VOLUME 16 NUMBER 2 | FEBRUARY 2008 | www.obesityjournal.org
behavior and psychology
nature publishing group
Effect of Group Racial Composition on Weight
Loss in African Americans
Jamy D. Ard1,2, Shiriki Kumanyika3, Victor J. Stevens4, William M. Vollmer4, Carmen Samuel-Hodge5,
Betty Kennedy6, Debra Gayles7, Lawrence J. Appel7, Phillip J. Brantley8, Catherine Champagne6, Jeanne
Charleston7 and Laura P. Svetkey9
Objective: We do not know how racial composition of a group influences behavior change for African Americans
(AAs) in group-based weight loss programs. We tested the hypothesis that AA who participate in all AA weight loss
intervention groups will lose more weight than AA who participate in mixed race groups.
Methods and Procedures: This observational study was ancillary to Phase 1 of the Weight Loss Maintenance Study,
a multi-center trial of strategies to maintain weight loss after a 20-week behavior modification program. Three of
four centers recruited several all-AA intervention groups. Remaining groups were combinations of AA and non-AA
participants. All participants received the same weight loss intervention. Change in weight was the primary outcome,
comparing participants of all-AA groups with AA participants of mixed race groups conducted by the same AA
interventionists. Secondary outcomes included measures of intervention adherence and behavior change.
Results: Participants of all-AA groups (n = 271) were comparable to other AA participants (n = 106). The mean
proportion of AA in mixed race groups was 56%. All-AA group participants had similar weight loss as those in mixed
groups (−4.2 vs. −4.2 kg, P = 0.97). There were no differences between the groups in mean number of sessions
attended or changes in dietary intake.
Discussion: Significant weight loss was observed in both groups, with no effect of group composition on adherence
or weight loss outcomes. Special logistics to accommodate all-AA groups may not be necessary. Despite varying
instructional environments, AA appeared to respond positively to intervention messages with significant changes in
dietary intake, physical activity (PA), and weight.
Obesity (2008) 16, 306–310. doi:10.1038/oby.2007.49
Large multi-center randomized controlled clinical trials have
demonstrated the effectiveness of lifestyle interventions for
weight loss, blood pressure reduction, and diabetes control
(1–5). However, closer inspection reveals that the actual amount
of weight loss achieved in these and similar trials was less for
African Americans than whites (6–10). This was also observed in
the recent PREMIER trial, in which African Americans assigned
to active behavioral intervention lost an average of 3.5 kg at
6 months, compared to ~6 kg in non-African Americans (pre-
dominantly non-Hispanic whites) (11).
It has been suggested that a lack of cultural appropriate-
ness may partly explain the smaller weight losses in African
American compared with other participants in studies such
as the ones noted above (12). There is a strong theoretical
rationale for adapting weight control programs to be appro-
priate to the relevant sociocultural context. Such programs
may be more acceptable and more effective than standard
programs when the audience of interest is a racial/ethnic
minority population with higher than average rates of obes-
ity, as is the case for African Americans. The substrate for
weight-related behavior changes includes food preferences,
symbolic psychosocial meanings of food, food related social
roles and interactions, and attitudes about physical activity
(PA), health, and body size. These factors are all culturally
influenced, with documented differences for ethnic minor-
ity compared with the majority population in the United
To our knowledge, the question of whether and how much
various cultural adaptations improve weight losses in African
1Department of Nutrition Sciences, University of Alabama at Birmingham, Birmingham, Alabama, USA; 2Department of Medicine, University of Alabama at Birmingham,
Birmingham, Alabama, USA; 3Department of Biostatistics and Epidemiology, University of Pennsylvania, Philadelphia, Pennsylvania, USA; 4Kaiser Permanente, Center
For Health Research, Portland, Oregon, USA; 5Department of Nutrition, University of North Carolina at Chapel Hill, Chapel Hill, North Carolina, USA; 6Pennington
Biomedical Research Center, Dietary Assessment and Food Analysis Core, Baton Rouge, Louisiana, USA; 7Johns Hopkins University, Welch Center for Prevention,
Epidemiology, and Clinical Research, Baltimore, Maryland, USA; 8Pennington Biomedical Research Center, Division of Education, Baton Rouge, Louisiana, USA; 9Duke
University Medical Center, Sarah Stedman Nutrition and Metabolism Center, Durham, North Carolina, USA. Correspondence: Jamy D. Ard (firstname.lastname@example.org)
Received 6 February 2007; accepted 15 June 2007. doi:10.1038/oby.2007.49
Obesity | VOLUME 16 NUMBER 2 | FEBRUARY 2008 307
behavior and psychology
Americans over and above what can be obtained without these
adaptations has heretofore not been addressed directly. Studies
labeled as “culturally adapted” have reported wide ranges of
weight loss, including weight reductions that are small or neg-
ligible and some that are relatively large (16–18). Differences in
study length and design make it difficult to make even indirect
comparisons of results of culturally adapted studies. Moreover,
the use of multiple adaptations within any one study does not
permit separate assessment of the effects of different types of
cultural adaptations. Most studies have involved delivery of a
group intervention in which all participants and the instruc-
tors were the same ethnicity, (e.g., all African Americans) along
with other cultural modifications, e.g., culturally appropriate
recipes, addressing attitudes about exercise, and including
family members (16).
The use of all African-American intervention groups is one
adaptation that appears to be most uniformly applied through-
out studies of culturally appropriate weight loss interven-
tions. The effect of the group’s racial composition on outcome
measures in these studies is unclear. Based on feedback from
African-American participants in prior studies, we identi-
fied the group composition as a potentially important factor
in successful behavior change. The ethnic mix could alter the
dynamics of the group interaction and the amount and type of
social support that participants receive, ultimately improving
weight loss outcomes.
We investigated the “value added” of this specific cultural
adaptation as part of a multisite trial of weight loss mainte-
nance. Specifically, as part of an initial weight loss program, we
compared the impact of all-African-American vs. mixed race
intervention groups on weight loss in our African-American
participants. We hypothesized that African Americans who
participated in all-African-American groups would lose more
weight than African Americans who participated in mixed race
groups. Our secondary hypothesis was that African Americans
who participated in the all-African- American groups would:
have higher attendance at group sessions, adhere better to
study dietary recommendations, and exhibit a greater increase
in PA than African American participants in the mixed race
Methods and procedures
Weight Loss Maintenance (WLM) is a multi-center clinical trial of strat-
egies for maintaining weight loss after it has been achieved. The study
consisted of an initial, non-randomized weight loss phase in which all
participants participated in a 20-week, group-based lifestyle change
intervention designed to help them achieve at least 4 kg of weight loss.
This was followed by a second, maintenance phase in which partici-
pants were randomly assigned to one of three intervention arms. This
report is limited to the Phase 1 results.
Three of the WLM clinical centers (Baltimore, Baton Rouge, and
Durham) participated in an ancillary study in which some of the
weight loss groups were structured to include only African-American
participants to evaluate the impact of racial mix on weight loss results in
African-American participants. Assignment to the groups was not made
at random. Rather, for logistical reasons each site arbitrarily designated
one “wave” of intervention groups to be their all African-American wave
and only recruited African-American participants for those groups.
Study participants were aged 25 years or older, had BMIs between 25
and 45 kg/m2, and had to be taking prescription medication for either
dyslipidemia or hypertension. Those with treated diabetes or with a
major cardiovascular disease event in the past year were excluded from
participation (Table 1).
Participants were recruited from the local communities using a variety
of means, including mass media advertising, targeted mailings, and flyers.
For this ancillary, participating sites developed a targeted recruitment
plan to recruit only African-American participants for one of the waves
of intervention classes.
phase 1 intervention
Participants proceeded through the Phase 1 intervention as speci-
fied in the WLM protocol. No alterations in materials, intervention
delivery or assessments were attempted. Participants met weekly for
20 weeks, in groups of 15–25 participants led by an interventionist.
The recruitment of all-African-American cohorts allowed the crea-
tion of all-African-American intervention groups at each study site.
The only differences between the all-African American groups and the
mixed race groups were: (i) the racial composition of the group (100%
African American vs. a variable proportion of African Americans),
and (ii) the ethnicity of the interventionist. Only African American
interventionists led the all-African-American groups. Mixed race
groups were led by African American or white interventionists. All
interventionists received training on culturally appropriate delivery of
the Phase 1 program.
Weight loss was promoted through calorie reduction, change in dietary
pattern, and increased energy expenditure with PA. The specific interven-
tion targets included reducing weight by 4.5–6.8 kg (or more if desired),
engaging in moderate PA for 180 min/week, and following a healthy,
reduced-calorie dietary pattern (the Dietary Approaches to Stop Hyper-
tension diet)aimed at reduction of cardiovascular disease risk factors (19).
Some key strategies used to promote weight loss included self-monitoring
of diet and PA; reducing portion sizes, substituting alternative foods, and
modifying the original items to be lower in calories and fat; increasing
fruit, vegetable, and fiber intake; increased PA; identifying problematic
situations for undesired behavior and developing and rehearsing specific
plans of action to deal with those situations; and developing core food-
The primary outcome for this analysis was change in weight (in kg)
from study enrollment through the end of phase 1 (20 weeks). Weight
was measured in light indoor clothes without shoes to the ~0.1 kg
using a calibrated balance beam or digital scale. Height was measured
to the ~0.1 cm using a wall-mounted stadiometer. BMI was calculated
as the Quetelet index (kg/m2). Missing end weights were imputed as
Secondary outcome measures included assessments of adherence to
study recommendations for diet and PA and group session attendance.
table 1 Key WLM phase 1 eligibility criteria
Aged 25 years or older
BMI of 25–45 kg/m2
Currently taking prescription medication for hypertension and/or
Not taking medication for diabetes
Negative stress test required if diabetic or if history of CVD
No CVD event within past 12 months
Willing to lose 4 kg in phase 1
CVD, cardiovascular disease.
VOLUME 16 NUMBER 2 | FEBRUARY 2008 | www.obesityjournal.org
behavior and psychology
The Block Food Frequency Questionnaire was used to determine die-
tary intake. The Stay Healthy, accelerometer (RT3) was used to measure
PA objectively. Participants wore the monitor for a minimum of 4 days,
including at least 1 weekend day, with at least 6 h of wearing time. The
weekly PA measure was re-scaled to 5 weekdays and 2 weekend days for
a total of 7 days. Participant attendance at group sessions was recorded
by study interventionists each week. No secondary outcome data were
imputed for this analysis.
A total of 678 African-American participants were enrolled at one of the
three study sites. Each study site had one (African American) interven-
tionist who led the all-African-American groups. To better control for
potential interventionist effects, mixed race group participants are only
included in the analysis if their group was led by the same intervention-
ist who led the all-African-American groups at that site. At Baltimore,
this interventionist led three all-African-American groups with a total
of 74 participants and one mixed race group that included a total of
16 African Americans. In similar fashion, the Baton Rouge sample
included 90 individuals from five all-African-American groups and 22
African Americans from three mixed race groups led by the same inter-
ventionist, while the Durham sample included 107 individuals from six
all-African-American groups and an additional 68 African Americans
from four mixed race groups. Therefore, the analysis includes a sample
of 377 participants—271 from 14 all-African-American groups and 106
from 8 mixed race groups. The remaining 301 African-American par-
ticipants were in mixed race groups led by interventionists who did not
lead the all-African-American groups at their respective sites and are
thereby excluded from the analysis.
Of the 377 individuals included in the analysis, 342 (91%) had
follow-up weights at the end of phase 1. We used a multiple imputation
procedure for the 35 individuals with missing follow-up weights (20,21).
Five separate imputation samples were generated using SAS PROC MI,
each with 1,000 burn-in iterations and 500 iterations between data draws.
Subsequent data analyses are done separately on each of the five complete
datasets so generated. The resulting parameter estimates are averaged
according to Rubin’s rules for combining results (20). The s.e. are simi-
larly averaged and then adjusted to account for the between-imputation
variation, while the degrees of freedom are adjusted to obtain unbiased
P-values. As they account for the added variability of the imputation
process itself, the resulting s.e. are typically inflated relative to what would
be seen using only a single imputation sample.
We used one-way ANOVA (continuous variables) and Pearson chi-
square tests (categorical variables) for unadjusted analyses that compared
both weight change and the secondary outcomes between the all-African
American and mixed race groups. An adjusted regression analysis of
the primary outcome, weight change, included main effects to account
for the potentially confounding effects of age, initial BMI, gender, and
interventionist. This analysis also included intervention group as a clus-
In general, individuals who participated in the all-African-
American groups were similar to those in the mixed race
groups (Table 2). The average age was 52 years, and 30% were
male. Over 93% reported at least some college education and
almost 50% reported annual household incomes in excess of
≥$60,000. Ninety-three percent were taking antihyperten-
sive medications and approximately one-quarter were taking
medication for hyperlipidemia. The proportion of African
Americans in the mixed race groups averaged 56% and varied
markedly by site, ranging from 23 to 38% at Baton Rouge, from
71 to 90% at Durham, and 53% for the one mixed race group at
the Baltimore site.
table 2 characteristics of study sample
N = 271
Mixed race group
N = 106
51.6 ± 9.351.8 ± 9.2
Male 26.6% 37.7%
Weight (kg)99.4 ±15.8 97.5 ± 18.9
35.4 ± 4.834.3 ± 5.1
Continuous variables expressed as mean ± s.d.
table 3 Intervention attendance and lifestyle change
N = 271
N = 106
% Attending ≥17
# Sessions attendeda
13.6 ± 6.512.8 ± 6.20.27
N = 148
N = 43
Servings of fruits and
vegetables per daya
5 ± 2.7
8 ± 4.4
3 ± 4.3
4.3 ± 2.4
7.7 ± 4.4
3.4 ± 4.8
% Calories fata
37.9 ± 6.6
31.4 ± 7.3
−6.5 ± 7.6
39.3 ± 6.3
31.8 ± 6.6
−7.5 ± 7.9 0.46
Fiber intake (g/d)a
17.5 ± 8.1
21.1 ± 10
3.5 ± 9.3
15.4 ± 6.1
18.8 ± 8.2
3.4 ± 10.20.94
N = 159
N = 52
% Achieving goal of
≥180 min/week of PA
PA, physical activity.
aContinuous variables expressed as mean ± s.d. bP values based on one-
way ANOVA for continuous variables and Pearson chi-square for categorical
Obesity | VOLUME 16 NUMBER 2 | FEBRUARY 2008 309
behavior and psychology
process measure outcomes
Group session attendance and several indicators of lifestyle
change are summarized in Table 3. Although participants in
the all- African-American groups tended to attend more group
sessions than did those in the mixed groups, none of the dif-
ferences in Table 3 were statistically significant. Both groups
increased fruit and vegetable intake by about three servings
per day and fiber intake by ~3.5 g/day, while decreasing % cal-
ories from fat by ~7%. Both also more than doubled the pro-
portion who reported at least 180 min of moderate-vigorous
PA per week.
Similarly, we found no evidence of a statistically significant
difference in weight loss between participants in the all-
African-American and mixed race groups, although the trend
was for greater weight loss for participants in the all- African-
American groups. This was true both for the unadjusted
(Table 4) and the adjusted (Table 5) analyses. Mean weight
loss was 4.2 kg, and 60% achieved the minimum weight loss of
4 kg to qualify for randomization into phase 2. BMI declined
by 1.5 kg/m2. As seen in Table 5, African American men lost
~1.2 kg more than did African American women, even after
adjusting for initial BMI.
During phase 1 of the WLM trial, targeted recruitment strat-
egies created the opportunity to observe the effect of group
racial composition, all-African-American vs. mixed race,
on weight loss outcomes. These observations are of inter-
est for a number of reasons. African Americans, particularly
women, can benefit to a greater extent from effective lifestyle
interventions due to the high prevalence of obesity and related
comorbidities that are improved by weight loss (1–5). African
Americans who engage in lifestyle interventions typically lose
less weight than non-African Americans (6–10), suggesting a
need for strategies to make standard programs more effective
in African Americans. African American participants with
whom we have worked in previous lifestyle interventions have
suggested that their experience in a racially homogeneous
group may be decidedly different because of a shared cultural
perspective. This anecdotal observation is consistent with the
common practices of attempting to create culturally specific
interventions by recruiting African Americans only and eth-
nically matching the intervention leader. This study formally
evaluated the potential value of these approaches. Group com-
position was varied and the ethnicity of the intervention leader
was held constant.
The lack of a statistically significant effect of group composi-
tion on intervention attendance, the lifestyle change variables,
and weight loss suggests that special logistics to accommodate
all-African American groups may not be worthwhile. Whether
this finding only holds true when the group leader is African
American cannot be determined from these data. The results
of this study suggest that, regardless of the racial mix of the
groups, African Americans appear to respond positively to
well-designed lifestyle change interventions designed to pro-
mote healthier eating, increased PA, and weight loss.
There are several factors that may have contributed to our
negative findings. This study was not a randomized trial, and
despite our efforts to control for potential confounding there
may still be some residual factors confounding the analysis.
Second, the percentage of African-American participants in
our mixed race groups was high, >50% at one site and 70%
for all groups at another site. It is possible that a critical mass
of African-American participants led to a change in the group
dynamics and interactions, creating a group environment that
is similar in nature to one that is comprised of only African
Americans and thus blurring the experimental contrast. We
attempted to look at the association between the percent of
African-American participants in a group and weight loss, but
it was too highly correlated with site (and hence intervention-
ist) to be evaluated after adjusting for the latter effect. Third,
our participants were specifically recruited because they were
at high risk for future cardiovascular events. Thus they may
have been particularly motivated to adopt behavior change to
improve their health. Fourth, in an effort to increase the effi-
cacy of the intervention for all-African American participants,
the study investigators expended considerable effort to develop
a culturally appropriate intervention. This included a minority
implementation committee specifically charged with develop-
ing trial wide procedures to enhance minority participation and
success, and cross-cultural training for all study personnel that
included specific, specialized instruction in cultural sensitivity
and motivational interviewing techniques for study interven-
tionists. These tactics may have contributed to an increase in
the overall effectiveness of the intervention for all participants,
thus limiting the potential impact that group racial mix might
table 4 unadjusted weight change results
N = 271
N = 106
Weight change (kg)−4.2 + 4.6−4.2 + 8.6 0.97
BMI change (kg/m2) −1.5 + 1.7−1.5 + 2.9 0.84
% >4 kg weight loss61.2%55.7% 0.33
aTwo-tailed P values based on one-way ANOVA with imputed weight data.
table 5 results of multiple regression analysis for weight
BMI (kg/m2)0.049 0.12
Male gender0.56 0.03
All-African American group0.54 0.73
Site A 0.460.63 0.47
Site B 0.430.57 0.45
aIncludes imputed weight data from five replicate multiple imputation process.
bNegative coefficient indicates greater weight loss compared to the reference
category (indicator variables) or for a unit increase (continuous variables).
310 Download full-text
VOLUME 16 NUMBER 2 | FEBRUARY 2008 | www.obesityjournal.org
behavior and psychology
have had. To some degree, this latter conjecture is supported by
the fact that African Americans in WLM had a greater mean
weight loss in the 20 weeks than has generally been reported in
similar multi-center lifestyle change interventions (8,22).
One final consideration is the lack of statistically significant
results may reflect inadequate statistical power for the primary
outcome and the process measures. Due to the nature of the
overall study design, it was not possible to fix in advance the
total number of African-American participants or to constrain
the number of participants in the all-African-American and
mixed race groups to be the same. Using the s.e. of the effect
size estimators in our multivariate analyses, we estimate that
we would have had 79% power to detect a weight difference
of 1.5 kg between the two groups and 90% power to detect a
1.75 kg difference. The 95% confidence interval for the observed
difference of −0.19 kg was (−1.25 kg, 0.87 kg).
Our analysis did not address the impact of the group com-
position on weight loss for African Americans relative to non-
African Americans. Rather we chose this research question
because we wanted to assess the incremental impact of the
group composition within African Americans. As a key first
step, our primary intent was to isolate specific components of
culturally specific interventions to estimate the value added of
a particular component. By understanding the value of a spe-
cific intervention component, we may better determine what
causes an intervention to be culturally appropriate. Once we
have defined the effective components of culturally specific
interventions, studying the impact of multiple intervention
components on weight loss for African Americans relative to
non-African Americans seems a logical progression.
In the larger scope of behavioral weight loss research, it is
apparent that additional specific studies are needed to isolate and
identify components of interventions that will be most effective
for African American populations. Standardizing and evaluating
explicit strategies for targeting minority populations could pro-
vide useful knowledge in the collective effort to improve inter-
vention effectiveness for everyone and reduce health disparities.
This study demonstrates that a behavioral approach to weight
loss can be effective in overweight/obese African Americans
with cardiovascular disease risk factors, regardless of whether
or not the intervention groups are all-African-American.
Weight Loss Maintenance is supported by National Heart, Lung, Blood
Institute grants 5-U01 HL68920, 5-U01 HL68734, 5-U01 HL68676, 5-U01
HL68790, and 5-HL68955. J.D.A. is supported by grants from the Robert
Wood Johnson Foundation (grant 51894) and the National Institutes of
Health (grant DK68223).
The authors declared no conflict of interest.
© 2008 The Obesity Society
1. Knowler WC, Barrett-Connor E, Fowler SE et al. Reduction in the incidence
of type 2 diabetes with lifestyle intervention or metformin. N Engl J Med
2. Elmer PJ, Grimm R Jr, Laing B et al. Lifestyle intervention: results of
the Treatment of Mild Hypertension Study (TOMHS). Prev Med
3. Stevens VJ, Corrigan SA, Obarzanek E et al. Weight loss intervention in
phase 1 of the trials of hypertension prevention. The TOHP Collaborative
Research Group. Arch Intern Med 1993;153:849–858.
4. Wing RR, Blair E, Marcus M, Epstein LH, Harvey J. Year-long weight
loss treatment for obese patients with type II diabetes: does including
an intermittent very-low-calorie diet improve outcome? Am J Med
5. Whelton PK, Appel LJ, Espeland MA et al. Sodium reduction and weight
loss in the treatment of hypertension in older persons: a randomized
controlled trial of nonpharmacologic interventions in the elderly (TONE).
TONE Collaborative Research Group. JAMA 1998;279:839–846.
6. Stevens VJ, Obarzanek E, Cook NR et al. Long-term weight loss and
changes in blood pressure: results of the Trials of Hypertension Prevention,
phase II. Ann Intern Med 2001;134:1–11.
7. Wing RR, Hamman RF, Bray GA et al. Achieving weight and activity goals
among diabetes prevention program lifestyle participants. Obesity Res
8. Kumanyika SK, Obarzanek E, Stevens VJ, Hebert PR, Whelton PK.
Weight-loss experience of black and white participants in NHLBI-sponsored
clinical trials. Am J Clin Nutr 1991;53(Suppl 6):1631S–1638S.
9. Wing RR, Anglin K. Effectiveness of a behavioral weight control program for
blacks and whites with NIDDM. Diabetes Care 1996;19:409–413.
10. Kumanyika SK, Espeland MA, Bahnson JL et al. Ethnic comparison of
weight loss in the trial of nonpharmacologic interventions in the elderly.
Obesity Res 2002;10:96–106.
11. Svetkey LP, Erlinger TP, Vollmer WM et al. Effect of lifestyle modifications
on blood pressure by race, sex, hypertension status, and age. J Hum
12. Kumanyika SK, Morssink C, Agurs T. Models for dietary and weight
change in African-American women: identifying cultural components.
Ethn Dis 1992;2:166–175.
13. Airhihenbuwa CO, Kumanyika S, Agurs TD et al. Cultural aspects of African
American eating patterns. Ethn Health 1996;1:245–260.
14. Airhihenbuwa CO, Kumanyika S, Agurs TD, Lowe A. Perceptions and beliefs
about exercise, rest, and health among African-Americans. Am J Health
15. Bennett G, Wolin K, Goodman M et al. Attitudes regarding overweight,
exercise, and health among blacks (United States). Cancer Causes Control
16. Bronner Y, Boyington JE. Developing weight loss interventions for
African-American women: elements of successful models. J Natl Med Assoc
17. Karanja N, Stevens VJ, Hollis JF, Kumanyika SK. Steps to soulful living
(steps): a weight loss program for African-American women. Ethn Dis
18. Ard JD, Rosati R, Oddone EZ. Culturally-sensitive weight loss program
produces significant reduction in weight, blood pressure, and cholesterol
in eight weeks. J Natl Med Assoc 2000;92:515–523.
19. Appel LJ, Moore TJ, Obarzanek E et al. A clinical trial of the effects of
dietary patterns on blood pressure. DASH Collaborative Research Group.
N Engl J Med 1997;336:1117–1124.
20. Little RJA, Rubin DB. Statistical Analysis With Missing Data, 2nd edn, Wiley:
Hoboken, NJ, 2002.
21. Collins LM, Schafer JL, Kam CM. A comparison of inclusive and restrictive
strategies in modern missing data procedures. Psychol Methods
22. Appel LJ, Champagne CM, Harsha DW et al. Effects of comprehensive
lifestyle modification on blood pressure control: main results of the PREMIER
clinical trial. JAMA 2003;289:2083–2093.