Substituting dietary monounsaturated fat for saturated fat is associated

with increased daily physical activity and resting energy expenditure

and with changes in mood
13
C Lawrence Kien, Janice Y Bunn, Connie L Tompkins, Julie A Dumas, Karen I Crain, David B Ebenstein, Timothy R Koves,
and Deborah M Muoio
ABSTRACT
Background: The Western diet increases risk of metabolic disease.
Objective: We determined whether lowering the ratio of saturated
fatty acids to monounsaturated fatty acids in the Western diet would
affect physical activity and energy expenditure.
Design: With the use of a balanced design, 2 cohorts of 18 and 14
young adults were enrolled in separate randomized, double-masked,
crossover trials that compared a 3-wk highpalmitic acid diet (HPA;
similar to the Western diet fat composition) to a lowpalmitic acid
and higholeic acid diet (HOA; similar to the Mediterranean diet fat
composition). All foods were provided by the investigators, and the
palmitic acid (PA):oleic acid (OA) ratio was manipulated by adding
different oil blends to the same foods. In both cohorts, we assessed
physical activity (monitored continuously by using accelerometry)
and resting energy expenditure (REE). To gain insight into a possible
mood disturbance that might explain changes in physical activity,
the Prole of Mood States (POMS) was administered in cohort 2.
Results: Physical activity was higher during the HOA than during
the HPA in 15 of 17 subjects in cohort 1 (P = 0.008) (mean: 12%
higher; P = 0.003) and in 12 of 12 subjects in the second, conr-
matory cohort (P = 0.005) (mean: 15% higher; P = 0.003). When
the HOA was compared with the HPA, REE measured during the
fed state was 3% higher for cohort 1 (P ,0.01), and REE was 4.5%
higher in the fasted state for cohort 2 (P = 0.04). POMS testing
showed that the anger-hostility score was signicantly higher during
the HPA (P = 0.007).
Conclusions: The replacement of dietary PA with OA was associ-
ated with increased physical activity and REE and less anger. Be-
sides presumed effects on mitochondrial function (increased REE),
the dietary PA:OA ratio appears to affect behavior. The second co-
hort was derived from a study that was registered at clinicaltrials.
gov as R01DK082803. Am J Clin Nutr 2013;97:68997.
INTRODUCTION
During the past several decades the incidences of obesity and
associated health problems, such as type 2 diabetes, have greatly
increased (1, 2). This obesity epidemic is especially evident in
westernized countries where sedentary lifestyles, the over-
consumption of calorically dense foods, and a high saturated fat
intake impede optimal health (2, 3). Some data have suggested
that an excessive energy intake, not abnormally decreased energy
expenditure, is the principal cause of the obesity epidemic (1).
However, obesity is fundamentally a disorder of energy-balance
regulation, and low resting energy expenditure (REE)
4
or sed-
entary behavior also can contribute to weight gain (4, 5). Despite
considerable industry investment in the formulation of phar-
maceutical agents that promote energy wasting, efforts in the
development of antiobesity drugs continue to fall short (6). Phys-
ical activity still remains the safest and most commonly prescribed
means of raising energy expenditure.
The 2 most prevalent fatty acids (FAs) in the Western diet are
the SFA palmitic acid (PA; 16:0) and the MUFA oleic acid (OA;
C18:1), each of which is present in approximately equal amounts
as a percentage of dietary energy (SFA: 13.7% energy; MUFA:
11.7%) (7). Although an increased dietary PA:OA ratio has been
linked to elevated serum LDL concentrations and ischemic heart
disease (8), the widespread and effective use of statin drugs might
lessen the concern about lowering SFA intake. However, we have
been interested in how SFA compared with MUFA in the diet
affects risk of diabetes and obesity (9); our previous studies
suggested that lowering the dietary PA:OA ratio increased daily
energy requirements for weight maintenance as well as the en-
ergy cost of physical activity (10, 11). In this article, we asked
whether a shift in the FA composition of the diet might affect
physical activity and REE.
Dietary FAmay inuence physical activity via effects on mood
(eg, anger and hostility) (12), but a meta-analysis has also shown
that low-intensity exercise improves the self-rated effect (13).
1
Fromthe Departments of Pediatrics (CLK), Medicine (CLK, KIC, and
DBE), Medical Biostatistics (JYB), Rehabilitation and Movement Sciences
(CLT), and Psychiatry (JAD), University of Vermont, Burlington, VT, and
the Departments of Medicine and Pharmacology & Cancer Biology, Sarah W
Stedman Nutrition and Metabolism Center, Duke University, Durham, NC
(TRK and DMM).
2
This study was supported by the NIH (grants R01 DK073284 and R01
DK082803) and conducted at the University of Vermont General Clinical
Research Center funded by the National Center for Research Resources,
NIH, United States Public Health Service (grant RR00109).
3
Address correspondence to CL Kien, University of Vermont College of
Medicine, Room #175A, University of Vermont Colchester Research Facil-
ity, 208 South Park Drive, Colchester, VT 05446. E-mail: cl.kien@uvm.edu.
4
Abbreviations used: FA, fatty acid; GCRC, General Clinical Research
Center; HOA, higholeic acid diet; HPA, highpalmitic acid diet; OA, oleic
acid; PA, palmitic acid; POMS, Prole of Mood States; REE, resting energy
expenditure; TMD, total mood disturbance.
ReceivedSeptember 24, 2012. Accepted for publication January 15, 2013.
First published online February 27, 2013; doi: 10.3945/ajcn.112.051730.
Am J Clin Nutr 2013;97:68997. Printed in USA. 2013 American Society for Nutrition 689

b
y

g
u
e
s
t

o
n

S
e
p
t
e
m
b
e
r

1
3
,

2
0
1
4
a
j
c
n
.
n
u
t
r
i
t
i
o
n
.
o
r
g
D
o
w
n
l
o
a
d
e
d

f
r
o
m


b
y

g
u
e
s
t

o
n

S
e
p
t
e
m
b
e
r

1
3
,

2
0
1
4
a
j
c
n
.
n
u
t
r
i
t
i
o
n
.
o
r
g
D
o
w
n
l
o
a
d
e
d

f
r
o
m


b
y

g
u
e
s
t

o
n

S
e
p
t
e
m
b
e
r

1
3
,

2
0
1
4
a
j
c
n
.
n
u
t
r
i
t
i
o
n
.
o
r
g
D
o
w
n
l
o
a
d
e
d

f
r
o
m


b
y

g
u
e
s
t

o
n

S
e
p
t
e
m
b
e
r

1
3
,

2
0
1
4
a
j
c
n
.
n
u
t
r
i
t
i
o
n
.
o
r
g
D
o
w
n
l
o
a
d
e
d

f
r
o
m


b
y

g
u
e
s
t

o
n

S
e
p
t
e
m
b
e
r

1
3
,

2
0
1
4
a
j
c
n
.
n
u
t
r
i
t
i
o
n
.
o
r
g
D
o
w
n
l
o
a
d
e
d

f
r
o
m


b
y

g
u
e
s
t

o
n

S
e
p
t
e
m
b
e
r

1
3
,

2
0
1
4
a
j
c
n
.
n
u
t
r
i
t
i
o
n
.
o
r
g
D
o
w
n
l
o
a
d
e
d

f
r
o
m


b
y

g
u
e
s
t

o
n

S
e
p
t
e
m
b
e
r

1
3
,

2
0
1
4
a
j
c
n
.
n
u
t
r
i
t
i
o
n
.
o
r
g
D
o
w
n
l
o
a
d
e
d

f
r
o
m


b
y

g
u
e
s
t

o
n

S
e
p
t
e
m
b
e
r

1
3
,

2
0
1
4
a
j
c
n
.
n
u
t
r
i
t
i
o
n
.
o
r
g
D
o
w
n
l
o
a
d
e
d

f
r
o
m


b
y

g
u
e
s
t

o
n

S
e
p
t
e
m
b
e
r

1
3
,

2
0
1
4
a
j
c
n
.
n
u
t
r
i
t
i
o
n
.
o
r
g
D
o
w
n
l
o
a
d
e
d

f
r
o
m


b
y

g
u
e
s
t

o
n

S
e
p
t
e
m
b
e
r

1
3
,

2
0
1
4
a
j
c
n
.
n
u
t
r
i
t
i
o
n
.
o
r
g
D
o
w
n
l
o
a
d
e
d

f
r
o
m


b
y

g
u
e
s
t

o
n

S
e
p
t
e
m
b
e
r

1
3
,

2
0
1
4
a
j
c
n
.
n
u
t
r
i
t
i
o
n
.
o
r
g
D
o
w
n
l
o
a
d
e
d

f
r
o
m


b
y

g
u
e
s
t

o
n

S
e
p
t
e
m
b
e
r

1
3
,

2
0
1
4
a
j
c
n
.
n
u
t
r
i
t
i
o
n
.
o
r
g
D
o
w
n
l
o
a
d
e
d

f
r
o
m


b
y

g
u
e
s
t

o
n

S
e
p
t
e
m
b
e
r

1
3
,

2
0
1
4
a
j
c
n
.
n
u
t
r
i
t
i
o
n
.
o
r
g
D
o
w
n
l
o
a
d
e
d

f
r
o
m

Thus, we examined the relation between dietary FA, physical
activity behavior, and mood.
SUBJECTS AND METHODS
Subjects, screening, and overall design
Two cohorts of adult volunteers were derived from 2 separate,
double-masked, crossover trials that used an identical dietary
protocol for the determination of effects of changes in the dietary
PA:OA ratio. Each study (cohorts 1 and 2) used a crossover
design; subjects were randomly assigned into 2 groups that
differed with respect to the treatment order; the trial lasted for 2
treatment periods. For cohort 1, the rst volunteer began the study
on 30 August 2007; therefore, the study was not registered as
a clinical trial. However, cohort 2 was derived from a study that
was registered at clinicaltrials.gov as R01DK082803.
The procedures followed for the 2 clinical studies were in
accordance with the ethical standards of the relevant institutional
committees. Cohort 1 consisted of healthy, sedentary, nonobese
healthy men (n = 9) and women (n = 9) aged 1840 y. Some
salient characteristics of these subjects are shown in Table 1.
Exclusion criteria included prescription medication, regular
aerobic exercise training, dyslipidemia (14), evidence of type 2
diabetes (15), and women with abnormal ovulation. In addition,
subjects were excluded who manifested an abnormally high
value (.4.65) for the HOMA-IR (16). Subjects in cohort 1
participated in a trial in which comprehensive transcriptional
and metabolomic proling was used to assess the effects of
dietary FA composition on metabolic outcomes, such as insulin
sensitivity (17).
To conrm results from the rst cohort, we used data collected
fromcohort 2 (Table 1). This cohort was comprised of 12 subjects
(6 men and 6 women) with physical activity measurements; in
one of these subjects, REE was not measured because of technical
issues. Mood data were collected for the last 10 subjects studied.
We were able to study mood only in 2 additional subjects (for
a total of 6 men and 6 women with mood data). Thus, a total of 14
individual subjects contributed to data on physical activity, mood,
and/or REE for cohort 2.
The recruitment strategy for cohort 2 was identical to that for
cohort 1, except for 2 factors. First, in contrast to in cohort 1, we
allowed women in cohort 2 who were receiving hormonal con-
traception. Second, subjects were recruited in 2 ranges of BMI
(in kg/m
2
) of 1825 and $30. However, after an initial clinical
evaluation, we evaluated body composition during screening in 2
male subjects with respective BMIs of 31.7 and 27.0; these men
were reassigned respectively to the lean and obese categories after
their BMI was corrected for an empirical relation between BMI
and percentage of body fat (18). On this basis, there were 2 of 6
men and 2 of 6 women who were judged to be obese. Cohort 2
participated in a trial designed in part to investigate how dietary
FA composition affects palmitate metabolism.
Diets
Detailed screening of cohort 1 indicated a habitual intake of
37% of kilocalories from total fat, 14.5% of kilocalories from
TABLE 1
Demographic and metabolic characteristics
1
Cohort 1 Cohort 2
M W Lean M Obese M Lean W Obese W
Age (y) 28.9 6 2.5 30.0 6 2.2 30.5 6 2.8 38.0 6 0.5 24.1 6 1.1 26.8 6 4.6
Body weight (kg) 77.9 6 3.6 66.9 6 4.0 74.9 6 9.6 87.6 6 12.5 61.1 6 2.6 115.7 6 27.8
Height (cm) 180.8 6 2.3 171.1 6 1.3 176.6 6 4.1 174.8 6 8.0 169.0 6 2.1 169.1 6 2.8
BMI (kg/m
2
) 23.8 6 0.9 22.8 6 1.3 23.7 6 2.2 28.5 6 1.5
2
21.4 6 0.9 40.2 6 8.4
Abdominal circumference (cm) 81.0 6 3.8 80.6 6 3.9 84.3 6 5.1 94.4 6 6.7 76.6 6 2.8 107.8 6 14.2
Body fat (%)
3
17.8 6 2.7 30.6 6 3.1 22.0 6 2.8 32.2 6 0.7 27.3 6 2.2 49.1 6 0.2
FFM (kg) 64.4 6 3.0 46.5 6 1.5 59.5 6 7.8 59.4 6 8.2 44.5 6 1.8 59.5 6 13.9
TC (mmol/L) 3.7 6 0.2 3.9 6 0.3 4.9 6 0.2 5.1 6 0.0 4.3 6 0.5 5.0 6 0.8
LDL (mmol/L) 2.1 6 0.2 2.0 6 0.2 2.8 6 0.2 3.2 6 0.2 2.3 6 0.5 3.3 6 0.7
HDL (mmol/L) 1.3 6 0.1 1.6 6 0.1 1.7 6 0.2 1.1 6 0.2 1.7 6 0.3 1.1 6 0.0
LDL:HDL ratio 1.7 6 0.3 1.2 6 0.1 1.7 6 0.2 3.0 6 0.5 1.6 6 0.5 3.1 6 0.6
TAG (mmol/L) 0.6 6 0.1 0.6 6 0.1 0.8 6 0.1 1.6 6 0.9 0.8 6 0.1 1.4 6 0.3
Fasting insulin (pmol/L) 18.9 6 3.6 24.2 6 5.1 27.1 6 4.7 53.8 6 24.0 26.0 6 2.3 87.5 6 6.3
Fasting glucose (mmol/L) 4.4 6 0.1 4.3 6 0.1 4.5 6 0.2 4.9 6 0.2 4.2 6 0.1 4.3 6 0.0
HOMA-IR 0.5 6 0.1 0.7 6 0.2 0.8 6 0.2 1.6 6 0.7 0.7 6 0.1 2.4 6 0.2
ALT (U/L) 30.4 6 5.3 22.6 6 1.5 29.8 6 3.2 33.0 6 2.0 24.8 6 6.5 29.0 6 2.0
Hemoglobin (g/dL) 14.4 6 0.4 12.8 6 0.5 14.9 6 0.1 14.2 6 0.4 12.6 6 0.2 13.1 6 0.3
NEFAs (mmol/L) 0.4 6 0.1 0.3 6 0.1 ND ND ND ND
1
All values are means 6 SEMs. n = 18 in cohort 1 and n = 14 in cohort 2 (includes subjects with physical activity, Prole of Mood States, or resting
energy-expenditure data presented in Results and described in Subjects and Methods). There were 2 obese men and 2 obese women in cohort 2. ALT, serum
concentration of alanine aminotransferase; FFM, fat-free mass; HDL, serum concentration of HDL cholesterol; LDL, serum concentration of LDL cholesterol;
ND, nondetermined; NEFA, nonesteried fatty acid; TAG, serum concentration of triacylglycerol (triglycerides); TC, serum total cholesterol concentration.
2
See Subjects and Methods for an explanation of why one man with BMI (in kg/m
2
) .25 but ,30 was included in our obese group because his
percentage of body fat was exceptionally high for his BMI.
3
Data were obtained at the time of screening except for the percentage of body fat, FFM, and NEFAs, which were obtained at the end of the 1-wk
baseline diet.
690 KIEN ET AL

b
y

g
u
e
s
t

o
n

S
e
p
t
e
m
b
e
r

1
3
,

2
0
1
4
a
j
c
n
.
n
u
t
r
i
t
i
o
n
.
o
r
g
D
o
w
n
l
o
a
d
e
d

f
r
o
m

saturated fat, and 12% of kilocalories from monounsaturated fat,
which was consistent with the usual American diet (7). The
habitual diet of cohort 2 was similar at 35.8%of kilocalories from
total fat, 12.0% of kilocalories from saturated fat, and 13.0% of
kilocalories from monounsaturated fat. All subjects in both co-
horts ingested solid-food diets until after completion of physical
activity measurements (as well as mood assessment in cohort 2).
All foods and beverages, except water, were provided to subjects
at an energy intake sufcient to maintain body weight. After
screening, a low-fat and low-PA baseline-control diet was fed for
7 d (protein: 19.7% of kilocalories; carbohydrate: 51.6% of ki-
localories; fat: 28.4% of kilocalories; PA: 5.3% of kilocalories;
and OA: 15.9% of kilocalories) (14). Subjects participated in
a crossover study of 3-wk diet periods (separated by another 1 wk
of the control diet) that consisted of a diet that resembled the
habitual diet and was a highpalmitic acid diet (HPA; similar to
Western diet fat composition) (fat: 40.4% of kilocalories; PA:
16.0% of kilocalories; and OA: 16.2% of kilocalories) or a low-
PA and higholeic acid diet (HOA; similar to the Mediterranean
diet fat composition) (fat: 40.1% of kilocalories; PA: 2.4% of
kilocalories; and OA: 28.8% of kilocalories) (Table 2). For all 3
diets, the FA composition was varied by adding oil blends to the
6 precisely formulated meals that composed the control diet and
the 9 meals that composed experimental diets.
Foods, including chicken and turkey (the only sources of
meat), were all very low in fat. Thus, FAs were mainly provided
by vegetable-oil blends appropriate to each diet (Natural Oils In-
ternational Inc). The HPA and HOA otherwise contained the exact
same foods with a 3-d rotating menu. These oils, at roomtemperature,
were mixed with foods that had been warmed; thus, these oils were
not used for cooking. The oil blend for the control diet consisted of
palm oil (36.9%), high-oleic sunower oil (19.3%), and hazelnut oil
(43.8%). The HPA oil blend consisted of palm oil (89%), peanut oil
(6.75%), and virgin olive oil (4.25%), and the HOA blend consisted
only of hazelnut oil. Except for the virgin olive oil used only in the
HPA, all natural oils were rst extracted (eg, centrifugation) and
rened [alkaline rening (crude oil treated with alkali to separate
impurities from triglycerides) and ltering to remove impurities such
as seed fragments]. The HOA and HPA had identical, low glycemic
loads (10.7; average of 3 d of menus) (19, 20).
Diets were prescribed in a random order, stratied by sex, and
separated by a 1-wk period of the baseline-control diet. All foods
and drinks, except water, were provided by the General Clinical
Research Center (GCRC), and body-energy balances were main-
tained throughout the study as previously described (21). For cohort
1, 10 of 17 subjects ingested the HPA rst, and in cohort 2, equal
numbers of subjects ingested the HPA or HOA rst. Thus, boredom
or fatigue from the study was not weighted toward the HPAbecause
fewer subjects in the combined cohorts ingested the HPA last.
However, the diet order was considered in the statistical analysis.
Subjects in the rst cohort ate breakfast in the GCRC from
Sunday to Friday, but most subjects chose to eat their 2 remaining
meals each day at home. Subjects in the second cohort ate breakfast
in the GCRC from Monday to Friday and at home on Saturday and
Sunday. Each meal was packaged and ready to be reheated by using
either an oven or microwave. Because the foods, themselves, were
practically devoid of fat, subjects also were given containers of oil to
add to each meal after it had been reheated. Subjects also were given
instructions regarding convenient ways to add the oils to various
food items on the menu. Each day, subjects completed and signed
a questionnaire that attested to their having eaten all of the food (and
food oil) and to not having consumed any food or drink, except
water, not on the menu. On Sunday (cohort 1) or Monday (cohort 2),
volunteers completed questionnaires pertaining to the previous 1 or
2 d. All food and oil containers were inspected each day to be sure all
food and oil was consumed. Subjects were given instructions to use
spatulas that were provided to help scrape all oil from its container
but, ultimately, were instructed to lick the oil container to nally
empty it. Occasionally, there was evidence of incomplete food or oil
consumption or a subject would admit to the ingestion, usually
inadvertently, of, eg, a cookie or mint. However, any consistent
noncompliance was grounds for removal from the study. Fortu-
nately, only one subject left the study on day 3 because of an
aversion to the food (ie, a dislike of fat-free cottage cheese). Because
completely following the diet was a dichotomous issue, and subjects
were queried each day about this, we did not use a detailed diet
history during the study. Because the food was provided ready to eat
to subjects, except for the reheating of the nonfat component, and
food and oil containers were returned for daily inspection, there was
no need to use photography, eg, to showwhat food was offered to the
subject or was not eaten. For the rst cohort, the average number of
days when food was returned during the HPA and HOA was 1.33
and 1.67 d, respectively, and the average daily consumption of
oil for the HPA and HOA as a percentage of the total oil ad-
ministered (127.8 and 127.6 g/d, respectively) was 99.9% and
99.2%, respectively.
Assessment of physical activity
Because both cohorts participated in studies in which discrete
outcomes were being assessed for the entire diet period, we
wanted an integrated measurement of physical activity over this
TABLE 2
Composition of experimental diets
1
HPA HOA
Percentage of kcal
Protein 16.8 16.89
Carbohydrate 42.85 43.51
Fat 40.45 40.13
Fatty acid prole (g/100 g)
Palmitic 40.29 4.59
Oleic 39.95 74.8
Linoleic 10.37 14.29
Stearic 4.22 2.81
a-Linolenic 0.18 0.14
Myristic 0.97 0
Palmitoleic 0.18 0.12
Eicosapentaenoic 0 0
Docosahexaenoic 0 0
Arachidonic 0 0
Percentage of kcal
12:0 0 0
14:0 0.42 0.05
16:0 16.02 2.37
18:0 1.8 1.27
18:1 16.23 28.75
18:3 0.16 0.19
18:2 4.97 6.36
1
HOA, higholeic acid diet; HPA, highpalmitic acid diet.
DIETARY FATTY ACIDS AND PHYSICAL ACTIVITY 691

b
y

g
u
e
s
t

o
n

S
e
p
t
e
m
b
e
r

1
3
,

2
0
1
4
a
j
c
n
.
n
u
t
r
i
t
i
o
n
.
o
r
g
D
o
w
n
l
o
a
d
e
d

f
r
o
m

same period. Thus, data were derived from recordings during
days 37 of the control diet and days 221 of each experi-
mental diet for cohort 1 and during days 36 of the control diet
and days 218 of each experimental diet for cohort 2. For
cohort 1, because of the dysfunction of the activity monitor,
data presented in this article concern only 8 men. For both
cohorts, physical activity (counts $ min
21
$ d
21
) was assessed
daily during nonsleeping hours by using an activity monitor
that was worn at the hip (belt at the waist) (ActiGraph GT1M;
ActiGraph).
The activity monitor is a small, electromechanical device that
records acceleration and deceleration of movement. The activity
monitor records these accelerations and decelerations as activity
counts and provides data for different intensities of physical
activity. These counts are linearly related to the intensity of the
movement performed by the participant and are summed over
a set period of time (epoch). Because activity monitors do not rely
on self-report, they provide an objective measure of physical
activity, which correlates with total and activity-related energy
expenditure measured by using the doubly labeled water tech-
nique (22).
The reliability of this activity monitor increases when subjects
wear them $3 compared with 1 d (23). The reported intradevice
and interdevice reliabilities of this activity monitor are 2.9% and
3.5%, with a mechanical shaker used to induce movement (24).
The actual use of this particular monitor, which is worn at the
hip, has been shown to account for w7679% of the variability
in activity-related energy expenditure, which is measured by
using a respiratory room calorimeter with positive predictive
values for sedentary, light, moderate, and vigorous activities of
80%, 66%, 69%, and 74%, respectively (25).
With the use of the framework of Treuth et al (26), we also
analyzed minutes spent and average physical activity at
sedentary, light, or moderate-vigorous physical activity levels
(,100, 1012999, and $3000 counts $ min
21
$ d
21
, re-
spectively) for cohort 1 only. One of the denitions of the
word sedentary is sitting. Because we showed data related to
the sedentary range of activity, we note that sedentary activ-
ities included resting and watching TV in the research by
Treuth et al (26).
Indirect calorimetry
For cohort 1, on day 20 of each experimental diet period (HPA
and HOA), after an evening meal at 1800 (one-third of the daily
energy intake), we carried out indirect calorimetry overnight in
both fed and fasted states, as previously described (10). Except
for bedside bathroom privileges, subjects remained in bed for the
duration of the indirect calorimetry study. Measurements of
oxygen consumption (
_
VO
2
), and carbon dioxide production
(
_
VCO
2
) (Vmax SPECTRA 29; Sensor Medics Corp) were used
for 20 min each time at 60-min intervals postmeal for 7 h (fed
state) (1740, 1900, 2000, 2100, 2200, 2300, 2400, and 0100)
and at 120-min intervals until 11 h postmeal (fasted state)
(0300 and 0500) (10). Urine was collected during the entire
interval, and protein oxidation was estimated from the rate of
urea nitrogen for both the fed (17200120) period and the fasted
period (01200720) (10). REE and substrate utilization were
calculated according to standard procedures by using urine urea
nitrogen measurements as estimates of protein oxidation (10, 27,
28) and Weirs equation (29) as follows:
Total kilocalories per unit of time eg; REE

3:941 3
_
V O
2

1:106 3
_
V CO
2

22:17 urinary urea nitrogen excretion

For cohort 2, indirect calorimetry in the fed state was carried out
on day 21 of each experimental diet as part of a [1-
13
C]-acetate
recovery study to correct for labeled carbon dioxide recovery during
palmitate tracer studies on the 2 d before the acetate test. The
protocol consisted of a 9-h administration (every 20 min) of a liquid
diet that was formulated on the basis of either experimental diet
(each dose was one-twenty-seventh or 3.7% of the daily energy
intake). Fed measurements were obtained at 360 and 420 min after
the initiation of feedings; subjects were allowed to rest in bed or sit
in a chair between indirect calorimetry measurements, which,
however, were conducted in the supine position. One measurement
of REE (50 min) in the fasted state was made 10.3 h after the last
20-min dose of formula (05000550).
Assessment of mood (cohort 2 only)
On day 4 of the control diet and day 12 or 13 of experimental
diets, subjects were asked to complete the Prole of Mood States
(POMS) (30), which is a self-rating questionnaire. Subscale scores
for 5 negative mood states (tension-anxiety, depression-dejection,
anger-hostility, fatigue-inertia, and confusion-bewilderment states)
and one positive mood state (vigor-activity) were calculated. A total
mood disturbance (TMD) score was calculated by subtracting the
one positive mood state from the sum of the 5 negative states.
Body composition
On the rst day of the baseline diet, body composition was
assessed by using dual-energy X-ray absorptiometry (GE Lunar
Prodigy Densitometer, version 5.6)
Metabolite assays
The FA composition of serum phospholipids was analyzed by
using recently described methods (17, 21). Nonesteried FAs in
serum were assessed by using capillary gas chromatographymass
spectrometry (31). Serum concentrations of total cholesterol, HDL
cholesterol, and triacylglycerols (triglycerides) were measured at
the Clinical Chemistry Laboratory at Fletcher Allen Health Care,
which is an afliate of the University of Vermont, by using a col-
orimetric method (Vitros 5.1 FS Chemistry System; Ortho-Clinical
Diagnostics). LDL cholesterol was calculated. The following
measurements were also determined at the Clinical Chemistry
Laboratory at Fletcher Allen Health Care (methods): serum glucose
(by using colorimetric reectance spectrophotometry); insulin (by
using a chemiluminescent immunoassay), hemoglobin (by using an
automated cell counter), and serum concentrations of alanine
aminotransferase (by using rate reectance spectrophotometry).
Statistics
All data are expressed as means 6 SEMs. Analyses were
performed with SAS software (version 9.2; SAS Institute Inc).
692 KIEN ET AL

b
y

g
u
e
s
t

o
n

S
e
p
t
e
m
b
e
r

1
3
,

2
0
1
4
a
j
c
n
.
n
u
t
r
i
t
i
o
n
.
o
r
g
D
o
w
n
l
o
a
d
e
d

f
r
o
m

Each study (cohorts 1 and 2) used a crossover design; subjects
were randomly assigned into 2 groups that differed with respect
to treatment order, and the trial lasted for 2 treatment periods.
Diet effects were analyzed by using a repeated-measures ANOVA,
including sequence and treatment effects, with the baseline
value as a covariate. In addition, Wilcoxons signed-rank test
was used to examine potential diet-associated increases (de-
creases) in variables of interest, without regard to the magnitude
of the increase (decrease).
RESULTS
Body weight
For both cohorts, there were no diet effects on body-weight
changes during either experimental diet.
Serum phosphatidylcholine
To determine whether diets provided to subjects had their
presumed effects on body lipids, we examined the diet effect on
the PA:OAratio for serumphosphatidylcholine in the fasted state.
For both cohorts, every subject exhibited a higher PA:OA ratio
during the HPA than HOA (P #0.001 for the diet effect by using
Wilcoxons signed-rank test for cohorts 1 and 2). In addition, as
reported previously (17), in cohort 1, the mean value for the ratio
was higher during the HPA than HOA (3.13 6 0.10 compared
with 1.49 6 0.08, respectively; P , 0.01); similar results were
shown for cohort 2 (3.62 6 0.17 compared with 1.94 6 0.14,
respectively; P , 0.0001).
Physical activity
As an index of physical activity behavior, we monitored
physical movement by using accelerometry. Physical activity was
signicantly higher during the HOA in 15 or 17 subjects in cohort
1 (P = 0.008) (Figure 1A) and in all subjects in cohort 2 (P =
0.005) (Figure 1B). For cohorts 1 and 2, mean physical activity
was 12% (P = 0.01) and 15% (P = 0.003) higher with the HOA
than HPA, respectively; differences between means (HOA
HPA) for the 2 cohorts were almost identical at 40 and 43
counts $ min
21
$ d
21
, respectively (Figure 2, C and D). As
noted in Subjects and Methods, for cohort 1, we explored
whether diet affected the amount of time per day spent at dif-
ferent levels of physical activity, but no such effect was detected
(26).
REE
For cohort 1, the HOA was associated with a 3% higher av-
erage REE measured in the fed state (P ,0.01) (Figure 1E). For
cohort 2, there was no diet difference in REE in the fed state
(1.3% higher during the HOA), but in the fasted state, REE was
4.5% greater during the HOA (P = 0.04) (Figure 1 F).
For the assessment of effects of diets on REE within cohorts,
we did not correct for fat-free mass during the diet to avoid bias as
a result of possible effects of the diets on body composition (17).
However, because, as shown in Figure 1, especially values for
REE in the fed state were higher in cohort 2 (daytime mea-
surements) than cohort 1 (supine, evening measurements), we
compared the 2 cohorts by expressing REE per fat-free mass. On
this basis, there was no difference between cohorts in REE in the
fasted state or diet change (HOA HPA) in REE during either
the fed or fasted state. However, for the fed state, values for
cohort 2 were signicantly higher than those for cohort 1 during
either the HOA or HPA (P = 0.012 and P ,0.001, respectively).
The history and physical examination and laboratory measure-
ments used at screening and during the dietary trial did not
detect subjects who manifested malabsorption, but we did not
collect feces during the 2 diets to quantify the excretion of fat.
However, our extensive lipidomic analysis performed in cohort 1
subjects suggested that the diets were well absorbed because the
PA and OA content of blood and muscle lipids increased to an
extent commensurate with the 3-wk diet and single meals (17).
POMS
After the results on physical activity from cohort 1 became
available, we studied cohort 2 to explore whether the composition
of dietary FA might affect mood. Ten of 12 men and women
(cohort 2) exhibited a lower anger-hostility score with the HOA
(P = 0.005) (Figure 2A). Eight of 12 men and women exhibited
a lower TMD score with the HOA (P = 0.06) (Figure 2B). The
mean score on the anger-hostility subscale was signicantly
lower (P = 0.007) during the HOA than HPA, but there also was
a trend for a lower mean TMD score during the HOA (P =
0.096) (Table 3). This study had .80% power to detect a dou-
bling of scores for individual scales of the POMS as signicant,
with a type I error rate of 5%. We showed no signicant cor-
relations between the change in any mood score, including the
anger-hostility or TMD score (HOA HPA) and change in
physical activity. In addition, there was no correlation between
the change in physical activity and any mood score assessed on
the baseline diet.
DISCUSSION
The alteration of the dietary SFA:MUFA ratio in the same
individuals affected physical activity and mood and to a mild
extent REE also. Our study design could not determine whether
these outcomes were primarily determined by increased dietary
OA per se or decreased PA. Despite obvious environmental
triggers, such as advertising, that promote excessive eating or
less physical activity (1, 3234) as well as potentially innate
differences in REE (4), cognition also plays a role in behaviors
relevant to physical activity and food selection and consumption
(35). We showed that 27 of 29 subjects in the 2 cohorts increased
their physical activity during the low-SFA diet, which resulted in
mean 1215% increases in physical activity compared with the
habitual, high-SFA diet. These results seemingly conict with
our previous observation that the variation of the dietary PA:OA
ratio did not affect physical activity (10). However, in the earlier
study, physical activity was monitored for only 1 wk by using
the wrist position for the accelerometer, and the dietary FA
composition was manipulated by using liquid diets without
a crossover design. Although the observed increments in phys-
ical activity could have had a substantial effect on energy bal-
ance if not compensated by changes in food intake, we observed
no signicant changes in body weight (as was our design).
Moreover, in our recent article (17), we did not report a corre-
lation between diet change in physical activity and change in
DIETARY FATTY ACIDS AND PHYSICAL ACTIVITY 693

b
y

g
u
e
s
t

o
n

S
e
p
t
e
m
b
e
r

1
3
,

2
0
1
4
a
j
c
n
.
n
u
t
r
i
t
i
o
n
.
o
r
g
D
o
w
n
l
o
a
d
e
d

f
r
o
m

insulin sensitivity. However, our study pointed to the possibility
that dietary SFA compared with MUFA may affect brain func-
tion (cognition) because both physical activity behavior and
mood were differentially affected by the diets.
Apersuasive argument has been made that an excessive dietary
energy intake, not physical activity, is a major contributor to
obesity (1). However, Levine et al (5) reported that obese
compared with nonobese people differed in the time per day that
they stood compared with sat, and these posture differences may
affect mortality risk (36). Although there is no clear evidence that
the intake of either total FA or SFA has been temporally asso-
ciated with the epidemic of obesity, the latter is a risk factor for
type 2 diabetes (37). Although more than one-half of Americans
ingest more SFA than has been recommended for optimal health
(7), more physically t individuals tend to eat lower amounts of
SFA (38). In the current study, we showed that, in 2 separate
cohorts, the lowering of the PA:OA ratio of the diet caused
modest increases in REE, but the effect was detected only under
conditions when subjects rested quietly in bed (fed state in cohort
1 and fasted state in cohort 2). We also acknowledge that, in our
previous, parallel-group trial, there was no trend toward a diet-
induced difference (10). Both these previous data (10) and the
differing effects of the diets on REE in the 2 cohorts of the current
study emphasized that interindividual variation and differences in
FIGURE 1. Physical activity and REE. A and B: Effects of diet condition on individual changes in physical activity during the 2 experimental diets. Diet
effects were analyzed by using Wilcoxons signed-rank test in cohorts 1 (n = 17; P = 0. 008) (A) and 2 (n = 12; P = 0.005) (B). Lean subjects are designated by
circles, and obese subjects are designated by triangles. Effects of diet condition on average (mean 6 SEM) physical activity were calculated in cohorts 1
(HPA: 327 6 26 counts $ min
21
$ d
21
; HOA: 367 6 33 counts $ min
21
$ d
21
) (C) and 2 (HPA: 280 6 36 counts $ min
21
$ d
21
; HOA: 323 6 38 counts $
min
21
$ d
21
) (D). Diet effects were analyzed by using a repeated-measures ANOVA, including sequence and treatment effects, with the baseline value as
a covariate. Effects of diets on mean (6SEM) REE (kcal/min) in the fasted and fed state were calculated in cohorts 1 (n = 18; HPA: fasted, 1.05 60.03 kcal/min;
fed, 1.22 6 0.04 kcal/min; HOA: fasted, 1.06 6 0.04 kcal/min; fed, 1.25 6 0.04 kcal/min) (E) and 2 (n = 11; HPA: fasted, 1.04 6 0.06 kcal/min; fed,
1.41 60.08 kcal/min; HOA: fasted, 1.09 60.06 kcal/min; fed, 1.43 60.10 kcal/min) (F). *
,
**Diet effect, *P #0.05, **P #0.01. HOA, higholeic acid diet;
HPA, highpalmitic acid diet; REE, resting energy expenditure.
694 KIEN ET AL

b
y

g
u
e
s
t

o
n

S
e
p
t
e
m
b
e
r

1
3
,

2
0
1
4
a
j
c
n
.
n
u
t
r
i
t
i
o
n
.
o
r
g
D
o
w
n
l
o
a
d
e
d

f
r
o
m

experimental design may alter subtle dietary effects on REE.
Because individuals spend w17 h/d eating or within 7 h of
eating (the thermic effect of food), the increment in REE with
the HOA for cohort 1 accounted for w20 kcal/d. However, the
biological relevance of the REE ndings likely relates to a proof
of concept that these diets might have subtle differential effects
on mitochondrial function because REE was perturbed during
strict resting conditions. We reported that the HPA was associ-
ated with a higher serum concentration of ceramide (17), and the
inhibition of muscle synthesis of ceramide has been associated
with decreased REE (39). There is also evidence that OA may
increase mitochondrial uncoupling, especially after exercise
training (40).
It is generally assumed that higher, executive-type brain
function mediates the desire to exercise per se (35). In this regard,
Sartorius et al (41) recently reported that feeding mice a high-PA
diet prevented the stimulation of locomotor activity otherwise
caused by the intracerebroventricular application of insulin seen
in mice fed a low-PA and high-OA diet. Sartorius et al (41) also
studied the effect of diets supplemented with either canola oil
(low PA and high OA) or dairy fat (high PA) in humans and
showed that the high-PA diet caused a decrease in hippocampal
and cortical activity, which was assessed by using functional
magnetic resonance imaging techniques. Studies that have used
functional magnetic resonance imaging also have shown that
patients with metabolic syndrome have an abnormal cerebro-
vascular response to a cognitive challenge (42), which implies
a link between nutritional status and brain function. Although the
progress toward a mechanistic understanding of these ndings
will require additional investigation, it is tempting to speculate
that the FA composition of the diet might affect neural circuitries
that inuence physical activity. Relevant to this hypothesis,
changes in the dietary FA composition can affect the lipid
composition of the brain, which is accompanied by alterations in
signaling and inammation in the hypothalamus (43). Subjects
were not told the identity of the 2 study oils, and the investigators
were very careful not to know the order of the diets in each
subject. Subjects displayed no curiosity about which diet they
consumed; anecdotally, when asked, subjects did not seem to
know which diet they were ingesting. However, it is possible that
physical differences between oils (eg, the melting point) might
have altered behavior in some way.
Our studies with the use of the POMS instrument were
intended to discern if consumption of the HPA was associated
with depression or lethargy, which we thought, a priori, might
induce sedentary behavior. However, instead, we showed that the
anger-hostility score was higher during the HPA, and there was
also a trend for an increase in the total mood disturbance during
this same diet. At present, we do not know if diet-related effects
on mood, in turn, induced changes in physical activity or whether
changes in physical activity affected mood, but we showed no
correlations between changes in mood scores on the POMS and
changes in physical activity. On the basis of the literature, it was
plausible that the reduced anger-hostility score with the HOAwas
actually a consequence of increased physical activity on that diet;
a walking program in female college students was shown to
reduce the anger-hostility score measured by using the POMS
(44). However, in vitro studies have linked PA (but not OA) to
increased secretion of TNF-a (45, 46), and in turn, TNF-a has
been linked to feelings of anger and hostility (47, 48). Thus, the
cellular PA:OA ratio also may be directly related to feelings of
anger because of the secretion of TNF-a by peripheral blood
monocytes or, perhaps, synthesis within the brain, which would
imply a central inammatory process (43, 49).
In conclusion, data from 2 rigorously controlled, crossover
studies suggest that lowering the saturated fat content of a typical
Western diet by replacing PA with OA increased daily physical
activity and REE and also affected mood, particularly anger and
hostility. Because posture can contribute to differences in energy
expenditure in lean compared with obese adults (5), these results
FIGURE 2. Changes in anger-hostility and TMD scores during the HPA
and HOA. Diet effects were analyzed by using Wilcoxons signed-rank test.
A: Ten of 12 men and women exhibited a lower anger-hostility score with
the HOA (P = 0.005). Two pairs of subjects showed identical, anger-hostility
scores with both the HPA and HOA (5 and 2 for 2 subjects and 4 and 2 for 2
subjects, respectively), which resulted in 2 pairs of overlapping points. Thus,
only 10 distinct data points are visible, although all data points were in-
cluded in the statistical analysis. B: Eight of 12 men and women exhibited
a lower TMD score with the HOA (P = 0.06). Two subjects (one lean subject
and one obese subject) had a score of 10 with the HPA; we labeled this data
point for the obese subject. Lean subjects are designated by circles, and
obese subjects are designated by triangles. HOA, higholeic acid diet;
HPA, highpalmitic acid diet; TMD, total mood disturbance.
TABLE 3
Results from the POMS questionnaire
1
HPA HOA P
Tension-anxiety 3.1 6 0.7 3.3 6 0.7 0.400
Depression-dejection 4.6 6 1.0 2.7 6 0.5 0.132
Anger-hostility 4.7 6 1.0 2.2 6 0.5 0.007
Fatigue-inertia 5.3 6 1.0 3.7 6 0.9 0.330
Confusion-bewilderment 3.6 6 0.4 2.9 6 0.5 0.441
Vigor-activity 8.2 6 1.1 7.8 6 1.1 0.801
Total mood disturbance 13.1 6 3.2 7.0 6 2.9 0.096
1
All values are means 6 SEMs of POMS raw scores (n = 6 men and
n = 6 women). The total mood disturbance score was calculated by subtract-
ing the one positive mood state from the sum of the 5 negative states. Diet
effects were analyzed by using a repeated-measures ANOVA, including
sequence and treatment effects, with the baseline value as a covariate.
HOA, higholeic acid diet; HPA, highpalmitic acid diet; POMS, Prole
of Mood States.
DIETARY FATTY ACIDS AND PHYSICAL ACTIVITY 695

b
y

g
u
e
s
t

o
n

S
e
p
t
e
m
b
e
r

1
3
,

2
0
1
4
a
j
c
n
.
n
u
t
r
i
t
i
o
n
.
o
r
g
D
o
w
n
l
o
a
d
e
d

f
r
o
m

might have relevance to an individuals propensity for weight
gain. Our ndings suggest that a high consumption of PA might
dampen motivation for physical activity. In the face of envi-
ronmental, sociologic, and cultural pressures that discourage
obligatory physical activity, even small changes in posture and
lifestyle activity could inuence energy balance and body hab-
itus. We surmise that studies designed to examine the role of
specic dietary FAs in the regulation of physical activity pa-
tterns are worthy of additional consideration. Furthermore,
we speculate that our studies of physical activity and mood raise
the possibility that the type of fat we eat may alter cognitive
function.
We thank our subjects for their patience and hard work in enduring our
rigorous protocols. We also are very grateful to the staff of the University
of Vermont General Clinical Research Center for dietary, nursing, and infor-
matics support.
The authors responsibilities were as followsCLK and DMM: designed
the research; CLK, KIC, CLT, DBE, TRK, and DMM: conducted the re-
search; JYB: analyzed data; CLK, JYB, CLT, JAD, and DMM: wrote the
manuscript; and CLK: had primary responsibility for the nal content of the
manuscript. None of the authors had a conict of interest.
REFERENCES
1. Swinburn BA, Sacks G, Lo SK, Westerterp KR, Rush EC, Rosenbaum
M, Luke A, Schoeller DA, DeLany JP, Butte NF, et al. Estimating the
changes in energy ux that characterize the rise in obesity prevalence.
Am J Clin Nutr 2009;89:17238.
2. Ng SW, Popkin BM. Time use and physical activity: a shift away from
movement across the globe. Obes Rev 2012;13:65980.
3. Berglund L, Lefevre M, Ginsberg HN, Kris-Etherton PM, Elmer PJ,
Stewart PW, Ershow A, Pearson TA, Dennis BH, Roheim PS, et al.
Comparison of monounsaturated fat with carbohydrates as a re-
placement for saturated fat in subjects with a high metabolic risk
prole: studies in the fasting and postprandial states. Am J Clin Nutr
2007;86:161120.
4. Ravussin E, Lillioja S, Knowler WC, Christin L, Freymond D, Abbott
WGH, Boyce V, Howard BV, Bogardus C. Reduced rate of energy
expenditure as a risk factor for body-weight gain. N Engl J Med 1988;
318:46772.
5. Levine JA, Lanningham-Foster LM, McCrady SK, Krizan AC, Olson
LR, Kane PH, Jensen MD, Clark MM. Interindividual variation in
posture allocation: possible role in human obesity. Science 2005;307:
5846.
6. Vetter ML, Faulconbridge LF, Webb VL, Wadden TA. Behavioral and
pharmacologic therapies for obesity. Nat Rev Endocrinol 2010;6:578
88.
7. Roger VL, Go AS, Lloyd-Jones DM, Adams RJ, Berry JD, Brown TM,
Carnethon MR, Dai S, de Simone G, Ford ES, et al. Heart disease and
stroke statistics2011 update: a report from the American Heart As-
sociation. Circulation 2011;123:e18209.
8. Astrup A, Dyerberg J, Elwood P, Hermansen K, Hu FB, Jakobsen MU,
Kok FJ, Krauss RM, Lecerf JM, Legrand P, et al. The role of reducing
intakes of saturated fat in the prevention of cardiovascular disease:
where does the evidence stand in 2010? Am J Clin Nutr 2011;93:
6848.
9. Kien CL. Dietary interventions for metabolic syndrome: role of mod-
ifying dietary fats. Curr Diab Rep 2009;9:4350.
10. Kien CL, Bunn JY, Ugrasbul F. Increasing dietary palmitic acid
decreases fat oxidation and daily energy expenditure (PMCID:
PMC1314972). Am J Clin Nutr 2005;82:3206.
11. Brsheim E, Kien CL, Pearl WM. Differential effects of dietary intake
of palmitic acid and oleic acid on oxygen consumption during and after
exercise. Metabolism 2006;55:121521.
12. Fontani G, Corradeschi F, Felici A, Alfatti F, Bugarini R, Fiaschi AI,
Cerretani D, Montorfano G, Rizzo AM, Berra B. Blood proles, body
fat and mood state in healthy subjects on different diets supplemented
with Omega-3 polyunsaturated fatty acids. Eur J Clin Invest 2005;35:
499507.
13. Reed J, Ones DS. The effect of acute aerobic exercise on positive
activated affect: a meta-analysis. Psychol Sport Exerc 2006;7:477514.
14. Executive Summary of The Third Report of The National Cholesterol
Education Program (NCEP) Expert Panel on Detection. Evaluation,
And Treatment of High Blood Cholesterol In Adults (Adult Treatment
Panel III). JAMA 2001;285:248697.
15. American Diabetes Association. Diagnosis and classication of di-
abetes mellitus. Diabetes Care 2012;35(suppl 1):S6471.
16. Stern SE, Williams K, Ferrannini E, Defronzo RA, Bogardus C, Stern
MP. Identication of individuals with insulin resistance using routine
clinical measurements. Diabetes 2005;54:3339.
17. Kien CL, Bunn JY, Poynter ME, Stevens R, Bain J, Ikayeva O,
Fukagawa NK, Champagne CM, Crain KI, Koves TR, et al. A lip-
idomics analysis of the relationship between dietary fatty acid com-
position and insulin sensitivity in young adults. Diabetes (Epub ahead
of print 13 December 2012).
18. Gallagher D, Heymseld SB, Heo M, Jebb SA, Murgatroyd PR, Sa-
kamoto Y. Healthy percentage body fat ranges: an approach for de-
veloping guidelines based on body mass index. Am J Clin Nutr 2000;
72:694701.
19. Louie JC, Flood V, Turner N, Everingham C, Gwynn J. Methodology
for adding glycemic index values to 24-hour recalls. Nutrition 2011;27:
5964.
20. Foster-Powell K, Holt SH, Brand-Miller JC. International table of
glycemic index and glycemic load values: 2002. Am J Clin Nutr 2002;
76:556.
21. Kien CL, Everingham KI, Stevens D, Fukagawa NK, Muoio DM.
Short-term effects of dietary fatty acids on muscle lipid composition
and serum acylcarnitine prole in human subjects. Obesity (Silver
Spring) 2011;19:30511.
22. Ekelund U, Sjostrom M, Yngve A, Poortvliet E, Nilsson A, Froberg K,
Wedderkopp N, Westerterp K. Physical activity assessed by activity
monitor and doubly labeled water in children. Med Sci Sports Exerc
2001;33:27581.
23. Janz KF, Witt J, Mahoney LT. The stability of childrens physical ac-
tivity as measured by accelerometry and self-report. Med Sci Sports
Exerc 1995;27:132632.
24. Silva P, Mota J, Esliger D, Welk G. Technical reliability assessment of
the Actigraph GT1M Accelerometer. Meas Phys Educ Exerc Sci 2010;
14:7991.
25. Puyau MR, Adolph AL, Vohra FA, Zakeri I, Butte NF. Prediction of
activity energy expenditure using accelerometers in children. Med Sci
Sports Exerc 2004;36:162531.
26. Treuth MS, Schmitz K, Catellier DJ, McMurray RG, Murray DM,
Almeida MJ, Going S, Norman JE, Pate R. Dening accelerometer
thresholds for activity intensities in adolescent girls. Med Sci Sports
Exerc 2004;36:125966.
27. Garrow JS. Energy balance and obesity in man. Amsterdam, Nether-
lands: North-Holland Publishing Co, 1974.
28. Frayn KN. Calculation of substrate oxidation rates in vivo from gas-
eous exchange. J Appl Physiol 1983;55:62834.
29. Weir JB. New methods for calculating metabolic rate with special
reference to protein metabolism. J Physiol 1949;109:19.
30. McNair DM, Lorr M, Droppleman LF. Prole of Mood State manual
(revised). San Diego, CA: Educational and Industrial Testing Service,
1992.
31. Newgard CB, An J, Bain JR, Muehlbauer MJ, Stevens RD, Lien LF,
Haqq AM, Shah SH, Arlotto M, Slentz CA, et al. A branched-chain
amino acid-related metabolic signature that differentiates obese and
lean humans and contributes to insulin resistance. Cell Metab 2009;9:
31126.
32. Ello-Martin JA, Ledikwe JH, Rolls BJ. The inuence of food portion
size and energy density on energy intake: implications for weight
management. Am J Clin Nutr 2005;82:236S41S.
33. Drewnowski A, Specter SE. Poverty and obesity: the role of energy
density and energy costs. Am J Clin Nutr 2004;79:616.
34. Hruschka DJ. Do economic constraints on food choice make people
fat? A critical review of two hypotheses for the poverty-obesity para-
dox. Am J Hum Biol 2012;24:27785.
35. McAuley E, Mullen SP, Szabo AN, White SM, Wojcicki TR, Mailey
EL, Gothe NP, Olson EA, Voss M, Erickson K, et al. Self-regulatory
processes and exercise adherence in older adults: executive function
and self-efcacy effects. Am J Prev Med 2011;41:28490.
696 KIEN ET AL

b
y

g
u
e
s
t

o
n

S
e
p
t
e
m
b
e
r

1
3
,

2
0
1
4
a
j
c
n
.
n
u
t
r
i
t
i
o
n
.
o
r
g
D
o
w
n
l
o
a
d
e
d

f
r
o
m

36. Katzmarzyk PT, Lee IM. Sedentary behaviour and life expectancy in
the USA: a cause-deleted life table analysis. BMJ Open 2012;2:
e000828.
37. Salmeron J, Hu FB, Manson JE, Stampfer MJ, Colditz GA, Rimm EB,
Willett WC. Dietary fat intake and risk of type 2 diabetes in women.
Am J Clin Nutr 2001;73:101926.
38. Brodney S, McPherson RS, Carpenter RS, Welten D, Blair SN. Nu-
trient intake of physically t and unt men and women. Med Sci Sports
Exerc 2001;33:45967.
39. Ussher JR, Koves TR, Cadete VJ, Zhang L, Jaswal JS, Swyrd SJ,
Lopaschuk DG, Proctor SD, Keung W, Muoio DM, et al. Inhibition of
de novo ceramide synthesis reverses diet-induced insulin resistance and
enhances whole-body oxygen consumption. Diabetes 2010;59:2453
64.
40. Tonkonogi M, Krook A, Walsh B, Sahlin K. Endurance training in-
creases stimulation of uncoupling of skeletal muscle mitochondria in
humans by non-esteried fatty acids: an uncoupling-protein-mediated
effect? Biochem J 2000;351:80510.
41. Sartorius T, Ketterer C, Kullmann S, Balzer M, Rotermund C, Binder S,
Hallschmid M, Machann J, Schick F, Somoza V, et al. Monounsaturated
fatty acids prevent the aversive effects of obesity on locomotion,
brain activity, and sleep behavior. Diabetes 2012;61:166979.
42. Hoth KF, Gonzales MM, Tarumi T, Miles SC, Tanaka H, Haley AP.
Functional MR imaging evidence of altered functional activation in
metabolic syndrome. AJNR Am J Neuroradiol 2011;32:5417.
43. Cintra DE, Ropelle ER, Moraes JC, Pauli JR, Morari J, Souza CT,
Grimaldi R, Stahl M, Carvalheira JB, Saad MJ, et al. Unsaturated fatty
acids revert diet-induced hypothalamic inammation in obesity. PLoS
ONE 2012;7:e30571.
44. Sakuragi S, Sugiyama Y. Effects of daily walking on subjective
symptoms, mood and autonomic nervous function. J Physiol Anthropol
2006;25:2819.
45. Wen H, Gris D, Lei Y, Jha S, Zhang L, Huang MT, Brickey WJ, Ting
JP. Fatty acid-induced NLRP3-ASC inammasome activation in-
terferes with insulin signaling. Nat Immunol 2011;12:40815.
46. Coll T, Eyre E, Rodriguez-Calvo R, Palomer X, Sanchez RM, Merlos
M, Laguna JC, Vazquez-Carrera M. Oleate reverses palmitate-induced
insulin resistance and inammation in skeletal muscle cells. J Biol
Chem 2008;283:1110716.
47. Suarez EC, Lewis JG, Kuhn C. The relation of aggression, hostility,
and anger to lipopolysaccharide-stimulated tumor necrosis factor
(TNF)-alpha by blood monocytes from normal men. Brain Behav
Immun 2002;16:67584.
48. Lotrich FE, Ferrell RE, Rabinovitz M, Pollock BG. Labile anger during
interferon alfa treatment is associated with a polymorphism in tumor
necrosis factor alpha. Clin Neuropharmacol 2010;33:1917.
49. Soliman ML, Puig KL, Combs CK, Rosenberger TA. Acetate reduces
microglia inammatory signaling in vitro. J Neurochem 2012;123:
55567.
DIETARY FATTY ACIDS AND PHYSICAL ACTIVITY 697

b
y

g
u
e
s
t

o
n

S
e
p
t
e
m
b
e
r

1
3
,

2
0
1
4
a
j
c
n
.
n
u
t
r
i
t
i
o
n
.
o
r
g
D
o
w
n
l
o
a
d
e
d

f
r
o
m

Letters to the Editor
Vitamin supplements and mortality in older people
Dear Sir:
Macpherson et al (1) carried out a meta-analysis of multivitamin
and multimineral (MVMM) tablet trials and found no effect of
MVMMs on average mortality. However, their study may suffer
from ecological fallacy. Ecological fallacy means that study-level
(group-level) analysis can lead to different conclusions than do
corresponding individual-level analyses (2). For this reason, exam-
ination of individual-level data is recommended, whenever feasible,
to avoid the potential for the ecological fallacy introduced by study-
level analyses (2).
Macpherson et al (1) calculated that the average age of the par-
ticipants in the studies was 62 y. However, ages ranged from 17 to
86 y in the included trials (1). It is probable that the effects of all
vitamins and minerals are not identical at the lower and upper ends
of such a wide age range. Therefore, pooling diverse trials with
young and old people to a single average MVMM effect may cam-
ouage effects of some individual vitamins or minerals, for exam-
ple, on the oldest people. In the case of vitamin E there is strong
empirical evidence of effect modication by age.
In an individual-level analysis of the Alpha-Tocopherol, Beta-
Carotene Cancer Prevention (ATBC) Study data, we found that among
participants aged 5062 y at baseline with a dietary vitamin C intake
above the median, vitamin E increased mortality by 19% (95% CI:
5%, 35%; based on 1021 deaths). However, among participants aged
6669 y at baseline with a dietary vitamin C intake above the median,
vitamin E decreased mortality by 41% (95% CI: 21%, 56%; based on
195 deaths) (3).
Furthermore, because the follow-up time in the ATBCStudy was up
to 8 y, the participants became substantially older during the trial so
that the baseline age was not a proper way to characterize them over
the entire follow-up period. Therefore, the modication of vitamin E
effects was also analyzed by using the follow-up age as the time vari-
able (4). Among 10,837 ATBC Study participants who contributed
follow-up time past the age of 65 y, the survival curves of the vitamin
E and novitamin E participants signicantly diverged at 71 y. Vita-
min E extended life span by ;0.5 y at the upper limit of the follow-
up age span (4).
Macpherson et al (1) write that in a meta-regression the estimate
of the effect of MVMMs was not associated with the duration of
supplementation. In the ATBC Study, the harm from vitamin E in
the young participants was restricted to the supplementation period
after 3.3 y, indicating that there can be a lag period of several years
before the effects of some vitamins appear (3). Macpherson et al
used the study-level average durations, which provide a poor basis
for analyzing supplementation timedependent effect modica-
tions. Proper analysis of time-dependent effects requires individual-
level data.
It is possible that some vitamins and minerals are benecial for
specic subpopulations. For example, age, sex, smoking, diet, and
exercise might modify the effects of some vitamins and minerals,
so that some restricted population groups might benet (and some
might be harmed). Such subgroups can be explored by analyzing
individual-level data, whereas pooling study-level averages provides
no information on relevant narrow subpopulations.
The meta-analysis by Macpherson et al (1) is important in dis-
couraging ordinary middle-aged people from taking MVMMs. Nev-
ertheless, their study should not be interpreted as evidence that none
of the vitamins and minerals included in the MVMM tablets have
effects on males and females in the age range of 1786 y. It is possible
that some vitamins, such as vitamin E, are useful for restricted groups
of older people. Individual-level data analyses are needed for explor-
ing such a possibility.
The author did not declare any conicts of interest.
Harri Hemila
Department of Public Health
University of Helsinki
Helsinki, FIN-00014
Finland
E-mail: harri.hemila@helsinki.
REFERENCES
1. Macpherson H, Pipingas A, Pase MP. Multivitamin-multimineral sup-
plementation and mortality: a meta-analysis of randomized controlled
trials. Am J Clin Nutr 2013;97:43744.
2. Berlin JA, Santanna J, Schmid CH, Szczech LA, Feldman HI; Anti-
Lymphocyte Antibody Induction Therapy Study Group. Individual patient-
versus group-level data meta-regressions for the investigation of treatment
effect modiers: ecological bias rears its ugly head. Stat Med 2002;21:371
87.
3. Hemila H, Kaprio J. Modication of the effect of vitamin E supplemen-
tation on the mortality of male smokers by age and dietary vitamin C.
Am J Epidemiol 2009;169:94653.
4. Hemila H, Kaprio J. Vitamin E may affect the life expectancy of men,
depending on dietary vitamin C intake and smoking. Age Ageing 2011;
40:21520.
doi: 10.3945/ajcn.113.064204.
Reply to H Hemila
Dear Sir:
We thank Hemila for his interest in our article entitled Multivitamin-
multimineral supplementation and mortality: a meta-analysis of ran-
domized controlled trials (1). Our primary nding was that, across
a pooled sample of 91,074 participants, multivitamin-multimineral
502 Am J Clin Nutr 2013;98:50212. Printed in USA. 2013 American Society for Nutrition
(MVMM) supplementation had no signicant effect on the risk of
all-cause mortality, mortality due to cancer, or mortality due to
cardiovascular disease.
Despite our overall nding, Hemila asserts that some vitamins and
minerals may be benecial for specic subpopulations. We concur
with his suggestion that variables such as age, sex, and lifestyle fac-
tors might modify the effects of some vitamins, such that differential
effects may emerge in different subpopulations. However, as pointed
out by Hemila, we were unable to perform subanalyses to examine
the modifying effect of these different variables given that only trial-
level data were available.
If individual-level data were accessible we could have performed
any number of subanalyses. A limitation of this approach is that each
subanalysis involves an additional statistical comparison and thus
a greater risk of a type I error. Furthermore, subgroup analysis
based on post hoc examination of data can lead to erroneous con-
clusions (2). The ndings discussed by Hemila, relating to vitamin
E mortality risk across different age groups, still require replication
for this reason. To avoid these issues, we used a limited number of
prespecied analyses to determine the overall effects of MVMM
supplementation in the general population, rather than in specic
subpopulations.
Our results were strengthened by the large number of trials included
in our analyses, generating a large pooled sample size. Although there
are several advantages to undertaking an individual-level data meta-
analysis, such an analysis is not always feasible. For example, we
excluded 7 relevant trials from our analysis simply because trial-level
data were unobtainable. Given the difculty in obtaining raw data
from chief investigators (especially when many of the trials included
in our analysis were more than a decade old), undertaking a patient-
level meta-analysis would have further diminished the number of
trials included in our analysis.
Hemila states that our meta-analysis is important in discouraging
ordinary middle-aged people fromtaking MVMMs. We are not sure
how this conclusion was derived from our work given that our meta-
analysis did not specically focus on middle-aged adults. Moreover,
whereas we found no effect of MVMMs on mortality across adults of all
ages, this does not rule out other possible benets to health or well-being.
Before our investigation, information on the association of
MVMM use and mortality had frequently been obtained from obser-
vational studies (3). Our meta-analysis showed that, across ran-
domized controlled trials, MVMM supplementation had no effect
on mortality (1). Although we acknowledge that vitamins may
have different effects in different subpopulations, it was rst nec-
essary to investigate the overall effects of MVMM supplementa-
tion in the general population. Identifying a harmful effect of
MVMM use across all adults would have shown greater implica-
tions than identifying a harmful effect in one of many narrow
subgroups. As discussed in our meta-analysis, we call for further
research into the effects of MVMM use on all aspects of human
health (1). This includes examination of MVMM use in specic
subpopulations.
MPP is funded by a Menzies Foundation Scholarship in Allied Health
Sciences. AP receives funding from Swisse Wellness Pty Ltd for ongoing
research projects. HM holds a Postdoctoral Fellowship, which is funded
by Swisse Wellness Pty Ltd. There were no other potential conicts of
interest.
Helen Macpherson
Andrew Pipingas
Matthew P Pase
Centre for Human Psychopharmacology
Swinburne University of Technology
Advanced Technology Building
427451 Burwood Road
Hawthorn, Victoria 3122
Australia
E-mail: matthewpase@gmail.com
REFERENCES
1. Macpherson H, Pipingas A, Pase MP. Multivitamin-multimineral sup-
plementation and mortality: a meta-analysis of randomized controlled
trials. Am J Clin Nutr 2013;97:43744.
2. Oxman AD, Guyatt GH. A consumers guide to subgroup analyses. Ann
Intern Med 1992;116:7884.
3. Chang SM. Should meta-analyses trump observational studies? Am J
Clin Nutr 2013;97:2378.
doi: 10.3945/ajcn.113.064709.
Limitations to the use of plasma osmolality as
a hydration biomarker
Dear Sir:
In some laboratories, plasma osmolality (P
osm
) is used as the
gold standard for detecting dehydration (1), without consider-
ation of its limitations; however, published data dispute this tech-
nique (2, 3), which prompts us to write in response to the recent
article by Cheuvront et al (4) with regard to quantitative dehy-
dration assessment. This article correctly states that P
osm
is the
key regulated variable in uid balance, which means that P
osm
is
constantly regulated toward a central set point as the kidneys
modify urine concentration and water excretion in response to
diet and daily activities. We believe that this controlled regulation
limits the efcacy of P
osm
as an index of hydration change in
many experimental designs. This article (4) also states that the
criticisms for adopting P
osm
as a gold standard for dehydration
assessment are minimal (p 460). We disagree and write to de-
scribe several limitations to the use of P
osm
as a gold standard for
dehydration.
First, individuals who lose a large amount of body water (reported as
% body mass loss relative to a beginning euhydrated state) may exhibit
a decreased P
osm
, contrary to anticipated hemoconcentration. For ex-
ample, a summary of 2 studies (5) reported that the P
osm
of 6 in-
dividuals (out of 39) decreased after they lost 38% of body mass.
In a different study, men and women who consumed a 500-mL bolus
of uid acutely exhibited an increased P
osm
, contrary to anticipated
hemodilution (1); that is, after 90 min of rest, 4 of 30 P
osm
mea-
surements increased. These values show that P
osm
may not reect
widely accepted physiologic principles, and that variance of P
osm
measurements may be large.
Second, evidence suggests that P
osm
changes are time- and
protocol-specic. Unpublished observations (CX Mun oz, EC
Johnson, JK DeMartini, et al, 2012) show that dehydration equiv-
alent to 2% of body mass resulted in P
osm
changes that were
twice as large during mild cycling exercise (2.3 h; DP
osm
of 9
mOsm/kg) compared with a passive exposure (5.0 h; DP
osm
of 4
mOsm/kg); participants consumed no water during either trial in
LETTERS TO THE EDITOR 503
a 36C environment. It is likely that this difference occurred
because exercise increased intracellular osmolality (6) and in-
creased extracellular uid tonicity, causing water to move into
muscle tissue.
Third, Kenney et al (3) reported that mean (6SE) P
osm
values in
7 resting, euhydrated young male subjects decreased from 281 6 3
at baseline to 276 6 2 mOsm/kg at 60 min after they had consumed
1.9 L of water. However, the mean P
osm
value returned to baseline
(282 6 2 mOsm/kg) at 90 min postingestion. These ndings chal-
lenge our understanding of the interactions between intracellular-
extracellular uid shifts (6) and renal compensatory mechanisms;
they also suggest that further research into the time course of acute
P
osm
changes is warranted.
Fourth, 2 recent publications (7, 8) showed that a single P
osm
or
serum osmolality measurement was a poor predictor of changes in
hydration status when a single, fasted morning blood sample is
collected. The former article (7) involved modied uid intake in
habitually low-volume drinkers and habitually high-volume
drinkers, with the outcome that P
osm
was constant across days in
men and women, whereas urinary biomarkers reected modied
water consumption. The latter publication (8) showed that serum
osmolality was a poor predictor (r
2
0.01) of 24-h water retention-
clearance by the kidneys. Furthermore, the NHANES (19881994)
reported that serum osmolality values were constant across a wide
range of uid intakes (9). Men exhibited similar mean P
osm
values
(range: 279281 mOsm/kg) regardless of total daily uid intake,
which ranged from 1.7 to 7.9 L; women exhibited similar P
osm
values (range: 276278 mOsm/kg) across a total daily uid intake
range of 1.36.1 L. These studies argue that P
osm
is not appropriate
in clinical settings, in which a single blood sample is collected
during an ofce visit.
Furthermore, Cheuvront et al (4) recommended that a P
osm
value of 301 6 5 mOsm/kg be used clinically as the threshold
of dehydration (p 460), as determined statistically. However, pre-
viously published data (10) show that a P
osm
value of 301 6 5
mOsm/kg represents a body mass loss of ;4.5% in healthy, young
males; this marked level of dehydration is hardly a threshold for
dehydration.
Finally, serum samples contain numerous substances (eg, so-
dium, chloride, potassium, bicarbonate, urea, glucose) that consti-
tute 95% of total osmolarity. Even though they are found in small
amounts (45%), proteins inuence total osmolality considerably.
Thus, the water content in a serum sample is less per unit volume
than in a calibration solution, and to obtain an accurate measurement
of osmolality, the empirical value should be mathematically cor-
rected. Furthermore, normal intraindividual differences in serum
protein concentration (range: 6.08.5 g/dL) and within-individual
changes in serum protein concentration induced by factors such
as physical training and heat acclimation (11) increase the statisti-
cal variance and difculty of interpreting the meaning of P
osm
as a
hydration index.
We recommend that scientists use P
osm
as a marker of dehydration
cautiously, with careful consideration of experimental protocol (ie,
dehydration compared with hypohydration, exercise compared with
rest) and tight control of dietary total osmolar load and uid volume
(2, 8, 10). We recommend that P
osm
not be used in clinical settings
as a gold standard for dehydration assessment (2, 7, 8). The limi-
tations (described above) reect the dynamic and complex regula-
tion of human uid-electrolyte balance (2), which does not lend
itself to generalizations.
All authors were involved in the writing of this letter, reviewed its content,
and approved the nal version. None of the authors claimed a conict of interest.
Lawrence E Armstrong
Human Performance Laboratory
University of Connecticut
Storrs, CT 06269-1110
E-mail: lawrence.armstrong@uconn.edu
Ronald J Maughan
School of Sport and Exercise Sciences
Loughborough University
Loughborough
United Kingdom
Leo C Senay
Department of Pharmacologic and Physiologic Sciences
St Louis University
St Louis, MO
Susan M Shirreffs
GlaxoSmithKline
Brentford
United Kingdom
REFERENCES
1. Sollanek KJ, Keneck RW, Cheuvront SN, Axtell RS. Potential impact
of a 500-mL water bolus and body mass on plasma osmolality dilution.
Eur J Appl Physiol 2011;111:19992004.
2. Armstrong LE. Assessing hydration Status: the elusive gold standard.
J Am Coll Nutr 2007;26:575S84S.
3. Kenney WL, Tankersley CG, Newswanger DL, Hyde DE, Puhl SM, Turner
NL. Age and hypohydration independently inuence the peripheral vascu-
lar response to heat stress. J Appl Physiol 1990;68:19028.
4. Cheuvront SN, Keneck RW, Charkoudian N, Sawka MN. Physiologic
basis for understanding quantitative dehydration assessment. Am J Clin
Nutr 2013;97:45562.
5. Sawka MN, Montain SJ, Latzka WA. Body uid balance during exer-
cise-heat exposure. In: Buskirk ER, Puhl SM, eds. Body uid balance in
exercise and sport. Boca Raton, FL: CRC Press, 1996:13958.
6. Sjgaard S, Saltin B. Extra- and intracellular water spaces in muscles of man
at rest and with dynamic exercise. Am J Physiol 1982;243:R27180.
7. Perrier E, Desmazieres A, Girard N, Pross N, Osbild D, Metzger D,
Guelinckx I, Klein A. Circadian variation and responsiveness of hydra-
tion biomarkers to changes in daily water intake. Eur J Appl Physiol
2013;Apr 23 (Epub ahead of print; DOI:10.1007/s00421-013-2649-0).
8. Armstrong LE, Johnson EC, McKenzie AL, Munoz CX. Interpreting
common hydration biomarkers on the basis of solute and water excre-
tion. Eur J Clin Nutr 2013;67:24953.
9. Institute of Medicine, Food and Nutrition Board. Dietary Reference In-
takes for water, potassium, sodium, chloride, and sulfate. Washington,
DC: National Academies Press, 2004:269423.
10. Armstrong LE, Maresh CM, Gabaree CV, Hoffman JR, Kavouras SA,
Keneck RW, Castellani JW, Ahlquist LE. Thermal and circulatory re-
sponses during exercise: effects of hypohydration, dehydration, and wa-
ter intake. J Appl Physiol 1997;82:202835.
11. Senay LC, Kok R. Effects of training and heat acclimatization on blood
plasma contents of exercising men. J Appl Physiol 1977;43:5919.
doi: 10.3945/ajcn.113.065466.
504 LETTERS TO THE EDITOR
Reply to LE Armstrong et al
Dear Sir:
We have great respect for the authors who have expressed in-
terest in our article, and we appreciate the opportunity to reply to
their letter; however, we nd little convincing evidence for their
concerns.
First and foremost we wish to emphasize 2 important points from
our article that were left out of the quote taken from page 460 (1). We
were very careful in our review to outline why plasma osmolality
(P
osm
) should be considered a gold standard for assessing dehydration,
dened as intracellular dehydration (or hypertonic-hypovolemia),
and not extracellular dehydration (or isotonic-hypovolemia or vol-
ume depletion). We also point out the criticality of considering the
dehydration magnitude. With these 2 very important points in mind,
the criticisms that we describe as minimal on page 460 relate
directly to articles that have neglected these important points in their
misguided assertions about the limitations of using P
osm
for assessing
dehydration.
The criticisms of our review on dehydration assessment seem to
involve 3 major points: 1) disparate research ndings, 2) a P
osm
threshold of 301 6 5 mmol/kg for dehydration, and 3) the contri-
bution of protein to P
osm
.
Disparate research ndings
Six published articles or reports were used when trying to re-
fute our review. Curiously, only 2 of those studies were designed
to produce dehydration and only one directly described the po-
tential for using P
osm
to quantify dehydration (2). Although the
remaining studies referenced do describe the normal, and ex-
tremely well-documented, physiologic response to both normal
and overconsumption of water (water intake water losses),
when carefully read they do not in any way refute the perspec-
tives presented in our article. As a matter of interpretation, we
would also suggest that the composite gure from Sawka et al
(2) shows that P
osm
responded to dehydration exactly as expected
in 33 of 39 volunteers (85%). In a recent study from our labo-
ratory (3) in which baseline values were very carefully controlled,
P
osm
increased in 36 of 36 volunteers (100%) who became de-
hydrated by 2.25.8% of body mass via sweating (exercise-heat
stress). Nowhere in our article do we generalize or make claims
that P
osm
is perfect. We do argue, however, that P
osm
is the best
currently available assessment measure (gold standard) for one
specic type of dehydration (intracellular).
P
osm
threshold of 301 6 5 mmol/kg
A full appreciation for the genesis of the 301 6 5-mmol/kg
threshold for dehydration requires knowledge of biological var-
iation and diagnostic decision making, which goes well beyond
the scope of this letter. We encourage interested readers to seek
Cheuvront et al (1, 4, 5) for details. Briey, the nosological
sensitivity of P
osm
is modest but superior to all other common
body uids used to assess dehydration. When the variance term
for P
osm
is properly considered, the range of P
osm
values that
indicate dehydration (2% body mass) agree extremely well
with many independently published observations and commonly
accepted clinical thresholds for dehydration (4). Change values
are better when it is practical to make 2 measures, but here again
P
osm
does extremely well (4, 5). The DP
osm
remains sensitive
even when water loading is used (urine osmolality:P
osm
,1.5).
This practice is often adopted in research where assurance of
euhydration is desired; however, it is important to recognize
that it also decreases the nosological sensitivity of the 301 6
5-mmol/kg threshold (4). Under said circumstances, a
1 5-mmol/kg change in P
osm
still affords 80% probability that
intracellular dehydration has occurred (4, 5), which is remark-
ably consistent with the well-taught osmotic change threshold
(;2% or 16 mmol/kg) for renal compensation and water ac-
quisition (thirst) (1).
Contribution of protein to P
osm
In all of our articles on P
osm
(1, 35), and more in press or
forthcoming, we recognize and discuss its complexity. A reduc-
tion in plasma water increases the concentration of all dissolved
substances. It is, of course, well known that plasma protein con-
centration increases linearly as plasma water is reduced (6). When
assessing the potential for dehydration, the question can only
be why it increases. The concentration of P
osm
reects the
loss of water from the plasma and it describes the loss of body
water very well (3). Both inter- and intraindividual variation in
plasma protein concentrations are already a part of inter- and
intraindividual P
osm
variation (4). Therefore, plasma protein
variation is already taken into consideration in the 301 6
5-mmol/kg threshold. Thus, unless there is good reason to
believe that circumstances have produced a grossly dispropor-
tionate increase in protein beyond that expected from plasma
water loss, there would be no need for corrections. Studies
from our laboratory and Senays pioneering research have
shown that plasma protein can be added by heat exposure as
well as lost with dehydration. We acknowledge that some ux
of total circulating proteins occurs, but as previously stated such
protein uxes are already part of the observed variance and di-
agnostic error. Any acute inuence of protein ux due to exercise
would also be remedied by allowing proper recovery (1). In other
words, the potential for plasma protein to confound the appropriate
use of P
osm
for assessing dehydration is marginal at best.
In our review article (1), we carefully described the true lim-
itations of using P
osm
for dehydration assessment on page 460.
The concerns expressed in the letter by Armstrong et al are
clearly but curiously misplaced. We must therefore regard the
limitations inferred by the title of their letter as false.
All of the authors were involved in the writing of this letter, reviewed its
content, and approved the nal version. The opinions or assertions contained
herein are the private views of the authors and should not be construed as
ofcial or reecting the views of the US Army or the US Department of De-
fense. None of the authors claimed a conict of interest.
Samuel N Cheuvront
Robert W Keneck
Nisha Charkoudian
US Army Research Institute of Environmental Medicine
Thermal and Mountain Medicine Division
Kansas Street, Building 42
Natick, MA 01760
E-mail: samuel.n.cheuvront.civ@mail.mil
Michael N Sawka
Georgia Institute of Technology
Atlanta, GA
LETTERS TO THE EDITOR 505
REFERENCES
1. Cheuvront SN, Keneck RW, Charkoudian N, Sawka MN. Physiologic
basis for understanding quantitative dehydration assessment. Am J Clin
Nutr 2013;97:45562.
2. Sawka MN, Montain SJ, Latzka WA. Body uid balance during exercise-
heat exposure. In: Buskirk ER, Puhl SM, eds. Body uid balance in
exercise and sport. Boca Raton, FL: CRC Press, 1996:13958.
3. Cheuvront SN, Keneck RW, Sollanek KJ, Ely BR, Sawka MN. Water-
decit equation: systematic analysis and improvement. Am J Clin Nutr
2013;97:7985.
4. Cheuvront SN, Ely BR, Keneck RW, Sawka MN. Biological variation
and diagnostic accuracy of dehydration assessment markers. Am J Clin
Nutr 2010;92:56573.
5. Cheuvront SN, Fraser CG, Keneck RW, Ely BR, Sawka MN. Reference
change values for dehydration monitoring. Clin Chem Lab Med 2011;
49:10337.
6. Eisenman AJ, Mackenzie LB, Peters JP. Protein and water of serum and
cells of human blood, with a note on the measurement of red cell volume.
J Biol Chem 1936;116:335.
doi: 10.3945/ajcn.113.065482.
No and low alcohol intake may have differential
effects on risk of overall and cause-specic
mortality
Dear Sir:
We read with great interest the article by Vergnaud et al (1) on the
relation between adherence to the World Cancer Research Fund
(WCRF)/American Institute for Cancer Research (AICR) guidelines
and risk of death in Europe. This well-crafted, large-scale study
conducted in participants in the European Prospective Investigation
into Cancer and Nutrition (EPIC) cohort offers valuable data re-
garding the impact of the WCRF/AICR recommendations on re-
ducing total and cause-specic mortality and suggests that the
utility of these guidelines may extend beyond the scope of cancer
prevention. We are, however, keen on gaining additional under-
standing of the results presented in their Table 4: namely, the risk
of death associated with alcohol consumption.
The authors found that adherence to the WCRF/AICR recommen-
dation for daily alcohol intake (2 drinks for men and 1 drink for
women) was protective against all-cause mortality in men but not
in women. This result was based on a scoring system that operation-
alized this alcohol-specic guideline into 3 categories of ethanol
intake: 20, .20 to 30, and .30 g/d for men and 10, .10 to
20, and .20 g/d for women. Among the 257,421 male study
participants, the men whose ethanol intake was .20 to 30 g/d
had a signicantly reduced risk of death compared with men whose
consumption exceeded 30 g/d (HR: 0.80), as did men who limited
their intake to 20 g/d compared with the same referent (HR: 0.89).
However, signicant associations between risk of death and the
alcoholic drinks component of the WCRF/AICR recommendations
were not observed among the 121,443 female study participants.
We are highly curious both to learn whether making the distinction
between no and lowethanol intake would alter the results of this anal-
ysis and to see the stratication of HRs by cause of death. Whereas it
is widely acknowledged that, unlike in cardiovascular disease, the
lowest alcohol-related cancer risk is in fact conferred in the absence
of alcohol consumption (2), there remains uncertainty regarding
whether the protective effect of abstinence on cancer risk translates
to survival outcomes. The most current estimate of alcohol-attributable
cancer mortality in the United States to our knowledge suggests that
alcohol consumption at any level not only increases cancer risk but,
more critically, is a major factor behind cancer-related death in men
and women (3). Interestingly, the number of alcohol-attributable
deaths was highest for female breast cancer in this investigation. A
meta-analysis by Bagnardi et al (4) that included 222 articles concern-
ing alcohol consumption and cancer found that light alcohol drinking
(1 drink/d) was associated with breast cancer death. In contrast and
illustrative of the ambiguity related to drinking and cancer mortality,
another recent study reported that any alcohol consumption either
before or after breast cancer diagnosis had no adverse impact on
survival from breast cancer, cardiovascular disease, or other cause,
and that moderate consumption may even have a survival benet (5).
The robust data set of Vergnaud et al presents an opportunity for
additional analyses that could shed further light on the advantages
or lack thereof of teetotaling in the prevention of cancer or other
chronic diseases. As such, we appreciate the authors consideration
of our request that they both reoperationalize the alcohol-specic
WCRF/AICR score such that 0 g/d of ethanol intake is assigned
its own category and evaluate alcohol-specic mortality by cause
of death and share these results.
Support for this letter was provided by the University of Alabama at Bir-
mingham Cancer Prevention and Control training grant R25 CA047888. The
authors had no conicts of interest to disclose.
Emily Falk Libby
Department of Nutrition Sciences
University of Alabama at Birmingham
VH G017, 1670 University Boulevard
Birmingham, AL 35294-0019
E-mail: elibby@uab.edu
Michelle S Williams
Department of Health Behavior
University of Alabama at Birmingham
Birmingham, AL
Will L Tarver
Department of Health Care Organization and Policy
University of Alabama at Birmingham
Birmingham, AL
Wendy Demark-Wahnefried
Department of Nutrition Sciences
University of Alabama at Birmingham
Birmingham, AL
REFERENCES
1. Vergnaud AC, Romaguera D, Peeters PH, van Gils CH, Chan DS, Romieu
I, Freisling H, Ferrari P, Clavel-Chapelon F, Fagherazzi G, et al. Adher-
ence to the World Cancer Research Fund/American Institute for Cancer
Research guidelines and risk of death in Europe: results from the Euro-
pean Prospective Investigation into Nutrition and Cancer cohort study. Am
J Clin Nutr 2013;97:110720.
2. Kushi LH, Doyle C, McCullough M, Rock CL, Demark-Wahnefried W,
Bandera EV, Gapstur S, Patel AV, Andrews K, Gansler T. American
Cancer Society Guidelines on nutrition and physical activity for cancer
506 LETTERS TO THE EDITOR
prevention: reducing the risk of cancer with healthy food choices and
physical activity. CA Cancer J Clin 2012;62:3067.
3. Nelson DE, Jarman DW, Rehm J, Greeneld TK, Rey G, Kerr WC,
Miller P, Shield KD, Ye Y, Naimi TS. Alcohol-attributable cancer deaths
and years of potential life lost in the United States. Am J Public Health
2013;103:6418.
4. Bagnardi V, Rota M, Botteri E, Tramacere I, Islam F, Fedirko V, Scotti
L, Jenab M, Turati F, Pasquali E, et al. Light alcohol drinking and
cancer: a meta-analysis. Ann Oncol 2013;24:3018.
5. Newcomb PA, Kampman E, Trentham-Dietz A, Egan KM, Titus LJ,
Baron JA, Hampton JM, Passarelli MN, Willett WC. Alcohol consump-
tion before and after breast cancer diagnosis: associations with survival
from breast cancer, cardiovascular disease, and other causes. J Clin
Oncol 2013;31:193946.
doi: 10.3945/ajcn.113.066217.
Reply to E Falk Libby et al
Dear Sir:
We thank Falk Libby et al for their interest in our article. We
acknowledge the need for more detailed analysis of the association
between individual components of the World Cancer Research
Fund/American Institute for Cancer Research (WCRF/AIRC)
score, including alcohol consumption and cause-specic mortal-
ity. The association between pattern of lifetime alcohol use and
cause of death in the European Prospective Investigation into Can-
cer and Nutrition (EPIC) study has been addressed in detail by
Manuela M Bergmann et al in a manuscript currently under sub-
mission. Results cannot be displayed before publication, so we en-
courage Falk Libby et al to pay attention to the release of this
article, which will provide a comprehensive answer to their
requests.
None of the authors had a conict of interest.
Anne-Claire Vergnaud
Dora Romaguera
Petra HM Peeters
Department of Epidemiology and Biostatistics
School of Public Health
Imperial College London
Medical Building, Room VC8, Norfolk Place
St Marys Campus
London, W2 1PG
United Kingdom
E-mail: a.vergnaud@imperial.ac.uk
Carla H van Gils
Julius Center for Health Sciences and Primary Care
University Medical Center Utrecht
Utrecht
The Netherlands
Doris SM Chan
Department of Epidemiology and Biostatistics
School of Public Health
Imperial College London
London
United Kingdom
Manuela M Bergmann
Division of Epidemiology
German Institute of Human Nutrition
Potsdam-Rehbrucke
Germany
Teresa Norat
Department of Epidemiology and Biostatistics
School of Public Health
Imperial College London
London
United Kingdom
On behalf of the EPIC coauthors
doi: 10.3945/ajcn.113.066324.
The challenge of complexity and arginine
metabolism
Dear Sir:
Chapeau! to Mariotti et al (1) for their attempt to put order to
complexity by giving a dimension to arginine uxes in its metabolism.
But, complexity is both a challenge and a burden. An important
question relates to the lack of computation of the possible effects that
arginine-derived and naturally produced inhibitors of enzymes dealing
with arginine metabolism, such as asymmetric-di-methyl-arginine
(ADMA), may have on peripheral tissue activity of arginases (2). Do the
authors have data on acute effects of arginine ingestion on ADMA?
Indeed, it has been reported that long-term ingestion of arginine sup-
plements increases ADMA (3) and inhibition of arginases was efcient
in maintaining nitric oxide (NO) production and in preventing damage
related to impaired NO production in peripheral tissues (4).
Also, the expression and activity of arginases, and thus their con-
tribution to plasma and urea by red blood cells, were not sufciently
stressed by Mariotti et al in their text or in the supplemental data.
Peculiarly, in capillaries red blood cells may dramatically control
and blunt arginine concentrations in plasma (5, 6) and this should
also be included in a model that focuses on clusters of peripheral
needs, even if the said model groups together sums of activities by
different compartments. Moreover, habitual dietary arginine intake by
controlling arginase expression may rule uxes of arginine toward
availability for protein syntheses or catabolism producing urea. Urea
production may become misleading in evaluating adequate nitrogen
intake if this is calculated on the basis of urinary urea excretion (7).
The author did not declare any conicts of interest.
LETTERS TO THE EDITOR 507
Francesco Saverio Dioguardi
Department of Clinical Sciences and Community Health
University of Milan
via Sannio 18
20137 Milan-I
Italy
E-mail: fsdioguardi@gmail.com
REFERENCES
1. Mariotti F, Petzke KJ, Bonnet D, Szezepanski I, Bos C, Huneau J-F,
Fouillet H. Kinetics of the utilization of dietary arginine for nitric oxide
and urea synthesis: insight into the argininenitric oxide metabolic sys-
tem in humans. Am J Clin Nutr 2013;97:9729.
2. Masuda H. Signicance of nitrogen oxide and its modulation mecha-
nisms by endogenous nitrogen oxide synthase inhibitors and arginase
in the micturition disorders and erectile dysfunction. Int J Urol 2008;
15:12834.
3. Jabecka A, Ast J, Bogdaski P, Drozdowski M, Pawlak-Lemaska K,
Cielewicz AR, Pupek-Musialik D. Oral L-arginine supplementation in
patients with mild arterial hypertension and its effect on plasma level of
asymmetric dimethylarginine, L-citrulline, L-arginine and antioxidant
status. Eur Rev Med Pharmacol Sci 2012;16:166574.
4. Shemyakin A, Kovamees O, Rafnsson A, Bohm F, Svenarud P, Settergren
M, Jung C, Pernow J. Arginase inhibition improves endothelial function in
patients with coronary artery disease and type 2 diabetes mellitus. Circu-
lation 2012;126:294350.
5. El-Hady SB, Farahat MH, Atfy M, Elhady MA. Nitric oxide metabolites
and arginase I levels in b-thalassemic patients: an Egyptian study. Ann
Hematol 2012;91:1193200.
6. Carvalho DR, Brand GD, Brum JM, Takata RI, Speck-Martins CE,
Pratesi R. Analysis of novel ARG1 mutations causing hyperargininemia
and correlation with arginase I activity in erythrocytes. Gene. 2012;
509(1):12430.
7. Dioguardi FS. To give or not to give? Lessons from arginine paradox.
J Nutrigenet Nutrigenomics. 2011;4:908.
doi: 10.3945/ajcn.113.065011.
Reply to FS Dioguardi
Dear Sir:
We appreciated the congratulations and comments received from
Dioguardi regarding our recently published article, which was the rst
attempt to delineate the metabolism of dietary arginine, including its
bioavailability and utilization for the competitive pathways that are
arginase and nitric oxide (NO) synthase (1). The objective of model
development was to determine the minimal structure for this nutri-
tional system that could solve the isotopic metabolic data at hand and
provide an insight into the key metabolic/compartmental structuring
that explains how the body deals structurally with arginine intake.
According to the design and process of this modeling study, the
effects of any potential changes in arginase or NO synthase activity
during the postprandial phase (the potential existence of which was
suggested by Dioguardi) are embedded in the isotopic (urea and ni-
trate) metabolic data and are therefore computed in the model pre-
dictions for the uxes of urea and NO production. In the model, both
urea and NOproduction indeed originate fromboth a plasma compart-
ment and another compartment that aggregates all other possible sour-
ces of arginine entry into the NO synthase and arginase pathways.
Because the parsimony principle was applied when developing the
model, we selected the minimum structure that would include just
the main features of the system to reduce model complexity to a man-
ageable level (2, 3), and we did not represent all of the compartments
of physiologic interest, such as the red blood cells mentioned by
Dioguardi. In other words, a higher-order model with a more detailed
structure was not required to analyze the data and the main features of
the system. As Dioguardi will understand, this does not mean that red
blood cells are not physiologically important with respect to arginase
activity, and, as he suggested, peripheral arginase activity, which we
estimated mainly as urea synthesis from plasma dietary arginine,
may in part be ascribed to this specic compartment. However, once
again, any contribution of red blood cells to the dynamics of post-
prandial arginine metabolism is both embedded in the data and solved
by the model. Of course, our model, like all models, remains a sim-
plication of the system but has proved to be the simplest way to
understand the dynamic behavior of the arginine nutritional system.
To answer the direct question posed by Dioguardi with regard to
plasma asymmetric-di-methyl-arginine (ADMA), we do have these
data on effects after the ingestion of arginine in this setting, and we
did not observe that plasma ADMA changed after ingestion (4). Of
note, Dioguardi cited a reference that reported an increase in plasma
ADMA with long-term arginine supplementation, whereas our re-
sults, and those of other groups, indicated no increase in different
populations and at different doses (eg, 59).
However, from a general standpoint, we agree that little is known
about the possible changes in arginine metabolism with regard to
NO compared with urea in individuals given large amounts of argi-
nine over the long term, and that changes in arginase activity have
emerged as a critical determinant of arginine-NO homeostasis and
vascular health (10). Our study was not designed to address these
potential long-term effects or to analyze the related underlying pos-
sible mechanisms. By using the integrative methodology detailed
here, future studies may be able to investigate whether, and to what
extent, the key parameters of the system are affected by a long-term
increase in arginine intake and should also be able to determine
how the system is altered in prepathological conditions (such as
with the metabolic syndrome) and in different dietary and nutri-
tional situations.
The authors declared no conicts of interest.
Franc xois Mariotti
Jean-Franc xois Huneau
UMR914 Nutrition Physiology and Ingestive Behavior
CRNH-IdF
AgroParisTech
F-75005 Paris
France
E-mail: francois.mariotti@agroparistech.fr
Hele`ne Fouillet
UMR914 Nutrition Physiology and Ingestive Behavior
CRNH-IdF
INRA
F-75005 Paris
France
REFERENCES
1. Mariotti F, Petzke KJ, Bonnet D, Szezepanski I, Bos C, Huneau JF,
Fouillet H. Kinetics of the utilization of dietary arginine for nitric oxide
508 LETTERS TO THE EDITOR
and urea synthesis: insight into the arginine-nitric oxide metabolic sys-
tem in humans. Am J Clin Nutr 2013;97:9729.
2. Cobelli C, Caumo A. Using what is accessible to measure that which
is not: necessity of model of system. Metabolism 1998;47:100935.
3. Juillet B, Saccomani MP, Bos C, Gaudichon C, Tome D, Fouillet H.
Conceptual, methodological and computational issues concerning the
compartmental modeling of a complex biological system: postprandial
inter-organ metabolism of dietary nitrogen in humans. Math Biosci
2006;204:282309.
4. Mariotti F, Huneau JF, Szezepanski I, Petzke KJ, Aggoun Y, Tome D,
Bonnet D. Meal amino acids with varied levels of arginine do not affect
postprandial vascular endothelial function in healthy young men. J Nutr
2007;137:13839.
5. West SG, Likos-Krick A, Brown P, Mariotti F. Oral L-arginine improves
hemodynamic responses to stress and reduces plasma homocysteine in
hypercholesterolemic men. J Nutr 2005;135:2127.
6. Boger GI, Rudolph TK, Maas R, Schwedhelm E, Dumbadze E, Bierend
A, Benndorf RA, Boger RH. Asymmetric dimethylarginine determines
the improvement of endothelium-dependent vasodilation by simvastatin:
effect of combination with oral L-arginine. J Am Coll Cardiol 2007;
49:227482.
7. Schwedhelm E, Maas R, Freese R, Jung D, Lukacs Z, Jambrecina A,
Spickler W, Schulze F, Boger RH. Pharmacokinetic and pharmacody-
namic properties of oral L-citrulline and L-arginine: impact on nitric
oxide metabolism. Br J Clin Pharmacol 2008;65:519.
8. Bode-Boger SM, Muke J, Surdacki A, Brabant G, Boger RH, Frolich JC.
Oral L-arginine improves endothelial function in healthy individuals
older than 70 years. Vasc Med 2003;8:7781.
9. Walker HA, McGing E, Fisher I, Boger RH, Bode-Boger SM, Jackson G,
Ritter JM, Chowienczyk PJ. Endothelium-dependent vasodilation is in-
dependent of the plasma L-arginine/ADMA ratio in men with stable an-
gina: lack of effect of oral L-arginine on endothelial function, oxidative
stress and exercise performance. J Am Coll Cardiol 2001;38:499505.
10. Morris SM Jr. Recent advances in arginine metabolism: roles and reg-
ulation of the arginases. Br J Pharmacol 2009;157:92230.
doi: 10.3945/ajcn.113.065474.
Describing a taxonomy of cognitive processes for
clinical trials assessing cognition
Dear Sir:
Stonehouse et al (1) reported that DHA supplementation im-
proved both memory and reaction time in healthy, young adults.
This randomized, placebo-controlled, double-blind clinical trial had
many strengths and was, for the most part, technically sound. How-
ever, we question the atheoretical manner in which the cognitive tests
were grouped into broader cognitive abilities.
In an accompanying editorial, Dangour and Allen (2) questioned
the applicability of the cognitive tests used by Stonehouse et al (1).
They stated that considerable variability exists in the cognitive tests
used between clinical trials and that this signicantly hampers com-
parisons between studies (2). Dangour and Allen proposed that ex-
perts in the eld should urgently agree on a set of cognitive tests to be
used consistently across clinical trials (2). We agree that efforts need
to be made to facilitate cross-study comparisons. Yet, consensus as to
a standardized set of cognitive tasks is unlikely to be agreed on given
the plethora of cognitive tests available and the fact that individual
preferences for specic cognitive tests vary greatly. Moreover, be-
cause different cognitive tests are suited to different populations and
interventions, cognitive tests are often appropriately selected on a
case-by-case basis. We propose a less radical solution to aid cross-
study comparisons in this area.
Even if researchers cannot agree on the cognitive tests used, con-
sensus should be reached on the types of cognitive functions that ex-
ist. This would then enable reviewers and readers of published
studies to better understand the scope of the tests chosen against
the full spectrum of cognitive processes that have been reliably dis-
covered. At present, many clinical trials combine cognitive tests into
broader cognitive abilities without justication from existing litera-
ture or factor analytic investigation. This appears to be the case in the
study by Stonehouse et al (1), whereby cognitive tests are combined
into cognitive domains of episodic memory, working memory, at-
tention, and processing speed without explicit justication for this
grouping. This signicantly hampers comparisons between studies
because the cognitive composites are seemingly arbitrary and may
never be created again in the same way. We suggest that a standard-
ized and evidence-based approach to grouping cognitive test data
will aid comparisons between studies. An empirically supported model
for grouping cognitive test data already exists but seems to be ignored
by the eld of clinical nutrition.
On the basis of 70 y of factor analytical work on cognition, Carroll (3)
published a seminal book on human cognitive abilities. Through ex-
tensive factor analysis of .460 data sets, his work provides a solid
empirical and science-based approach to better understanding the struc-
ture of cognition. Such is the signicance of this publication to the area
of applied psychometrics that it has been compared in importance to
Sir Isaac Newtons Mathematical Principles of Natural Philosophy (4).
FIGURE 1. The structure of cognitive abilities based on the work of Carroll (3). Note that the gure is designed to give a snapshot of the model and only
some of the 69 narrow cognitive abilities are shown. Adapted with permission from Cambridge University Press.
LETTERS TO THE EDITOR 509
Carrolls work provides an empirically veried taxonomy of
human cognitive abilities (4). In essence, Carroll (3) outlined a 3-strata
hierarchical model of cognitive ability (Figure 1). At the broadest
level, stratum 3 consists of a general intelligence factor, which sub-
sumes the following 2 strata. The second stratum includes 8 broad
cognitive abilities. Stratum 1 includes a group of 69 narrow, well-
dened abilities. All of the cognitive abilities can be classied as
belonging to one of the following domains: language, reasoning,
memory and learning, visual perception, auditory perception, idea
production, cognitive speed, knowledge and achievement, and mis-
cellaneous abilities (3). These cognitive abilities can also be bro-
ken down into additional narrow abilities. For example, memory
and learning can be further broken down into associative memory,
meaningful memory, free recall memory, visual memory, and
learning abilities. It is easy to group cognitive test scores into
these true cognitive abilities because the taxonomy was derived
through extensive factor analysis of existing cognitive tests used
throughout the past century. Carroll also provides descriptions of
each cognitive ability. We therefore suggest that researchers use
this taxonomy to group cognitive test score data or at least report
how their measures map onto this framework. This will allow
signicantly better comparison across clinical studies assessing
cognition.
The ndings reported by Stonehouse et al (1) are of great in-
terest, but as pointed out by others, heterogeneity in cognitive
outcomes between studies is signicantly limiting advancements
in this eld. It is surprising that researchers continue to group
cognitive tasks into seemingly arbitrary cognitive abilities when
a comprehensive evidence-based approach exists. Carrolls work
provides a common nomenclature for professional communica-
tion (4). From a practice perspective, this nomenclature allows
for comparison and grouping of cognitive tests across studies.
This cognitive taxonomy is widely accepted and used in the eld
of psychology, and we suggest that it also be appropriately ap-
plied in clinical trial research.
Neither of the authors had a conict of interest.
Matthew P Pase
Con Stough
Centre for Human Psychopharmacology
Swinburne University of Technology
400 Burwood Road
Hawthorn, Victoria, 3122
Australia
E-mail: matthewpase@gmail.com
REFERENCES
1. Stonehouse W, Conlon C, Podd J, Hill SR, Minihane AM, Haskell C,
Kennedy D. DHA supplementation improved both memory and reaction
time in healthy young adults: a randomized controlled trial. Am J Clin
Nutr 2013;97:113443.
2. Dangour AD, Allen E. Omega-3 fats boost brain function in adults? Are
we any closer to an answer? Am J Clin Nutr 2013;97:90910.
3. Carroll JB. Human cognitive abilities: a survey of factor analytic studies.
New York, NY: Cambridge University Press, 1993.
4. Flanagan DP, Harrison PL, eds. A history of intelligence assessment. 3rd
ed. New York, NY: The Guilford Press, 2012.
doi: 10.3945/ajcn.113.065532.
Reply to MP Pase and C Stough
Dear Sir:
In an editorial commenting on our recently published study (1), which
showed benecial cognitive effects as a consequence of 6 mo of sup-
plementation with DHA in healthy, young adults, Dangour and Allen
(2) expressed major concerns over the heterogeneity of the tests being
used to assess the cognitive function of adults in clinical trials. They
illustrated their point by noting that a wide selection of cognitive tests
had been used across the 10 relevant studies published in the Journal in
20112012, and that no 2 studies had adopted the same primary end-
points. What they failed to mention was that only one of these studies
used any form of computer-administered cognitive testing. The other
studies collected data in written or verbal form. In our own case, we
used a sophisticated computerized cognitive assessment system (COM-
PASS; Northumbria University) that was purpose designed to deliver
multiple, parallel versions of a wide range of classic, standard, and
bespoke cognitive tasks. The tasks used in the study were then chosen
with reference to the recommendations of the European Food Safety
Authoritys recent guidance on the cognitive tests that are suitable for
assessing the effects of nutritional interventions (3) and previous work
in this area by an expert panel under the auspices of the International
Life Sciences Institute (4). The potential benets of assessing cognitive
function with a computer are self-evident and include the collection of
accurate information on the speed of performing tasks and responding
to stimuli. This information represents a fundamental measure of brain
function and is always either equally informative or complementary to
information on the accuracy of task performance. Beyond this, on a
purely practical level, computerized testing also allows the standardized
presentation of properly randomized stimuli, it removes the person-
to-person interactions with a researcher that can bias and obfuscate
data, and it allows the closely controlled collection of a large amount
of data within a short period of time. We are literally surrounded in our
everyday lives by powerful personal computers, and computerized cog-
nitive testing can be readily adopted both in the laboratory and in more
ecologically valid environments. Given the above, it is somewhat baf-
ing that our own study was picked out for the editorial observations
on the heterogeneity of testing across the eld.
Dangour and Allen concluded their editorial by suggesting that
experts in cognitive testing urgently need to reach a consensus on
a small set of outcomes to use across future trials. Pase and Stough,
in response, suggest that because consensus in this regard is unlikely,
Carrolls Three Stratum Theory (5) could provide a taxonomy for
cognitive processes that could then inform a standardized and
evidence based approach to grouping cognitive test data. By Pase
and Stoughs account, our own atheoretical collapsing of task
outcomes into arbitrary composite scores (which, in reality, were
based on a previous factor analysis of a similar group of tasks) could
be replaced by simply grouping or describing the task outcomes
from a study with reference to the 8 broad cognitive ability domains
and 69 narrow, well-dened abilities in Carrolls model. Whereas
this seems, on the face of it, to be a plausible suggestion, there are
actually several major obstacles standing in the way of adopting this
approach. From a purely practical perspective, a major problem would
be deciding how a given task outcome maps onto one or more of
Carrolls factor analysisderived abilities. Presumably, this process
would require further factor analysis of multiple data sets. From
a more theoretical perspective, Carrolls model could also best be
described as a work in progress. As he himself noted in his preface,
the model was merely a starting point for future investigators and was
formulated by looking backward. Carroll also acknowledged the in-
adequacy of some of the data that he had to work with. For instance,
510 LETTERS TO THE EDITOR
he noted that the literature on memory and learning leaves much to
be desired and listed the many gaps in the data that would need to be
lled to arrive at a complete picture of this domain. In consequence,
Carrolls model has not been the xed and stationary taxonomy that
Pase and Stough would seem to be suggesting. Rather, it has been in
a continuous state of modication since its initial publication. More
recently, it has, for instance, been integrated with other models and
has been modied and added to as new data and analytic techniques
have become available (6). As an example, up to 6 new broad cog-
nitive ability domains have been suggested as additions to Carrolls
original 8 domains (6). It is also notable that Carroll started work on
his opus magnum in 1979 and worked on it for 14 y, synthesizing the
ndings of factor analyses from a vast body of data. Although he
himself was a pioneer in the application of computer technology to
his complex analyses, the data that he worked with were collected
without the benet of any such technology.
As McGrew noted recently (6), Carrolls work represented a tip-
ping point that provided the rst working map of the human cognitive
ability terrain, a terrain warranting additional exploration and rened
cartographic efforts. McGrew went on to urge the integration of cur-
rent and future research into the emerging taxonomy. However, in this
task we still seem to be laboring, certainly within the clinical
trials eld, with the astrolabes, quadrants, and verniers of the
early map makers. Simply adopting the ubiquitous technology of
our own age would necessarily make for much more accurate map-
ping tools, and therefore better maps. Although I applaud the am-
bition of Pase and Stoughs suggestion, I think the necessary rst
step toward their ultimate goal, and indeed greater standardization
of cognitive tests, is the wider adoption of sensitive computerized
testing techniques within the clinical trials eld. The resulting data
can then contribute to the factor-analytic process of further rening
the map of human cognitive ability.
The author had no conicts of interest.
David O Kennedy
Brain, Performance and Nutrition Research Centre
Northumbria University
Newcastle, NE1 8ST
United Kingdom
E-mail: david.kennedy@northumbria.ac.uk
REFERENCES
1. Stonehouse W, Conlon CA, Podd J, Hill SR, Minihane AM, Haskell C,
Kennedy D. DHA supplementation improved both memory and reaction
time in healthy young adults: a randomized controlled trial. Am J Clin
Nutr 2013;97:113443.
2. Dangour AD, Do Allen E. Omega-3 fats boost brain function in adults?
Are we any closer to an answer? Am J Clin Nutr 2013;97:90910.
3. European Food Safety Authority. Guidance on the scientic require-
ments for health claims related to functions of the nervous system, in-
cluding psychological functions. EFSA J 2012;10:2816.
4. Westenhoefer J, Bellisle F, Blundell JE, de Vries J, Edwards D, Kallus W,
Milon H, Pannemans D, Tuijtelaars S, Tuorila H. PASSCLAIM
mental state and performance. Eur J Nutr 2004;43:II85117.
5. Carroll JB. Human cognitive abilities: a survey of factor analytic studies.
New York, NY: Cambridge University Press, 1993.
6. McGrew KS. CHC theory and the human cognitive abilities project:
standing on the shoulders of the giants of psychometric intelligence
research. Intelligence 2009;37:110.
doi: 10.3945/ajcn.113.065730.
Erratum
Gibert A, Kruizinga AG, Neuhold S, Houben GF, Canela MA, Fasano A, Catassi C. Might gluten traces in wheat substitutes pose
a risk in patients with celiac disease? Apopulation-based probabilistic approach to risk estimation. AmJ Clin Nutr 2013;97:10916.
Because of a copyediting error, data are missing in Table 3 for Distribution under Scenario 3. In the rst 2 columns, under
Combination of the 4 countries, the Mean 6SD value should be 0.18 60.04, and the 95th Percentile value should be 0.24.
doi: 10.3945/ajcn.113.065615.
Erratum
Kien CL, Bunn JY, Tompkins CL, Dumas JA, Crain KI, Ebenstein DB, Koves TR, Muoio DM. Substituting dietary monoun-
saturated fat for saturated fat is associated with increased daily physical activity and resting energy expenditure and with changes
in mood. Am J Clin Nutr 2013;97:68997.
On page 693, the second sentence in the third paragraph of the Results section contains a copyediting error in which the word
or was mistakenly used: 15 or 17 subjects should read 15 of 17 subjects instead.
doi: 10.3945/ajcn.113.066282.
LETTERS TO THE EDITOR 511
Erratum
Schernhammer ES, Bertrand KA, Birmann BM, Sampson L, Willett WC, Feskanich D. Consumption of articial sweetener and
sugar-containing soda and risk of lymphoma and leukemia in men and women. Am J Clin Nutr 2012;96:141928.
The supplemental data for this article were inadvertently missed during production and were therefore not posted online. The
supplemental data le (Table 1) is now available online.
doi: 10.3945/ajcn.113.066365.
Erratum
Wilkinson SB, Tarnopolsky MA, MacDonald MJ, MacDonald JR, Armstrong D, Phillips SM. Consumption of uid skim milk
promotes greater muscle protein accretion after resistance exercise than does consumption of an isonitrogenous and isoenergetic
soy-protein beverage. Am J Clin Nutr 2007;85:103140.
On page 1039, an error appears in the legend to Figure 5. The solid circle line should represent skimmilk, and the open circle line
should represent the soy-protein beverage. The rst sentence of the gure legend should read as follows: Mean (6SEM) total
amino acid (TAA) chemical net balance (NB) after consumption of a nonfat milk-protein beverage (d) or an isonitrogenous,
isoenergetic, macronutrient-matched (750 kJ, 18.2 g protein, 1.5 g fat, and 23 g carbohydrate) soy-protein beverage (s).
doi: 10.3945/ajcn.113.067389.
Erratum
Clemente-Postigo M, Queipo-Ortuno MI, Boto-Ordonez M, Coin-Araguez L, Roca-Rodriguez MdM, Delgado-Lista J, Cardona
F, Andres-Lacueva C, Tinahones FJ. Effect of acute and chronic red wine consumption on lipopolysaccharide concentrations.
Am J Clin Nutr 2013;97:105361.
On page 1053, footnote 2 should include the following additional funding information: The study was also supported by CP07/
00095 from the ISCIII, and MdMR-R was a recipient of a fellowship from ISCIII (Rio Hortega CM11/00030), Spanish Ministry
of Economy and Competitiveness, Madrid, Spain.
doi: 10.3945/ajcn.113.066357.
512 LETTERS TO THE EDITOR