Body Mass Index and Alcohol Consumption:
Family History of Alcoholism as a Moderator
Ashley N. Gearhardt and William R. Corbin
Recent research suggests that excess food consumption may be conceptualized as an addictive behavior.
Much of the evidence comes from neurobiological similarities between drug and food consumption. In
addition, an inverse relation between alcohol consumption and body mass index (BMI) has been
observed. Previous research has hypothesized that this inverse relation is attributable to competition
between food and alcohol for similar neurotransmitter receptors. The current study explored this
neurobiological hypothesis further by examining the influence of an indicator of biological risk associ-
ated with alcohol problems (family history of alcoholism) on the relationship between alcohol and food
intake. Data from 37,259 participants in the National Epidemiologic Survey on Alcohol and Related
Conditions (NESARC) were included in the study. BMI, family history of alcoholism, gender, and
race/ethnicity were assessed as predictors of typical drinking frequency and estimated blood alcohol
concentration (BAC). An inverse relationship between alcohol consumption and BMI was demonstrated.
An attenuation of family history effects on drinking behavior was evident for obese compared to
nonobese participants. The results suggest a neurobiological link between alcohol use and food con-
sumption, consistent with theories characterizing excess food consumption as an addictive behavior.
Keywords: alcohol, BMI, family history, addiction, food
Although the “obesity epidemic” has recently received consid-
erable media and research attention, the international rate of obe-
sity continues to grow to epidemic proportions and shows no signs
of slowing. Approximately one-third of American adults are obese
(Yach, Stucker, & Brownell, 2006), and obesity has now topped
alcohol to become the second leading cause of preventable death in
the United States, surpassed only by tobacco use (Mokdad, Marks,
Stroup, & Gerberding, 2004). The increased rates of obesity are
accompanied by an increase in personal and societal spending to
combat this problem. Although billions of dollars are spent on
weight loss products and countless research studies are conducted
to achieve sustainable weight loss, no successful solution has been
found (Anderson, Stokholm, Backer, & Quaade, 1988). Recently,
similarities between palatable food consumption and addictive
substances have led to the theory that excess food consumption
may be appropriately conceptualized as an addictive behavior
(Gearhardt, Corbin, & Brownell, in press; Kleiner, Gold, Frost-
Pineda, Lenz-Brunsman, Perri, & Jacobs, 2004).
Both biological and behavioral evidence suggest that food may
be addictive. First, neuroimaging and animal model research has
found that excess food consumption is associated with neurobio-
logical changes in the opiate and dopamine systems that parallel
changes caused by drugs of abuse (Hajnal, Smith, & Norgren,
2004; Hoebel, Rada, Mark, & Pothos, 1999; Mark, Smith, Rada, &
Hoebel, 1994; Nieto, Wilson, Cupo, Roques, & Noble, 2002;
Volkow et al., 2003). Behaviorally, animals given intermittent
access to sugar develop symptoms of tolerance and withdrawal
(Hoebel et al., 1999; Nieto et al., 2002), both of which are
hallmarks of addiction. Many of the closest connections between
food and addictive substances have been drawn between alcohol
and high-fat, high-sugar foods. In addition to producing behavioral
reinforcement through the same neurobiological pathway (Hajnal
et al., 2004; Hoebel et al., 1999; Mark et al., 1994; Nieto et al.,
2002) both high-fat sweets and alcohol are frequently used to
regulate emotions (Canetti, Bachar, & Berry, 2001; Cooper, Frone,
Russelll, & Mudar, 1995). Research on human eating habits has
also found behavioral evidence that maps onto substance depen-
dence criteria, such as loss of control, continued use despite
negative consequences, and an inability to reduce consumption of
calorie dense foods (Gearhardt et al., in press; Kleiner et al., 2004).
Although theories of food addiction are relatively recent, re-
searchers have been interested in the relationship between alcohol
and weight gain for some time. Alcohol is a commonly consumed,
calorie-dense substance, which led early theorists to predict that
heavy alcohol consumption would increase the likelihood of being
overweight or obese (Gruchow, Sobocinski, Barboriak, &
Scheller, 1985). In fact, Gruchow et al. (1985) found that, although
consumption of alcohol increases the overall consumption of cal-
ories, men and women drinkers do not appear to be at increased
risk for obesity (Gruchow et al., 1985). More recent studies have
actually found an inverse relationship between alcohol consump-
tion and body mass index (BMI) (Lahti-Koski, Pietinen, Helio-
vaara, & Vartiainen, 2002; Liu, Serdula, Williamson, Mokdad, &
Byers, 1994; Rohrer, Rohland, Denison, & Way, 2005). In other
words, as BMI increases, alcohol consumption decreases. Al-
though, this inverse relationship does not support the hypothesis
Ashley N. Gearhardt and William R. Corbin, Department of Psychol-
ogy, Yale University.
Correspondence concerning this article should be addressed to Ashley
Gearhardt, Department of Psychology, Yale University, Box 208205, New
Haven, CT 06520-8205. E-mail: Ashley.firstname.lastname@example.org
Psychology of Addictive Behaviors
2009, Vol. 23, No. 2, 216–225
© 2009 American Psychological Association
0893-164X/09/$12.00 DOI: 10.1037/a0015011
that alcohol use contributes to obesity, it is consistent with the
theory that food may be addictive. From a food addiction frame-
work, researchers began to consider the possibility that alcohol and
food’s shared biological pathways might cause competition be-
tween the substances. In other words, when a pathway is occupied
by one of the behaviors (i.e., food consumption or alcohol con-
sumption), it would block the other (Kleiner et al., 2004).
These results are consistent with studies using pharmacotherapy
to block opiate receptors. When opiate blockers, such as naltrex-
one, are used to block reward pathways, binge eaters reduce their
consumption of sweet high-fat foods (Drewnowski, Krahn, Demi-
track, Nairn, & Gosnell, 1995), and alcohol dependent participants
reduce their consumption of alcohol (O’Malley, Krishnan-Sarin,
Farren, Sinja, & Kreek, 2002). Working from this conceptual
model, Kleiner et al. (2004) evaluated the relationship between
BMI and alcohol use among individuals awaiting bariatric weight
reduction surgery. They found that, in a female overweight and
obese population, BMI was inversely associated with alcohol
consumption, with 35.4% of extremely obese participants consum-
ing alcohol in the last year compared with 62.5% of participants
who were not obese.
Although suggestive of a shared biological pathway, the results
of the Kleiner et al. study (2004) do not speak directly to the issue
of biological vulnerability. Demonstrating a relation between bio-
logical risk for one behavior (e.g. alcohol use) and risk for exces-
sive engagement in the other (e.g. overeating) would provide a
more compelling argument for a shared biological vulnerability.
There is, in fact, some evidence to suggest that such a relation may
exist. For example, a personal or family history of alcohol prob-
lems can affect one’s food choices. Alcoholics and those with a
positive family history demonstrate a greater preference for higher
concentrations of sweeteners than nonalcoholics or negative fam-
ily history participants (Kampov-Polevoy, Garbutt, & Janowsky,
1999). These results suggest that risk for alcohol use may increase
the hedonic value of foods that are implicated in obesity
(Drewnowski, Kurth, & Rahaim, 1991). However, no studies to
date have examined family history of alcoholism with respect to
the relation between BMI and alcohol use.
In the current study, we hoped to fill this gap in the literature by
using data from a large nationally representative data set (the
National Epidemiologic Survey on Alcohol and Related Condi-
tions; NESARC). By using a nationally representative sample, we
were able to explore the risk factor of a family history of alcohol
problems as a possible moderator of the relationship between
alcohol consumption and body mass index. We hypothesized that
measures of alcohol consumption would be significantly lower
among obese and severely obese individuals relative to nonobese
individuals (underweight, normal weight, and overweight). In ad-
dition, individuals with a positive family history of alcohol prob-
lems were expected to consume alcohol more frequently, and at
higher typical levels (Merikangas et al., 1998). Finally, consistent
with the idea that consumption of one substance can block reward
pathways for the other, we expected the increased risk for elevated
levels of alcohol consumption associated with a family history of
alcoholism to attenuate at high levels of BMI (obese and morbidly
In addition to the primary study hypotheses, we hoped to ex-
plore the impact of other potentially important moderators, includ-
ing gender and race/ethnicity. Previous studies of the relationship
between alcohol consumption and BMI have found different pat-
terns of results for men and women. Among women, an inverse
relation between BMI and alcohol consumption has been found for
both frequency and quantity of alcohol consumption (Lahti-Koski
et al, 2002; Liu et al., 1994; Rohrer, Rohland, Denison, & Way,
2005). For men, studies of the association between BMI and
alcohol use have not shown a consistent relationship in either
direction. Frequency of alcohol use has not been associated with
BMI, and although average alcohol consumption has been found to
relate positively to BMI in some studies (Lahti-Koski et al., 2002),
others have found no association (Colditz et al., 1991; Liu et al.,
1994; Williamson et al., 1987).
The null findings in men may be attributable to aspects of the
study designs used in prior research. Previous studies have often
failed to report participants’ ranges of body mass indices (Colditz
et al., 1991; Gruchow et al., 1985; Liu et al., 1994; Williamson et
al., 1987) or have had very few obese participants in the study
samples. To the extent that decreased alcohol use is associated
specifically with obesity (as opposed to overweight), restricted
range of weights may have limited the ability to detect effects.
Gender differences in weight may also have contributed to a lack
of effects within samples of men. Drinking quantity is likely to
differ by BMI simply because heavier individuals need to consume
more alcohol to reach comparable blood alcohol concentrations
(BACs), and this may be especially true for men who typically
have higher average weights. Thus, it is important to control for
the impact of weight when examining the quantity of alcohol
consumed. For this reason, estimated BACs may provide the best
index of drinking quantity for studies of the relation between
alcohol use and BMI. In the current data set, we were able to
explore the relation between BMI and BAC for all weight classi-
fications and therefore hypothesized that the inverse relationship
would be present for both male and female participants.
Although there is a lack of prior research on other factors that
may moderate the relation between BMI and alcohol use, other
important variables known to impact both alcohol consumption
and risk for obesity should be considered. For example, racial/
ethnic differences in rates of alcohol use/abuse and obesity are
well documented. Although Caucasians often have higher rates of
alcohol consumption (Higuchi, Parrish, Dufour, Towle, & Harford,
1993), Hispanic and African American populations typically have
higher rates of obesity (Kumanyika, 1993). Although there is
sufficient reason to suspect that race/ethnicity might serve as a
moderator (based on known associations with both alcohol use and
BMI), there is no prior empirical work to guide hypotheses about
the nature of possible interactions. Thus, relative to gender, anal-
yses of race/ethnicity as a possible moderator of the relation
between BMI and alcohol use were more exploratory in nature. We
also controlled for socioeconomic status (SES) and age in all
analyses. It is possible that racial/ethnic group differences in
alcohol use and food consumption are at least partially attributable
to differences in SES. Thus, to accurately assess the role of
race/ethnicity in the relation between alcohol use and BMI, it is
necessary to control for the impact of SES. Age is another impor-
tant characteristic associated with both alcohol use and obesity. As
participants get older they are less likely to drink or to drink
heavily (Johnston, O’Malley, Bachman & Schulenberg, 2006), but
are more likely to be overweight or obese (Romeis, Grant, Knopik,
Pedersen, & Heath, 2004). Thus, the inverse relation between BMI
BODY MASS INDEX, FAMILY HISTORY, AND ALCOHOL USE
and alcohol consumption could alternatively be explained by the
The NESARC, which was designed and sponsored by NIAAA,
is a longitudinal survey that began in 2001–2002. The target
population of the NESARC is the civilian noninstitutionalized
population, 18 years and older, residing in the United States and
the District of Columbia. Data were collected from 43,093 Amer-
icans in the first wave of the survey. (Grant, Kaplan, Shepard, &
Moore, 2003). A total of 37,259 participants had complete data on
the variables of interest for the current study.
The subset of participants used in this study had an average age
of 46.55 years (SD ? 18.24) and an average family income of
$35,000 to $39,999. The mean weekly alcohol consumption was
3.11 drinks per week (SD ? 9.24), with the average participant
drinking 0.99 days per week (SD ? 1.81). Mean quantity per
drinking episode was 1.59 drinks (SD ? 2.22) with participants
reaching an average BAC of .03 g% (SD ? .04). Not including the
35.78% of participants who did not drink, mean weekly alcohol
consumption was 4.88 drinks per week (SD ? 11.19), with an
average of 1.55 drinking days a week (SD ? 2.05) and 2.49 drinks
(SD ? 2.34) per drinking episode. Drinkers reported reaching an
average BAC of .04 g% (SD ? .05).
With respect to BMI, the majority of participants were in the
normal weight category (40.4%). The largest racial/ethnic group
was Caucasian (60.4%), although there were also relatively large
numbers of African American (19.2%) and Hispanic (20.4%)
participants because of oversampling of these groups. The gender
breakdown was relatively even, although there were more women
(56.5%) than men in the sample (43.5%). With respect to family
history status, 77.0% reported no family history of alcohol prob-
lems, and 23.0% reported a family history of alcohol problems.
Table 1 provides detailed demographics for the study sample.
The Census Supplementary Survey (C2SS), in combination with
the Census 2000 Group Quarters Inventory, formed the sampling
frame for the NESARC. One sample adult was selected for inter-
view in each household (Grant et al., 2003). Data were collected in
face-to-face computer-assisted personal interviews that were con-
ducted in respondents’ homes. The NESARC oversampled African
American and Hispanic participants at the design phase of the
survey and oversampled adults 18 to 24 years of age at the
BMI for each participant was computed by using the standard
formula: [(weight (lb)/(height (in)2] ? 703. Participants were then
classified by BMI into the following weight categories: under-
weight ?17.99, normal weight ? 18–24.9, overweight ? 25–29.9,
obese ? 30–39.9, and severely obese ?40.
A family member’s history of alcohol problems was defined
as that person having the following: physical or emotional prob-
lems because of drinking; problems with a spouse, family, or
friends because of drinking; problems at work or school; legal
problems (e.g. drunk driving arrests); or having to spend a lot of
time drinking or being hungover. Participants were given this
definition and were asked to evaluate whether each of their first-
degree relatives was an alcoholic or problem drinker. In the current
study, a positive family history of alcohol problems was indicated
by either a paternal or maternal alcohol problem using this crite-
rion. A negative family history of alcohol problems was indicated
by neither parent meeting the criteria for an alcohol problem. The
use of parental history of alcohol problems as an index of family
Alcohol Consumption by Weight Class, Family History Status, Gender, and Race/Ethnicity
Weekly frequency of
drinking daysTypical BAC
Average quantity of
nM SDnM SDnM SD
Underweight (average BMI ? 16.93)
Normal weight (average BMI ? 22.29)
Overweight (average BMI ? 27.23)
Obese (average BMI ? 33.38)
Severely obese (average BMI ? 44.65)
BMI ? body mass index.
GEARHARDT AND CORBIN
history status is a common approach and one that has been used in
prior studies from the NESARC data set (Grant et al., 2006;
Stonenberg, Mudd, Blow, & Hill, 1998). Excluding those with
alcohol-dependent mothers had little effect on outcomes, suggest-
ing that fetal alcohol effects were not responsible for the observed
Participants were asked to respond to separate questions regard-
ing the frequency (number of drinking days in the past year) and
quantity (drinks per drinking day) of alcohol use. Participants were
shown a series of flashcards with life-sized picture representations
of different alcoholic beverages and the corresponding number of
ounces to assist them in estimating standard drink size (Grant et
al., 2003). The number of drinking days the participant reported
was used to indicate frequency of alcohol consumption, whereas
quantity was used to compute typical blood alcohol levels. Typical
blood alcohol levels were computed using an empirically validated
calculation (Hustad & Carey, 2005) for estimating BACs from
naturally occurring drinking episodes [Women; (drinks per drink-
ing day)/2] ? [9/(weight in pounds)] – [.017/number of hours over
which drinking occurred]; Men; [(drinks per a drinking day)/2] ?
[7.5/(weight in pounds)] – [.017/number of hours over which
drinking occurred]. No information was available in the data set
regarding the time over which alcohol consumption took place.
Thus, three different time periods were initially examined; one
hour, two hours, and three hours. Because the same pattern of
results was found for each time frame, we report results only for
the one-hour time frame as it produced the best distributional
properties for BAC. Because the formula can occasionally return
negative values (e.g. a 250-lb male consumes one drink over a
one-hour period), any negative values were recoded as zero values.
Data Analytic Plan
All analyses were performed using SPSS version 15.0 (2007).
Prior to conducting the primary analyses, all variables were ex-
amined for missing values. From the original 43,093 participants,
1,439 individuals did not provide sufficient information to calcu-
late BMI and an additional 2,453 did not provide information on
their parents’ history of alcohol consumption. In regards to the
dependent variable, 205 participants did not provide information
on drinking frequency and 238 participants did not provide infor-
mation on drinking quantity. In addition, only Caucasian, African
American, and Hispanic participants were included because these
racial/ethnic categories had sufficient sample sizes to provide
appropriate power for group comparisons. Thus, an additional
2,033 participants of other ethnic/racial backgrounds were not
included in the analyses, resulting in a total of 37,259 participants
in analyses testing the primary study hypotheses.
In addition, all variables were examined for outliers and nor-
mality. Outliers for the BAC distribution were identified as values
that were extreme relative to the distribution (?3 SD above or
below the mean) and/or were very unlikely to reflect true values.
Because values greater than .30 g% are generally incapacitating
and unlikely to occur on average, values above this cutoff were
recoded to .30 g%. These values were more than 4 SDs above the
mean, and only .07 % of participants had an average BAC above
this cutoff. Frequency of alcohol consumption and typical BAC
levels (after resetting extreme values) were significantly skewed
(values of 5.35 and 2.77 respectively). Log transformation resulted
in a skewness value of 1.44 for frequency of alcohol consumption,
but transformations made little difference for BAC. Thus, the
original BAC values were kept for ease of interpretation. BAC
analyses were also conducted using nonparametric tests (i.e.
Kruskal-Wallis), and the same pattern of results was found. Sep-
arate 2 (Family History Status) ? 2 (Gender) ? 5 (Weight Clas-
sification) ? 3 (Race) Analyses of Covariance (ANCOVAs) were
conducted for each outcome measure, with age and socioeconomic
status included as covariates. The BAC analyses were only con-
ducted for participants who reported drinking alcohol (n ?
23,928), whereas analyses for drinking frequency included the full
sample (n ? 37, 259). To test the study hypotheses, we specifically
assessed main effects for each independent variable and interac-
tions between weight classification and each of the other indepen-
dent variables. All significant interactions were decomposed by
using the simple slopes procedures outlined by Aiken & West
(1991). Table 2 provides a full correlation matrix of all variables
included in the study.
Frequency of Alcohol Consumption
Main effects were observed for SES, F(1, 37,502) ? 485.228, p ?
.001, partial ?2? .013, age, F(1, 37,502) ? 60.987, p ? .001, part-
?2? .000, weight class, F(4, 37,502) ? 34.469, p ? .001, partial
Correlations Between the Study Variables
Blood alcohol concentration
?Correlation is significant at the .05 level (2-tailed).
Frequency ? Frequency of alcohol consumption.
??Correlation is significant at the .01 level (2-tailed).
BODY MASS INDEX, FAMILY HISTORY, AND ALCOHOL USE
?2? .004, gender, F(1, 37,502) ? 193.291, p ? .001, partial ?2?
.005 and race, F(2, 37,502) ? 16.159, p ? .001, partial ?2? .001.
The results suggested that higher socioeconomic status and younger
age were associated with more frequent drinking. Men and family
history positive participants also drank significantly more frequently
than women or family history negative participants, respectively.
With respect to race, Caucasian participants consumed alcohol more
frequently than African American or Hispanic participants, who had
relatively similar frequencies of consumption. Finally, obese partici-
pants consumed alcohol significantly less frequently than normal and
overweight participants, and severely obese participants consumed
alcohol significantly less frequently than normal weight, overweight,
and obese participants (see Table 1).
Several 2-way interactions relating to the study hypothesis were
also observed; family history by weight classification; F(4,
37,502) ? 6.798, p ? .01, partial ?2? .001, gender by weight
classification, F (4, 37,502) ? 7.279, p ? .001, partial ?2? .001,
and race/ethnicity by weight classification, F(8, 37,502) ? 14.827,
p ? .001, partial ?2? .003. Graphic depiction of the significant
two-way interactions revealed likely moderation of family history
and race effects related to obese/nonobese status. Thus, a 2-level
variable consisting of nonobese (underweight, normal weight, and
overweight) and obese (obese, morbidly obese) groups was created
to simplify interpretation. Consistent with study hypotheses, the
family history by weight classification interaction resulted from a
significant simple main effect of family history for nonobese
participants, F(1, 28,544) ? 97.260, p ? .001, ?2? .003, but not
for obese participants, F(1, 8,971) ? 2.577, p ? .28, ?2? .000.
Family history effects for all five weight classifications are de-
picted in Figure 1.
The race/ethnicity by weight classification interaction was also
driven by larger effects for nonobese participants, partial ?2?
.011, than for obese participants, partial ?2? .001, although the
simple main effects of race were significant for both groups (see
Figure 2 for racial/ethnic groups differences within each of the five
weight classes). Caucasian participants drank more frequently than
African American or Hispanic participants in both obese and
nonobese groups. However, the rank ordering of Hispanic and
African Americans for alcohol use differed by weight status.
Hispanic participants had the lowest frequency of alcohol use
among the nonobese participants, whereas African Americans had
the lowest frequency among the obese participants.
Unlike the 2-way interactions by weight class for family history
status and race, the gender by weight classification interaction did
not appear to be because of clear differences between obese and
nonobese participants. Rather, inspection of the graph of the in-
teraction suggested that it was driven by the two extreme ends of
the BMI continuum (the underweight and morbidly obese catego-
ries). Additional analyses exploring simple main effects of gender
within each weight classification confirmed our impressions based
on the graph, with a reduced, although still significant effect of
gender for underweight (partial ?2? .017) and morbidly obese
participants (partial ?2? .032). The effect of gender was signif-
icant for all other weight classifications and resulted in a partial
?2? .059 for normal weight participants, partial ?2? .076 for
overweight participants, and partial ?2? .070 for obese partici-
pants (see Figure 3).
Average Blood Alcohol Levels on Drinking Occasions
Once again, main effects were observed for socioeconomic
status, F(1, 23,901) ? 611.275, p ? .001, partial ?2? .025, age,
F(1, 23,901) ? 1706.301, p ? .001, partial ?2? .067, family
history, F(1, 23,901) ? 51.566, p ? .001, partial ?2? .002,
weight class, F(4, 23,901) ? 122.734, p ? .001, partial ?2? .020,
gender, F(1, 23,901) ? 9.225, p ? .01, partial ?2? .000, and race,
F(2, 23,901) ? 10.267, p ? .001, partial ?2? .001. The direction
of the results was identical to the model for frequency of alcohol
use for all variables other than socioeconomic status and race.
Results in the BAC analyses suggested that lower socioeconomic
status was associated with higher blood alcohol levels. With re-
spect to race, Hispanic participants reached significantly higher
blood alcohol levels than African American participants, with
intermediate blood alcohol levels among Caucasian participants
(see Table 1).
Two-way interactions between the predictor variables were ob-
served for family history by weight classification, F(4, 23,901) ?
11.314, p ? .001, partial ?2? .002, and race/ethnicity by weight
Weight Category (BMI)
Logged Frequency of Consumption
Family History Negative
Family History Positive
Weight class by family history interaction for frequency of alcohol use.
GEARHARDT AND CORBIN
classification, F(8, 23,901) ? 4.059, p ? .001, partial ?2? .001.
Once again, graphic depiction of the significant 2-way interactions
revealed likely moderation of family history and race effects
stemming from obese/nonobese status. Thus, a 2-level variable of
obesity status was used to decompose the interactions. Simple
main effects of family history were significant for both nonobese
and obese participants, although the effect of family history was
larger for nonobese, partial ?2? .012, than for obese participants,
partial ?2? .001 (see Figure 4). As in the analyses for drinking
frequency, simple main effects of race/ethnicity were significant
for both obese, F(2, 8,960) ? 51.878, p ? .001, partial ?2? .011,
and nonobese, F(2, 28,480) ? 186.077, p ? .001, partial ?2?
.013, participants. African Americans had the lowest blood alcohol
levels for both groups, but BACs for Caucasian and Hispanic
participants differed by obesity status. In the nonobese group,
Caucasian and Hispanic participants had similar blood alcohol
levels, whereas Hispanic participants demonstrated the highest
BACs in the obese group (see Figure 5).
Many of the main effects identified in the current study were
consistent with the extant literature. Men, younger participants,
and those with a family history of alcohol problems drank more
frequently and had higher typical BACs than women, older par-
ticipants, or family history negative participants, respectively
(Johnston et al., 2006; Merikangas et al., 1998; Wilsnack,
Vogeltanz, Wilsnack, & Harris, 2000). Also consistent with pre-
vious literature, Caucasian participants consumed alcohol more
frequently than African American or Hispanic participants (Higu-
chi et al., 1993). In contrast to the findings for frequency, Hispanic
participants had higher average BACs than both Caucasian and
Weight Category (BMI)
Logged Frequency of Consumption
Weight class by race/ethnicity interaction for frequency of alcohol use.
Weight Category (BMI)
Logged Frequency of Consumption
Weight class by gender interaction for frequency of alcohol use.
BODY MASS INDEX, FAMILY HISTORY, AND ALCOHOL USE
African American participants. Although low SES has generally
been considered a risk factor for alcohol-related problems (Van
Oers, Bongers, Van De Goor & Garretsen, 1999), more recent
research has shown increased risk at very low and very high levels
of SES (Hanson & Chen, 2005). The results of the current study
were consistent with this more recent literature with higher fre-
quency of consumption among high SES individuals and higher
BACs among low SES individuals. The results for both race/
ethnicity and SES highlight the importance of disaggregating
frequency and quantity of alcohol consumption when evaluating
group differences in drinking behavior.
Most important to the hypotheses of the current study, obese
participants consumed alcohol significantly less frequently and
had lower typical BACs than normal weight and overweight par-
ticipants, and severely obese participants consumed alcohol sig-
nificantly less frequently and had lower typical BACs than normal
weight, overweight, and obese participants. In contrast to some
previous studies, the inverse relation between alcohol use and BMI
was evident for both men and women. The identification of a
consistent inverse relationship between alcohol use and BMI
among men in the current study may be related to differences in
methodology from previous studies (Colditz et al., 1991; Lahti-
Koski et al., 2002; Liu et al., 1994; Williamson et al., 1987). In the
current study, we had a sufficient sample size to explore the
relation between alcohol consumption and BMI across the full
range of BMI, and frequency of alcohol use was disaggregated
from quantity. In addition, the measure of drinking quantity (BAC)
took into account weight-related differences on the impact of
quantity consumed. Although replication of the current results is
important, the strengths of the methodology used provide compel-
ling evidence that the inverse relation between alcohol use and
BMI is not exclusive to women.
In addition to main effects, there were important interactions
between weight classification and the demographic variables as-
sessed. Although there was an interaction between gender and
weight classification for frequency of consumption, the same gen-
eral pattern of results was seen for both men and women. The
interaction was driven by a lack of gender differences among
Weight Category (BMI)
Typical Blood Alcohol Concentration
Family History Negative
Family History Positive
Weight class by family history interaction for typical BAC.
Weight Category (BMI)
Typical Blood Alcohol Concentration
Weight class by race/ethnicity interaction for typical BAC.
GEARHARDT AND CORBIN
underweight and severely obese participants, with robust gender
differences in all other weight classes. Perhaps the lack of gender
differences in the underweight group relate to higher levels of
alcohol consumption associated with certain eating disorders (e.g.,
bulimia nervosa) that occur disproportionately among women
(Dansky, Brewerton, & Kilpatrick, 2000). In regards to the se-
verely obese participants, it is possible that the reduced impact is
because of floor effects associated with low frequency of con-
sumption for women.
Race/ethnicity by weight classification interactions were ob-
served for both drinking frequency and BAC. Although the nature
of the interaction differed somewhat across the two outcomes, in
both cases the interaction was driven by the sample of Hispanic
participants. For frequency of consumption, the Hispanic sample
was at lowest risk in the nonobese sample and at intermediate risk
in the obese sample and for BAC, the Hispanic sample was at
intermediate risk in the nonobese sample and at highest risk in the
obese sample. In both cases, the result is that obesity status had
less of an impact on alcohol use for Hispanic participants than for
Caucasian or African American participants. The smaller effect for
Hispanic participants was not expected, and it is unclear through
what mechanism this effect may be operating. Biological differ-
ences in response to alcohol do indeed exist among different
racial/ethnic groups and such differences may have important
implications for health and well-being (Chan, 1986;Galarza, Diaz,
Guzma ´n, Caballero, & Martı ´nez, 1997). It is also quite possible
that racial/ethnic group effects identified in the current study are a
result of cultural or environmental influences.
Of central importance to the shared biological vulnerability
hypothesis, an interaction between family history status and weight
classification was found for both measures of alcohol use. The
influence of a family history of alcohol problems was moderated
by BMI, such that family history effects on frequency of alcohol
consumption were only present among nonobese individuals. Al-
though family history was a significant predictor of BAC for both
obese and nonobese participants, the effect was over three times
larger for nonobese participants. Although these results do not
provide direct evidence for shared neurobiological pathways, they
are quite consistent with the hypothesis that food occupies neuro-
biological pathways related to the reinforcement value of alcohol.
Although it is not entirely clear what biological systems might be
associated with the increased risk associated with a family history
of alcohol problems, the opioid and dopamine systems have been
identified as likely candidates (Gianoulakis, 2001; Sander et al.,
1995). Thus, the blocking of these neurobiological pathways may
be responsible for the attenuation of family history effects on
drinking behavior. Although our results are consistent with this
conclusion, future studies would benefit from the direct examina-
tion of the dopamine and opioid systems to confirm the blocking
hypothesis. Regardless of the mechanism, this study supports the
burgeoning literature suggesting that excessive food intake may be
similar to other addictive behaviors, specifically compulsive alco-
hol use (Drewnowski et al., 1995; Gearhardt et al., in press; Gold,
Frost-Pineda, & Jacobs, 2003; Kleiner et al., 2004).
Although the results were consistent with the blocking hypoth-
eses, other possible explanations for the results must also be
considered. For example, it may be that obese and severely obese
participants drink less because they would need to consume a
prohibitively large quantity of alcohol to reach a BAC level that is
reinforcing. Because participants with a family history of alcohol-
ism have been shown to have an innate tolerance to alcohol
(Schuckit & Smith, 2000), obese individuals with a positive family
history would need to consume even greater quantities of alcohol
to achieve reinforcing BACs. Although plausible, there are both
conceptual and empirical reasons counter to this potential alterna-
tive explanation. Conceptually, the increased quantity of alcohol
necessary to achieve an intoxicating BAC does not appear to be
substantial enough to prevent obese individuals from drinking to
intoxication. For example, during a 2-hour period, an obese 5’6”
woman (weighing 210 lbs) would need to consume less than one
additional alcoholic beverage to reach a BAC of .05, relative to an
overweight 5’6” woman (weighing 170 lbs). The data presented in
Figure 4 also argue against this alternate explanation. If the diffi-
culty of consuming enough alcohol were the driving force behind
reduced consumption among obese participants, one would expect
the effects to be most pronounced among the morbidly obese
participants. However, the attenuation of alcohol consumption was
seen clearly among both the obese and morbidly obese groups, and
attenuation of family history effects was also evident in both
groups. Thus, it seems unlikely that difficulty consuming sufficient
alcohol to reach reinforcing BACs was responsible for the pattern
of results in the current study.
An additional alternative social explanation of the current re-
sults is worth considering. Consumption of alcohol frequently
takes place in social contexts, such as bars or parties. Obese
individuals are often subjected to discrimination based on weight
bias (Puhl & Brownell, 2001). The experience of weight-based
discrimination could result in reduced participation in social ac-
tivities, including activities in which alcohol consumption is com-
mon. Thus, it is possible that the reduction in alcohol consumption
may be due to negative social experiences. Although this is pos-
sible, it is less likely that this possibility would explain the reduced
impact of a family history of alcohol problems. In fact, one might
expect that the negative feelings associated with weight bias would
increase alcohol consumption, especially for those with the added
risk factor of a positive family history.
Although there were a number of strengths of the research
methodology, several important limitations should be considered
when interpreting the results. The study was strengthened by the
use of a large, nationally representative sample, but the NESARC
oversampled African American and Hispanic participants, as well
as individuals between the ages of 18 and 24 years. In this study,
we addressed this issue by controlling for age in our analyses and
by examining the effect of weight on alcohol consumption within
each racial/ethnic category. Nonetheless, we cannot say with con-
fidence that the results will generalize to the general U.S. popu-
lation. In addition, the NESARC depended upon self-report data
for all of the variables used in the study. Thus, incorrect reporting
of alcohol consumption, body weight, or family history of alcohol
problems may have affected the results. This may be especially
pertinent, as weight was a variable for which there was significant
missing data. Heavier participants may be less likely than normal
weight participants to report their weight, such that overweight and
obese participants may have been underrepresented in this study.
Future studies would benefit from direct assessment of weight and
collateral reports of alcohol use to reduce reporting error.
Although the calculation of BACs in the current study was a
relative methodological strength, participants were not asked to
BODY MASS INDEX, FAMILY HISTORY, AND ALCOHOL USE
report the amount of time over which the alcohol was consumed.
Thus, it was necessary to estimate typical time of consumption.
Although the pattern of results was consistent across three differ-
ent possible time periods, future studies would benefit from as-
sessing typical time of consumption. Finally, the effect sizes in the
current study were small in magnitude. Small effects are not
surprising for behaviors as complex as food and alcohol consump-
tion. It is unlikely that the relationship between obesity and alcohol
consumption will explain a large amount of the total variability in
Despite the relatively small effect sizes, we believe there are
both public health and theoretical reasons that the findings are
important. First, both alcohol use disorders and obesity are major
health concerns. Even minor impacts on the consumption of either
calorie-dense foods or alcohol use may have major societal impli-
cations. Secondly, as Kazdin (2003) points out, small effect sizes
in the company of theoretical underpinnings can have a major
impact on the understanding of a phenomenon of interest. In the
current study, the relationship between alcohol consumption, obe-
sity, and family history is accompanied by a theoretical explana-
tion regarding the addictive properties of food. The finding that a
risk factor for excess alcohol consumption can be attenuated by
excess food consumption has important implications for the con-
cept of food addiction.
If the characterization of excess food consumption as a possible
addictive behavior is accurate, it may have important implications
for the prevention and treatment of excessive food consumption.
Perhaps the most important implication for food’s possible addic-
tive qualities is the potential impact of the “toxic environment”
(Brownell, 2004, p. 7) of highly available energy-dense foods. The
widespread availability and aggressive advertising of unhealthy
foods may play on cue-triggered relapse to derail public health
interventions to decrease consumption of these unhealthy foods.
Thus, public policy interventions designed to limit exposure to
“toxic foods” for both adults and children may prove to be effec-
tive in reducing excess food consumption. With respect to treat-
ment, empirically validated approaches for substance dependence
and binge eating share important similarities. Both treatments
include identification of triggers and other relapse prevention
strategies (Agras, 1993; Witkiewitz & Marlatt, 2004). In addition,
numerous treatments based on the 12-step Alcoholics Anonymous
model, such as Overeaters Anonymous, are currently available. If
food is addictive, this would suggest that future obesity and binge
eating treatments should continue to explore methods used to treat
substance dependence. For example cue-exposure and identifica-
tion of alternative reinforcers (Monti et al., 1993) may be worth
considering in the treatment of obesity and binge eating disorder.
Although food’s addictive nature is far from established, the re-
sults of the current study, in combination with previous research,
suggest that further exploration of this possibility is warranted.
Agras, W. S. (1993). Short-term psychological treatments for binge eating.
In C. G. Fairburn & G. T. Wilson (Eds.), Binge eating: Nature, assess-
ment, and treatment (pp. 280–286). New York: The Guilford Press.
Aiken, L. S., & West, S. G. (1991). Multiple regression: Testing and
interpreting interactions. Newbury Park, CA: Sage Publications.
Anderson, T., Stokholm, K. H., Backer, O. G., & Quaade, F. (1988).
Long-term (5-year) results after either horizontal gastroplasty or very-
low-calorie diet for morbid obesity. International Journal of Obesity, 12,
Brownell, K. (2004). Food fight: The inside story of the food industry,
America’s obesity crisis and what we can do about it. New York:
Canetti, L., Bachar, E., & Berry, E. M. (2001). Food and emotion. Behav-
ioural Processes, 60, 157–164.
Chan, A. (1986). Racial differences in alcohol sensitivity. Alcohol and
Alcoholism, 1, 93–104.
Colditz, G. A., Giovannucci, E., Rimm, E. B., Stampfer, M. J., Rosner, B.,
Speizer, F. E., et al. (1991). Alcohol intake in relation to diet and obesity
in women and men. American Journal of Clinical Nutrition, 54, 49–55.
Cooper, M. L., Frone, M. R., Russelll, M., & Mudar, P. (1995). Drinking
to regulate positive and negative emotions: A motivational model of
alcohol use. Journal of Personality and Social Psychology, 69, 990–
Dansky, B. S., Brewerton, T. D., & Kilpatrick, D. G. (2000). Comorbidity
of bulimia nervosa and alcohol use disorders: Results from the National
Women’s Study. International Journal of Eating Disorders, 27, 180–
Drewnowski, A., Kurth, C. L., & Rahaim, J. E. (1991). Taste preferences
in human obesity: Environmental and familial factors. American Journal
of Clinical Nutrition, 54, 635–641.
Drewnowski, A., Krahn, D. D., Demitrack, M. A., Nairn, K., & Gosnell,
B. A. (1995). Naloxone, an opiate blocker, reduces the consumption of
sweet high-fat foods in obese and lean female binge eaters. American
Journal of Clinical Nutrition, 61, 1206–1212.
Galarza, N. J., Diaz, R. D., Guzma ´n, F., Caballero, J. A., & Martı ´nez, A. J.
(1997). The use of naltrexone to treat ambulatory patients with alcohol
dependence. Boletı ´n de la Asociacio ´n Me ´dica de Puerto Rico, 89,
Gearhardt, A. N., Corbin, W. R., & Brownell, K. D. (in press). Food
addiction: An examination of the diagnostic criteria for dependence.
Journal of Addictive Medicine.
Gianoulakis, C. (2001). Influence of the endogenous opioid system on high
alcohol consumption and genetic predisposition to alcoholism. Journal
of Psychiatry and Neuroscience, 206, 304–318.
Gold, M. S., Frost-Pineda, K., & Jacobs, W. S. (2003). Overeating, being
eating, and eating disorders as addictions. Psychiatric Annals, 33, 117–
Grant, B. F., Harford, T. C., Muthe ´n, B. O., Yi, H.-y., Hasin, D. S., &
Stinson, F. S. (2006). DSM-IV alcohol dependence and abuse: Further
evidence of validity in the general population. Drug and Alcohol De-
pendence, 86, 154–166.
Grant, B. F., Kaplan, K., Shepard, J., & Moore, T. (2003). Source and
accuracy statement for wave 1 of the 2001–2002 National Epidemio-
logic Survey on Alcohol and Related Conditions (NESARC). Bethesda,
MD: National Institute on Alcohol Abuse and Alcoholism. Retrieved
March 28, 2007, from http://niaaa.census.gov/
Gruchow, H. W., Sobocinski, K. A., Barboriak, J. J., & Scheller, J. G.
(1985). Alcohol consumption, nutrient intake and relative body weight
among US adults. The American Journal of Clinical Nutrition, 42,
Hajnal, A., Smith, G. P., & Norgren, R. (2004). Oral sucrose stimulation
increases accumbens dopamine in the rat. American Journal Physiology—
Regulatory Integrative and Comparative Physiology, 286, R31–R37.
Hanson, M. D., & Chen, E. (2005). Socioeconomic status and substance
use behaviors in adolescents: The role of family resources versus family
social status. Journal of Health Psychology, 12, 31–35.
Higuchi, S., Parrish, K. M., Dufour, M. C., Towle, L. H., & Harford, T. C.
(1993). Relationship between age and drinking patterns and drinking
problems among Japanese, Japanese-Americans, and Caucasians. Alco-
holism: Clinical and Experimental Research, 18, 305–310.
GEARHARDT AND CORBIN
Hoebel, B., Rada, P. V., Mark, G. P., & Pothos, E. (1999). Neural systems
for reinforcement and inhibition of behavior: Relevance to eating, ad-
diction and depression. In D. Kahneman, E. Diener, & N. Schwarz
(Eds.), Well being: Foundations of hedonic psychology (pp. 558–572).
New York, NY: Russell Sage Foundation.
Hustad, J. T. P., & Carey, K. B. (2005). Using calculations to estimate
blood alcohol concentrations for naturally occurring drinking episodes:
A validity study. Journal of Studies on Alcohol, 66, 130–138.
Johnston, L. D., O’Malley, P. M., Bachman, J. G., & Schulenberg, J. E.
(2006). Monitoring the future: national survey results on drug use,
1975–2005. Volume II: College students and adults ages 19–45 (NIH
Publication No. 06–5884). Bethesda, MD: National Institute on Drug
Kampov-Polevoy, A. B., Garbutt, J. C., & Janowsky, D. S., (1999).
Association between preference for sweets and excessive alcohol intake:
A review of animal and human studies. Alcohol, 34, 386–395.
Kazdin, A. E., (2003). Research design in clinical psychology. Boston:
Allyn & Bacon.
Kleiner, K. D., Gold, M. S., Frost-Pineda, K., Lenz-Brunsman, B., Perri,
M. G., & Jacobs, W. S. (2004). Body mass index and alcohol use.
Journal of Addictive Diseases, 23, 105–118.
Kumanyika, S. K. (1993). Special issues regarding obesity in minority
populations. Annals of Internal Medicine, 119, 650–654.
Lahti-Koski, M., Pietinen, P., Heliovaara, M., & Vartiainen, E. (2002).
Associations of body mass index and obesity with physical activity, food
choices, alcohol intake, and smoking in the 1982–1997 FINRISK stud-
ies. American Journal of Clinical Nutrition, 75, 809–817.
Liu, S., Serdula, M. K., Williamson, D. F., Mokdad, A. H., & Byers, T.
(1994). A prospective study of alcohol intake and change in body weight
among US adults. American Journal of Epidemiology, 140, 912–920.
Mark, G. P., Smith, S. E., Rada, P. V., & Hoebel, B. G. (1994). An
appetitively conditioned taste elicits a preferential increase in mesolim-
bic dopamine release. Pharmacology Biochemistry and Behaviour, 48,
Merikangas, K. R., Stolar, M., Stevens, D. E., Goulet, M. S., Preisig, M. A.,
Fenton, B., et al. (1998). Familial transmission of alcohol use disorders.
Archives of General Psychiatry, 55, 973–979.
Mokdad, A. H., Marks, J. S., Stroup, M. F., & Gerberding, J. L. (2004).
Actual causes of death in the United States. Journal of the American
Medical Association, 291, 1238–1245.
Monti, P. M., Rohsenow, D. J., Rubonis, A. V., Niaura, R. S., Sirota, A. D.,
Colby, S. M., et al., (1993). Alcohol cue reactivity: Effects of detoxifi-
cation and extended exposure. Journal of Studies on Alcohol, 54, 235–
Nieto, M., Wilson, J., Cupo, A., Roques, B. P., & Noble, F. (2002).
Chronic morphine treatment modulates the extracellular levels of en-
dogenous enkephalins in rat brain structures involved in opiate depen-
dence: A microdialysis study. Journal of Neuroscience, 22, 1034–1041.
O’Malley, S. S., Krishnan-Sarin, S., Farren, C., Sinja, R., & Kreek, M.
(2002). Naltrexone decreases craving and alcohol self-administration in
alcohol-dependent subjects and activates the hypothalamo–pituitary–
adrenocortical axis. Psychopharmacology, 160, 19–29.
Puhl, R. M., & Brownell, K. D. (2001). Bias, discrimination, and obesity.
Obesity Research, 9, 788–805.
Rohrer, J. E., Rohland, B. M., Denison, A., & Way, A. (2005). Frequency
of alcohol use and obesity in community medicine patients. BMC Family
Practice, 6, 17.
Romeis, J. C., Grant, J. D., Knopik, V. S., Pedersen, N. L., & Heath, A. C.
(2004). The genetics of middle-age spread in middle-class males. Twin
Research, 7, 596–602.
Sander, T., Harms, H., Podschus, J., Finckh, U., Nickel, B., Rolfs, A., et al.
(1995). Dopamine D1, D2, and D3 receptor genes in alcohol depen-
dence. Psychiatric Genetics, 5, 171–176.
Schuckit, M. A., & Smith, T. L. (2000). The relationships of a family
history of alcohol dependence, a low level of response to alcohol and six
domains of life functioning to the development of alcohol use disorders.
Journal of Studies on Alcohol, 61, 827–835.
Stonenberg, S. F., Mudd, S. A., Blow, F. C., & Hill, E. M. (1998).
Evaluating measures of family history of alcoholism: Density versus
dichotomy. Addiction, 93, 1511–1520.
Van Oers, J., Bongers, I., Van De Goor, L., & Garretsen, H. (1999).
Alcohol consumption, alcohol-related problems, problem drinking, and
socioeconomic status. Alcohol and Alcoholism, 34, 74–88.
Volkow, N. D., Wang, G.-J., Maynard, L., Fowler, J. S., Zhu, W., Logan,
J., et al. (2003). Brain dopamine is associated with eating in humans.
International Journal of Eating Disorders, 33, 136–143.
Williamson, D. F., Forman, M. R., Binkin, N. J., Gentry, E. M., Reming-
ton, P. L., & Trowbridge, F. L. (1987). Alcohol and body weight in
United States adults. American Journal of Public Health, 77, 1324–
Wilsnack, R. W., Vogeltanz, N. D., Wilsnack, S. C., & Harris, T. R.
(2000). Gender differences in alcohol consumption and adverse drinking
consequences: Cross-cultural patterns. Addiction, 95, 251–265.
Witkiewtz, K., & Marlatt, G. A. (2004). Relapse prevention for alcohol and
drug problems: That was Zen, this is Tao. American Psychologist, 59,
Yach, D., Stucker, D., & Brownell, K. D. (2006). Epidemiologic and
economic consequences of the global epidemics of obesity and diabetes.
Nature, 12, 62–65.
Received May 9, 2008
Revision received October 22, 2008
Accepted December 3, 2008 ?
BODY MASS INDEX, FAMILY HISTORY, AND ALCOHOL USE