Ethical approval was obtained from the National Health Sciences Research Committee, Malawi (NHSRC reference #784), prior to commencing the study.
Two Extension Planning Areas (EPAs) in Malawi were studied: Zombwe EPA in Mzuzu Agricultural Development Division (ADD) in the north and Mikalango EPA in Shire Valley ADD in the south. Zombwe EPA is characterised by low pH soils (median 5.2, n = 11), particularly Haplic Lixisols (70% of EPA area), which have a low soil-to-maize transfer of Se and median grain Se concentration of 22 μg Se kg−1
fresh weight (FW)10
. Mikalango EPA is characterised by areas of calcareous Eutric Vertisols (median pH = 7.8, n = 16; 10% of EPA area) with a high soil-to-grain transfer of Se and a median grain Se concentration of 342 μg Se kg−1
. Six villages were selected from each EPA for participant recruitment. In Zombwe EPA, villages were located in Zombwe I Section: Bandawe Tembo, Kenani, Msekeni, Ngayiwona, Yesaya Jere, Yolamu. In Mikalango EPA, villages were located in Nyamichimba Section: Billy, Chamwaka, Chifundo, Chimkango, Moses, Mozyenti.
In total, a sample of 120 apparently healthy adult females aged 18–50 years were recruited to participate in the study, from six villages per EPA, most of whom were subsistence farmers. Only one volunteer per family was recruited to avoid multiple sampling of participants with similar dietary patterns. Exclusion criteria for participants included known pregnancy, smokers, those diagnosed with chronic physical or mental disease or long term illness requiring treatment, acute or chronic infection, those taking regular prescribed medicine including oral contraceptive pills, those taking regular dietary supplements and those who had lived in the study area for less than six months. Sensitisation visits were held in September 2010 by the scientific team and Extension Planning Officers from the Malawi Ministry of Agriculture and Food Security, and in March 2011 by the Ministry of Health and Traditional Village Authorities. Informed consent of all volunteers was obtained prior to the collection of the duplicate diet composites and tissue samples by a trained member of the Lilongwe University of Agriculture and Natural Resources study team. No payment was made for participation or the use of tissue samples. Participants were reimbursed at cost-price for the food and beverages in the duplicate diet composites taken for analyses.
Socio-economic status of participants
Background data were collected using questionnaires administered by trained interviewers in the homes. A socio-economic status (SES) index was developed for each participant based on categories used earlier in the Nutrition Collaborative Research Support Program (NCRSP33
), and in a later study of pregnant women in rural Malawi34
. Information collected included house quality, sanitation, water source, household size, occupation and schooling. The index has a maximum theoretical numerical value of 14 (high SES). The numerical value is based on derived numerical equivalents calculated for each of the string variables in STATA (StataCorp LP, College Station, Texas, USA). The SES index and background data on participants are given in Supplementary Table 1
Duplicate diet composites
Duplicate diet composites were collected over a full day (24 h) for each participant. Food records and diet composites samples were collected between 13 and 16 March 2011 in Mikalango EPA, and between 20 and 23 March 2011 in Zombwe EPA. Research assistants (RAs) were recruited to reside in the household of each participant to weigh and sample all food and beverage items (including drinking water and snack items consumed away from the household) during the sampling date using a 750 mL polyethylene jug and a kitchen scale accurate to ~1 g. For each food and beverage item consumed, an exact duplicate sample was weighed and collected in a double-lined trace-element-free sealable polyethylene bag. Composites were placed in a cooler box with ice packs, transferred to a central laboratory within 24 h of collection, and blended to a homogeneous slurry using a domestic blender. Approximately 50 mL of homogenate was subsampled using a trace-element-free pipette (7017, Corning, Costar, Amsterdam, The Netherlands) and this was divided into two trace-element-free universal tubes (Bibby Sterilin, Stone, Staffordshire, UK). Homogenate subsamples were frozen at −20°C. Weighed food samples were recorded (data not shown) to provide details on the composite diet samples containing animal products, to comply with import licence requirements.
One set of frozen diet composite subsamples (n = 115) was transported to the UK on dry ice for Se analyses, and the remaining set was retained in Malawi. Subsamples were freeze-dried in their original universal tubes with muslin caps. Subsample volumes were recalculated to correspond to the loss of mass on freeze-drying plus the residual mass of solid material, assuming that the dry food had a material density of 1.0 and was wholly in suspension. The mean volume of each subsample was 24.7 mL: typically 20–23 mL and 1–4 g dry material. Freeze-dried diet composites were manually ground to homogenise the samples and crush seeds etc. Approximately 0.3 g of each sample was microwave-digested in 3.0 mL of 70% Trace Analysis Grade (TAG) HNO3, 2.0 mL H2O2 and 3.0 mL milli-Q water (18.2 MΩ cm; Fisher Scientific UK Ltd, Loughborough, UK). Elemental analysis was by inductively coupled plasma-mass spectrometry (ICP-MS; X-SeriesII, Thermo Fisher Scientific Inc., Waltham, MA, USA). Acid digests were analysed by ICP-MS (i) using Collision Cell Technology with Kinetic Energy Discrimination (CCT-KED) mode for Na, Mg, K, Ca, Al, Ti, V, Cr, Mn, Fe, Co, Ni, Cu, Zn, As, Rb, Sr, Mo, Cd, Cs, Ba, Pb and U (using Sc, Ge, Rh and Ir internal standards), and (ii) in H2-reaction cell mode for 78Se alone (using Ge and Rh internal standards) with 2% methanol added to samples to maximise sensitivity for Se determination. Three certified reference materials (CRMs) from NIST (National Institute of Standards and Technology, Gaithersburg, MD, USA) were included: NIST 1573a (tomato leaves, certified value 0.054 mg Se kg−1 dry weight, DW, 123% recovery), NIST 1577c (bovine liver, 2.031 mg Se kg−1 DW, 96% recovery) and NIST 1567a (wheat flour, 1.1 mg Se kg−1 DW, 100% recovery). In addition, four inter-lab flour samples supplied by Dave Hart, Institute of Food Research, UK yielded an average recovery of 102%. Operational blanks (n = 10) were run to determine practical Limit of Detection (LOD; 3*SD) and Limit of Quantification (LOQ, 10*SD) values for Se of 0.0018 and 0.0059 mg kg−1 DW, respectively, in composite diet samples. These and subsequent data collected from individuals were analysed by ANOVA with ‘EPA/Village’ as a treatment factor and no blocking terms using GenStat (15th Edition, VSN International, Hemel Hempstead, UK).
Blood and urine samples
Whole blood was collected from consenting volunteers by trained staff from the Community Health Sciences Unit, Malawi. Samples were collected on Saturday 12 March 2011 in Mikalango EPA and Saturday 19 March 2011 in Zombwe EPA. Each volunteer was seated, their skin cleaned with alcohol at the site of the antecubital vein, and their arm restricted with a tourniquet for <1 minute. Blood was drawn into two 7 mL trace-element-free blue-top evacuated tubes (NH119 I.U. ref 367735, Becton Dickinson UK Ltd, Plymouth, UK). Disposable polyethylene powder-free gloves were used for all blood handling steps. The tourniquet was loosened during blood collection and before the needle was removed from the arm. Immediately after the blood was collected the tube was inverted 10 times to mix the contents before placing the samples in a cooler box with ice packs for transport to the nearest laboratory for processing (Zombwe EPA, Mzuzu Central Hospital; Mikalango EPA, Zomba District Hospital). For one of the 7 mL samples, plasma was separated by centrifugation at 1500 g for 10 min and transferred into trace-element-free tubes (n = 4 aliquots for each sample) and frozen at −80°C. Plasma samples were transported to the UK on dry ice for Se analysis using ICP-MS, and to determine Se-dependent glutathione peroxidase activity, as described below. The other 7 mL samples were used for whole-blood analyses (data not shown). Casual urine (i.e. samples unrelated to eating/sleeping patterns) samples were collected on the same day as the blood sampling. These were collected in sterile pre-labelled tubes (30 mL sterile universal tubes, Bibby Sterilin, UK) and placed in a cooler with ice before transfer to the nearest laboratory for processing. Two 8 mL subsamples were frozen at −20°C and transported to the UK using dry ice for ICP-MS analysis.
Selenium concentration in blood plasma was determined by ICP-MS in H2-cell mode. Samples and standards (SPEX CertiPrep Inc., Metuchen, NJ, USA) were diluted 1-in-20 in 1% TAG HNO3 containing 0.1% of a non-ionic surfactant (‘Triton X-100’ + ‘antifoam-B’, Sigma-Aldrich Company Ltd., Dorset, UK) and 2% methanol. Internal standards containing Ir (5 μg L−1), Rh (10 μg L−1), Ga (25 μg L−1) and Sc (50 μg L−1) were included. Calibrations were in the range 0–50 μg Se L−1. Quality control CRMs included Seronorm™-1 (107 μg Se L−1, 109% recovery), Seronorm™-2 (163 μg Se L−1, 102% recovery) (Sero AS, Billingstad, Norway) and UTAK-66816 Lot 7528 (certified value 115 μg Se L−1, 106% recovery) (UTAK Labs Inc., Valencia, CA, USA) CRMs. Sample blanks (n = 10) were run to determine LOD and LOQ values of 0.055 and 0.182 μg L−1, respectively, for blood plasma. Urine Se was determined by ICP-MS in H2-cell mode for samples diluted 1-in-10 with 2% HNO3 and calibrated using Seronorm™-1 (certified value 58.6 μg Se L−1) as a matrix-matched calibration standard. Urine assays were corrected for creatinine by dividing urine Se concentration (μg L−1) by creatinine concentration (g L−1). Sample blanks (2% HNO3; n = 10) were run to determine LOD and LOQ values of 0.095 and 0.316 μg L−1, respectively, for urine. Creatinine was analysed using a commercial kit on a RX IMOLA Chemistry Analyzer (Randox Laboratories, Belfast, UK) according to the manufacturer's instructions.
Quantification of plasma GPx3 activity
Plasma glutathione peroxidase (GPx3) activity was quantified using a high throughput 96 well spectrophotometric enzyme assay35
. All plasma samples were analysed in triplicate. The enzyme assay mixture contained 100 mM Tris-HCl pH 7.4, 3 mM glutathione, 0.25 mM NADPH, 1U glutathione reductase and 0.1% triton X100 with hydrogen peroxide as the substrate. The rate of decrease in absorbance at 340 nm was measured at 10 s intervals for 15 min at 37°C using a plate reader with kinetic capability (FLUOstar Omega plate reader, BMG Labtech, Germany). The GPx3 activities were calculated from the initial rates of reaction36
, with one unit (U) defined as 1 nmol NADPH oxidised per minute.
Estimating dietary selenium supply from food balance sheets
supply of Se was estimated for all African countries as the product of food supply data available for human consumption at a retail level and food Se concentration using previously described methodology19
. Food supply was based on Food Balance Sheet (FBS) data sourced from 46 Food and Agriculture Organization (FAO) member countries in Africa19
. Food Se concentration data were sourced from published food composition tables21,22,37,38
and three food Se concentration databases were generated: East (E) Africa, Southern (S) Africa (also used for North (N) Africa) and West (W) Africa (also used for Mid (M) Africa). All Se concentration data are expressed as μg Se 100 g−1
fresh weight (FW) edible portion. Food Se concentration data, literature sources and best-fit FBS categories are shown in Supplementary Table 3
. Per capita
Se supply data were used to infer risk of dietary Se inadequacy using the EAR cut-point method19,20
. The US Estimated Average Requirement (EAR) of 45 and 23 μg Se person−1
was used for those aged >10 and <10 yr, respectively, and 49 and 59 μg Se person−1
for pregnant and lactating women, respectively39
, with a conservative inter-individual coefficient of variation (CV) of Se intake of 25%. As discussed previously19
, the EAR cut-point method is highly sensitive to CV and also to food concentration data reported within the food composition tables. For example, Stadlymayr et al.21
reported 810 μg Se 100 g−1
FW edible portion in raw coconuts, resulting in high intake estimates for some West African countries (see Results).
Estimating disability-adjusted life years (DALYs) lost due to micronutrient deficiency in Malawi
It is not straightforward to determine the existing burden of disease due to Se deficiency in any country at present because an accepted framework for Se is lacking. However, it is possible to illustrate the impact of mineral deficiencies in Malawi using zinc deficiency (ZnD) as an example. Thus, the number of DALYs lost due to ZnD in Malawi was estimated using previously developed methods24,40,41,42
. Population size was taken from World Bank sources43
and demographic data were obtained from the 2008 Population and Housing Census Results of the National Statistical Office of Malawi44
; the crude birth rate was obtained from the 2008 Population and Housing Census Results Main Report45
. Infant mortality rate was obtained from the 2008 Population and Housing Census Results Mortality Report46
, while the under-five mortality rate was obtained from the 2012 Malawi Population Data Sheet47
. Average remaining life expectancies were obtained from the World Health Organization life tables48
, while stunting rates (height/age - 2 SD) were taken from the WHO Dataset on Child Growth and Malnutrition for Malawi49
. The estimated number of episodes of diarrhoea and pneumonia in infants and children were taken from Stein et al.40
and confirmed by local expert opinion (AA Kalimbira, Lilongwe University of Agriculture and Natural Resources). Per capita
income figures were taken from the World Development Indicators of the World Bank43
; DALYs were valued in accordance with established methods50
and using a discount rate of 3%40,41
. To convert DALYs lost to monetary terms, a standard figure for developing countries of US $1,000 was used which corresponds approximately to triple the annual per capita
income in Malawi.