Upload
others
View
3
Download
0
Embed Size (px)
Citation preview
obesity
reviews
© 2005 The International Association for the Study of Obesity.
obesity
reviews
6
, 43–65
43
Blackwell Science, LtdOxford, UKOBRobesity reviews1467-78812004 The International Association for the Study of Obesity. 6
14365
Review Article
Psychosocial predictors of weight loss P. J. Teixeira
et al.
Address reprint requests to: PJ Teixeira,
Faculdade de Motricidade Humana, Estrada
da Costa, Cruz Quebrada, 1495-688 Lisboa,
Portugal. E-mail: [email protected]
A review of psychosocial pre-treatment predictors of weight control
P. J. Teixeira
1
, S. B. Going
2
, L. B. Sardinha
1
and T. G. Lohman
3
1
Department of Exercise and Health, Faculty of
Human Movement, Technical University of
Lisbon, Lisbon, Portugal and Departments of
2
Nutritional Sciences and
3
Physiology, Body
Composition Research Laboratory, The
University of Arizona, Tucson, AZ, USA
Received 7 May 2004; revised 13 September
2004; accepted 15 September 2004
Summary
Prompted by the large heterogeneity of individual results in obesity treatment,many studies have attempted to predict weight outcomes from information col-lected from participants before they start the programme. Identifying significantpredictors of weight loss outcomes is central to improving treatments for obesity,as it could help professionals focus efforts on those most likely to benefit, suggestsupplementary or alternative treatments for those less likely to succeed, and helpin matching individuals to different treatments. To date, however, research effortshave resulted in weak predictive models with limited practical usefulness. The twoprimary goals of this article are to review the best individual-level psychosocialpre-treatment predictors of short- and long-term (1 year or more) weight loss andto identify research needs and propose directions for further work in this area.Results from original studies published since 1995 show that few previous weightloss attempts and an autonomous, self-motivated cognitive style are the bestprospective predictors of successful weight management. In the more obese sam-ples, higher initial body mass index (BMI) may also be correlated with largerabsolute weight losses. Several variables, including binge eating, eating disinhibi-tion and restraint, and depression/mood clearly do not predict treatment out-comes, when assessed before treatment. Importantly, for a considerable numberof psychosocial constructs (e.g. eating self-efficacy, body image, self-esteem, out-come expectancies, weight-specific quality of life and several variables related toexercise), evidence is suggestive but inconsistent or too scant for an informedconclusion to be drawn. Results are discussed in the context of past and presentconceptual and methodological limitations, and several future research directionsare described.
Keywords:
Correlates, moderators, obesity, treatment.
Introduction
During the last three decades, predictors of outcomes ofobesity treatment have been periodically investigated inoriginal research trials, and summarized in review articles(1,2) and book chapters (3,4). Collectively, these studieshave concluded that predicting weight loss with acceptableaccuracy is difficult to impossible. The large number ofvariables potentially involved, each explaining only a very
small share of the variance in weight change, and the com-plex interactions among predictors, behaviours and typesof treatment are possible reasons for the failure to builduseful predictive profiles/models (5,6). However, a numberof conceptual and methodological limitations may partiallyexplain the disappointing results (7). Since the last majorreview on this topic was published (8) the number ofresearch studies conducted on obesity treatment hasincreased substantially, treatment programmes have
obesity
reviews (2005)
6
, 43–65
44
Psychosocial predictors of weight loss
P. J. Teixeira
et al.
obesity
reviews
© 2005 The International Association for the Study of Obesity.
obesity
reviews
6
, 43–65
evolved and new variables have been tested as potentialpredictors of weight loss. New analytical methods havealso been introduced. For these reasons, and consideringthe likely usefulness of prospectively predicting outcomesin obesity treatment (1), an update on pre-treatment pre-dictors of weight loss is warranted.
The heterogeneity of obesity and weight loss results
Obesity is a heterogeneous condition and individualresponses to standardized protocols leading to weightchange are highly variable (9). In real life settings such asobesity treatment programmes, physiological and psycho-logical individual factors, some of which may carry astrong genetic influence (10), interact with environmentalfactors in a complex manner, producing a wide range orindividual responses in both the magnitude and the rate ofweight changes. At a psychological level, the principalfocus of this review, individual variability is the norm.Some people perceive their excess weight as emotionallydistressing, while others of similar weight appear unaf-fected (11,12). Individual perceptions of body weight andshape may determine a person’s body image (13) and qual-ity of life (14) more than weight itself. A subgroup ofindividuals presenting for obesity treatment displays dys-functional eating behaviours while others display a normalrelationship with food (15). Finally, a negative psycholog-ical impact of dieting appears to be present in some, butdefinitely not all dieting individuals (16).
A large degree of weight change variability is present inweight loss treatments such as hypocaloric balanced dietsand behaviour modification (17), very-low calorie diets(VLCD) (18), dietary regimens without exercise (19), exer-cise without dieting (20) and various other therapeuticmodalities (21–23). Regarding long-lasting weight control,while current prevalence data (24) and other reports (25)suggest widespread failure in available treatments, datafrom prospective trials (26,27), retrospective studies(28,29), as well as findings from self-selected groups of thepopulation (30) suggest success is possible and indeedoccurs for a sizeable number of people.
The source of the variability in response to obesity treat-ment remains unclear; researchers have not been able toadequately explain why some people adopt attitudes andbehaviours needed for weight loss while others do not (31).In the last few decades, the majority of treatment studieshave been focused on improving and contrasting treatmentmodalities in randomly grouped subjects, while investigat-ing which treatments are more effective to which individ-uals, a long-standing need (32,33), has received lessattention. Indeed, concerns have been raised that psycho-logical, biological and environmental differences amongthe obese are too large for any single treatment to ever beeffective in this population, if considered as a homogeneous
group (34–37). Any treatment will be useful to some sub-jects, but no treatment will be effective for all, suggestingthe need for a better matching between treatment featuresand patients’ needs (32).
Several patient-matching treatment algorithms, primarilybased on biological or morphological variables [e.g. degreeof overweight, waist circumference, cardiovascular disease(CVD) risk factors] are now available (38–43), highlightingthat not all overweight/obese persons are the same and thatemploying similar treatment protocols in diverse groups isnot an effective approach. These guidelines also indicatethat psychological factors such as personal preferences,expectations, attitudes towards physical activity, emotionaldistress, depression and body image should be consideredwhen making treatment decisions (38–40,42). Unfortu-nately, empirical evidence supporting the ability of these orother variables to consistently predict outcomes is notwidely available. Retrospective studies have identifiedbehavioural (e.g. low levels of physical activity, lack of self-monitoring, emotional eating, a high-fat diet) as well aspsychological (e.g. dichotomous thinking, body weight-and shape-based self-worth, unrealistic weight goals, lowself-efficacy, higher anxiety and depression) correlates ofunsuccessful long-term weight loss (29,30,44). However,thus far only 20–35% of the variance in subsequent weightloss has been predicted from baseline variables (17,45–51).
Why study predictors of weight loss?
The question of whether all individuals who
want
to loseweight
can
or
should
engage in weight reduction pro-grammes at a particular time in their lives has been raisedseveral times in the past (8,16,32,52). The assessment ofparticipants’ characteristics and level of readiness as theyenter a programme has been justified in two primary ways.First, to optimize the intervention’s efficacy for each indi-vidual or groups with similar profiles through a bettermatching of treatments to participants’ measured charac-teristics (8,40). This is a long-standing goal in obesity treat-ment, but one that has rarely been practised (51,53–55). Asecond rationale is to screen some candidates out of treat-ment, if their likelihood of success is estimated as very low(36,52). This would spare them further distress upon arepeated failed attempt, save limited resources and permita higher focus on individuals who stand a better chance ofsuccess. Under the title ‘Ready or Not: Predicting WeightLoss’, an expert panel on the evaluation and treatment ofoverweight and obesity observed that ‘predicting a patient’sreadiness for weight loss and identifying potential variablesassociated with weight loss success is an important step inunderstanding the needs of patients’ (p. 21) (56).
This article reviews what is known about predictors,emphasizes how conceptual and methodological factorshave limited what has been studied and discovered and
obesity
reviews
Psychosocial predictors of weight loss
P. J. Teixeira
et al.
45
© 2005 The International Association for the Study of Obesity.
obesity
reviews
6
, 43–65
suggests a new generation of studies to advance the field.We will focus on pre-treatment questionnaire-based psy-chosocial variables and other indices such as weight loss/dieting history and initial weight, all of which are easilyaccessible in most clinical, commercial and researchsettings.
Methods and study description
In the present review, articles published after 1995 report-ing associations between baseline/pre-treatment individualfactors and subsequent weight loss, within the context ofadult obesity treatment programmes, were identified andanalysed. The option of choosing only studies publishedsince 1995 is based on the existence of several reviewpapers and book chapters on this topic published up to,but not after 1995, as described later in this text. Indepen-dent measures were primarily psychosocial variablesassessed by questionnaire. Because initial weight has beenfrequently considered a predictor of treatment outcomesand was available in many of the studies we selected, it wasalso included in this review. Other demographic variablessuch as age, gender, marital status, income and educationallevel were not analysed (as predictors) in the large majorityof studies we selected, and were not included in this review.
Physical activity, diet, and behaviour therapy are cur-rently recommended for all obese subjects (56). Thus, weincluded only intervention studies that involved a lifestylebehaviour modification programme generalized aimed atchanging diet, or diet and physical activity. Studies usingmeal replacements, hypocaloric balanced diets (HBD, forthe purpose of this review defined as an intake of approx-imately 1200–1500 kcal d
-
1
), low-calorie diets (LCD,approximately 800–1200 kcal d
-
1
), very-low-calorie diets(VLCD, approximately
<
800 kcal d
-
1
), and treatment forbinge eating disorder (BED) were included as long asweight loss was a clear outcome of the study and a behav-ioural intervention was present. Whenever an LCD orVLCD was not mentioned in the text (or could not beassumed from caloric intake data reported), but some formdietary therapy appeared to be a part of the intervention,an HBD was assumed. In these cases, we cannot be surethat subjects were prescribed a diet lower than 1500 kcald
-
1
but some degree of energy restriction can probably beassumed (‘non-dieting’ studies were not covered in thisreview). Only one study was identified that analysed psy-chosocial pre-treatment predictors of success using phar-macotherapy for weight loss (57); because no behaviouralcomponent was described as part of this study’s interven-tion and considering its singularity compared to all otherstudies, it was not included. Surgery studies and interven-tions to prevent weight gain were also excluded.
Twenty-nine studies were identified and included in thisreview’s primary analysis, reflecting 38 predictive models
for weight loss and/or maintenance from variables assessedat baseline (Table 1). A model is defined by one or severalpredictors associated with a weight loss outcome for agiven time period. Several studies included more than onemodel (i.e. if they covered different time periods). InTable 1, models are organized by the use of HBD/LCD(Table 1a) or VLCD (Table 1b). Programmes with VLCDare usually more intensive and induce larger averageweekly weight losses, and are typically employed withgroups displaying higher initial levels of obesity, whichjustifies the separate analysis. Within each category (HBD/LCD or VLCD), predictive models are sorted according tothe duration of the study. Models with no follow-up/main-tenance period or lasting less than 1 year are listed firstand considered as
weight loss
studies (sorted by durationof initial treatment). They are followed by predictivemodels including weight loss and a follow-up/maintenancephase and reaching 1 year or more in duration(treatment
+
follow/up), which were considered
weight lossand maintenance
studies. One year is has been indicated byan expert panel on obesity as an appropriate thresholdbetween weight loss and the maintenance of the weight lost(8). Finally, some models evaluated predictors of changesoccurring only during the follow-up/maintenance period.They are listed last (
weight maintenance
studies).
Study description
All studies reviewed included female subjects while 15(54%) had mixed-gender samples. Initial mean body massindex (BMI) for participants varied from
~
29 to
~
46 kgm
-
2
and sample size ranged from 42 to 444 subjects. Theaverage duration of the weight loss phase was 21 weeks forstudies with HBD/LCD and 22 weeks for VLCD trials,although the actual VLCD feeding period was shorter insome cases. In three VLCD studies, only a subgroup of thetotal sample underwent VLCD but outcomes were notshown separately regarding weight loss predictors.Approximate average weekly weight loss was 0.80 kg forVLCD studies and 0.58 kg for non-VLCD studies (weightloss was 0.52 kg week
-
1
when only HBD studies were con-sidered, i.e. when studies using LCD and treatment for BEDwere excluded). It should be noted that weight changes arereported for subjects completing the study. Intent-to-treatanalyses, where data for all starting participants areincluded (which was not reported in the overwhelmingmajority of studies), may have yielded lower averageweekly weight losses.
Average follow-up/maintenance phase in the 13 studiesthat included such a phase was approximately 63 weeks.During maintenance or follow-up, average yearly weightchange was
+
4.9 kg and
+
1.1 kg for VLCD (six studies)and HBD/LCD (13 studies) respectively (overall average,
+
1.9 kg year
-
1
). Attrition rates were variable, ranging from
46
Psychosocial predictors of weight loss
P. J. Teixeira
et al.
obesity
reviews
© 2005 The International Association for the Study of Obesity.
obesity
reviews
6
, 43–65
Tab
le 1
Pre
dic
tion
mod
els
from
stu
die
s us
ing
beh
avio
ur m
odifi
catio
n an
d (
a) h
ypoc
alor
ic o
r lo
w-c
alor
ie d
iets
/(b)
very
-low
-cal
orie
die
ts a
s a
mea
ns t
o w
eig
ht lo
ss
Mod
el (
refe
renc
e)S
amp
leTr
eatm
ent,
wei
ght
loss
pha
seFo
llow
-up
/mai
nten
ance
pha
se
Out
com
e (d
epen
den
t va
riab
le)
pre
dic
ted
fro
m
bas
elin
e va
riab
les
Siz
e
1
Gen
der
BM
ITy
pe
Dur
atio
nW
t. lo
ssA
ttriti
onD
urat
ion
Wt.
chan
ge
2
Attr
ition
3
(a)
Pos
ton_
BL
(205
)10
2M
+
F39
BM
+
LC
D8
wee
ks
~
12 k
gN
RW
eig
ht c
hang
e fro
m b
asel
ine
to 8
wee
ksB
ryan
_BL
(107
)42
F34
BM
+
HB
D12
wee
ks7.
9 kg
~
9%W
eig
ht c
hang
e fro
m b
asel
ine
to 1
2 w
eeks
Font
aine
_BL
(206
)17
7M
+
F
~
42B
M
+
LC
D14
wee
ks
4
13.8
kg
11%
5
Wei
ght
cha
nge
from
bas
elin
e to
14
wee
ksFo
ntai
ne_B
L2 (
86)
109
M
+
F42
BM
+
LC
D15
wee
ks
6
13.9
kg
~
11%
7
Wei
ght
cha
nge
from
bas
elin
e to
15
wee
ksS
mith
_BL
(207
)54
F32
BM
+
BE
D15
wee
ks4.
5 kg
24%
Wei
ght
cha
nge
from
bas
elin
e to
15
wee
ksTe
ixei
ra_B
L (1
7)11
2F
31B
M
+
HB
D16
wee
ks5.
4 kg
21%
Wei
ght
cha
nge
from
bas
elin
e to
16
wee
ksW
ing
_BL
(61)
166
M
+
F31
BM
+
HB
D17
wee
ks
~
8 kg
10%
Wei
ght
cha
nge
from
bas
elin
e to
17
wee
ksTr
aver
so_B
L (8
7)50
M
+
F33
BM
+
HB
D24
wee
ks11
.3 k
gN
RW
eig
ht c
hang
e fro
m b
asel
ine
to 2
4 w
eeks
She
rwoo
d_B
L (7
1)44
4F
~
31B
M
+
HB
D26
wee
ks
~
8 kg
8
14%
Wei
ght
cha
nge
from
bas
elin
e to
26
wee
ksE
ldre
dg
e_B
L (1
24)
47M
+
F39
BM
+
BE
D36
wee
ks
~
3 kg
21%
Wei
ght
cha
nge
from
bas
elin
e to
36
wee
ks
9
Den
nis_
BL
(83)
109
F31
BM
+
HB
D39
wee
ks
~
9 kg
50%
Wei
ght
cha
nge
from
bas
elin
e to
39
wee
ksS
mith
_BL2
(20
7)54
F32
BM
+
BE
D15
wee
ks4.
5 kg
24%
41 w
eeks
+
0.2
kg30
%W
eig
ht c
hang
e fro
m b
asel
ine
to 4
1 w
eeks
Win
g_B
L2 (
61)
166
M
+
F31
BM
+
HB
D17
wee
ks
~
8 kg
10%
26 w
eeks
~+
1 kg
18%
Wei
ght
cha
nge
from
bas
elin
e to
43
wee
ksLi
nné_
BL
(208
)10
0M
+
F
~
41B
M
10
+
HB
D11
wee
ks
~
9 kg
NR
33 w
eeks
11
~+
1 kg
18%
Wei
ght
cha
nge
from
bas
elin
e to
44
wee
ks
12
Gla
dis
_BL
(50)
118
F36
BM
+
LC
D48
wee
ks
~
15 k
g17
%W
eig
ht c
hang
e fro
m b
asel
ine
to 4
8 w
eeks
, fo
r fo
ur g
roup
s b
ased
on
BE
D s
tatu
s at
bas
elin
eK
iern
an_B
L (4
9)17
7M
+
F
~
29B
M
+
HB
D52
wee
ksN
R14
%W
eig
ht c
hang
e fro
m b
asel
ine
to 5
2 w
eeks
(tw
o ca
teg
orie
s,
-
2 B
MI
units
as
crite
rion)
Sm
ith_B
LM (
207)
54F
32B
M
+
BE
D15
wee
ks4.
5 kg
24%
52 w
eeks
+
1.2
kg37
%W
eig
ht c
hang
e fro
m b
asel
ine
to 6
7 w
eeks
Teix
eira
_BLM
(64
)15
8F
31B
M
+
HB
D16
wee
ks5.
1 kg
14%
52 w
eeks
+
0.5
kg30
%W
eig
ht c
hang
e fro
m b
asel
ine
to 6
8 w
eeks
She
rwoo
d_B
LM (
71)
444
F
~
31B
M
+
HB
D26
wee
ks
~
8 kg
8
14%
52 w
eeks
11
~-
4 kg
8
21%
Wei
ght
cha
nge
from
bas
elin
e to
78
wee
ksLe
ibb
rand
_BLM
(12
0)13
8M
+
F45
BM
+
HB
D10
wee
ks
~
7 kg
NR
78 w
eeks
13
-
8.0
kg21
%W
eig
ht c
hang
e fro
m b
asel
ine
to 8
8 w
eeks
Gla
dis
_BLM
(50
)11
8F
36B
M
+
LC
D48
wee
ks
~
15 k
g17
%52
wee
ks
~+
7 kg
36%
Wei
ght
cha
nge
from
bas
elin
e to
100
wee
ks,
for
four
gro
ups
bas
ed o
n B
ED
sta
tus
at b
asel
ine
Jeffe
ry_B
LM (
93)
130
M
+
F31
BM
+
HB
D78
wee
ksN
RN
R52
wee
ks
~-
2 kg
20%
Wei
ght
cha
nge
from
bas
elin
e to
130
wee
ksN
ir_B
LM (
123)
66F
30B
M
+
HB
D10
wee
ks6.
6 kg
NR
74–2
04
wee
ks
+
3.1
kg30
%W
eig
ht c
hang
e fro
m b
asel
ine
to e
nd o
f fo
llow
-up
; d
urat
ion
of w
eig
ht s
tab
ility
per
iod
Pos
ton_
BM
(20
5)10
2M
+
F39
BM
+
LC
D8
wee
ks
~
12 k
gN
R52
wee
ks
~+
3 kg
NR
Wei
ght
cha
nge
dur
ing
fol
low
-up
(tw
o ca
teg
orie
s,
≥
1 kg
as
crite
rion)
Gla
dis
_BM
(50
)11
8F
36B
M
+
LC
D48
wee
ks
~
15 k
g17
%52
wee
ks~+
7 kg
36%
Wei
ght
cha
nge
dur
ing
fol
low
-up
, fo
r fo
ur g
roup
s b
ased
on
BE
D s
tatu
s at
bas
elin
eC
untz
_BM
(13
1)13
8M
+ F
~46
BM
+ H
BD
10 w
eeks
6.9
kgN
R78
wee
ks-1
.5 k
g13
%W
eig
ht c
hang
e d
urin
g f
ollo
w-u
p (
thre
e ca
teg
orie
s,±2
BM
I un
its a
s cr
iterio
n)(b
)Fo
gel
holm
_VL
(18)
85F
34B
M +
VLC
D12
wee
ks13
.5 k
g4%
40 w
eeks
14+0
.4 k
g6%
Wei
ght
cha
nge
from
bas
elin
e to
12
wee
ksR
aym
ond
_VL
(209
)17
4F
~36
BM
+ V
LCD
24 w
eeks
~17
kg12
%W
eig
ht c
hang
e fro
m b
asel
ine
to 2
4 w
eeks
Sm
ith_V
L (2
10)
289
F35
BM
+ V
LCD
15~2
5 w
eeks
~11
kg45
%W
eig
ht c
hang
e fro
m b
asel
ine
to 2
0–30
wee
ksFo
ster
_VL
(75)
223
F37
BM
+ V
LCD
26 w
eeks
16.7
kg
NR
Wei
ght
cha
nge
from
bas
elin
e to
26
wee
ks
obesity reviews Psychosocial predictors of weight loss P. J. Teixeira et al. 47
© 2005 The International Association for the Study of Obesity. obesity reviews 6, 43–65
Will
iam
s_V
L (5
9)12
8M
+ F
41B
M +
VLC
D26
wee
ks~2
4 kg
8%W
eig
ht c
hang
e fro
m b
asel
ine
to 2
6 w
eeks
Car
gill
_VL
(211
)15
9M
+ F
~42
BM
+ V
LCD
1626
wee
ksN
RN
RW
eig
ht c
hang
e fro
m b
asel
ine
to 2
6 w
eeks
Dra
pki
n_V
L (5
8)93
M +
F17
37B
M +
VLC
D16
52 w
eeks
~12
kg15
%W
eig
ht c
hang
e fro
m b
asel
ine
to 5
2 w
eeks
Will
iam
s_V
LM (
59)
128
M +
F41
BM
+ V
LCD
26 w
eeks
~24
kg8%
~73
wee
ks~+
6.4
kg55
%W
eig
ht c
hang
e fro
m b
asel
ine
to 1
00 w
eeks
Pek
karin
en_V
LM (
62)
62M
+ F
36B
M +
VLC
D17
wee
ks14
.9 k
g5%
104
wee
ks+9
.1 k
g8%
Thre
e ca
teg
orie
s of
wei
ght
loss
fro
m b
asel
ine
to 1
02 w
eeks
(10
% lo
ss a
s cr
iterio
n18)
Wes
tert
erp
-Pla
teng
a_V
M(2
12)
57F
31B
M +
VLC
D8
wee
ks19
8.5
kgN
R52
wee
ks19
+6.1
kg
53%
Wei
ght
reg
ain
dur
ing
fol
low
-up
(t
hree
cat
egor
ies,
50%
reg
ain
as c
riter
ion20
)R
aym
ond
_VM
(20
9)17
4F
~36
BM
+ V
LCD
24 w
eeks
~17
kg12
%52
wee
ks~+
12 k
g27
%W
eig
ht c
hang
e d
urin
g f
ollo
w-u
pP
asm
an_V
M (
63)
67F
32B
M +
VLC
D8
wee
ks9.
7 kg
NR
60 w
eeks
+7.0
kg
NR
Wei
ght
cha
nge
dur
ing
fol
low
-up
Mod
el (
refe
renc
e)S
amp
leTr
eatm
ent,
wei
ght
loss
pha
seFo
llow
-up
/mai
nten
ance
pha
se
Out
com
e (d
epen
den
t va
riab
le)
pre
dic
ted
fro
m
bas
elin
e va
riab
les
Siz
e1G
end
erB
MI
Typ
eD
urat
ion
Wt.
loss
Attr
ition
Dur
atio
nW
t. ch
ang
e2A
ttriti
on3
In th
e fir
st c
olum
n: B
, beh
avio
ur m
odifi
catio
n p
lus
hyp
ocal
oric
bal
ance
d d
iet (
HB
D, e
nerg
y in
take
gen
eral
ly b
etw
een
1200
and
150
0 kc
al d
–1)
or lo
w-c
alor
ie d
iet (
LCD
, ene
rgy
inta
ke g
ener
ally
bet
wee
n 80
0 an
d 1
200
kcal
d–1
); V
, beh
avio
ur m
odifi
catio
n p
lus
VLC
D (
ener
gy
inta
ke g
ener
ally
<80
0 kc
al d
–1);
L, w
eig
ht lo
ss m
odel
; LM
, wei
ght
loss
and
mai
nten
ance
mod
el; M
, wei
ght
mai
nten
ance
mod
el (
see
text
for
det
ails
). I
n ot
her
colu
mns
: M
, m
ales
; F,
fem
ales
; B
MI,
mea
n b
ody
mas
s in
dex
(kg
m–2
) at
bas
elin
e; B
ED
, tre
atm
ent
for
bin
ge
eatin
g d
isor
der
; N
R,
not
rep
orte
d (
whe
n at
triti
on r
ates
wer
e no
t re
por
ted
, d
ata
wer
e of
ten
rep
orte
d f
or c
omp
lete
rs o
nly)
. W
hene
ver
abse
nt,
initi
al B
MI
or w
eig
ht c
hang
es w
ere
estim
ated
fro
m a
vaila
ble
dat
a (e
.g.
initi
al w
eig
ht,
BM
I ch
ang
es)
usin
g a
ref
eren
ce h
eig
ht o
f 1.
70 m
. S
ome
stud
ies
wer
e lis
ted
mor
e th
an o
nce,
ind
icat
ing
mul
tiple
pre
dic
ting
mod
els.
1 In m
ost
case
s, it
rep
rese
nts
initi
al s
amp
le s
ize.
In
som
e ca
ses,
whe
n at
triti
on w
as n
ot r
epor
ted
, it
rep
rese
nts
com
ple
ters
onl
y.2 In
dic
ates
wei
ght
cha
nges
for
sub
ject
s th
at c
omp
lete
d t
he m
aint
enan
ce/fo
llow
-up
per
iod
s. B
ecau
se o
f at
triti
on d
urin
g t
his
per
iod
, th
ese
sub
ject
s ar
e no
t th
e sa
me
ind
ivid
uals
as
(i.e.
the
y ar
e a
sub
set
of)
the
gro
up t
hat
und
erw
ent
and
com
ple
ted
the
tre
atm
ent
pha
se.
3 Ind
icat
es o
vera
ll at
triti
on r
ate
(i.e.
incl
udin
g t
reat
men
t an
d f
ollo
w-u
p p
erio
ds)
.4 R
epre
sent
s m
ean
treat
men
t d
urat
ion.
Ind
ivid
ualiz
ed t
reat
men
ts la
sted
bet
wee
n 0
and
33
wee
ks.
5 Per
cent
age
of p
artic
ipan
ts w
ho f
aile
d t
o co
mp
lete
at
leas
t 6
wee
ks o
f tre
atm
ent.
Ove
rall
attr
ition
was
not
rep
orte
d.
6 Rep
rese
nts
mea
n tre
atm
ent
dur
atio
n. I
ndiv
idua
lized
tre
atm
ents
last
ed b
etw
een
0 an
d 3
4 w
eeks
.7 E
stim
ated
attr
ition
rat
e, b
ased
on
stud
y b
y Fo
ntai
ne &
Che
skin
(20
6).
8 Wei
ght
cha
nges
wer
e ca
lcul
ated
by
aver
agin
g d
ata
from
11
diff
eren
t in
terv
entio
n g
roup
s (w
ith s
amp
le s
izes
wei
ghe
d)
and
thu
s re
pre
sent
a r
oug
h es
timat
e of
ove
rall
outc
omes
.9 A
naly
ses
at 1
2 an
d 2
4 w
eeks
rev
eale
d a
sim
ilar
pat
tern
as
at 3
6 w
eeks
.10
Die
tary
tre
atm
ent
not
wel
l des
crib
ed;
the
abse
nce
of V
LCD
is a
ssum
ed.
11In
clud
ed a
wei
ght
mai
nten
ance
pro
gra
mm
e w
ith m
onth
ly g
roup
mee
ting
s.12
Rep
rese
nts
mea
n tre
atm
ent
dur
atio
n. I
ndiv
idua
lized
tre
atm
ents
last
ed b
etw
een
6 an
d 1
21 w
eeks
(to
tal d
urat
ion)
.13
Ind
icat
es o
vera
ll at
triti
on r
ate
(i.e.
incl
udin
g t
reat
men
t an
d f
ollo
w-u
p p
erio
ds)
.14
Incl
uded
a w
eig
ht m
aint
enan
ce p
rog
ram
me.
15O
nly
a su
bg
roup
of
all p
artic
ipan
ts (
51%
) re
ceiv
ed V
LCD
tre
atm
ent
(30
wee
ks),
the
rem
aini
ng r
ecei
ving
a lo
w-c
alor
ie d
iet
(LC
D)
last
ing
20
wee
ks.
16O
nly
a su
bg
roup
of
all p
artic
ipan
ts r
ecei
ved
VLC
D t
reat
men
t (n
umb
er o
f su
bje
cts
not
rep
orte
d,
assu
med
to
be
app
roxi
mat
ely
half)
, th
e re
mai
ning
rec
eivi
ng a
HB
D.
17A
ll p
artic
ipan
ts h
ad t
ype
II d
iab
etes
.18
Mai
nten
ance
of
at le
ast
10%
wei
ght
lost
fro
m b
asel
ine
was
the
crit
erio
n fo
r su
cces
s an
d w
eig
ht g
ain
from
bas
elin
e w
as t
he c
riter
ion
for
non-
succ
ess.
19S
tud
y in
clud
ed t
wo
cons
ecut
ive
per
iod
s of
VLC
D (
8 w
eeks
) an
d f
ollo
w-u
p (
52 w
eeks
); w
eig
ht c
hang
es d
urin
g t
reat
men
t an
d f
ollo
w-u
p w
ere
sim
ilar
in b
oth
per
iod
s an
d w
ere
aver
aged
in t
his
tab
le20
Twic
e a
reg
ain
<50%
of
wei
ght
lost
was
the
crit
erio
n fo
r su
cces
s an
d t
wic
e a
reg
ain
>50%
was
the
crit
erio
n fo
r no
n-su
cces
s.
Tab
le 1
Con
tinue
d
48 Psychosocial predictors of weight loss P. J. Teixeira et al. obesity reviews
© 2005 The International Association for the Study of Obesity. obesity reviews 6, 43–65
6% to 55% but they were not reported or were incom-pletely reported in many cases. The most typical outcomevariable was weight loss, expressed either in absolute (kg)or relative (% of initial weight) terms. In five cases, thedependent measure was a categorical variable, based onspecific weight loss/maintenance criteria (e.g. reaching achange of 2 BMI units).
Table 2 shows all predictive models, organized by pre-dictor (left column) and separated by the type and directionof the association observed. Independent variables aresorted by the overall number of predictive models available(i.e. variables with more evidence on top). Models areidentified by the name of first author in the source article,and by one to two letters identifying the duration/design ofthe predictive model (see legend for Table 1 for additionaldetails). For some models we also include the value corre-sponding to the per cent variance in weight loss accountedby the predictor (R2 ¥ 100), based on a bivariate correla-tion. These values, which unfortunately were not availablein all studies, can generally be compared across predictivemodels.
Results and discussion
An expert panel for the Institute of Medicine (IOM) haspreviously identified several pre-treatment individual pre-dictors of weight loss, gathering information from severalearlier reviews, book chapters and original research studies(8). In the 1995 peer-reviewed IOM report1, variables withevidence suggesting the existence of a relationship predic-tive of weight loss were initial weight and eating self-efficacy (positive association with weight loss) and repeatedweight loss attempts and perceived stress (negative associ-ation). Other variables such as binge eating, restricted eat-ing, psychopathology and personality factors were listed asnon-predictors.
Initial weight/BMI2
Sixteen studies reviewed in this article addressed the asso-ciation between initial weight or BMI and change inweight/BMI after treatment. Six studies reported a positiveassociation (i.e. higher initial BMI, larger weight losses)and two studies reported a negative association, with the
eight remaining studies showing no association. Two addi-tional studies appeared to show a significant associationbut the direction of the relationship could not be ascer-tained (58,59). Previous reviews have generally indicatedthat an higher initial weight is associated with higher abso-lute losses during treatment (8).
There was a slight tendency for longer duration studiesto show either no association or for subjects who wereinitially heavier to be less successful; of the 10 models witha weight maintenance phase, only two showed a positiveimpact of a higher initial BMI on long-term results. Acrossstudies, the sample’s overall degree of obesity seemed to beof influence. The weighted average for initial (mean) BMIfor models showing a positive association (i.e. heavier indi-viduals losing more weight) was ~37 kg m-2 while for allother models (negative or no association) the average initial(mean) BMI was ~31 kg m-2. Thus, a threshold of initialweight may be necessary to observe initial weight as asignificant predictor of results.
Previous dieting and repeated weight loss attempts
Overweight individuals participating in research-basedweight loss programmes may have different characteristicsthan the general overweight population, some of whichmay make the former group more resistant to successfulweight management (34,60). Previous reviews, includingthe IOM report previously cited, have concluded that whilea history of weight cycling is not predictive of weightoutcomes, frequent prior participation in weight loss pro-grammes and a history of repeated attempts at weight lossare negative prospective correlates of weight loss (3,6,8).Specific mechanisms have not been identified and it cannotbe ruled out that a history of failed attempts may alsoreflect a physiological resistance to weight loss, whether itdeveloped through the years or it is the expression of aparticular genetic trait.
For the present review, five studies published since 1995were identified that reported on the association of previousweight loss attempts, participation in previous pro-grammes, or a history of repeated weight loss with weightchanges during treatment. Of these, two studies showed noassociation and the three remaining reports showed a neg-ative association between previous dieting and weight losssuccess. Of the two studies showing no association, oneasked for past participation in organized weight loss pro-grammes in a dichotomous yes/no fashion (61) while theother showed no effect of the number of ‘previous weightloss attempts’ on weight loss at the 1- and 2-year timepoints (62). Kiernan and colleagues observed that a historyof weight cycling was part of an unsuccessful profile forBMI change at 1 year (49), while Pasman et al. used a four-item Likert scale (never to always) to classify individualsregarding their self-reported frequency of previous dieting,
1Table 2 includes a reference to the evidence collected in the IOM review
(indicated by the letters ‘IOM’).2Considering the potentially very large number of studies reporting on the
association between baseline weight/BMI and change in weight/BMI, we
did not specifically searched for such studies. Instead, we report these
associations only when they were found in studies containing information
on other predictors also reported in this study.
obesity reviews Psychosocial predictors of weight loss P. J. Teixeira et al. 49
© 2005 The International Association for the Study of Obesity. obesity reviews 6, 43–65
Tab
le 2
Ind
ivid
ual p
red
icto
rs/c
orre
late
s of
suc
cess
in s
ubse
que
nt w
eig
ht lo
ss a
nd/o
r m
aint
enan
ce a
nd p
red
ictiv
e m
odel
s
Con
stru
ctN
umb
erof M
odel
s
No
asso
ciat
ion
Sig
nific
ant
asso
ciat
ion
Pos
itive
Neg
ativ
e
Bin
ge/
emot
iona
l eat
ing
19IO
M,
Car
gill
_VL
(211
), C
untz
_BM
(13
1),
Gla
dis
_BL
(50)
, G
lad
is_B
M (
50),
Gla
dis
_BLM
(50
), L
eib
bra
nd_B
LM (
120)
, P
ekka
rinen
_VLM
, 2.
0 (6
2),
Ray
mon
d_V
L (2
09),
Ray
mon
d_V
M (
209)
, S
herw
ood
_BL
(71)
,S
mith
_BL,
1.7
(20
7),
Sm
ith_B
L2,
1.7
(207
)1 , S
mith
_BLM
, 1.
7 (2
07),
Teix
eira
_BL,
0.3
(17
), T
eixe
ira_B
LM,
0.5
(64)
Bry
an_B
L, 1
1.6
(107
), G
lad
is_B
L (5
0)Fo
gel
holm
_VL
(18)
, S
herw
ood
_BLM
(71
)
Initi
al w
eig
ht/B
MI
16C
untz
_BM
(13
1),
Jeffe
ry_B
LM (
93),
Kie
rnan
_BL
(49)
, P
ekka
rinen
_VLM
(62)
, P
osto
n_B
M (
205)
, Te
ixei
ra_B
L (1
7),
Teix
eira
_BLM
, 2.
0 (6
4),
Wes
tert
erp
-Pla
teng
a_V
M (
212)
IOM
, Fo
ster
_VL
(75)
, G
lad
is_B
L (5
0),
Gla
dis
_BLM
(50
), L
eib
bra
nd_B
LM(1
20),
Tra
vers
o_B
L (8
7),
Win
g_B
L (6
1)
Nir_
BLM
(12
3),
Pas
man
_VM
, 5.
8 (6
3)
Dep
ress
ion/
moo
d,
psy
chop
atho
log
y10
IOM
, B
ryan
_BL
(107
), C
arg
ill_V
L (2
11),
Cun
tz_B
M (
131)
, G
lad
is_B
L (5
0),
Leib
bra
nd_B
LM (
120)
, P
osto
n_B
L (2
05),
Pos
ton_
BM
(20
5),
Teix
eira
_BL,
2.3
(17
), T
eixe
ira_B
LM,
1.7
(64)
Eld
red
ge_
BL
(124
)
Eat
ing
dis
inhi
biti
on,
exte
rnal
eat
ing
10B
ryan
_BL
(107
), C
untz
_BM
(13
1),
Fost
er_V
L (7
5),
Leib
bra
nd_B
LM (
120)
,P
ekka
rinen
_VLM
, 3.
2 (6
2),
Teix
eira
_BL,
0.4
(17
), T
eixe
ira_B
LM,
0.3
(64)
, Tr
aver
so_B
L (8
7),
Wes
tert
erp
-Pla
teng
a_V
M (
212)
Pas
man
_VM
, 6.
8 (6
3)
Bod
y im
age,
bod
y si
ze s
atis
fact
ion
9Le
ibb
rand
_BLM
(12
0),
Teix
eira
_BL,
3.2
(17
), T
eixe
ira_B
LM,
1.2
(64)
, Tr
aver
so_B
L (8
7)2
Kie
rnan
_BL
(49)
, Te
ixei
ra_B
L, 6
.8 (
17),
Te
ixei
ra_B
LM,
4.8
(64)
,Tr
aver
so_B
L (8
7)
Trav
erso
_BL
(87)
Per
ceiv
ed h
ung
er9
Cun
tz_B
M (
131)
, Fo
ster
_VL
(75)
, Le
ibb
rand
_BLM
(12
0),
Pek
karin
en_V
LM,
0.1
(62)
, Te
ixei
ra_B
L, 0
.5 (
17),
Tei
xeira
_BLM
,0.
2 (6
4),
Trav
erso
_BL
(87)
Fog
elho
lm_V
L (1
8)P
asm
an_V
M,
5.3
(63)
Cog
nitiv
e (e
atin
g)
rest
rain
t, ch
roni
c d
ietin
g9
IOM
, B
ryan
_BL
(107
), C
untz
_BM
(13
1),
Leib
bra
nd_B
LM (
120)
,P
ekka
rinen
_VLM
, 0.
2 (6
2),
Teix
eira
_BL,
0.5
(17
), T
eixe
ira_B
LM,
0.3
(64)
, Tr
aver
so_B
L (8
7),
Wes
tert
erp
-Pla
teng
a_V
M (
212)
Fost
er_V
L, 2
.3 (
75)
Eat
ing
sel
f-ef
ficac
y7
Car
gill
_VL
(211
), F
onta
ine_
BL2
, 0.
4 (8
6),
Sm
ith_B
LM (
207)
3 ,S
mith
_BL2
(20
7)3 ,
Teix
eira
_BL,
0.8
(17
),IO
M,
Sm
ith_B
L (2
07)3 ,
Teix
eira
_BLM
,4.
4 (6
4)P
revi
ous
die
ting
/wei
ght
lo
ss a
ttem
pts
6P
ekka
rinen
_VLM
(62
), W
ing
_BL
(61)
IOM
, K
iern
an_B
L (4
9),
Pas
man
_VM
, 21
.2 (
63),
Tei
xeira
_BL,
13.
7 (1
7),
Teix
eira
_BLM
, 5.
8 (6
4)
Per
sona
lity,
gen
eral
co
gni
tive
styl
e6
IOM
, B
ryan
_BL,
13.
7 (1
07)5 ,
Font
aine
_BL,
0.5
(20
6)4 ,
Trav
erso
_BL
(87)
6B
ryan
_BL
(107
)5 , P
osto
n_B
M (
205)
6 , Tr
aver
so_B
L (8
7)71
2
Sel
f-es
teem
5B
ryan
_BL
(107
), E
ldre
dg
e_B
L (1
24),
Tei
xeira
_BLM
, 1.
0 (6
4)N
ir_B
LM,
4.4
(123
)8 , Te
ixei
ra_B
L, 4
.4(1
7)In
tern
al lo
cus
of c
ontro
l95
Bry
an_B
L, 1
2.3
(107
)10,
Will
iam
s_V
L, 0
.3 (
59),
Will
iam
s_V
LM,
1.4
(59)
Bry
an_B
L (1
07)10
, N
ir_B
LM (
123)
Sel
f-m
otiv
atio
n,
auto
nom
y or
ient
atio
n,
gen
eral
effi
cacy
5W
illia
ms_
VL,
0.3
(59
)D
enni
s_B
L (8
3)11
, Te
ixei
ra_B
L, 7
.8(1
7),
Teix
eira
_BLM
, 2.
3 (6
4),
Will
iam
s_V
LM,
10.2
(59
)P
erce
ived
soc
ial s
upp
ort
4K
iern
an_B
L (4
9)12
, Te
ixei
ra_B
L, 0
.3 (
17),
Tei
xeira
_BLM
, 0.
6 (6
4),
Win
g_B
L2 (
61)13
Wei
ght
loss
g
oals
/exp
ecta
tions
, ou
tcom
es e
valu
atio
n
4Je
ffery
_BLM
(93
), L
inné
_BL
(208
)Te
ixei
ra_B
L, 7
.8 (
17),
Tei
xeira
_BLM
, 7.
3(6
4)
50 Psychosocial predictors of weight loss P. J. Teixeira et al. obesity reviews
© 2005 The International Association for the Study of Obesity. obesity reviews 6, 43–65
Cog
nitiv
e st
yle
reg
ard
ing
d
iet
lap
ses
2S
mith
_V3L
(21
0)14
Dra
pki
n_V
L (5
8)15
Exe
rcis
e se
lf-ef
ficac
y2
Teix
eira
_BL,
3.6
(17
), T
eixe
ira_B
LM,
6.3
(64)
Exe
rcis
e p
erce
ived
bar
riers
2Te
ixei
ra_B
L, 4
.4 (
17),
Tei
xeira
_BLM
, 4.
8(6
4)E
xerc
ise
soci
al s
upp
ort
2Te
ixei
ra_B
L, 2
.3 (
17),
Tei
xeira
_BLM
, 0.
4 (6
4)Q
ualit
y of
life
(ob
esity
-sp
ecifi
c)2
Teix
eira
_BL,
3.6
(17
), T
eixe
ira_B
LM,
2.3
(64)
Qua
lity
of li
fe (
gen
eral
, S
F-36
)2
Teix
eira
_BL,
2.9
(17
), T
eixe
ira_B
LM,
1.4
(64)
Per
ceiv
ed s
tress
, an
xiet
y1
Kie
rnan
_BL
(49)
Cog
nitiv
e p
erfo
rman
ce1
Bry
an_B
L (1
07)
Per
ceiv
ed a
uton
omy
(soc
ial)
sup
por
t1
Will
iam
s_V
LM (
59)
Bul
imic
beh
avio
ur1
Trav
erso
_BL
(87)
Con
stru
ctN
umb
erof M
odel
s
No
asso
ciat
ion
Sig
nific
ant
asso
ciat
ion
Pos
itive
Neg
ativ
e
IOM
, rel
atio
nshi
p re
por
ted
in a
com
pre
hens
ive
rep
ort b
y th
e In
stitu
te o
f Med
icin
e p
ublis
hed
in 1
995
(8).
See
leg
end
in T
able
1 fo
r mor
e d
etai
ls o
n le
tters
/num
ber
s us
ed to
iden
tify
pre
dic
tive
mod
els.
Whe
neve
r p
rese
nt,
the
num
ber
afte
r ea
ch m
odel
rep
rese
nts
the
biv
aria
te a
ssoc
iatio
n w
ith w
eig
ht c
hang
e, e
xpre
ssed
as
per
cen
t va
rianc
e ac
coun
ted
by
the
pre
dic
tor
varia
ble
(R
2 ¥ 1
00).
Whe
neve
r th
e sa
me
mod
el
is li
sted
in d
iffer
ent
colu
mns
with
in t
he s
ame
cons
truc
t, d
iffer
ent
scal
es w
ere
used
for
sim
ilar
cons
truc
ts (
see
num
ber
ed f
ootn
otes
).1 Th
e in
tera
ctio
n b
etw
een
bin
ge
eatin
g a
nd p
erce
ived
con
trol w
as a
sig
nific
ant
pre
dic
tor.
It ap
pea
rs t
hat
bin
ge
eatin
g lo
aded
neg
ativ
ely
and
con
trol p
ositi
vely
, re
gar
din
g w
eig
ht lo
ss.
2 Exc
ept t
he F
eelin
g F
at s
cale
, all
othe
r fiv
e d
imen
sion
s of
the
Bod
y A
ttitu
des
Que
stio
nnai
re (
113)
, nam
ely
Dis
par
agem
ent,
Fitn
ess
and
Stre
ngth
, Sal
ienc
e of
Sha
pe,
Attr
activ
enes
s, a
nd L
ower
Bod
y Fa
tnes
s,
wer
e no
t as
soci
ated
with
wei
ght
loss
.3 Q
uest
ions
rep
rese
nted
per
ceiv
ed b
ehav
iour
al c
ontro
l, as
the
‘lik
elih
ood
to
over
com
ing
sp
ecifi
c b
arrie
rs t
o lo
sing
wei
ght
and
eat
ing
in h
ealth
y m
anne
r’, w
hich
are
phr
ased
sim
ilarly
to
eatin
g s
elf-
effic
acy
que
stio
nnai
res.
4 Gen
eral
op
timis
m,
asse
ssed
by
the
revi
sed
Life
Orie
ntat
ion
Test
(21
3).
5 The
Imp
ress
ion
Man
agem
ent
and
Vul
nera
bili
ty s
cale
s of
the
Dys
func
tiona
l Atti
tud
es S
cale
(21
4) p
ositi
vely
pre
dic
ted
wei
ght
loss
; th
e re
mai
ning
sca
les
(Ap
pro
val,
Cat
astro
phi
sing
, Im
per
ativ
es,
Nee
d t
o S
ucce
ss,
and
Ple
asin
g O
ther
s) w
ere
not
sig
nific
ant
pre
dic
tors
.6 M
uscu
lar T
ensi
on (p
ositi
ve),
Mon
oton
y A
void
ance
(pos
itive
), S
usp
icio
n (n
egat
ive)
and
Gui
lt (n
egat
ive)
wer
e th
e on
ly s
igni
fican
t (‘m
odes
t’) p
red
icto
rs o
f wei
ght
loss
, fro
m a
ll 15
sca
les
pre
sent
in th
e K
arol
insk
a S
cale
s of
Per
sona
lity
(215
)7 Th
e B
ody
Dis
satis
fact
ion
and
Int
erp
erso
nal D
istr
ust
dim
ensi
ons
of t
he E
atin
g D
isor
der
s In
vent
ory
(112
) w
ere
the
only
sca
les
asso
ciat
ed w
ith w
eig
ht lo
ss;
all o
ther
six
sca
les
(Bul
imia
, D
rive
for
Thin
ness
, In
tero
cep
tive
Aw
aren
ess,
Ine
ffect
iven
ess,
Mat
urity
Fea
rs,
and
Per
fect
ioni
sm)
wer
e no
t.8 B
asel
ine
self-
este
em w
as a
pos
itive
pre
dic
tor
of m
onth
s of
wei
ght
sta
bili
ty,
not
over
all w
eig
ht lo
ss.
9 Alth
oug
h lo
cus
of c
ontro
l was
not
men
tione
d in
the
IO
M r
evie
w,
an in
tern
al lo
cus
of c
ontro
l has
bee
n id
entifi
ed in
oth
er r
evie
ws
as a
pos
itive
and
sig
nific
ant
pre
dic
tor
of w
eig
ht lo
ss (
5,7)
.10
The
Die
ting
Bel
iefs
Sca
le (
104)
was
rel
ated
with
wei
ght
loss
whi
le t
he W
eig
ht L
ocus
of
Con
trol S
cale
(10
5) w
as n
ot.
11Ite
ms
in t
his
que
stio
nnai
re s
tart
ed w
ith ‘W
hen
it co
mes
to
my
wei
ght
. . .
’, b
ut w
ere
phr
ased
in a
gen
eral
sen
se,
rep
rese
ntin
g a
n ov
eral
l sen
se o
f co
nfid
ence
and
sel
f-m
otiv
atio
n (e
.g.
‘I’m
con
fiden
t in
my
abili
ty t
o re
ach
my
goa
l’, ‘I
mak
e d
ecis
ions
as
wel
l as
anyo
ne e
lse’
) an
d d
id n
ot in
clud
e re
fere
nce
to a
ny s
pec
ific
wei
ght
-con
trol b
ehav
iour
suc
h as
eat
ing
or
exer
cise
.12
Did
not
ent
er p
red
ictio
n m
odel
usi
ng s
igna
l det
ectio
n m
etho
dol
ogy
and
thu
s as
sum
ed a
s no
n-si
gni
fican
t p
red
icto
r.13
In t
hree
of
the
four
exp
erim
enta
l gro
ups
in s
tud
y; in
fou
rth
gro
up,
a m
odes
t ne
gat
ive
asso
ciat
ion
was
ob
serv
ed.
14A
ttitu
des
reg
ard
ing
die
t la
pse
s w
ith t
hree
gro
ups
wer
e d
efine
d:
obje
ctiv
e, q
uant
itativ
e th
inke
rs (
‘mat
ter-o
f-d
egre
e’),
rat
iona
l thi
nker
s an
d r
igid
, d
icho
tom
ous
thin
kers
(‘a
ll-or
-not
hing
’).15
The
per
ceiv
ed n
umb
er o
f hi
gh-
risk
situ
atio
ns (
for
a d
ieta
ry la
pse
) in
whi
ch a
par
ticip
ant
gen
erat
ed a
cop
ing
res
pon
se w
as a
pos
itive
pre
dic
tor.
Tab
le 2
Con
tinue
d
obesity reviews Psychosocial predictors of weight loss P. J. Teixeira et al. 51
© 2005 The International Association for the Study of Obesity. obesity reviews 6, 43–65
showing that the higher the score (i.e. the more frequentthe previous dieting) the larger the weight regain during14 months after treatment was over (63). Finally, tworelated studies reported a negative association between thenumber of diet attempts during the 12 months prior totreatment and weight change at 4 months (17) and at a 16-month follow-up (64). Four to five attempts in the yearprior to treatment was a good cut-off point, with very fewsuccessful participants reporting such a high level of previ-ous dieting (64). Thus, most results published before(46,65–67) and after 1995 suggest that previous participa-tion in weight loss programmes and previous dietingattempts are predictive of worse weight loss outcomes dur-ing and after treatment for obesity.
Binge, restricted and disinhibited eating
Binge eating and (cognitive) eating restraint and disinhibi-tion are among the most studied variables in the contextof weight loss and management. In this review, with onlya few exceptions, data show little or no associationbetween weight changes during treatment and baselinebinge eating [most frequently assessed by the Binge EatingScale (68)], or variables derived from the popular EatingInventory (69). Perhaps the findings regarding binge/disor-dered eating, high levels of which are commonly reportedby many weight loss candidates (70), are the most counter-intuitive. Emotional overeating (and cognitive disinhibitionof eating), especially in the presence of binge eating epi-sodes, is commonly viewed as a negative factor in weightcontrol and is associated with more perceived barriers toweight loss (71). Two factors may account for the non-significant relationships observed. First is the high hetero-geneity within any of the particular eating characteristicsmeasured, regarding other relevant variables that maywork as confounders or mediators. Restrained eating, forinstance, has been found to be present in flexible vs. rigidpatterns in groups presenting for treatment (72), with dis-tinct implications for long-term weight loss (more flexiblecontrol, better results). Restrictive dieting may exacerbatea tendency to counter-regulatory overeating, but only insusceptible overweight individuals (73). Second, normaliza-tion of eating behaviours, for example, by addressing emo-tional triggers to eating or improving cognitive control overexternal situations likely to induce overeating, is a centralintervention goal in the majority of lifestyle treatment pro-grammes [e.g. (74)]. To the extent improvements in thisarea do occur during treatment (75), and assuming thatsubjects with the worst initial values will improve the most,then pre-treatment scores may not carry a strong influenceon outcomes; an equalizing phenomenon concerning disor-dered eating may take place during the programme,rendering baseline scores largely irrelevant for overallresults.
Some evidence indicates that individuals with the highestbinge eating scores may drop out at a higher rate than non-binge eaters (71,75–77). Differences in the assessment ofbinge eating notwithstanding, this suggests that some bingeeaters who cannot adhere to the new eating patterns pro-posed by the behavioural interventions may leave the pro-gramme before its completion, whereas the remainingindividuals within this subgroup are on average as success-ful as their non-binging counterparts. In contrast, otherstudies have shown similar attrition rates between subjectswith binge eating disorder (clinical or subclinical) and with-out this disorder, using different measures to characterizebinge eating patterns (50).
Self-efficacy and self-motivation
Eating self-efficacy has been frequently assessed with theEating Self-Efficacy Scale [ESES (78)] and the Weight Effi-cacy Life-Style Questionnaire [WEL (79)]. Studies previousto 1995 (46,80,81) have led to the general conclusion thathigh self-efficacy towards eating behaviours is perhaps theonly reliable predictor of subsequent weight loss (39).However, the collective evidence since 1995 does notentirely confirm the earlier findings (see Table 2). It is note-worthy that in the original validation study for one of themore popular self-efficacy questionnaires (the ESES), Glynn& Ruderman had already shown that eating self-efficacy,whether assessed at baseline or at any other time point, wasnot predictive of subsequent weight loss (78).
Confounding factors (e.g. sample characteristics, inter-vention features) and the relatively small number of studieswith comparable methodologies limit general conclusions.For example, the extent to which treatment is successful inactually increasing participants’ self-efficacy during theprogramme could reduce the impact of pre-treatment self-efficacy on weight loss. However, changes in both eatingand exercise self-efficacy during treatment are consistentlyassociated with weight reduction (65,78,82,83). It is pos-sible that earlier reviews have considered eating self-effi-cacy more generally (e.g. as a mediator of success; changesin self-efficacy predicting changes in weight) rather thanexclusively as a pre-treatment predictor.
When addressing self-efficacy, it is important to distin-guish between self-regulatory efficacy specific to a particu-lar situation or behaviour (e.g. resisting overeating whenfeeling depressed) and more general constructs such as anoverall sense of assurance or self-confidence. Although ahigh level of generalized efficacy and autonomy could pos-itively influence situation-specific self-efficacy, these con-structs are not interchangeable (47). Interestingly, severalstudies in the present review seem to indicate that general-ized measures of efficacy may be more predictive of out-comes (17,59,64,83–85) than scales that target perceivedself-efficacy for specific behaviours, especially eating-
52 Psychosocial predictors of weight loss P. J. Teixeira et al. obesity reviews
© 2005 The International Association for the Study of Obesity. obesity reviews 6, 43–65
related (17,65,78,86,87). Although self-efficacy theoryendorses the importance of task specificity for the predic-tive ability of self-efficacy measures (88), self-motivation,often assessed by the Self-motivation Inventory [SMI (89)],and an autonomous orientation to one’s motivation, bothof which are general attributes, have shown consistentlypositive results as predictors of subsequent success inweight control (17,59,64,84). In the case of the SMI, it hasalso been shown to correlate with eating-related variablesduring weight loss (90) and to significantly predict subse-quent exercise behaviour (89).
A more in-depth discussion of the mixed findings foreating self-efficacy as a predictor in weight management isbeyond the scope of this article. Nevertheless, it is possiblethat the true details about the eating behaviours in question(e.g. what it really takes to consistently limit intake ofcertain foods or eat less in certain situations) are largelyunknown to participants at the time of initial assessments,leading to wrongly formed efficacy expectations; thesewould weaken the predictive ability of the eating-relatedmeasures in question (80). It may also be that the specificeating behaviours covered by the self-efficacy items ques-tionnaires (e.g. cognitive restriction of eating in a varietyof situations) are in themselves not strongly related toweight control, particularly in the long term. Conversely,more general measures (e.g. self-motivation) may be suffi-ciently broad to cover the various dimensions involved inweight control (behavioural, environmental, psychological)and well-suited to predict compliance to the multitude ofbehaviours actually necessary for long-term success.
Outcome expectancies
Before a weight loss programme, participants are sometimesasked ‘how much weight do you expect to lose?’ (47) or‘how likely are you to reach your goal weight?’ (91). Otherquestions have also been used in this context, namely ‘whatwould an acceptable (or unacceptable, or “happy”) weightbe for you, by the end of this program?’ (92) or ‘how muchwould you like to weigh?’ (93). Although these relatedquestions refer to somewhat different types of expectations/evaluations, they are henceforth addressed together.
Studying outcome expectancies as a predictor of weightoutcomes is particularly relevant in the context of thedebate concerning the role of realistic (vs. unrealistic)expectations for weight loss prior to treatment (94–97).Excessively optimistic expectations are the norm in obesitytreatment-seeking individuals, at least in American samples(92) and a great value is placed on reaching desired weights(98). Jeffery and colleagues showed that men and womenwith more modest absolute weight loss goals were morelikely to achieve their goals, and that those who achievedtheir weight goals had better weight maintenance 2.5 yearsafter beginning the trial (93); however, desired weight loss
did not directly predict actual weight loss. Using the Goalsand Relative Weights Questionnaire (92) it was recentlyobserved that overweight/obese women with less demand-ing pre-treatment evaluations of what ‘acceptable’ or‘happy’ weights would be after treatment (relative to initialweight) lost significantly more weight than women indicat-ing they would only find larger weight losses as acceptable/‘happy’ ones (17,64). Others have shown that positiveexpectations (i.e. higher reported likelihood of reachinggoal weight) predict short-term weight loss, especiallywhen subjects reveal a low level of fantasizing and day-dreaming about potential beneficial consequences of largeweight loss; in the long-run, positive initial expectations,low/negative fantasy thoughts and the interaction of thetwo were all predictive of weight loss (91). Still otherstudies have shown positive outcome expectations (i.e.larger weight loss goals) to positively precede better weightreduction results (47,85). More research is needed to clarifythe impact of outcome expectancies/evaluations on andduring weight reduction efforts. At this moment, it appearsthat positive and realistic expectations foretell the bestresults, especially if accompanied by a strong sense of self-assurance (85).
Locus of control
Locus of control refers to the degree to which peoplebelieve their own behaviours determine the outcomes oftheir lives (internal locus of control), as opposed to chanceor the impact of other people and external events (externallocus of control). Several instruments are available to mea-sure this construct: general, non-specific measures such asthe popular Internal-External scale [I-E (99)] or adapta-tions of it (100,101), and more specific scales, such as theHealth Locus of Control scale [HLC (102)], the Multidi-mensional Health Locus of Control scale [MHLC (103)],the Dieting Beliefs Scale [DBS (104)] and the Weight Locusof Control Scale [WLOC (105)]. All have been used topredict weight loss in the past.
Despite being one of the most frequently studied vari-ables among predictors of weight loss, especially in the1970–80s, locus of control was not specifically mentionedin some of the previous reviews on this topic (3,8), possiblybecause it was included in the category of personality fac-tors and classified as a non-predictor (8). Nevertheless,locus of control has a good theoretical rationale (99) andacceptable empirical support as a correlate of weight loss.In a ‘mini meta-analysis’ of locus of control as a predictorof weight change, Allison & Engel reviewed 11 studiespublished prior to 1995, concluding that locus of controlwas significantly related to weight loss (mean for all stud-ies, adjusted for sample size, was 0.19) and that health- andweight-specific locus of control appeared to be more pre-dictive than more general measures (7).
obesity reviews Psychosocial predictors of weight loss P. J. Teixeira et al. 53
© 2005 The International Association for the Study of Obesity. obesity reviews 6, 43–65
We identified three more recent studies that have usedlocus of control measures to predict weight loss results, oneof which used two different scales for that purpose, thusyielding a total of four predictive models (see Table 2). Twomodels showing non-significant associations used theMHLC and the WLOC scale, a four-item instrument devel-oped specifically for weight control, whereas the DBS andan I-E general scale resulted in significant predictive mod-els. Given the heterogeneity of programme characteristicsand the diversity of scales utilized, it is not surprising thatconflicting results have been reported. Nevertheless, a pre-vious review (7) and results from some of the more recenttrials indicate that an internal locus of control is a beneficialtrait regarding weight management; not one study to datehas shown external (as opposed to internal) beliefs to bepredictive of better results. Additionally, the importance orthe value that individuals place on the expected outcomesmay be a relevant concurrent assessment with locus ofcontrol, as its effect on behaviour (i.e. the effect of locusof control) may be mediated by value cognitions (99,105).Whether general or specific measures result in closer rela-tionships with weight outcomes is not clear, especially con-sidering the very few published studies that have usedweight-specific locus of control measures. The DBS in par-ticular is a promising tool (104,106,107), deserving furthervalidation and investigation as a predictor of weight loss.Additionally, exercise-specific locus of control is also in itsexploration phase (108,109) and could be used in weightmanagement in the future.
Body image and self-esteem
Whether low starting levels of self- and body-esteem limitor ease weight management efforts has been investigatedbefore, with mixed findings. While general self-esteem is arelatively well-defined construct, most frequently measuredwith the Rosenberg Self-esteem/Self-concept scale [RSE(110)], body image is considered a multidimensional trait,involving cognitive/attitudinal, perceptual and behaviouraldimensions (111). Five recent studies have examined asso-ciations of initial body image variables and subsequentweight loss. In some cases, multiple measures were usedwithin the same study (17,64,87). Six psychometric instru-ments were used to assess body image, namely the EatingDisorders Inventory [EDI (112)], the Body Attitude Ques-tionnaire [BAQ (113)], the Body Shape Questionnaire[BSQ (114,115)], the Physical Self-Perception Scale [PSPP(116)], the Body Cathexis questionnaire (117) and theBody Image Assessment Questionnaire (118). As could beexpected from such a broad measurement setting, resultswere mixed, with some predictive models showing no asso-ciation, some showing a positive association (better bodyimage, more weight loss) and one model showing a nega-tive relationship (see Table 2).
Several studies show that the BSQ, one of the mostpsychometrically sound instruments for the assessment ofdistress related to body image (119), does not significantlypredict subsequent weight loss (17,64,87,120). In contrast,significant relationships were found for the EDI’s BodyDissatisfaction scale (49,87) for the Body Image Assess-ment Questionnaire (difference between self and idealsilhouettes representing a range of body shape/sizes) andfor the Body Cathexis questionnaire (17,64). For example,Kiernan and colleagues used a novel statistical procedure(signal detection methodology) to identify baseline charac-teristics of successful vs. unsuccessful weight losers after1 year, and found lower body dissatisfaction (from the EDI)and no history of repeated weight loss to significantly pre-dict success (49).
The diversity of assessments for this construct (in fact,the ‘diversity’ of the construct itself) makes a summaryconclusion difficult to reach. Collectively, it appears thatelevated levels of body distress do hinder attempts toweight loss in some cases, although the degree to whichthese issues are addressed during treatment, as well as otherfactors, may confound this association. Some studies havealso shown that age of onset on obesity and depression mayaffect the impact of negative body image on obesity(121,122).
Regarding self-esteem, recent evidence is scarce. A rela-tionship with subsequent weight loss, which was significantin the short term (17) but not in the long term (64), hasbeen reported. Nir & Neumann found baseline self-esteemto be correlated with the duration of weight loss mainte-nance but not with weight loss per se (123). Two otherstudies have found non-significant associations (107,124).While previous review papers have not highlighted a sta-tistically significant role of self-esteem as a predictor ofweight loss (4,8), more current data and some previousstudies (125) lend support, albeit modest, to a positiveinfluence of self-esteem in weight management outcomes.
Psychological health and perceived stress
In virtually all studies analysed for this review measuresof psychological well-being or psychopathology (mood,depression, personality disorders, etc.) were not found tosignificantly predict weight loss. The Beck DepressionInventory [BDI (126,127)] is by far the most commonlyused measure in this domain and it is safe to assert that itdoes not adequately identify subjects with low likelihoodof success for weight management. The adoption of exer-cise (128), improvements in body image (e.g. followingsome initial weight loss) (13) and the regular social contactwith a supportive group and intervention team (61) aresome factors likely to improve mood and psychologicaladjustment during weight loss. As mentioned for othervariables (e.g. self-efficacy, binge eating), improvements in
54 Psychosocial predictors of weight loss P. J. Teixeira et al. obesity reviews
© 2005 The International Association for the Study of Obesity. obesity reviews 6, 43–65
depressive symptoms during weight loss treatment occur(50,129) and tend to covary with weight changes(130,131). Consequently, baseline assessments may lack animportant or lasting impact.
Considering the face-validity of high levels of stress andanxiety as limiting factors in obesity behavioural treatment(which has lead to the inclusion of stress reduction compo-nents into treatment programmes), it was surprising to findonly one study published since 1995 evaluating the impactof pre-treatment levels of perceived stress on outcomes(49). One logical explanation may be that stress is lookedupon as a transient state and considered primarily as aprocess factor (3); its influence assumed the strongest assubjects go through weight loss, not as a relevant pre-treatment aspect (when optimism is generally the highest).Trait anxiety is a more stable construct but it has largelybeen ignored in the context of weight loss predictors. Inthe past (132), pre-treatment anxiety (and depression) hasyielded some significant predictive results, using the Min-nesota Multiphasic Personality Inventory [MMPI (133)]and the Taylor Manifest Anxiety Scale (134), but theseresults have not since been confirmed or refuted.
Our interpretation of the literature on psychologicalhealth and subsequent weight loss is that mood and depres-sion, as well as general indicators of psychopathology, aregenerally not appropriate measures to predict subsequentweight loss, at least considering heterogeneous clinical sam-ples; some subgroups may be particularly vulnerable (135).This said, many studies initially screen out subjects withclinical depression or clinically diagnosable psychopathol-ogy, which may render some analyses inappropriatebecause of lack of variance in the variable of interest.Regarding state and especially trait anxiety and stress, noinformed conclusion can be made from the available evi-dence. Cross-sectionally, higher levels of stress have beenshown to predict unhealthier behaviours that can contrib-ute to obesity such as a higher fat diet and less exercise(136). More studies are encouraged, especially regardingthe most stable traits, which may impact behaviours in thelong run.
Social support
Despite its apparent face-validity and some suggestiveresults (137,138), we found no evidence that social supportmeasured before treatment is predictive of subsequentweight outcomes. Williams and colleagues observed thatthe perceived type of support (provided by the interventionteam, not by family/peers as more commonly assessed)measured 5–10 weeks into the programme was importantfor later success (59); when support was viewed as fosteringautonomy as opposed to being controlling, participantsobtained significantly better results. This distinctionbetween types of support may be important for other types
of social influences, such as those arising from close familyand significant others. It has been our experience thatspouses’ attitudes (and their behaviours) towards theirpartners’ treatment process can be very influential but arelargely unpredictable. Many women we have worked withhave reported surprise by their spouse’s reactions (positiveor negative) during the programme. This suggests thatbaseline assessments of social support, as opposed to mea-surements during or after treatment, may be assessingsupport too generally or measuring wrong estimations offuture support. A questionnaire to measure social supportspecifically for weight management has been recently devel-oped (139) but no data for its association with weight lossare available.
Quality of life
As with social support, health-related quality of life(HRQL) variables have been analysed as pre-treatment pre-dictors of weight management in only a limited number ofstudies. HRQL measures psychological, physical and inter-individual functioning in daily life and can be divided intogeneral (context-unspecific) measures and those that targetquality of life associated with a particular condition. Sev-eral instruments are available to measure both general (e.g.SF-36) and specific HRQL, including various obesity-specific instruments (140,141). We are aware of only twostudies that have measured HRQL before obesity treatmentand reported associations between these scores and subse-quent weight outcomes (17,64). Results from these relatedstudies suggest that the Impact of Weight on Quality ofLife-lite (IWQOL-lite) may have potential as a predictor ofweight loss and maintenance, and particularly as a corre-late of early drop-out. Women reporting lower HRQL atprogramme start, particularly in the areas of work, healthand self-esteem, were more likely to finish the study amongthe least successful group or to drop out. These results wereessentially the same when predicting short- and longer-termweight loss (17,64).
Exercise-related variables
The adoption of physical activity is regarded as criticallyimportant in the process of weight management, particu-larly for long-term success (142). Besides its direct contri-bution to energy expenditure during and after the activitybout, exercise may also contribute to improved dietarycompliance (143), possibly through decreased feelings ofhunger and eating disinhibition and improved psychologi-cal well-being (144). However, research on exercise-relatedpsychosocial variables in the context of obesity treatmentis considerably scarcer when compared to other areas suchas exercise and eating behaviours, body image, or psycho-logical health. In non-obese samples, exercise participation
obesity reviews Psychosocial predictors of weight loss P. J. Teixeira et al. 55
© 2005 The International Association for the Study of Obesity. obesity reviews 6, 43–65
can be predicted by several questionnaire-based variablessuch as self-efficacy, perceived barriers, social support, per-ceived stress and feelings of enjoyment and competence inphysical activities (145–147). In fact, motivational modelsfor exercise intention and behaviour have been developedto some detail and with relative success (148), and psycho-logical correlates of exercise behaviours have been easierto establish than comparable constructs for eating behav-iours [cf. (149–152)].
Despite the sound rationale, whether exercise psychoso-cial pre-treatment variables are able to predict exercisebehaviour or actual weight loss during obesity treatmenthas been investigated in only a few studies (17,64). Theserecent studies have analysed scores from the Exercise Self-efficacy Scale [ESES (153)], the Exercise Perceived BarriersScale [EPBS (154)] and the Exercise Social Support Scale[ESSS (153)] as predictors of weight loss. Higher perceivedself-efficacy for exercise, measuring an individual’s convic-tion that he/she can ‘stick with’ an exercise programme forat least 6 months under varying circumstances (e.g. whentime is short, when undergoing a major life change, etc.)and lower perceived barriers to exercise were significantlyassociated with short- and long-term weight loss results(17,64). Baseline social support for exercise behaviours didnot predict subsequent weight loss.
The questionnaires used for exercise psychosocialvariables are relatively short, have generally displayedgood psychometric properties, are simple to interpretand can easily be used in the context of obesity treat-ment at baseline, concurrently with the intervention(e.g. to monitor exercise attitudes throughout the pro-gramme), or as retrospective correlates of weight lossand exercise behaviours (44). More research is needed
to analyse attitudes and cognitions regarding exerciseand physical activity behaviours, in the context of obe-sity treatment.
Summary of predictors of weight control
Table 3 shows all predictors, organized based on thestrength and consistency of their association with weightoutcomes and also considering the relative amount of evi-dence available for each case (i.e. number of publishedstudies). Less previous dieting or little history of repeatedweight loss attempts, low self-reported self-motivation (orgeneral efficacy) and autonomy were the most consistentpredictors of subsequent weight loss. There is clearly moreevidence for previous dieting than for the latter variables,but the consistency of recent findings (all positive and sig-nificant) and a sound theoretical rationale led us to classifya self-motivated, autonomous style among the strongestpredictive traits. The SMI in particular, for general motiva-tion/efficacy attributions, and instruments that assess cau-sality orientations [v (59)] deserve more research regardingeating and physical activity adherence, and overall successrates in weight loss. A number of other variables, includingpoor body image, low eating self-efficacy and unrealisticweight loss expectations may correlate negatively withsubsequent weight loss, though not as consistently. A highinitial BMI may be associated with more absolute weightloss but only in samples including more obese individuals(i.e. mean BMI > 35 kg m-2). Finally, there are several vari-ables for which preliminary studies suggest the existence ofan association, but at the moment too few studies areavailable. Exercise variables and weight-specific quality oflife are two examples.
Table 3 Summary of pre-treatment variables reviewed as potential predictors of weight management
Predictors Non-predictors
Consistent Less previous dieting, fewer weight loss attempts Binge/emotional eatingevidence Self-motivation, general efficacy, autonomy Depression, psychopathology, mood
Eating disinhibition, external eatingCognitive (eating) restraint, chronic dietingPerceived hunger
Mixed Initial BMI/weight Personality, general cognitive styleevidence Body image, body size satisfaction Perceived social support
Self-esteemEating self-efficacyRealistic weight loss goals/expectationsInternal locus of control
Suggestive Exercise self-efficacy Low perceived stress, anxietyevidence Few exercise perceived barriers Exercise social support
Quality of life (obesity-specific) Cognitive performancePerceived autonomy (social) support Quality of life (general, SF-36)Absence of bulimic behaviourCognitive style regarding diet lapses
56 Psychosocial predictors of weight loss P. J. Teixeira et al. obesity reviews
© 2005 The International Association for the Study of Obesity. obesity reviews 6, 43–65
A similar logic was applied to non-predictors, for whomconclusions should be seen as more dependable at the topon the table (e.g. binge eating and perceived hunger arereliable non-predictors) and less dependable on the bottom(e.g. there is only preliminary/suggestive evidence thatscores from the SF-36 quality of life questionnaire do notpredict weight loss).
With the exception of suggestive results for initial weight(described before), we were unable to detect differentialeffects for each of the various predictors when contrastingVLCD vs. HBD/LCD programmes. This is not surprisingconsidering the small number of predictors for whichseveral studies were available for the different dietaryinterventions. More importantly, not a single study wasidentified that directly (i.e. within the same research design)compared the two/three types of dietary approachesregarding effects of pre-treatment variables, which furtherlimits our conclusions. In any case, it should be consideredthat many VLCD studies also included a subsequent phasewithout VLCD. Thus, it is conceivable that a pre-treatmentvariable could influence outcomes by affecting adherenceto the VLCD phase or alternatively by affecting compliancewith the non-VLCD phase that frequently follows. Bothwould be reflected on long-term weight loss results.
Matching treatments to participants
The rationale for studying predictors of weight loss is asstrong today as it was in the past, when the followingstatement, now almost 20 years old, was written: ‘thesearch continues because the potential benefits of identify-ing reliable predictors are great. Such information couldprovide an index for prognosis of treatment so that clinicalefforts could be targeted to those most likely to succeed. Inthe likely possibility that some treatments are more effec-tive for certain subgroups of patients, predictor variableswould permit a match between individuals and treatments’(p. 543) (1). Many authors have since expressed similaropinions (11,16,34,37,155–159). However, considering thelarge heterogeneity of individuals presenting for weight lossprogrammes and also the diversity of treatment modalitiesavailable it is perhaps the interaction (i.e. the effectivematch) between individual and treatment characteristics,an aspect largely ignored in the present obesity treatmentliterature, that holds the highest hope for critical answersto the problem of low success rates in obesity treatment.Certain individual profiles (biological, psychological, etc.)may show positive results only under particular types oftreatment.
For example, Dennis & Goldberg showed that severalself-efficacy profiles can be identified before and duringtreatment, and concluded that ‘the distinct differences inthe types of self-efficacy beliefs (. . . ) indicates that it is notappropriate to place all obese women in the same kind of
treatment and expect them to have similar outcomes’(p.114) (83). Rodin et al. observed that an interaction termbetween confidence to reach weight loss goals and pro-gramme type (medication plus cognitive-behavioural ther-apy) was a significant predictor of weight change both aftertreatment (20 weeks) and after a 6-month follow-up (160).Kiernan et al. showed that a particular programme type(diet and physical activity, as opposed to diet-only) com-bined with specific individual characteristics (better bodyimage and no history of repeated weight loss) produced thebest overall weight loss results (49). In another study,VLCD (as opposed to behavioural treatment only) wasshown to be more effective in men than in women, andespecially beneficial in persons living in households withfewer family members (159). Finally, Beliard and cowork-ers reported impressive treatment results (of 173 startingparticipants, 65% achieved moderate to high success after30–70 weeks), using an intensive pre-treatment testing pro-tocol, including individual interview and questionnaires, tomatch participants to one of three treatment modalities(individual counselling, group therapy, or a combinedapproach) (51). Although these and more recent findings(161) have direct relevance to the study of predictors ofweight loss outcomes, few studies have been designed andconducted under this conceptual framework, and not allhave showed matching to be a successful strategy (54).
A moderator of treatment is a particular type of predic-tor defined as a pre-treatment variable that interacts withtreatment modality to produce significantly improvedresults (162,163)3. Analysing individual moderators ofweight control is especially important when we considerthe large and increasing number of treatment modalitiesavailable, which vary in caloric restriction, structure, typeand level of support, philosophy and conceptual model andin several other ways (33). Low-carbohydrate diets (19),meal replacement plans (164), exercise-only (20) and non-dieting approaches (165) are available, as are self-help(166), Internet- (167) and doctor-delivered modalities(168,169), individual, psychologically intense programmes(170), short-term interventions (171), long-term care (172)and various combinations of the above [e.g. (173)]. It islikely that each of these approaches will be beneficial forsome participants; knowing who will benefit from eachapproach is the challenge.
3Besides moderators of obesity treatment (correlated with outcomes only
under specific treatments), other variables may predict weight loss for all
individuals irrespective of treatment type. These have been named non-
specific pre-treatment predictors (162). Since the large majority of the
studies reviewed here did not analyse the interaction between predictors
and treatments to produce weight loss, this distinction is not critical for
this review; the term pre-treatment predictor was used generically to
describe all variables analysed.
obesity reviews Psychosocial predictors of weight loss P. J. Teixeira et al. 57
© 2005 The International Association for the Study of Obesity. obesity reviews 6, 43–65
The problem of attrition
Lack of completion with treatment and/or follow-up mea-surements represents an important limitation in weightmanagement trials. Generalizability of results is particu-larly sensitive to bias caused by analysing data only forparticipants who complete the study, a subgroup of allstarting participants sometimes reaching only 40–50% ofthe initial sample (see Table 1). In studying baseline deter-minants of weight loss success, completer-only analyses areparticularly inadequate as participants presenting with themost barriers to success (i.e. lower readiness) are also likelyto drop out. Dropping out of treatment is often associatedwith lower weight loss, before participants drop out, com-pared with later completers (82,85,174). Hence, to theextent that predictors of weight loss and of attrition aresimilar, not using baseline data for non-completers cancause a substantial reduction in the variance, and predictiveability, of pre-treatment variables regarding weight loss,resulting in lower statistical power.
Procedures have been proposed to decrease drop-outrates, which is undoubtedly the best possible solution(175), and to account for missing data in statistical analyses[e.g. (176)]. The latter can be accomplished through vari-ous data imputation methods, such as baseline or lastobservation carried forward, multiple imputation methods,or using mixed effects models (which do not require dataimputation). More information on this topic, which isbeyond the scope of the present article, can be found else-where (177,178). At a minimum, the most common intent-to-treat model, where baseline (or last observation) valuesare used as final scores for missing subjects, should be usedin predictor studies, in addition to reporting data for com-pleters only.
Evaluating predictors of programme completion (vs.dropping out) is an alternative or complementary procedure(37,179–182). Dropping out of a study is often the worsecase scenario for any given participant, who is unlikely toachieve substantial weight loss and who will not benefitfrom the exposure to the full treatment. Because attritionand smaller weight changes are related and because currentstatistical methods can include drop-outs’ data in analysesand partially mitigate the impact of attrition, it is likely thatsome redundancy is present when comparing pre-treatmentpredictors of drop-out and predictors of weight loss (3,64).In addition, when the initial sample size is small (and/ordrop-out rates are low) the reduced number of non-compl-eters may render statistical comparisons between completersand non-completers underpowered and uninformative.
Conclusions and future research directions
Our major goal was to review the most recent studieson prospective predictors of weight loss, with the aim of
improving the understanding of the impact of pre-treatment factors in weight control. After a thoroughreview of the available data, only a few variables emergedas significant pre-treatment predictors of subsequentweight management results in behavioural interventions.Conversely, sufficient evidence exists for a larger numberof additional constructs, which do not appear to predictweight outcomes. For a majority of variables, however,published research data are not sufficient for a valid con-clusion one way or the other. Thus, based on current evi-dence, it appears inappropriate to refuse treatment tooverweight/obese candidates who so desire, if they are ingood physical and psychological health, considering base-line information alone. As so few data are available anal-ysing true moderators of success (i.e. patient–treatmentinteractions), no definitive conclusion can be drawn fromthis review regarding the future matching of subjects totreatments. This notwithstanding, some qualifications tothe present findings should be highlighted, suggesting dif-ferent interpretations of, and implications from the currentevidence. We now address some of them.
First, predicting levels of success may be different thanpredicting individual weight lost. In other words, anticipat-ing if a person, or a group presenting similar characteris-tics, will likely be among the most or, more critically,among the least successful participants by treatment’s endmay be easier to accomplish than estimating exactly howweight will change. In fact, some evidence suggests catego-ries of success can be predicted from baseline data(17,49,53,64,131). This could also facilitate a simplematching of treatment to patients, where, for instance,treatment groups containing pre-defined numbers of themost and least likely to succeed would be set up to worktogether throughout the programme (or separately, ifdeemed more appropriate), instead of the common proce-dure of having randomly assigned intervention groups.Very recent studies have started to use similar approachesof manipulating a priori group formation, with promisingresults (183).
Second, broader definitions of success in weight manage-ment should be considered (184–186). By considering addi-tional biological (e.g. body composition, cardio-respiratoryfitness), behavioural (e.g. programme completion, meetingactivity and dietary intake goals) and psychosocial out-comes (e.g. quality of life, self-worth, mental health), it isconceivable that models based on these bio-psycho-socialprofiles stand a better chance of statistical and clinicalsignificance, than what has been achieved thus far.
Third, we have informally observed that many partici-pants in weight reduction programmes, mostly femaleadults, seem to enjoy and/or benefit personally from theprocess of filling out questionnaires regarding their atti-tudes, perceptions and knowledge on issues such as bodyimage, emotional eating, exercise, social connectedness,
58 Psychosocial predictors of weight loss P. J. Teixeira et al. obesity reviews
© 2005 The International Association for the Study of Obesity. obesity reviews 6, 43–65
among others. Subjects often mention increased awareness,motivation and improved understanding of personal fac-tors critical to their obesity status (and potentially to treat-ment) upon filling out the psychometric battery. Takingparticipants’ subjective views into greater account whenmaking treatment decisions in obesity treatment (personalpreferences, attitudes regarding the best course of treat-ment, specific goals and reasons to present for treatment,etc.) can be a worthy avenue of investigation. Pre-treatmentpsychosocial assessments could be used to prompt partici-pants to reflect upon issues related to available treatmentchoices or decisions where their input would be used.
Fourth, although at this time elaborate evidence-basedmatching decisions regarding psychosocial profiles cannotbe recommended, this does not mean some degree of indi-vidualization (or tailoring) cannot be incorporated imme-diately and tested against less individualized approaches.One example is the pre-treatment identification of individ-uals with the most unrealistic expectations for weight loss,who could benefit from a more directed interventionaimed at addressing weight and other goals, their originsand their implications for treatment (170). This particularcognitive restructuring protocol would not have to be usedwith participants with more realistic goals, for whomintervention resources could be used in different ways.The same logic could be applied for other characteristicsthat evidence suggests may be detrimental to long-termsuccess, such as having many perceived exercise barriers,low levels of autonomy and self-motivation, an externallocus of control and low self- and body-esteem. In practi-cal terms, this would represent a middle way along thecontinuum from individual therapy (170) to the mostcommon group, ‘one-size-fits-all’ approaches. Even if notwholly successful, some examples on how to improve thedegree of individualization within formal obesity treat-ment programmes are available and can lead the way(51,53–55,156,187–189).
Fifth, we recommend the evaluation of pre-treatmentassessments within predictive models that include pre-treatment as well as process factors, particularly events andvariables assessed early in the programme (i.e. first 4–6 months, roughly corresponding to the extent of pro-grammes’ most intense phase or active weight loss period).Predictive models including baseline and process variables,such as early weight change, attendance, changes in exer-cise and eating attitudinal/behavioural measures, bodyimage, among others, may predict long-term success orfailure better than models with pre-treatment data only(18,71,190). If this proves correct, important implicationscould be derived in terms of resource allocation, interven-tion design (duration, focus, etc.), matching criteria andother decisions regarding long-term weight maintenanceprogrammes, rather than having indefinite treatment as theonly (partially) effective solution (172).
Sixth, although this review deliberately focused on psy-chosocial and other easily accessible variables, investigatorsshould continue to explore less accessible prognostic mark-ers of effective weight control (genetic, neuroendocrine,metabolic, environmental, etc.). This has been attemptedbefore with some success (48,191) and it should proceed,in the hope that less-accessible variables become increas-ingly available to researchers, clinicians, dietitians andother professionals.
Next, a frequent criticism of research using psychologicalpredictors of behaviour is the lack of a sound theoreticalbackground (7). Theory-based formulations in behaviouralweight loss research are clearly needed and are progres-sively being investigated and applied (59,192,193). How-ever, the remarkable difficulty in producing a singletheoretical representation predictive of an outcome as com-plex as human weight change over time must be recog-nized, and it should not stand in the way of continuing topursue more exploratory research lines, if well-justified andmethodologically sound (162,194). This is an area whereboth hypotheses-testing and hypotheses-generating studiesare needed.
Eighth, most constructs are clearly understudied aspredictors of weight loss, especially considering the largevariability in methodologies among the studies that areavailable (e.g. psychometric instruments, treatment modal-ities). For about half of the predictors we identified, onlyone to four studies/predictive models are available in theliterature (that we could identify, see Table 2), which isinsufficient for a firm conclusion to be drawn.
Ninth, it is likely that many studies in this review wereunderpowered to detect significant correlations. In thebehavioural sciences, effect sizes (in terms of R2 ¥ 100 or% variance accounted) for moderate and strong associa-tions can be set at 5% and 10%, respectively, the equivalentof correlation coefficients between 0.22 and 0.31(194,195). Sample sizes required to detect these associa-tions as significant (a = 0.05 level, 1 - b = 0.8) would be160 and 79 subjects respectively. In Table 1, the majorityof studies had a low level of power to detect significantmoderate associations (r around 0.2), and many might notdetect correlations around 0.3 as significant, attributed tosample size alone.
Tenth, the issue of appropriately using cumulative evi-dence to test the strength of association for single predic-tors, when a sufficient number of studies are available,should also be raised. As Allison pointed out (7), it may bethat more research is not needed in some cases, but insteada more efficient way to analyse all the evidence available isrequired. Meta-analytical and related procedures may beable to detect significant associations between a predictorand an outcome, even when a majority of isolated studiesshow non-significant associations. For example, initialBMI, depression (e.g. with the BDI), eating behaviours (e.g.
obesity reviews Psychosocial predictors of weight loss P. J. Teixeira et al. 59
© 2005 The International Association for the Study of Obesity. obesity reviews 6, 43–65
with scales from the Eating Inventory) and binge eating(e.g. with the Binge Eating Scale) could now be exploredin this fashion.
Lastly, in the reality of weight management programmes,the association between baseline psychosocial factors andweight loss success, particularly over longer periods oftime, is probably moderated and mediated by a series offactors, most of which this review could not adequatelyaddress. It is apparent that more studies on the variouspredictor–treatment combinations are needed for a moreaccurate evaluation. Other potentially important treatmentvariables were not considered or not properly analysed,such as treatment group size, type of exercise recommen-dations, occurrence/characteristics of maintenance pro-grammes, among others. Also, potential moderators suchas gender, genetic/biological factors, family/work circum-stances and other psychosocial factors were not reviewedhere. Moreover, as the large majority of the studiesreviewed used a ‘bivariate model’ to study predictors, it islikely that important interactions among predictors exist.For example, the effects of depression on success rates maybe contingent on the presence of a binge eating pattern [orvice versa (135)], on attitudes/behaviours regarding physi-cal activity [which improves depression (196)], or otherfactors such as external stressors or social support. Simi-larly, there is increasing evidence for interacting effectsbetween body image, psychological distress and eatingbehaviours (121,197–199).
Summary of limitations and recommended research directions
The following points highlight some of the most salientlimitations in the literature on predictors of weight man-agement, and summarize possible lines of investigation toovercome them. Many have been alluded to in previoussections and are described henceforth in brief format.
Large heterogeneity in intervention programmesMore studies are needed with standardized behaviouralapproaches, which can be adequately compared relative tovariables assessed at baseline.
Few studies available for a large number of constructsMore research is particularly recommended for: self-motivation and autonomy, body image and self-esteem,goals and expectations, general and weight-specific locusof control, social support, exercise, quality of life, andstress/anxiety.
Lack of studies evaluating the interaction between individual and programme characteristicsMore such studies are encouraged, preferably using clearlydistinct treatment modalities (e.g. individual vs. group;
treatments with vs. without drugs/meal replacements;aggressive dieting vs. non-dieting, etc.), comprehensive psy-chosocial batteries, and studying the interaction amongbiological, psychological and behavioural individualvariables.
Homogeneity of current research design paradigmsStudies where profile-based or more individualized inter-ventions are tested against one-size-fits-all approaches arestrongly encouraged. More qualitative research [e.g. (29)],an area not covered in this review, is also strongly sup-ported as are retrospective investigations that analysesubjects’ subjective perceptions and attitudes as potentialmoderators of success.
Too many instruments for similar constructsResearchers are encouraged to use only the most psycho-metrically sound and previously used instruments. Guide-lines are provided throughout this review and are alsoavailable elsewhere (140,200–203).
Lack of statistical powerAppropriate a priori power analyses are encouraged, con-sidering the low-to-moderate effect sizes typically observedin the behavioural sciences. Some solutions include usingimputation techniques for missing data, collapsing data-bases from different research centres (or from multiplecomparable published studies) to maximize sample size,more adequate/powerful statistical procedures and decreas-ing attrition.
(The assumption of high need for) very large sample size and very long follow-up periodsThe need for appropriate sample size and power are notcontradictory with smaller-scale studies testing very spe-cific variables in well-defined settings, innovative workinghypotheses and alternative/creative research models.Fresh ideas are needed and many types of studies can beuseful.
Too high attrition rates and completers-only analysesNew ideas are also needed to improve rates of completion,which are too small in many obesity treatment studies. Ata minimum, strategies to ensure weight changes can becaptured over time for non-compliers, even in less-than-ideal conditions that should be attempted (e.g. self-reportby phone or email).
Too simple statistical analysesMeta-analytical, multivariate and other more powerful/alternate statistical procedures (e.g. structural equationsmodels, recursive partitioning techniques) are encouragedto complement more common techniques. Outcomes andchange variables can also be explored and expressed in
60 Psychosocial predictors of weight loss P. J. Teixeira et al. obesity reviews
© 2005 The International Association for the Study of Obesity. obesity reviews 6, 43–65
alternative ways such as evaluating data in a disaggregatedfashion (e.g. looking at subgroups based on multiple-pointindividual change vectors) rather than focusing exclusivelyon group mean outcomes (204).
Theoretical limitationsTheory-based and theory development work is particularlyimportant and strongly encouraged, including psychosocialas well as physiological and other (genetic, environmental,socio-cultural) variables. Extensions of the Theory ofPlanned Behaviour and social ecological models are twoareas that have been highlighted as particularly promising(194).
Lack of body composition dataWhenever possible, accounting for body composition, inaddition to weight changes is recommended, if the analyt-ical methods used are sensitive to intraindividual change.
Exclusive or excessive focus on weight as an outcomeWeight and other body habitus changes are only a part ofthe many health-related effects of weight management pro-grammes. Outcome variables can also include quality oflife, specific healthy behaviours, body image, self-esteem,social functioning and many other variables.
Weight loss and weight maintenance outcomes not differentiatedPredictive models should be time-specific, in recognition ofthe differences that exist between losing weight and keep-ing it off. Also, investigation into demographic and othertypes of treatment moderators (age, gender, socio-economicstatus, etc.) should be supported.
Multidisciplinary, international, cross-ethnicity and cross-cultural studies underrepresentedA wide gap in information exists regarding cultural, ethnicand cross-country differences in response to obesity treat-ment and factors that correlate with success. Also, studiesthat cross disciplinary fields (e.g. cognitive science and evo-lutionary psychology, anthropology, sociology, molecularand evolutionary biology) would be welcomed in obesityresearch.
Acknowledgements
The authors are grateful to Drs Kelly Brownell andMichaela Kiernan for their insightful comments to earlydrafts of this manuscript. Additionally, the first authorwishes to thank the Faculty of Human Movement, Techni-cal University of Lisbon for its institutional support duringthe preparation of this article. The authors have no con-flicts of interest to report regarding information presented
in this manuscript. This work was possible thanks in partto the NIH/NIDDK Grant #57453.
References
1. Brownell KD. Behavioral, psychological, and environmentalpredictors of obesity and success at weight reduction. Int J Obes1984; 8: 543–550.2. Foreyt JP, Goodrick GK. Factors common to successful therapyfor the obese patient. Med Sci Sports Exerc 1991; 23: 292–297.3. Wadden TA, Letizia KA. Predictors of attrition and weight lossin patients treated by moderate to severe caloric restriction. In:Wadden TA, VanItallie TB (eds). Treatment of the Seriously ObesePatient. The Guilford Press: New York, NY, 1992, pp. 383–410.4. Wilson GT. Behavioral and psychological predictors of treat-ment outcome in obesity. In: Allison DA, Pi-Sunyer FX (eds).Obesity Treatment: Establishing Goals, Improving Outcomes, andReviewing the Research Agenda. Plenum Press: New York, 1995,pp. 183–189.5. Weiss AR. Characteristics of successful weight reducers: a briefreview of predictor variables. Addict Behav 1977; 2: 193–201.6. Foreyt JP, Goodrick GK. Prediction of weight managementoutcome: Implications for practice. In: Allison DA, Pi-Sunyer FX(eds). Obesity Treatment: Establishing Goals, Improving Out-comes, and Reviewing the Research Agenda. Plenum Press: NewYork, 1995, pp. 199–205.7. Allison DB, Engel CN. Predicting treatment outcome: why havewe been so unsuccessful? In: Allison DA, Pi-Sunyer FX (eds).Obesity Treatment: Establishing Goals, Improving Outcomes, andReviewing the Research Agenda. Plenum Press: New York, 1995,pp. 191–198.8. Institute of Medicine. Weighing the Options: Criteria for Eval-uating Weight-Management Programs. National Academy Press:Washington, DC, 1995.9. Levine JA, Eberhardt NL, Jensen MD. Role of nonexerciseactivity thermogenesis in resistance to fat gain in humans. Science1999; 283: 212–214.10. Hainer V, Stunkard AJ, Kunesova M, Parizkova J, Stich V,Allison DB. Intrapair resemblance in very low calorie diet-inducedweight loss in female obese identical twins. Int J Obes Relat MetabDisord 2000; 24: 1051–1057.11. Fitzgibbon ML, Stolley MR, Kirschenbaum DS. Obese peoplewho seek treatment have different characteristics than those whodo not seek treatment. Health Psychol 1993; 12: 342–345.12. Friedman MA, Brownell KD. Psychological correlates of obe-sity: moving to the next research generation. Psychol Bull 1995;117: 3–20.13. Sarwer DB, Wadden TA, Foster GD. Assessment of bodyimage dissatisfaction in obese women: specificity, severity, andclinical significance. J Consult Clin Psychol 1998; 66: 651–654.14. Burns CM, Tijhuis MA, Seidell JC. The relationship betweenquality of life and perceived body weight and dieting history inDutch men and women. Int J Obes Relat Metab Disord 2001; 25:1386–1392.15. Marchesini G, Natale S, Chierici S, Manini R, Besteghi L, DiDomizio S, Sartini A, Pasqui F, Baraldi L, Forlani G, MelchiondaN. Effects of cognitive-behavioural therapy on health-related qual-ity of life in obese subjects with and without binge eating disorder.Int J Obes Relat Metab Disord 2002; 26: 1261–1267.16. Brownell KD, Rodin J. The dieting maelstrom. Is it possibleand advisable to lose weight? Am Psychol 1994; 49: 781–791.17. Teixeira PJ, Going SB, Houtkooper LB, Cussler EC, MartinCJ, Metcalfe LL, Finkenthal NR, Blew RM, Sardinha LB, Lohman
obesity reviews Psychosocial predictors of weight loss P. J. Teixeira et al. 61
© 2005 The International Association for the Study of Obesity. obesity reviews 6, 43–65
TG. Weight loss readiness in middle-aged women: psychosocialpredictors of success for behavioral weight reduction. J Behav Med2002; 25: 499–523.18. Fogelholm M, Kukkonen-Harjula K, Oja P. Eating controland physical activity as determinants of short-term weight main-tenance after a very-low-calorie diet among obese women. IntJ Obes Relat Metab Disord 1999; 23: 203–210.19. Foster GD, Wyatt HR, Hill JO, McGuckin BG, Brill C,Mohammed BS, Szapary PO, Rader DJ, Edman JS, Klein S. Arandomized trial of a low-carbohydrate diet for obesity. N EnglJ Med 2003; 348: 2082–2090.20. Donnelly JE, Hill JO, Jacobsen DJ, Potteiger J, Sullivan DK,Johnson SL, Heelan K, Hise M, Fennessey PV, Sonko B, Sharp T,Jakicic JM, Blair SN, Tran ZV, Mayo M, Gibson C, WashburnRA. Effects of a 16-month randomized controlled exercise trial onbody weight and composition in young, overweight men andwomen: the Midwest Exercise Trial. Arch Intern Med 2003; 163:1343–1350.21. Skender ML, Goodrick GK, Del Junco DJ, Reeves RS, DarnellL, Gotto AM, Foreyt JP. Comparison of 2-year weight loss trendsin behavioral treatments of obesity: diet, exercise, and combina-tion interventions. J Am Diet Assoc 1996; 96: 342–346.22. Heshka S, Anderson JW, Atkinson RL, Greenway FL, HillJO, Phinney SD, Kolotkin RL, Miller-Kovach K, Pi-Sunyer FX.Weight loss with self-help compared with a structured commercialprogram: a randomized trial. Jama 2003; 289: 1792–1798.23. Wadden TA, Berkowitz RI, Sarwer DB, Prus-Wisniewski R,Steinberg C. Benefits of lifestyle modification in the pharmacologictreatment of obesity: a randomized trial. Arch Intern Med 2001;161: 218–227.24. Flegal KM, Carroll MD, Ogden CL, Johnson CL. Prevalenceand trends in obesity among US adults, 1999–2000. Jama 2002;288: 1723–1727.25. Crawford D, Jeffery RW, French SA. Can anyone successfullycontrol their weight? Findings of a three year community-basedstudy of men and women. Int J Obes Relat Metab Disord 2000;24: 1107–1110.26. Anderson JW, Konz EC, Frederich RC, Wood CL. Long-termweight-loss maintenance: a meta-analysis of US studies. Am J ClinNutr 2001; 74: 579–584.27. Melin I, Karlstrom B, Lappalainen R, Berglund L, MohsenR, Vessby B. A programme of behaviour modification and nutri-tion counselling in the treatment of obesity: a randomised 2-yclinical trial. Int J Obes Relat Metab Disord 2003; 27: 1127–1135.28. McGuire MT, Wing RR, Hill JO. The prevalence of weightloss maintenance among American adults. Int J Obes Relat MetabDisord 1999; 23: 1314–1319.29. Byrne S, Cooper Z, Fairburn C. Weight maintenance andrelapse in obesity: a qualitative study. Int J Obes Relat MetabDisord 2003; 27: 955–962.30. Klem ML, Wing RR, McGuire MT, Seagle HM, Hill JO. Adescriptive study of individuals successful at long-term mainte-nance of substantial weight loss. Am J Clin Nutr 1997; 66: 239–246.31. Wing RR. Behavioral interventions for obesity: recognizingour progress and future challenges. Obes Res 2003; 11(Suppl. 1):3S–6S.32. Brownell KD, Wadden TA. Etiology and treatment of obesity:understanding a serious, prevalent, and refractory disorder. J Con-sult Clin Psychol 1992; 60: 505–517.33. Schwartz MB, Brownell KD. Matching individuals to weightloss treatments: a survey of obesity experts. J Consult Clin Psychol1995; 63: 149–153.
34. Fitzgibbon ML, Kirschenbaum DS. Heterogeneity of clinicalpresentation among obese individuals seeking treatment. AddictBehav 1990; 15: 291–295.35. Fitzgibbon ML, Kirschenbaum DS. Distressed binge eaters asa distinct subgroup among obese individuals. Addict Behav 1991;16: 441–451.36. Brownell KD, Kramer FM. Behavioral management of obe-sity. In: Blackburn GL, Kanders BS (eds). Obesity: Pathophysiol-ogy, Psychology, and Treatment. Chapman & Hall: New York,1994, pp. 231–252.37. Clark MM, Guise BJ, Niaura RS. Obesity level and attrition:support for patient-treatment matching in obesity treatment. ObesRes 1995; 3: 63–64.38. Brownell KD, Wadden TA. The heterogeneity of obesity: fit-ting treatments to individuals. Behav Ther 1991; 22: 153–177.39. USDHHS. Clinical Guidelines on the Identification, Evalua-tion, and Treatment of Overweight and Obesity in Adults. NIH –National Heart, Lung, and. Blood Institute: Bethesda, MD, 1998.40. Wadden TA, Brownell KD, Foster GD. Obesity: respondingto the global epidemic. J Consult Clin Psychol 2002; 70: 510–525.41. Yanovski SZ, Yanovski JA. Obesity. N Engl J Med 2002; 346:591–602.42. Serdula MK, Khan LK, Dietz WH. Weight loss counselingrevisited. JAMA 2003; 289: 1747–1750.43. Anderson JW, Pi-Sunyer FX, Danforth E, Dujovne CA,Greenway F, Hill JO, Lucas CP, O’Neil PM, Smith DK. Clinicaltrial design for obesity agents: a workshop report. Obes Res 1998;6: 311–315.44. Wing RR, Hill JO. Successful weight loss maintenance. AnnuRev Nutr 2001; 21: 323–341.45. Wilson GT. Psychological prognostic factors in the treatmentof obesity. In: Hirsch J, Van Ittalie TB (eds). Recent Advances inObesity Research. Libbey: London, 1985, pp. 301–311.46. Jeffery RW, Bjornson-Benson WM, Rosenthal BS, LindquistRA, Kurth CL, Johnson SL. Correlates of weight loss and itsmaintenance over two years of follow-up among middle-aged men.Prev Med 1984; 13: 155–168.47. Edell BH, Edington S, Herd B, O’Brien RM, Witkin G. Self-efficacy and self-motivation as predictors of weight loss. AddictBehav 1987; 12: 63–66.48. Wadden TA, Foster GD, Wang J, Pierson RN, Yang MU,Moreland K, Stunkard AJ, VanItallie TB. Clinical correlates ofshort- and long-term weight loss. Am J Clin Nutr 1992; 56: 271S–274S.49. Kiernan M, King AC, Kraemer HC, Stefanick ML, Killen JD.Characteristics of successful and unsuccessful dieters: an applica-tion of signal detection methodology. Ann Behav Med 1998; 20:1–6.50. Gladis MM, Wadden TA, Vogt R, Foster G, Kuehnel RH,Bartlett SJ. Behavioral treatment of obese binge eaters: do theyneed different care? J Psychosom Res 1998; 44: 375–384.51. Beliard D, Kirschenbaum DS, Fitzgibbon ML. Evaluation ofan intensive weight control program using a priori criteria todetermine outcome. Int J Obes Relat Metab Disord 1992; 16:505–517.52. Ogden J. The correlates of long-term weight loss: a groupcomparison study of obesity. Int J Obes Relat Metab Disord 2000;24: 1018–1025.53. Feuerstein M, Papciak A, Shapiro S, Tannenbaum S. Theweight loss profile: a biopsychosocial approach to weight loss. IntJ Psychiatry Med 1989; 19: 181–192.54. Renjilian DA, Perri MG, Nezu AM, McKelvey WF, ShermerRL, Anton SD. Individual versus group therapy for obesity: effects
62 Psychosocial predictors of weight loss P. J. Teixeira et al. obesity reviews
© 2005 The International Association for the Study of Obesity. obesity reviews 6, 43–65
of matching participants to their treatment preferences. J ConsultClin Psychol 2001; 69: 717–721.55. Fuller PR, Perri MG, Leermakers EA, Guyer LK. Effects of apersonalized system of skill acquisition and an educational pro-gram in the treatment of obesity. Addict Behav 1998; 23: 97–100.56. USDHHS. The Practical Guide to the Identification, Evalua-tion, and Treatment of Overweight and Obesity in Adults. NIH –National Heart, Lung, and Blood Institute: Bethesda, MD, 2000.57. Womble LG, Williamson DA, Greenway FL, Redmann SM.Psychological and behavioral predictors of weight loss during drugtreatment for obesity. Int J Obes Relat Metab Disord 2001; 25:340–345.58. Drapkin RG, Wing RR, Shiffman S. Responses to hypotheti-cal high risk situations: do they predict weight loss in a behavioraltreatment program or the context of dietary lapses? Health Psy-chol 1995; 14: 427–434.59. Williams GC, Grow VM, Freedman ZR, Ryan RM, Deci EL.Motivational predictors of weight loss and weight-loss mainte-nance. J Pers Soc Psychol 1996; 70: 115–126.60. Kolotkin RL, Crosby RD, Williams GR. Health-related qual-ity of life varies among obese subgroups. Obes Res 2002; 10: 748–756.61. Wing RR, Jeffery RW. Benefits of recruiting participants withfriends and increasing social support for weight loss and mainte-nance. J Consult Clin Psychol 1999; 67: 132–138.62. Pekkarinen T, Takala I, Mustajoki P. Two year maintenanceof weight loss after a VLCD and behavioural therapy for obesity:correlation to the scores of questionnaires measuring eating behav-iour. Int J Obes Relat Metab Disord 1996; 20: 332–337.63. Pasman WJ, Saris WH, Westerterp-Plantenga MS. Predictorsof weight maintenance. Obes Res 1999; 7: 43–50.64. Teixeira PJ, Going SB, Houtkooper LB, Cussler EC, MetcalfeLL, Blew RM, Sardinha LB, Lohman TG. Pretreatment predictorsof attrition and successful weight management in women. Int JObes Relat Metab Disord 2004; 28: 1124–1133.65. Forster JL, Jeffery RW. Gender differences related to weighthistory, eating patterns, efficacy expectations, self-esteem, andweight loss among participants in a weight reduction program.Addict Behav 1986; 11: 141–147.66. French SA, Jeffery RW, Forster JL, McGovern PG, Kelder SH,Baxter JE. Predictors of weight change over two years among apopulation of working adults: the Healthy Worker Project. IntJ Obes Relat Metab Disord 1994; 18: 145–154.67. Hoiberg A, Berard S, Watten RH, Caine C. Correlates ofweight loss in treatment and at follow-up. Int J Obes 1984; 8:457–465.68. Gormally J, Black S, Daston S, Rardin D. The assessment ofbinge eating severity among obese persons. Addict Behav 1982; 7:47–55.69. Stunkard AJ, Messick S. The three-factor eating questionnaireto measure dietary restraint, disinhibition and hunger. J Psycho-som Res 1985; 29: 71–83.70. Spitzer RL, Yanovski S, Wadden T, Wing R, Marcus MD,Stunkard A, Devlin M, Mitchell J, Hasin D, Horne RL. Bingeeating disorder: its further validation in a multisite study. Int J EatDisord 1993; 13: 137–153.71. Sherwood NE, Jeffery RW, Wing RR. Binge status as a pre-dictor of weight loss treatment outcome. Int J Obes Relat MetabDisord 1999; 23: 485–493.72. Westenhoefer J, Stunkard AJ, Pudel V. Validation of the flex-ible and rigid control dimensions of dietary restraint. Int J EatDisord 1999; 26: 53–64.73. Lowe MR, Foster GD, Kerzhnerman I, Swain RM, WaddenTA. Restrictive dieting vs. ‘undieting’ effects on eating regulationin obese clinic attenders. Addict Behav 2001; 26: 253–266.
74. Brownell KD. The LEARN Program for Weight Management.American Health Publishing Co: Dallas, TX, 2000.75. Foster GD, Wadden TA, Swain RM, Stunkard AJ, Platte P,Vogt RA. The Eating Inventory in obese women: clinical correlatesand relationship to weight loss. Int J Obes Relat Metab Disord1998; 22: 778–785.76. Yanovski SZ, Gormally JF, Leser MS, Gwirtsmann HE,Yanovski JA. Binge eating disorder affects outcome of comprehen-sive very-low-calorie diet treatment. Obes Res 1994; 2: 205–212.77. LaPorte DJ. Treatment response in obese binge eaters: prelim-inary results using a very low calorie diet (VLCD) and behaviortherapy. Addict Behav 1992; 17: 247–257.78. Glynn SM, Ruderman AJ. The development and validation ofan eating self-efficacy scale. Cognitive Ther Res 1986; 10: 403–420.79. Clark MM, Abrams DB, Niaura RS, Eaton CA, Rossi JS. Self-efficacy in weight management. J Consult Clin Psychol 1991; 59:739–744.80. Bernier M, Avard J. Self-efficacy, outcome and attrition in aweight-reduction program. Cognitive Ther Res 1986; 10: 319–338.81. Prochaska JO, Norcross JC, Fowler JL, Follick MJ, AbramsDB. Attendance and outcome in a work site weight control pro-gram: processes and stages of change as process and predictorvariables. Addict Behav 1992; 17: 35–45.82. Karlsson J, Hallgren P, Kral J, Lindroos AK, Sjostrom L,Sullivan M. Predictors and effects of long-term dieting on mentalwell-being and weight loss in obese women. Appetite 1994; 23:15–26.83. Dennis KE, Goldberg AP. Weight control self-efficacy typesand transitions affect weight-loss outcomes in obese women.Addict Behav 1996; 21: 103–116.84. Clifford PA, Tan SY, Gorsuch RL. Efficacy of a self-directedbehavioral health change program: weight, body composition,cardiovascular fitness, blood pressure, health risk, and psychoso-cial mediating variables. J Behav Med 1991; 14: 303–323.85. Pratt CA, McLaughlin GW, Gaylord C. A multivariateanalysis of weight-loss behavior. Psychol Rep 1992; 71: 1075–1084.86. Fontaine KR, Cheskin LJ. Self-efficacy, attendance, andweight loss in obesity treatment. Addict Behav 1997; 22: 567–570.87. Traverso A, Ravera G, Lagattolla V, Testa S, Adami GF.Weight loss after dieting with behavioral modification for obesity:the predicting efficiency of some psychometric data. Eat WeightDisord 2000; 5: 102–107.88. Bandura A. Self-efficacy: toward a unifying theory of behav-ioral change. Psychol Rev 1997; 84: 191–215.89. Dishman RK, Ickes W. Self-motivation and adherence to ther-apeutic exercise. J Behav Med 1981; 4: 421–438.90. Bjorvell H, Aly A, Langius A, Nordstrom G. Indicators ofchanges in weight and eating behaviour in severely obese patientstreated in a nursing behavioural program. Int J Obes Relat MetabDisord 1994; 18: 521–525.91. Oettingen G, Wadden TA. Expectation, fantasy, and weightloss: is the impact of positive thinking always positive? CognitiveTher Res 1991; 15: 167–174.92. Foster GD, Wadden TA, Vogt RA, Brewer G. What is areasonable weight loss? Patients’ expectations and evaluations ofobesity treatment outcomes. J Consult Clin Psychol 1997; 65: 79–85.93. Jeffery RW, Wing RR, Mayer RR. Are smaller weight lossesor more achievable weight loss goals better in the long term forobese patients? J Consult Clin Psychol 1998; 66: 641–645.94. Polivy J, Herman CP. If at first you don’t succeed. False hopesof self-change. Am Psychol 2002; 57: 677–689.
obesity reviews Psychosocial predictors of weight loss P. J. Teixeira et al. 63
© 2005 The International Association for the Study of Obesity. obesity reviews 6, 43–65
95. Polivy J. The false hope syndrome. unrealistic expectations ofself-change. Int J Obes Relat Metab Disord 2001; 25(Suppl. 1):S80–S84.96. Foster GD, Wadden TA, Phelan S, Sarwer DB, Sanderson RS.Obese patients’ perceptions of treatment outcomes and the factorsthat influence them. Arch Intern Med 2001; 161: 2133–2139.97. Linde JA, Jeffery RW, Finch EA, Ng DM, Rothman AJ. Areunrealistic weight loss goals associated with outcomes for over-weight women? Obes Res 2004; 12: 569–576.98. O’Neil PM, Smith CF, Foster GD, Anderson DA. The per-ceived relative worth of reaching and maintaining goal weight. IntJ Obes Relat Metab Disord 2000; 24: 1069–1076.99. Rotter JB. Generalized expectancies for internal versus exter-nal control of reinforcement. Psychol Monogr 1966; 80: 1–28.100. Reid DW, Ware EE. Multidimensionality of internal-externalcontrol: implications for past and future research. Can J Behav Sci1974; 5: 264–271.101. Gurin P, Gurin G, Lao RC, Beattie M. Multidimensional I-E scale. In: Robinson JP, Shaver PR (eds). Measures of SocialPsychological Attitudes. Survey Research Center Institute forSocial Research: Ann Arbor, MI, 1995, pp. 218–226.102. Wallston BS, Wallston KA, Kaplan GD, Maides SA. Devel-opment and validation of the health locus of control (HLC) scale.J Consult Clin Psychol 1976; 44: 580–585.103. Wallston KA, Wallston BS, DeVellis R. Development of theMultidimensional Health Locus of Control (MHLC) scales.Health Educ Monogr 1978; 6: 160–170.104. Stotland S, Zuroff DC. A new measure of weight locus ofcontrol: the Dieting Beliefs Scale. J Pers Assess 1990; 54: 191–203.105. Saltzer EB. The weight locus of control (WLOC) scale: aspecific measure for obesity research. J Pers Assess 1982; 46: 620–628.106. Holt CL, Clark EM, Kreuter MW. Weight locus of controland weight-related attitudes and behaviors in an overweight pop-ulation. Addict Behav 2001; 26: 329–340.107. Bryan J, Tiggemann M. The effect of weight-loss dieting oncognitive performance and psychological well-being in overweightwomen. Appetite 2001; 36: 147–156.108. Markland D, Hardy L. On the factorial and construct valid-ity of the Intrinsic Motivation Inventory: conceptual and opera-tional concerns. Res Q Exerc Sport 1997; 68: 20–32.109. Rose EA, Markland D, Parfitt G. The development andinitial validation of the Exercise Causality Orientations Scale. JSports Sci 2001; 19: 445–462.110. Rosenberg M. Society and the Adolescent Self-Image. Prin-ceton University Press: Princeton, NJ, 1965.111. Rosen JR. Improving body image in obesity. In: ThompsonJK (ed.). Body Image, Eating Disorders, and Obesity. An Integra-tive Guide for Assessment and Treatment. American PsychologicalAssociation: Washington, DC, 1996, pp. 425–440.112. Garner MG, Olsmted MP, Polivy J. Development and vali-dation of a multidimensional eating disorder inventory for anor-exia nervosa and bulimia. Int J Eat Disord 1984; 2: 15–34.113. Ben-Tovim DI, Walker MK. The development of the Ben-Tovim Walker Body Attitudes Questionnaire (BAQ), a new mea-sure of women’s attitudes towards their own bodies. Psychol Med1991; 21: 775–784.114. Rosen JC, Jones A, Ramirez E, Waxman S. Body ShapeQuestionnaire: studies of validity and reliability. Int J Eat Disord1996; 20: 315–319.115. Cooper PJ, Taylor MJ, Cooper Z, Fairburn CG. The devel-opment and validation of a body shape questionnaire. Int J EatDisord 1987; 6: 485–494.116. Fox KR. The Physical Self-Perception Profile Manual.Northern Illinois University: DeKalb, IL, 1990.
117. Secord S, Jourard P. The appraisal of body cathexis: bodycathexis and the self. J Consult Psych 1953; 17: 343–347.118. Williamson DA, Gleaves DH, Watkins PC, Schlundt DG.Validation of a self-ideal body size discrepancy as a measure ofbody size dissatisfaction. J Psychol Behav Assess 1993; 15(Suppl.1): 57–68.119. Thompson JK. Assessing body image disturbance. measures,methodology, and implementation. In: Thompson JK (ed.). BodyImage, Eating Disorders, and Obesity. An Integrative Guide forAssessment and Treatment. American Psychological Association:Washington, DC, 1996, pp. 49–82.120. Leibbrand R, Fichter MM. Maintenance of weight loss afterobesity treatment: is continuous support necessary? Behav ResTher 2002; 40: 1275–1289.121. Wardle J, Waller J, Rapoport L. Body dissatisfaction andbinge eating in obese women: the role of restraint and depression.Obes Res 2001; 9: 778–787.122. Wardle J, Waller J, Fox E. Age of onset and body dissatis-faction in obesity. Addict Behav 2002; 27: 561–573.123. Nir Z, Neumann L. Relationship among self-esteem, inter-nal-external locus of control, and weight change after participa-tion in a weight reduction program. J Clin Psychol 1995; 51: 482–490.124. Eldredge KL, Agras WS. The relationship between per-ceived evaluation of weight and treatment outcome amongindividuals with binge eating disorder. Int J Eat Disord 1997;22: 43–49.125. Nir Z, Neumann L. Self-esteem, internal-external locus ofcontrol, and their relationship to weight reduction. J Clin Psychol1991; 47: 568–575.126. Beck AT, Ward CH, Mendelson M, Mock J, Erbaugh J. Aninventory for measuring depression. Arch Gen Psych 1961; 4:561–571.127. Beck AT, Steer RA. Manual for the Beck Depression Inven-tory. Psychological Corporation: New York, 1987.128. Biddle SJ. Emotion, mood, and physical activity. In: BiddleSJ, Fox KR, Boutcher SH (eds). Physical Activity and Psycholog-ical Well-Being. Routledge: London, 2000, pp. 63–87.129. Rapoport L, Clark M, Wardle J. Evaluation of a modifiedcognitive-behavioural programme for weight management. Int JObes Relat Metab Disord 2000; 24: 1726–1737.130. Wadden TA, Vogt RA, Andersen RE, Bartlett SJ, Foster GD,Kuehnel RH, Wilk J, Weinstock R, Buckenmeyer P, Berkowitz RI,Steen SN. Exercise in the treatment of obesity: effects of fourinterventions on body composition, resting energy expenditure,appetite, and mood. J Consult Clin Psychol 1997; 65: 269–277.131. Cuntz U, Leibbrand R, Ehrig C, Shaw R, Fichter MM.Predictors of post-treatment weight reduction after in-patientbehavioral therapy. Int J Obes Relat Metab Disord 2001;25(Suppl. 1): S99–S101.132. Shipman WG, Plesset MR. Predicting the outcome for obesedieters. J Am Diet Assoc 1963; 42: 383–386.133. Hathaway SR, McKinley JC. The Minnesota MultiphasicPersonality Inventory. University of Minnesota Press: Minneapo-lis, MN, 1943.134. Taylor JA. A personality scale of manisfest anxiety. JAbnorm Soc Psych 1953; 48: 285–290.135. Faith MS, Matz PE, Jorge MA. Obesity–depression associa-tions in the population. J Psychosom Res 2002; 53: 935–942.136. Ng DM, Jeffery RW. Relationships between perceived stressand health behaviors in a sample of working adults. Health Psy-chol 2003; 22: 638–642.137. Black DR, Gleser LJ, Kooyers KJ. A meta-analytic evaluationof couples weight-loss programs. Health Psychol 1990; 9: 330–347.
64 Psychosocial predictors of weight loss P. J. Teixeira et al. obesity reviews
© 2005 The International Association for the Study of Obesity. obesity reviews 6, 43–65
138. Wing RR, Gorin AA. Behavioral techniques for treating theobese patient. Prim Care 2003; 30: 375–391.139. Rieder S, Ruderman A. The development and validation ofthe weight management support inventory. Obes Res 2003;11(Suppl.): A96 (abstract 394-P).140. Wadden TA, Phelan S. Assessment of quality of life in obeseindividuals. Obes Res 2002; 10(Suppl. 1): 50S–57S.141. Kolotkin RL, Meter K, Williams GR. Quality of life andobesity. Obes Rev 2001; 2: 219–229.142. Jakicic JM. The role of physical activity in prevention andtreatment of body weight gain in adults. J Nutr 2002; 132: 3826S–3829S.143. Jakicic JM, Wing RR, Winters-Hart C. Relationship of phys-ical activity to eating behaviors and weight loss in women. MedSci Sports Exerc 2002; 34: 1653–1659.144. Hulens M, Vansant G, Claessens AL, Lysens R, Muls E,Rzewnicki R. Health-related quality of life in physically active andsedentary obese women. Am J Human Biol 2002; 14: 777–785.145. Sallis JF, Hovell MF. Determinants of exercise behavior.Exerc Sport Sci Rev 1990; 18: 307–330.146. Sherwood NE, Jeffery RW. The behavioral determinants ofexercise: implications for physical activity interventions. Annu RevNutr 2000; 20: 21–44.147. King AC, Kiernan M, Oman RF, Kraemer HC, Hull M, AhnD. Can we identify who will adhere to long-term physical activity?Signal detection methodology as a potential aid to clinical decisionmaking. Health Psychol 1997; 16: 380–389.148. Vallerand RJ. A hierarchical model of intrinsic and extrinsicmotivation in sport and exercise. In: Roberts GC (ed.). Advancesin Motivation Is Sport and Exercise. Human Kinetics: Champaign,IL, 2001, pp. 263–319.149. Conner M, Norman P, Bell R. The theory of planned behav-ior and healthy eating. Health Psychol 2002; 21: 194–201.150. Bryan AD, Rocheleau CA. Predicting aerobic versus resis-tance exercise using the theory of planned behavior. Am J HealthBehav 2002; 26: 83–94.151. Baranowski T, Cullen KW, Baranowski J. Psychosocial cor-relates of dietary intake: advancing dietary intervention. Annu RevNutr 1999; 19: 17–40.152. Godin G. Theories of reasoned action and planned behavior:usefulness for exercise promotion. Med Sci Sports Exerc 1994; 26:1391–1394.153. Sallis JF, Grossman RM, Pinski RB, Patterson TL, NaderPR. The development of scales to measure social support for dietand exercise behaviors. Prev Med 1987; 16: 825–836.154. Steinhardt MA, Dishman RK. Reliability and validity ofexpected outcomes and barriers for habitual physical activity. JOccup Med 1989; 31: 536–546.155. Wadden TA, Sarwer DB. Behavioral treatment of obesity:new approaches to an old disorder. In: Goldstein DJ (ed.). TheManagement of Eating Disorders and Obesity. Humana Press Inc.:Totowa, NJ, 1999, pp. 173–199.156. Schlundt DG, Taylor D, Hill JO, Sbrocco T, Pope-Cordle J,Kasser T, Arnold D. A behavioral taxonomy of obese femaleparticipants in a weight-loss program. Am J Clin Nutr 1991; 53:1151–1158.157. Kayman S, Bruvold W, Stern JS. Maintenance and relapseafter weight loss in women: behavioral aspects. Am J Clin Nutr1990; 52: 800–807.158. Yass-Reed EM, Barry NJ, Dacey CM. Examination of pre-treatment predictors of attrition in a VLCD and behavior therapyweight-loss program. Addict Behav 1993; 18: 431–435.159. Torgerson JS, Lissner L, Lindroos AK, Kruijer H, SjostromL. VLCD plus dietary and behavioural support versus support
alone in the treatment of severe obesity. A randomised two-yearclinical trial. Int J Obes Relat Metab Disord 1997; 21: 987–994.160. Rodin J, Elias M, Silberstein LR, Wagner A. Combinedbehavioral and pharmacologic treatment for obesity: predictors ofsuccessful weight maintenance. J Consult Clin Psychol 1988; 56:399–404.161. Heshka S, Zhu S, Miller-Kovach K. Predictive value of moti-vation scales for weight loss. Obes Res 2003; 11(Suppl.): A103.162. Kraemer HC, Wilson GT, Fairburn CG, Agras WS. Media-tors and moderators of treatment effects in randomized clinicaltrials. Arch General Psychiatry 2002; 59: 877–883.163. Bauman AE, Sallis JF, Dzewaltowski DA, Owen N. Towarda better understanding of the influences on physical activity: therole of determinants, correlates, causal variables, mediators, mod-erators, and confounders. Am J Prev Med 2002; 23: 5–14.164. Heymsfield SB, van Mierlo CA, van der Knaap HC, Heo M,Frier HI. Weight management using a meal replacement strategy:meta and pooling analysis from six studies. Int J Obes Relat MetabDisord 2003; 27: 537–549.165. Foster GD. Nondieting approaches. In: Fairburn CG,Brownell KD (eds). Eating Disorders and Obesity: A Comprehen-sive Handbook. Guilford Press: New York, 2002, pp. 604–609.166. Latner JD, Wilson GT, Stunkard AJ, Jackson ML. Self-helpand long-term behavior therapy for obesity. Behav Res Ther 2002;40: 805–812.167. Tate DF, Jacikic JM, Wing RR. The relationship betweenbaseline depression, exercise participation, and long term weightloss. Med Sci Sports Exerc 2000; 32: S300.168. Weigle DS. Pharmacological therapy of obesity: past,present, and future. J Clin Endocrinol Metab 2003; 88: 2462–2469.169. Kushner RF. Medical management of obesity. Semin Gas-trointest Dis 2002; 13: 123–132.170. Cooper Z, Fairburn CG, Hawker DM. Cognitive-BehavioralTreatment of Obesity: A Clinician’s Guide. Guilford: New York,2003.171. Mustajoki P, Pekkarinen T. Maintenance programmes afterweight reduction – how useful are they? Int J Obes Relat MetabDisord 1999; 23: 553–555.172. Perri MG, Corsica JA. Improving the maintenance of weightlost in behavioral treatment of obesity. In: Wadden TA, stunkardAJ (eds). Handbook of Obesity Treatment. Guilford Press: NewYork, NY, 2002, pp. 357–379.173. Phelan S, Wadden TA. Combining behavioral and pharma-cological treatments for obesity. Obes Res 2002; 10: 560–574.174. Wadden TA, Foster GD, Letizia KA, Stunkard AJ. A multi-center evaluation of a proprietary weight reduction program forthe treatment of marked obesity. Arch Intern Med 1992; 152:961–966.175. Greenway FL, Bray GA, Marlin RL. Methods to maximizeretention in weight loss studies. Obes Res 1999; 7: 593–596.176. Ware JH. Interpreting incomplete data in studies of diet andweight loss. N Engl J Med 2003; 348: 2136–2137.177. Liu G, Gould AL. Comparison of alternative strategies foranalysis of longitudinal trials with dropouts. J Biopharm Stat2002; 12: 207–226.178. Gadbury GL, Coffey CS, Allison DB. Modern statisticalmethods for handling missing repeated measurements in obesitytrial data: beyond LOCF. Obes Rev 2003; 4: 175–184.179. Davis MJ, Addis ME. Predictors of attrition from behavioralmedicine treatments. Ann Behav Med 1999; 21: 339–349.180. Honas JJ, Early JL, Frederickson DD, O’Brien MS. Predic-tors of attrition in a large clinic-based weight-loss program. ObesRes 2003; 11: 888–894.
obesity reviews Psychosocial predictors of weight loss P. J. Teixeira et al. 65
© 2005 The International Association for the Study of Obesity. obesity reviews 6, 43–65
181. Clark MM, Niaura R, King TK, Pera V. Depression, smok-ing, activity level, and health status: pretreatment predictors ofattrition in obesity treatment. Addict Behav 1996; 21: 509–513.182. Ek A, Andersson I, Barkeling B, Rossner S. Obesity treat-ment and attrition: no relationship to obesity level. Obes Res1996; 4: 295–296.183. Jeffery RW, Wing RR, Sherwood NE, Tate DF. Physicalactivity and weight loss: does prescribing higher physical activitygoals improve outcome? Am J Clin Nutr 2003; 78: 684–689.184. Rossner S. Factors determining the long-term outcome ofobesity treatment. In: Bjorntorp P, Brodoff BN (eds). Obesity.Lippincott: Philadelphia, PA, 1992, pp. 712–719.185. Kushner RF, Foster GD. Obesity and quality of life. Nutri-tion 2000; 16: 947–952.186. Hill JO, Billington CJ. It’s time to start treating obesity. AmJ Cardiol 2002; 89: 969–970.187. Kiernan M, Kraemer HC, Winkleby MA, King AC, TaylorCB. Do logistic regression and signal detection identify differentsubgroups at risk? Implications for the design of tailored interven-tions. Psychol Methods 2001; 6: 35–48.188. Kreuter MW, Bull FC, Clark EM, Oswald DL. Understand-ing how people process health information: a comparison of tai-lored and nontailored weight-loss materials. Health Psychol 1999;18: 487–494.189. Sbrocco T, Nedegaard RC, Stone JM, Lewis EL. Behavioralchoice treatment promotes continuing weight loss: preliminaryresults of a cognitive-behavioral decision-based treatment for obe-sity. J Consult Clin Psychol 1999; 67: 260–266.190. Pekkarinen T, Mustajoki P. Comparison of behavior therapywith and without very-low-energy diet in the treatment of morbidobesity. A 5-year outcome. Arch Intern Med 1997; 157: 1581–1585.191. Astrup A, Buemann B, Gluud C, Bennett P, Tjur T, Chris-tensen N. Prognostic markers for diet-induced weight loss in obesewomen. Int J Obes Relat Metab Disord 1995; 19: 275–278.192. Cooper Z, Fairburn CG. A new cognitive behaviouralapproach to the treatment of obesity. Behav Res Ther 2001; 39:499–511.193. Senekal M, Albertse EC, Momberg DJ, Groenewald CJ,Visser EM. A multidimensional weight-management program forwomen. J Am Diet Assoc 1999; 99: 1257–1264.194. Baranowski T, Cullen KW, Nicklas T, Thompson D, Bara-nowski J. Are current health behavioral change models helpful inguiding prevention of weight gain efforts? Obes Res 2003;11(Suppl. 1): 23S–43S.195. Gould JE. Concise Handbook of Experimental Methods forthe Behavioral and Biological Sciences. CRC Press: Boca Raton,FL, 2002.196. Fox KR. The influence of physical activity on mental well-being. Public Health Nutr 1999; 2: 411–418.197. Friedman KE, Reichmann SK, Costanzo PR, Musante GJ.Body image partially mediates the relationship between obesityand psychological distress. Obes Res 2002; 10: 33–41.198. Adami GF. The influence of body weight on food and shapeattitudes in severely obese patients. Int J Obes Relat Metab Disord2001; 25(Suppl. 1): S56–S59.
199. Wardle J, Steptoe A, Oliver G, Lipsey Z. Stress, dietaryrestraint and food intake. J Psychosom Res 2000; 48: 195–202.200. Allison DA (ed.). Handbook of Assessment Methods forEating Behaviors and Weight-Related Problems: Measures, The-ory and Research. Sage Publications: Thousand Oaks, CA, 1995.201. St. Jeor ST (ed.). Obesity Assessment: Tools, Methods, Inter-pretations. Chapman & Hall Series in Clinical Nutrition. (ed.Chernoff R). Chapman & Hall: New York, NY, 1997.202. Faith MS, Allison DB. Assessment of psychological statusamong obese persons. In: Thompson JK (ed.). Body Image, EatingDisorders, and Obesity. An Integrative Guide for Assessment andTreatment. American Psychological Association: Washington, DC,1996, pp. 365–388.203. Thompson JK, Tantleff-Dunn S. Assessment of body imagedisturbance in obesity. Obes Res 1998; 6: 375–377.204. Kraemer HC. Sample size: when is enough enough? Am JMed Sci 1988; 296: 360–363.205. Poston WS II, Ericsson M, Linder J, Nilsson T, GoodrickGK, Foreyt JP. Personality and the prediction of weight loss andrelapse in the treatment of obesity. Int J Eat Disord 1999; 25:301–309.206. Fontaine KR, Cheskin LJ. Optimism and obesity treatmentoutcomes. J Clin Psychol 1999; 55: 141–143.207. Smith MC, Sondhaus E, Porzelius LK. Effect of binge eatingon the prediction of weight loss in obese women. J Behav Med1995; 18: 161–168.208. Linne Y, Hemmingsson E, Adolfsson B, Ramsten J, Ross-ner S. Patient expectations of obesity treatment-the experiencefrom a day-care unit. Int J Obes Relat Metab Disord 2002; 26:739–741.209. Raymond NC, de Zwaan M, Mitchell JE, Ackard D, ThurasP. Effect of a very low calorie diet on the diagnostic category ofindividuals with binge eating disorder. Int J Eat Disord 2002; 31:49–56.210. Smith CF, O’Neil PM, Rhodes SK. Cognitive appraisals ofdietary transgressions by obese women: associations with self-reported eating behavior, depression, and actual weight loss. IntJ Obes Relat Metab Disord 1999; 23: 231–237.211. Cargill BR, Clark MM, Pera V, Niaura RS, Abrams DB.Binge eating, body image, depression, and self-efficacy in an obeseclinical population. Obes Res 1999; 7: 379–386.212. Westerterp-Plantenga MS, Kempen KP, Saris WH. Determi-nants of weight maintenance in women after diet-induced weightreduction. Int J Obes Relat Metab Disord 1998; 22: 1–6.213. Scheier MF, Carver CS, Bridges MW. Distinguishing opti-mism from neuroticism (and trait anxiety, self-mastery, and self-esteem): a reevaluation of the Life Orientation Test. J Pers SocPsychol 1994; 67: 1063–1078.214. Dyck MJ. Subscales of the Dysfunctional Attitudes Scale. BrJ Clin Psychol 1992; 31: 333–335.215. Schalling D, Edman G. The Karolinska Scales of Personality(KSP) Manual: An Inventory for Assessing TemperamentalDimensions Associated with Vulnerability for Psychosocial Devi-ance. Department of Psychiatry, Karolinska Institute: Stockholm,Sweden, 1993.