Heshka S, Anderson JW, Atkinson RL, Greenway FL, Hill JO, Phinney SD, Kolotkin RL, Miller-Kovach K, Pi-Sunyer FX. Weight Loss With Self-help Compared With a Structured Commercial ProgramA Randomized Trial. JAMA. 2003;289(14):1792–1798. doi:10.1001/jama.289.14.1792
Author Affiliations: New York Obesity Research Center, St Luke's/Roosevelt Hospital and Columbia University, New York, NY (Drs Heshka and Pi-Sunyer); VA Medical Center, Lexington, and University of Kentucky (Dr Anderson); Nutritional Sciences, University of Wisconsin, Madison (Dr Atkinson); Pennington Biomedical Research Center, Louisiana State University, Baton Rouge (Dr Greenway); Health Sciences, University of Colorado, Boulder (Dr Hill); Department of Medicine, University of California at Davis (Dr Phinney); Obesity and Quality of Life Consulting, Durham, NC (Dr Kolotkin); and Weight Watchers International, Woodbury, NY (Ms Miller-Kovach).
Context Although commercial weight loss programs provide treatment to millions
of clients, their efficacy has not been evaluated in rigorous long-term trials.
Objective To compare weight loss and health benefits achieved and maintained through
self-help weight loss vs with a structured commercial program.
Design and Setting A 2-year, multicenter randomized clinical trial with clinic visits at
12, 26, 52, 78, and 104 weeks conducted at 6 academic research centers in
the United States between January 1998 and January 2001.
Participants Overweight and obese men (n = 65) and women (n = 358) (body mass index,
27-40) aged 18 to 65 years.
Intervention Random assignment to either a self-help program (n = 212) consisting
of two 20-minute counseling sessions with a nutritionist and provision of
self-help resources or to a commercial weight loss program (n = 211) consisting
of a food plan, an activity plan, and a cognitive restructuring behavior modification
plan, delivered at weekly meetings.
Main Outcome Measures Weight change was the primary outcome measure. Secondary outcomes included
waist circumference, body mass index, blood pressure, serum lipids, glucose,
and insulin levels.
Results At 2 years, 150 participants (71%) in the commercial group and 159 (75%)
in the self-help group completed the study. In the intent-to-treat analysis,
mean (SD) weight loss of participants in the commercial group was greater
than in the self-help group at 1 year (−4.3 [6.1] kg vs −1.3 [6.1]
kg, respectively; P<.001) and at 2 years (−2.9
[6.5] kg vs −0.2 [6.5] kg, respectively; P<.001).
Waist circumference (P = .003) and body mass index
(P<.001) decreased more in the commercial group.
Changes in blood pressure, lipids, glucose, and insulin levels were related
to changes in weight in both groups, but between-group differences in biological
parameters were mainly nonsignificant by year 2.
Conclusion The structured commercial weight loss program provided modest weight
loss but more than self-help over a 2-year period.
An increasing number of men and women in the United States are attempting
to lose weight.1,2 Formal diet
and exercise programs provide weight loss treatment to approximately 13% of
these women and 5% of the men.3 However, obese
participants in diet and exercise programs rarely achieve a body mass index
(BMI, calculated as weight in kilograms divided by the height in meters squared)
in the normal range (20-25) and because weight lost is often regained, there
is a perception that nothing is accomplished by participation in such programs.
Nevertheless, numerous studies have documented improvements in indicators
of health with modest weight loss,4- 7 which,
if maintained over several years, may be clinically significant.
Commercial weight loss programs have been recommended as the second
step in a graded, 4-step model for weight management5,8 but
their efficacy has rarely been rigorously evaluated.9- 13 We
conducted a 2-year, multicenter randomized trial to evaluate the largest provider
of commercial weight loss services in the United States.14 A
report of the 6-month results of this trial has been published previously.15
The study was a randomized, parallel-group, 2-year trial conducted at
6 US clinical centers (see author affiliations on the first page) between
January 1998 and January 2001 with scheduled visits to the study center at
weeks 0, 12, 26, 52, 78, and 104. The number of participants to be enrolled
was selected to provide a power of 0.84 to detect a difference of 2.5 kg between
treatment and control groups in the amount of weight lost at the end of 2
years based on an estimated SD of 6.5 kg in change scores, with a type I error
rate of .05, 2-tailed, if 120 participants remained in each group.
Participants were recruited from existing clinic records or by advertising
a long-term nonmedication weight loss study for moderately overweight persons.
All participants provided written informed consent. A medical history was
taken and physical examination and electrocardiography were performed at the
screening visit. Men and women with a BMI of 27 to 40, aged 18 to 65 years,
including persons with health problems for which weight reduction is a medically
accepted therapy, were eligible for the study.
Exclusion criteria were fasting glucose higher than 140 mg/dL (7.8 mmol/L),
triglycerides higher than 1000 mg/dL (11.3 mmol/L), liver function test results
(aspartate aminotransferase, alanine aminotransferase, alkaline phosphatase,
lactate dehydrogenase, γ-glutamyltransferase, and bilirubin) more than
2 times the upper normal limit, and serum creatinine higher than 1.4 mg/dL
Also excluded were potential participants using systemic or inhaled
corticosteroids or lithium, having a history of alcohol abuse within the past
year, or having a history or presence of a significant psychiatric disorder
or any condition that, in the investigator's judgment, would interfere with
participation in the trial. Potential participants who initiated a new drug
therapy within 30 days of randomization, who were already participating in
a weight loss program, or who took prescription weight loss or investigational
medications within 90 days of randomization were also excluded. If clinically
significant abnormalities were found at screening, individuals were referred
to their personal physician. The protocol was reviewed and approved annually
by the institutional review board at each site.
After all screening test results were reviewed and the candidate's eligibility
confirmed, a randomization envelope prepared by the data coordinating center
was opened and the participant was assigned to self-help or the commercial
program (Figure 1). Participants
were block randomized as determined by a random number table. A different
block and randomization sequence was prepared for each site. Participants
and investigators at each site were blind to the assignment condition until
the envelope was opened.
Participants assigned to the self-help group received 20-minute consultations
with a dietitian at the week 0 (baseline) and week 12 visits and were given
publicly available printed material orienting them to dietary principles and
exercise guidelines for safe weight loss.16,17 Other
resources such as public library materials, Web sites, and telephone numbers
of health promotion organizations offering free weight control information
were drawn to their attention.
Participants assigned to the commercial program were given vouchers
entitling them to attendance at sessions of Weight Watchers, and the locations
of available sites of this commercial program were reviewed with them. The
vouchers enabled participants to attend sessions at no cost and, during the
study period, had a retail value of approximately $9 per voucher. The commercial
program consists of a food plan, an activity plan, and a behavior modification
plan focused primarily on cognitive restructuring. The food plan is a nutritionally
balanced, moderate-deficit diet designed to produce a weight loss of up to
0.9 kg/wk. The activity plan follows current National Institutes of Health
guidelines.18 Weekly group meetings of approximately
an hour's duration are led by successful program graduates who act as role
models and provide written educational materials, a weekly weigh-in, and social
Compliance with treatment assignment in the commercial group was assessed
at each clinic visit by self-reported attendance at group meetings. The use
of concomitant medications for weight loss, including herbal preparations,
The primary outcome measure was change in body weight measured at each
visit. Secondary measures to assess change in adiposity were BMI, waist circumference
(at weeks 26, 52, and 104), and body fat measured by bioimpedance analysis
(Tanita Bodyfat Analyzer, model TBF 105 or 305, Tanita Corporation, Arlington
Heights, Ill). Other secondary measures were blood pressure (measured at each
visit); total cholesterol, high-density lipoprotein (HDL) cholesterol, triglycerides,
insulin, and glucose (measured at 1-year intervals); and quality of life measured
using the Medical Outcomes Study Short-Form 36 Health Survey (SF-36)19 and Impact of Weight on Quality of Life Questionnaire
Adverse events, vital signs, and a standard panel of 20 biochemical
indicators were monitored for safety. Blood samples were sent to a commercial
laboratory (Quest Diagnostics, San Juan Capistrano, Calif) except for insulin
levels, which were assayed at the New York Obesity Research Center, New York
Differences between the treatment groups at randomization and characteristics
of completers vs dropouts were tested using independent t tests for continuous variables and the Fisher exact test for categorical
variables. Values are presented as mean (SD) except where otherwise noted.
Statistical significance was set at P<.05.
Three analyses were performed on weight change and related outcome variables:
an intent-to-treat (ITT) analysis including all randomized participants (missing
values were imputed by last-observation-carried-forward or linear interpolation
and participants who made no follow-up visits were assumed to remain at baseline
value); a modified ITT analysis including all participants who made at least
1 clinic visit after randomization; and a completers analysis using only participants
who completed the study.
A mixed-model repeated-measures analysis of covariance was used to assess
the significance of changes in the outcome variable, with treatment group,
clinic site, visit, and visit × group interaction as fixed effects,
and initial weight as a covariate (Proc Mixed, SAS version 8, SAS Institute
Inc, Cary, NC). Similar mixed regression models were used for analysis of
changes in secondary outcome measures. Initial values of variables such as
age, sex, and BMI were included as covariates if they contributed significantly
to the model.
In the completers analysis of biological indices participants taking
blood pressure, lipid, or glucose lowering medication were excluded from analyses
on those measures, respectively, but included in all others. Analyses on changes
in biological indices were run once including all values and a second time
excluding values more than 3 SDs from the mean. Results were similar in all
substantive aspects and the latter are presented here as being more typical
of the interventions. The relation between weight changes and biological indicators
of health was examined by including weight change and interactions of weight
change and treatment group in linear regression models.
For the tabulation of cases by percentage of weight lost, separate χ2 tests were conducted for year 1 and year 2 to test the hypothesis
that the distributions in the self-help group were not different from those
in the commercial group. Following rejection of the null hypothesis, frequencies
within each percentage of weight loss category were tested against the null
hypothesis that successes were distributed in the same proportion as the number
of participants in the 2 groups.
Characteristics of participants at baseline are shown in Table 1. Two participants developed lymphoma during the trial and
were excluded from the completers analysis. Two participants assigned to the
commercial group reported using a weight loss medication and were excluded
from completers efficacy analyses. In the self-help group almost all participants
reported attempting to change diet and increase physical activity, 14 reported
using weight loss medications, another 6 tried herbal products, 10 enrolled
in some form of structured commercial program (TOPS [Take Off Pounds Sensibly],
Jenny Craig, 5 in Weight Watchers), and 9 mentioned following an alternative
diet plan (protein, Atkins, the Zone) at some point during the 2-year study.
All were retained in the analyses.
Participants who dropped out of the study were younger (age, 42.0 [10.6]
vs 45.8 [9.9] years), had a higher BMI (34.3 [3.5] vs 33.4 [3.5]), greater
percentage of body fat (46.6% [8.8%] vs 44.1% [8.3%]), were more likely to
be smokers, and reported slightly lower income than participants who completed
the study (for all comparisons P<.05). The characteristics
of dropouts did not differ by treatment group.
Bonferroni-adjusted statistical tests on the standard safety panel of
20 biochemistry analytes at randomization found only 1 difference: mean alanine
aminotransferase values in the commercial group were statistically higher
than in the self-help group, although both were within the normal range and
the difference is not clinically significant (25.7 [13.8] vs 21.7 [11.2] U/L, P<.001). There were no clinically significant changes
in safety parameters over the course of the study in either group and no participants
were removed from the study for adverse changes in biochemistry parameters.
Participants' change in weight ranged from −28 to +12 kg in the
commercial group and −26 to +15 kg in the self-help group at 1 year,
and −23 to +21 kg in the commercial group and −26 to +30 kg in
the self-help group at the end of year 2.
Results of ITT, modified ITT, and completers analyses of weight changes
and related body composition variables are shown in Table 2. Because of the low attrition rate, the results are not
greatly different among these 3 analyses.
Mean weight changes from baseline for each of the clinic visits for
all participants who made the visit are shown in Figure 2. At each visit, weight in the commercial group was significantly
lower than at baseline and the amount of loss was greater than in the self-help
group. Weight in the self-help group was significantly less than baseline
weight until week 52 but not thereafter. There were differences among sites
in the amount of weight loss (P<.01); however,
the magnitude of the difference in weight loss between commercial and self-help
groups at the 6 sites was similar (the treatment × site interaction
was not significant). The difference from baseline decreased over time in
both groups after week 26 (P<.001) and the rates
of regain were not different (the treatment × time interaction was not
significant). Amount of weight lost by men compared with women was not significantly
different, although the average weight of men was greater at baseline (106.7
[11.9] vs 91.2 [12.7] kg; P<.001). Weight loss
and gain as a percentage of initial weight at 1 and 2 years is shown in Table 3.
Compliance with treatment assignment in the commercial group was assessed
by self-reported attendance at meetings. The median number of sessions attended
for 6-month periods ending at weeks 26, 52, 78, and 104 were 21, 19, 17, and
13, respectively. Weight loss differed as a function of self-reported attendance
level (P<.05) (Figure 3). Some weight regain was seen between year 1 and 2 at all
attendance levels (P<.001).
Analysis of changes in BMI yielded results similar to those for weight
loss. After 1 year 56% and after 2 years 52% of the commercial group participants
were more than 1 BMI unit below starting BMI, compared with 31% after 1 year
and 29% after 2 years in the self-help group. The differences between commercial
and self-help groups in waist circumference reduction at 1 and 2 years were
statistically significant (Table 2).
The amount of reduction declined over time in both groups but a difference
of about 2 cm remained at the end of 2 years. There were also significant
differences between commercial and self-help groups in average amount of fat
loss maintained over the last 18 months (3.2 [6.3] kg vs 1.6 [6.4] kg, respectively; P = .04). Mean fat loss by men was not significantly different
from that of women.
Although there were sustained differences in mean weight loss between
the 2 groups over the duration of the study, the wide range in degree of individual
success in each group and the low correlation between amount of weight loss
and improvement in biological parameters resulted in mainly nonsignificant
differences between treatment groups (Table
4). Two indices, diastolic blood pressure and serum insulin, were
statistically improved in the commercial group compared with self-help at
year 1 but only insulin was significantly different at year 2. Total cholesterol
and the HDL/total cholesterol ratio improved in both groups, whereas fasting
glucose levels increased in both groups.
Weight changes were reliably correlated with changes in all 8 biological
indices in the commercial group and in 5 of the 8 indices in the self-help
group (all P<.05). Changes in waist circumference
were also significantly related to changes in all indices (all P<.02). Change in total cholesterol and triglycerides was 3 or more
times as large per unit change in waist circumference for men as for women.
There were no significant differences between groups in amount of improvement
on quality of life scales. Weight loss and percentage of weight loss were
positively correlated with improvements in all scales of the SF-36 at both
year 1 and year 2 in both groups (data not shown). Using a Bonferroni-adjusted
significance level of P = .006, at 1 year weight
loss was significantly correlated with improvements in scales for general
physical health (r = 0.25) and vitality (r = 0.26). At year 2, scales for physical function (r = 0.22) and vitality (r = 0.21) were significantly
correlated with weight loss. A similar pattern was seen for scales of the
IWQOL-Lite (data not shown). Weight strongly predicted total score and all
subscale scores, with the strongest relationships for public distress, physical
function, and total score. The relationship was in the expected direction,
with lower weight predicting better quality of life, and this did not differ
The self-help group was able to lose and maintain approximately 1.3
to 1.4 kg for the first year, after which weight tended to increase until
it returned to baseline at 2 years. The commercial group maintained a weight
loss of 4.3 to 5.0 kg at the end of the first year and was 2.7 to 3.0 kg lower
than baseline weight at the end of the second year. Participants who attended
78% or more of the commercial group sessions maintained a mean weight loss
of almost 5 kg at the end of the 2-year study. These weight losses were achieved
despite a general tendency to gain weight during most of the adult lifespan.21 After 2 years, 52% of the commercial group and 29%
of the self-help group had a BMI 1 unit or more below baseline BMI, an amount
that the Institute of Medicine report on obesity treatment described as significant
long-term weight loss.22 Waist circumference,
an independent cardiovascular risk factor,23,24 was
also reduced in the commercial group by about 4.5 cm at 1 year and 2.5 cm
at 2 years. It is noteworthy that these results were obtained not in academic
or research settings but in the context of a regular ongoing commercial program.
The correlations of changes in biological parameters with weight changes
were generally similar to those seen in other studies with modest amounts
of weight loss.4,25- 29 The
magnitude of the group change may depend on characteristics of the study sample
at randomization and the precise nature of the intervention (eg, diet, exercise,
sodium content of diet, amount of weight lost). The reasons for the low correlations
of individual improvement in biological indices with amount of weight lost
are not known. Additional analyses are being conducted to establish whether
individual participant factors, such as the nature of the dietary and lifestyle
changes, might account in part for the degree of improvement.
Despite sustained differences in mean weight loss between the 2 groups,
the mean improvement in most biological parameters was not statistically different.
In part, this is a statistical issue because the study was powered to detect
differences in weight loss, not biological indices. However, the wide variability
in attendance at the commercial sessions and success at weight loss highlights
the importance of compliance with the assigned treatment. In this case, attendance
at group meetings may provide the conditions necessary to produce and sustain
weight loss and may increase the likelihood of improvements in biological
indicators. However, even among those with the highest degree of adherence,
there was weight regain during the second year of the study.
Self-selection bias—the tendency for successful participants to
remain in the trial—may be a factor influencing these weight loss results,
although that influence is likely to be relatively small because the high
retention rate (73% after 2 years) limited the opportunity for self-selection
to occur. Also, the correlation between attendance and success at weight loss
is supported by other research showing that length of the treatment period,
and consequently the number of intervention meetings, has an independent effect
on success.30 It is also possible that characteristics
of individuals willing and motivated to participate in long-term randomized
clinical trials may have influenced the trial results. Finally, it should
be emphasized that this study reports on a particular commercial program with
many unique aspects. Our results should not be taken as representative of
all commercial programs, many of which use other interventions, such as proprietary
liquid formulas or diets that are not balanced.
In summary, this 2-year trial provides information on weight loss in
an ongoing structured commercial weight loss program in comparison with self-help
attempts to lose weight. The results show that this program provides modest
weight loss but is more effective than brief counseling and self-help for
overweight and obese adults.