Predictors of Weight Loss Maintenance following an Insurance-Sponsored Weight Management Program

Abstract

Intentional weight loss among overweight and obese adults (body mass index ≥ 25 kg/m(2)) is associated with numerous health benefits, but weight loss maintenance (WLM) following participation in weight management programming has proven to be elusive. Many individuals attempting to lose weight join formal programs, especially women, but these programs vary widely in focus, as do postprogram weight regain results. We surveyed 2,106 former participants in a community-based, insurance-sponsored weight management program in the United States to identify the pre, during, and post-intervention behavioral and psychosocial factors that lead to successful WLM. Of 835 survey respondents (39.6% response rate), 450 met criteria for inclusion in this study. Logistic regression analyses suggest that interventionists should assess and discuss weight loss and behavior change perceptions early in a program. However, in developing maintenance plans later in a program, attention should shift to behaviors, such as weekly weighing, limiting snacking in the evening, limiting portion sizes, and being physically active every day.

Full text

Research Article
Predictors of Weight Loss Maintenance following an
Insurance-Sponsored Weight Management Program
Christiaan G. Abildso,1Olivier Schmid,2Megan Byrd,3Sam Zizzi,3
Alessandro Quartiroli,4and Sean J. Fitzpatrick5
1Department of Social and Behavioral Sciences, West Virginia University School of Public Health, P.O. Box 9190,
Morgantown, WV, 26506-9190, USA
2Institute of Sport Science, University of Bern, 3012 Bern, Switzerland
3College of Physical Activity and Sport Sciences, West Virginia University, Morgantown, WV 26506-6116, USA
4Department of Psychology, University of Wisconsin-La Crosse, La Crosse, WI 54601, USA
5College of Graduate and Professional Studies, John F. Kennedy University, Pleasant Hill, CA 94523-4817, USA
Correspondence should be addressed to Christiaan G. Abildso; cgabildso@hsc.wvu.edu
Received  October ; Revised  January ; Accepted  January ; Published  March 
Academic Editor: Aron Weller
Copyright ©  Christiaan G. Abildso et al. is is an open access article distributed under the Creative Commons Attribution
License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly
cited.
Intentional weight loss among overweight and obese adults (body mass index ≥ kg/m2)isassociatedwithnumeroushealth
benets, but weight loss maintenance (WLM) following participation in weight management programming has proven to be elusive.
Many individuals attempting to lose weight join formal programs, especially women, but these programs vary widely in focus, as
do postprogram weight regain results. We surveyed , former participants in a community-based, insurance-sponsored weight
management program in the United States to identify the pre, during, and post-intervention behavioral and psychosocial factors
that lead to successful WLM. Of  survey respondents (.% response rate),  met criteria for inclusion in this study. Logistic
regression analyses suggest that interventionists should assess and discuss weight loss and behavior change perceptions early in
a program. However, in developing maintenance plans later in a program, attention should shi to behaviors, such as weekly
weighing, limiting snacking in the evening, limiting portion sizes, and being physically active every day.
1. Introduction
Intentional weight loss among overweight and obese adults
(body mass index ≥ kg/m2)isassociatedwithnumer-
ous health benets. Reviews of the literature suggest that
diet-plus-physical activity weight loss interventions produce
greater weight losses than diet-only interventions [,]. How-
ever, weight loss maintenance (WLM) continues to be the
Achilles heel of many such interventions, with postprogram
weight regains in diet-plus-physical activity lifestyle interven-
tions of generally % by one year aer intervention [,]. A
systematic review of studies published between  and 
suggests that  to % of lifestyle intervention participants
achieve intentional WLM [,]. is wide variation in rate of
“successful losers” is primarily accounted for by inconsistent
denitions of WLM, which is commonly conceptualized as
a combination of achieving a specied minimum weight loss
and sustaining it over a certain period of time [,].
Various criteria have been used in the literature to deter-
mine successful WLM, including the duration of the active
weight loss and weight maintenance phases, the amount of
weight loss during the active and maintenance phases, the
types of interventions, and the times of assessment. Main-
taining a –% weight loss has been shown to have clinically
signicant health benets [,]andanincreaseby%,the
likelihood of successful maintenance over ve years []. An
even lower amount of initial weight loss may have additional
benets for WLM, as higher amounts of weight loss do not
improve the prediction of WLM []andmaybeassociated
with weight regain, cycling, yo-yo dieting, and ill health
Hindawi Publishing Corporation
Journal of Obesity
Volume 2014, Article ID 736080, 12 pages
http://dx.doi.org/10.1155/2014/736080

Journal of Obesity
[–]. Stevens and colleagues [] further recommended
that a weight change of ±% is to be considered weight
maintenance, weight changes ranging from % to % is to be
considered small weight uctuations, and weight loss of >%
is to be considered clinically signicant. Despite the lack of
denitional consensus, adopting more inclusive denitions of
weight loss maintenance that allows for some regain following
loss appears to provide participants with the most health
benets [,].
Methodological concerns notwithstanding, a variety of
behavioral and psychosocial predictors have been identied
to account for successful WLM. Individuals who have suc-
cessfully achieved self-directed WLM have been found to be
more physically active during their period of weight loss than
their unsuccessful peers [,]. In addition, eating behaviors
such as consuming breakfast regularly, reducing portion size,
and limiting snacking have been found to predict lower
caloric intake [–]. Self-monitoring strategies, such as
keeping a food and exercise log and frequent weighing,
have also been found to be critical for WLM [,,,
,]. Psychosocial predictors have commonly included
receivingsocialsupportfromaweightmaintenancegroup
or friends, but the benecial impact of spousal participation
has remained inconsistent []. In addition to the ndings
about social support, autonomy and self-reliance have also
predicted successful WLM [].
National Weight Control Registry research suggests that
themajorityofsuccessfulweightlossmaintainers,especially
women, participate in a formal program to achieve initial
weight loss [] and keep using the behavior change strategies
learned during the interventions aer the intervention is
completed []. Large corporations and health insurance
companies worldwide have a key role to play in incentivizing
participation in weight management programming and have
started investing in such initiatives [,]. erefore, den-
ing the most eective in-program strategies to prevent weight
regain following weight loss is critical for weight management
interventions []. e purpose of this study was to iden-
tify the pre-, during, and postintervention behavioral and
psychosocial factors that lead to successful WLM following
participation in an insurance-sponsored diet-plus-physical
activity community-based intervention.
2. Methods
2.1. Participants. West Virginia Public Employees Insurance
Agency (PEIA) members that enrolled in PEIA’s weight
management program (WMP) benet between April , ,
andJune,,wererecruitedtocompleteaprogram
evaluation and postprogram health behavior survey in Febru-
ary  (𝑁 = 2,106). e enrollment dates were chosen
to ensure that all participants contacted had the time to
have completed at least six months of the WMP by the
time study recruitment began. A full evaluation and details
of the WMP [,]areavailable.Briey,theWMPisan
insurance benet that provides access to exercise and nutri-
tion professionals for a small monthly copayment at private
exercise facilities with intervention services decreasing as
participants progress through the program of up to two years
(see Table  for details). Facilities are reimbursed by PEIA
for services provided using a predetermined fee schedule,
and participant progress is tracked by care management
nurses. A -pound weight loss is expected of participants
by the end of month  of the WMP. Otherwise, no weight
loss, calorie intake, or physical activity goal is mandated or
strictly enforced. Participants may also be removed from the
program for noncompliance with the following behavioral
expectations: exercising at their site at least twice per week;
turning in food logs periodically; attending appointments
with the exercise physiologist, registered dietitian, and per-
sonal trainer; and having monthly body measurements taken
by site sta. Professional exercise and nutrition services are
provided following relevant guidelines for weight loss and
maintenance (e.g., American College of Sports Medicine,
American Dietetic Association).
is study was approved by the West Virginia University
Institutional Review Board. Using a modied version of
Dillman’s [] recruiting method; eligible participants (𝑁=
2,106) were contacted by mail and/or email up to ve times
overthecourseofsixweekstocompleteaprogramevaluation
and postprogram health behavior survey (see Figure ). All
, eligible participants were sent a letter in February 
notifyingthemthatasurveywouldbeforthcomingandwere
sent a follow-up by email (𝑛 = 1,056)ormail(𝑛 = 1,050)
with a link to, or a hard copy of, the survey depending on
the availability of a valid email address. ose with a valid
email address were sent two reminders before being mailed
ahardcopyofthesurvey.Surveysweremailedtothosewith
invalid email addresses (𝑛 = 332). ese participants, and
those without an email address, were sent a follow-up letter
within three weeks of receiving the hard copy of the survey if
theyhadnotlledoutandreturnedthesurvey.Toencourage
participation, the opportunity to enter a random drawing
for  recipients to receive a free health-related book was
oered.
2.2. Instrumentation. Participants were asked to complete a
program evaluation survey containing a mix of open-ended
and closed-ended items in sections categorized chronolog-
ically as they related to the WMP (i.e., pre-, during, and
postprogram). Each section had a prompt to ensure the
respondent was evaluating the correct time period (e.g., “e
next set of questions asks you about your participation in the
Program”). e survey sections and items pertinent to this
study of WLM are described in detail below.
2.2.1. Preprogram Factors. e rst section of the survey con-
tained items assessing preprogram factors including demo-
graphic information, physical activity and weight loss history,
and bariatric surgery intention.
Demographic Information. Survey items assessed demo-
graphicinformationincludingage,gender,maritalstatus,and
number of dependents in the home (i.e., caregiver status).
Basedonresponsedistribution,agewascategorizedas<45
years,45–54.9 years,or≥55 years; marital status was cate-
gorized as married or unmarried (single/divorced/widowed).

Journal of Obesity 
T : Minutes of services per participant and monthly reimbursement made by the insurer during the weight management program.
Service Phase I (months –) Phase II (months –) Phase III (months –)
M M M M M M M M M M M M M M–
Registered dietitian  —  — —  — —  — — —  —
Fitness assessment  —  — —  — —  — — —  —
Personal training              /mo
Member copayment              maxa maxa
Agency payment to facility . . .           maxa maxa
a: member copayment (and insurance agency payment to the facility) during months – is one-half of the facility’s maximum published private membership
fee up to a maximum of .
Premailing
N = 2106
Email 1
n = 1056
Email returned
n = 332
Mailed survey (n = 1382)
Survey complete (n = 420; 30.4%)
Survey complete (n = 3; 8.3%)
Mailed survey (n=36)
Email 1received (n=726)
Survey complete (n = 246; 33.9%)
Survey incomplete-mailed (n = 17; 2.3%)
Asked for mailed survey (n = 19; 2.6%)
No response ( n = 444; 61.2%)
Survey complete (n=0)
Mailed survey (n=4)
Email 2received (n = 444)
Survey complete (n = 25; 5.6%)
Survey incomplete-mailed (n = 3; 0.7%)
Asked for mailed survey (n = 1; 0.2%)
No response ( n = 412; 92.8%)
Opted out (n = 3; 0.7%)
Survey complete (n=48;13.0%)
Mailed survey (n = 368)
Email 3received (n=412)
Survey complete (n = 59; 14.3%)
Survey incomplete-mailed (n = 16; 3.9%)
No response ( n = 352; 85.4%)
Opted out (n = 3; 0.7%)
Survey complete (n = 471; 26.3%)
Total emailed (n= 724)
Survey complete (n = 330; 45.6%)
Survey incomplete-mailed (n = 36; 4.7%)
Asked for mailed survey (n = 20; 2.8%)
Opted out (n = 6; 0.8%)
Survey complete (n = 801; 38.0%)
Total mailed (n=1790)
Total (N = 2106)
F : Sample phases and response rates.

Journal of Obesity
Caregiver status was determined by using the number of
dependents in the home item to categorize the respondent as
acaregiver (one or more dependents in the home) or noncar-
egiver (zero dependents in the home). Race and employment
statuses were not used as predictors because over % of
participants are white (reective of the population of West
Virginia) and full-time employees eligible for this insurance
benet.
Physical Activity and Weight Loss History.Physicalactivity
was determined using condensed versions of Behavioral
Risk Factor Surveillance System (BRFSS) physical activity
module items []. Separate items queried participants to
retrospectively assess the number of days in a usual week
that they did  or more minutes of moderate physical
activity (MPA) and  or more minutes of vigorous physical
activity (VPA) in the six months prior to entering the WMP.
Becauseoftheretrospectivenatureofthisitem,responses
were categorized into sedentary ( z e r o M PA and VPA) ver s u s
any activity (nonzero MPA or VPA). e number of weight
loss attempts was used to assess weight loss history. Responses
were categorized into quartiles for analysis (<5, 5–9, 10–19, or
≥20). Participants were also asked if they were considering
bariatric surgery before joining the WMP (yes/no).
2.2.2. In-Program Factors. In the second section of the sur-
vey, respondents were asked to evaluate in-program factors
including perceptions of weight loss, eort, and success and
diculty of health behavior change and maintenance as they
progressed through Phase I of the program (months –) and
beyond.
Perception of Weight Loss, Eort, and Success. To understand
perception of initial weight loss, participants were asked to
rate their weight loss during Phase I as Excellent, Good,
Acceptable, Poor,orDisappointing. is was condensed based
on response distribution as Excellent/Good, Acceptable, or
Poor/Disappointing. In addition, they were asked to provide
numerical ratings for their perceived eort during Phase
Ifrom(least)to(most) and success during Phase 
from  (worst)to(most). Based on prior research [],
these responses were compared and condensed into three
categories for analysis (success >eort,success = eort,or
success <eort).
Perceived Diculty of Health Behavior Change and Mainte-
nance. Perceived diculty of initial health behavior change
was assessed using multiple items to rate the diculty
of losing weight, changing diet routine, and starting an
exercise routine during Phase I on a six-point scale from 
(extremely easy)to(extremely dicult). Perceived diculty
of maintaining these health behavior changes was assessed
similarly, using items to rate the diculty of sticking with diet
changes and continuing an exercise routine beyond Phase I
on a six-point scale from  (extremely easy)to(extremely
dicult). Responses were split at the midpoint to dichotomize
the variables to easy (responses –) versus dicult (responses
–).
2.2.3. Postprogram Factors. In the nal section of the survey,
we assessed postprogram factors (i.e., current health behav-
iors). ese included current physical activity level, weight
management behaviors, food management strategies, and
current height and weight.
Current Physical Activity. Physical activity was assessed using
condensed versions of items from the physical activity mod-
ule of the BRFSS []. Separate items asked the respondent
to assess the number of days in a usual week; they did
 or more minutes of MPA and  or more minutes of
VPA. Responses were categorized into sedentary (zero MPA
and VPA), insuciently active (not meeting MPA or VPA
guidelines), or suciently active (meeting MPA and/or VPA
guidelines).
Weight M anagement Beha v i o r s .Behaviorsassociatedwith
WLM were assessed in the instrument, including frequency
of self-weighing (never, <1timeperweek,weeklybutnotdaily,
and daily), current method of weight loss (not currently trying
to lose weight, activity or diet alone, and activity and diet in
combination), frequency of eating breakfast (daily, not daily),
logging physical activity (yes, no), and currently exercising at
a gym, or WMP facility (yes, no).
Food Management Behaviors. Seven behavioral food strate-
gies to maintain weight associated with WLM were also
assessed in the instrument by asking the respondent to
endorse which strategies they were currently using. ese
strategies included counting calories, limiting the amount
of fat consumed, eating out less oen, limiting portion size
at meals, keeping a food log or journal, limiting soda and
sweetened drinks, and limiting snacking in the evening. All
were coded as yes/no based on respondent endorsement or
not.
2.3.WeightData,LengthofTimeinProgram,andLength
of Time aer Program. Fitness and exercise professionals at
facilities measure participant data monthly, including height,
weight, and body mass index (BMI). Each site determines its
measurement protocols on the basis of available instruments
and sta training. While protocols and instrumentation may
vary across facilities, they do not vary within facilities over
time. Data are entered into a secure database from which
data were extracted for the current study. Baseline and nal
program measurements were used to calculate baseline BMI
(25–29.9, 30–34.9, 35–39.9, and ≥40 kg/m2)andpercentage
of baseline weight lost during the program. In-program
weight loss was categorized as clinically signicant (≥5%)
or nonclinically signicant (<5%) [–]. Further, because
eachdatapointhasadateassociatedwithit,thesedata
were used to calculate the length of time each participant
remained in the program. Six months is generally the point
at which habits are formed [], the common length of
weight management interventions, and the point at which
weight loss peaks in these interventions [–]. Further,
the WMP moves to a minimal “maintenance” intervention
period (Phase III) aer the th month. us, length of time
in the program was classied as ≤6months,>6–12 months,or

Journal of Obesity 
>12 months. Lastly, the nal measurement date and the date of
thesurveyresponsewereusedtocalculatethepostprogram
time, classied as ≤6months,>6–12 months,>12–24 months,
or >24 months.
2.4. Analyses. Statistical analyses were conducted using SPSS
version .. Comparisons of successful maintainers (SM)
and unsuccessful maintainers (UM) of weight loss were
conducted using independent samples 𝑡-tests for continuous
dependent variables or chi-square analyses for categorical
dependent variables. Forward stepwise logistic regression
analysis,aneectiveexploratorytechnique[,], was
conducted to identify the predictors of WLM, our outcome
of interest. We operationalized WLM as any participant that
met the following criteria: (a) lost any amount of weight
during the WMP, (b) maintained that weight loss or regained
<% of postprogram weight during the time from program
end to survey completion, and (c) achieved overall weight
loss during the preprogram to survey completion time point.
As called for in recent literature, this is a very inclusive
operationalization of WLM which allows for moderate short-
term losses that may lead to greater losses over the longer
term, excludes extreme weight loss changes, and allows for
minimal regain postintervention [,].
Four regression models were run to determine factors to
include in a nal predictive model. Repeated contrasts were
used for each predictor variable in each model. is method
compares each category of a predictor (except the rst) to the
previous category. us, contrasts include categories  versus
,  versus , and  versus , rather than the simple contrasts
of categories  versus ,  versus ,  versus , and so on. is
allows for pinpointing specic frequencies of behaviors, such
as self-weighing and amount of PA, predictive of WLM.
Model A (preprogram) included four demographic fac-
tors (age, gender, marital status, and caregiver status), phys-
ical activity level, whether weight loss surgery was being
considered or not, and objectively measured baseline BMI.
Model B (in-program) consisted of nine factors, including
perception of weight loss, dierence between perceived eort
and success, ve perceived diculties of health behavior
change items, and objectively measured percentage weight
loss and length of time in the program. Model C (postpro-
gram ) included seven factors, specically length of time
from program end to survey completion date, frequency
of self-weighing, current method of weight loss, frequency
of eating breakfast, logging physical activity, and currently
exercising at a gym or WMP facility. Model D (postprogram
) consisted of seven food management behavioral factors
including counting calories, limiting the amount of fat con-
sumed, eating out less oen, limiting portion size at meals,
keeping a food log or journal, limiting soda and sweetened
drinks, and limiting snacking in the evening.
Predictors signicant at the 𝑃 < 0.05 level from Models
A–D were included in the nal model. Odds ratios (ORs) and
% condence intervals (CI) are reported for successfully
achieving WLM as operationalized in this study. Because
we used repeated contrasts for each predictor variable, ORs
should be interpreted as the change in the likelihood of
being a successful maintainer (SM) that results in a one-unit
increase in the predictor variable. us an OR >shouldbe
interpreted as an increase, and OR < should be interpreted
as a decrease, in the likelihood of being a SM with a one-unit
increase in the predictor.
3. Results
3.1. Response and Baseline Data. Atotalofsurveyswere
received (.% response rate),  of which were complete.
From these,  completed surveys were removed because
they did not have a weight measurement following baseline
(𝑛=26), had a baseline BMI < kg/m2(𝑛=3), became
pregnant during the program (𝑛=4), had bariatric surgery
postprogram (𝑛=7), did not report a current weight (𝑛=
21), were still active in the program when they completed
the survey (𝑛 = 154), were <monthaerprogramatthe
time of survey completion (𝑛=39), were duplicate entries
from the same individual across survey platforms (𝑛=2),
or had gained weight during the program (𝑛=95). e
resulting analytic sample size was 𝑁 = 450. Our sample was
largely females (.%), married (.%), and  years or older
(.%).
Nearly half of the respondents were successful at WLM
(𝑛 = 210, .%). Independent samples 𝑡-tests showed that
SM and UM did not achieve signicantly dierent percentage
weight loss during the program (.% versus .%; 𝑃 = 0.157)
but did achieve signicantly dierent weight change aer
programandoverallfrompreprogramtocurrenttime(𝑃<
0.001). In fact, SM lost .% of end program weight and lost
.% of preprogram weight overall, a clinically signicant
loss []. In comparison, UM gained .% postprogram and
gained .% from preprogram. A greater percentage of SM
were meeting PA guidelines aer program than UM (.%
versus .%), and fewer SM than UM were insuciently
active (.% versus .%) or sedentary (.% versus .%)
aer program (𝜒2= 19.000;𝑃 < 0.001).
3.2. Predictors of WLM. Items included in, and results of,
Models A–D are presented in Tables ,,,and. Please
note that the size of the analytic sample in each model varies
because SPSS performs a listwise deletion of missing data
when running logistic regression. us, if there is a missing
value for any variable in the model, the entire case is excluded
from the analysis. Tables –present all potential predictor
variables in the order in which the repeated contrasts were
conducted. e Wald chi-square statistic, which indicates
whether 𝛽for each variable is signicantly dierent than
zero and the variable is a signicant predictor of weight loss
maintenance [],isreportedforallvariables,butanORis
only reported for signicant predictors. Larger values of the
Wald statistic indicate a variable more likely to be a signicant
predictor of the outcome.
Preprogram physical activity level was the only signicant
predictor of WLM from regression Model A, with SM more
likely to have been getting any physical activity before WMP
than UM (OR = ., % CI = .–.). Regression Model
B revealed that respondents completing >– months of
the program were less likely to be a SM than those that

Journal of Obesity
T : Preprogram predictors of weight loss maintenance (𝑁 = 404)—Model A.
𝑛𝛽Wald 𝜒2OR (% CI)
Marital status
Single/divorced/widowed  —
Married  .
Gender
Female  —
Male  .
Caregiver
No  —
Yes  .
Considering bariatric surgery
No  —
Yes  .
Preprogram MVPA
None (sedentary)  — .
Any activity  . . . (.–.)∗
Age at the program start, years
+  —
–.  .
<  .
Baseline body mass index, kg/m
Obese III (+)  —
Obese II (–.)  .
Obese I (–.)  .
Overweight (–.)  .
Weight loss attempts
≥  —
–  .
–  .
<  .
Note: MVPA: moderate-to-vigorous physical activity ∗𝑃 < 0.05;∗∗𝑃 < 0.01;∗∗∗𝑃 < 0.001.
e Wald 𝜒2statistic, which indicates whether 𝛽for each variable is signicantly dierent than zero, and the variable is a signicant predictor of weight loss
maintenance and is reported for all variables, but an OR is only reported for signicant predictors. Each variable is presented in the order in which therepeated
contrasts were conducted. us within each variable, each level moving down the rows of the table should be compared with the level of the variable in the row
immediately above it. us, ORs should be interpreted as the change in the likelihood of being a successful maintainer (SM) that results in a one-unit increase
in the predictor variable represented by a move one row down in the table.
completed at least  months (OR = ., % CI = .–
.); respondents rating Phase I weight loss as acceptable
were more likely to be SM than respondents rating weight
loss as good or excellent (OR = ., % CI = .–.);
respondents indicating it was easy to stick with diet changes
(OR=.,%CI=.–.)andeasytocontinuea
regular exercise routine (OR = ., % CI = .–.) were
more likely to be SM than those rating those changes as
dicult. Signicant predictors of SM from Model C indicate
that respondents >– months aer program were more
likely be SM than respondents >– months aer program
(OR = ., % CI = .–.); respondents insuciently
active were less likely than their suciently active peers to
be SM (OR = ., % CI = .–.); and respondents
weighing themselves less than once per week were less likely
to achieve SM than respondents weighing themselves at least
onceperweekbutnotdaily(OR=.,%CI=.–
.). Model D produced two food management behaviors
predictive of SM: limiting portion sizes (OR = ., % CI
= .–.) and limiting snacking in the evening (OR = .,
% CI = .–.).
In the nal model, six factors signicantly predicted SM
(see Table  for details), including being >– months aer
program compared to > months aer program (OR = .,
% CI = .–.); being >– months aer program
compared to >– months aer program (OR = .,
% CI = .–.); self-weighing less than once per week
compared with weekly but not daily (OR = ., % CI =
.–.); limiting snacking in the evening (OR = ., %
CI = .–.); limiting portion sizes (OR = ., % CI =
.–.); rating Phase I weight loss as acceptable compared
with good/excellent (OR = ., % CI = .–.); and

Journal of Obesity 
T : In-program predictors of weight loss maintenance (𝑁 = 428)—Model B.
𝑛𝛽Wald 𝜒2OR (% CI)
In-program weight loss
Not clinically signicant (<%)  —
Clinically signicant (≥%)  .
Months in the program
>  — .
>–  −. . . (.–.)∗
≤−. . . (.–.)
Perceived Phase I weight loss
Good/excellent  — .
Acceptable  . . . (.–.)∗∗
Poor/disappointing  −. . . (.–.)
Perceived Phase I eort/success balance
Success <eort  —
Success =eort  .
Success >eort  .
Perceived diculty to
Start an exercise routine
Dicult to extremely dicult  —
Easy to extremely easy  .
Change diet
Dicult to extremely dicult  —
Easy to extremely easy  .
Lose weight
Dicult to extremely dicult  —
Easy to extremely easy  .
Continue regular exercise routine
Dicult to extremely dicult  — .
Easy to extremely easy  . . . (.–.)∗∗
Stick with diet changes
Dicult to extremely dicult  — .
Easy to extremely easy  . . . (.–.)∗∗
Note: ∗𝑃 < 0.05;∗∗𝑃 < 0.01;∗∗∗𝑃 < 0.001.
e Wald 𝜒2statistic, which indicates whether 𝛽for each variable is signicantly dierent than zero, and the variable is a signicant predictor of weight loss
maintenance and is reported for all variables, but an OR is only reported for signicant predictors. Each variable is presented in the order in which therepeated
contrasts were conducted. us, within each variable, each level moving down the rows of the table should be compared with the level of the variable in therow
immediately above it. us, ORs should be interpreted as the change in the likelihood of being a successful maintainer (SM) that results in a one-unit increase
in the predictor variable represented by a move one row down in the table.
perceiving it to be easy to continue a regular exercise routine
as compared with dicult (OR = ., % CI = .–.).
4. Discussion
In agreement with published research [], results from this
study suggest that the likelihood of successfully maintaining
weight loss diminishes over time, peaking in our survey
respondents in the –-month postprogram timeframe and
decreasing in a stepwise fashion over time.
Preprogram physical activity level signicantly predicted
WLM, but only in the regression model that included prepro-
gram predictors. However, results from the comprehensive
predictor model of our study suggest no signicant prepro-
gram predictors of WLM. is is a positive nding from a
population-based perspective in that it shows the program
works similarly in a real-world environment with people of
varying demographic characteristics, weight loss histories,
BMI, and PA.
Respondents who perceived their early program weight
loss as acceptable were more than twice as likely to achieve
WLMasthosewhoratedtheirweightlossasgood or
excellent. is nding is similar to prior research that suggests
that unrealistic weight loss expectations are associated with
dropout from weight management programs [–]and
that program completers achieve results that closely match
preprogram expectations []. Interventionists should fre-
quently assess and discuss perception of in-program weight
loss, especially early, to make sure that participants perceive
that they are gradually meeting modest, realistic weight loss

Journal of Obesity
T : Postprogram predictors of weight loss maintenance (𝑁 = 404)—Model C.
𝑛𝛽Wald 𝜒2OR (% CI)
Months postprogram
>  —
>–  . . . (.–.)
>–  . . . (.–.)∗∗∗
≤  . . . (.–.)
Self-weighing frequency
At least once e very day  —
At least once p er week but not daily  −. . . (.–.)
Less than once per week  −. . . (.–.)∗∗∗
Never  . . . (.–.)
Current weight loss method
Using both physical activity and diet  —
Using physical activity or diet alone  .
Not currently trying to lose weight  .
Current level of physical activity
Meeting guidelines  —
Insuciently active  −. . . (.–.)∗∗
Sedentary  . . . (.–.)
Eating breakfast daily
No  —
Yes  .
Keeping a physical activity log
No  —
Yes  .
Currently exercising at a gym or WMP site
No  —
Yes  .
Note: WMP: weight management program. ∗𝑃 < 0.05;∗∗ 𝑃 < 0.01;∗∗∗𝑃 < 0.001.
e Wald 𝜒2statistic, which indicates whether 𝛽for each variable is signicantly dierent than zero, and the variable is a signicant predictor of weight loss
maintenance and is reported for all variables, but an OR is only reported for signicant predictors. Each variable is presented in the order in which therepeated
contrasts were conducted. us within each variable, each level moving down the rows of the table should be compared with the level of the variable in the row
immediately above it. us, ORs should be interpreted as the change in the likelihood of being a successful maintainer (SM) that results in a one-unit increase
in the predictor variable represented by a move one row down in the table.
goals. In contrast, individuals that perceive weight loss as
good or excellent may believe the behavior change process
to be easy, underestimating the vigilance and cognitive
restraint [] needed to maintain such changes, leading to
overcondence, dropout, and weight regain/cycling.
e perception that maintaining a regular exercise rou-
tine was easy (compared with dicult)waspredictiveof
WLM, suggesting another target for intervention. is con-
rms other research ndings [] that an individual’s self-
ecacy, or belief in their ability to accomplish a behavior
[] - in this case exercise—is important for sustaining weight
loss. SM were more likely to meet PA guidelines. However,
contrary to prior research [,,], postprogram PA level was
not predictive of WLM in the nal regression model though
our measurement of PA was dissimilar to prior studies.
Postprogram behaviors were predictive of WLM and
should be considered as education components that are
incorporated later in WMPs, reinforced with participants
upon completing WMPs, and targeted for “booster” post-
program interventions. ese included weekly weighing,
limiting snacking in the evening, and limiting portion sizes.
In concordance with prior studies [], self-weighing was pre-
dictive of SM, specically self-weighing at least once per week
as compared to less frequent weighing. In addition, it would
be benecial to work with participants to develop long-term
strategies for limiting evening snacking (e.g., brushing teeth
immediately aer dinner, drinking water instead of snacking,
and limiting the availability of snacks in the home) and
limiting portion sizes (e.g., using portion control dishware,
learning to measure portions accurately, and immediately
putting half of a dinner in a to-go box when eating out). ese
strategies can easily be gleaned from surveys or interviews
with individuals successful at WLM.
It is critical that the ndings of the current study be
viewed in the appropriate context. e program we evaluated
is a community-based, public insurance benet for working
adults in a rural state in the USA (West Virginia) that
has some of the highest rates of chronic disease in the
country. Its development was informed by evidence-based
programs (i.e., Diabetes Prevention Program) but adapted to

Journal of Obesity 
T : Postprogram predictors of weight loss maintenance (𝑁 = 450)—Model D.
𝑛𝛽Wald 𝜒2OR (% CI)
Limiting snacking in the evening
No  —
Yes  . . . (.–.)∗∗
Limiting amount of fat consumed
No  —
Yes  .
Eating out less oen
No  —
Ye s     .  
Limiting portion size at meals
No  —
Yes  . . . (.–.)∗∗∗
Keeping a food log or journal
No  —
Ye s    .  
Limiting soda or sweetened drinks
No  —
Yes  .
Counting calories
No  —
Yes  .
Note: ∗𝑃 < 0.05;∗∗𝑃 < 0.01;∗∗∗𝑃 < 0.001.
e Wald 𝜒2statistic, which indicates whether 𝛽for each variable is signicantly dierent than zero, and the variable is a signicant predictor of weight loss
maintenance and is reported for all variables, but an OR is only reported for signicant predictors. Each variable is presented in the order in which therepeated
contrasts were conducted. us, within each variable, each level moving down the rows of the table should be compared with the level of the variable in therow
immediately above it. us, ORs should be interpreted as the change in the likelihood of being a successful maintainer (SM) that results in a one-unit increase
in the predictor variable represented by a move one row down in the table.
be contextually appropriate and sustainable. Recent reviews
and meta-analyses of randomized, controlled trials (e.g.,
[]) have elucidated ndings from such work. us, context-
specic ndings of our evaluation that are incongruent with
others’ (e.g., nonsignicant PA-WLM relationship) may be a
result of the dierent setting, intervention approach, and/or
assessment methods of our work from that of RCTs, and
the limitations of our study are discussed in what follows.
First, we used an inclusive denition of WLM as the outcome
variable in agreement with recent recommendations [,]. A
more conservative denition of WLM in weight loss amount
and/or length of maintenance such as those used by the
NWCR and IOM [,] may have yielded dierent results.
Second, preprogram factors were assessed retrospectively.
ough half of the respondents began the program within
two years of this project, they may have begun up to four
years prior to the survey. is time delay may have limited
the ability of respondents to accurately assess some variables.
To address this limitation, we categorized preprogram PA
responses as none or some because it is likely that people were
able to recall the dierence between doing any or no PA rather
than specic minutes of PA in a week. ird, the majority
of predictor variables and current weight used to categorize
WLM were assessed via a self-report. Such subjective reports
of perceptions are relevant as intervention targets, but self-
report of weight and PA has inherent weaknesses. Further,
some items in our evaluation survey were taken from the
current evidence base and are specic to this evaluation but
not yet validated. Additionally, generalizability of the ndings
is limited to participants in similar insurance-sponsored
programs because a large portion of survey respondents
are white, full-time employed, and married women over 
years old. However, these demographics are similar to studies
related to other formal programs and the NWCR sample
[]. Also, our response rate (.%) may be considered less
thanideal,butitisconsistentwithameta-analysisofsurvey
research (.%) []. Sampling error was controlled by
inviting all members of the population to complete the survey,
and nonresponse error appears low because responders did
not dier from nonresponders in any of the key variables such
as BMI, program completion, or % weight loss rate.
5. Conclusions
Despite the aforementioned limitations, modiable percep-
tions and behaviors predictive of WLM that could be targets
of future interventions were identied in this study. Many
SM in the current study were able to achieve and maintain
a clinically meaningful amount of weight loss, providing
valuable guidance for other programs. e results suggest
that weight loss interventionists should change intervention
targets as participants move through a WMP, in concordance

 Journal of Obesity
T : Final model predicting weight loss maintenance (𝑁 = 428).
𝑛𝛽Wald 𝜒2OR (% CI)
Preprogram MVPA
None (sedentary)  —
Any activity  .
Perceived Phase I weight loss
Good/excellent  — .
Acceptable  . . . (.–.)∗∗
Poor/disappointing  −. . . (.–.)
Months in the program
>  —
>–  .
≤  .
Perceived diculty of sticking with diet changes
Dicult to extremely dicult  —
Easy to extremely easy  .
Perceived diculty of continuing exercise routine
Dicult to extremely dicult  — .
Easy to extremely easy  . . . (.–.)∗∗∗
Current level of physical activity
Meeting guidelines  —
Insuciently active  .
Sedentary  .
Months aer program
>  — .
>–  . . . (.–.)∗
>–  . . . (.–.)∗∗∗
≤  . . . (.–.)
Self-weighing frequency
At least once e very day   — . 
At least once p er week but not daily  −. . . (.–.)
Less than once per week  −. . . (.–.)∗∗∗
Never  . . . (.–.)
Limiting snacking in the evening
No  — .
Yes  . . . (.–.)∗∗
Limiting portion size at meals
No  — .
Yes  . . . (.–.)∗∗
Note: MVPA =moderate-to-vigorous physical activity. ∗𝑃 < 0.05;∗∗𝑃 < 0.01;∗∗∗𝑃 < 0.001.
e Wald 𝜒2statistic, which indicates whether 𝛽for each variable is signicantly dierent than zero, and the variable is a signicant predictor of weight loss
maintenance and is reported for all variables, but an OR is only reported for signicant predictors. Each variable is presented in the order in which therepeated
contrasts were conducted. us within each variable, each level moving down the rows of the table should be compared with the level of the variable in the row
immediately above it. us, ORs should be interpreted as the change in the likelihood of achieving weight loss maintenance that results in a one-unit increase
in the predictor variable represented by a move one row down in the table.
with the shi from cognitive to behavioral processes of
change as individuals progress through the stages of change
[]. Early in a program, interventionists may want to assess
and discuss weight loss and behavior change perceptions
as these may reect participant self-ecacy and, ultimately,
participant retention. As individuals progress through a
program and shi toward maintenance of weight loss, inter-
ventionists are encouraged to focus attention on behaviors
in developing maintenance plans, such as weekly weighing,
limiting snacking in the evening, limiting portion sizes, and
being physically active every day.
Conflict of Interests
e authors declare that there is no conict of interests
regarding the publication of this paper.

Journal of Obesity 
Acknowledgment
is research was funded by the State of West Virginia Public
Employees Insurance Agency (PEIA). PEIA assisted with the
development of the survey and provided the books used as
a survey completion incentive. PEIA was not involved in the
data collection or analysis and did not have to approve the
paper prior to submission.
References
[] C. C. Curioni and P. M. Lourenc¸o, “Long-term weight loss aer
diet and exercise: a systematic review,” International Journal of
Obesity,vol.,no.,pp.–,.
[] T. Wu, X. Gao, M. Chen, and R. M. van Dam, “Long-term
eectiveness of diet-plus-exercise interventions vs. diet-only
interventions for weight loss: a meta-analysis: obesity manage-
ment,” Obesity Reviews,vol.,no.,pp.–,.
[] J.C.M.Barte,N.C.W.TerBogt,R.P.Bogersetal.,“Mainte-
nance of weight loss aer lifestyle interventions for overweight
and obesity, a systematic review,” Obesity Reviews, vol. , no. ,
pp.–,.
[] K. Elag and S. R¨
ossner, “Who succeeds in maintaining weight
loss? A conceptual review of factors associated with weight loss
maintenance and weight regain,” Obesity Reviews,vol.,no.,
pp.–,.
[]G.Blackburn,“Eectofdegreeofweightlossonhealth
benets,” Obesity Research, vol. , pp. s–s, .
[] National Institutes of Health, Clinical Guidelines on the Identi-
cation, Evaluation, and Treatment of Overweight and Obesity
inAdults:eEvidenceReport, National Heart Lung and Blood
Institute, .
[] R. R. Wing and J. O. Hill, “Successful weight loss maintenance,”
Annual Review of Nutrition,vol.,pp.–,.
[] G. Cutter, S. S. Jeor, R. Brunner et al., “Methodological issues in
weight cycling,” Annals of Behavioral Medicine,vol.,no.,pp.
–, .
[] G. D. Foster, D. B. Sarwer, and T. A. Wadden, “Psychological
eects of weight cycling in obese persons: a review and research
agenda,” Obesity Research, vol. , no. , pp. –, .
[] A.-C. Vergnaud, S. Bertrais, J.-M. Oppert et al., “Weight uc-
tuations and risk for metabolic syndrome in an adult cohort,”
International Journal of Obesity,vol.,no.,pp.–,.
[] J. Stevens, K. P. Truesdale, J. E. McClain, and J. Cai, “e
denition of weight maintenance,” International Journal of
Obesity,vol.,no.,pp.–,.
[] T. A. Wadden, V. L. Webb, C. H. Moran, and B. A. Bailer,
“Lifestyle modication for obesity: new developments in diet,
physical activity, and behavior therapy,” Circulation,vol.,no.
, pp. –, .
[] M.L.Klem,R.R.Wing,M.T.McGuire,H.M.Seagle,andJ.O.
Hill, “A descriptive study of individuals successful at long-term
maintenance of substantial weight loss,” e American Journal
of Clinical Nutrition, vol. , no. , pp. –, .
[] R.W.Jeery,W.M.Bjornson-Benson,B.S.Rosenthal,C.L.
Kurth, and M. M. Dunn, “Eectiveness of monetary contracts
with two repayment schedules on weight reduction in men and
women from self-referred and population samples,” Behavior
erapy,vol.,no.,pp.–,.
[] M. Katahn, J. Pleas, M. ackrey, and K. A. Wallston, “Rela-
tionship of eating and activity self-reports to follow-up weight
maintenance in the massively obese,” Behavior erapy,vol.,
no. , pp. –, .
[] S. Kayman, W. Bruvold, and J. S. Stern, “Maintenance and
relapse aer weight loss in women: behavioral aspects,” e
American Journal of Clinical Nutrition,vol.,no.,pp.–
, .
[] A. J. Stunkard and S. Messick, “e three-factor eating question-
naire to measure dietary restraint, disinhibition and hunger,”
Journal of Psychosomatic Research,vol.,no.,pp.–,.
[] R. C. Baker and D. S. Kirschenbaum, “Self-monitoring may be
necessary for successful weight control,” Behavior erapy,vol.
,no.,pp.–,.
[] M.T.McGuire,R.R.Wing,M.L.Klemetal.,“Whatpredicts
weight regain in a group of successful weight losers?” Journal of
Consulting and Clinical Psychology,vol.,pp.–,.
[] G.C.Williams,V.M.Grow,Z.R.Freedman,R.M.Ryan,andE.
L. Deci, “Motivational predictors of weight loss and weight-loss
maintenance,” Journal of Personality and Social Psychology,vol.
, no. , pp. –, .
[] World Economic Forum Workplace Wellness Alliance, Making
the Right Investment: Employee Health and the Power of Metrics,
World Economic Forum Workplace Wellness Alliance, Geneva,
Switzerland, .
[] National Institute for Health Care Management Foundation,
Health Plans Emerging as Pragmatic Partners in Fight against
Obesity, National Institute for Health Care Management Foun-
dation, Washington, DC, USA, .
[] C. G. Abildso, S. J. Zizzi, and B. Reger-Nash, “Evaluating an
insurance-sponsored weight management program with the
RE-AIM Model, West Virginia, –,” Preventing Chronic
Disease,vol.,no.,p.A,.
[] C. G. Abildso, S. J. Zizzi, and S. J. Fitzpatrick, “Predictors of
clinically signicant weight loss and participant retention in an
insurance-sponsored community-based weight management
program,” Health Promotion Practice, vol. , pp. –, .
[] D. A. Dillman, MailandInternetSurveys:eTailoredDesign
Method, John Wiley & Sons, Hoboken, NJ, USA, nd edition,
.
[] R.C.Brownson,D.A.Jones,M.Pratt,C.Blanton,andG.W.
Heath, “Measuring physical activity with the behavioral risk
factor surveillance system,” Medicine and Science in Sports and
Exercise,vol.,no.,pp.–,.
[] C. Abildso, S. Zizzi, D. Gilleland, J. omas, and D. Bonner, “A
mixed methods evaluation of a -week insurance-sponsored
weight management program incorporating cognitive-
behavioral counseling,” Journal of Mixed Methods Research,vol.
, no. , pp. –, .
[] R. R. Wing, R. Koeske, L. H. Epstein,M. P. Nowalk, W.G ooding,
and D. Becker, “Long-term eects of modest weight loss in type
II diabetic patients,” Archives of Internal Medicine,vol.,no.
, pp. –, .
[] I. L. Mertens and L. F. van Gaal, “Overweight, obesity, and
blood pressure: the eects of modest weight reduction,” Obesity
Research,vol.,no.,pp.–,.
[] L. F. van Gaal, M. A. Wauters, and I. H. de Leeuw,“ e benecial
eects of modest weight loss on cardiovascular risk factors,”
International Journal of Obesity,vol.,no.,pp.S–S,.
[] J. O. Prochaska and W. F. Velicer, “ e transtheoretical model of
health behavior change,” American Journal of Health Promotion,
vol.,no.,pp.–,.

 Journal of Obesity
[] R. R. Wing, Behavioral Weight Control,eGuilfordPress,New
Yo r k , N Y, U S A ,     .
[] M. J. Franz, J. J. VanWorme r, A. L. Crain et al., “Weig ht-loss
outcomes: a systematic review and meta-analysis of weight-loss
clinical trials with a minimum -year follow-up,” Journal of the
American Dietetic Association,vol.,no.,pp.–,
.
[] T. A. Wadden, M. L. Butr yn, andK. J. Byrne, “Ecacy of lifestyle
modication for long-term weight control,” Obesity Research,
vol. , pp. S–S, .
[] A. P. Field, Discovering Statistics Using SPSS,Sage,London,UK,
rd edition, .
[] B. G. Tabachnick and L. S. Fidell, Using Multivariate Statistics,
Allyn & Bacon, Boston, Mass, USA, th edition, .
[] National Heart Lung Blood Institute, “Obesity education ini-
tiative expert panel: clinical guidelines on the identication,
evaluation, and treatment of overweight and obesity in adults—
the evidence report,” Obesity Research,vol.,supplement,pp.
S–S, .
[] R. W. Jeery, L. H. Epstein, G. T. Wilson et al., “Long-term
maintenance of weight loss: current status,” Health Psychology,
vol.,no.,pp.–,.
[] J. Polivy and C. Peter Herman, “If at rst you don’t succeed: false
hopes of self-change,” American Psychologist,vol.,no.,pp.
–, .
[] S. R. Sears and A. L. Stanton, “Expectancy-value constructs
and expectancy violation as predictors of exercise adherence in
previously sedentary women,” Health Psycholog y,vol.,no.,
pp. –, .
[] F. Jones, P. Harris, and H. Waller, “Expectations of an exercise
prescription scheme: an exploratory study using repertory
grids,” British Journal of Health Psychology,vol.,no.,pp.–
, .
[] F. Jones, P. Harris, H. Waller, and A. Coggins, “Adherence to
an exercise prescription scheme: the role of expectations, self-
ecacy, stage of change and psychological well-being,” British
JournalofHealthPsychology,vol.,no.,pp.–,.
[] R. R. Wing and S. Phelan, “Long-ter m weight loss maintenance,”
e American Journal of Clinical Nutrition,vol.,no.,pp.
S–S, .
[] A. Bandura, “Self-ecacy: toward a unifying theory of behav-
ioral change,” Psychological Review,vol.,no.,pp.–,
.
[] J. W. Anderson, E. C. Konz, R. C. Frederich, and C. L. Wood,
“Long-term weight-loss maintenance: a meta-analysis of US
studies,” e American Journal of Clinical Nutrition,vol.,no.
, pp. –, .
[] K. Casazza, K. R. Fontaine, A. Astrup et al., “Myths, presump-
tions, and facts about obesity,” e New England Journal of
Medicine,vol.,pp.–,.
[] Institute of Medicine (IOM), Weighing the Options: Criteria for
Evaluating Weight-Management Programs, National Academies
Press, Washington, DC, USA, .
[] C. Cook, F. Heath, and R. L. ompson, “A meta-analysis of
responseratesinWeb-orinternet-basedsurveys,”Educational
and Psychological Measurement,vol.,no.,pp.–,
.