Menu
Menu
Menu
Menu
In 2011, Tanzania mandated the fortification of edible oil with vitamin A to help address its vitamin A deficiency (VAD) public health problem. By 2015, only 16% of edible oil met the standards for adequate fortification. There is no evidence on the cost-effectiveness of the fortification of edible oil by small- and medium-scale (SMS) producers in preventing VAD. The MASAVA project initiated the production of sunflower oil fortified with vitamin A by SMS producers in the Manyara and Shinyanga regions of Tanzania. A quasi-experimental nonequivalent control-group research trial and an economic evaluation were conducted. The household survey included mother and child pairs from a sample of 568 households before the intervention and 18 months later. From the social perspective, the incremental cost of fortification of sunflower oil could be as low as $0.13, $0.06, and $0.02 per litre for small-, medium-, and large-scale producers, respectively, compared with unfortified sunflower oil. The SMS intervention increased access to fortified oil for some vulnerable groups but did not have a significant effect on the prevention of VAD due to insufficient coverage. Fortification of vegetable oil by large-scale producers was associated with a significant reduction of VAD in children from Shinyanga. The estimated cost per disability-adjusted life year averted for fortified sunflower oil was $281 for large-scale and could be as low as $626 for medium-scale and $1,507 for small-scale producers under ideal conditions. According to the World Health Organization thresholds, this intervention is very cost-effective for large- and medium-scale producers and cost-effective for small-scale producers.
The research component of the MASAVA project included a quasiexperimental nonequivalent control‐group trial, with one control and three intervention districts in each of the Manyara and Shinyanga regions. Control districts were selected based on the municipal divisions. Children 6–59 months of age and their lactating mothers were the units of observation for this research trial. Local nutrition officers and health care personnel at health centres helped identify eligible mothers to participate in the household survey. Eligible mothers were randomized within districts from stratified groups by age, geographic area, and income to obtain a heterogeneous sample. Inclusion criteria for the index child in each household were being between 6 and 59 months of age and being the oldest child under five of a lactating mother (at baseline). The household survey questionnaire contained multiple modules. The DHS programme methodology served as a template for the modules on household demographics, wealth index (including property, housing facilities, and consumer goods), and mother and child nutrition and feeding behaviours (Measure Demographic and Health Survey, 2014). The survey also included modules on food security, food consumption, and dietary diversity (Food and Agriculture Organization, 2013) as well as oil fortification coverage based on the Global Alliance for Improved Nutrition’s fortification assessment coverage tool (Aaron et al., 2017). A Helen Keller International questionnaire was used as the template for the vitamin A knowledge, attitude, and practices module (Helen Keller International, 2012). Data collection also included dried blood finger prick samples for the mother and child pair, anthropometric measurements of the index child (height and weight), and a sample of the available cooking oil from each household. All modules were collected both at baseline and at end line, with the exception of the asset variables. The National Institute for Medical Research in Tanzania and the University of Waterloo both provided Research Ethics Board approval. The baseline data for this study were collected between May and July 2015, and the end line survey conducted approximately 18 months later between November 2016 and January 2017. Other related publications and reports contain additional information about the MASAVA research trial (Horton et al., 2017; Walters et al., 2018). The household survey sample size calculation, which estimated that a sample size of 385 would be required, assumed a maximum variability in the population (p = 0.50), a 5% margin of error, and a 95% confidence interval (CI; Wu, Corbett, Horton, Saleh, & Mosha, 2019). To account for the high uncertainty of attrition at end line, the sample size was inflated to include 568 households. Assuming an expected mean retinol level of 14.84 μg/ml in children of the group unexposed to fortification (based on the baseline mean retinol in Manyara), the unbalanced sample size design (N ratio = 0.55) provided sufficient power to detect a 10% change in mean serum retinol when comparing the intervention and control groups (β = 0.88). Among the 568 households that participated in the baseline survey, there were 366 mother and child pairs in the intervention group and 202 in the control group. Attrition in the total number of households participating in the survey at end line compared with baseline was 13%. Moreover, dried blood samples could not be collected at end line for 21% of mothers and 24% of children surveyed at baseline; therefore, a total of 412 children participated at end line. Comparison of the participant characteristics and vitamin A status of attrited versus nonattrited children suggested that it was unlikely that attrition introduced any systematic bias (Walters et al., 2018). A multivariate regression analysis of access to fortified oil was undertaken using the end line data. Because consumption of adequately fortified oil was small at baseline, we do not present these results here. The underlying theoretical model was informed by the prevailing conceptual frameworks for the determinants of malnutrition (Bhutta et al., 2013; Black et al., 2013) and economic models of health investment and household allocation of resources in low‐resource settings (Alderman, Chiappori, Haddad, Hoddinott, & Kanbur, 1995; Strauss & Thomas, 1998). Both ordinary least squares (OLSs) and quantile regression analyses were used to examine the effect of independent variables representing geography, socio‐economic status, food and oil consumption, and knowledge of vitamin A, on the dependent variable representing household access to fortified oil, namely, oil retinol levels in milligram per kilogram measured from the samples collected. Independent variables in the model included household location (whether the household was located in an intervention or control district), region, and urban or rural, as well as a household wealth index score, maternal dietary diversity, knowledge of the benefits of vitamin A, and primary oil‐type consumed. The OLS regression only measures the effect of the independent variables on the dependent variable at the mean of the distribution, whereas the quantile regression analysis allows for a more flexible, non‐linear relationship between independent and dependent variables at the median and various percentiles of interest (Hill, Griffiths, & Lim, 2011). In the quantile regression, the dependent variable quantile rank is the proportion of values in the oil retinol level distribution that are greater or equal to the percentile. The quantile regression was conducted for the 10th, 25th, 50th, 75th, and 90th percentiles, and statistical significance at p < 0.05 and 0.01 levels was reported. Stata V.14.2 was used for all univariate, bivariate, and multivariate analyses. The Fiedler and Afidra (2010) economic evaluation of vitamin A fortification in Uganda and the relevant guidelines for the economic evaluation in health programmes informed the methodological design of this costing and cost‐effectiveness analysis (Bill and Melinda Gates Foundation, 2014; Drummond, Sculpher, Torrance, O'Brien, & Stoddart, 2005). The costing analysis of fortified oil in Tanzania compared with unfortified oil, and within each of the three producer scale categories, was conducted from the societal perspective for a 1‐year time period and included both private and public sector costs. Private sector cost components included start‐up costs, which were annuitized, for the purchase and installation of equipment to mix the fortificant with the oil, initial training of staff, and redesign of oil labels. Private sector recurrent costs for the purchase of the vitamin A premix, additional energy use and labour for mixing, quality assurance testing, and packaging of fortified oil if different from unfortified oil were included. Management and distribution fees as well as the government value‐added tax were included as a proportion of the direct costs where applicable. Monthly monitoring data provided by the MASAVA project SMS partners and consultant reports were the primary sources of fortification unit‐cost data (Market Axis Limited, 2016; Randall, 2016). Public sector costs for regulation, social marketing, monitoring, and programme management were crudely estimated based on the previous estimates from a prospective economic evaluation of fortification in Tanzania (National Food Fortification Alliance, 2009). All costs were expressed in 2017 U.S. dollars (US$), and the 90‐day average exchange rate on December 5, 2017, of 2,325 Tanzania shillings (TZS) to one US$ was used for necessary conversions. For the cost‐effectiveness analysis, the estimated health effect was drawn from the results of a differences‐in‐differences (DIDs) regression analysis of the reduction of the prevalence of VAD in children and women (Horton et al., 2017; Walters et al., 2018; Walters, 2018). This study assumed a 1:1 serum retinol to retinol‐binding protein ratio (e.g., the threshold for VAD was below 0.7 or 14.44 μg/ml in children and below 1.05 or 26.04 μg/ml in women; Namaste et al., 2017). This health effect was factored into a population‐attributable fraction equation using the published relative risks of mortality from childhood diarrhoea and measles associated with VAD (Stevens et al., 2015) and global burden of disease data on morbidity associated with VAD to calculate the hypothetical total number of child deaths and disability‐adjusted life years (DALYs) averted (Institute for Health Metrics and Evaluation, 2017). The total number of DALYs averted due to fortification of oil with vitamin A was then combined with the above‐mentioned estimates on the incremental cost of fortification to estimate the cost per DALY averted for each of small‐, medium‐, and large‐scale producers. The health effect of fortification observed in the Shinyanga region, which relied more on consumption of oil fortified by large‐scale producers, was extrapolated to the entire country for this calculation. Sensitivity analysis was also conducted on costing and cost‐effectiveness results by varying values for key cost drivers including production volume, the cost of fortification equipment, and the cost of not using recycled oil containers. The health effect estimate of fortification was varied to both low and higher values derived from the DID regression analysis (Walters et al., 2018).
N/A
