As one of the project provinces of the China Health and Nutrition Survey (CHNS) which is a longitudinal and an ongoing survey covered eight provinces in China starting in 1989 and follows the human subjects approval procedures of both the University of North Carolina School of Public Health and the Chinese Academy of Preventive Medicine (CAPM, now renamed as China CDC), HLJ, replacing Liaoning, has performed the project since 1997. A multi-stage, random-cluster, weighted sampling scheme was used to draw the sample surveyed in the province according to CHNS' protocol(₁, ₂, ₆). There were six sampling spots chosen, namely two cities Harbin (provincial capital) and Shuangyashan (lower-income city), and four counties, Shuangcheng, Anda, Fujin and Jixian. Within each county, the town where the county government is located was selected, and then three other villages were chosen randomly. While within each city, urban and suburban neighbourhoods were selected randomly. The sample design was consisted of 24 primary sampling units (villages, towns or sub-districts from the six cities and counties), including 480 households, and covering 1470 individuals aged 0-92 (31.2±17.8) in 1997 and 1418 aged 0-87 (33.0±17.8) in 2000, respectively, after the exclusion of those respondents who were not examined physically or interviewed for dietary information. To ensure adequate sample size, new household were added in later surveys to replace those who were lost to follow up mainly due to migration or removal out of the community.

The data were obtained from CHNS-based surveys, rounds of the 1997—2000 in HLJ, with two methods for collecting dietary data – measurements of household food inventory and individual intake from a free-living population. Detailed household food consumption (HFC) data and individual dietary intake (IDI) data were collected for three consecutive days. For the former, HFC was determined from inventory change by weighing all foods in stock (including edible oils, sugar, salt and other condiments) from the start to the end of each day; foods purchased or produced at home during the survey period should be weighed and added into the inventory. Food preparation waste or those used for any other purpose except eaten by family members should also be weighed and subtracted from the inventory. The data collected herein were used to supplement the IDI data(₇). For the latter, IDI from the same 3 consecutive days was collected from all family members based on a 24-hour recall by face-to-face interviews at each participant's house, except children aged below 12 years whose food consumption was recalled by asking the child's mother or a mother substitute who handled food preparation and feeding in the household.

Dietary intake was expressed by the consumption per capita per day. The dietary data from both household inventory and individual 24-h recall were converted to nutrients according to the 1991 Chinese food consumption tables (FTC) (₈).

In this study, anthropometric measurements were made following standardized procedures(₃, ₆, ₉, ₁₀). Specific training on anthropometric assessment was provided at the beginning of each survey. Height and weight were measured (with study participants in light indoor clothes and without shoes) in centimeter and kilogram respectively during the course of dietary surveys by trained health workers, from which the body mass index (BMI) was calculated [BMI= weight (kg)/height (m²)]. The World Health Organization's cut-off points (₃, ₁₁,₁₂,₁₃) for BMI were used to defined underweight (BMI<18.5 kg m^-2) and overweight (BMI?25 kg m^-2 ), of which obesity was defined as BMI?30 kg m^-2.

Population characteristics were examined in the descriptive analysis. We attempted to compare any differences between/among gender, residence and household income. The changes in diet and nutrients, either macronutrients or micronutrients, during the 1997—2000 period were examined in the study sample stratified by gender, time, place of residence, sample spots, household income. The data input program was designed and offered by School of Public Health, the University of North Carolina. All data management and analyses were performed by using STATA (Version 8.0, College Station, TX).

There were 1,470 (male: 736; female: 734) participants in 1997's survey and 1,418 (male: 706 & female: 712) in 2000's. There was no significant difference in size between two genders in the two surveys (Pearson chi² =0.0035, P=0.953) (see table 1). There was difference in average ages between the two surveys (P<0.01) and the age proportion was different too (Pearson chi²= 9.6808; P=0.021) in two surveys, among which majority of participants were under 60 (about 92%). There was no difference in mean ages between genders no matter whether it was considered separately (P=0.37 in 1997 and P=0.52 in 2000) or as a whole (P=0.27). This was because that almost all the participants (n=1091) were nearly three years older in 2000 than they were investigated in 1997.

Figure 1

Table 1: Sociodemographic and characteristics of study sample population

Of the study samples, there were twice as many participants who lived in rural areas as those lived in urban and the proportion remained almost the same over time (P=0.431). From the point of monthly household income, most of them (75%) were at a level of medium or high in both survey years.

379 out of 1470 (25.8%) participants withdraw from the follow-up study, because of removal of one community at Anda after 1997 and the growing number of away-from-home students and farmers. There were no significant difference statistically in mean food consumption and in dietary patterns between the drop-off and those still remained in the survey in 1997, and so was that between the newcomers and the remainders in 2000. It can be said from this point of view that the withdrawal of the drop-off and the participation of the newcomers would not have much influence on making analyses and conclusions. Due to incomplete information on household income in 2000, only 1182 out of 1418 questionnaires, with satisfying income information, were used in analysis.

The percentage of participants who were underweight (BMI >18.5 kgm^-2), overweight (BMI = 25.0-29.9kgm^-2) or obese (BMI ? 30.0 kgm^-2) was calculated by using recommended international cutoffs. Among the participants, the mean BMIs were 21.5 (±4.1) in 1997 and 22.1(±4.2) in 2000, respectively. The percentages of underweight, overweight and adiposity were 22.3, 17.4, 2.6 in 1997 and 18.7, 23.6, 3.6 in 2000, respectively, which showed the underweight in the population was getting smaller and the numbers of overweight or obesity were increasing (Pearson chi²=19.0528; P<0.001).

We regrouped food into more than 12 groups, i.e. cereals, tofu (and its products), bean, potato, vegetables, fruit, meat (and its products), poultry (and its products), milk (and its products), egg (and its products), fish (aquatic products), alcohol (including beer, wine and all kinds of drinks containing alcohol) and others which were omitted in the analysis. Consumption of these food groups was compared over time, between genders, residence and among household income (see table 2).

Figure 2

Table 2: Comparisons of foods consumed among different subpopulations and time

In general, the average food intakes of the study population decreased over time (P<0.01) with the exception of increased intakes of bean, meat and milk (P <0.01), and of statistically unchanged intakes of tofu, poultry and eggs (P ?0.08). This would be more clarified if the consumed amounts of food groups in 1997 were taken as a baseline and then the comparison was made (figure 1). The transitions of both men's and women's food intakes were consistent with those of the whole study population, and almost so were transitions of food consumptions in the subpopulations in different residences or with different incomes. But there were a few distinctions. Instead of bean in urban and milk in low-and medium-income subpopulations, they consumed more tofu, and less milk and more eggs over time respectively.

Figure 3

Figure 1: Changes in Mean Food Consumption during 3 Years

There were not statistically significant differences in food intakes between male and female participants, except that females consumed more fruit and vegetables than males did. Urban subpopulation had few intakes of cereals, potatoes and vegetables, but more intakes from other food groups. People tended to consume more food, but not cereals and potatoes which were less eaten in meals, with increased family incomes.

In this study, energy, protein, fat, fibre and carbohydrate were taken into consideration to figure out their changes in people's dietary transitions over time and in different subpopulations. Of concern are these macronutrients which are the integrated responders of food value, in any studies related to nutrition. From 1997 to 2000, intakes of energy, proteins, fibres and carbohydrates reduced with the exception of fat which showed a bit elevation in average intake over time, but there was no statistical significance (P =0.62) (see table 3). In spite of decreases in intake, the intake proportions of energy, protein and fat from animal food rose. The results were mostly consistent when we explored the changes of macronutrients among subgroups with different genders, residences and family incomes. The urban participants reduced their intakes of fat and percentages of energy, proteins and fat from animal food, but the rural participants increased their fat intakes.

Figure 4

Table 3: Comparisons of macronutrients consumed among different subpopulations and time

Men's intakes of macronutrients were greater than women's (P <0.01). Urban people had more intakes of protein and fat, especially from animal food, than their rural peers, but the latter had greater intakes of carbohydrates and fibres. There were no statistical discrepancies in energy consumption in both surveys (P >0.18) and in fibre intakes in 2000 (P =0.26), between urban and rural subpopulations, but the former tended to have larger portions of energy from animal food (P<0.01). Higher income group ingested more energy, proteins and fat, whereas lower income group ingested more carbohydrates (P<0.01). There were no statistical differences in fibre intakes in both years (P>0.06) and in energy consumption in 1997 (P=0.39).

It could be found that the intakes of energy, protein, fibre and carbohydrate decreased (P=0.000), while that of vitamins increased (P=0.003) and those of fat and minerals remained almost the same in 2000, compared with those in1997. There existed, between genders, very significant differences in intakes of all macronutrients (P=0.000-0.002) except that of vitamins (P=0.121), that is, the male participants took more of these nutrients than their female partners did (table 3).

The discrepancy did also exist among participants between urban and rural residences, and in households with different income. In general, urban participants incepted more proteins, fat, minerals and vitamins (P=0.000-0.012), and few fibres and carbohydrates (P<0.001) from diet than rural ones did, while energy taken were almost the same between the two groups (P=0.129). It was found that the same rules applied to household income, when the correlation between the consumed macronutrients and the income was addressed. The subjects in higher income households took in more proteins, fat, and vitamins (P=0.000-0.002), and those in lower income households incepted more carbohydrates (P<0.001) from diet with getting nearly the same amount of energy and minerals as the other two groups (P=0.112-0.505).

As addressing the issue of energy sources, the description (table 4) was similar with percentage of energy supplied from protein, fat and carbohydrates being over 99.0%. The biggest proportion of energy, 53.4 – 70.1%, was derived from that of carbohydrates, while energy from proteins and fat composed 10.4 – 12.0% and 19.2 – 34.3%, respectively, of the total energy. It was clear that the energy from fat increased and that from carbohydrate decreased over time between genders, locations and with enhancement of income.

Figure 5

Table 4: Comparisons of foods consumed among different subpopulations and time

The research was completed with financial assistance from the US National Institutes of Health (NIH) (R01-HD30880 and R01-HD38700) and statistical analysis was supported by the Alumni Association of Sasakawa Medical Scholarship and Chronic disease Division, School of Medicine, the University of Queensland.

We appreciate the Institute of Nutrition and Food Safety, China CDC for the training and instruction and thank all health workers who took part in the surveys for their hard work.