Study location and population
Yen Bai province, one of 64 provinces and cities in Vietnam, is a mountainous region with a largely rural economy, widespread poverty and diverse ethnic groups. Two districts, Yen Binh and Tran Yen, were chosen for the study. At the time each district had a total population of approximately 26,000 WRA. Women of reproductive age were defined in this setting as aged between 15 and 45 years and all were eligible for the intervention.
Before the intervention, two staff of each district Department of Preventive Medicine, two nurses from each commune health station and all village health workers in the two districts (a total of 680 health personnel) were trained about the causes, health risks, treatment and prevention of anemia and hookworm infection and received promotional and educational materials for the women. Commencing May 2006, weekly supplementation and 4 monthly deworming was implemented by active distribution through the existing health structure. Iron/folic acid and albendazole tablets were distributed from the provincial implementing agency, the Yen Bai Malaria Control Program office, through the district Preventive Medicine centres, to the Commune Health Stations (for albendazole treatment) and on to the Village Health Workers for distribution of the iron/folic acid tablets to the women on a monthly basis. In addition, health workers distributed simple educational materials. All WRA were encouraged to collect packs of four ferrous sulphate/folic acid tablets (60 mg/0.4 mg, UNICEF, Copenhagen) from their village health worker each month. Albendazole (400 mg, UNICEF, Copenhagen) was administered as observed treatment on locally designated days either at the commune health station or supervised in the village by a commune health worker.
A stratified multi-stage cluster sampling design was used for the baseline survey, which was conducted in November 2005. Primary sampling units (villages) were chosen using a 'probability proportional to size' random sampling method separately within each district, with half the target sample of villages taken from each district. Secondary sampling units (individual women) were selected randomly from each village using provincial lists. Sample size was calculated on an expected hemoglobin range of 90 g/L to 140 g/L with a population standard deviation of 10 g/L. Allowing for the clustered nature of the surveys a sample size of 250 was estimated to be sufficient to detect a 3 g/L change in the mean hemoglobin. Allowing for refusals to participate and potentially inaccessible villages, 34 villages were selected and 12 women from each village randomly selected, a total of 408 women.
Post-implementation surveys were conducted after three and 12 months of intervention using the same sampling design for choosing participants. We used a cross sectional survey approach for participant selection at each time point in order to examine the population effect of the intervention in the target districts.
Samples and testing
All surveys were conducted by the same teams and included trained phlebotomists, stool preparation and analysis technicians, a demographic recorder and a supervisor. Sample collection and laboratory analysis has been previously reported  and was the same in each of the surveys. Briefly, Hb was assessed at the field site using a HemoCue 201+ (HemoCue AB, Angelholm, Sweden). A 3 mL sample of venous blood was also collected using a closed collection system into tubes containing fast clotting agent. Serum ferritin was measured using a sandwich immunoenzymatic assay (IEA; Beckman Coulter Access Reagents, Fullerton, CA). Soluble transferrin receptor was evaluated using enzyme-linked immunoassay (ELISA; IT; Orion Diagnostica, Espoo, Finland). The ratio of transferrin receptor to log (base 10) serum ferritin (TfR-F index) was calculated from these results. Anemia was defined as a hemoglobin concentration of <120 g/L and iron deficiency as serum ferritin of < 15 μg/L, except where otherwise stated, in concordance with WHO recommendations for women of reproductive age . Transferrin receptor levels of 2.3 mg/L or above were considered abnormal, based on the manufacturer's reference interval (0.8–2.3 mg/L). The TfR-F index implies depletion of iron stores once the ratio exceeds 1.8 .
Faecal samples were examined microscopically for soil transmitted helminth eggs at the field site using standard Kato-Katz methodology . Classification of hookworm infection was based on WHO guidelines of 0 eggs per gram (epg) as no infection, >0 and <2000 epg as mild infection and >2000 as moderate or severe infection. Soil transmitted helminth (STH) infection was categorized as positive if eggs of hookworm (Ancylostoma duodenale and Necator americanus), Ascaris lumbricoides or Trichuris trichiura were detected in the stool sample.
In July 2006, just prior to the 3-month post-implementation survey, the Research and Training Centre for Community Development (RTCCD), an independent Vietnamese non-Government Organization, monitored compliance in the two districts. Sampling was as described for the surveys above. Monitoring was conducted as one-to-one interviews with women and included questions about their understanding of anemia, iron deficiency and helminthiasis, the availability of the intervention and their recall of accessing the deworming treatment and taking the weekly iron tablets during the previous 10 weeks.
Data entry and checking
Hb and helminth egg counts were entered into an Excel (Microsoft Office 2003) spreadsheet at the field site. Team supervisors crosschecked entries each day.
The study sample was defined as randomly selected groups at baseline and three and twelve months post-implementation. Hb values were approximately normally distributed. Serum ferritin and soluble transferrin receptor values were right-skewed and so log-transformed for analysis. For certain multi-level ordered categorical variables (e.g. education level, place of work), some categories were collapsed to identify relevant differences. Robust confidence intervals were derived by clustering for village, the primary sampling unit.
Linear regression analysis was used for continuous variables while binary variables were analysed by logistic regression. For continuous variables the differences in means between time points were calculated in absolute values for Hb, and for the log transformed variables (serum ferritin, soluble transferrin receptor and the TfR/F-Index) as the ratio of the geometric means between post-implementation and baseline surveys. Binary prevalence variables were analysed by proportion estimation adjusted for robust SEs with clusters as villages. Change over time was analysed as the risk ratio between post-implementation and baseline. Statistical analysis was performed using Stata vs. 10 (StataCorp, 2005, College Station, Texas).
Extensive consultation was undertaken between the project team and community leaders, as well as liaison with village, district and provincial health staff. Village health workers provided participants with information regarding the surveys and signed informed consent was documented at the time of enrolment for the surveys. The survey team assisted the village health workers where participants expressed concerns or uncertainty relating to any aspect of their participation. The project was approved by the Human Research Ethics Committee of the National Institute of Malariology, Parasitology and Entomology (Hanoi, Vietnam), the Walter and Eliza Hall Institute of Medical Research (Melbourne, Australia) and Melbourne Health (Melbourne, Australia).