Comparison of the effectiveness of a milk‐free soy‐maize‐sorghum‐based ready‐to‐use therapeutic food to standard ready‐to‐use therapeutic food with 25% milk in nutrition management of severely acutely malnourished Zambian children: an equivalence non‐blinded cluster randomised controlled trial

Abstract

Community‐based Management of Acute Malnutrition using ready‐to‐use therapeutic food (RUTF) has revolutionised the treatment of severe acute malnutrition (SAM). However, 25% milk content in standard peanut‐based RUTF (P‐RUTF) makes it too expensive. The effectiveness of milk‐free RUTF has not been reported hitherto. This non‐blinded, parallel group, cluster randomised, controlled, equivalence trial that compares the effectiveness of a milk‐free soy–maize–sorghum‐based RUTF (SMS‐RUTF) with P‐RUTF in treatment of children with SAM, closes the gap. A statistician randomly assigned health centres (HC) either to the SMS‐RUTF (n = 12; 824 enrolled) or P‐RUTF (n = 12; 1103 enrolled) arms. All SAM children admitted at the participating HCs were enrolled. All the outcomes were measured at individual level. Recovery rate was the primary outcome. The recovery rates for SMS‐RUTF and P‐RUTF were 53.3% and 60.8% for the intention‐to‐treat (ITT) analysis and 77.9% and 81.8% for per protocol (PP) analyses, respectively. The corresponding adjusted risk difference (ARD) and 95% confidence interval, were −7.6% (−14.9, 0.6%) and −3.5% (−9,6., 2.7%) for ITT (P = 0.034) and PP analyses (P = 0.257), respectively. An unanticipated interaction (interaction P < 0.001 for ITT analyses and 0.0683 for PP analyses) between the study arm and age group was observed. The ARDs were −10.0 (−17.7 to −2.3)% for ITT (P = 0.013) and −4.7 (−10.0 to 0.7) for PP (P = 0.083) analyses for the <24 months age group and 2.1 (−10.3,14.6)% for ITT (P = 0.726) and −0.6 (−16.1, 14.5) for PP (P = 0.939) for the ≥24 months age group. In conclusion, the study did not confirm our hypothesis of equivalence between SMS‐RUTF and P‐RUTF in SAM management.

Introduction

Severe acute malnutrition (SAM) affects over 19 million children under the age of 5 and is associated with 1–2 million preventable child deaths each year (Collins et al. 2006a; Black et al. 2008). Over the last decade following the introduction of the Community‐based Management of Acute Malnutrition (CMAM) using ready‐to‐use therapeutic foods (RUTF), there have been major improvements in the survival of children with SAM treated in these outpatient therapeutic programmes (Collins et al. 2006a,b; WHO et al. 2007). To date, the internationally accepted RUTF has been made from peanut paste, dried skimmed milk oil, sugar and mineral vitamin mix (Manary 2006) and the vast majority of this product has been made in France.

The use of this peanut/milk‐based RUTF (P‐RUTF), coupled with increasing access to CMAM services, has been associated with high recovery rates, lower case fatalities and greater weight gain of children with SAM (Ciliberto et al. 2005; Chaiken et al. 2006; Collins et al. 2006a,b; Linneman et al. 2007; Lapidus et al. 2009). In addition, providing RUTF in a dosage tailored to body weight has been shown to increase catch‐up growth in children with SAM (Diop et al. 2003a). Follow‐up studies from Malawi of children discharged as recovered from CMAM programmes have also demonstrated that over 85% of these children maintained a normal weight for height at 6 months (Manary et al. 2004) and 15 months (Bahwere et al. 2008) after discharge.

Despite the effectiveness of P‐RUTF in the treatment of SAM, its cost is high at $3.5–$4.00 kg⁻¹, with dried skimmed milk powder accounting for over 50% of the ingredient costs (Manary 2006). In addition, the absence of locally produced milk powder in almost all of the countries where SAM is most prevalent and the considerable difficulties in purchasing peanuts that meet the strict United Nations standards from Aflatoxin levels both pose a considerable barrier to the manufacture of RUTF in these countries; increasing costs, working capital requirements and the lead times for RUTF manufacture and procurement (UNICEF 2011). These factors combine to limit the availability of RUTF to needy children and there is a need for a new RUTF formulation that can be more safely and cheaply manufactured in the countries where it is needed. Such a product would need to use locally available foodstuffs such as legumes/pulses (e.g. soybean, chickpea, lentils) and grains (e.g. rice, maize, sorghum, millet), and ideally limit or eliminate the milk content in order to reduce costs.

This study tests the effectiveness of a new RUTF formulation [soy–maize–sorghum‐based (SMS)‐RUTF] for the outpatient treatment of SAM without complications. The SMS‐RUTF was made in Kenya from soya, maize, sorghum, oil, sugar and a mineral vitamin mix and contained no milk powder or peanuts. At the time of the study using Kenyan market prices, the cost of ingredients for producing one metric ton of SMS‐RUTF was USD $1583 compared to an ingredients costs of USD $2393 for one ton of P‐RUTF.

Key messages

• The milk‐free soy–maize–sorghum‐based ready‐to‐use therapeutic food may not be equivalent to the standard peanut‐based and milk‐based RUTF in severe acute malnutrion (SAM) management but the per protocol analyses and the findings in children aged 24 months or more suggest that a milk‐free RUTF could in the future be an option for improving cost‐effectiveness of programmes treating SAM.

• Research aimed at identifying suitable cheaper milk‐free RUTF that is as effective as the standard RUTF in treating children less or older than 24 months with SAM is still needed. This research should be conducted in a research setting designed to best facilitate regular follow ups and should include a direct assessment of product intake and an assessment of body composition.

Methods

Setting

The study was conducted between June 2009 and August 2010 in the health care clinics run by the Lusaka District Health Management Team in Lusaka, Zambia. To be eligible for inclusion in the study, the clinic needed to have the status of a health centre (HC), have a well‐established outpatient therapeutic programme (OTP) that had been running for at least 6 months and had already treated a caseload of over 100 children with SAM in the OTP programme. Twenty‐four of the 26 HCs met the criteria and these HCs were recruited to participate in the study. All the 24 HCs recruited completed the study.

Study design

This was a non‐blind, parallel group, cluster‐randomised equivalence trial with the randomisation to either SMS‐RUTF or P‐RUTF arms occurring at the level of the HC. The study could not be blind because of the differences in packaging and taste between the SMS‐RUTF and the P‐RUTF, while the cluster design was chosen to try to address the risk that caregivers and health workers would be biased in favour of the already well‐known P‐RUTF. The equivalence hypothesis was assumed given that results of nutrition therapy of human immunodeficiency virus (HIV) wasting adults with chickpea‐sesame‐based RUTF, another cereals and pulse‐based RUTF, was associated with a weight gain and recovery rate similar to that observed with the standard P‐RUTF (Bahwere et al. 2011; Ndekha et al. 2009).

Study population

All children were aged between 6 and 59 months and had been diagnosed as suffering from SAM without complications at one of the 24 HCs. The diagnostic criteria for SAM was a mid‐upper arm circumference (MUAC) <11.0 cm or pitting oedema of grade 1 (+) or 2 (++) (Collins et al. 2006b). Complications were defined as either medical or the absence of appetite. Medical complications were diagnosed using the World Health Organization's (WHO) Integrated Management of Childhood Illness (IMCI) standard definitions (WHO & UNICEF. 2005). Dehydration was diagnosed if the caregiver reported the occurrence of more than three watery stools per day and sunken eyes that started after the commencement of the watery stools. Appetite was assessed by asking the mother to sit quietly with the child for 15 min during which time she offered RUTF. If the mother reported that the child ate the RUTF then appetite was assessed as good, if not then the appetite was assessed as poor.

Children with SAM who presented with complication were referred to one of the four inpatient stabilisation units and were not eligible for this study. Children previously discharge from the study with a recovered outcome that later relapsed and presented again at the one of the participating HCs with a new episode of SAM were also not eligible for enrolment in the study a second time.

Randomisation

Stratified cluster randomisation was used to allocate the 24 HCs into intervention (SMS‐RUTF) and control (P‐RUTF) arms. Randomisation was stratified according to the time at which HCs started their CMAM service, and by their total number of patient visits. The classification of the HCs according to these criteria was carried out by the principal investigator (AHI) who had a good knowledge of the Lusaka CMAM programme. Using the sampling frame prepared by AHI, the epidemiologist (MBO) with no prior knowledge of the Lusaka programme, randomly allocated intervention arms to HCs in block of four using randomisation software. Twelve HCs were randomised to the P‐RUTF arm and 12 HCs to the SMS‐RUTF arm. Children who initially accepted the SMS‐RUTF but who at a later point in their treatment subsequently refused to eat the SMS‐RUTF were transferred to the P‐RUTF. These children were kept in the SMS‐RUTF group for the intention‐to‐treat analyses (ITT) but were excluded from the sample for the per protocol (PP) analyses.

Data collection and follow‐up

A modified version of the standard CMAM patient monitoring tool was used for data collection. It contained background information, anthropometric information, medical history and physical examination, and follow‐up sections. HIV and tuberculosis (TB) status were recorded. HIV testing was offered to untested children and their mothers using an opt‐in approach. Those who were willing to be tested were referred to an HIV testing room where they were counselled and tested. All confirmed HIV‐positive children and caregivers were linked with the HIV treatment clinic.

The study used nurses from the HCs as enumerators. In the week preceding the start of data collection, they received refresher training on acute malnutrition assessment and on CMAM protocols. Children presenting to the clinics were assessed to determine their nutritional status using MUAC, weight and bilateral pedal pitting oedema. MUAC was measured by trained health workers using graduated MUAC tapes on the left upper arm of each child. This was done twice for each child by two different enumerators and the average recorded at the nearest 0.1 cm. Weight in minimal clothing was measured to the nearest 100 g using Salter scales. Bilateral pitting oedema was assessed on the dorsum of the foot by pressing for 3 s twice, once at the screening stage and then again at admission. Once SAM was established, a medical history and physical examination was undertaken by a trained nurse. Children found to have medical complications and/or no appetite, were referred to the nearest inpatient stabilisation centre. All other SAM cases were admitted to the OTP and included in the trial.

Treatment protocols

After admission into the study, all children received a 5‐day course of amoxicillin, a single 100 mg dose of mebendazole, a 1‐week ration of RUTF and health and nutrition advice. The RUTF ration was calculated to provide 200 kcal kg⁻¹ day⁻¹. All children were asked to return to the HC for a follow‐up visit once each week until they were discharged from the programme. At each follow‐up visit, MUAC, oedema and weight were recorded and the children were screened for medical problems and the presence of appetite. Caregivers were interviewed at each visit about the acceptability of the RUTF and a repeat 1 week ration of RUTF was provided at the same dosage rate. Children were also asked about whether they had eaten an RUTF formulation other than the one to which they had been allocated. No other means was used to assess for non‐adherence to allocated RUTF.

Outcomes

Children exited the study in one of five ways: recovery (cure), death, default, transfer out of the catchment area and non‐recovery. For children admitted because of a MUAC <11.0 cm, recovery was defined as a weight gain of at least 18% and MUAC >11.0 cm and no medical complication and the absence of bilateral pitting oedema. In the case of children admitted because of bilateral pitting oedema, recovery was defined as the absence of bilateral pitting oedema, and clinically well and a MUAC >11.0 cm.

A child was considered to have defaulted if he was absent for three consecutive visits. Defaulters were followed‐up and invited back into the programme by trained volunteers and those who returned were given a new outcome based on their status when they exited the programme. At the end of the study, all remaining defaulters were traced and classified as alive, dead or in the case of those not found, as lost‐to‐follow‐up (LTFU). Verbal autopsy to confirm death and to assess the possible cause of the death was undertaken for children who were reported as having died.

Children whose condition deteriorated in the course of the outpatient treatment were referred to one of the four in‐patient stabilisation centres located in Lusaka. Once stabilised, the children returned to the HC to complete their treatment and an outcome was allocated based on their exit status from the OTP. Those children who died in an inpatient unit were given an outcome status of ‘death’. Those who remained in an in‐patient unit until they met the recovery criteria or were still in the inpatient unit at the end of the study period were given an outcome status of ‘not returned from inpatient’.

Food products

The SMS‐RUTF and P‐RUTF content are presented in Table 1 and details of P‐RUTF and SMS‐RUTF processing and acceptability have been published separately (Owino et al. 2013). All the main ingredients for SMS‐RUTF preparation, namely soybean, maize and sorghum were whole, non‐defatted, non‐dehulled grains/seeds and were processed based on an extrusion cooking technique. The level of some of the nutrients in the SMS‐RUTF was higher than the United Nations specifications for RUTF in order to compensate for the increased levels of anti‐nutrients present in the plant‐based ingredients (WHO et al. 2007; Golden 2009; Michaelsen et al. 2009). The iron content was increased to compensate for iron absorption inhibitors including phytic acid and polyphenols and also because it has been shown that the current P‐RUTF formulation may increase the risk of anaemia during rapid catch‐up growth (Hurrell 2002; Diop et al. 2003b; Gibson et al. 2010). The level of iron fortification used was also guided by the WHO recommendations of 100 mg of ferrous sulphate for 1000 Kcal for the fortification of therapeutic milk F100 (WHO 1999). The zinc content was increased following the WHO recommendations on micronutrient fortification in food with phytic acid to ensure that the molar phytic acid/zinc ratio remains close to that of the P‐RUTF (WHO & FAO 2006). The content of niacin was increased because the current specification is for RUTF containing milk and milk is a substrate that can be used to synthesise niacin in vivo, a metabolic route not available when a non‐milk SMS‐RUTF is used. (Henriksen et al. 2000; Ferguson et al. 2008).

Table 1.

Ingredients and nutrients of the study foods

Ingredients/Nutrients	SMS‐RUTF*, †	P‐RUTF ‡ , §	United Nations specifications ¶
Ingredients
Soybean (g 100 g−1)	29.7	0.0
Maize (g 100 g−1)	18.2	0.0
Sorghum (g 100 g−1)	6.5	0.0
Dried Skim Milk (g 100 g−1)	0.0	25.0
Sugar (g 100 g−1)	14.6	27.4
Peanut paste (g 100 g−1)	0.0	26.0
Palm Oil (g 100 g−1)	22.4	0.0
Soybean oil (g 100 g−1)	0.0	20.0
Palm stearin (g 100 g−1)	5.6	0.0
Vitamin and minerals Premix (g 100 g−1)	3.0	1.6
Nutrients
Energy (Kcal 100 g−1)	521	530	520–550
Protein/Energy ratio (%)	8.5	12	10–12
Fat/Energy ratio (%)	57.0	56.0	45–60
Omega‐6/Energy ratio (%)	10.4		3–10
Omega‐3/Energy ratio (%)	1.1		0.3–2.5
Omega‐6/Omega‐3 ratio	9.6		5–9
Vitamin A (μg 100 g−1)	1852	910	810–1100
Vitamin C (mg 100 g−1)	139	53	≥50
Vitamin D (μg 100 g−1)	14	16	15–20
Vitamin E (mg 100 g−1)	139	20	≥20
Thiamin (Vitamin B1) (mg 100 g−1)	1.4	0.6	≥0.5
Riboflavin (Vitamin B2) (mg 100 g−1)	1.9	1.8	≥1.6
Niacin (Vitamin B3) (mg 100 g−1)	19	5.3	≥5
Pantothenic acid (Vitamin B5) (mg 100 g−1)	8.3	3.1	≥3
Pyridoxine (Vitamin B6) (mg 100 g−1)	1.4	0.6	≥0.6
Biotin (Vitamin B7) (μg 100 g−1)	56	65	≥60
Folates (Vitamin B9) (μg 100 g−1)	370	210	≥200
Cobalamin (Vitamin B12) (μg 100 g−1)	2.3	1.8	≥1.6
Vitamin K (μg 100 g−1)	14	21	15–30
Calcium (mg 100 g−1)	463	315	300–600
Phosphorus (mg 100 g−1)	380	370	300–600
Magnesium (mg 100 g−1)	74	86	60–140
Potassium (mg 100 g−1)	704	1140	1100–1400
Copper (mg 100 g−1)	0.9	1.7	1.4–1.8
Iodine (μg 100 g−1)	417	100	70–140
Iron (mg 100 g−1)	52.5	12	10–14
Zinc (mg 100 g−1)	18.5	11.1	11–14
Anti‐nutrients
Phytic acid (mg 100 g−1)	475	255	<100
Phytic acid/Zinc ratio	2.5	2.2	<15
Phytic acid/Iron ratio	0.8	1.9	<1

*SMS‐RUTF, soya‐maize‐sorghum‐based ready‐to‐use therapeutic food; ^†Values calculated by linear programming using the foods composition tables from Nutrisurvey database, a database regularly updated by the World Food programme; ^‡P‐RUTF, peanut paste based ready‐to‐use therapeutic food; ^§Values obtained from the manufacturer of the standard P‐RUTF; ^¶Obtained from references Golden 2009, Michaelsen 2009, and WHO, UNICEF & SCN, 2007.

The SMS‐RUTF was packaged in 250 g clear plastic screw top pots and the P‐RUTF in 92 g, branded, laminated foils sachets.

Sample size calculation

To calculate the sample size, we defined equivalence as being when the recovery rates of SMS‐RUTF and P‐RUTF were within the 10% margin of equivalence. The calculation was premised on the study being implemented in 24 HCs with 12 unequal sized clusters per arm and an intra‐class correlation (ICC) coefficient of 0.02. Based on the analysis of Lusaka CMAM data from the 2 years immediately preceding the study, we assumed a recovery rate of >80% for P‐RUTF and a 10% loss to follow‐up. With this margin, we expected the two RUTFs to have a minimum recovery rate of equal or superior to the internationally set SPHERE standard cut‐off of quality of 70% (SPHERE project team 1999). To demonstrate equivalence at a 5% significance level and with an 80% power, 1604 children equally divided between the two arms were required (Jones et al. 1996). It was assumed that this sample would be sufficient for both the ITT and the PP analyses.

Data management and statistical analysis

Patient data were entered into a Microsoft Access Software™ (vista 2007) database prepared for this study. Data were entered by two enumerators and cleaned and coded data were then exported to STATA‐11 (StataCorp LP, College Station, TX, USA) for analysis. We analysed data at individual level rather than at cluster level and all analyses were adjusted for clustering.

In accordance with recommendations for analysing and reporting equivalence studies, both ITT and PP analyses were performed (Le Henanff et al. 2006; Piaggio et al. 2006). The ITT analyses included all children enrolled in the study. The PP analyses excluded children who defaulted, transferred out of the programme, were lost to follow up after inpatient transfer or switched RUTF.

Summary enrolment characteristics for each arm were calculated as mean ± standard deviation or median (interquartile range, IQR) for continuous or discrete measures and as n (%) for categorical measures. Means were compared using a clustered t‐test, the median using the adjusted Mantel‐Haenszel test and proportions using the clustered Chi‐square test. Recovery rate was calculated as a percentage of total recovered children divided by total exits. Binomial regression was used to estimate the adjusted risk difference between the two RUTF arms for each programme outcome. This was estimated using Huber‐White robust adjustment of errors to account for intra‐cluster correlation of outcomes and for strata randomisation. We used the analysis of variance estimator to calculate the ICC for all the primary outcomes.

Rates of weight gain during the entire period of follow‐up were estimated in g kg⁻¹ day⁻¹ and compared between the study arms. They were calculated by dividing the weight gain (weight at exit – weight at admission) expressed in grams by the weight at admission (in kilograms) and the length of stay (in days). We used logistic regression to test for interactions between the recovery rate and other variables. These variables were sex, age group (<24 months vs. ≥24 months), immunisation status (fully immunised vs. not fully immunised), child HIV status, mother HIV status, travelling time from home to the HC (>30 min vs. ≤30 min), number of children below 5 years in the households (≤2 vs. >2), admission criterion, presence of oedema and cluster. We used linear regression to test the interaction between weight gain and the variables listed above and appropriate multivariable analyses (logistic regression for recovery rate and linear regression for weight gain) to examine for confounding due to imbalances in baseline characteristics.

Ethical considerations

Permission for the trial to be conducted was obtained from the Lusaka District Health Management Team (LDHMT) and the Zambian Ministry of Health prior to randomisation. No additional authorisation was sought at HC level before or after randomisation. At the time of admission, each child's parent or carer was informed about the nature and purpose of the study and about the study arm to which the HC was allocated and asked for their informed written consent for their child to be included in the study and for their medical information to be used for research purposes. The study was approved by the University of Zambia Biomedical Research Ethics Committee.

Results

Figure 1 presents movement of children from preliminary screening to data analysis. A total of 2462 children were screened, of whom 1927 were eligible and consented for study enrolment while 535 were ineligible. The eligible children received either P‐RUTF (n = 1103) or SMS‐RUTF (n = 824) based on the HCs they attended for treatment. Forty‐three children of the SMS‐RUTF arm were switched to P‐RUTF while none of the P‐RUTF was switched to SMS‐RUTF.

Figure 1.

Trail flow diagram. ITT, intention to treat; PP, per protocol; P‐RUTF, peanut‐based ready‐to‐use therapeutic food; SMS‐RUTF, soya‐maize‐sorghum ready‐to‐use therapeutic food.

Table 2 compares the baseline characteristics of children included in the ITT analyses in the two study arms. There was no difference in terms of age, weight, time to the clinic, distribution of sex, child HIV status, mother HIV status, number of children below 5 years of age and that of HIV infected children on antiretroviral therapy. For children included in ITT analyses, a greater proportion of children in the SMS‐RUTF arm had oedema, diarrhoea, dehydration or were undergoing TB treatment on admission but these differences were not statistically significant. The median MUAC was significantly higher in the SMS‐RUTF arm than in the P‐RUTF arm. Similar imbalances were observed for children included in the PP analyses (see Supporting Information Table S1).

Table 2.

Baseline characteristics of children included in the intention‐to‐treat analysis of intervention (SMS‐RUTF) and control (P‐RUTF) arms

Criteria	P‐RUTF		SMS‐RUTF
N	1103		824
Socio‐demographic parameters
Male, n (%)	576	(52.1)	397	(48.2)
Age (months), median(IQR)	17.0 (12–22)		17.0 (12–22)
n children <5 years in HH, median (IQR)	2 (1–3)		2 (1–3)
Time to clinic, (minutes), median (IQR)*	30 (20–45)		30 (15–40)
Mother alive, n (%)	1011	(91.7)	771	(93.6)
Medical history
Fully immunised, n (%)	588	(62.6)	482	(66.8)
Child HIV status, n (%)
Positive	162	(14.7)	117	(14.2)
Negative	580	(52.6)	456	(55.3)
Not known/not tested	361	(32.7)	251	(30.5)
Mother HIV status, n (%)
Positive	321	(29.1)	263	(31.9)
Negative	510	(46.2)	374	(45.4)
Not known/not tested	272	(24.7)	187	(22.7)
ART initiated, n (%) †
Yes	50	(30.9)	43	(36.7)
No	69	(42.6)	41	(37.3)
Not known	43	(26.5)	33	(28.2)
On Anti‐TB, n (%)
Yes	51	(4.6)	20	(2.4)
No	856	(77.6)	658	(79.8)
Not known	196	(17.8)	146	(17.7)
Diarrhoea, n (%)	327	(29.6)	285	(34.6)
Dehydration n (%)	51	(6.2)	63	(13.8)
Nutrition status
MUAC (cm),median(IQR)	11.0 (10.5–12.5)		11.5 (10.5–12.7)
Weight (Kg), median (IQR)	7.0 (6.0–8.5)		7.1 (6.0–8.3)
Oedema, n (%)
None	424	(38.4)	248	(30.1)
Plus −1	388	(35.2)	284	(34.5)
Plus‐2	291	(26.4)	292	(35.4)
Admission category, n (%)
MUAC <110 mm	424	(38.4)	248	(30.1)
Oedema	585	(53.0)	509	(61.8)
MUAC <110 mm + oedema	94	(8.5)	67	(8.1)

ART, antiretroviral therapy; HH, household; HIV, human immunodeficiency virus; IQR, interquartile range; MUAC, mid‐upper arm circumference; P‐RUTF, peanut paste based ready‐to‐use therapeutic food; SMS‐RUTF, soya‐maize‐sorghum‐based ready‐to‐use therapeutic food; TB, tuberculosis; *n = 1519 (830 for P‐RUTF arm and 689 for SMS‐RUTF arm; ^†only children with confirmed HIV infection included.

Overall, in both the ITT and PP analyses, the recovery rate in the SMS‐RUTF was lower than the recovery rate in the P‐RUTF group. The recovery rates were 60.8% in the P‐RUTF arm and 53.3% in the SMS‐RUTF arm in the ITT analysis, with an adjusted risk difference (ARD) of −7.6 [95% confidence interval (CI) –14.6, −0.6]% (Table 3). In the PP analysis, the recovery rates were 81.9% in the P‐RUTF arm and 78.1% in the SMS‐RUTF arm, with an ARD of −3.8 (95% CI‐10.2, 2.6)%. The ICC for the ITT and the PP were 0.015 and 0.008, respectively. However, in terms of the equivalence margin of ±10% set in the study design, the results are inconclusive as the 95% CI for the observed recovery rates overlap this equivalence margin (the lower limit of the zone of equivalence) for both the ITT and the PP analyses (Table 3).

Table 3.

Treatment outcome and adjusted risk difference of enrolled children at the end of the study

Outcome	P‐RUTF (n, %)		SMS‐RUTF (n, %)		Adjusted RD (%, 95% CI)	P‐value*
All patients
Intention‐to‐treat (ITT) analysis	n = 1103		n = 824
Recovered	671	(60.8)	439	(53.3)	−7.6 (−14.6,−0.6)	0.034
Dead	138	(12.5)	113	(13.7)	1.1 (−0.3, 5.8)	0.597
Default	278	(25.2)	233	(28.3)	2.7 (−4.9, 10.2)	0.470
Non‐recovered	10	(0.9)	10	(1.2)	0.3 (−0.4, 1.1)	0.424
Other †	6	(0.5)	29	(3.5)	3.0 (1.3, 4.6)	0.001
Per protocol (PP) analysis ‡	n = 819		n = 534 §
Recovered	671	(81.9)	419	(78.5)	−3.5 (−9.6, 2.7)	0.257
Dead	138	(16.8)	106	(19.8)	3.0 (−3.5, 9.5)	0.350
Non‐recovered	10	(1.2)	9	(1.7)	0.5 (−0.8, 1.7)	0.454
Subgroup analyses ¶
<24 months
ITT analysis	n = 894		n = 681
Recovered	546	(61.1)	352	(51.7)	−9.4 (−16.5, −2.3)	0.012
Dead	113	(12.6)	99	(14.5)	1.8 (−2.1, 5.7)	0.357
Default	220	(24.6)	197	(28.9)	4.3 (−2.6, 11.1)	0.212
Non‐recovered	9	(1.0)	9	(1.3)	0.3 (−0.6, 1.2)	0.487
Other †	6	(0.7)	24	(3.5)	2.9 (1.1, 4.6)	0.003
PP analysis	n = 668		n = 435**
Recovered	546	(81.7)	335	(77.0)	−4.7 (−10.5, 1.0)	0.102
Dead	113	(16.9)	92	(21.2)	4.2 (−1.8, 10.2)	0.160
Non‐recovered	9	(1.4)	8	(1.8)	0.5 (−0.8, 1.8)	0.433
≥24 months
ITT analysis	n = 201		n = 135
Recovered	122	(60.7)	84	(62.2)	1.5 (−9.4, 13.0)	0.785
Dead	23	(11.4)	12	(8.9)	−2.5 (−11.5, 6.3)	0.558
Default	55	(27.4)	33	(24.4)	−2.9 (−14.1, 8.2)	0.593
Non‐recovered	1	(0.5)	1	(0.7)	0.2 (−1.5, 2.0)	0.487
Other †	0	(0.0)	5	(3.7)	3.7 (0.9, 6.5)	0.011
PP analysis	n = 146		n = 94 ††
Recovered	122	(83.6)	81	(86.2)	2.6 (−8.7, 13.9)	0.636
Dead	23	(15.7)	12	(12.8)	−3.0 (−14.7, 8.8)	0.603
Non‐recovered	1	(0.7)	1	(1.1)	0.4 (−2.1, 2.8)	0.753

CI, confidence interval; RD, risk difference (SMS‐RUTF minus PN‐RUTF); P‐RUTF, peanut paste based ready‐to‐use therapeutic food; SMS‐RUTF, soya‐maize‐sorghum‐based ready‐to‐use therapeutic food; *Linear regression, adjusted for intra‐cluster correlation of outcomes and randomisation strata, ^†Other includes the categories transfer out and lost to follow‐up after inpatient transfer; ^‡Children of the exit categories default and others not included; ^§The 43 children who switched RUTF are not included (20 recovered, seven died, 15 defaulted and one discharged as non‐recovered); ^¶Age unknown for 16 children (eight in each arm for the ITT analysis and five in each arm for the PP analysis) including two who switched RUTF; **The 38 children who switched RUTF are not included (17 recovered, seven dead, 13 defaulters and one non‐recovered); ^††Three children who switched RUTF not included (three recovered).

The subgroup analysis demonstrated an interaction between the age group and recovery rate (interaction P < 0.001 for ITT analyses and 0.0683 for PP analyses). The ARD and CIs for recovery rates for children >24 months of age at enrolment showed that the recovery rate in the SMS‐RUTF arm was lower than the recovery rate in the P‐RUTF in both the ITT (ICC = 0.017) and PP (ICC = 0.004) analyses (Table 3). By contrast, in children ≥24 months at enrolment, the recovery rate in the SMS‐RUTF arm was higher than the recovery rate in P‐RUTF arm (Table 3). In both these age groups, this study is again inconclusive as to whether the two RUTF were equivalent with respect to recovery rates. In children below 24 months of age, the 95% CI for the observed effects overlaps the equivalence margin at the lower limit of the zone of equivalence. For children aged 24 months or above at enrolment (Table 3) the 95% CI for the observed effects overlap the equivalence margin at the upper limit of the zone of equivalence.

The results for the mortality, default and non‐recovery rates demonstrated no difference between the two study arms (Table 3). Table 4 shows average weight gain of the children classified by their admission status and treatment arm. Children in the SMS‐RUTF arm had a lower weight gain than those in P‐RUTF arm (P = 0.007) in both oedematous (P = 0.018) and non‐oedematous (P = 0.091) cases. There was no statistically significant interaction between weight gain and any other baseline variables.

Table 4.

Mean rate of weight gain (g kg⁻¹ d⁻¹) by study group and by the form of severe acute malnutrition

Category	P‐RUTF	n	SMS‐RUTF	n	Difference	P‐value*
	Weight gain (95%CI) g kg−1 d−1		Weight gain (95%CI) g kg−1 d−1		Weight gain (95%CI) g kg−1 d−1
All discharged
All forms of SAMb	3.2 (2.9, 3.5)	667	2.2 (1.9, 2.5)	407	−1.0 (−1.7,−0.3)	0.007
Non‐oedematous cases	4.5 (4.0, 5.0)	222	3.4 (2.8, 4.1)	109	−1.1 (−2.3, 0.2)	0.091
Oedematous cases	2.6 (2.2, 2.9)	445	1.7 (1.4, 2.1)	298	−0.8 (−1.5, −0.2)	0.018
Discharged per protocol
All forms of SAM	3.4 (3.1, 3.7)	617	2.2 (1.8. 2.5)	361	−1.2 (−1.8, −0.6)	<0.001
Non‐oedematous cases	5.0 (4.4, 5.6)	199	3.7 (3.0, 4.4)	88	−1.3 (−2.4, −0.3)	0.016
Oedematous cases	2.6 (2.3, 3.0)	418	1.7 (1.3, 2.0)	273	−0.9 (−1.6, −0.3)	0.004
Discharged cured
All forms of SAM	3.5 (3.2, 3.8)	594	2.3 (2.0, 2.6)	370	−1.2 (−1.8, −0.6)	<0.001
Non‐oedematous cases	5.4 (4.8, 5.9)	186	3.9 (3.2, 4.5)	91	−1.5 (−2.6, −0.4)	0.008
Oedematous cases	2.7 (2.4, 3.1)	408	1.8 (1.4, 2.1)	279	−0.9 (−1.6, −0.3)	0.005

CI, confidence interval; P‐RUTF, peanut paste based ready‐to‐use therapeutic food; SAM, severe acute malnutrition; SMS‐RUTF, soya‐maize‐sorghum‐based ready‐to‐use therapeutic food. *Linear regression, adjuster for intra‐cluster correlation of outcomes and randomisation strata.

The median length of stay for children recruited into P‐RUTF arm was 35 days (IQR, 23–49 days) and for SMS‐RUTF it was 35 days (IQR, 21–56), (Kruskal–Wallis: X₁ ² = 0.476, P = 0.494). For those who were discharged as recovered, the median length of stay was 35 days (IQR, 28–52) and 47 days (IQR, 29–70) for those in P‐RUTF arm and those in the SMS‐RUTF, respectively (Kruskal–Wallis: X12 = 34.69, P < 0.001).

As part of sensitivity analysis, we performed as treated analysis in which children of the SMS‐RUTF cluster who took P‐RUTF were included in the P‐RUTF group. This as treated analysis gave the same results as the planned ITT and PP analyses for cure rate, weight gain and length of stay (data not shown).

Discussion

In this study, we aimed to compare the effectiveness of two different RUTFs in the treatment of SAM without complications in Zambian children. The RUTFs were a milk‐free SMS‐RUTF and the United Nations recommended standard P‐RUTF with 25% milk. Overall recovery rates, the primary outcome for this study, were higher in the P‐RUTF than in the SMS‐RUTF arm but the study was inconclusive and did not confirm our hypothesis of equivalence between the two different RUTFs in the treatment of SAM. However, there was evidence of a possible heterogeneity of treatment effect (HTE) between children <24 months and those ≥24 months, with a suggestion that although the SMS‐RUTF might be inferior to P‐RUTF for children <24 months, this might not apply to children aged >24 months.

This was an effectiveness study implemented as part of the routine operations of a primary health care programme treating SAM and the results need to be interpreted in the context of the difficult and unpredictable environment within which the study was carried out. Cholera and measles epidemics occurred between December 2009 and May 2010 during the implementation study and much of Lusaka was affected by extensive flooding between January and February 2010. These factors made it more difficult for caregivers to travel to the health clinics; delaying presentation and increasing the proportion of defaulters across all study sites. These incidents also disrupted the activities of health workers and outreach workers delivering routine services, undermining the routine management of SAM cases and the follow‐up of enrolled children. The poor public health environment increased the likelihood of children with SAM suffering from underlying acute infection, as evidenced by the high proportion of children in both arms having diarrhoea at admission. Taken together, these external factors increased the risk of default and death while at the same time decreased the ability of the staff to manage the programme and trace defaulted children. The resultant negative impact meant that both study arms suffered similarly high default rates and neither the P‐RUTF (the effectiveness of which is well documented), nor the SMS‐RUTF, attained a recovery rate of 75%, the level considered acceptable for CMAM services (SPHERE project team 1999). This is very different to the 3 years that preceded the study, when the defaulter rate of the Lusaka CMAM programme was below 20% and the mortality rate was below 5% (Mbwili‐Muleya, C, International CMAM workshop, April 2008). The PP analysis suggests that in the absence of this increased defaulter rate, the recovery rate for both the RUTFs would likely have been above 70%.

This study had several limitations. It was implemented as part of an ongoing routine programme operated by the LDHMT and nurses who were not directly part of the research team recruited the children. The two RUTFs were packaging differently and it was therefore not possible to blind staff or carers to the type of RUTF allocated. This inability to blind the study opened the possibility of bias resulting from a preference for the packaging or product on the part of the care providers or patients. P‐RUTF was packaged in attractively branded foil sachets and had already been in use in Lusaka for several years meaning that the health workers and some caretakers were already familiar with and were appreciative of this product. By contrast, the SMS‐RUTF was new to Lusaka and was packaged in plastic screw top pots with a stick‐on label. A preference for P‐RUTF over SMS‐RUTF was mentioned by a small number of health workers and caregivers during the study and 43 children of the SMS‐RUTF arm were switched to P‐RUTF as a direct result of this preference. The greater number of cases enrolled into the P‐RUTF arm in this study also indicates that this preference may have biased some health workers and caregivers against SMS‐RUTF leading to preferential referral to or attendance at the sites of the P‐RUTF arm. As a prior acceptability study among Zambian children had revealed no difference in acceptability and tolerance between the two products, the study did not plan for direct actual food intakes, compliance and resource restrictions precluded such monitoring.

Another major limitation was the large proportion of patients for whom the final outcome was unknown. Despite substantial efforts to track down children who defaulted from the programme, the problems affecting Lusaka and its health service mentioned above, combined with high intra‐city migration, meant that many could not be traced. The high default rates observed in this study are comparable to those reported in other urban settings with a high HIV prevalence (Bachou et al. 2006; Sadler et al. 2008; Kerac et al. 2009). Although default rates were statistically similar between the two study arms and are therefore unlikely to have biased the results, the large number of lost of follow‐ups reduced the power of the study.

In accordance with the Lusaka CMAM guidelines, height was not measured in our study and the lack of height data is a potential limitation for this study. Some have argued that height gain is the best anthropometric proxy indicator of increase in lean mass during nutrition rehabilitation and that weight gain without linear catch‐up may just indicate increase in body fat (Oakley et al. 2010) However, we believe that this limitation is relatively unimportant as several other studies have demonstrated that during recovery weight gain precedes height gain by 3 to 4 months and that increases in height start only after the recovering child has achieved at least 85% of expected weight‐for‐length (Brown et al. 1982; Walker & Golden 1988; Maleta et al. 2003) by which time they would have been discharged from this study. The expected rates of height gain during the treatment of SAM are approximately 0.2 mm day⁻¹ (Manary et al. 2004; Ciliberto et al. 2005, 2006) meaning that with the average 45‐day length of stay in this programme, measuring height change with sufficient precision would have posed considerable technical difficulties. Despite the lag between weight and height increments, studies have shown that nutrition therapy with a diet providing a balanced amount of energy proteins and micronutrients was associated with deposition of more lean mass than fat increase (Fjeld et al. 1989; Walker & Golden 1988; Branca et al. 1992; Oakley et al. 2010).

The limited number of allocated units in a cluster randomised study made it more difficult to match each study arm than in an individually randomised trial (Hahn et al. 2005). In this study, we found that many important baseline characteristics, such prevalence of oedema, average MUAC, the presence of diarrhoea and presence of dehydration, differed between the two arms. Although we performed multivariable analyses to assess the effect of these differences on our study outcomes and found that these variables did not appear to interact with the outcomes, we cannot categorically rule out the possibility that differences between the two arms other that diet, might have affected our findings.

To the best of our knowledge, no previous studies have compared P‐RUTF to alternative milk‐free RUTF processed from non‐peanut cereals and legumes. A recent randomised controlled study from Malawi compared standard P‐RUTF with 25% milk vs. P‐RUTF with only 10% milk. Recovery rates with both RUTFs were above 80%, well above the 75% international standard for therapeutic feeding programmes. However, the study found that the 10% milk RUTF had recovery rates that were 7% lower at 4 weeks, reducing to 3% lower at 8 weeks at which point children were discharged (SPHERE project team 2004; Oakley et al. 2010). With no predefined equivalence margin, the authors' conclusion of lower effectiveness was based solely on a statistical test of differences and did not take into account the clinical relevance of any difference (Oakley et al. 2010), nor the possible public health benefit of increasing access to RUTF through lowering the cost of RUTF.

The availability of financial resources to purchase RUTF is a major barrier to the scale up of CMAM and the impetus behind this study was to search for alternative RUTF formulations that are cheaper to manufacture and can be made more safely and easily in the developing countries where they are used. The SMS‐RUTF is a good candidate recipe as it does not contain milk and can be manufactured almost entirely out of locally grown ingredients. The elimination of milk powder and the inclusion of locally available grains and pulses reduces the cost of ingredients by USD$ 810 per ton (−33%) and increases the potential for the manufacturer to provide an economic boost in the countries where it is manufactured. The absence of imported milk and the elimination of peanuts that are highly prone to aflatoxin contamination also decrease the working capital needed for at scale manufacture, and ease the manufacturing challenges of producing safe RUTF in developing countries' factories. Therefore, despite the non‐equivalence in recovery rates in the overall study result, we believe that the similar mortality rates in each arm, the 70% recovery rate and the equivalence to P‐RUTF observed in the PP analysis for children above 23 months, merit further investigation.

However, those receiving P‐RUTF had higher rates of weight gain and reduced length of stay compared to those receiving SMS‐RUTF. This decreased the total quantity of P‐RUTF required to achieve a cure, offsetting some of the advantages of using the cheaper SMS‐RUTF product. Future studies on the public health impact of new RUTF recipes should include a comprehensive cost effectiveness analysis to assess the relative importance of the different cost parameters.

There are several possible reasons for the lower rates of weight gain observed in the SMS‐RUTF group. The absolute protein content in the SMS‐RUTF was lower than that in the P‐RUTF and this might have increased the proportion of new tissue deposited as fat rather than lean tissue. As the energy requirement for the synthesis of fat (35 kJ g⁻¹) is far greater than that required to deposit lean tissue (5.6 kJ g⁻¹) (Jackson & Wootton 1989) this could have decreased the rate of weight gain in this study. Future studies into new RUTFs should include a direct assessment of body composition.

Another possibility is that the intake of SMS‐RUTF was lower than then P‐RUTF. Although the acceptability trial conducted before the start of this study in Zambian children did not show any difference in acceptability (Owino et al. 2013), intake or side effects including a depression of appetite through the production of flatulence could have been different in the children with SAM. The absence of food intake data in this study prevents us from analysing the impact of intake on rates of weight gain and future studies into new RUTFs should include a direct assessment of intake.

The micronutrient levels of the SMS‐RUTF are unlikely to be the cause as these were increased to counteract the effect of anti‐nutrients and the protein digestibility–corrected amino acid score of SMS‐RUTF was above 70% (Dibari et al. 2012) making it unlikely that protein quality contributed to the decreased rates of weight gain in the SMS‐RUTF group.

As mentioned above, we believed that the internal validity of our findings have been affected by the operational constraints faced during study implementation. Similarly, despite good adherence and acceptability of the intervention by health workers, carers and their children, we consider that the generalisability of our findings is limited to settings facing similar constraints as those faced during the present study. The comparable defaulter rates and the completion of the study by all the HCs suggest that both interventions were feasible and that results could have been generalisable to the whole Lusaka CMAM programme.

Conclusion

This study was inconclusive and did not confirm our hypothesis of equivalence between SMS‐RUTF and P‐RUTF in SAM nutrition management. However, the similar mortality rates between the two arms, the PP analyses that minimised the effect of the unexpected natural hazards that occurred during the study, and the findings in children aged 24 months or more suggest that a milk‐free RUTF could in the future be an option for improving cost‐effectiveness of programmes treating SAM. This merits further investigation.

Source of funding

Irish Aid (IA) provided funding for the study.

Conflict of interest

Valid Nutrition designed and produced the SMS‐RUTF. VOO is an employee of Valid Nutrition. SC is the unpaid director of Valid Nutrition. Valid International is the sister company of Valid Nutrition. Irish Aid had no say on the design, implementation and interpretation of the results. Valid Nutrition administered the study grant, while staff from Valid International implemented the study. A statistician from the University of East Anglia did the final analysis.

Contributions

SC and PB conceived the study idea and provided technical oversight throughout the trial including data collection, data analysis and preparation of this manuscript. AHI, VOO and MOB designed the study protocol and implemented data collection, entry and analysis. FD designed SMS‐RUTF formulation in collaboration with VOO and participated in manuscript preparation. EID implemented initial data collection and participated in the manuscript write up. CMM supervised data collection and reviewed the manuscript. All authors contributed to the write up of the manuscript. All authors have read and approved the manuscript.

Supporting information

Table S1. Baseline characteristics of children included in the per protocol analysis in each arm.

Irena, A. H. , Bahwere, P. , Owino, V. O. , Diop, EH. I. , Bachmann, M. O. , Mbwili‐Muleya, C. , Dibari, F. , Sadler, K. , and Collins, S. (2015) Comparison of the effectiveness of a milk‐free soy‐maize‐sorghum‐based ready‐to‐use therapeutic food to standard ready‐to‐use therapeutic food with 25% milk in nutrition management of severely acutely malnourished Zambian children: an equivalence non‐blinded cluster randomised controlled trial. Matern Child Nutr, 11: 105–119. doi: 10.1111/mcn.12054.