An article in the New England Journal of Medicine, July 21 issue (The spread of obesity in a large social network over 32 years, N.A. Christakis and J.H. Fowler, pp. 370-379, 2007) dropped like a bombshell into the medical community, exploding many long-held assumptions and beliefs.
What was the question this research attempted to answer?
To quote the authors:
“The prevalence of obesity has increased substantially over the past 30 years. We performed a quantitative analysis of the nature and extent of the person-to-person spread of obesity as a possible factor contributing to the obesity epidemic.”
The italics are mine, to emphasize the fact that the authors set out to explore a quite revolutionary concept: obesity, like infections, spreads by person-to-person contact.
How the study was done
The authors took advantage of a famous study run by the Federal Government in Framingham, Mass. Basically, the study consisted of medical researchers taking exhaustive medical history, blood tests, physical examination, and x-rays on a massive scale; most of the townspeople volunteered for the study, and since its inception in 1970 to date the original subjects, their children and grandchildren, have been subjected to repeated periodic examinations. The main objective of the Framingham Heart Study was, as the name implies, to identify all the factors that contribute to heart disease. But because of the extensive database, including data that could be important in other diseases, such as smoking and lung cancer, this study became a rich trove of information for medical researchers in many diverse fields.
For this study, an evaluation was made of a densely interconnected social network of 12,067 people assessed repeatedly from 1971 to 2003. The BMI (body mass index) was available for all subjects. Obesity was defined as BMI ≥ 30. The authors used statistical models that had been developed to analyze other network systems (for instance, person-to-person spread of the influenza virus, or effect of peer influence on alcoholism in schools) to examine whether weight gain in one person was associated with weight gain in his or her friends, siblings, spouse, and neighbors. Before we go to results, here are a couple of definitions used in network studies:
Ego: The person whose behavior is being analyzed.
Alter: A person connected to the ego who may influence the behavior of the ego.
And the results were…
Surprising would be an understatement. Here is a summary:
- If an ego stated that an alter was his or her friend, the ego’s chances of becoming obese increased by 57% if the alter became obese. The type of friendship appears to be important. Between mutual friends (both ego and alter state that they are friends), the ego’s risk of obesity increased by 171% (!) if an alter became obese. In other words, if your friend becomes obese, you might as well just give up and start gorging. In contrast, there was no statistically meaningful relationship when the friendship was perceived by the alter but not by the ego.
- How can we talk about any subject without invoking sex? So here it is: When the sample was restricted to same-sex friendship, the probability of obesity in the ego increased by 71% if the alter became obese. But that’s not all. A male-male friendship increased the ego’s chances of becoming obese by 100% if the alter became obese, whereas the female-to-female spread of obesity was not significant.
- What is the “reach” of this relationship of friendship and obesity? When an ego becomes obese, his/her friend (or alter) has a 45% higher chance of becoming obese (as compared to a random network). A friend of the friend (2 degrees of separation from the ego) has a 20% increased probability of becoming obese, and a friend of a friend of a friend of the ego (3 degrees of separation from the ego) has a 10% increased probability of becoming obese. At four degrees of separation from the ego, there is no increase in probability of becoming obese (thank God!)
- Geographic distance made no difference. A distant alter had just as strong an influence on the ego as the friend next door. Reminiscent of chaos theory predicting that a butterfly in California may cause a hurricane in Alabama.
- Neighbors, on the other hand, had no influence on the ego’s obesity.
- How about family relationships? Surprisingly weaker than friendships. Obese husband effect on wife—44% increase in probability; Obese wife effect on husband—37% increased likelihood of obesity. Siblings of the same sex—55%; Siblings of the opposite sex—27%. But caution, before you blame your same-sex sibling for your obesity, the difference between the 55% and 27% increase in likelihood of obesity was not statistically significant.
So what does it all mean?
I must admit, I am baffled. The obvious conclusion is that obesity can spread just like any other infectious disease. Amazing, if true.
When I read claims that are in apparent clash with reality as I know it, or that stretch my credulity, I approach such claims with an increased index of suspicion. Space does not permit an in-depth analysis of the work, but here is one objection: The social relationship (ego-alter) was determined on the basis of 0.7 relationships per ego. Does this truly reflect the social milieu of an individual? Would say 3 or 4 social relationships per person (much more realistic) completely negate this finding? In other words, despite the enormous number of people participating, the sample of total relationships is just too small. That could lead to spurious results. For example, if you had a sample of 2 and you added another 1, that addition can have a large influence on the result. If you had a sample of 100 and added another 1, the likely influence is quite small. Here, the sample size is 0.7! adding 1 or more relationships could upend the analysis and its conclusions.
So am I skeptical of the study? Yes and no. And the reason I hope I am wrong will be the subject of another posting.