Weight will be measured to the nearest 0.1 kg using a Seca digital portable scale (model 813 Seca gmbh & Co., Hamburg, Germany) with the participant wearing minimal clothing. Participants will be briefed about the importance of taking accurate weight measurements. HIV-infected people lose a considerable amount of weight before they seek medical care and clothes become very loose. To keep clothes in shape, many women wear several layers of clothing and if left unchecked, these could add considerably to the body weight. Each participant will be given a pre-weighed piece of cloth (two yard) to tie around their body at the time of measurement. All other clothes and accessories, with exception of underwear, will be removed.
Height will be measured to the nearest 0.1 cm using a free-standing Seca stadiometer (model 225 Seca gmbh & Co., Hamburg, Germany). A standard procedure for measuring height, without shoes and socks, will be followed. Height, if accurately measured, should remain constant across the duration of the study. However, it is common for African women to have elaborate hair styles that cannot be undone for the purpose of taking height measurements, therefore, where necessary, extra care will be taken in the estimation of height to account for such hair styles.
Circumference and skinfold thickness measurements will be done simultaneously since similar procedures are used to locate the sites. Three circumference measurements: mid-upper arm, abdominal at the level of the umbilicus, and hip will be taken using a Seca plastic measuring tape. Skinfold thickness will be measured at five sites: bicep, triceps, subscapular, iliac crest and medial calf using a Harpenden skinfold caliper (British Indicators Ltd., London, UK) with all measurements taken on the right side of the body. Standard procedures for the location of sites and taking measurements will be followed . All measurements will be taken in triplicate. Circumference and skinfold thickness measures provide an indication of regional fat distribution. FM as a percentage of total body weight will be calculated from the sum of skinfold thickness measurements (biceps, triceps, subscapular and suprailiac sites = ∑4SF) and prediction equations .
Bioelectrical impedance analysis
Impedance will be measured using a multi-frequency BIA (SFB7 model, ImpediMed Limited, Queensland, Australia) device according to the manufacturer's instructions. This will include ensuring that the battery is fully charged and the equipment is calibrated; participants will be asked not to exercise or take alcohol during the 24 hours before measurement. Participants will rest in a supine position for 5-10 min on a medical examination couch prior to measurement. A very mild current (50 ohms), imperceptible to the participant, passes through the body's water compartments. The SFB7 device takes measurements at 256 discrete frequencies, logarithmically spaced from 4 kHz to 1000 kHz. The software in the device enables the prediction of TBW and by extrapolation, FFM and FM. These prediction equations are population-specific hence there is a need to validate the results using the isotopic technique before the BIA device can be used routinely on HIV-infected people. Prior to analysis, each participant's height, weight, age and sex will be entered into the device.
Total Body Water
A procedural requirement is that the deuterium oxide dose be administered to a fasted individual or at least two hours after taking a meal. Prior to administering the dose, a baseline saliva sample will be collected from each participant and used to determine the natural baseline enrichment of the isotope in the body. Participants will also be asked whether they have fasted and if not, the time they had their last meal will be noted. Participants will report to the clinic at 08.00 h and the deuterium dose will be administrated between 10.30 h and 11.00 h (as an additional precaution against participants having eaten recently). A pre-weighed dose of 30 g deuterium oxide will be administered orally to each participant regardless of body weight with the aid of a straw, followed by a rinse of the bottle (approximately 50 mL of mineral water) to ensure consumption of the full dose. Participants will be told not to eat, drink or void their bladder within one hour of ingesting the dose and 30 min before collecting the saliva sample. During the equilibration period, the consumption of drinks and emptying the bladder will be restricted. If this is necessary, volumes will be noted. However, due to the extended period of fasting, and period of assessment, participants will be provided with a standard lunch and drink and the volume of the drink will be noted. The first post-dose saliva sample will be collected at four hours after taking the dose and the second post-dose sample at five hours to ensure equilibration.
Saliva samples (4.5 mL) will be collected using a cotton wool ball and with the aid of a syringe and transferred into pre-labelled 5 mL vials and subsequently, pre-labelled zip-lock bags. Saliva samples will be transported in an insulated box cooled with freeze packs to the laboratory at NISIR and stored at -20°C for later analysis.
Determination of TBW
Deuterium oxide enrichment of the saliva samples will be measured using a Fourier Transform Infrared (FTIR 8400S; Shimadzu, Vienna, Austria) spectrometer. This FTIR is available at the NISIR laboratories in Lusaka, Zambia. Prior to enrichment measurement of the saliva the FTIR will be calibrated using a 99.9 atom % deuterium oxide standard solution of known composition. The standard will be prepared according to the method described in the IAEA guidelines  by weighing deuterium oxide and diluting with local drinking water (bottled mineral water) to give a solution of about 1000 mg/kg by weight enrichment. This enrichment of the FTIR calibration standard will be verified using an Isotope Ratio Mass Spectrometer (IRMS) at a reference laboratory (Iso-Analytical Laboratory, UK). A range of enrichments will likely be encountered during analysis of saliva collected from the participants; the accuracy of deuterium oxide analysis over the range will therefore need to be checked by preparing a calibration curve. The 1000 mg/kg enrichment represents the maximum expected body water enrichment. The calibration procedure will involve preparation of deuterium oxide by gravimetric dilution (between 0 and 2000 mg/kg) and measuring their enrichment by the FTIR. Accuracy will be determined from the gradient and linear regression through the data of the calibration curve should be close to 1.
The FTIR will be set up according to IAEA guidelines [15–17]. Once ready to take measurements, the analytical precision of the FTIR will be checked before and after completing the measurements for the day by first taking measurements of local drinking water and the calibrator standard enrichment solution 3 times each and ensuring that the coefficient variation (CV) of less than 1% is achieved. Once the precision is ascertained the enrichment of the saliva samples for the participants will be measured. The spectrum of local drinking water will be obtained first as a background followed by measuring the spectrum of the standard in the range 2300-2800 cm-1. Thereafter the spectrum of the saliva will be taken by first obtaining the spectrum of the pre-dose saliva as background then followed by that of the post-dose samples (collected at 4 and 5 h). Prior to obtaining the spectra of the saliva, the samples will be thawed to room temperature and then centrifuged (Model 5702R, Eppendorf AG, Hamburg, Germany) at 3000 rpm for 10 minutes. Once the spectrum has been obtained it will be exported as a text file and further processed using the "isotope.exe" programme developed by the Medical Research Centre (MRC) . This program enables the translation of the spectrum intensity into excess concentration of deuterium oxide in the saliva sample. The dilution space will be calculated from the concentration of deuterium oxide in the saliva using a standard formula. TBW (kg) will be calculated from the dilution space after correcting for non aqueous dilution of deuterium oxide. The dilution space is 1.04 times TBW. The FFM will be derived from TBW using a hydration coefficient, that is the fraction of FFM comprised of water and this is constant in adult, 0.732. Once FFM has been estimated, FM and percentage body fat will be calculated.
Quality control will be done in accordance with the IAEA guidelines . The equation proposed by the IAEA for use as a quality control measure in TBW (kg) = 7.4 × height3 (m3), where TBW from each post-dose sample is used with the height of each participant. Since this equation will not be derived from our population in particular, a range incorporating the 95% confidence interval for the above relationship (< 5.7 × height3 or > 9.6 × height3) also proposed by the IAEA for assessment of accuracy, will be considered. The TBW samples falling outside this range will be excluded from the analysis. The variation in TBW values depends on both the degree of analytical precision and natural biological variation.
Physical examinations will be undertaken by a medical practitioner and fasting blood sample collected. At baseline and at 12-month assessments, three vials of blood will be collected, for CD4 count, viral load and micronutrient analysis. At 4-month and 8-month follow-ups, all measures except micronutrient analysis will be undertaken. Part of the blood sample for CD4 count will be used to measure full blood count which will be determined at the clinic laboratory. Blood samples will be processed at the clinic laboratory to separate the plasma from the blood cell for CD4 and viral load and the serum for micronutrient analysis. The plasma for viral load and serum for micronutrient determination will immediately be stored at -20°C at the clinic until shipment, while the plasma for CD4 count will be stored at room temperature, packaged and couriered by FedEx to the TDRC in Ndola.
Dietary intake study
A 3-day weighed food record coupled with a 24-h dietary recall technique will be used to determine the dietary pattern and the nutrient intake of the participants. Ten individuals with knowledge of nutrition will be recruited from various government institutions as "field staff". Training will include pre-testing on both 3-day weighed food records and 24-h dietary recall methodologies. Twenty participants will be recruited for the pre-test from another clinic in Lusaka offering ART services but not involved in the main study.
It is common practice for HIV-infected people to seek care and support from a health facility far from their residential catchment area. Therefore, it will be expected that many participants may be located in a wide range of residential locations of Lusaka including urban, peri-urban and rural areas. A well-designed house location plan will be required. Final updates on the residential address of each participant will be collected when they come for the 8-month follow-up visit. A participants' register to aid clustering based on their residential location will be prepared to assist in assigning participants within close proximity to the same field staff. In addition to telephone and house numbers, detailed directions to the participant's residence will be taken, including names of nearby landmarks such as schools, churches, and shops. It is also important to note the local names used to identify participants in their local communities. For example, it is common for women in Zambia to be called by their children or husband's names.
Visiting participants in their home might not be acceptable as some participants might not wish to disclose their HIV status to some or all members of their household. Participants will be assured that the all dietary intake data will be collected by non-medical personnel. This component of the study will be led by the National Food and Nutrition Commission which is well known as an institution responsible for promoting the nutrition agenda in the country which should help to minimize any potential stigma.
A total of 200 participants will be recruited for the study and a drop-out rate of 20% is anticipated. Accordingly, we anticipate that dietary intake data will be collected from approximately 160 participants. Each field staff member will be assigned 2 participants per day therefore we anticipate that data collection will take about two months to complete. Each participant will be contacted a week before data collection to remind them of the visit by field staff. Wherever possible, an understanding of level of literacy will be established beforehand. For participants who are unable to read and write, somebody in the household will need to be identified to record for them.
Field staff will train two participants per day, one in the morning and the second in the afternoon. This will be done on Mondays while Tuesday, Thursday and Saturday (or Sunday for Seventh Day Adventists) will be the 3 days during which the participants will be recording their food intake. On the second day (Wednesdays) the field staff will visit the participant's home to conduct a 24-h dietary recall. This will be done to assess the correct observance of the diet records, after keeping the record for one week day. The results of which will be immediately compared with those recorded for one day and will be the basis for reinforcing what will be discussed during the training. All forms and the food scales will be collected on Sundays, or Mondays for Seventh Day Adventists. Five supervisors will pay random visits to participant's homes to check on the accuracy and completeness of recording.
A questionnaire to assess demographic information will be completed by each participant on the day of recruitment and include questions on age, educational level, marital and socioeconomic status, and length of time they have known about their HIV status.
Data management and entry
All measurements will be entered on forms specifically designed for the project. Due to the number of measurements, forms of different colours will be used for easy identification. Completed forms will be checked for completeness by the PI or any other NISIR project staff present at the clinic. Each clinic will maintain separate files with all files stored safely by the PI.
Clinical assessment data including micronutrient analysis and socio-demographic assessment will be entered using EPI-INFO software (Centres for Disease Control and Prevention, Atlanta GA) while BIA, skinfold thickness, circumference and TBW measurements will be entered in Excel data sheets. Dietary intake forms will be checked for completeness by the supervisors at NFNC. A spread sheet in Excel 2007 will be created in which raw data from the record forms consisting of the foods and amounts consumed will be entered. Data bases containing nutrients to be used in the analysis will be created using food composition tables for Zambia , Food and Agriculture Organisation  and Mali [20, 21]. Recipe databases will be created using values from the 2009 NFNC study and the Uganda/HarvestPlus databases. Then the data base containing nutrients from the food composition tables will be merged with those containing the foods and amounts consumed to calculate the nutrient intake.
Sample size calculation
The sample size calculations were based on detecting a 5% weight change between groups over the 12-month period. Assuming a 10% variance in the within-group body weight, 37 participants are needed in each group to detect this difference with an alpha of 0.05 and a power of 90%. If the attrition is 25% over the year, a minimum of 48 participants are required in each group. Assuming FFM is 35% of the weight gain , 44 participants are needed in each group to detect this difference with an alpha of 0.05 and a power of 90%. If the attrition is 25% over the year, a minimum of 58 participants are required in each group. It is uncertain what magnitude of change is expected in CD4 counts and viral load. For a moderate effect size (Cohen's d = 0.4) 66 participants are needed in each group to detect this difference with an alpha of 0.05 and a power of 90%. If the attrition is 25% over the year, a minimum of 88 participants are required in each group.
Statistical Analysis plan
Statistical analyses will be performed using SAS software (SAS Institute, Cary, NC, USA). Descriptive statistics will be provided for the nutritional status, body composition, clinical and socio-demographic characteristics of the participants at baseline, 4, 8 and 12 months follow-up. At baseline, the characteristics of participants on food supplement and those without food supplement will be compared using t-test for continuous variables and Fisher's Exact Test for categorical variables. Repeated measures analysis of variance (RM-ANOVA) will be employed to compare the primary outcome variables (nutritional status and body composition) in participants with and without food supplements. Where confounding variables may need to be account for, adjustment will be made by covariance.