Setting and survey design
Tibet is located on the highest plateau of the world, with 86% of the region at an altitude of at least 4000 meters. Most Tibetans live at altitudes between 2500 and 4500 meters , and the main economical activities are farming and animal husbandry . The 2003 population census indicated that there were approximately 2.59 million inhabitants in Tibet, with 96% being of Tibetan nationality . The rural population accounted for approximately 85% of the total population . The study was conducted in rural areas surrounding Lhasa, the capital of Tibet, which is located on the Qinghai-Tibet Plateau. The mean altitude of Lhasa is 3685 meters above sea level.
A cross-sectional survey was designed to document the health status of children aged under 24 months and their mothers' dietary and nutrient intakes. Participants were selected using simple random sampling and were interviewed face-to-face by trained professional interviewers between May and August 2008. The study was approved by the Ethics Review Committee, Xi'an Jiaotong University College of Medicine. All participants signed a consent form before completing the interview.
Since the purpose of our study was to measure the health status of children aged under 24 months and the dietary intakes of mothers, we first calculated the sample size based on the rates of diarrhea (67.2%) and upper respiratory infection (67.3%) reported in the 2002 NNHS and the rate of breastfeeding (85%) in the Development of Chinese Children (2001-2010). The sample size equations were as follows: α = 0.05, δ = |p-π| = 0.05, followed by . The required sample sizes were 339, 338, and 196, respectively. We chose the maximum sample size (339), and expected a 20% no-response rate. The final sample size was 406. Furthermore, there was a lack of precise statistics for the nutrient intakes of Tibetan mothers with young children. However, a sample size of 382 was used for lactating mothers' nutrient intakes in the 2002 NNHS ; we used a sample size of 406 to describe Tibetan mothers' nutritional status. We selected the participants using simple random sampling based on the list of children who were under 24 months old and their mothers in the study area. A total of 386 participants completed the study (95.1%).
Socio-demographics and anthropometry
Standardized interview questions were used to collect socio-demographic information, including age, years of education, maternal occupation, and family size, as described in previous studies in Tibet [8, 9]. Subjects were asked to remove all heavy clothes and shoes and undo hair styles and accessories in a preparation area. Next, trained staff weighed the subjects on a calibrated electronic scale (Tanita HD-305, Tanita (Shanghai) Trading Co., Ltd.) and recorded the value to the nearest 0.1 kilogram. Standing height was measured using a height-measuring tape (LD-SG01, Ningbo Land cooperation) suspended from the wall. The subjects were asked to stand erect with their shoulders level, hands at their sides, thighs together, and heels comfortably together. Subjects also kept the upper back, buttocks, and heels in contact with the wall and the head aligned in the Frankfort plane during the height measurement. The height values were recorded to the nearest 0.1 centimeter. All anthropometric measurements were taken by a single trained staff member. Body mass index (BMI) was calculated as weight (in kilograms) divided by height (in meters) squared. Women were classified into BMI less than 18.5 kg/m2, 18.5-24.9 kg/m2, 25-29.9 kg/m2, or over 30 kg/m2 .
The mothers' nutritional status was determined from the dietary data collected using a food-frequency questionnaire (FFQ). The FFQ has adequate reproducibility and validity for assessing dietary intake in rural China [11, 12]. Since this survey was conducted in Tibet, we added some Tibetan foods to the questionnaire, such as Zanba, Tibetan milk tea, and Tibetan salt cream tea. Before being used in the study, the FFQ was revised twice with the assistance of local nutrition staff. After each revision, the FFQ was re-tested to determine whether the items fitted local daily food consumption patterns. The final version of the FFQ included 92 items, covering cereals, meat, vegetables, fruits, egg products, nuts, fish, beverages, alcoholic beverages, and snacks. Typical dishes were displayed in the booklet, along with average portion sizes [13, 14]. Options for the serving size were 0.5, 1, 1.5, or 2, with the displayed size as reference (average size) for most of the food items. More detailed options were offered for portion sizes of Tibetan foods. Participants were asked how often, on average, they consumed each food item during the past 12 months.
Intakes were calculated for energy, protein, fat, carbohydrate, vitamin A, vitamin B1, vitamin C, vitamin E, calcium, iron, and zinc. Nutrients were calculated according to the 2002 and 2004 China Food Composition Tables [15, 16]. Food items were divided into six different subgroups, following previous experience of nutritional investigations and dietary habits in the rural area of western China [11, 12].
Nutrient reference values
Nutrient reference values were based on national standards . Since these were rural participants, we used Chinese DRIs (dietary reference intakes) with moderate PAL (physical activity level). Reference values were RNI (recommended nutrient intake) or AI (adequate intake) for lactating mothers. These recommended values are: energy 11721 kJ, protein 90 g, fat at 20-30% of total energy (63.4-95.0 g), and carbohydrate at 55-65% of total energy (379.2-448.2 g). We chose the median of the range for fat (79 g) and carbohydrate (414 g) as the recommended value. Other nutrient reference values were as follows: vitamin A 1200 μg (retinol equivalents), vitamin B1 1.8 mg, vitamin B2 1.7 mg, vitamin C 130 mg, vitamin E 14 mg, calcium 1200 mg, iron 25 mg, and zinc 21.5 mg. In the present analysis, we compared our results with the daily nutrient intakes of lactating mothers in the 2002 NNHS. In that survey, dietary intake data were collected using 24-hour recall over 3 consecutive days . The nutrient intakes of lactating mothers in the 2002 NNHS were as follows: energy 10036 kJ, protein 70 g, fat 77 g, carbohydrate 355 g, vitamin A 524 μg, vitamin B1 1.1 mg, vitamin B2 0.8 mg, vitamin C 96 mg, vitamin E 37 mg, calcium 385 mg, iron 24 mg, and zinc 11.6 mg.
The interviewers (physicians from the Department of Public Health, School of Medicine, Tibet University) were trained to standardize administration of the questionnaire and anthropometric measurements. They were trained in the field for at least 1 week prior to commencing the surveys. A pilot survey was performed before the formal survey, with all data for analysis being from the formal survey. Two investigation teams were established, each consisting of four members and a supervisor. At least two members were Tibetan and were able to communicate in both Tibetan native language and Chinese. During the survey, a data checking system was employed, which involved all interviewers checking their own data, each interviewer checking data with another interviewer, and data checking by supervisors. Subjects were re-interviewed when answers to key questions or missing values were identified. The accurate age of the child was based on the Permanent Residence Registration and/or Record of Scheming Immunization documents, where the birth date is recorded. The Tibetan lunar calendar dates were converted to Gregorian calendar dates.
A database was established by using Epi Data version 3.1 (The EpiData Association Odense, Denmark), and double data entry was performed. We used the Shapiro-Wilk test to determine whether nutrient intakes had a normal distribution. The data for nutrient intakes had a non-normal distribution; therefore, median values were used to describe continuous variables. We calculated the consumption frequency of all foods, and arranged the five most frequently consumed foods from each food subgroup in descending order. We used the Wilcoxon signed rank test to determine whether our data were significantly different from 2002 NNHS values. To compare the differences between different socio-demographic groups, we used the Kruskal-Wallis test. To compare the differences between two subgroups within each socio-demographic group, we used the Wilcoxon two-sample test (adjusted P < 0.017). Statistical significance was determined at P < 0.05 (two-tailed tests). SAS version 9.2 (SAS Institute Inc., Cary, NC, United States) was used for all analyses.