International Journal of Obesity (2013) 37, 1407–1412
& 2013 Macmillan Publishers Limited All rights reserved 0307-0565/13
www.nature.com/ijo
ORIGINAL ARTICLE
Association of elevation, urbanization and ambient temperature
with obesity prevalence in the United States
JD Voss1, P Masuoka1, BJ Webber1, AI Scher1 and RL Atkinson2,3,4
BACKGROUND: The macrogeographic distribution of obesity in the United States, including the association between elevation and
body mass index (BMI), is largely unexplained. This study examines the relationship between obesity and elevation, ambient
temperature and urbanization.
METHODS AND FINDINGS: Data from a cross-sectional, nationally representative sample of 422 603 US adults containing BMI,
behavioral (diet, physical activity, smoking) and demographic (age, sex, race/ethnicity, education, employment, income) variables
from the 2011 Behavioral Risk Factor Surveillance System were merged with elevation and temperature data from WorldClim and
with urbanization data from the US Department of Agriculture. There was an approximately parabolic relationship between mean
annual temperature and obesity, with maximum prevalence in counties with average temperatures near 18 1C. Urbanization and
obesity prevalence exhibited an inverse relationship (30.9% in rural or nonmetro counties, 29.2% in metro counties with o250 000
people, 28.1% in counties with population from 250 000 to 1 million and 26.2% in counties with 41 million). After controlling for
urbanization, temperature category and behavioral and demographic factors, male and female Americans living o500 m above sea
level had 5.1 (95% confidence interval (CI) 2.7–9.5) and 3.9 (95% CI 1.6–9.3) times the odds of obesity, respectively, as compared
with counterparts living X3000 m above sea level.
CONCLUSIONS: Obesity prevalence in the United States is inversely associated with elevation and urbanization, after adjusting for
temperature, diet, physical activity, smoking and demographic factors.
International Journal of Obesity (2013) 37, 1407–1412; doi:10.1038/ijo.2013.5; published online 29 January 2013
Keywords: altitude; geographic; GIS; urbanization; hypoxia; BRFSS
INTRODUCTION
Although prevalence has stabilized in recent years,1,2 obesity
remains a top public health concern in the United States. Regional
differences in body mass index (BMI) become evident upon
cursory examination of state-level US maps published by the
Centers for Disease Control and Prevention (CDC).3 Obesity
appears most prevalent in the Southeast and Midwest states
and less prevalent in the Mountain West. Despite significant
research into the environmental determinants of obesity, including the built environment, the explanation for these macrogeographic differences is unclear.
Differences in elevation provide a biologically plausible explanation for regional variation.4 Potential mechanisms include
increased metabolic demands, altered leptin signaling secondary
to hypoxia,5–14 reduced birth weight,15–17 reduced childhood
growth18,19 and increased sympathetic tone.20 Cross-sectional
studies of the relationship between BMI and elevation, however,
have produced conflicting results. Among 617 Tibetans, waist
circumference, waist-to-hip ratio and BMI were inversely related to
elevation, in a range from 1200 to 3700 m above sea level.21
Similarly, in an endogamous Indian population, women living in
the plains were overweight, whereas those living at elevations
above 2400 m were of normal weight.22 Similarly, dogs were
found to have lower rates of obesity in the Mountain West than in
lower elevation areas of the United States.23 An opposite pattern,
however, was observed for childhood overweight and obesity in
Saudi Arabia24 and for metabolic syndrome in Peru, although the
latter did not achieve statistical significance.25
Urbanization and mean annual temperature also demonstrate
regional variation. Rural residence is a known risk factor for poor
diet,26 and cold ambient temperature has been described as
catabolic.21 In this study we assessed the geographic distribution
of obesity in the United States as it relates to elevation,
temperature and urbanization, after correcting for known
behavioral and demographic covariates.
MATERIALS AND METHODS
Data sources
The Behavioral Risk Factor Surveillance System (BRFSS) is a nationwide
telephone health survey with a well-defined sampling strategy that
permits extrapolation to the noninstitutionalized US adult population
using sampling weights provided in the data set. Methods of data
collection and limitations are described elsewhere.27 Unlike 2010, the 2011
data set includes information on diet and physical activity recommendation compliance, in addition to the demographic questions (age, sex,
race/ethnicity, education and income) collected annually.
In brief, we evaluated obesity as a function of macrogeographic
independent variables (elevation, mean annual temperature and urbanization) after adjusting for known demographic and lifestyle predictors. For
this analysis, education was dichotomized at the college degree level
(ocollege degree, Xcollege degree), annual income was trichotomized at
o$20 000 and X$75 000 levels and self-reported race/ethnicity was recategorized as white, black, Hispanic, Asian, other and missing. Degree of
1
Department of Preventive Medicine and Biometrics, Uniformed Services University of the Health Sciences, Bethesda, MD, USA; 2Virginia Commonwealth University, Richmond,
VA, USA and 3Obetech Obesity Research Center, Richmond, VA, USA. Correspondence: Dr JD Voss, Department of Preventive Medicine and Biometrics, Uniformed Services
University of the Health Sciences, 4301 Jones Bridge Road, Bethesda, MD 20814, USA.
E-mail: [email protected]
4
Editor at the International Journal of Obesity.
Received 5 October 2012; revised 10 December 2012; accepted 19 December 2012; published online 29 January 2013
Elevation and obesity
JD Voss et al
1408
urbanization was categorized using modified Beale codes (BC) as follows:
counties with Z1 million residents (BC ¼ 1); 250 000 to 1 million (BC ¼ 2);
those in metro areas, but with o250 000 (BC ¼ 3); and those in nonmetro
or rural areas (BC ¼ 4–9).28 Obesity was defined as BMI X30 kg m 2 and
was classified as missing for pregnant women and for those with a BMI
value X99.99 or p12.00 kg m 2. In order to account for the sampling
strategy employed by CDC, the final weighting variable (_LLCPWT) was
used as an inverse probability weight. Complex survey design can reduce
precision due to stratification, but analysis of this data set demonstrated
that the effect was negligible for our study.
Elevation above sea level, mean annual temperature (degrees centigrade) and urbanization for subjects were based on county of residence
reported in the 2011 survey. Mean elevation and annual temperature for
3134 administrative areas (counties) in the United States were obtained
through publicly available data sets. Elevation and temperature data were
obtained from WorldClim (www.worldclim.org) and were processed using
ArcGIS version 10.0 (ESRI, Redlands, CA, USA). WorldClim provides weather
data that are interpolated from average monthly weather station data to
1 km resolution grids using well-described methodology.29 WorldClim
elevation data are resampled to 1 km resolution from the Shuttle Radar
Topography Mission altitude data. County administrative boundaries were
downloaded from the Global Administrative Areas website
(www.gadm.org). The ArcGIS program, Zonal Statistics as Table, was
used to calculate mean annual temperature and mean elevation by county
(Supplementary Figure 1). The merged ArcGIS outputs combined via
Microsoft Access and Excel with the federal information processing
standards (FIPS) codes and county typology codes (such as urbanization)
obtained from US Department of Agriculture (USDA) Economic Research
Service were matched using state and county name.28 The combined data
set was then merged by state and county FIPS codes with the 2011 BRFSS
data. All subsequent analyses were performed using SAS version 9.3 (SAS
Institute Inc., Cary, NC, USA) and STATA version 12.0 (StataCorp, College
Station, TX, USA). The CDC used a modified sampling strategy to include
cell phones where the place of residence for each cell phone observation
was based on self-report rather than location of telephone service, as 5090
individuals reported living in a different state. For 31 of these (o1%), the
recorded county FIPS code did not exist within the reported state of
residence; these 31 observations were excluded.
Population
There were 504 408 observations in the data set. Aside from those in Puerto
Rico (n ¼ 6613), missing county codes (n ¼ 52 972, including 888 codes
n ¼ 1489) and the 31 observations reported above, every unique FIPS code
(n ¼ 2231) contained within the BRFSS data set for the remaining 444 792
observations matched with a corresponding USDA FIPS code. Elevation and
temperature data derived from WorldClim for three cities (Baltimore, MD, St
Louis, MO, and Fairfax, VA) did not merge with the USDA file because of
FIPS code discrepancies between them and their surrounding counties of
the same name. The FIPS codes corresponding to these cities were
manually assigned the temperature and elevation derived from their
respective counties. Excluding invalid BMI as explained above, the final data
set included 422 603 subjects representative of B207 million Americans.
Statistical analyses
We considered whether elevation, urbanization and ambient temperature
were associated with obesity (generalized estimating equation (GEE)) or
median BMI (quantile regression) after adjusting for lifestyle (smoking,
physical activity and diet) and demographics (age, sex, race/ethnicity,
education, employment status and income). A hierarchal analysis was
necessary to account for the differing unit of observation from the county
level (elevation, urbanization and temperature) to individual observations
(from BRFSS). Thus, GEE was employed with an exchangeable correlation
matrix, logit link and the repeating unit defined as the FIPS code. Results
reported separately by sex were analyzed with stratification by sex. To
calculate the number of Americans represented, frequency weighting was
used by rounding the final weight to the nearest integer.
RESULTS
Elevation
Compared with the lowest elevation category of o500 m above
sea level (322 681 observations representing B170 million
individuals), subjects in the highest elevation category of
International Journal of Obesity (2013) 1407 – 1412
X3000 ms above sea level (236 observations representing 41 271
individuals) included lower proportions of smokers and higher
proportions of diet and physical activity recommendation compliance with other protective demographic characteristics (Table 1).
Obesity prevalence decreased with increasing elevation by
200 m increments (Figure 1). Variance in BMI was wider at lower
elevations (Figure 2), particularly in rural areas. Multiple measures
of the overall distribution of the weighted crude BMI data (outside
values, upper and lower adjacent, upper and lower hinge),
displayed by box plot (Figure 2), demonstrated a progressive
decrement in BMI with increasing elevation category.
After controlling for urbanization, temperature category and
behavioral and demographic factors, male and female Americans
living o500 m above sea level had 5.1 (95% confidence interval
(CI) 2.7–9.5) and 3.9 (95% CI 1.6–9.3) times the odds of obesity,
respectively, as compared with counterparts living X3000 m
above sea level (Table 2). The data are also presented using
o500 m as the referent group, as this is a more common exposure
(Table 2). When modeling elevation as a continuous variable with
nonhierarchical first-degree fractional polynomial regression, after
controlling for demographics, lifestyle, temperature and urbanization, the odds of obesity were B7.5% lower for Americans at
1000 m of elevation as compared with their counterparts at sea
level. Although the reduced odds of obesity were modest at the
low dose of 1000 m, the coefficient of reduced odds increased to
the squared power with increasing elevation beyond 1000 m. This
can be expressed by the formula pOe ¼ 0.925 (E2 0.2) pOc,
where pOe is predicted odds of obesity, E is elevation in km and
pOc is predicted odds of obesity based on other covariates.
Using quantile regression, there was an inverse dose response
between median BMI and elevation, after controlling for demographics, lifestyle and urbanization; those living X3000 m above
sea level had a median BMI B2.4 BMI units lower than those living
o500 m above sea level (Table 3). This nonparametric, nonhierarchical analysis is a fully adjusted point estimate with similar
results as the unadjusted graphical representation of the
distribution of BMI by altitude category seen in Figure 3. The
relationship between elevation and BMI was not substantially
altered when adjusted for diet and physical activity.
Urbanization
The crude prevalence of obesity decreased with increasing
urbanization: 30.9% in rural and nonmetro counties; 29.2% in
metro counties with o250 000 people; 28.1% in counties with
population between 250 000 and 1 million; and 26.2% in population
of 41 million. Compared with metro counties with 41 million
residents, obesity was more prevalent in metro counties with
populations of 250 000 to 1 million residents (odds ratio (OR) 1.08;
95% CI 1.02–1.15), metro counties with o250 000 (OR 1.11;
95% CI 1.05–1.18) and nonmetro or rural counties (OR 1.17; 95%
CI 1.12–1.23), after controlling for demographics, lifestyle, elevation
and temperature category using GEE (Table 1). Median BMI showed
the same trend according to urbanization (Table 3).
Ambient temperature
The relationship between temperature and obesity prevalence
was approximately parabolic when plotting mean annual temperature in 1/4 1C increments (Figure 3). Florida accounted for
over 80% of lean inhabitants residing at a mean annual
temperature of 422 1C. Among lean Floridians living at 422 1C,
25% were X65 years old. In the fully adjusted GEE model
controlling for elevation, urbanization, demographics and lifestyle,
all temperature categories (5 1C increments) were not statistically
significantly different than the highest temperature category, but
extremes of temperature category trended to the lowest odds
(Table 1). Median BMI by quantile regression was similar across
& 2013 Macmillan Publishers Limited
Elevation and obesity
JD Voss et al
1409
Table 1. Prevalence of obesity-related risk factors by lowest vs highest
elevation regions; odds ratio (OR) for association of risk factor with
obesity in multilevel analysis controlling for elevation category
X2
Elevation
N ¼ 322 681
N ¼ 236
o500 m
X3000 m
N ¼ 422 603
P-valuea
OR (95%
confidence
interval (CI))
12.22
16.29
17.67
19.52
15.86
18.45
8.63
11.70
18.99
22.30
17.92
20.46
o0.001
Sex
Women
Men
50.13
49.87
50.11
49.89
¼ 0.929
Referent
1.06 (1.03–1.10)
Employment status
Unemployed
Employed
Missing
15.59
84.05
0.36
8.39
91.61
0.00
o0.001
1.38 (1.32–1.44)
Referent
1.11 (0.86–1.43)
26.01
42.82
o0.001
0.71 (0.68–0.73)
73.81
55.55
Referent
0.18
1.63
0.72 (0.45–1.15)
17.99
43.81
25.20
13.00
5.40
47.34
31.92
15.34
o0.001
1.27 (1.19–1.34)
1.24 (1.18–1.30)
1.04 (0.98–1.10)
Referent
68.78
88.40
o0.001
Referent
12.86
0.00
1.55 (1.48–1.63)
10.96
3.51
2.89
1.00
4.15
1.03
5.66
0.77
1.16
0.31
1.26
0.98
Income
o$20k
$20k–$75k
$75k þ
Missing
Race/ethnicity
White, NonHispanic
Black, NonHispanic
Hispanic
Asian
Other
Missing
Daily servings of fruits and vegetables
o1.00
7.47
5.24
Serving
1.00–2.99
40.35
26.46
3.00–4.99
27.80
35.94
5.00 þ
15.56
21.53
Servings
Missing
8.83
10.83
Physical activity
Highly active
Active
Insufficiently
active
Inactive
Missing
o0.001
0.53 (0.49–0.56)
1.09 (1.03–1.15)
1.34 (1.28–1.40)
1.48 (1.42–1.54)
1.45 (1.40–1.50)
Referent
(1.10–1.24)
(0.27–0.37)
(1.17–1.36)
(0.87–1.11)
1.16 (1.08–1.24)
1.18 (1.13–1.24)
1.06 (1.02–1.11)
Referent
0.97 (0.91–1.04)
o0.001
29.14
17.89
19.20
52.53
9.83
16.74
25.63
8.14
7.48
13.42
20.86
78.70
0.45
11.23
87.48
1.29
o0.001
0.68 (0.65–0.71)
Referent
0.73 (0.60–0.89)
Mean annual temperature
o5 1C
0.86
5–9.9 1C
27.94
10–14.9 1C
35.63
15–19.9 1C
22.64
20 þ 1C
12.93
100.00
0.00
0.00
0.00
0.00
o0.001
0.96 (0.85–1.08)
1.03 (0.93–1.13)
1.00 (0.92–1.09)
1.03 (0.94–1.13)
Referent
Current smoker
Yes
No
Missing
& 2013 Macmillan Publishers Limited
0.59 (0.57–0.61)
0.69 (0.66–0.73)
0.84 (0.81–0.88)
Referent
0.77 (0.72–0.82)
X2
Elevation
OR
Age
18–24
25–34
35–44
45–54
55–64
65 þ
Educational level
College
degree
No college
degree
Missing
Table 1. (Continued )
Population size
Pop 1
million þ
250k to 1
million
Metro,
o250k
Rural/
nonmetro
N ¼ 322 681
N ¼ 236
o500 m
X3000 m
OR
N ¼ 422 603
P-valuea
o0.001
OR (95%
confidence
interval (CI))
53.03
0.00
Referent
20.55
0.00
1.08 (1.02–1.15)
9.86
0.00
1.11 (1.05–1.18)
16.56
100.00
1.17 (1.12–1.23)
Proportions based on frequency-weighted final weight variable rounded
to nearest integer. To categorize by age, the Centers for Disease Control
and Prevention (CDC) imputed values if necessary. The odds ratios are
exponentiated from the generalized estimating equation (GEE) model
adjusted for all variables above as well as the elevation category.
a
The P-value is based on w2 test of homogeneity using frequency weights.
temperature categories with suggestion of lower median BMIs at
the extremes of temperature category (Table 3).
DISCUSSION
Based on a cross-sectional and nationally representative data set
of 422 603 observations, representing B207 million Americans, an
inverse dose-response relationship between elevation and prevalent obesity was evident in both unadjusted and fully adjusted
models. In other words, Americans living at high elevation are less
likely to be obese. Moreover, variance in BMI narrows with
increasing elevation. This is consistent with findings from an
interventional study, in which amount of weight loss during a 33day period on Kunlun Mountain (elevation: 4678 m, location:
China) was correlated to baseline body weight (r ¼ 0.677,
Po0.01).30 Others have demonstrated that sensitivity to known
obesity risk factors is ‘more potent with increasing adiposity.’31
This study identifies low elevation and rural residence as
additional risk factors subject to this phenomenon.
There are several plausible mechanisms relating elevation and
obesity, including hypoxia, leptin signaling, metabolic demands,
norepinephrine levels and fetal/childhood growth. Relative
hypoxia is an attractive mechanism as it could link decreased
odds of obesity found in smoking (chronic carbon monoxide
exposure), urbanization (chronic air pollution exposure) and
higher elevation (hypobaric atmospheric conditions). Hypoxia also
presents a solution that is consistent with the exponential dose
response between elevation and obesity prevalence identified in
this study. Additionally, this has implications for therapy as
artificial hypoxia can be obtained at any elevation using an
altitude chamber. Prior studies have evaluated the way
oxygen saturation varies by elevation.32,33 In the future, use of
oxygen saturation prediction models or pressure altitude could be
used rather than topographical elevation to better model the
underlying constructs.
Protection against obesity in hypoxic environments may be
biologically adaptive, whereby hypoxia alters the near term
survival tradeoff between the benefit of increased energy storage
and the cost of excess body weight. Numerous observational
studies have reported that lowlanders traveling to various
elevations above 2650 m experience acute anorexia, decreased
caloric intake and weight loss.4–7,30,34,35 In a clinical trial, obese
subjects randomized to low-intensity physical activity in hypoxic
International Journal of Obesity (2013) 1407 – 1412
Elevation and obesity
JD Voss et al
1410
leptin levels, hypoxia may also improve leptin signaling through
increased production of leptin receptor. In mouse hepatocytes
exposed to chronic hypoxia, the leptin receptor gene is upregulated
more than any other gene.9 This may explain why hypoxic
conditions induce anorexia in rats, including polycythemic rats,
while other environmental conditions are held constant.37 Obese
rats exposed to artificial pulsatile hypoxia (8% O2), moreover,
decrease ad libitum food consumption for 1 month, signifying
perhaps a therapeutic potential of hypoxia for human obesity.38
Previous literature has suggested that reduced temperature at
increased elevation may lead to weight loss through catabolic
effects.21 This is biologically plausible based on animal39,40 and
human41 studies; a theoretical thermodynamic model predicted
annual caloric expenditure equivalent to 5.1–7.3 kg of adipose
tissue to maintain core temperature during a 5 1C reduction in
ambient temperature.42 There is also evidence of decreased
energy balance beyond either side of a ‘thermoneutral zone,’ at
which increased metabolic expenditures are required to cope with
either very hot or very cold temperatures.43 Temperature,
however, did not account for the effect of elevation on obesity
seen in this study.
Norepinephrine has also been implicated in the relationship
between elevation and weight loss.20 Among six subjects who
endured a simulated 40-day ascent, plasma norepinephrine
concentrations tripled, whereas epinephrine concentrations
remained stable.44 This increased sympathetic tone during ascent
alters blood flow to the gut and thereby diminishes appetite.45
Similarly, depending on ancestry, blood flow is also altered during
pregnancy at high elevation,46 and birth weight is reduced by a
curvilinear dose response.15,16,47 The lower birth weight is
attributed to reduced subcutaneous tissue.47 Decreased childhood
growth and stunting are more common at elevation.18,19
Furthermore, birth weight is correlated with BMI in adulthood.17
In multiple experimental animal models, however, reducing
prenatal nutrition increases adiposity.43 Therefore, although the
relationship between elevation and birth weight may be
mechanistic, an alternative explanation is that the relationship
between elevation and BMI exists throughout all stages of human
life. The complex relationship between elevation, hypoxia, leptin,
thermoregulation and norepinephrine underscores the need for
further research and the potential for new obesity treatments.
The inverse dose-response relationship between urbanization
and obesity was also a notable finding. This may represent
increased food security, enhanced walkability or better diet.
Objective calculation of walkability is a function of population
density,48 and rurality has been associated with poor diet.26
Although urbanization is associated with increased obesity in
India,49 our findings are consistent with data from the National
Health and Nutrition Examination Survey showing higher
prevalence of obesity in rural US residents.50 Our study adds to
this research by showing a dose-response relationship while
correcting for other geographic factors.
conditions achieved greater weight loss than those randomized to
the same activity in sham hypoxia.36 Given that many human
studies have demonstrated that elevation causes anorexia and
weight loss in the short term, our study is significant because it
also suggests a role for elevation in weight homeostasis over the
long term.
The impact of hypoxia on concentration of plasma leptin—a
hormone secreted by adipose tissue that produces negative
feedback on appetite—is complex and controversial.10,11 Some
studies have found that plasma leptin concentrations rise at
elevation5,7 or in hypoxic conditions6,14 in both traveling and
acclimatized subjects, whereas others have found unchanged or
decreased levels.8,13 Hypoxia may modulate leptin levels through
hypoxia-inducible transcription factor, which regulates both iron
metabolism and leptin gene expression.10,12 Regardless of serum
Proportion of obese
0.30
0.20
0.10
0.00
0
1
2
3
altkm rounded to nearest 0.2
Assessing Linearity Assumption −− Proportions
Figure 1. Proportion obese by elevation in 200 m increments
(unweighted).
100
bmi
80
60
40
20
0
1
2
3
4
5
6
Figure 2. Box plots of BMI distribution by elevation category using
inverse probability weight (0 ¼ o0.50 km of elevation above sea
level; 1 ¼ 0.50–0.99 km; 2 ¼ 1.00–1.49 km; 3 ¼ 1.50–1.99 km; 4 ¼ 2.00–
2.49 km; 5 ¼ 2.50–2.99; 6 ¼ X3.00 km).
Table 2.
Adjusted odds of obesity by elevation category, stratified by sex
Strata
Overall OR (95% confidence interval (CI))
Men OR (95% CI)
Women OR (95% CI)
Low referent (95% CI)
4.62 (2.20–9.69)
4.41 (2.10–9.28)
3.93 (1.86–8.26)
3.72 (1.77–7.83)
3.33 (1.58–7.02)
1.69 (0.75–3.82)
Referent
5.08 (2.72–9.48)
4.62 (2.47–8.65)
4.27 (2.26–8.06)
3.98 (2.12–7.45)
3.68 (1.95–6.95)
1.56 (0.75–3.23)
Referent
3.86 (1.61–9.28)
3.86 (1.60–9.29)
3.39 (1.41–8.19)
3.24 (1.34–7.80)
2.71 (1.12–6.56)
1.69 (0.65–4.40)
Referent
Referent
0.96 (0.90–1.02)
0.85 (0.79–0.92)
0.81 (0.74–0.87)
0.72 (0.65–0.80)
0.37 (0.26–0.52)
0.22 (0.10–0.45)
0.00–0.50 km
0.50–0.99 km
1.00–1.49 km
1.50–1.99 km
2.00–2.49 km
2.50–2.99 km
3.00–3.49 km
Odds ratios (ORs) calculated using generalized estimating equation (GEE), controlling for urbanization, temperature category, diet, physical activity, smoking
status, age group, income, college education, employment status and race/ethnicity.
International Journal of Obesity (2013) 1407 – 1412
& 2013 Macmillan Publishers Limited
Elevation and obesity
JD Voss et al
1411
Table 3. Adjusted median body mass index (BMI) by elevation,
urbanization and temperature category
BMI units
95% Confidence interval (CI)
Elevation
0.00–0.50 km
0.50–0.99 km
1.00–1.49 km
1.50–1.99 km
2.00–2.49 km
2.50–2.99 km
3.00–3.49 km
Referent
0.18
0.40
0.61
0.95
1.93
2.42
—
0.25
0.52
0.73
1.16
2.66
3.79
—
0.12
0.28
0.49
0.74
1.21
1.05
Urbanization
Nonmetro
Metro, o250k
250k to 1 million
41 million
Referent
0.22
0.36
0.67
—
0.29
0.43
0.72
—
0.14
0.30
0.61
Temperature
0.0–4.9 1C
5.0–9.9 1C
10.0–14.9 1C
15.0–19.9 1C
20.0–24.9 1C
Referent
0.20
0.19
0.16
0.14
—
0.02
0.01
0.02
0.05
—
0.38
0.37
0.34
0.33
CONFLICT OF INTEREST
The authors declare no conflict of interest.
ACKNOWLEDGEMENTS
Calculated using quantile regression, controlling for diet, physical activity,
smoking status, age group, race/ethnicity, sex, income, college education
and employment status. All calculations are inverse probability weighted
(analytic weight) without accounting for imprecision associated with
hierarchical data.
Matthew Gilbert provided help with Microsoft Access. Robert Smalley provided help
with Microsoft Excel and suggested quantile regression. Daniel Burnett first
presented and discussed BRFSS obesity maps with principal investigator and
discussed the referent group. Roger Gibson reviewed the project and discussed with
Principal Investigator. Cara Olsen provided biostatistical support and Tzucheg Kao
provided base code for GEE in SAS. The USUHS School of Medicine Office of Research
approved this study as non-human subjects research. The project was completed
while the principal investigator was in training within the USUHS Public Health and
General Preventive Medicine Residency.
DISCLAIMER
This work is the sole responsibility of the authors and does not represent the
official views of the Uniformed Services University of the Health Sciences,
Department of Defense or Virginia Commonwealth University.
AUTHOR CONTRIBUTIONS
0.40
Proportion of obese
observations in this analysis came from the counties X3000 m in
elevation. Finally, self-reporting of body weight in BRFSS is likely
underestimated among some individuals.51
In summary, this study demonstrates that obesity prevalence in
the United States is inversely associated with elevation and
urbanization after controlling for temperature, behavioral factors
and demographic factors.
Study concept and design: J Voss and P Masuoka; acquisition of data: J Voss, P
Masuoka and AI Scher; analysis and interpretation of data: J Voss, P Masuoka, BJ
Webber, AI Scher and RL Atkinson; drafting of the manuscript: J Voss, P
Masuoka and BJ Webber; critical revision of the manuscript for important
intellectual content: P Masuoka, AI Scher and RL Atkinson; statistical analysis:
J Voss, P Masuoka, BJ Webber, AI Scher and RL Atkinson; administrative,
technical or material support: P Masuoka and AI Scher; study supervision:
AI Scher and RL Atkinson.
0.20
0.00
30
10
20
t1 rounded to nearest 0.25
Assessing Linearity Assumption −− Proportions
0
Figure 3. Relationship between temperature and proportion of
population obese.
The major strength of this study is the combination of precise
geographic information with a large data set representative of
B207 million Americans. Extensive demographic and behavioral
data, furthermore, allow for evaluation of potential confounders
for obesity. However, the findings should be interpreted in light of
their limitations. Although the geographic pattern of elevation
existed before the development of macrogeographic disparities in
obesity, this study is best characterized as a cross-sectional
analysis. Although sensitivity analyses removing either
USDA-defined retirement or population loss FIPS codes from a
fully adjusted logistic model did not meaningfully decrease the
effect of elevation, person-specific duration of residence would be
preferable; however, this was not feasible with the available
data. Furthermore, the observations regarding elevation should
be interpreted with caution as the sampling and weighting
methodology was not designed to detect differences in elevation.
Observations in the highest elevation category (43000 m) were
limited to counties in only one state (Colorado) and despite
including over 500 000 observations in this data set, only 236
& 2013 Macmillan Publishers Limited
REFERENCES
1 Flegal KM, Carroll MD, Kit BK, Ogden CL. Prevalence of obesity and trends in the
distribution of body mass index among US adults, 1999–2010. JAMA 2012; 307:
491–497.
2 Ogden CL, Carroll MD, Kit BK, Flegal KM. Prevalence of obesity and trends in body
mass index among US children and adolescents, 1999–2010. JAMA 2012; 307:
483–490.
3 Centers for Disease Control and Prevention. Overweight and obesity 2012, [cited
2012 June 15]; Available from http://www.cdc.gov/obesity/data/adult.html.
4 Hamad N, Travis SP. Weight loss at high altitude: pathophysiology and practical
implications. Eur J Gastroenterol Hepatol 2006; 18: 5–10.
5 Shukla V, Singh SN, Vats P, Singh VK, Singh SB, Banerjee PK. Ghrelin and leptin
levels of sojourners and acclimatized lowlanders at high altitude. Nutr Neurosci
2005; 8: 161–165.
6 Lippl FJ, Neubauer S, Schipfer S, Lichter N, Tufman A, Otto B et al. Hypobaric
hypoxia causes body weight reduction in obese subjects. Obesity (Silver Spring)
2010; 18: 675–681.
7 Tschop M, Strasburger CJ, Hartmann G, Biollaz J, Bartsch P. Raised leptin concentrations at high altitude associated with loss of appetite. Lancet 1998; 352:
1119–1120.
8 Vats P, Singh VK, Singh SN, Singh SB. High altitude induced anorexia: effect of
changes in leptin and oxidative stress levels. Nutr Neurosci 2007; 10: 243–249.
9 Baze MM, Schlauch K, Hayes JP. Gene expression of the liver in response to
chronic hypoxia. Physiol Genomics 2010; 41: 275–288.
10 Yingzhong Y, Droma Y, Rili G, Kubo K. Regulation of body weight by
leptin, with special reference to hypoxia-induced regulation. Intern Med 2006; 45:
941–946.
11 Sierra-Johnson J, Romero-Corral A, Somers VK, Johnson BD. Effect of altitude on
leptin levels, does it go up or down? J Appl Physiol 2008; 105: 1684–1685.
International Journal of Obesity (2013) 1407 – 1412
Elevation and obesity
JD Voss et al
1412
12 Ambrosini G, Nath AK, Sierra-Honigmann MR, Flores-Riveros J. Transcriptional
activation of the human leptin gene in response to hypoxia. Involvement of
hypoxia-inducible factor 1. J Biol Chem 2002; 277: 34601–34609.
13 Vats P, Singh SN, Shyam R, Singh VK, Singh SB, Banerjee PK et al. Leptin may not
be responsible for high altitude anorexia. High Alt Med Biol 2004; 5: 90–92.
14 Snyder EM, Carr RD, Deacon CF, Johnson BD. Overnight hypoxic exposure and
glucagon-like peptide-1 and leptin levels in humans. Appl Physiol Nutr Metab
2008; 33: 929–935.
15 Gonzales GF, Tapia V. Birth weight charts for gestational age in 63,620 healthy
infants born in Peruvian public hospitals at low and at high altitude. Acta Paediatr
2009; 98: 454–458.
16 Moore LG, Shriver M, Bemis L, Vargas E. An evolutionary model for identifying
genetic adaptation to high altitude. Adv Exp Med Biol 2006; 588: 101–118.
17 Sorensen HT, Sabroe S, Rothman KJ, Gillman M, Fischer P, Sorensen TI. Relation
between weight and length at birth and body mass index in young adulthood:
cohort study. BMJ 1997; 315: 1137.
18 Yip R, Binkin NJ, Trowbridge FL. Altitude and childhood growth. J Pediatr 1988;
113: 486–489.
19 Dang S, Yan H, Yamamoto S. High altitude and early childhood growth retardation: new evidence from Tibet. Eur J Clin Nutr 2008; 62: 342–348.
20 Barnholt KE, Hoffman AR, Rock PB, Muza SR, Fulco CS, Braun B et al. Endocrine
responses to acute and chronic high-altitude exposure (4300 meters): modulating
effects of caloric restriction. Am J Physiol Endocrinol Metab 2006; 290: E1078–E1088.
21 Sherpa LY, Deji, Stigum H, Chongsuvivatwong V, Thelle DS, Bjertness E. Obesity in
Tibetans aged 30-70 living at different altitudes under the north and south faces
of Mt. Everest. Int J Environ Res Public Health 2010; 7: 1670–1680.
22 Tyagi R, Tungdim MG, Bhardwaj S, Kapoor S. Age, altitude and gender differences
in body dimensions. Anthropol Anz 2008; 66: 419–434.
23 Lund EAP, Kirk C, Klausner J. Obesity in adult dogs from private US veterinary
practices. Intern J Appl Res Vet Med 2006; 4: 10.
24 Khalid Mel H. Is high-altitude environment a risk factor for childhood overweight
and obesity in Saudi Arabia? Wilderness Environ Med 2008; 19: 157–163.
25 Baracco R, Mohanna S, Seclen S. A comparison of the prevalence of metabolic
syndrome and its components in high and low altitude populations in peru.
Metab Syndr Relat Disord 2007; 5: 55–62.
26 Lutfiyya MN, Chang LF, Lipsky MS. A cross-sectional study of US rural adults’
consumption of fruits and vegetables: do they consume at least five servings
daily? BMC public health 2012; 12: 280.
27 Centers for Disease Control and Prevention (CDC). Behavioral Risk Factor
Surveillance System Survey Data. Atlanta, Georgia: US Department of Health and
Human Services, Centers for Disease Control and Prevention, 2011.
28 Parker T. 2004 ERS County Typology Codes. In USDAUSDA, Economic Research
Service, 2009.
29 Hijmans R, Cameron S, Parra J, Jones P, Jarvis A. Very high resolution interpolated
climate surfaces for global land areas. Int J Climatol 2005; 25: 1965–1978.
30 Ge RL, Wood H, Yang HH, Liu YN, Wang XJ, Babb T. The body weight loss during
acute exposure to high-altitude hypoxia in sea level residents. Sheng Li Xue Bao
2010; 62: 541–546.
31 Williams PT. Evidence that obesity risk factor potencies are weight dependent, a
phenomenon that may explain accelerated weight gain in western societies. PLoS
One 2011; 6: e27657.
32 Goldberg S, Buhbut E, Mimouni FB, Joseph L, Picard E. Effect of moderate elevation above sea level on blood oxygen saturation in healthy young adults.
Respiration 2012; 84: 207–211.
33 Tannheimer M, Thomas A, Gerngross H. Oxygen saturation course and altitude
symptomatology during an expedition to broad peak (8047 m). Int J Sports Med
2002; 23: 329–335.
34 Wagner PD. Operation Everest II. High Alt Med Biol 2010; 11: 111–119.
35 Aeberli I, Erb A, Spliethoff K, Meier D, Gotze O, Fruhauf H et al. Disturbed eating at
high altitude: influence of food preferences, acute mountain sickness and satiation hormones. Eur J Nutr 2012; e-pub ahead of print 11 May 2012.
36 Netzer NC, Chytra R, Kupper T. Low intense physical exercise in normobaric
hypoxia leads to more weight loss in obese people than low intense physical
exercise in normobaric sham hypoxia. Sleep Breath 2008; 12: 129–134.
37 Norese MF, Lezon CE, Alippi RM, Martinez MP, Conti MI, Bozzini CE. Failure of
polycythemia-induced increase in arterial oxygen content to suppress the anorexic
effect of simulated high altitude in the adult rat. High Alt Med Biol 2002; 3: 49–57.
38 Quintero P, Milagro FI, Campion J, Martinez JA. Impact of oxygen availability on
body weight management. Med Hypotheses 2010; 74: 901–907.
39 Vaanholt LM, Daan S, Schubert KA, Visser GH. Metabolism and aging: effects of
cold exposure on metabolic rate, body composition, and longevity in mice.
Physiol Biochem Zool 2009; 82: 314–324.
40 Zhao ZJ, Chi QS, Cao J, Han YD. The energy budget, thermogenic capacity and
behavior in Swiss mice exposed to a consecutive decrease in temperatures. J Exp
Biol 2010; 213(Pt 23): 3988–3997.
41 Wijers SL, Schrauwen P, Saris WH, van Marken Lichtenbelt WD. Human skeletal
muscle mitochondrial uncoupling is associated with cold induced adaptive
thermogenesis. PLoS One 2008; 3: e1777.
42 Hansen JC, Gilman AP, Odland JO. Is thermogenesis a significant causal factor in
preventing the "globesity" epidemic? Med Hypotheses 2010; 75: 250–256.
43 McAllister EJ, Dhurandhar NV, Keith SW, Aronne LJ, Barger J, Baskin M et al. Ten
putative contributors to the obesity epidemic. Crit Rev Food Sci Nutr 2009; 49:
868–913.
44 Young PM, Rose MS, Sutton JR, Green HJ, Cymerman A, Houston CS. Operation
Everest II: plasma lipid and hormonal responses during a simulated ascent of Mt.
Everest. J Appl Physiol 1989; 66: 1430–1435.
45 Loshbaugh JE, Loeppky JA, Greene ER. Effects of acute hypobaric hypoxia on
resting and postprandial superior mesenteric artery blood flow. High Alt Med Biol
2006; 7: 47–53.
46 Wilson MJ, Lopez M, Vargas M, Julian C, Tellez W, Rodriguez A et al. Greater
uterine artery blood flow during pregnancy in multigenerational (Andean) than
shorter-term (European) high-altitude residents. Am J Physiol Regul Integr Comp
Physiol 2007; 293: R1313–R1324.
47 Galan HL, Rigano S, Radaelli T, Cetin I, Bozzo M, Chyu J et al. Reduction of
subcutaneous mass, but not lean mass, in normal fetuses in Denver, Colorado. Am
J Obstet Gynecol 2001; 185: 839–844.
48 Hankey S, Marshall JD, Brauer M. Health impacts of the built environment: withinurban variability in physical inactivity, air pollution, and ischemic heart disease
mortality. Environ Health Perspect 2012; 120: 247–253.
49 Ebrahim S, Kinra S, Bowen L, Andersen E, Ben-Shlomo Y, Lyngdoh T et al.
The effect of rural-to-urban migration on obesity and diabetes in India: a crosssectional study. Plos Med 2010; 7: e1000268.
50 Befort CA, Nazir N, Perri MG. Prevalence of obesity among adults from rural and
urban areas of the United States: findings from NHANES (2005-2008). J Rural
Health 2012; 28: 392–397.
51 Ezzati M, Martin H, Skjold S, Vander Hoorn S, Murray CJ. Trends in national and
state-level obesity in the USA after correction for self-report bias: analysis of
health surveys. J R Soc Med 2006; 99: 250–257.
Supplementary Information accompanies the paper on International Journal of Obesity website (http://www.nature.com/ijo)
International Journal of Obesity (2013) 1407 – 1412
& 2013 Macmillan Publishers Limited