Physical activity and mood during pregnancy.

Abstract

To compare three physical activity (PA) measures in pregnant women and examine mood correlates of physical activity changes during pregnancy.
A sample of 12 pregnant women (PW) was recruited during their first trimester and 12 nonpregnant women (NPW) (baseline age = 30.7 +/- 4.4 yr) were matched to the PW (29.8 +/- 4.2 yr). Once monthly, for seven consecutive months, total weekly energy expenditure (TWEE) was assessed using a physical activity diary (PAD) and a recall interview (PAR). Accelerometers were worn for 3 d each month. Mood was also assessed. A two-factor mixed-model ANOVA was used to analyze the data (2 groups [PW vs NPW] x 7 times across pregnancy).
Mean PAR TWEE and daily mean MTI counts were positively and moderately correlated with mean PAD TWEE (PW: r = 0.40 and NPW: r = 0.50, P < 0.001). The MTI counts decreased significantly in PW by 23% from gestational weeks 12 to 36 compared with a 5% decrease in NPW. Fatigue and vigor scores improved from gestational weeks 12 to 16 and worsened from weeks 32 to 36 in PW. Changes in PA were not correlated to changes in mood in either group.
The results provide some evidence for the validity of the 7-d PAR and accelerometer as measures of physical activity in pregnancy. Healthy women who maintain an above average level of physical activity during the second and third trimesters can enjoy mood stability.

Full text

Page 1

Physical Activity and Mood during Pregnancy
MELANIE S. POUDEVIGNE1, and PATRICK J. O’CONNOR2
1Clayton State University; Health Care Management, Morrow, GA; and2University of Georgia, Exercise Science
Department, Athens, GA
ABSTRACT
POUDEVIGNE, M. S., and P. J. O’CONNOR. Physical Activity and Mood during Pregnancy. Med. Sci. Sports Exerc., Vol. 37, No.
8, pp. 1374–1380, 2005. Purpose: To compare three physical activity (PA) measures in pregnant women and examine mood correlates
of physical activity changes during pregnancy. Methods: A sample of 12 pregnant women (PW) was recruited during their first
trimester and 12 nonpregnant women (NPW) (baseline age ? 30.7 ? 4.4 yr) were matched to the PW (29.8 ? 4.2 yr). Once monthly,
for seven consecutive months, total weekly energy expenditure (TWEE) was assessed using a physical activity diary (PAD) and a recall
interview (PAR). Accelerometers were worn for 3 d each month. Mood was also assessed. A two-factor mixed-model ANOVA was
used to analyze the data (2 groups [PW vs NPW] ? 7 times across pregnancy). Results: Mean PAR TWEE and daily mean MTI counts
were positively and moderately correlated with mean PAD TWEE (PW: r ? 0.40 and NPW: r ? 0.50, P ? 0.001). The MTI counts
decreased significantly in PW by 23% from gestational weeks 12 to 36 compared with a 5% decrease in NPW. Fatigue and vigor scores
improved from gestational weeks 12 to 16 and worsened from weeks 32 to 36 in PW. Changes in PA were not correlated to changes
in mood in either group. Conclusions: The results provide some evidence for the validity of the 7-d PAR and accelerometer as measures
of physical activity in pregnancy. Healthy women who maintain an above average level of physical activity during the second and third
trimesters can enjoy mood stability. Key Words: ACCELEROMETER, DIARY, RECALL, SELF-WORTH
M
ommendation of bed rest for approximately 20% of preg-
nant women, and both retrospective and prospective studies.
For example, a retrospective survey of 9953 pregnant
women found that the percent of the sample who reported
exercising for 30 min or more at least three times per week
changed from 55% before pregnancy to 42% after preg-
nancy. This change was observed despite 7% of the women
who were inactive before pregnancy becoming active during
pregnancy (27).
One weakness of the available evidence regarding
changes in physical activity during pregnancy is the crude
recall methods that typically have been used to assess phys-
ical activity. Most studies with pregnant women have used
physical activity measures with no published evidence of
their reliability or validity despite the likelihood that the
accurate assessment of physical activity is more difficult in
pregnant women (23). Studies that have used established
physical activity measures often have been characterized by
other limitations, including infrequent assessments (e.g.,
only one to two consecutive days of activity measured only
two to three times during pregnancy) and a failure to employ
any women reduce their physical activity during
pregnancy. Evidence for this statement stems
from common clinical practices, such as the rec-
nonpregnant controls (5,9,12,23,26). Only one study has
compared several methods for measuring physical activity
in pregnant women (23). Pregnancy is associated with in-
creased psychological distress for many women, including
increased anxiety, depression, and fatigue. A prospective
investigation of more than 14,000 pregnant women found
that depression symptom scores were higher at 18 and 32 wk
of pregnancy than at 8 wk and 8 months postpartum (7). A
prospective study of 117 pregnant women found that symp-
toms were most common during the third trimester and 97%
of the women reported fatigue at some point during their
pregnancy (28).
Given the known links between inactivity and reduced
mental health, a relationship between pregnancy-related re-
ductions in physical activity and psychological distress is
plausible. Only a small number of investigations have ex-
amined the relationship between changes in physical activ-
ity during pregnancy and changes in mood. These studies
found that low physical activity was associated with higher
scores on anxiety (5,9), depression (5,12) and fatigue scales
(26). Conclusions from these investigations are tempered by
methodological limitations such as the absence of control
groups or the failure to use physical activity measures with
supportive validity evidence.
The primary purposes of this prospective investigation
were to describe physical activity patterns during the second
and third trimester of pregnancy using three different, es-
tablished measures and make comparisons among the mea-
sures. A secondary purpose was to learn whether the
changes in physical activity during pregnancy were associ-
ated with changes in mood. The research was designed to
extend the current literature by making frequent (monthly)
assessments, by measuring physical activity using three
Address for correspondence: Me ´lanie S. Poudevigne, Ph.D., Clayton State
University, Office of Health Care Management, 5900 North Lee St.,
Morrow GA 30260; E-mail: mpoudevigne@yahoo.com.
Submitted for publication September 2004.
Accepted for publication March 2005.
0195-9131/05/3708-1374/0
MEDICINE & SCIENCE IN SPORTS & EXERCISE®
Copyright © 2005 by the American College of Sports Medicine
DOI: 10.1249/01.mss.0000174907.27818.ff
1374

Page 2

commonly used instruments, and by comparing the findings
to a group of nonpregnant women controls. It was hypoth-
esized that: 1) there would be a moderate positive correla-
tion between energy expenditure assessed with a criterion
measure of physical activity (7-d diary) and energy expen-
diture estimated using a seven day physical activity recall,
and that this relationship would be stronger for the controls
than for pregnant group; 2) there would be positive corre-
lations of weak-to-moderate strength between both these
measures and an objective measure of physical activity (i.e.,
acceleration counts), and that these relationships would be
stronger for the controls than the pregnant group; 3) the
moods of anxiety, depression, fatigue, and vigor would
worsen at the end of pregnancy compared with both the
beginning of the second trimester and to controls; and 4)
moderate associations would be found between changes in
physical activity and changes in mood.
METHODS
Recruitment
Potential participants were recruited through newspaper
announcements, by flyers placed at women clinics and ob-
stetricians offices, and at maternity-related businesses. Preg-
nant participants were required to be healthy women with-
out a complicated pregnancy who were between the ages of
18 and 40, and who were able to read and understand
English. Women were recruited into the study during their
first trimester. All participants provided written informed
consent after reviewing a document that had been approved
by the human research review committee at the University
of Georgia. Each woman was given $70 for her time.
A total of 30 pregnant women inquired about the study
and 18 completed the first testing session. The reasons
expressed by the 12 women who did not engage in the study
included distance to the laboratory, time commitment re-
quired, and inadequate monetary or other incentives. No
data were obtained from these women. After the first testing
session, six women dropped out, two because of relocation
and four because of the time demands (primarily with the
diary). After the remaining group of 12 pregnant partici-
pants completed all the testing sessions, a group of 12 age-
(? 2 yr), height- (? 5 cm), weight- (? 2.5 kg), and race-
matched nonpregnant/nonlactating control women were re-
cruited. The mean age (SD) at enrollment was 29.8 yr (4.2)
for the pregnant group and 30.7 yr (4.4) for the controls. The
mean baseline weight (SD) was 68.9 kg (11.9) for the
pregnant group and 66.5 kg (13.5) for the controls. Forty
two percent of the pregnant group was parous, with a me-
dian parity of one (range, 0–3). Ten women worked
throughout their pregnancy. Ninety-two percent of the preg-
nant women took prenatal vitamins and 42% of the controls
took birth control pills. Eighty five percent of the partici-
pants in each group were white. None of the participants
took psychiatric medication or was a competitive athlete.
Power Analysis
Thirteen subjects per group provided a priori statistical
power of 0.80 for testing the presence of a large interaction
effect (d ? 1.1) using a mixed model with two groups
(pregnant vs nonpregnant) and seven repeated trials (24;
Table 3). This calculation assumes a correlation between
repeated trials of 0.90 and a two-sided test at an alpha of
0.05. Because hypothesis 3 was one-sided, 12 participants
per group provided an a priori power of at least 0.80.
Procedures
Physical activity, body weight, and mood were measured
at gestational weeks 12, 16, 20, 24, 28, 32, and 36. The
women were asked to refrain from eating, drinking caffeine,
and exercising for at least 2 h before coming to the labora-
tory. Also, the women were asked to not deviate from their
normal activities of daily living.
Measurements
Demographic information was collected at the beginning
of the study. Weight was measured at each time point using
a double balance beam scale. Physical activity was mea-
sured using a 7-d diary, a 7-d physical activity recall, and 3
d of motion sensor recordings.
Physical activity diary. A diary of daily physical
activities was obtained for seven continuous days before
each laboratory testing session using the 3-d physical-activ-
ity record method (2). With this method the day is divided
into 96 periods of 15 min each. For each 15-min period, the
participants indicated the dominant physical activity per-
formed. An activity card was provided that gave examples
of activities of different intensities divided into nine cate-
gories. When a participant completed an activity that was
not listed, she was instructed to apply the categorical value
that seemed closest in intensity. Uncertainties were reported
to the study coordinator for proper classification at the time
of the next laboratory visit. Administration of the diary and
calculation of estimated energy expenditure followed Bou-
chard et al. (2). Before actual data collection, each woman
practiced completing the diary until she became proficient
with it. Estimated energy expenditure using this diary has
been found to be in close agreement with doubly labeled
water. Energy expenditure estimates from this diary have
been found to be significantly correlated (r range from 0.24
to 0.86) with other measures of energy expenditure based on
questionnaires and accelerometry (21). The diary was con-
sidered to be the criterion measure of physical activity in
this investigation.
Physical activity recall. An interviewer-administered
7-d physical activity recall was used to assess physical
activities performed during the week preceding each testing
session (1). Participants were asked to estimate the number
of hours spent each day (starting with the current day and
working backwards) in sleep, moderate, hard, and very hard
physical activities. Any time not accounted for by these
activities was categorized as light activity. Details of the
PHYSICAL ACTIVITY AND MOOD IN PREGNANCYMedicine & Science in Sports & Exercise?
1375

Page 3

interview procedures and the methods used to estimate
energy expenditure from the physical activity reports have
been described elsewhere (13). The 7-d physical activity
recall is one of the most commonly used measures of phys-
ical activity. Energy expenditure estimates from the 7-d
recall have been found to be significantly correlated (r ?
0.20–0.86) with other measures of energy expenditure
based on diary, questionnaire, heart rate, acceleration count,
and doubly labeled water data (1,6,19,21).
Motion sensor. The MTI®accelerometer (Manufac-
turing Technology Inc., Fort Walton Beach, FL) was used to
obtain an objective measure of physical activity. The MTI is
a uniaxial piezo-electric accelerometer. Accelerations are
measured within the range of human movements and con-
verted to counts. One count is equivalent to 16 mg?s?1. The
MTI counts have been found to be highly correlated (r ?
0.85) to the speed of both treadmill walking and running as
well as oxygen consumption during walking (3,8).
The MTI was worn on the left hip, clipped to a waist
band. Participants were instructed to remove it while swim-
ming or showering. The participants were instructed to wear
the MTI for two weekdays and one weekend day from the
time they woke up to the time they went to bed. If the MTI
record revealed that these instructions were not followed the
participant was asked to wear the MTI for three additional
full days, including one weekend day, to minimize missing
data. Three women were asked to wear the device again.
MTI counts were recorded and stored every minute. The
number of counts recorded each day was totaled and divided
by the number of minutes the MTI was worn each day to
yield a daily average number of counts per minute. The
criterion physical activity measure from the MTI was the
3-d mean of the daily average counts per minute.
Mood. Mood was measured using the 65-item Profile of
Mood States (POMS) questionnaire. The POMS measures
six mood states: tension, depression, anger, vigor, fatigue,
and confusion (15). The POMS was used in this study
because it is among the most accepted and widely used
measures of mood available and it has extensive evidence
supporting its reliability and validity (17), including data
from large samples of pregnant women (25). The partici-
pants described how they had been feeling “during the past
week including today” by rating the intensity of their feeling
using five categories: “not at all,” “a little,” “moderately,”
“quite a bit,” and “extremely.” There is a large body of
factor analytic, correlational, and experimental evidence
supporting the idea that scores from this questionnaire can
be interpreted as six separate measures of transitory mood
states (15,16). For example, in a sample of 400 adults, the
POMS scores correlated highly (r ? 0.54–0.70) with cor-
responding visual analog mood scales (16). Completion of
this questionnaire took 10 min.
Statistical Analysis
Data were entered into two spreadsheets by two separate
investigators. Comparing the two spreadsheets facilitated
the identification of data entry errors.
The control group had no missing data. A small amount
of data was missing for the pregnant group primarily be-
cause of a failure to attend one of the seven testing days.
Missing data from the pregnant group involved five of 252
daily MTI records and three of 84 7-d recall, physical
activity diary, and mood measures. Missing data were esti-
mated by interpolation from scores obtained during the
testing session before or after (or both) the testing session
from which the data were missing.
Statistical analyses were conducted with SPSS Statistical
Software version 11.5 (SPSS Inc., Chicago, IL). The as-
sumptions for analysis of variance were tested; normality
and independence assumptions were met. When sphericity
was not met, degrees of freedom were adjusted using the
Huynh–Feldt adjustments. A box plot analysis revealed no
outliers for any variable. For hypothesis 1 and 2, Pearson
correlations were computed with the pregnant and control
groups separately (12 participants per group ? 7 trials ? 84
data points). Hypothesis three was tested using a mixed-
model ANOVA, with repeated measures on the time factor.
A significant group and time interaction was hypothesized.
For hypothesis four, Spearman correlations were performed
on the changes in physical activity between weeks 16 and 36
and the corresponding changes in the four measures of
mood, which the literature suggests change with pregnancy
(anxiety, depression, vigor and fatigue). Spearman rank–
order correlations were used as a nonparametric measure of
association based on the rank of the data values because the
change scores were not normally distributed, even after log
transformation.
RESULTS
Physical activity. Table 1 reports body weight data
and the time spent both sleeping and in different intensities
of physical activities based on the 7-d recall. No significant
group main effects or group ? time interactions were found
for any of the variables regarding sleep or time spent in
different intensities.
Figure 1 illustrates the relation between energy expendi-
ture estimated with the diary and recall methods. Group and
time main effects and group ? time interactions were not
statistically significant for either the diary or the recall
method. Averaged across the seven testing sessions, the
recall method underestimated (insignificantly) energy ex-
penditure by an average of 37 kcal?kg?1?wk?1for the preg-
nant women, and 23 kcal?kg?1?wk?1for the controls. Figure
2 shows a scatterplot of the relationship between diary and
recall estimates of energy expenditure. The correlation be-
tween the diary and recall estimates of energy expenditure
for pregnant women was r ? 0.41 (N ? 84). The correlation
between the diary and recall estimate of energy expenditure
for the controls was r ? 0.52 (N ? 84).
The pregnant group wore the physical activity monitors
for an average of 14.1 h across the seven testing sessions
(group mean range 13.6–14.5). The controls wore the phys-
ical activity monitors for an average of 15.1 h across the
seven testing sessions (range of 14.6–16). Figure 3 shows
1376
Official Journal of the American College of Sports Medicine http://www.acsm-msse.org

Page 4

the averaged daily MTI counts across time for the pregnant
and nonpregnant groups. A mixed-model two-way ANOVA
on counts per minute showed insignificance for the group ?
time interaction and for the main effect for group. The main
effect for time was significant (F ? 2.51; df ? 4.6, 101.0;
P ? 0.039; ?2? 0.10). Univariate analysis showed a
significant difference in the time factor for the pregnant
women (F ? 2.60; df ? 6.66; P ? 0.025, ?2? 0.19). A
series of paired t-tests showed differences in time for the
pregnant group between sessions 1 and 6 (P ? 0.038), 2 and
6 (P ? 0.035), 2 and 7 (P ? 0.032), 4 and 6 (P ? 0.025),
and 4 and 7 (P ? 0.042). The level of MTI activity counts
decreased by 23% in the pregnancy group and 5% in the
controls between gestational weeks 12 and 36. Figure 4
shows a scatterplot of the relationship between energy ex-
penditure estimated from the diary and physical activity
measured by MTI counts. The correlation between these two
variables for the pregnant group was r ? 0.41 and for the
controls was r ? 0.50. The correlation between the MTI
counts and energy expenditure estimated from the seven day
recall was r ? 0.23 for the pregnant group and r ? 0.30 for
the controls.
Mood. The mood data are provided in Table 2. No
significant group main effects or group ? time interactions
were found for any of the mood variables. A significant time
main effect was found for the moods of fatigue (F ? 4.3; df
? 4.4; 97.4; P ? 0.002; ? (2) ? 0.16) and vigor (F ? 3.56;
df ? 6, 132; P ? 0.003; ? (2) ? 0.14). For the pregnant
women the largest changes in fatigue and vigor occurred
during weeks 12–16 and weeks 32–36. Fatigue decreased
and vigor increased from weeks 12 to 16. Fatigue increased
and vigor decreased from weeks 32 to 36. The controls
showed similar changes in fatigue scores during these
weeks. In general, Spearman correlations between changes
in physical activity and mood were not significant for the
pregnant or control groups (range from r ? ?0.55 to 0.35;
the highest negative correlation was between fatigue and
physical activity assessed by recall).
DISCUSSION
This study compared three different methods for assess-
ing physical activity in a group of pregnant women because
methods for measuring physical activity rarely have been
validated in pregnant women.
No prior studies of pregnant women have compared es-
timates of energy expenditure from a physical activity diary
to the 7-d recall. The range of mean weekly energy expen-
diture values from both the diary and recall methods for the
pregnant and control women in this study were consistent
with prior studies of nonobese adult women in which
weekly energy expenditures of 250–305 kcal?kg?1?wk?1
have been reported using doubly labeled water and indirect
calorimetry (20). Weekly energy expenditures of a similar
range (230–300 kcal?kg?1?wk?1) also have been reported
FIGURE 1—Seven-day diary and 7-d recall
estimates of energy expenditure (mean ? SD)
in 12 pregnant women during gestational
weeks 12–36 and 12 nonpregnant controls.
Pregnant women diary, closed square; non-
pregnantdiary,
open
women recall, closed triangle; nonpregnant
women recall, open triangle.
square; pregnant
TABLE 1. Body weight and weekly duration (h ? SD) in sleep, light, moderate, hard, and very hard physical activity intensities in 12 pregnant and 12 nonpregnant women during
gestational weeks 12–36.
Gestational Week12 16
SleepNonpregnant women 54 ? 6 57 ? 5
Pregnant women60 ? 560 ? 5
Light activitiesNonpregnant women 106 ? 10102 ? 10
Pregnant women 100 ? 7 95 ? 7
Moderate activitiesNonpregnant women 5.5 ? 7.4 7.5 ? 9.7
Pregnant women 7.5 ? 6.610.5 ? 8.4
Hard activities Nonpregnant women 1.1 ? 1.7 1.4 ? 2.0
Pregnant women0.87 ? 0.670.85 ? 0.84
Very hard activitiesNonpregnant women0.40 ? 0.54 0.33 ? 0.65
Pregnant women0.50 ? 0.14 0.42 ? 0.81
Body weightNonpregnant women66.5 ? 13.5 66.5 ? 13.2
Pregnant women68.9 ? 11.970.1 ? 11.6
20 24283236
57 ? 5
59 ? 5
101 ? 9
100 ? 7
8.5 ? 8.8
8.0 ? 6.2
0.67 ? 1.1
0.83 ? 0.11
0.46 ? 0.86
0.25 ? 0.59
66.3 ? 13.0
70.3 ? 11.9
56 ? 6
57 ? 4
105 ? 7
103 ? 9
5.0 ? 4.8
7.75 ? 9.5
0.71 ? 0.9
0.84 ? 0.97
0.31 ? 0.50
0.08 ? 0.29
66.1 ? 12.8
72.8 ? 12.7
60 ? 10
59 ? 6
99 ? 12
96 ? 7
7.0 ? 6.2
12.5 ? 6.4
0.77 ? 1.2
0.86 ? 0.80
0.54 ? 1.72
0.08 ? 0.29
66.0 ? 13.8
74.7 ? 12.8
53 ? 4
58 ? 5
105 ? 9
97 ? 10
8.75 ? 9.0
11.25 ? 9.8
1.4 ? 4.9
0.84 ? 0.73
0.27 ? 0.62
0.06 ? 0.15
66.0 ? 13.4
76.3 ? 13.2
53 ? 7
58 ? 5
103 ? 13
102 ? 8
11.0 ? 13.8
8.0 ? 8.1
0.90 ? 1.6
0.82 ? 0.13
0.19 ? 0.47
0.00 ? 0.00
66.0 ? 13.9
78.8 ? 13.5
PHYSICAL ACTIVITY AND MOOD IN PREGNANCY Medicine & Science in Sports & Exercise?
1377

Page 5

for college students and community samples using the 7-d
recall (1,6).
The 7-d recall results showed that the pregnant group, on
average, was as active during their second and third trimes-
ters as 566 nonpregnant women (25–34 yr old) who partic-
ipatedinalargecommunity
kcal?kg?1?wk?1, respectively) (1). Nonetheless, weekly en-
ergy expenditure was underestimated to a small and insig-
nificant degree with the 7-d recall compared with the diary
by 37 kcal?kg?1?wk?1in the pregnant group. This underes-
timate was 14 kcal?kg?1?wk?1greater than the underesti-
mate found for the control group (23 kcal?kg?1?wk?1). In-
accuracies in memory represent a plausible explanation for
this small underestimation of energy expenditure with the
recall method. The recall required the women to rely on
their memory of activities performed during the prior week.
study(252 vs 256
The women may have forgotten certain physical activities,
especially short duration, moderate intensity, or infrequent
activities performed at the beginning of the recall period.
Irwin and Ainsworth (10) suggested that the estimation of
total energy expenditure from the 7-d recall was more vari-
able than from the 7-d diary in a study of 24 males because
the 7-d recall overestimated by 68 kcal?kg?1?wk?1com-
pared with the 7-d diary. Conversely, Richardson and col-
leagues (19) reported that the 7-d recall underestimated total
energyexpenditurecompared
kcal?kg?1?wk?1in 50 women of child-bearing age. Prior
studies with nonpregnant samples suggest that the 7-d recall
is most accurate in distinguishing activity extremes (6).
The 7-d recall method resulted in weekly energy expen-
diture values for pregnant women that were moderately
correlated with the criterion diary method (r ? 0.52), and
this correlation was similar to that found for the nonpregnant
control group (r ? 0.40). Correlations of both similar (11)
and higher (6) magnitude have been reported in studies of
nonpregnant samples. An examination of the scatterplot of
the relationship in this study (i.e., Fig. 2) reveals that the
slopes of the regression lines appear to differ for the preg-
nant compared with the control group. A statistical test for
a difference in the slopes of these regression lines was not
performed because one of the assumptions underlying the
test (independence) could not be met. Nonetheless, the re-
gression lines appear to show greater underestimation of
energy expenditure associated with higher physical activity
in the pregnant compared to the control group.
The 7-d recall method resulted in weekly energy expen-
diture values for pregnant women that were weakly corre-
lated with the accelerometer counts (r ? 0.23), and this
correlation was similar to that found for the nonpregnant
control group (r ? 0.30). Correlations of higher magnitude
have been reported in studies of nonpregnant samples
to thediaryby21
FIGURE 2—The relationship between weekly energy expenditures
from the diary and the recall in 12 pregnant women at gestational
weeks 12–36 and 12 nonpregnant women. Pregnant women, closed
square/solid line; nonpregnant women, open circle/dashed line.
FIGURE 3—Physical activity in the 12 pregnant women during gestational week 12–36 and 12 nonpregnant women controls. Data are expressed
in average daily acceleration counts per minute (mean ? SD). Pregnant women, closed triangle; nonpregnant women, open triangle.
1378
Official Journal of the American College of Sports Medicine http://www.acsm-msse.org