by Wei Perng, PhD
Obesity is not transmitted through direct contact or biological fluids the same way that a virus is, but it is socially contagious via behaviors and social norms.
You can imagine that if your work buddies attend Monday pizza, beer, and wings night at the local bar, you might join them. This behavior could lead to weight gain if it becomes habitual. Individuals within a social group might also serve as a standard or norm by which others compare themselves. The 10-lb weight gain by your roommate gained over Christmas break could decrease your motivation to address the extra 5 that you are sporting. Some individuals within a network may be more influential than others. For example, in Chinese work cultures with rigid social hierarchies, the CEO’s favorite hobbies are quickly adopted by his underlings in effort to gain his/her favor. In such a network, the CEO would be considered a key player from which social norms and behaviors spread. Theoretically, interventions targeted at highly-connected persons could appreciably alter behaviors, and thus health outcomes of their social network.
Intervene on one person’s risk of obesity and intervene on the risk of 20 others? Music to my ears! Unfortunately, social interactions are not so cut-and-dry. A recent simulation study based on data from the Health Surveys for England 1999 and 2004 and Christakis and Fowler’s study suggests that obesity interventions aimed at well-connected individuals within a network are not more effective than implementing the intervention at random. The authors speculated that this could be due to a “rebound effect” where the target individual who receives the intervention is subsequently re-exposed to the disease. In other words, even if well-connected individuals lose weight, their continued contact with the same network could promote weight re-gain. It may also be difficult to maintain weight loss because of the biological mechanisms that decrease metabolic rate and increase hunger.
These findings may seem discouraging, but there are additional network measures to consider. The spread of behaviors through a network is complex and depends on a number of factors. In addition to an individual’s connectivity, his or her relative importance within the network could influence behavioral transmission. Let’s take the example of the all-powerful CEO in China. If he enjoys golf, his employee may pick up golf to please him, but it is unlikely that the employee’s hobbies will influence the CEO’s behavior, making the transmission of golf-playing behavior unidirectional. The concept of directionality between individuals could have important implications for phenomena like the rebound effect observed in the above-mentioned study. Other key network metrics to consider include the structure of cliques, which are clusters or groups of similar individuals within a network, and whether or not there are bridges between cliques. One might imagine that transmission could end in a clique with no links out to other parts of the network.
And as always, there is research yet to be done. Chronic diseases like obesity tend to develop slowly, and data on interpersonal relationships are not typically collected (at least not yet!). This lack of data limits the ability to test these concepts. However, simulation models with input from real-world data can provide insight into emergent social phenomena that may otherwise take months or years to play out. Simulations still operate on some rigid and, at times, unrealistic assumptions that limit validity (e.g. lack of social or residential mobility) and, thus, aren’t perfect. Future work in social epidemiology and refinement of network analysis techniques will improve design and interpretation of these types of studies.