This study shows that the prevalence of overweight (BMI ≥ 25 kg/m2) among Vietnamese adults aged 25-64 years in 2005 was almost twice as compared to 2000. The estimated average increase in prevalence of overweight amounted to 0.6%/year, which was almost twice as that of 0.3%/year over the period of 1992-2002 . Based on our data, the estimated number of subjects aged 25-64 years with overweight in 2005 amounted to about 2.6 million (BMI≥ 25 kg/m2) or 6.5 million (BMI≥ 23 kg/m2).
The pattern of overweight and obesity prevalence across population groups defined by age, gender and/or areas of residence were similar between the two periods (1992-2002 and 2000-2005), with a higher prevalence among women, urban residents, and older age-groups. The higher estimated increase in the nationwide prevalence of overweight and obesity and the bigger differences between urban and rural areas observed in 1992-2002 and 2000-2005 highlights the increasing problem of overweight in Vietnam, particularly in the urban areas.
The increasing trend of overweight and obesity is not only observed in Vietnam but also in other countries in the Asian region as well as worldwide. Fortunately, the magnitude of the problem in Vietnam is still much less than in many of these countries, such as 29% (1996-1997) in Hongkong  and 26.7% (1998) in Korea , which might be due to lower level of economic development in Vietnam. In Thailand, the National Health Examination Survey II showed a prevalence of overweight (BMI ≥ 25 kg/m2) in adults aged 20-59 years of 28%, with highest values in women (33.9%) and in the urban population (34.8%) . The problem of overweight and obesity is also rapidly increasing in China in all gender and age groups and in geographical areas, particularly in the urban area, with overall reported prevalence rates of 15% in 1992 and 22% in 2002 . The higher prevalence of overweight and obesity among women and urban residents in Thailand and China were similar to the situation found in the present study.
The prevalence of urban overweight (BMI ≥ 25 kg/m2) in our nation-wide samples is in concordance with previous smaller studies done in Hanoi and Ho Chi Minh City which reported overweight in 17.2% to 18.5% of adults aged 20-60 years [14–16]. In the past, the percentage of overweight in the adults, was also higher in urban areas than in rural areas (4.8% vs. 1.2% in 1998 and 9.6% vs. 3.5% in 2002) . This higher prevalence in the urban area may be explained by the faster economic growth. Over the period 1993-2004 Vietnam was considered to be one of the best performers in the world in terms of economic growth . As a result, poverty rates were halved in the same period. The general poverty rates decreased from 37% to 20%, while the food poverty rate went down from 13% to 7%. In parallel with the economic growth, the urbanization went up and the rate of urbanization is expected to remain above 3% per annum until 2020. It was estimated that the urban area accounted for 70% of the growth while containing only 25% of the population . We used food expenditure as a proxy indicator of income, and this was indeed also associated with higher overweight rates. But independent of food expenditure and other demographic factor overweight was still twice as high in urban areas compared to rural areas. In addition to higher income, urbanization has brought changes in lifestyle and food consumption habit which may also contribute to the higher prevalence of overweight. In urban areas of developing countries, food scarcity may no longer be the driving force behind energy intake. Instead, the availability of cheap, energy-dense foods (including those from street vendors and fast food restaurants) may facilitate the consumption of more calories. Widespread access to television would favour an indoor, sedentary lifestyle, further reducing the average daily energy expenditure . Those changes lead to an obesogenic environment.
The prevalences of overweight and obesity differed across the eight ecological regions but were all higher in 2005 as compared to 2000. There were several reasons for these differences and changes, but they were likely to be closely related to socio-economic status. Household poverty status significantly influences food consumption and food patterns . There were considerable disparities in regional poverty and poverty reduction . The South East region had the lowest poverty rate, which reduced from 12% in 1998 to 5% in 2004, and the same region was shown to have the highest and fastest increase in the prevalence of overweight and obesity. The Northern mountains (North East and North West), the North Central Coast and the Central Highland all have high poverty rates (50% and above in 1998 and still over 30% in 2004) and accordingly have lower prevalence rates of overweight and obesity.
In contrast with the increasing problem of overnutrition, undernutrition showed a decreasing trend. The estimated average annual reduction rate was 0.8%/year in the period 2000-2005, after an earlier reduction of underweight in adults from 31.2% in 1992 to 24.3% in 2002 . This reduction is probably thanks to the economic development and the considerable achievement in nutrition policy and intervention in Vietnam . In our study, food expenditure level was inversely associated with underweight. However, despite substantial improvements in rural living standards, poverty levels were still remarkably high in the rural area  in addition to the high prevalence of underweight. In the past decades, the available data showed that the prevalence of underweight and stunting among children aged under 5 years were very high, e.g. 51.5% and 59.7% in 1985, 44.9% and 46.9% in 1994, 31.9% and 34.8% in 2001 . In the earlier decades, a similar or even worse situation probably existed. This early childhood malnutrition situation maycontribute to the adult overweight nowadays . Maternal and child malnutrition control should be strengthened to reduce child undernutrition in order to prevent adulthood underweight and overweight, as well as the related chronic diseases.
In terms of age, the highest prevalence of overweight was observed in the age category of 45-54 years. Only for rural women, the age pattern was somewhat different, with the highest prevalence observed in the oldest category (55-64 yrs) in 2005. The general pattern of overweight and obesity by age agrees with survey findings from other Asian countries [28, 29]. The prevalence of underweight by age differed between urban and rural areas. In the urban area underweight was more prevalent in the youngest group of 25-34 years, while in the rural areas it was more prevalent in the oldest age groups. This may be explained by the immigration of young labour force from rural to urban areas due to rapid urbanization. Those young workers are mainly unskilled, having heavy manual works with low income. People who move from rural to urban areas usually lose the ability to grow their own food and thus become dependent for their calories on a cash market .
Interestingly, women were more likely to be both underweight and overweight as compared to men. This pattern was also reported among Indian women  and among Bangladesh rural and urban poor women . This may reflect various disadvantages which women face, such as poor nutrition care, heavy work load, physiological characteristics, and a high prevalence of early childhood undernutrition . Because of women's cyclical loss of iron and childbearing, their nutritional status is particularly vulnerable to deficiencies in diet, care, and health or sanitation services. Gender inequality exacerbates infectious diseases among the less affluent through the pathway of childhood undernutrition. At the same time, it exacerbates the new regime of chronic diseases among the relatively more affluent, possibly through a pathway that has come to be known as "the Barker hypothesis". Gender inequality thus leads to a double jeopardy, aggravating the double nutrition burden .
A steady shift is shown between underweight and overweight prevalence according to food expenditure, independent of age, gender and education. These findings confirm the association of economic growth with food consumption and nutritional status, particularly in developing countries where more than 50% of income is spent on food . Interestingly, the observed higher prevalence of overweight and lower prevalence of underweight in the highly educated group was accounted for by the other demographic factors and food expenditure in the logistic regression analysis.
Our results indicate that it is timely and necessary to take immediate action for effective control of underweight and early prevention of the spread of overweight and obesity problem in Vietnam. However, programs and interventions should take the double nutrition burden into consideration to avoid sharpening the severity of underweight when spending efforts in reducing overweight. Appropriate interventions are needed for specific population subgroups. Some important interventions for reducing the rate of undernutrition may also be beneficial in terms of reducing the burden of obesity are promoting breast-feeding, improving nutritional status of women of reproductive age, and reducing the rates of fetal growth retardation  and low birth weight . Improving the obesogenic environment in urban area by nutritional education, information and communication for promoting healthy eating and physical activity and monitoring food market should be intensively implemented in order to reducing underweight and preventing overweight . Reducing gender inequalities should be paid attention in improving double burden of nutrition among women in particular and in the whole population in general . Promoting household food production with the existing successful VAC model (i.e. the Vegetation, Aquaculture and Cattle-breeding model), particularly encouraging small-scale farmers and, especially women, to grow and utilize a wide variety of food crops toward improving household food security and dietary diversity can be an effective way for combating double nutrition burden in rural area .
The present study has some limitations. Data on diet, physical activity and smoking were not available for both datasets and thus could not be adjusted for. However, with data from the two largest recent nationally representative nutrition surveys, conducted by well-trained personnel according to a standardized protocol, the shift in the double burden of malnutrition was clearly demonstrated.