Fermat's Library | The Spread of Obesity in a Large Social Network over 32 Years annotated/explained version.

This 2007 study made a big splash when it came out and has now be...

> Nicholas A. Christakis is a sociologist and a physician known for...

> The **Framingham Heart Study** is a long-term, ongoing cardiovasc...

> Network phenomena appear to be relevant to the biologic and behav...

Todays numbers look a bit different (even higher). The percentage o...

Many other potential confounders not mentioned: environmental affec...

This data essentially unlocks this study: >These sheets contain v...

Original Kamada-Kawai algorithm paper: https://citeseerx.ist.psu.ed...

> Our study suggests that obesity may spread in social networks in ...

special article

T h e

n e w e n gl a n d j o u r n a l

o f

m e d i c i n e

n engl j med 357;4 www.nejm.org july 2 6, 2007

370

The Spread of Obesity in a Large Social

Network over 32 Years

Nicholas A. Christakis, M.D., Ph.D., M.P.H., and James H. Fowler, Ph.D.

From the Department of Health Care

Policy, Harvard Medical School, Boston

(N.A.C.); the Department of Medicine, Mt.

Auburn Hospital, Cambridge, MA (N.A.C.);

the Department of Sociology, Harvard Uni-

versity, Cambridge, MA (N.A.C.); and the

Department of Political Science, Univer-

sity of California, San Diego, San Diego

(J.H.F.). Address reprint requests to Dr.

Christakis at the Department of Health

Care Policy, Harvard Medical School, 180

Longwood Ave., Boston, MA 02115, or at

christakis@hcp.med.harvard.edu.

N Engl J Med 2007;357:370-9.

A B S T R AC T

Background

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.

Methods

We evaluated a densely interconnected social network of 12,067 people assessed

repeatedly from 1971 to 2003 as part of the Framingham Heart Study. The body-

mass index was available for all subjects. We used longitudinal statistical models to

examine whether weight gain in one person was associated with weight gain in his

or her friends, siblings, spouse, and neighbors.

Results

Discernible clusters of obese persons (body-mass index [the weight in kilograms

divided by the square of the height in meters], ≥30) were present in the network at

all time points, and the clusters extended to three degrees of separation. These

clusters did not appear to be solely attributable to the selective formation of social

ties among obese persons. A person’s chances of becoming obese increased by 57%

(95% confidence interval [CI], 6 to 123) if he or she had a friend who became obese

in a given interval. Among pairs of adult siblings, if one sibling became obese, the

chance that the other would become obese increased by 40% (95% CI, 21 to 60). If

one spouse became obese, the likelihood that the other spouse would become

obese increased by 37% (95% CI, 7 to 73). These effects were not seen among neigh-

bors in the immediate geographic location. Persons of the same sex had relatively

greater influence on each other than those of the opposite sex. The spread of smok-

ing cessation did not account for the spread of obesity in the network.

Conclusions

Network phenomena appear to be relevant to the biologic and behavioral trait of

obesity, and obesity appears to spread through social ties. These findings have

implications for clinical and public health interventions.

The New England Journal of Medicine

Downloaded from nejm.org on April 16, 2019. For personal use only. No other uses without permission.

The Spr e ad of Obesit y in a Large Social Net wor k Over 32 Ye a r s

n engl j med 357;4 www.nejm.org july 2 6, 2007

371

he prevalence of obesity has in-

creased from 23% to 31% over the recent

past in the United States, and 66% of adults

are overweight.

1,2

Proposed explanations for the

obesity epidemic include societal changes that

promote both inactivity and food consumption.

The fact that the increase in obesity during this

period cannot be explained by genetics

4,5

and

has occurred among all socioeconomic groups

provides support for a broad set of social and

environmental explanations. Since diverse phe-

nomena can spread within social networks,

6-10

we conducted a study to determine whether obe-

sity might also spread from person to person,

possibly contributing to the epidemic, and if so,

how the spread might occur.

Whereas obesity has been stigmatized in the

past, attitudes may be changing.

11,12

To the extent

that obesity is a product of voluntary choices or

behaviors, the fact that people are embedded in

social networks and are influenced by the evident

appearance and behaviors of those around them

suggests that weight gain in one person might

influence weight gain in others. Having obese

social contacts might change a person’s tolerance

for being obese or might influence his or her

adoption of specific behaviors (e.g., smoking, eat-

ing, and exercising). In addition to such strictly

social mechanisms, it is plausible that physiolog-

ical imitation might occur; areas of the brain

that correspond to actions such as eating food

may be stimulated if these actions are observed

in others.

Even infectious causes of obesity are

conceivable.

14,15

We evaluated a network of 12,067 people who

underwent repeated measurements over a period

of 32 years. We examined several aspects of the

spread of obesity, including the existence of clus-

ters of obese persons within the network, the

association between one person’s weight gain and

weight gain among his or her social contacts, the

dependence of this association on the nature of

the social ties (e.g., ties between friends of differ-

ent kinds, siblings, spouses, and neighbors), and

the influence of sex, smoking behavior, and geo-

graphic distance between the domiciles of per-

sons in the social network.

M e t h od s

Source Data

The Framingham Heart Study was initiated in

1948, when 5209 people were enrolled in the orig-

inal cohort.

The Framingham Offspring Study

began in 1971, when most of the children of

members of the original cohort and their spouses

were enrolled in the offspring cohort.

There has

been almost no loss to follow-up other than death

in this cohort of 5124 people; only 10 people left

the study. In 2002, the third-generation cohort,

consisting of 4095 children of the offspring co-

hort, was initiated. All participants undergo phys-

ical examinations (including measurements of

height and weight) and complete written ques-

tionnaires at regular intervals.

Net work Ascertainment

For our study, we used the offspring cohort as

the source of 5124 key subjects, or “egos,” as they

are called in social-network analysis. Any persons

to whom the egos are linked — in any of the

Framingham Heart Study cohorts — can, how-

ever, serve as “alters.” Overall, 12,067 living egos

and alters were connected at some point during

the study period (1971 to 2003).

To create the network data set, we entered in-

formation about the offspring cohort into a com-

puter. This information was derived from archived,

handwritten administrative tracking sheets that

had been used since 1971 to identify people close

Glossary

Ego: The person whose behavior is being analyzed.

Alter: A person connected to the ego who may influence the behavior of the ego.

Node: An object that may or may not be connected to other objects in a net-

work. In this study, nodes represent people in the Framingham Heart

Study cohorts.

Tie: A connection between two nodes that can be either one-way (directed)

or two-way (bilateral). In this study, all family ties (e.g., between siblings

and parents) as well as marital ties are bilateral, but friendship ties are di-

rectional since a subject may identify someone as a friend who does not

necessarily identify that person as a friend in return.

Degree of separation: The social distance between two people as measured

by the smallest number of intermediaries between an ego and other

members of the network. For a given ego, alters are degree 1, since they

are directly connected to the ego. Nodes that are connected to the alters

but not to the ego are degree 2 (alters’ alters). Nodes that are connected

to the alters’ alters but not to the ego are degree 3, and so on.

Homophily: The tendency for people to choose relationships with people

who have similar attributes.

Induction: The spread of a behavior or trait from one person to another.

Cluster: A group of nodes, each of which is connected to at least one other

node in the group.

Connected component: Part of a social network in which all persons have

a social tie to at least one other person and no person is connected to a

member of any other component of the network.

The New England Journal of Medicine

Downloaded from nejm.org on April 16, 2019. For personal use only. No other uses without permission.

T h e

n e w e n gl a n d j o u r n a l

o f

m e d i c i n e

n engl j med 357;4 www.nejm.org july 2 6, 2007

372

to the study participants to facilitate follow-up.

These sheets contain valuable, previously unused

social-network information because they system-

atically and comprehensively identify relatives and

friends named by the ego. The tracking sheets

provide complete information about all first-order

relatives (parents, spouses, siblings, and children),

whether they are alive or dead, and at least one

“close friend” at each of seven examinations be-

tween 1971 and 2003. The examinations took

place during 3-year periods centered in 1973, 1981,

1985, 1989, 1992, 1997, and 1999. Detailed home

addresses were also recorded at each time point;

we used this information to calculate the geo-

graphic distance between people.

Many of the named alters on these sheets also

were members of Framingham Heart Study co-

horts. This newly computerized database thus

identifies the network links among participants

at each examination and longitudinally from one

examination to the next. As a person’s family

changed because of birth, death, marriage, or

divorce, and as contacts changed because of resi-

dential moves or new friendships, this informa-

tion was recorded. Furthermore, dates of birth

and death were available from separate Framing-

ham Heart Study files.

Overall, there were 38,611 observed social and

family ties to the 5124 egos, yielding an average

of 7.5 ties per ego (not including neighbors). For

example, 83% of the spouses of egos were direct-

ly and repeatedly observed at the time of exami-

nation, and 87% percent of egos with siblings

had at least one sibling in the network. For 10%

of the egos, an immediate neighbor also partici-

pated in the study; more expansive definitions of

neighbors yielded similar results.

A total of 45% of the 5124 egos were con-

nected through friendship to another person in

the network. There were 3604 unique, observed

friendships, for an average of 0.7 friendship tie per

ego. Because friendship identifications are direc-

tional, we studied three different kinds of friend-

ships: an “ego-perceived friendship,” in which an

ego identifies an alter as a friend; an “alter-per-

ceived friendship,” in which an alter identifies

an ego as a friend; and a “mutual friendship,” in

which the identification is reciprocal. We hypoth-

esized that a friend’s social influence on an ego

would be affected by the type of friendship, with

the strongest effects occurring in mutual friend-

ships, followed by ego-perceived friendships, fol-

lowed by alter-perceived friendships. Our reason-

ing was that the person making the identification

esteems the other person and may wish to emu-

late him or her.

We included only persons older than 21 years

of age at any observation point and subsequently.

At the inception of the study, 53% of the egos

were women, the mean age of the egos was 38

years (range, 21 to 70), and their mean educa-

tional level was 13.6 years (range, no education

to ≥17 years of education).

The study data are available from the Framing-

ham Heart Study. The study was approved by the

institutional review board at Harvard Medical

School; all subjects provided written informed

consent.

Statis tic al Analysis

We graphed the network with the use of the

Kamada–Kawai

algorithm in Pajek software.

We generated videos of the network by means of

the Social Network Image Animator (known as

SoNIA).

We examined whether our data con-

formed to theoretical network models such as the

small-world,

scale-free,

and hierarchical types

(see the Supplementary Appendix, available with

the full text of this article at www.nejm.org).

We defined obesity as a body-mass index (the

weight in kilograms divided by the square of

the height in meters) of 30 or more. Analyses in

which the body-mass index was a continuous var-

iable did not yield different results.

We considered three explanations for the clus-

tering of obese people. First, egos might choose

to associate with like alters (“homophily”).

21,23,24

Second, egos and alters might share attributes

or jointly experience unobserved contemporane-

ous events that cause their weight to vary at the

same time (confounding). Third, alters might

exert social influence or peer effects on egos

(“induction”). Distinguishing the interpersonal

induction of obesity from homophily requires

dynamic, longitudinal network information about

the emergence of ties between people (“nodes”)

in a network and also about the attributes of

nodes (i.e., repeated measures of the body-mass

index).

The basic statistical analysis involved the spec-

ification of longitudinal logistic-regression mod-

els in which the ego’s obesity status at any given

examination or time point (t + 1) was a function of

various attributes, such as the ego’s age, sex, and

The New England Journal of Medicine

Downloaded from nejm.org on April 16, 2019. For personal use only. No other uses without permission.

The Spr e ad of Obesit y in a Large Social Net wor k Over 32 Ye a r s

n engl j med 357;4 www.nejm.org july 2 6, 2007

373

educational level; the ego’s obesity status at the

previous time point (t); and most pertinent, the

alter’s obesity status at times t and t + 1.

used generalized estimating equations to account

for multiple observations of the same ego across

examinations and across ego–alter pairs.

assumed an independent working correlation

structure for the clusters.

26,27

The use of a time-lagged dependent variable

(lagged to the previous examination) eliminated

serial correlation in the errors (evaluated with a

Lagrange multiplier test

) and also substantial-

ly controlled for the ego’s genetic endowment and

any intrinsic, stable predisposition to obesity. The

use of a lagged independent variable for an alter’s

weight status controlled for homophily.

The

key variable of interest was an alter’s obesity at

time t + 1. A significant coefficient for this vari-

able would suggest either that an alter’s weight

affected an ego’s weight or that an ego and an

alter experienced contemporaneous events affect-

ing both their weights. We estimated these mod-

els in varied ego–alter pair types.

To evaluate the possibility that omitted vari-

ables or unobserved events might explain the as-

sociations, we examined how the type or direc-

tion of the social relationship between the ego

and the alter affected the association between the

ego’s obesity and the alter’s obesity. For example,

if unobserved factors drove the association be-

tween the ego’s obesity and the alter’s obesity,

then the directionality of friendship should not

have been relevant.

We evaluated the role of a possible spread in

smoking-cessation behavior as a contributor to

the spread of obesity by adding variables for the

smoking status of egos and alters at times t and

t + 1 to the foregoing models. We also analyzed

the role of geographic distance between egos

and alters by adding such a variable.

We calculated 95% confidence intervals by sim-

ulating the first difference in the alter’s contem-

Figure 1. Largest Connected Subcomponent of the Social Network in the Framingham Heart Study in the Year 2000.

Each circle (node) represents one person in the data set. There are 2200 persons in this subcomponent of the social

network. Circles with red borders denote women, and circles with blue borders denote men. The size of each circle

is proportional to the person’s body-mass index. The interior color of the circles indicates the person’s obesity status:

yellow denotes an obese person (body-mass index, ≥30) and green denotes a nonobese person. The colors of the

ties between the nodes indicate the relationship between them: purple denotes a friendship or marital tie and orange

denotes a familial tie.

The New England Journal of Medicine

Downloaded from nejm.org on April 16, 2019. For personal use only. No other uses without permission.

T h e

n e w e n gl a n d j o u r n a l

o f

m e d i c i n e

n engl j med 357;4 www.nejm.org july 2 6, 2007

374

33p9

C 1985

E 1995

1975

AUTHOR:

FIGURE:

JOB: ISSUE:

4-C

H/T

RETAKE

SIZE

ICM

CASE

Line

H/T

Combo

Revised

AUTHOR, PLEASE NOTE:

Figure has been redrawn and type has been reset.

Please check carefully.

REG F

Enon

1st

2nd

3rd

Christaki

2 of 4

07-26-07

ARTIST: ts

356xx

D 1990

F 2000

1980

Figure 2. Part of the Social Network from the Framingham Heart Study with Information about Body-Mass Index According to Year.

Each circle (node) represents one person in the data set. Circles with red borders denote women, and circles with blue borders denote men.

The size of each circle is proportional to the person’s body-mass index. The interior color of the circles indicates the person’s obesity status:

yellow denotes an obese person (body-mass index, ≥30) and green denotes a nonobese person. The colors of the ties between the circles

indicate the relationship between them: purple denotes a friendship or a marital tie and orange denotes a familial tie. The disappearance

of a circle from one year to another indicates the person’s death, and the disappearance of a tie between the circles indicates that the re-

lationship between the two persons no longer exists. The largest connected subcomponent of the whole network and the change in obe-

sity over the 32-year study period are shown in an animation that is available with the full text of this article at www.nejm.org.

The New England Journal of Medicine

Downloaded from nejm.org on April 16, 2019. For personal use only. No other uses without permission.

The Spr e ad of Obesit y in a Large Social Net wor k Over 32 Ye a r s

n engl j med 357;4 www.nejm.org july 2 6, 2007

375

poraneous obesity (changing from 0 to 1), using

1000 randomly drawn sets of estimates from the

coefficient covariance matrix and assuming mean

values for all other variables.

All tests were

two-tailed. The sensitivity of the results was as-

sessed with multiple additional analyses (see the

Supplementary Appendix).

R e s ult s

Figure 1 depicts the largest connected subcom-

ponent of the social network in the year 2000.

This network is sufficiently dense to obscure

much of the underlying structure, although re-

gions of the network with clusters of obese or

nonobese persons can be seen. Figure 2 illus-

trates the spread of obesity between adjoining

nodes in a part of the network over time. A video

(available with the full text of this article at www.

nejm.org) depicts the evolution of the largest

component of the network and shows the prog-

ress of the obesity epidemic over the 32-year study

period.

Figure 3A characterizes clusters within the

entire network more formally. To quantify these

clusters, we compared the whole observed net-

work with simulated networks with the same

network topology and the same overall preva-

lence of obesity as the observed network, but with

the incidence of obesity randomly distributed

among the nodes (in what we call “random body-

mass–index networks”). If clustering is occur-

ring, then the probability that an alter will be

obese, given that an ego is known to be obese,

should be higher in the observed network than

in the random body-mass–index networks. What

we call the “reach” of the clusters is the point, in

terms of an alter’s degree of separation from any

given ego, at which the probability of an alter’s

obesity is no longer related to whether the ego

is obese. In all of the examinations (from 1971

through 2003), the risk of obesity among alters

who were connected to an obese ego (at one de-

gree of separation) was about 45% higher in the

observed network than in a random network. The

100

Relative Increase in Probability of Obesity

in an Ego if Alter becomes Obese (%)

1 2 3 4 5

Social Distance between the Ego and the Alter (degree of separation)

100

Relative Increase in Probability of Obesity

in an Ego if Alter becomes Obese (%)

Geographic Distance between the Ego and the Alter (mileage group)

AUTHOR:

FIGURE:

JOB:

4-C

H/T

RETAKE

SIZE

ICM

CASE

Line

H/T

Combo

Revised

AUTHOR, PLEASE NOTE:

Figure has been redrawn and type has been reset.

Please check carefully.

REG F

Enon

1st

2nd

3rd

Christakis

3 of 4

07-26-07

ARTIST: ts

35704 ISSUE:

22p3

Examination 1

Examination 2

Examination 3

Examination 4

Examination 5

Examination 6

Examination 7

Figure 3. Effect of Social and Geographic Distance from

Obese Alters on the Probability of an Ego’s Obesity in

the Social Network of the Framingham Heart Study.

Panel A shows the mean effect of an ego’s social prox-

imity to an obese alter; this effect is derived by compar-

ing the conditional probability of obesity in the observed

network with the probability of obesity in identical net-

works (with topology preserved) in which the same

number of obese persons is randomly distributed. The

social distance between the alter and the ego is repre-

sented by degrees of separation (1 denotes one degree

of separation from the ego, 2 denotes two degrees of

separation from the ego, and so forth). The examina-

tion took place at seven time points. Panel B shows the

mean effect of an ego’s geographic proximity to an obese

alter. We ranked all geographic distances (derived from

geocoding) between the homes of directly connected

egos and alters (i.e., those pairs at one degree of sepa-

ration) and created six groups of equal size. This figure

shows the effects observed for the six mileage groups

(based on their average distance): 1 denotes 0 miles

(i.e., closest to the alter’s home), 2 denotes 0.26 mile,

3 denotes 1.5 miles, 4 denotes 3.4 miles, 5 denotes

9.3 miles, and 6 denotes 471 miles (i.e., farthest from

the alter’s home). There is no trend in geographic dis-

tance. I bars for both panels show 95% confidence in-

tervals based on 1000 simulations. To convert miles to

kilometers, multiply by 1.6.

The New England Journal of Medicine

Downloaded from nejm.org on April 16, 2019. For personal use only. No other uses without permission.

T h e

n e w e n gl a n d j o u r n a l

o f

m e d i c i n e

n engl j med 357;4 www.nejm.org july 2 6, 2007

376

risk of obesity was also about 20% higher for al-

ters’ alters (at two degrees of separation) and

about 10% higher for alters’ alters’ alters (at three

degrees of separation). By the fourth degree of

separation, there was no excess relationship be-

tween an ego’s obesity and the alter’s obesity.

Hence, the reach of the obesity clusters was three

degrees.

Figure 3B indicates that the effect of geo-

graphic distance is different from the effect of

social distance. Whereas increasing social dis-

tance appeared to decrease the effect of an alter

on an ego, increasing geographic distance did not.

The obesity of the most geographically distant

alters correlated as strongly with an ego’s obesity

as did the obesity of the geographically closest

alters. These results suggest that social distance

plays a stronger role than geographic distance in

the spread of behaviors or norms associated with

obesity.

We evaluated the extent of interpersonal asso-

ciation in obesity with the use of regression

analysis. Our models account for homophily by

including a time-lagged measurement of the

alter’s obesity. We evaluated the possible role of

unobserved contemporaneous events by separate-

ly analyzing models of subgroups of the data in-

volving various ego–alter pairings. Figure 4 sum-

marizes the associations.

If an ego stated that an alter was his or her

friend, the ego’s chances of becoming obese ap-

peared to increase by 57% (95% confidence in-

terval [CI], 6 to 123) if the alter became obese.

However, the type of friendship appeared to be

important. Between mutual friends, the ego’s risk

of obesity increased by 171% (95% CI, 59 to 326)

if an alter became obese. In contrast, there was

no statistically meaningful relationship when the

friendship was perceived by the alter but not the

ego (P = 0.70). Thus, influence in friendship ties

appeared to be directional.

The sex of the ego and alter also appeared to

be important. When the sample was restricted to

same-sex friendships (87% of the total), the prob-

ability of obesity in an ego increased by 71%

(95% CI, 13 to 145) if the alter became obese.

For friends of the opposite sex, however, there

was no significant association (P = 0.64). Among

friends of the same sex, a man had a 100% (95%

CI, 26 to 197) increase in the chance of becom-

ing obese if his male friend became obese, where-

as the female-to-female spread of obesity was

not significant (38% increased chance; 95% CI,

−39 to 161).

Among pairs of adult siblings, one sibling’s

chance of becoming obese increased by 40%

(95% CI, 21 to 60) if the other sibling became

obese. This phenomenon appeared to be more

marked among siblings of the same sex (55%;

95% CI, 26 to 88) than among siblings of the

opposite sex (27%; 95% CI, 3 to 54), although the

difference was not significant (P = 0.16). Among

brothers, an ego’s chance of becoming obese in-

creased by 44% (95% CI, 6 to 91) if his alter be-

came obese, and among sisters, an ego’s chance

of becoming obese increased by 67% (95% CI,

27 to 114) if her alter became obese. Obesity in

a sibling of the opposite sex did not affect the

chance that the other sibling would become

obese.

Among married couples, when an alter became

obese, the spouse was 37% more likely (95% CI,

7 to 73) to become obese. Husbands and wives

appeared to affect each other similarly (44% and

37%, respectively). Finally, we observed no effect

on the risk that an ego would become obese if

an immediate neighbor became obese.

22p3

0 100 200 300

Increase in Risk of Obesity in Ego (%)

Ego-perceived friend

Mutual friend

Alter-perceived friend

Same-sex friend

Opposite-sex friend

Spouse

Sibling

Same-sex sibling

Opposite-sex sibling

Immediate neighbor

Alter Type

AUTHOR:

FIGURE:

JOB:

4-C

H/T

RETAKE

SIZE

ICM

CASE

Line

H/T

Combo

Revised

AUTHOR, PLEASE NOTE:

Figure has been redrawn and type has been reset.

Please check carefully.

REG F

Enon

1st

2nd

3rd

Christakis

4 of 4

07-26-07

ARTIST: ts

35704 ISSUE:

Figure 4. Probability That an Ego Will Become Obese According to the Type

of Relationship with an Alter Who May Become Obese in Several Subgroups

of the Social Network of the Framingham Heart Study.

The closeness of friendship is relevant to the spread of obesity. Persons in

closer, mutual friendships have more of an effect on each other than persons

in other types of friendships. The dependent variable in each model is the

obesity of the ego. Independent variables include a time-lagged measure-

ment of the ego’s obesity; the obesity of the alter; a time-lagged measure-

ment of the alter’s obesity; the ego’s age, sex, and level of education; and

indicator variables (fixed effects) for each examination. Full models and

equations are available in the Supplementary Appendix. Mean effect sizes

and 95% confidence intervals were calculated by simulating the first differ-

ence in the contemporaneous obesity of the alter (changing from 0 to 1)

with the use of 1000 randomly drawn sets of estimates from the coefficient

covariance matrix and with all other variables held at their mean values.

The New England Journal of Medicine

Downloaded from nejm.org on April 16, 2019. For personal use only. No other uses without permission.

The Spr e ad of Obesit y in a Large Social Net wor k Over 32 Ye a r s

n engl j med 357;4 www.nejm.org july 2 6, 2007

377

We also investigated two factors that might

mediate or modify the effect of an alter’s weight

gain: his or her smoking behavior and geograph-

ic distance from the ego (see the Supplementary

Appendix). We added measures of smoking be-

havior for the ego and the alter at both the cur-

rent and previous examinations. The coefficient

for the effect of the alter’s obesity was virtually

unchanged; smoking behavior does not appear to

be instrumental to the spread of obesity. Models

that included the geographic distance between

the ego and alter corroborated the result shown

in Figure 3B: geographic distance did not modify

the intensity of the effect of the alter’s obesity on

the ego.

Di s c us sion

Our study suggests that obesity may spread in

social networks in a quantifiable and discernable

pattern that depends on the nature of social ties.

Moreover, social distance appears to be more im-

portant than geographic distance within these

networks. Although connected persons might

share an exposure to common environmental fac-

tors, the experience of simultaneous events, or

other common features (e.g., genes) that cause

them to gain or lose weight simultaneously, our

observations suggest an important role for a pro-

cess involving the induction and person-to-person

spread of obesity.

Our findings that the weight gain of immedi-

ate neighbors did not affect the chance of weight

gain in egos and that geographic distance did not

modify the effect for other types of alters (e.g.,

friends or siblings) helps rule out common expo-

sure to local environmental factors as an explana-

tion for our observations. Our models also con-

trolled for an ego’s previous weight status; this

helps to account for sources of confounding that

are stable over time (e.g., childhood experiences

or genetic endowment).

In addition, the control

in our models for an alter’s previous weight sta-

tus accounts for a possible tendency of obese

people to form ties among themselves. Finally,

the findings regarding the directional nature of

the effects of friendships are especially important

with regard to the interpersonal induction of

obesity because they suggest that friends do not

simultaneously become obese as a result of con-

temporaneous exposures to unobserved factors.

If the friends did become obese at the same time,

any such exposures should have an equally strong

influence regardless of the directionality of friend-

ship. This observation also points to the specifi-

cally social nature of these associations, since the

asymmetry in the process may arise from the fact

that the person who identifies another person as

a friend esteems the other person.

Finally, pairs of friends and siblings of the

same sex appeared to have more influence on

the weight gain of each other than did pairs of

friends and siblings of the opposite sex. This

finding also provides support for the social na-

ture of any induction of obesity, since it seems

likely that people are influenced more by those

they resemble than by those they do not. Con-

versely, spouses, who share much of their phys-

ical environment, may not affect each other’s

weight gain as much as mutual friends do; in the

case of spouses, the opposite-sex effects and

friendship effects may counteract each another.

Obesity in alters might influence obesity in

egos by diverse psychosocial means, such as chang-

ing the ego’s norms about the acceptability of

being overweight, more directly influencing the

ego’s behaviors (e.g., affecting food consump-

tion), or both. Other mechanisms are also pos-

sible. Unfortunately, our data do not permit a

detailed examination. However, some insight into

possible mechanisms can be gained from a con-

sideration of the roles of smoking and geograph-

ic distance in obesity. The tendency of persons

to gain weight when they stop smoking is well

known,

and the coincidence of a decrease in

smoking and an increase in obesity in the over-

all population has been noted.

However, the

present study indicates that regardless of wheth-

er smoking cessation causes weight gain in indi-

vidual persons, and regardless of whether smok-

ing-initiation or smoking-cessation behavior itself

spreads from person to person,

any spread in

smoking behavior is not a significant factor in

the spread of obesity. This finding indicates that

smoking behavior does not mediate the interper-

sonal effect in the spread of obesity. However, in

addition, it suggests that the psychosocial mech-

anisms of the spread of obesity may rely less on

behavioral imitation than on a change in an ego’s

general perception of the social norms regarding

the acceptability of obesity. This point is further

reinforced by the relevance of the directionality

of friendship.

Hence, an ego may observe that an alter gains

The New England Journal of Medicine

Downloaded from nejm.org on April 16, 2019. For personal use only. No other uses without permission.

T h e

n e w e n gl a n d j o u r n a l

o f

m e d i c i n e

n engl j med 357;4 www.nejm.org july 2 6, 2007

378

weight and then may accept weight gain in him-

self or herself. This weight gain in an ego might,

in turn, be determined by various behaviors that

an ego chooses to evince, and these behaviors

need not be the same behaviors that an alter

evinces. The observation that geographic distance

does not modify the effect of an alter’s obesity

also provides support for the concept that norms

may be particularly relevant here. Behavioral ef-

fects might rely more on the frequency of contact

(which one might reasonably expect to be attenu-

ated with distance), whereas norms might not.

The spread of obesity in social networks ap-

pears to be a factor in the obesity epidemic. Yet

the relevance of social influence also suggests

that it may be possible to harness this same force

to slow the spread of obesity. Network phenom-

ena might be exploited to spread positive health

behaviors,

34-36

in part because people’s percep-

tions of their own risk of illness may depend on

the people around them.

Smoking- and alcohol-

cessation programs and weight-loss interventions

that provide peer support — that is, that modify

the person’s social network — are more success-

ful than those that do not.

34,35,38,39

People are

connected, and so their health is connected.

40,41

Consequently, medical and public health interven-

tions might be more cost-effective than initially

supposed, since health improvements in one per-

son might spread to others.

The observation

that people are embedded in social networks sug-

gests that both bad and good behaviors might

spread over a range of social ties. This highlights

the necessity of approaching obesity not only as

a clinical problem but also as a public health

problem.

Supported by a grant from the National Institutes of Health

(NIH R-01 AG24448-01).

No potential conflict of interest relevant to this article was

reported.

We thank Laurie Meneades, Rebecca Joyce, Molly Collins,

Marian Bellwood, and Karen Mutalik for the expert assistance

required to build the data set, and Emelia Benjamin, Virginia

Chang, Scott Desposato, Felix Elwert, Peter Marsden, Joanne

Murabito, James O’Malley, Barbara McNeil, Mark Pachucki, Mark

Pletcher, Mason A. Porter, Darren Schreiber, Richard Suzman,

and Alan Zaslavsky for helpful comments.

Re fer enc e s

Chang VW, Lauderdale DS. Income

disparities in body mass index and obe-

sity in the United States, 1971-2002. Arch

Intern Med 2005;165:2122-8.

Hedley AA, Ogden CL, Johnson CL,

Carroll MD, Curtin LR, Flegal KM. Preva-

lence of overweight and obesity among US

children, adolescents, and adults, 1999-

2002. JAMA 2004;291:2847-50.

Hill JO, Peters JC. Environmental con-

tributions to the obesity epidemic. Science

1998;280:1371-4.

Stunkard AJ, Sorensen TI, Hanis C,

et al. An adoption study of human obesity.

N Engl J Med 1986;314:193-8.

Stunkard AJ, Harris JR, Pedersen NL,

McClearn GE. The body-mass index of

twins who have been reared apart. N Engl

J Med 1990;322:1483-7.

Newman MEJ. The structure and func-

tion of complex networks. SIAM Review

2003;45:167-256.

Nowak MA, Sigmund K. Evolution

of indirect reciprocity. Nature 2005;437:

1291-8.

Bearman PS, Moody J, Stovel K. Chains

of affection: the structure of adolescent

romantic and sexual networks. Am J Sociol

2004;110:44-91.

Liljeros F, Edling CR, Nunes Amaral

LA. Sexual networks: implications for the

transmission of sexually transmitted in-

fections. Microbes Infect 2003;5:189-96.

Watts DJ, Strogatz SH. Collective dy-

namics of ‘small-world’ networks. Nature

1998;393:440-2.

10.

Sobal J. The size acceptance move-

ment and the social construction of body

weight. In: Sobal J, Maurer D, eds. Weighty

issues: fatness and thinness as social

problems. New York: Aldine de Gruyter,

1999:231-49.

Chang VW, Christakis NA. Medical

modeling of obesity: a transition from

action to experience in a 20th century

American medical textbook. Sociol Health

Illness 2002;24:151-77.

Fogassi L, Ferrari PF, Gesierich B, Roz-

zi S, Chersi F, Rizzolatti G. Parietal lobe:

from action organization to intention

understanding. Science 2005;308:662-7.

Whigham LD, Israel BA, Atkinson RL.

Adipogenic potential of multiple human

adenoviruses in vivo and in vitro in ani-

mals. Am J Physiol Regul Integr Comp

Physiol 2006;290:R190-R194.

Turnbaugh PJ, Ley RE, Mahowald

MA, Magrini V, Mardis ER, Gordon JI. An

obesity-associated gut microbiome with

increased capacity for energy harvest. Na-

ture 2006;444:1027-31.

Dawber TR. The Framingham Study:

the epidemiology of atherosclerotic dis-

ease. Cambridge, MA: Harvard University

Press, 1980.

Feinleib M, Kannel WB, Garrison RJ,

McNamara PM, Castelli WP. The Framing-

ham Offspring Study: design and prelimi-

nary data. Prev Med 1975;4:518-25.

Kamada T, Kawai S. An algorithm for

drawing general undirected graphs. Infor-

mation Processing Letters 1989;31:7-15.

11.

12.

13.

14.

15.

16.

17.

18.

Batagelj V, Mrvar A. Pajek — analysis

and visualization of large networks. In:

Junger M, Mutzel P, eds. Graph drawing

software. Berlin: Springer, 2003:77-103.

Moody J, McFarland DA, Bender-

deMoll S. Visualizing network dynamics.

Am J Sociol 2005;110:1206-41.

Barabasi AL, Albert R. Emergence of

scaling in random networks. Science 1999;

286:509-12.

Ravasz E, Barabasi AL. Hierarchical

organization in complex networks. Phys

Rev E Stat Nonlin Soft matter Phys 2003;

67:026112.

McPherson M, Smith-Lovin L, Cook

JM. Birds of a feather: homophily in social

networks. Ann Rev Sociol 2001;27:415-44.

Sackett DL, Anderson GD, Milner R,

Feinleib M, Kannel WB. Concordance for

coronary risk factors among spouses. Cir-

culation 1975;52:589-95.

Carrington PJ, Scott J, Wasserman S.

Models and methods in social network

analysis. New York: Cambridge University

Press, 2005.

Liang K-Y, Zeger SL. Longitudinal data

analysis using generalized linear models.

Biometrika 1986;73:13-22.

Schildcrout JS, Heagerty PJ. Regression

analysis of longitudinal binary data with

time-dependent environmental covariates:

bias and efficiency. Biostatistics 2005;6:

633-52.

Beck N. Time-series–cross-section

data: what have we learned in the past few

years? Annu Rev Polit Sci 2001;4:271-93.

19.

20.

21.

22.

23.

24.

25.

26.

27.

28.

The New England Journal of Medicine

Downloaded from nejm.org on April 16, 2019. For personal use only. No other uses without permission.

The Spr e ad of Obesit y in a Large Social Net wor k Over 32 Ye a r s

n engl j med 357;4 www.nejm.org july 2 6, 2007

379

King G, Tomz M, Wittenberg J. Mak-

ing the most of statistical analyses: improv-

ing interpretation and presentation. Am J

Polit Sci 2000;44:341-55.

Atwood LD, Heard-Costa NL, Fox CS,

Jaquish CE, Cupples LA. Sex and age

specific effects of chromosomal regions

linked to body mass index in the Fram-

ingham Study. BMC Genet 2006;7:7.

Eisenberg D, Quinn BC. Estimating

the effect of smoking cessation on weight

gain: an instrumental variable approach.

Health Serv Res 2006;41:2255-66.

Philipson TJ, Posner RA. The long-run

growth in obesity as a function of tech-

nological change. Perspect Biol Med 2003;

46:Suppl 3:S87-S107.

Andrews JA, Tildesley E, Hops H, Li F.

29.

30.

31.

32.

33.

The influence of peers on young adult sub-

stance use. Health Psychol 2002;21:349-57.

Wing RR, Jeffery RW. Benefits of re-

cruiting participants with friends and in-

creasing social support for weight loss

and maintenance. J Consult Clin Psychol

1999;67:132-8.

Malchodi CS, Oncken C, Dornelas EA,

Caramanica L, Gregonis E, Curry SL. The

effects of peer counseling on smoking

cessation and reduction. Obstet Gynecol

2003;101:504-10.

Bruckner H, Bearman PS. After the

promise: the STD consequences of ado-

lescent virginity pledges. J Adolesc Health

2005;36:271-8.

Montgomery GH, Erblich J, DiLorenzo

R, Bovbjerg DH. Family and friends with

34.

35.

36.

37.

disease: their impact on perceived risk.

Prev Med 2003;37:242-9.

McKnight AJ, McPherson K. Evalua-

tion of peer intervention training for high

school alcohol safety education. Accid Anal

Prev 1986;18:339-47.

Wechsler H, Moeykens B, Davenport A,

Castillo S, Hansen J. The adverse impact

of heavy episodic drinkers on other college

students. J Stud Alcohol 1995;56:628-34.

Christakis NA, Allison PD. Mortality

after the hospitalization of a spouse.

N Engl J Med 2006;354:719-30.

Granovetter MS. The strength of weak

ties. Am J Sociol 1973;78:1360-80.

Christakis NA. Social networks and col-

lateral health effects. BMJ 2004;329:184-5.

38.

39.

40.

41.

42.

full

text

all

journal

articles

the

world

wide

web

Access to the complete text of the Journal on the Internet is free to all subscribers. To use this Web site, subscribers should go

to the Journal’s home page (www.nejm.org) and register by entering their names and subscriber numbers as they appear on

their mailing labels. After this one-time registration, subscribers can use their passwords to log on for electronic access to the

entire Journal from any computer that is connected to the Internet. Features include a library of all issues since January 1993

and abstracts since January 1975, a full-text search capacity, and a personal archive for saving articles and search results of

interest. All articles can be printed in a format that is virtually identical to that of the typeset pages. Beginning 6 months after

publication, the full text of all Original Articles and Special Articles is available free to nonsubscribers who have completed a

brief registration.

The New England Journal of Medicine

Downloaded from nejm.org on April 16, 2019. For personal use only. No other uses without permission.

Discussion

> Nicholas A. Christakis is a sociologist and a physician known for his research on social networks and on the socioeconomic, biosocial, and evolutionary determinants of human welfare (including the behavior, health, and capability of individuals and groups). Source: https://en.wikipedia.org/wiki/Nicholas_Christakis > James H. Fowler is is a social scientist specializing in social networks, cooperation, political participation, and genopolitics (the study of the genetic basis of political behavior). Source: https://en.wikipedia.org/wiki/James_H._Fowler Original Kamada-Kawai algorithm paper: https://citeseerx.ist.psu.edu/document?repid=rep1&type=pdf&doi=b8d3bca50ccc573c5cb99f7d201e8acce6618f04 This 2007 study made a big splash when it came out and has now been cited over 7000 times. Not everyone agrees with the conclusions and it has caused a bit of controversy: https://www.nytimes.com/2011/08/09/health/09network.html. NYTimes article about it when it came out: https://www.nytimes.com/2007/07/25/health/25cnd-fat.html Todays numbers look a bit different (even higher). The percentage of adults age 20 and older with obesity was 41.9% (2017-March 2020), and the percentage of adults age 20 and older with overweight, including obesity: 73.6% (2017-2018). Source: https://www.cdc.gov/nchs/fastats/obesity-overweight.htm > The **Framingham Heart Study** is a long-term, ongoing cardiovascular cohort study of residents of the city of Framingham, Massachusetts. The study began in 1948 with 5,209 adult subjects from Framingham, and is now on its third generation of participants…. Over 3,000 peer-reviewed scientific papers have been published related to the Framingham Heart Study. Source: https://en.wikipedia.org/wiki/Framingham_Heart_Study Many other potential confounders not mentioned: environmental affects on microbiome/metabolism for example. > Network phenomena appear to be relevant to the biologic and behavioral trait of obesity, and obesity appears to spread through social ties. These findings have implications for clinical and public health interventions. > Our study suggests that obesity may spread in social networks in a quantifiable and discernable pattern that depends on the nature of social ties. Moreover, social distance appears to be more important than geographic distance within these networks. Although connected persons might share an exposure to common environmental factors, the experience of simultaneous events, or other common features (e.g., genes) that cause them to gain or lose weight simultaneously, our observations suggest an important role for a pro-cess involving the induction and person-to-person spread of obesity This data essentially unlocks this study: >These sheets contain valuable, previously unused social-network information because they systematically and comprehensively identify relatives and friends named by the ego. The tracking sheets provide complete information about all first-order relatives (parents, spouses, siblings, and children), whether they are alive or dead, and at least one “close friend” at each of seven examinations be-tween 1971 and 2003. The examinations took place during 3-year periods centered in 1973, 1981, 1985, 1989, 1992, 1997, and 1999. Detailed home addresses were also recorded at each time point; we used this information to calculate the geographic distance between people.

Comments

Products

Project