Complementary feeding with cowpea reduces growth faltering in rural Malawian infants: a blind, randomized controlled clinical trial

ABSTRACT

Background: Growth faltering is common in rural African children and is attributed to inadequate dietary intake and environmental enteric dysfunction (EED).

Objective: We tested the hypothesis that complementary feeding with cowpea or common bean flour would reduce growth faltering and EED in 6-mo-old rural Malawians compared with the control group receiving a corn-soy blend.

Design: A prospective, double-blind, randomized controlled clinical trial was conducted in which children received daily feeding for 6 mo (200 kcal/d when 6–9 mo old and 300 kcal/d when 10–12 mo old). The primary outcomes were change in length-for-age z score (LAZ) and improvements in EED, as measured by percentage of lactulose excretion (%L). %L <0.2% was considered normal. Anthropometric measurements and %L through urine were compared between each legume group and the control group with Student's t test.

Results: Of the 355 infants enrolled, 291 infants completed the trial, and 288 were breastfed throughout the duration of the study. Cowpea and common bean added 4.6–5.2 g protein/d and 4–5 g indigestible carbohydrate/d to the diet. LAZ and weight-for-height z score were reduced in all 3 groups from 6 to 12 mo of age. The changes in LAZ [mean (95% CI)] for the cowpea, common bean, and control groups from 6 to 9 mo were −0.14 (−0.24, −0.04), −0.27 (−0.38, −0.16), and −0.27 (−0.35, −0.19), respectively. LAZ was reduced less in infants receiving cowpea than in those receiving control food from 6 to 9 mo (P = 0.048). The absolute value of %L did not differ between the dietary groups at 9 mo of age (mean ± SD: 0.30 ± 0.43, 0.23 ± 0.21, and 0.26 ± 0.31 for cowpea, common bean, and control, respectively), nor did the change in %L from 6 to 9 mo.

Conclusion: Addition of cowpea to complementary feeding in Malawian infants resulted in less linear growth faltering. This trial was registered at clinicaltrials.gov as NCT02472262.

INTRODUCTION

Stunting occurs in 35% of rural sub-Saharan African children, and its deleterious consequences are pervasive over an individual's lifetime. Stunting decreases life expectancy by 17% and lifetime income earned by 22%; it also reduces one's capacity for physical labor and diminishes one's cognitive potential (1). Stunting results from a combination of insults, most prominently inadequate dietary intake, repeated exposure to infectious agents, and subclinical chronic inflammation of the small bowel, also known as environmental enteric dysfunction (EED) (2). Most damage occurs during the ∼1000 d of development from conception through 2 y of age. In African children, growth falters most often during the ages of 6–15 mo, when complementary foods are first introduced and the child is exposed to a large variety of microbes in the maturing gastrointestinal tract. No known highly effective interventions are available to reduce faltering growth in the context of rural Africa (3).

Insufficient dietary protein intake and consumption of foods with low protein quality is associated with EED and stunting (4). The gut microbiota are an important determinant of gut health and are believed to play a role in EED and be associated with patterns of linear growth (5). Legumes, particularly cowpeas and common beans, are indigenous crops in sub-Saharan Africa and contain larger quantities of protein than cereals. These legumes also contain more indigestible complex carbohydrates than cereals; these carbohydrates have been associated with improved gut health in clinical scenarios and in animal and in vitro models (6–8).

EED is common in impoverished settings in the developing world, characterized by T-cell infiltration and villous blunting in the small-intestine mucosa and associated with reduced macronutrient and micronutrient absorption and increased intestinal permeability (9). The most commonly used noninvasive test for EED is the dual-sugar absorption test with lactulose and mannitol, in which known amounts of these 2 nonmetabolized sugars are administered orally and subsequently quantified in a timed urine collection to assess small-bowel barrier function and absorptive capacity (10).

We hypothesized that in this clinical trial, complementary feeding with either cowpea or common bean in infants 6–12 mo old would improve linear growth and reduce EED as measured by the lactulose-mannitol test (L:M test).

METHODS

Subjects

The trial was conducted in 8 villages surrounding Limera in the Machinga District and 9 villages surrounding Masenjere in the Nsjane District of Southern Malawi between July 2015 and October 2016. Typical homes are constructed from mud and thatch, and residents have little access to electricity. Water is obtained multiple times a day from boreholes or rivers. Subsistence farming is practiced by the majority of households and the staple crop is corn, which is harvested annually after the rainy season. Corn porridge is consumed universally as the primary complementary food in the first year of life, as infants wean from exclusive breastfeeding. All children aged 6 mo and living within walking distance of one of the village clusters were recruited. Exclusion criteria included severe or moderate acute malnutrition, receiving supplemental food from another intervention program, and having a chronic noninfectious disease or a congenital abnormality.

Ethical approval was obtained from the University of Malawi College of Medicine Research and Ethics Committee and the Human Research Protection Office at Washington University in St. Louis. Letters of support were obtained from the district health officers of the Machinga and Nsanje Districts. Before the start of the study, local chiefs of the villages surrounding the 2 study sites gave their consent for the study. The study was explained to each child's caregiver by a native Malawian Chichewa-speaking nurse, and verbal and written consent were obtained. Caregivers who could not sign their names documented consent with a thumbprint.

Study design

This was a prospective, double-blind, randomized controlled trial in which the intervention groups received daily complementary food composed of either cowpeas (Vigna unguiculata) or common beans (Phaseolus vulgaris); the control group received a corn-soy blend flour throughout the 24 wk of study. The study protocol was previously published (11). The trial is registered at clinicaltrials.gov as NCT02472262.

The primary outcomes were change in length-for-age z score (LAZ) and improvement in percentage of lactulose excretion (%L), a component of the L:M test. A total of 79 children were needed in each group to detect a difference in change in length of 1.1 cm (assuming an SD of 2.5 cm)—which corresponds to a change in LAZ of 0.45 at 12 mo of age—between a single randomly assigned food group and the control group. This sample size was calculated with a 2-tailed test through the use of G*Power 3.1.9.2 software (12) at a significance level of 0.05 and 80% power. To detect an effect size of 0.2%L after the legume intervention, again with 80% power and a significance level of 0.05, 64 children would be needed in each group. Thus, we aimed to enroll 300 children, assuming a 15–20% rate of failure; this would yield complete analysis for 240 children, or 80 in each group.

Study participants were randomly assigned through computer-generated block randomization into permuted blocks of 9. Intervention flours were packaged in identical jars with colored lids, which were coded to the 3 study arms. All research nurses, investigators, and laboratory personnel were blinded to the color code. While the color code was not revealed to the caregivers of the study participants, the 3 flours differed slightly in appearance, taste, and smell.

Participation

Children who were eligible for enrollment in the study were scheduled for the initial study visit at the age of 5.5–6.5 mo. Weight when nude was measured to the nearest 5 g; length and midupper arm circumference were measured to the nearest 1 mm. Caretakers of eligible subjects gave informed consent for study participation, and subjects completed the L:M test if they had had no diarrhea over the past 7 d. Baseline sociodemographic and health information was collected. Subjects were given a 6-wk supply of flour. Caregivers were educated by study nurses on how to prepare and feed the flour to their infants.

Additional study visits occurred in Limera or Masenjere when participants were about 6.5, 7.5, 9, 10.5, and 12 mo of age. Anthropometric measurements were taken, a symptom and compliance survey was completed, and stool was collected at every visit; the L:M test was conducted at 9 and 12 mo of age (Figure 1). Study nurses distributed intervention flour every 6 wk. Children who were found to have severe or moderate acute malnutrition during the course of the study were removed from the study and given ready-to-use therapeutic food or ready-to-use supplementary food to treat their malnutrition. The foods used to treat malnutrition were energy-dense peanut butter–based pastes with large amounts of milk powder—substantially different from the habitual diet.

FIGURE 1.

Consolidated Standards of Reporting Trials diagram of study participation. DD, developmental delay; MAM, moderate acute malnutrition; SAM, severe acute malnutrition.

Compliance was assessed through random home visits. Village health aides visited the home of each study participant at least twice during the study and recorded the lid color of the flour container in the home and the amount of flour remaining. Participants with little or no remaining flour were resupplied and educated about the proper use of the flour. These participants were visited again and excluded from the study if they were found to be noncompliant on >2 occasions.

Study foods

Cowpeas and common beans were purchased at local markets, roasted to an internal temperature of 120–130°C for 45 min, and milled into flour. The control group received a traditional complementary food made of extrusion cooked and milled corn and soybeans. All groups received from the supplements 200 kcal/d from 6 to 9 mo of age and 300 kcal/d from 9 to 12 mo of age, which is 40% of the recommended complementary food intake (13). Study nurses instructed caregivers to add either the legume or the control flour to the child's serving of porridge once each day, although several children in the study preferred to eat the legume flour plain.

Two 24-h dietary recalls were conducted among random subsets of 50 subjects in October 2015 and February 2016—times of relative food abundance and scarcity, respectively—to assess dietary intake. Caregivers were asked to recall the amounts of different foods eaten by the child over the past 24 h using food models (14). Nutrient intake was estimated through the use of this information.

L:M testing

L:M testing followed a rigorous protocol (15). Subjects were asked to take no food (water and breast milk were permitted) after 2200 the preceding day. On the day of the dual-sugar absorption test, children were given 20 mL of a solution containing 1 g mannitol and 5 g lactulose under close supervision to ensure complete ingestion. Children who coughed up, vomited, or spilled the solution were asked to return the next day. Adhesive urine bags were attached to the perineum, and urine output was monitored closely. Each time urine was noted in the bag, the bag was removed, a new bag was applied, and urine was deposited into a collection cup, unique to each child; the cups contained 10 mg merthiolate to prevent bacterial growth. Urine was collected until urine was passed ≥4 h after ingestion of the sugar solution. At the end of urine collection, the total volume of urine collected from each child was measured, 4 aliquots were transferred to cryovials, and samples were flash frozen in liquid nitrogen. Samples were stored at the University of Malawi at −80°C until analysis.

HPLC was used to quantify urine lactulose and mannitol as described previously by Trehan et al. (16). The assays used have been shown to be sensitive to the nearest 1 μg/mL for lactulose and mannitol. Assays were run on groups of 24 samples, with 3 duplicates and 3 reference samples. The CVs between duplicates were always <10%, and the average CV was <5%. The HPLC method was previously validated, with liquid chromatography–mass spectrometry showing excellent concordance between the analytic methods (r² = 0.85) (17). Based on previous studies in rural Malawian children, we designated children with %L <0.2% as not having EED, those with %L between 0.2% and 0.45% as having moderate EED, and children with %L >0.45% as having severe EED (18, 19).

Statistical analyses

Summary statistics were tabulated before the color code was shared. Anthropometric indexes were calculated based on the WHO's 2006 Child Growth Standards through the use of Anthro version 3.1 (WHO). Data from the 24-h dietary recalls were combined and analyzed with STATA version 12.1 (StataCorp). Urinary excretion of lactulose and mannitol was calculated as a percentage of the dose administered. The lactulose:mannitol ratio (L:M ratio) was calculated by dividing the concentration of lactulose by that of mannitol in each subject's aggregate urine specimen.

Data from each legume group and the control group were compared. Continuous outcomes measured at a single point were compared through the use of the Student t test; categorical outcomes were compared with the Fisher exact test because the subjects were randomly assigned by individual. (SPSS24; SPSS Inc.). All data with regard to growth, gut health, and survival that were collected or were available from the enrolled children were included in the statistical comparisons and in the tabular aggregate data. These included the anthropometric indexes.

To assess for adverse effects and amelioration of acute infectious symptoms, the presence of diarrhea, vomiting, or fever in the preceding 2 wk was surveyed at each study visit. The duration of diarrhea was also queried, and the percentage of possible days of diarrhea was calculated for each intervention group.

RESULTS

A total of 355 children were randomly assigned to the 3 study arms (Figure 1). Among the 64 children removed from the study after randomization, 42 were removed because they developed acute malnutrition and 15 were lost to follow-up (Figure 1). Among those who developed acute malnutrition, 13 of 117 were randomly assigned to the cowpea group; 17 of 120, to the common bean group; and 12 of 118, to the control group (P = 0.64). Baseline anthropometric and demographic information was similar between the 3 groups (Table 1). Households habitually consumed common beans <1 time/mo, on average.

TABLE 1.

Characteristics of Malawian children receiving cowpea, common bean, or control complementary foods¹

	Cowpea(n = 117)	Common bean(n = 120)	Control group(n = 118)
Girls, n	59 (50)	60 (50)	53 (45)
Age, mo	5.8 ± 0.3	5.8 ± 0.3	5.8 ± 0.4
Father lives in the home, n	97 (83)	106 (88)	97 (82)
Siblings, n	2.2 ± 1.9	2.6 ± 1.9	2.4 ± 1.9
Clean water source is available, n	88 (75)	86 (72)	84 (71)
Animals sleep with the child, n	35 (30)	43 (36)	34 (29)
Mother washes hands often when preparing food, n	70 (60)	71 (59)	59 (50)
Midupper arm circumference, cm	14.0 ± 0.9	13.9 ± 0.8	14.0 ± 1.0
z Scores
Weight for length	0.4 ± 0.9	0.2 ± 0.9	0.4 ± 0.9
Length for age	−1.3 ± 1.2	−1.1 ± 1.0	−1.2 ± 1.1
Weight for age	−0.7 ± 0.7	−0.6 ± 0.8	−0.6 ± 0.9
Diarrhea in past week, n	14 (12)	16 (13)	15 (13)
Fever in past week, n	43 (37)	35 (29)	35 (30)
Radio in home, n	42 (36)	36 (30)	40 (34)
Bicycle in home, n	58 (50)	56 (47)	61 (52)
Roof made of metal, n	30 (26)	31 (26)	20 (17)
Breastfed, n	112 (99)	110 (99)	106 (100)
Household consumption, times/mo
Common beans	0.9 ± 1.4	1.2 ± 2.9	0.9 ± 1.4
Cowpeas	0.6 ± 1.2	0.4 ± 1.2	0.7 ± 2.8

¹Values are means ± SDs or n (%).

Compliance was assessed with 663 home visits and surveys. No flour remained in the home at 5% of these visits, and the amount of flour remaining was, on average, 90% of what was expected based on correct use. On the compliance questionnaires completed at each follow-up visit, at 6% of visits caregivers reported missing ≥1 d of flour in the past week. The most common reasons included that the child had decreased appetite because of illness, the mother was away, or no flour remained.

Standard laboratory assessment of nutrients in legume flours used in the study found that cowpea contained 26% protein and 21% fiber (1% soluble), common bean contained 23% protein and 28% fiber (0% soluble), and the control flour contained 13% protein and 8% fiber. All of the flours had a similar energy content of ∼3.8 kcal/g.

All of the infants were breastfed ad libitum through 10.5 mo of age, and only 3 of 291 (1%) stopped breastfeeding by 12 mo; thus breast milk provided the major source of protein intake among the study population. Protein from complementary foods came primarily from corn, sweet potatoes, and pumpkin leaves. Only 7% of the dietary protein from complementary food was from animal sources. From the 24-h dietary recalls, nutrient intakes from complementary food, including dietary protein intake, seemed to be adequate in all dietary groups (Supplemental Table 1).

From 6 to 12 mo of age, LAZ and weight-for-length decreased in the study children (Figure 2,Supplemental Table 2). The changes in LAZ [mean (95% CI)] from 6 to 9 mo for children in the cowpea, common bean, and control groups were −0.14 (−0.24, −0.04), −0.27 (−0.38, −0.16), and −0.27 (−0.35, −0.19), respectively. Children receiving cowpea experienced less decrease in LAZ than the control group (P = 0.048). Changes in LAZ from 9 to 12 mo were similar in all food groups. At the age of 12 mo, the fraction of children in each dietary group experienced stunted growth: 35 of 99 receiving cowpea, 26 of 99 receiving common bean, and 25 of 93 receiving the control food (P = 0.30).

FIGURE 2.

Growth of study infants receiving complementary legumes. (A) Change in LAZ from 6 to 12 mo while receiving cowpea, common bean, or control complementary food daily. Infants receiving cowpea had a lesser decrease in LAZ than did those receiving the control cereal (P = 0.048, Student t test). The error bars around the cowpea points represent the 95% CIs. (B) All infants had linear faltering growth, as represented by progressive decreases in LAZ, from 6 to 12 mo of age. Differences between food groups were not significant. No error bars are shown because the measure of growth, a z score unit, was defined as 1 SD. (C) Weight-for-length z score decreased from 6 to 12 mo in healthy Malawian infants, concomitant with their linear growth. LAZ, length-for-age z score.

Of the 826 L:M tests conducted, 431 (53%) had %L <0.2%, representing those without EED, and 104 of 826 (13%) had %L >0.45%, representing severe EED. From 6 to 12 mo of age, children's %L and L:M ratio did not appreciably change (Figure 3,Supplemental Table 3). No differences in %L or L:M ratio were seen with the type of complementary feeding received. Among the 47 children with severe EED at 6 mo of age, 17 (36%) improved to no EED at 9 mo of age; 6 of 18 received cowpea, 4 of 11 received common bean, and 7 of 18 received control food (P = 0.94).

FIGURE 3.

Intestinal permeability in children receiving legume complementary foods. (A) The L:M ratio (A), %L (B), and %M (C) among children receiving 1 of 3 complementary feeding regimens: cowpea, common bean, or control. Bars indicate the median (line) and the 25th and 75th percentiles (bars above and below the lines). No statistically significant differences were seen between dietary groups at any given time point, nor for children receiving the same complementary food as they aged. L:M, lactulose:mannitol ratio; %L, percentage of lactulose excreted; %M, percentage of mannitol excreted.

Mothers reported episodes of fever in the preceding 2 wk for 20–32% of all study children from 6 to 12 mo, and diarrhea in 13–23% of all study children. No differences in diarrhea, fever, or vomiting were seen between the intervention and control groups (Supplemental Table 4).

DISCUSSION

In this carefully conducted randomized trial of either cowpea or common bean as a component of complementary feeding in rural African infants, cowpea reduced linear growth faltering. The effect was modest, estimated at +0.13 z scores. The benefit is particularly accrued from 6 to 9 mo of age. Neither legume demonstrated an effect on %L or L:M ratio. This is, to our knowledge, the first report of supplementation with a single plant-based food alone in a population at risk for stunting, showing an effect on LAZ.

While compliance was excellent and the population was extensively characterized, the study is limited by a small sample size of 355 children, among whom 42 developed acute malnutrition. The study design used random assignment by individual; thus the quality of the clinical evidence is higher than that provided by epidemiologic studies that associated stunting with the consumption of certain diets or foods. The potential of cowpea as an intervention to reduce stunting therefore warrants further investigation in larger trials. Supplementation with either cowpea or common bean did not reduce the high rate of acute malnutrition seen over the course of the study (>10% of those enrolled).

One challenge with nutritional intervention studies in which subjects may develop acute malnutrition during participation, as in ours, is that traditionally strong intention-to-treat analyses are more difficult to apply (20). In our study we chose to include the growth and gut health data up to the point that acute malnutrition developed, then we removed the child from the study and initiated a therapeutic diet, which supplied ample nutrients for recovery and growth, and collected no further data. Our study would have been strengthened if anthropometric follow-up had continued after a therapeutic diet was initiated, because these data would present a more complete picture of the nutritional outcome of the subject, and thereby of the study population as a whole. Removing subjects from any clinical trial, including this one, introduces uncertainty and possibly bias, as the subjects who are removed are different from the study population as a whole. The preventive dietary interventions may have exerted some effect, positive or negative, after the time when the child was removed, although we would therefore be unable to discern such an effect. This methodologic uncertainty should be considered by nutritionists, clinicians, and clinical trialists as they plan future studies in an effort to reduce ambiguity in data interpretation.

The cowpea and common bean supplements provided 5.2 and 4.6 g additional protein/d, respectively, in 6- to 8-mo-old infants—substantial amounts—and the protein digestibility–corrected amino acid scores of these legumes were about 50%, not of particularly high quality. Protein requirements for these infants are estimated to be 11 g/d, while their habitual breast milk intake is estimated to provide 7 g high-quality protein/d. Other dietary sources of protein elicited by the recalls were present in low quantities and were of low quality: corn containing 10% protein, sweet potatoes containing 1.5% protein, and pumpkin leaves containing 1% protein. The additional protein provided by the legumes might be necessary for better linear growth but might not be sufficient, as linear growth improved only in the children receiving cowpea. The 24-h dietary intake data do not suggest that protein intake was especially low in this study population; however, dietary recall data are limited in that they cannot be applied to individual subjects, just the study population as a whole (14). Furthermore, the 24-h dietary recall methodology has been shown to yield results that vary by 20–50% in the same individual for the same time period (21, 22), making it an imprecise measurement tool, and recall data alone should not be used to dismiss additional dietary protein intake as a strategy to reduce linear faltering growth.

Cowpeas provided considerably more indigestible carbohydrate, or fiber, to the diet than was consumed by the control group. The amount and duration of inflammatory stimuli to which a young child is exposed is inversely related to the child's linear growth (22). The gut is the most intimate interface between the environment and the growing child. The role of cowpea fiber in reducing inflammation mediated by the gut is uncertain. The evidence for legumes exerting an anti-inflammatory effect comes from elegant in vitro studies and epidemiologic associations (23, 24). The anti-inflammatory effect may be exerted through the promotion of protective biofilms that reduce exposure of the mucosa to more pathogenic microbes or that promote larger populations of symbiotic microbes that provoke less inflammation (25–27). The composition and effect of cowpea fiber on gut inflammation may provide insight into the mechanism by which this food reduces faltering growth.

The L:M test was conducted in a regimented, standard manner for all subjects, and our team has used it as a reproducible and practical measure of gut health for >2 decades. Our best measure of intestinal permeability is %L, as normal cell junctions are not permeable to lactulose and %L is inversely associated with linear growth in this population (28). We also report the traditional measure of gut health, L:M ratio (Figure 3). We have no data comparing histologic findings and %L. The %L and L:M ratio are dynamic parameters that can increase and decrease over several days in an individual; thus both the absolute measure of %L and the L:M ratio, and change in %L and the L:M ratio, are of interest. No differences in %L or the L:M ratio were seen in this trial with respect to type of complementary feeding, providing no direct evidence to support reduced gut inflammation as a mechanism by which cowpea improved LAZ.

Figure 2 demonstrates a clear pattern of faltering growth in the study population, with both weight-for-height z scores and LAZs falling from 6 to 12 mo. Yet our “measures” of dietary intake and EED—24-h dietary recall and the L:M test, respectively—did not suggest inadequate intake or increased inflammation. Did an inadequate diet and EED play minor roles in the observed faltering growth? Or does this observation highlight the need for better biomarkers of EED and dietary intake? In the 21st century, -omics technology allows for simultaneous assessment of multiple analytes with a minimum-sized biological sample, and recent reports describe the use of these technologies on novel analytes as tools to define dietary intake and gut health (19, 29). Use of such technology may well allow for more precise and accurate assessment of the causal factors that lead to stunting, and may direct us toward interventions to ameliorate it.

Effective strategies to reduce growth faltering in sub-Saharan Africa are likely to include a combination of interventions that address its multiple causes: suboptimal dietary intake, increased intestinal permeability, and an excess of inflammatory stimuli. Such strategies are desperately needed, as stunting occurs in 35% of these sub-Saharan African children, and the number of children with stunted growth is increasing in this region (30). It was estimated in 2013 that if all effective, scalable interventions to reduce stunting were in place, there would be only a 20% reduction in stunting (3). Our data suggest complementary feeding with cowpeas is a candidate component for such strategies, and further exploration of the utility of cowpea to this end is warranted.

ABBREVIATIONS

EED

environmental enteric dysfunction

L:M ratio

lactulose:mannitol ratio

L:M test

lactulose-mannitol test

LAZ

length-for-age z score

%L

percentage of lactulose excreted