To the Editor: In the report on the spread of obesity in a socialnetwork, by Christakis and Fowler, and the accompanying editorialby Barabási (July 26 issue),1,2 variables of social andeconomic status (SES) are not mentioned. Social networks conceala high degree of social homogeneity — that is, peopleare more likely to network with people at similar income andeducational levels — and we suspect that they may operatedifferently in different contexts. For example, the way a socialnetwork affects the incidence of obesity in a community wheremost people have a low educational level and cannot afford ahealthy diet is likely to differ from the way it affects anaffluent and well-informed community. In 2000, obesity in theUnited States was reported to be highest among black women andpeople who had not completed high school,3 and among black women,a low socioeconomic position predicts an early onset of obesity.4SES variables should be included in any interpretive model,and public health policies for the prevention of obesity shouldtake into account the SES characteristics of the communitiesthey are supposed to benefit.5
Christakis NA, Fowler JH. The spread of obesity in a large social network over 32 years. N Engl J Med 2007;357:370-379. [Free Full Text]
Barabási A-L. Network medicine -- from obesity to the "diseasesome." N Engl J Med 2007;357:404-407. [Free Full Text]
Freedman DS, Khan LK, Serdula MK, Galuska DA, Dietz WH. Trends and correlates of class 3 obesity in the United States from 1990 through 2000. JAMA 2002;288:1758-1761. [Free Full Text]
Bennett GG, Wolin KY, James SA. Lifecourse socioeconomic position and weight change among blacks: the Pitt County study. Obesity (Silver Spring) 2007;15:172-81.
Stafford M, Cummins S, Ellaway A, Sacker A, Wiggins RD, Macintyre S. Pathways to obesity: identifying local, modifiable determinants of physical activity and diet. Soc Sci Med (in press).
To the Editor: Christakis and Fowler describe how weight gainoccurs in social clusters and stress that people are influencedby the appearance and behaviors of others. In behavioral economics,this mechanism is referred to as "anchoring," meaning that judgmentsare based not on absolute values but on comparison with implicitreference points ("anchors").1 We collected data on the senseof urgency with respect to weight control that show how subtleand powerful anchoring can be. With the use of different anchorsfor the same scenario, 154 subjects were asked to judge theimportance of taking action concerning their weight if 35% ofthe population was heavier than they themselves were (scenarioA) versus the importance of taking action if 65% of the populationwas thinner than they were (scenario B). The importance of weightcontrol was judged to be significantly lower (P<0.001 bya paired t-test) in scenario A, in which subjects implicitlycompared themselves with heavier people. This finding illustrateshow anchoring may have contributed to the clustering of obesitybut also raises concern that increasing weight in the generalpopulation may increase our anchors — the implicit referencepoints for what we perceive as normal. This could turn overweightin society into a self-reinforcing process.
Stefan Knecht, M.D. Julia Reinholz, Ph.D. Peter Kenning, Ph.D. University of Muenster 48129 Muenster, Germany knecht{at}uni-muenster.de
References
Tversky A, Kahneman D. Judgment under uncertainty: heuristics and biases. Science 1974;185:1124-1131. [Free Full Text]
To the Editor: I believe that the article on the spread of obesityin a large social network has important public health implications.It is also of concern for those of us involved in health technologyassessment and firm believers in the advantages of randomized,controlled trials.
The study focused on obesity, but the findings probably holdfor nearly all types of lifestyle interventions. By using theconcept of etiologic fraction and using data on smoking behavioramong smokers and nonsmokers, I previously estimated that forevery 10 people who stop smoking, there will be another 2 whoin the long run do not smoke as a consequence,1 an indirecteffect of 20%. Trends toward a decline in smoking around theworld have surely been reinforced by social diffusion.
It is obvious that randomized, controlled trials with individualallocations do not take these indirect social effects into considerationand thereby underestimate the effects. Health technology assessmentshave so far not considered these indirect effects, and organizationssuch as the one I represent may have somewhat hampered the developmentof lifestyle interventions.
Måns Rosén, Ph.D. Swedish Council on Technology Assessment in Health Care SE-11486 Stockholm, Sweden rosen{at}sbu.se
References
Rosén M. On randomized controlled trials and lifestyle interventions. Int J Epidemiol 1989;18:993-994. [Free Full Text]
The authors reply: We investigated the interpersonal spreadof obesity as a possible factor in the obesity epidemic. Tamburliniand Cattaneo are concerned about the role of SES. Our reportedanalyses did account for education. Additional analyses (notreported in the article) also accounted for income. Adjustmentfor either or both of these measures did not alter our findingthat weight gain in one person was associated with weight gainin others. Moreover, we found no significant difference in thelikelihood of spread according to whether the level of educationwas high or low.
Knecht et al. provide very nice evidence for the effects ofboth "framing" and "anchoring." The experiment they describeshows the importance of how framing the context of an individualrespondent affects responses (i.e., the otherwise equivalentdescription of the group as being composed of 35% of peoplewho are heavier than the respondent versus 65% who are lighter).Even more telling results, we suspect, would arise if the scenarioswere manipulated so that 45% or 55% or 65% of the group wasdescribed as being lighter than the respondent.1 As suggestedin our report, we suspect that increasing weight gain in a populationmay affect reference points for what people perceive as "normal"weight,2 and this might be one mechanism for the interpersonalspread of obesity. Moreover, personal assessments of weightvary, perhaps in keeping with the sociodemographic group towhich a person belongs.3
We agree with Rosén both with respect to the likely relevanceof our findings to other health behaviors (e.g., smoking) andalso with respect to the relevance of our findings for technologyassessment. The existence of interpersonal health effects hassubstantial implications for the analysis of health policy,since outcomes in individuals to whom a person is connectedshould, in many situations, be included in enumerating costsand benefits of interventions.4 This is not limited solely tolifestyle interventions: replacing a hip, preventing a stroke,or curing a cataract in one person may reduce the disabilitynot only of that person but also of his or her spouse, for example.
Social networks are relevant to health and health care. Onepoint worth emphasizing, however, is that social support iswell known to be important.5 It is therefore unlikely that severingties with people on the basis of any of their particular traits— as some have supposed that our results might suggest— would necessarily be beneficial.
Nicholas A. Christakis, M.D., Ph.D. Harvard Medical School Boston, MA 02115 christakis{at}hcp.med.harvard.edu
James H. Fowler, Ph.D. University of California, San Diego La Jolla, CA 92093
References
Solnick SJ, Hemenway D. Is more always better? A survey of positional concerns. J Econ Behav Organ 1998;37:373-383. [CrossRef][Web of Science]
Burke MA, Heiland MA. Social dynamics of obesity. Econ Inq 2007;45:571-591. [Web of Science]
Chang VW, Christakis NA. Self-perception of weight appropriateness in the United States. Am J Prev Med 2003;24:332-339. [CrossRef][Web of Science][Medline]
Christakis NA. Social networks and collateral health. BMJ 2004;329:184-185. [Free Full Text]
Berkman LF, Syme SL. Social networks, host resistance, and mortality: a nine-year follow-up study of Alameda County residents. Am J Epidemiol 1979;109:186-204. [Free Full Text]
Previous
