Multilevel models are not new in public health; the concept stems from socioecological theories (6) that emphasize the importance of social and environmental factors in determining human behavior and health outcomes. However, the model has been interpreted to describe ecologic layers without elaborating on multiple sectors operating at multiple levels or including bidirectional interactions of factors (7). Glass and McAtee (8) present a multilevel model that is useful to address the complex, interacting contexts for obesity prevention. This model (Figure 1), which was a key focal point for the international conference, integrates biological (genes, cells, and organs) and socioenvironmental (economics, culture, social networks, and features of the physical environment) influences on behaviors such as eating and physical activity. Time, on the horizontal axis, is in the context of life course (conception to death) at the individual level or social change at the population level. The vertical axis depicts a nested hierarchy of systems including biological, social, and environmental influences (8). This model shows that the behaviors leading to health outcomes, not just health outcomes per se, are influenced by biological or socioenvironmental factors.

The model is consistent with economics and psychology in that people are assumed to engage in behaviors based on preferences and attitudes. It becomes multilevel in that a person is constrained by factors that exert regulatory control on those behaviors. For example, food choices are made not just on the basis of preferences but also on the basis of the price of food, the cultural meaning of food, the availability of food, and the biological responses to the reward value of food. The distribution of these parameters constitutes a behavioral niche or landscape, to which the person must adapt and respond according to particular goals and intentions. The movement in time of higher or lower rates of obesity is, therefore, the result of multiply-dependent and interlocking systems. There are 4 possible implications. First, a single cause of the obesity epidemic is unlikely. Second, the processes that give rise to increasing average body size probably involve combinations of factors at multiple levels of influence. Third, small changes in 1 or more key factors may have large and potentially nonlinear influences on distribution of body weight. Finally, both socioenvironmental factors and biological processes are involved in the expression of human behavior.

One problem with building a systems-oriented, multilevel framework for obesity is that key influences in the physical or economic environment may not fit conventional definitions of causes. Glass and McAtee contend that social factors, such as social inequity and poverty, are difficult to study from a traditional epidemiologic standpoint, in part because they do not fit the definition of a causal risk factor (8). An alternative view of these variables is to define them as risk regulators, or dynamic components of interconnected systems that influence obesity-related behaviors from the personal level to the public policy level (8). Systems of food distribution alter the probabilities at a population level that these causes will align in ways that lead to different rates of obesity (9). The search for a set of key risk regulators provides greater room to consider the social, physical, cultural, and economic environments that influence obesity.

The concept of risk regulators also may help overcome some of the disadvantages of conventional socio-ecological models, namely the lack of clarity on what is most important, where the key drivers are located, or what the optimal intervention points are. The multiple levels (individual vs community) require a bridging structure, which act as conduits between macro-level forces and the factors in the local environment that govern eating and activity. The temporally and spatially distal forces that operate at the macro level cascade through organizations, through systems of food distribution, through policies and pricing, and eventually shape the reality that people perceive in their lives. Examples of the bridging in the case of obesity could be cultural norms, social networks, local food availability, food prices and taxes, physical activity amenities, psychosocial stress, or economic insecurity. These might act through neurologic or epigenetic  regulatory pathways to affect behavior and to generate feedback loops higher in the system. Epigenetic pathways are phenotypic differences between individuals that are not a result of genetic composition per se but a result of alterations in genetic expression through the silencing of genes or interference with genetic transcription.

Diverse sectors of society operate at different levels to influence population energy balance (Figure 2) (2,10). Factors can range from the individual level to the international level, and the sectors of influence include education, agriculture, transportation, urban developments, and media, among others, in addition to the health sector. Research that cuts across these different levels and sectors can be undertaken (Figure 2).

The next-generation interventions for obesity should start at the community level or higher, with multiple stakeholders that connect people, families, schools, government, and the private sector. Intervention activities should include not only educational schemes but also environmental changes to shift norms and enable the adoption of healthy behaviors within everyday life. The family, schools, primary care settings, and municipalities can be targeted simultaneously as catchment sites to interface with children and parents. Media organizations and businesses (eg, food manufacturers, retailers, supermarkets, the transportation industry) can also help shift norms, effectively contributing to both the supply and demand sides of the energy balance equation.

Much can be learned from the North Karelia Project in Finland from the 1970s through the 1990s, where that country's public health agency transformed the lifestyle pattern of Finnish communities to reduce smoking rates and improve dietary practices. The Finland project did not rely exclusively on individually focused educational interventions. The government created incentives for farmers to switch from meat to fruit and vegetable production, and worked through social networks by using community organizations. There were also efforts to use regulatory changes to influence the nutrient content of food (eg, requiring sausage makers to lower the fat content of their products across the entire market). The result was a greater than 50% reduction in coronary heart disease mortality, as well as reductions in stroke, cancer, and other diseases, in the entire country within 20 years (39).

Although research on multilevel interventions advances slowly, actions are already being taken in many US and international communities. This parallel movement at the grassroots level needs to be taken advantage of with rigorous evaluations to determine the effect of community-initiated interventions (40). There is little research on the dissemination and diffusion requirements of multilevel interventions. As intervention and evaluation research continue, dissemination must be part of the strategic effort.

Multilevel research and interventions cannot be conducted or sustained if the agenda does not include a strong focus on building coalitions across societal sectors and increasing the capacity to tackle obesity (41). Specifically, public-private partnerships, leadership of national governments, and training of future multilevel researchers and policy makers are warranted.

In the world, approximately 22 million children younger than age 5 years are overweight. By 2015, an estimated 2.3 billion people aged 15 or older will be overweight and 700 million will be obese worldwide (49). By 2010, cardiovascular disease will be the leading cause of death in developing countries, and by 2030 more than 280 million people in developing countries will have type 2 diabetes (49). Key drivers of these numbers are transnational (globalization of markets and media, urbanization, trade, economic growth, food availability, marketing) (Figure 2), requiring a global perspective to address obesity. The increasing health effects related to obesity will pose substantial economic challenges as a result of cost and insufficient infrastructure in the world's health care systems (50). An unhealthy population leads to reduced economic productivity, which further exacerbates morbidity and mortality.

Experiences in the United States and other developed nations may serve as a starting point for understanding and combating obesity in developing countries. Nevertheless, factors may not all be equally relevant in different countries, and environmental, cultural, and sociopolitical influences within countries determine what types of solutions will be feasible and effective. More international research is needed to understand these differences. For example, the Seven Countries Study (51) on cardiovascular health provided great insight into the role of population-level variations in diet in heart disease risk. Although ecologic correlations are weak for supporting causal inference, this study was groundbreaking in showing population-level influences on disease rates and on preventive strategies. Obesity research can carry on these lessons. International research that capitalizes on the contrast on either differing obesity rates or differing socioenvironmental characteristics across contexts can be especially illuminating.

Current levels of obesity reflect complex social changes and biological susceptibilities, and their interactions, during the last 40 years. Individual behaviors such as eating and physical activity do not occur in a vacuum; rather, they are influenced by socioenvironmental factors and by powerful biological processes. Behavior change cannot be sustained if these drivers of behavior are not considered. A systems-oriented, multilevel framework encompassing science and research capacity-building is the way to generate solutions that deal with the complex system in which obesity arises. A multilevel research agenda is cross-disciplinary, bringing together expertise in traditionally disparate fields to pose cross-disciplinary hypotheses and to test those hypotheses collectively. The agenda also would extend conventional research boundaries by tackling structural aspects of the social, physical, and policy environment that affect obesity. Capacity building for global research is critical for sustaining a multilevel research agenda for obesity and chronic disease prevention.

Ultimately, interventions should strive to make healthy eating and physical activity a natural and easy way of life. Using the framework discussed here, one approaches the problem by first looking at the whole picture rather than immediately zeroing in on a detail. Having a view, even if not a full understanding, of the relations among factors that regulate energy balance, across individuals as well as populations, allows one to simultaneously consider multiple leverage points in the system within which obesity occurs that can or need to be modified to yield the desired outcomes (52). Focused studies can then be designed to confirm and quantify these relationships and to test their effects. By nature, this systems-oriented, multilevel approach is solution-oriented, underlining the philosophy that mechanistic and intervention studies are worthy only if they can improve population health in a sustainable way. Given where we are today, faced with the continued lack of effective and sustainable prevention strategies, there is a critical need to implement this multilevel approach. We can do this by extending the boundaries of biomedical research to fill the gaps across all the disciplines relevant to obesity, from biological and behavioral sciences to social and policy research.