI and each of its individual components with the time to first stunting, underweight and wasting was done using Cox proportional hazards regression methods with the Andersen–Gill data structure (Andersen & Gill 1982) (one observation per inter‐visit interval) to accommodate time‐varying covariates. The ICFI and its components were treated as time‐varying variables as were maternal body mass index (BMI), maternal CD4 cell count, maternal haemoglobin levels and child's age, child's CD4 cell percentage, child's HIV infection status and child's haemoglobin levels. For inclusion in the multivariate models, we considered the potential confounders and independent risk factors for the outcomes from a list of candidate variables; covariates with a p value < .2 in univariate analysis were included in multivariate models. Stepwise regression was performed to identify which of the five ICFI components drove the association between the index and each nutritional outcome. We tested for interaction between micronutrient supplementation and the ICFI, and HIV‐infection and the ICFI for each nutrition outcome. Children were censored at the end of 24 months if they were event‐free, or at their last visit. Effect estimates were hazard ratios (HR) and their corresponding 95% confidence intervals. A p < 0.05 was considered for statistical significance. All analyses were performed using the Statistical Analysis Systems statistical software package version 9.3 (SAS Institute Inc., Cary, NC, USA).

Institutional approval was granted by the Harvard T. H Chan School of Public Health Human Subjects Committee, the Muhimbili University of Health and Allied Sciences Committee of Research and Publications, the Tanzanian National Institute of Medical Research and the Tanzanian Food and Drugs Authority.

Results

Sample characteristics

A total of 2092 children were included in this analysis, more than half (54%) of whom were male (Table 2). At enrollment, 8% of the children were HIV infected and 41% were anaemic with less than 10 g/dL haemoglobin. Of 1919 children not known to be HIV‐infected at 6 weeks age, 72 (3.8%) were diagnosed with HIV infection by the end of follow‐up. Most of the children (86%) were exclusively breastfed (EBF); the median EBF duration was 3.6 months. Mean maternal age was 28 years ±5. At six weeks post‐partum, 12% of the mothers were in the WHO HIV stage III or IV.

Table 2.

Characteristics of HIV infected mothers and their children enrolled in A micronutrient supplementation clinical trial in Dar es Salaam, Tanzania (n = 2092) (mean ± SD or %)

	Frequency	Mean ± SD /%
Maternal characteristics
Age (years) mean ± SD		28 ± 5
Married/living with partner	1804	87.1
Education
None	148	7.1
1–7 years	1491	71.9
8+ years	435	21
Spouse's education
None	25	1.5
1–7 years	1032	59.9
8+ years	665	38.6
Employment
Housewife without income	1341	66.5
Housewife with income	187	9.3
Businesswoman	114	5.7
Other	375	18.6
Daily food expenditure per person < Tshs. 500*	1021	51.7
BMI group (kg/m2) †
<18.5	89	4.3
18.5 – <25	1154	55.9
25–30	623	30.2
≥30	200	9.7
CD4 count (cells/mm) †
<200	170	9.0
200 – <350	372	19.8
350+	1339	71.2
Haemoglobin <11 g/dL †	566	29.7
WHO disease stage†
I	982	73.5
II	194	14.5
III/IV	160	12.0
Used antiretroviral therapy during pregnancy‡	388	19.5
Number of prior pregnancies
None	450	21.7
1	699	33.7
2	457	22.1
3+	467	22.5
Child characteristics
Male sex	1130	54
Low birth weight (<2500 g)	129	6.4
Baseline HIV positive status‡	173	8.4
CD4% at baseline‡
<15%	37	2.2
15% – <25%	158	9.4
25% +	1487	88.4
Child's haemoglobin at baseline < 10 g/dL ‡	823	41.2
Underweight at baseline‡	159	7.7
Wasted at baseline‡	140	6.8
Stunted at baseline‡	165	8
Exclusively breast fed	1806	86.3
Exclusive breast feeding duration < 3 months	828	39.6

^* 1250 Tshs = ~ 1 US$ at the time of the study.

^† Maternal characteristics measured at 6 weeks postpartum.

^‡ Child characteristics measured at randomization.

Complementary feeding practices

The median ICFI score was 6 (Inter‐Quartile Range; IQR = 4–7). Breastfeeding decreased as the children grew older; breastfeeding was reported in 10% of visits at 6–8 months, 3% at 9–11 months and 1% at 12–24 months (Table 3). Consistency of complementary foods increased as the children grew older; a high food consistency score was reported in 31% of visits at 6–8 months, 96% at 9–11 months and 99% at 12–24 months. A dietary diversity of four food groups in the past 24 h was reported in 50% of visits at 6–8 months, 62% at 9–11 months and 49% at 12–24 months. A minimum meal frequency (measured by the high MFS tertile) based on WHO recommendations was reported in 40% of visits at 6–8 months, 33% at 9–11 months and 5% at 12–24 months.

Table 3.

Feeding practices as per the infant and child feeding index (ICFI) by age group

Variables	6–8 months N = 5666 child visits	9–11 months N = 4885 child visits	12–24 months N = 17 271 child visits
Currently breast feeding	577 (10.2%)	132 (2.7%)	106 (0.6%)
Food consistency
Low	168 (3.1%)	157 (3.2%)	144 (0.8%)
Medium	3559 (66.1%)	42 (0.9%)	85 (0.5%)
High	1657 (30.8%)	4646 (95.9%)	16 958 (98.7%)
Dietary diversity score (DDS) past 24 h
Low	1114 (19.7%)	378 (7.7%)	2390 (13.8%)
Medium	1726 (30.5%)	1458 (29.9%)	6464 (37.4%)
High	2825 (49.9%)	3049 (62.4%)	8416 (48.7%)
Food group frequency score (FGFS) past 7 days
Low	702 (13.1%)	382 (7.9%)	2434 (14.1%)
Medium	1643 (30.7%)	1567 (32.2%)	6622 (38.4%)
High	3006 (56.2%)	2915 (59.9%)	8179 (47.5%)
Meal frequency score (MFS) past 24 h
Low	81 (1.5%)	68 (1.4%)	7793 (45.4%)
Medium	3085 (58.2%)	3193 (66 %)	8568 (50 %)
High	2133 (40.3%)	1580 (32.6%)	792 (4.6%)
ICFI
Low	1949 (34.4%)	1932 (40%)	5901 (34%)
Medium	1839 (32.5%)	1883 (39%)	6043 (35%)
High	1878 (33.2%)	1070 (22%)	5327 (31%)

Relation of child feeding practices to nutritional status

In this study, infants were followed from 6 months age to 24 months age; median duration of follow‐up was 16.2 months (IQR = 10.4–17.5) and the mean inter‐visit interval was 1.1 months ± 0.6 The incidence rates for stunting, underweight and wasting were 27.1, 21.6 and 25.4 per 100 person‐years, respectively. In bivariate analysis (Table 4), compared with a low ICFI score, a high score was associated with 34% reduced risk of stunting (high vs. low tertile HR: 0.66; 95% CI: 0.53, 0.83; P < 0.01) while a medium score was associated with 26% reduced risk of stunting (medium vs. low tertile HR: 0.74; 95% CI: 0.60, 0.91; P < 0.01). In multivariate analysis, children with a high ICFI score had 28% reduced risk of stunting compared with those with a low score (high vs. low tertile HR: 0.72; 95% CI: 0.57, 0.91; P < 0.01) while those with a medium ICFI score had 22% reduced risk of stunting (medium vs. low tertile HR: 0.78; 95% CI: 0.63, 0.96; P < 0.01). High and medium ICFI scores were protective of underweight even after adjusting for potential confounders (medium vs. low tertile HR: 0.76; 95% CI: 0.59, 0.96; P = 0.02, high vs. low tertile HR: 0.79; 95% CI: 0.61, 1.02; P = 0.07). The ICFI was not associated with wasting.

Table 4.

Prospective association of the ICFI with subsequent risk of first episode of growth faltering among children during 6–24 months of life, Dar es Salaam, Tanzania

Outcome	Low ICFI	Medium ICFI				High ICFI
		Unadjusted HR (95% CI)	P *	Adjusted HR2 (95% CI)	P *	Unadjusted HR (95% CI)	P *	Adjusted HR2 (95% CI)	P *
Stunting (482 events)	Reference	0.74 (0.60, 0.91)	<0.01	0.78 (0.63, 0.96)	0.02	0.66 (0.53, 0.83)	<0.01	0.72 (0.57, 0.91)	<0.01
Underweight (386 events)	Reference	0.71 (0.56, 0.90)	0.01	0.76 (0.59, 0.96)	0.02	0.72 (0.56, 0.92)	0.01	0.79 (0.61, 1.02)	0.07
Wasting (456 events)	Reference	0.85 (0.68, 1.07)	0.16	0.89 (0.71, 1.12)	0.33	0.92 (0.73, 1.15)	0.46	1.01 (0.80, 1.27)	0.96

Adjusted for maternal: age, education level, partner's education level, food expenditure per person per day, Filmer–Pritchett wealth score, ARV use in pregnancy, WHO HIV disease stage at baseline and child's: sex, birth weight, EBF status, EBF duration and multivitamin supplementation; and the following time‐varying covariates: maternal BMI, maternal haemoglobin level, child's haemoglobin level, child's HIV infection status, child's CD4 % and acute illness (acute diarrheal disease and/or acute respiratory infection) in the child.

^* Hazard ratios (HR), 95% confidence intervals and corresponding P values were obtained from Cox proportional hazards models.

Micronutrient supplementation was not an effect modifier for the association between ICFI and undernutrition. HIV infection status was an effect modifier for the association between the ICFI and underweight (P = 0.01). Compared with low ICFI scores, high scores appeared to be more protective of underweight among HIV‐infected children (high vs. low tertile HR: 0.54; 95% CI: 0.27, 1.09; P = 0.08) than among HIV‐uninfected children (high vs. low tertile HR: 0.83; 95% CI: 0.63, 1.08; P = 0.16).

Breastfeeding at six months age and beyond was associated with stunting, underweight and wasting in bivariate analysis (P < 0.01) (Table 5). However, the significant association persisted only for wasting in multivariate analysis whereby children that were not currently being breastfed had reduced risk of wasting compared with their breastfed peers (HR: 0.66; 95% CI: 0.46, 0.95; P = 0.02). The association between breastfeeding and wasting did not differ by HIV infection status (P = 0.24).

Table 5.

Prospective association of the ICFI components with subsequent risk of first episode of growth faltering among children during 6–24 months of life, Dar es Salaam, Tanzania

ICFI components*	Stunting (482 events)				Underweight (386 events)				Wasting (456 events)
	Unadjusted HR (95% CI)	P‐value*	Adjusted HR2 (95% CI)	P‐value*	Unadjusted HR (95% CI)	P‐value*	Adjusted HR2 (95% CI)	P‐Value*	Unadjusted HR (95% CI)	P‐value*	Adjusted HR2 (95% CI)	P‐value*
Current breast feeding		<0.01		0.13		<0.01		0.48		<0.01		0.02
No	0.49 (0.34, 0.72)		0.73 (0.49, 1.09)		0.54 (0.36, 0.81)		0.86 (0.56, 1.32)		0.46 (0.33, 0.65)		0.66 (0.46, 0.95)
Yes	Reference		Reference		Reference		Reference		Reference		Reference
Food consistency		0.02		0.08		0.01		0.05		0.61		0.95
Low	1.45 (1.06, 1.98)		1.32 (0.96, 1.82)		1.53 (1.09, 2.14)		1.41 (1.00, 1.98)		1.08 (0.81, 1.44)		1.01 (0.75, 1.36)
High	Reference		Reference		Reference		Reference		Reference		Reference
Dietary diversity score (DDS) past 24 h
Low	1.93 (1.50, 2.49)	<0.01	1.59 (1.23, 2.07)	<0.01	1.83 (1.40, 2.41)	<0.01	1.48 (1.12, 1.96)	0.01	1.53 (1.19, 1.98)	<0.01	1.26 (0.97, 1.65)	0.08
Medium	1.45 (1.19, 1.76)	<0.01	1.33 (1.09, 1.63)	<0.01	1.15 (0.92, 1.45)	0.23	1.05 (0.83, 1.33)	0.67	0.95 (0.77, 1.18)	0.65	0.89 (0.72, 1.10)	0.28
High	Reference		Reference		Reference		Reference		Reference		Reference
Food group frequency score (FGFS) past 7 days
Low	2.07 (1.59, 2.68)	<0.01	1.70 (1.30, 2.22)	<0.01	1.88 (1.41, 2.50)	<0.01	1.54 (1.15, 2.07)	<0.01	1.49 (1.13, 1.97)	<0.01	1.25 (0.94, 1.66)	0.12
Medium	1.47 (1.21, 1.80)	<0.01	1.35 (1.11, 1.66)	<0.01	1.19 (0.95, 1.50)	0.14	1.10 (0.87, 1.39)	0.41	1.02 (0.82, 1.26)	0.88	0.96 (0.77, 1.19)	0.70
High	Reference		Reference		Reference		Reference		Reference		Reference
Meal frequency score (MFS) past 24 h
Low	Reference		Reference		Reference		Reference		Reference		Reference
Medium	1.04 (0.81, 1.33)	0.77	0.95 (0.74, 1.22)	0.69	1.07 (0.80, 1.43)	0.66	1.00 (0.74, 1.34)	0.99	1.32 (0.96, 1.79)	0.08	1.26 (0.92, 1.72)	0.15
High	1.23 (0.89, 1.70)	0.22	1.08 (0.78, 1.51)	0.64	1.30 (0.90, 1.88)	0.16	1.16 (0.79, 1.68)	0.45	1.61 (1.12, 2.32)	0.01	1.48 (1.02, 2.15)	0.04

^* Hazard ratios (HR), 95% confidence intervals and corresponding P values were obtained from Cox proportional hazards models.

In bivariate analysis, a low food consistency score was associated with increased risk of stunting (low vs. high tertile HR: 1.45; 95% CI: 1.06, 1.98; P = 0.02) and underweight (low vs. high tertile HR: 1.53; 95% CI: 1.09, 2.14; P = 0.01) but not with wasting. In multivariate analysis, low food consistency was associated with increased risk of underweight (low vs. high tertile HR: 1.40; 95% CI: 1.00, 1.98; P = 0.05).

Low DDS were associated with significantly higher risk of stunting (low vs. high tertile HR: 1.93; 95% CI: 1.50, 2.49; P < 0.01), underweight (low vs. high tertile HR: 1.83; 95% CI: 1.40, 2.41; P < 0.01) and wasting (low vs. high tertile HR: 1.53; 95% CI: 1.19, 1.98; P < 0.01) in bivariate analysis. The significant associations persisted for stunting (low vs. high tertile HR: 1.59; 95% CI: 1.23, 2.07; P < 0.01) and underweight (low vs. high tertile HR: 1.48; 95% CI: 1.12, 1.96; P = 0.01) in multivariate analysis. Similarly, low FGFS were associated with higher risk of stunting (low vs. high tertile HR: 1.70; 95% CI: 1.30, 2.22; P < 0.01) and underweight (low vs. high tertile HR: 1.54; 95% CI: 1.15, 2.07; P < 0.01) but not with wasting, after adjusting for potential confounders.

In bivariate analysis, high MFS were associated with higher risk of wasting (high vs. low tertile HR: 1.61; 95% CI: 1.12, 2.32; P = 0.01). The association was attenuated on adjusting for potential confounders (high vs. low tertile HR: 1.48; 95% CI: 1.02, 2.15; P = 0.04). MFS was not associated with stunting or underweight. The association between MFS and wasting did not differ by HIV infection status (P = 0.64).

Using stepwise regression we determined that the FGFS drove the association between the ICFI and stunting and underweight (p < 0.01), and that breastfeeding drove the association between the ICFI and wasting (p < 0.01). In this cohort, 70% of children were followed to growth endpoint, death or 24 months age. Not breastfeeding at six months age and beyond was protective of loss to follow‐up (P = 0.01) while low DDS were associated with loss to follow‐up (P = 0.03) (Supplementary table).

Discussion

This study employed a child feeding index to assess nutritional adequacy of complementary feeding and its relation with nutritional morbidity indices among HIV‐exposed children participating in a randomized controlled trial of micronutrients in an urban setting in sub‐Saharan Africa. Overall, complementary feeding practices of the children in this study were suboptimal and the incidence rates of stunting, underweight and wasting were high. Medium and high ICFI scores were protective of stunting and underweight. Micronutrient supplementation did not modify the association between complementary feeding and nutritional status in this cohort of children. Among its components, high 24‐h and 7‐day DDSs were significantly protective of stunting and underweight while breastfeeding at six months age and beyond and high MFSs were associated with wasting.

Our findings that high ICFI scores were protective of stunting and underweight in this population are in agreement with those of other studies based on national or multinational samples which reported positive associations between the ICFI and length for age Z scores (LAZ) and weight for age Z scores (WAZ) despite slight differences in the indices and age ranges used. Of these studies, three were based on HIV exposed infants data. Investigators in Southern Ethiopia (Haile et al. 2014a; Haile et al. 2014b) used both a cross‐sectional and longitudinal ICFI and found significant protective associations between both measures of the ICFI with stunting and underweight but not with wasting. Another prospective study among a cohort of HIV exposed infants in an urban setting in Cote d'Ivoire (Becquet et al. 2006) found a significant association between a low ICFI at 6 months and increased risk of stunting occurrence in the subsequent 12 months. Our findings are also similar to those from two longitudinal studies among HIV‐unexposed children in urban China and Madagascar which found that ICFI was significantly protective of stunting (Moursi et al. 2008; Ma et al. 2012) and underweight (Ma et al. 2012). However our results differ from those of studies in rural China and Senegal (Zhang et al. 2009; Ntab et al. 2005) which did not find an association between the ICFI and stunting. This discrepancy may be because of the different feeding practices in these settings as well as small sample sizes (500 children) used in these studies. Additionally, the study participants in China were aged 6–11 months; it has previously been shown that the association between feeding practices and stunting is generally weaker and less consistent in the first year of life and increases gradually with age because of the cumulative effects of previous feeding practices (Ruel & Menon 2002).

Our finding of a significant protective association between the ICFI and stunting and underweight but not with wasting may suggest that the ICFI is an indicator of long term feeding practices and thus illustrates the strengths of associations between complementary feeding practices and long‐term infant nutritional outcomes. Additionally, wasting was associated with breastfeeding and high meal frequency, two components that constitute the ICFI; this may partly explain the absence of a protective association between the ICFI and wasting. Our findings are consistent with those from studies in India, Madagascar and Ethiopia which reported absence of an association between the ICFI and wasting (Garg & Chadha 2009; Khatoon et al. 2011; Ma et al. 2012; Haile et al. 2014b).

Breastfeeding at six months age and beyond was associated with wasting even after adjusting for potential confounders. This association differed by age; not breastfeeding became more protective of wasting with increasing age. Children who were breastfed at six months and beyond may have been more likely to receive complementary foods of lower dietary diversity compared with non‐breastfed children. Mutually adjusting for dietary diversity attenuated the association between breastfeeding and wasting. Thus, low dietary diversity may have contributed to the association between breastfeeding and undernutrition in this cohort. Conversely, mutually adjusting for formula milk intake did not affect the association between breastfeeding and wasting. Thus, the lower risk for undernutrition among non‐breastfed infants was not likely explained by formula milk intake. Additionally, the WHO recommendations at the time of the study were that HIV exposed children be exclusively breastfed ‘for the first months of life’ followed by replacement feeding as soon as it was feasible, taking into account local circumstances and the individual woman's situation (World Health Organization 2001). Mothers who were unable to provide adequate replacement foods to their children may have continued breastfeeding as the main or only source of food after six months age resulting in undernutrition. Several studies have found negative associations between breastfeeding and LAZ scores among children aged between 12 and 36 months (Ntab et al. 2005; Sawadogo et al. 2006; Moursi et al. 2009; Bork et al. 2012). The authors hypothesized that these associations may have been because of reverse causality as it has previously been demonstrated that mothers may prolong breastfeeding for malnourished children (Simondon & Simondon 1998).

Our results suggest that complementary foods of thin consistency may increase young children's risk of stunting and underweight. This may be because complementary foods of thin consistency may have low energy density and because young children have small gastric capacities, it would be difficult for them to consume enough of these foods to meet their energy requirements. Similar findings were revealed in a study in urban China which found a positive association between consistency of complementary food at 6 months with LAZ and WAZ scores at 18 months (Ma et al. 2012). Our results may have failed to reach statistical significance because of majority of the children in our study being fed solid complementary foods, resulting in small sample sizes in the low food consistency category.

We measured dietary diversity using both 24‐h and 7‐day recall to include possible diet variations from one day to another. Stunting and underweight were negatively associated with dietary diversity, after adjusting for potential confounders. This suggests that children on more diverse diets may be more likely to meet their energy and nutrient requirements and thus have better nutritional outcomes. Several studies have shown that improved dietary diversity is important in improving children's nutritional status (Onyango et al. 1998; Tarini et al. 1999; Arimond & Ruel 2004; Sawadogo et al. 2006; Moursi et al. 2009; Bork et al. 2012; Ma et al. 2012; Darapheak et al. 2013).

We found an association between high meal frequency and wasting; children who were fed more meals in the preceding 24 h had a higher risk of wasting than those fed fewer meals. This association differed by age group; the risk of wasting decreased significantly with increasing age. Our findings differ from those of a study in rural China which found a significant positive association between meal frequency and WAZ and weight for length Z (WLZ) scores (Zhang et al. 2009). A possible reason for high MFS to be associated with wasting may be because of children being fed often but with inadequate amounts of food. Further analysis revealed that children with high MFS may have been more likely to receive complementary foods of thinner consistency. However, mutually adjusting for food consistency did not affect the association between MFS and wasting. Thus, it is unlikely that food consistency contributed significantly to this association. A cross‐sectional study in Burkina Faso observed lower mean HAZ scores in the class of children consuming a higher number of meals (Sawadogo et al. 2006) for which the authors hypothesized may be because of reverse causality.

The major strength of this study was the prospective nature of data collection which allows for a proper temporal relationship to be assessed between complementary feeding practices and nutritional status. Second, using 24‐h and 7‐day recalls of infant feeding may have reduced recall bias in the study as mothers were more likely to remember what their children had eaten in the previous week. Third, using a composite index to measure feeding practices enabled us to examine the cumulative effect of multiple feeding practices on child nutritional status. Furthermore, using the ICFI's components allowed us to expose associations between specific child feeding practices and the nutritional indices. Finally, controlling for ample confounders in multivariate regression analysis reduced bias in our results. Nonetheless, the study had some limitations. First, the ICFI was based on four food groups because we lacked 24‐h and 7‐day feeding data for legumes and nuts, eggs and Vitamin A‐rich fruits and vegetables; three food groups which the WHO includes in tabulating the dietary diversity indicator (World Health Organization 2005). Consequently, we may have underestimated the DDS and ICFI scores and their associations with the nutritional indices because misclassification was likely to be non‐differential in a prospective study like this. The loss of potentially useful information that arose from categorizing the ICFI and its components may have reduced this study's power to detect associations, leading to imprecise estimation of the risks. Because there was differential loss to follow‐up by DDS and breastfeeding at six months age and beyond, observed associations may be underestimated. Furthermore, clinic‐based reporting of nutrition behaviours may have led to under‐reporting of poor infant feeding behaviours and in turn led to exposure misclassification. However, because this was a prospective cohort study, exposure misclassification would likely be non‐differential. Finally, our results may not be generalizable to children born to mothers with a later HIV disease stage because only 12% of mothers in the present study were in the WHO HIV stage III or higher at enrolment.

In conclusion, we found that high DDS and ICFI scores were protective of stunting and underweight among children born to early‐stage HIV‐infected women in Tanzania who participated in a randomized controlled trial of micronutrients. Our results confirm existing literature that improved dietary diversity and optimal complementary feeding are important in improving young children's nutritional status. While the ICFI may not be a perfect tool for measuring complementary feeding, our findings shed some light on the cumulative effect of young child feeding practices on nutrition outcomes in an important subpopulation in sub‐Saharan Africa. We recommend for nutrition programmes in resource limited settings to emphasize educating HIV exposed children's caregivers on improving quality of complementary food, example, in terms of increasing dietary diversity, to facilitate optimal feeding and therefore improve nutritional status in this important subpopulation.

Source of Funding

This study was supported by Eunice Kennedy Shriver National Institute of Child Health and Human Development (NICHD R01 HD043688‐01 and K24DK104676). PK received support from the National Institute of Health's (NIH) Fogarty International Center programme 5D43 TW007886‐07 NIH/FIC.

Conflict of interest

The authors declare that they have no conflict of interest.

Contributions

PK: analysis plan, data analysis, interpretation of analysis results, and writing this article; CD, KPM, DS, WWF (Principal Investigator): contributed to the study design of the original trial of multivitamins study, the present study being a secondary analysis of data collected in the trial of multivitamins; EL, EH DS, and WWF: provided statistical guidance in data analysis; RK, RRK: study design, data collection and daily management of the field study; SA: contributed to the study implementation in the field, in the laboratory.

Supporting information

Supplementary Table: Comparison of the ICFI construction and findings across studies

Supplementary Table: Univariate and multivariate regression associations of the ICFI and its components with loss to follow‐up

Across Studies Infant Feeding Practices Form

Supporting info item

Supporting info item

Supporting info item

Kamenju, P. , Liu, E. , Hertzmark, E. , Spiegelman, D. , Kisenge, R. , Kupka, R. , Aboud, S. , Manji, K. P. , Duggan, C. , and Fawzi, W. W. (2017) Nutritional status and complementary feeding among HIV‐exposed infants: a prospective cohort study. Maternal & Child Nutrition, 13: e12358. doi: 10.1111/mcn.12358.