2.2. Data Collection and Laboratory Analysis
Prior to data collection, experienced field workers received intensive training consisting of classroom instruction and practice, laboratory practice, and field testing of all survey procedures. Training for phlebotomists included blood collection techniques, labeling of samples, and maintenance of the cold chain for transporting blood specimens. Tablet computers were used for direct data entry during interviews. Skip patterns were built into the electronic questionnaires, which sped up interviewing as well as minimized erroneous entries. Household and individual questionnaires were available in English, Uzbek, Russian, and Karakalpak. Interviews were conducted in the interviewee’s preferred language.
A short household questionnaire was administered to the head of the household or, if this person was not present, to another knowledgeable adult household member. The household questionnaire contained modules collecting data on demographic and socio-economic characteristics, household composition, and household food purchase and consumption. At each household, a flour sample (approximately 50 g) was requested, which was collected and stored in airtight containers for later analysis. Short questionnaires were administered to all recruited women, collecting information on minimum dietary diversity (24 hour recall), consumption of and knowledge about fortified and fortifiable foods, consumption of vitamin and mineral supplements, and demographic factors.
Anthropometry was conducted in all participating NPW using standard methods [16
] on a SECA (Hamburg, Germany) scale (UNICEF, #S0141021) and a standard wooden height board (UNICEF, #S0114540). Blood was collected from all NPW via venipuncture into 6-mL serum tubes (Becton Dickinson, Franklin Lakes, NJ, USA). Using a DIFF-Safe device (Becton Dickinson, Franklin Lakes, NJ, USA), a small amount of blood was extracted from the tubes onto a weighing boat to assess hemoglobin concentration using a portable hemoglobinometer (Hb301+, HemoCue, Angelsholm, Sweden). Quality control (QC) of the HemoCue devices was conducted and recorded daily using control materials (Eurotrol, Ede, Netherlands). Remaining whole blood was placed in a cool box containing cold packs to ensure storage at about 4 °C in the dark until further processing later the same day.
Every evening, blood and flour samples were transported to the regional blood transfusion centers for centrifugation and aliquotation of serum into cryovials appropriately labeled with the respondents’ identification numbers. Upon completion of field work in all regions, serum samples were collected and transported frozen to Tashkent, where they were stored at −80 °C until analysis. In the same transport, flour samples were taken to the ‘Donmahsulotlari LLC’ laboratory in Tashkent.
Sera were analyzed for ferritin, C-reactive protein (CRP), alpha1-acid-glycoprotein (AGP), retinol, vitamin B12, and folate. Ferritin, folate, and vitamin B12 were measured on the Siemens Immulite 2000 Xpi, while CRP and AGP were measured using the Siemens Dimension Xpand Plus. These biomarkers were analyzed at the Vitamed laboratories in Tashkent. Serum retinol was analyzed using reverse-phase high performance liquid chromatography (LC-20 Prominence with an Auto Sampler, Shimadzu, USA) at ‘MedStandart’ laboratories in Tashkent. Retinol was extracted into acetonitrile, centrifuged, and the supernatant was injected into the system using a Hypersil GOLD aQ Analytical HPLC Column, 3 µm, 4.6 × 100 mm (1202Y28, Thermo Fisher Scientific, Waltham, MA, USA), with a mobile phase of 83% acetonitrile, 0.1% trimethylamine, and 17% water by volume. Retinol was detected at 325 nm using a photo-diode array detector. Retinyl acetate was added as an internal standard before extraction (1716002, United States Pharmacopeia (USP) Reference Standard, Sigma-Aldrich (MERCK), USA). Purified retinol was used to construct the external standard curves (95144, BioXtra, ≥97.5% HPLC, Sigma-Aldrich (MERCK), USA).
Prior to analyzing survey samples, both laboratories underwent a series of external QC rounds conducted by the US Centers for Disease Control and Prevention, followed by capacity building, until laboratory performance was satisfactory. During the analyses of the survey samples, the laboratory conducted rigorous QC that was externally reviewed twice a week. Iron content in flour was assessed in a three-staged approach. All samples underwent qualitative testing using the iron spot test (AACC method 40–40). For samples containing iron, a semi-quantitative test was conducted (INCAP Method IV). Samples pre-determined to have ≥40 ppm iron underwent quantitative analysis using atomic absorption spectrophotometric method as an additional QC measure; there was good agreement between the semi-quantitative and quantitative method (R2 = 0.95), although the quantitative results were consistently higher by 1–7 ppm depending on the fortification levels, likely due to the fact that the quantitative analysis also measured intrinsic iron.
2.3. Parameters and Clinical Thresholds
Hemoglobin concentrations were adjusted for smoking status and altitude using World Health Organization (WHO) guidelines [17
]. Hemoglobin concentrations <120 g/L were used to classify NPW as anemic [17
Ferritin concentrations were adjusted for inflammation using CRP and AGP values via the method developed by the Biomarkers Reflecting Inflammation and Nutrition Determinants of Anemia (BRINDA) project [18
]. Ferritin concentrations <15 μg/L were used as cut-offs for ID in women [19
]. For retinol, a cutoff of 0.7 µmol/L was used to define vitamin A deficiency [20
]; however, because few women were found to be vitamin A deficient, we instead used vitamin A insufficiency (VAI) defined as serum retinol concentration <1.05 μmol/L [21
]. Retinol concentrations were not adjusted for inflammation [22
]. FD was defined as folate concentrations <10 nmol/L (<4.4 ng/mL), and vitamin B12 deficiency was identified when levels were <150 pmol/L (<203 ng/L) [23
]. Cut-offs for elevated CRP and AGP were >5 mg/ L and >1 g/L, respectively [24
Undernutrition and overnutrition in NPW were assessed using body mass index (BMI; kg/m2
). Undernutrition was defined as having a BMI less than 18.5, and overweight/obesity was defined as having a BMI of 25.0 or greater [25
2.4. Data Management and Statistical Analysis
Data were collected using Open Data Kit with built in checks and limits to minimize entry errors; additionally, the research team monitored data quality daily from a remote location outside of Uzbekistan. Laboratory data were double-entered using Microsoft Excel 2010. Data analysis was done using SPSS version 22 using the complex survey module. Standardized statistical weights calculated for households and women accounted for the unequal selection probability among the 14 strata. Because all NPW 15–49 years and all pregnant women in selected households were recruited for survey participation, the household sampling weights could be directly applied to each woman included in the survey sample.
Normality of the distribution of continuous data was checked using histograms and calculating skewness and kurtosis. Factors associated with anemia, ID, and FD were identified using bivariate analyses (see Supplementary Table S1
). All variables associated with a specific outcome in bivariate analyses with P
< 0.1 were included in a multivariable Poisson regression model after checking for co-linearity. The Poison regression produced adjusted risk ratios which were compared with unadjusted risk ratios calculated using the statistical weights. Variables included in the bivariate analysis were household variables (residence, region, household wealth quintile, household sanitation and access to safe drinking water source); woman’s physiology and nutrition (age, education, cigarette smoking, lactation, underweight, overweight/obesity, iron and folic acid supplement consumption, dietary diversity, consumption of iron- and folic acid-rich foods, additional iron and folic acid intake from wheat flour (as % of Reference Nutrient Intake (RNI)); woman’s micronutrient and inflammatory status (inflammation, iron, folate and vitamin B12 deficiency, vitamin A insufficiency).
Household socio-economic status was assessed using data on household characteristics and assets. Principal component analysis was used to calculate an index of household wealth, which was subsequently used to classify households into wealth quintiles [26
The weekly quantity of flour consumed in each household was calculated from the reported frequency of purchase and quantity usually purchased each time. The number of adult male equivalents (AMEs) in each household was calculated from the household roster information collected during the household interview [28
]. The AME is the proportion of a young adult male’s energy requirement needed by each age- and sex-specific group. The proportion of household flour consumed by an individual woman was considered equivalent to the proportion of total AME’s in the household represented by that woman. Estimates of daily flour consumption of >500 g/day were excluded from all analyses as physiologically implausible; this corresponded with the 95th centile of wheat flour consumption among Uzbek women in this survey. The calculation of mean flour intake included households reporting not having consumed wheat flour, but they represented only a small proportion and, thus, did not substantially affect the mean.
RNIs for iron and folate in women were obtained from WHO and the Food and Agriculture Organization (FAO) of the United Nations [30
]; for iron, an overall bioavailability of 12% was assumed to determine the target RNI. The additional amounts of iron or folic acid coming from household flour were calculated as a fraction of the RNI. Subsequently, %RNI categories were arbitrarily created using thresholds that would result in somewhat similarly sized groups of women with additional intake of iron and folic acid from fortified flour. For iron intake, the following RNI categories were created: 0% RNI (1461 women), 0.1%–39.9% RNI (308 women), and ≥40% RNI (268 women). For folic acid, the RNI categories were: 0% RNI (1093 women), 0.1%–69.9% RNI (265 women), and ≥70% RNI (192 women). Because the folic acid concentration in wheat flour was not directly measured, the folic acid levels were calculated as a proportion of the iron content in the wheat flour (using a 1:33 iron to folic acid ratio as found in the premix used for fortification).