Comparative Effectiveness of LifestyleIntervention Efforts in the CommunityResults of the Rethinking Eating and ACTivity (REACT) study

GRETCHEN A. PIATT, MPH, PHD1

MIRIAM C. SEIDEL, MS, RD, LDN2

ROBERT O. POWELL, MS2

JANICE C. ZGIBOR, RPH, PHD3

OBJECTIVEdTo determine the comparative effectiveness of three lifestyle intervention mo-dalities in decreasing risk for diabetes.

RESEARCHDESIGNANDMETHODSdFive hundred and fifty-five individuals (86.1%female, 95.1% white, and 55.8% obese) from eight rural communities were screened for BMI$25 kg/m2 and waist circumference .40 inches in men and .35 inches in women. Commu-nities with their eligible participants (n = 493; mean age 51 years, 87.6% female, 94.1% Cauca-sian) were assigned to four Group Lifestyle Balance (GLB) intervention groups: face to face (FF)(n = 119), DVD (n = 113), internet (INT) (n = 101), and self-selection (SS) (n = 101). SS partic-ipants chose the GLB modality. GLB is a comprehensive lifestyle behavior–change program.

RESULTSdA marked decline was observed in weight after the intervention in all groups (FF212.5 lbs, P = 0.01; DVD212.2 lbs, P, 0.0001; INT213.7 lbs, P, 0.0001; and SS214 lbs,P , 0.0001). Participants in SS experienced the largest average weight loss. Weight loss wassustained in .90% of participants in each group at 6 months (FF 90.7%, DVD 90.9%, INT92.1%, and SS 100%). All groups experienced improvements in the proportion of participantswith CVD risk factors. The proportion of individuals with CVD risk factors remained steadybetween 3 and 6 months in all groups and never returned back to baseline. All associationsremained after multivariate adjustment.

CONCLUSIONSdDespite the modality, the GLB intervention was effective at decreasingweight and improving CVD risk factor control. SS and FF participants experienced greaterimprovements in outcomes compared with other groups, establishing the importance of pa-tient-centered decision making and a support network for successful behavior change.

Diabetes Care 36:202–209, 2013

A s the obesity epidemic continues toworsen in the U.S., the health ofcommunities is increasingly in a

state of jeopardy. Approximately 65% ofAmericans are overweight or obese, 65million have prediabetes, and roughlyone-third of individuals in the U.S. havemetabolic syndrome (1). All of these con-ditions substantially increase the risk fordeveloping diabetes and/or cardiovasculardisease (CVD) (1). The economic burden ofthese conditions is staggering; costs related

to obesity alone currently exceed 147 bil-lion USD annually in the U.S (2).

Efficacy-based trials demonstrate anextensive amount of evidence for lifestyleinterventions aimed at preventing or de-laying diabetes (3–5). Moreover, the pos-itive outcomes observed initially in thesetrials appear sustainable in the long term.Data from the Da Qing Diabetes Preven-tion Study and the Finnish Diabetes Pre-vention Study demonstrate a 43% lowerincidence of diabetes for up to 14 and 7

years, respectively (6,7). Similarly, the Di-abetes Prevention Program OutcomesStudy (DPPOS) demonstrated a reductionin diabetes incidence by 34% in the lifestylegroup at 10-year follow-up (8). However,nearly all of these efficacy trials used indi-vidual behavior-change counseling in ahighly selected group of participants.

Multiple interventions adapted fromthese efficacy trials demonstrate the effec-tiveness of weight loss and diabetes and/orCVD risk reduction in real-world settings(9–16). Indeed, health professionals, pay-ers, and policy makers worldwide increas-ingly recognize the need for these efforts.However, for a meaningful impact onpublic health policy and clinical care forprimary prevention, determining the com-parative effectiveness of real-world primaryprevention efforts in community settings iscritical (17). This knowledge will assistconsumers, clinicians, purchasers, and pol-icy makers in making informed decisionsabout primaryprevention thatwill improvehealth care at both the individual andpopulation levels.

To address the need for evidence ofthe comparative effectiveness of differentlifestyle intervention modalities, we con-ducted a prospective, multisite, quasi-experimental study with four parallellifestyle intervention groups. Modalitiesincluded face-to-face group education(FF), DVD education, and internet edu-cation (INT). We aimed to determinewhich modality of lifestyle interventionwas most effective at reducing risk ofdiabetes and/or CVD. We further aimedto understand whether participants whowere given the option to choose themodality that best suited their lifestyleexperienced greater improvements inoutcomes compared with those in groupsin which the modality was predetermined.

RESEARCH DESIGN ANDMETHODS

Study setting and patientpopulationThe study population consisted of over-weight (BMI $25 kg/m2) adults withoutdiabetes who were abdominally obese

c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c c

From the 1Department of Medical Education, University of Michigan, Ann Arbor, Michigan; the 2University ofPittsburgh Diabetes Institute, Pittsburgh, Pennsylvania; and the 3Department of Epidemiology, Universityof Pittsburgh, Pittsburgh, Pennsylvania.

Corresponding author: Gretchen A. Piatt, piattg@umich.edu.Received 30 April 2012 and accepted 20 July 2012.DOI: 10.2337/dc12-0824The U.S. Government is authorized to reproduce and distribute reprints for governmental purposes not-

withstanding any copyright notation thereon. The views and conclusions contained herein are those of theauthors and should not be interpreted as necessarily representing the official policies or endorsements,either expressed or implied, of the Air Force Surgeon General’s office or the U.S. Government.

© 2013 by the American Diabetes Association. Readers may use this article as long as the work is properlycited, the use is educational and not for profit, and thework is not altered. See http://creativecommons.org/licenses/by-nc-nd/3.0/ for details.

202 DIABETES CARE, VOLUME 36, FEBRUARY 2013 care.diabetesjournals.org

C l i n i c a l C a r e / E d u c a t i o n / N u t r i t i o n / P s y c h o s o c i a l R e s e a r c hO R I G I N A L A R T I C L E

(waist circumference.40 inches inmalesand.35 inches in females) and who livedin eight rural underserved communitiesin southwestern Pennsylvania. We ex-cluded individuals who reported a diag-nosis of diabetes or had a currentprescription for glucose-lowering medi-cation, were pregnant, could not walkone-quarter mile without stopping, hadbariatric surgery, were currently usingweight loss medications, or could notprovide informed consent. In general,the eligibility criteria for the study weremore flexible than those typically used inefficacy trials (3,18–20). Additionally,

there was no requirement that partici-pants attend the group sessions. All studycommunities were socioeconomically de-pressed areas with high prevalence ratesof chronic disease.

Study designThe study was a prospective, multisite,quasi-experimental study with four par-allel lifestyle-intervention groups (Fig. 1).The community in which the participantlived determined the lifestyle-interven-tion group to which they were allocated.Allocation of communities to the inter-vention groups was based on community

characteristics (population size, ethnicmajority, education level, median house-hold income, persons below poverty, andthe age-adjusted prevalence of obesity[21]) and the community’s capacity tocarry out the respective interventions.For example, it was not practical for acommunity that did not have adequateaccess to high-speed internet to be allo-cated to the INT group. The study wasimplemented in three phases: phase 1,training and certification in the GroupLifestyle Balance (GLB) program and instandardized measurement techniques;phase 2, community-based screening to

Figure 1dFlow diagram. xThree-month assessment results carried forward to 6-month assessment if data were missing.

care.diabetesjournals.org DIABETES CARE, VOLUME 36, FEBRUARY 2013 203

Piatt and Associates

determine eligibility of individuals to takepart in the intervention; and phase 3, pro-vision of the intervention with 3- and6-month follow-up including clinical as-sessment. The three phases of the studyare described below.Phase 1: training and certification.Phase 1 consisted of GLB training providedby the University of Pittsburgh DiabetesPrevention Support Center (22). Trainedcommunity preventionists (registerednurses and dietitians fromeach communitysite) delivered all GLB content for the in-tervention, while lay health coaches (alsofrom the communities) were used in a sup-port role to aid preventionists and to pro-vide informational and emotional supportto study participants. Lay health coacheswere trained by the preventionists and re-search staff onGLB content, active listeningskills, research fundamentals, and TheHealth Insurance Portability and Account-ability Act requirements. For maintenanceof a high level of treatment fidelity in the

community, four research study coordina-tors were implemented to work closelywith the preventionists and lay healthcoaches at each community site. The re-search team received training in standard-ized measurement consistent with theDiabetes Prevention Program (DPP) bythe University of Pittsburgh Diabetes Pre-vention Support Center staff or their de-signees.Phase II: community-based screening.Individuals whomet the inclusion criteriawere targeted for screening between Oc-tober 2009 and June 2010 through re-cruitment flyers and local newspaper andradio station advertisements. Screeningswere offered at no cost at hospitals,churches, work sites, and other locationsthroughout the eight study communitiesand were facilitated by study staff (studycoordinators, preventionists, and layhealth coaches). Five hundred and fifty-five individuals were screened for BMI$25 kg/m2 and abdominal obesity at 1 of

44 screenings held in the study commu-nities. Four hundred and ninety-three in-dividuals were eligible andwere invited toparticipate in phase III of the study (life-style interventions). Those who were noteligible were referred to other exerciseand healthy-lifestyle programs in theircommunities (Fig. 1). For most partici-pants, the intervention visits were heldat the same site where they were screenedfor study eligibility. In instances when thescreening sites were not available or couldnot facilitate a group class, interventionsites in the same community as thescreening sites were chosen. Therewere a total of 26 intervention sites acrossthe eight study communities.Phase III: interventions. The GLB pro-gram is a comprehensive lifestyle behavior–change program adapted from the lifestyleintervention used in the DPP that focuseson healthy food choices, fat and calorieintake, and physical activity. Membersfrom the original DPP lifestyle team

Table 1dBaseline characteristics of REACT intervention participants

Characteristics All

Study group

PFF DVD INT SS

n 434 119 113 101 101Age (years) 51.1 6 11.3 50.9 6 11.3 52.4 6 10.9 48.7 6 9.7 52.3 6 12.7 0.07Weight (lbs) 215.3 6 44.4 216.7 6 46.4 215.2 6 47.4 218.9 6 43.7 209.9 6 39.2 0.52BMI (kg/m2) 36.3 6 6.6 37 6 6.9 36.2 6 7.2 36.1 6 6.4 34.9 6 5.7 0.08Waist circumference (inches) 44.2 6 6.0 44.1 6 5.9 44.7 6 6.7 44.3 6 5.7 43.5 6 5.4 0.58Glucose (mg/dL) 98.1 6 12.9 93.9 6 10.8 100.8 6 12.7 97.5 6 15.2 101.4 6 11.6 ,0.0001Women 374 (86.2) 106 (87.6) 96 (85.0) 89 (88.1) 98 (97) 0.45Non-Hispanic white 420 (96.8) 119 (100) 106 (93.8) 100 (99.1) 98 (97) 0.06Education#High school graduate 108 (24.9) 41 (34.5) 25 (22.1) 19 (18.8) 23 (22.8) 0.15Some college 189 (43.6) 46 (38.7) 51 (45.1) 51 (50.5) 41 (40.6)$College graduate 137 (31.6) 32 (26.9) 37 (32.7) 31 (30.7) 37 (36.6)

Income (USD),20,000 37 (9.0) 9 (8.0) 10 (8.9) 6 (6.1) 12 (13.2) 0.2520,000–39,999 90 (21.8) 27 (24.1) 31 (27.9) 18 (18.2) 14 (15.4)40,000–69,999 160 (38.7) 47 (42) 37 (33.3) 44 (44.4) 32 (35.2).70,000 126 (30.5) 29 (26) 33 (29.7) 31 (31.3) 33 (36.3)

Current smoker 40 (9.3) 11 (9.3) 10 (8.9) 14 (13.9) 5 (5.0) 0.19Family history of diabetes 270 (66) 74 (62.6) 70 (68.6) 66 (70.2) 62 (63.3) 0.58Class attendance/session viewsFF class attendance (range 1–12) 9.6 6 3.3DVD review session attendance (range 1–4) 2.9 6 1.2INT sessions viewed (range 0–12) 6.8 6 4.5SS FF class attendance (range 1–12) 8.4 6 3.3SS INT sessions viewed (range 3–12) 9.2 6 3.3

Medical conditionsHyperlipidemia 244 (63.4) 81 (69.8) 73 (70.1) 48 (54.6) 42 (53.9) 0.01Hypertension 123 (46.2) 43 (44.8) 25 (39.1) 21 (42.9) 34 (59.7) 0.04

Weight loss attempted in last 12 months 295 (68.9) 86 (72.9) 73 (65.8) 65 (65.7) 71 (71) 0.57

Data are means 6 SD or n (%) unless otherwise indicated. Boldface data indicate statistically significant differences between study groups (P , 0.05).

204 DIABETES CARE, VOLUME 36, FEBRUARY 2013 care.diabetesjournals.org

Comparative effectiveness of lifestyle intervention

collaborated to adapt and update theindividual intervention to a 12-weekgroup-based program (22). The GLB wasdelivered via three modalities (FF, DVD,and INT) for the purposes of this study.Despite the modality, all content was thesame. The theoretical framework for allmodalities was based on social cognitivetheory and incorporated behavioral self-management approaches designed tohelp participants set goals, problem solve,make behavior changes, and increase theirself-efficacy. The weight loss goal for theGLB is for participants to lose at least 5%oftheir body weight and/or decrease at leastone cardiovascular risk factor after the life-style intervention.

All participants, in all groups,received a copy of the GLB handouts, afat and calorie counter, self-monitoringbooks for keeping track of food andphysical activity, a pedometer, measuringcups and spoons, and a chart for record-ing weekly weights. Participants wereasked to self-monitor food intake andphysical activity throughout the 12-weekintervention and were given feedbackconcerning progress. A description ofeach GLB modality is outlined below.Face to face. One hundred and nineteenparticipants from two communities par-ticipated in FF, which consisted of 12group-education sessions that took placeover the course of 12–14 weeks. Par-ticipants met as a group for up to 90min/session each week. One trained

preventionist per community deliveredthe intervention. One lay health coachper community communicated with par-ticipants and identified barriers and solu-tions to promote program engagementand retention and provided informationaland emotional support to participantsupon their request. In addition, the layhealth coaches aided in the logistics ofthe study and shared relevant experiencesto initiate class discussion.DVD. One hundred and thirteen partic-ipants from two communities took part inthe DVD intervention. The DVD serieswas based on the GLB program, coveringall 12 weekly sessions. Professional actorsportrayed the preventionist and partici-pants in the sessions. Participants com-pleted the program over a 12- to 14-weekperiod. They were provided with a set ofDVDs that contained four weekly sessionsat a time. It was recommended thatparticipants watch one session per week,perhaps setting aside a specific day andtime each week to view each session.Participants then met as a group at fourtime points throughout the 12-week pe-riod to debrief the previous sessions theywatched and receive a new 4-week set ofDVDs. All activities were completed as ifthey were attending an FF group session,including keeping track of what they ateand their physical activity levels. Preven-tionists and lay health coaches calledparticipants weekly to offer information,support, and reminders as needed.

Internet. One hundred and one partic-ipants from two communities took part inthe INT intervention. INT was developedspecifically for Results of the RethinkingEating and ACTivity (REACT) and con-sisted of the use of the DVDs as describedabove but through internet access. Inaddition to viewing of the DVDs online,the INT incorporated behavioral toolssuch as e-mail prompts for online self-monitoring of eating patterns, physicalactivity, and weight and a graphing capa-bility to visualize progress made towardstated goals. Participants met as a group atbaseline and again after finishing the 12-to 14-week intervention. Preventionistsand lay health coaches supported partic-ipants via online counseling. If staff foundthat a participant did not log onto theREACT website for.1 week and did notrespond to an e-mail inquiry, a phone callwas made.Self-selection. One hundred and oneparticipants from two communities inthis study arm were able to self-selectthe intervention modality (as describedabove) of their choosing. Participantswere limited to one modality to avoidcontamination and bias in the results. Ofparticipants, 60% chose FF, 0% choseDVD, and 40% chose INT.Data collection. Eligibility, baseline, andfollow-up data were collected through in-person visits. Participants were asked tomake in-person visits at 3 and 6 monthsafter enrollment into the study. At eachvisit, anthropometric measures (height,weight, blood pressure, and waist cir-cumference) were collected. Weight wasmeasured on a high-quality calibrateddigital scale with the participant wearingclothes but no shoes. Height was mea-sured using a stadiometer. Blood pressurewas measured using the auscultatorymethod (23), and waist circumferencewas measured at the umbilical waist(24). Trained research staff completedthe measurements. Questionnaires thatassessed sociodemographic characteris-tics, quality of life, and other comorbidi-ties were completed. Physical and mentalfunctioning was measured with the Med-ical Outcomes Study Short Form 12 (25).All instruments are validated and aredeemed reliable in similar populations(25–27). Trained research staff per-formed the measurements. CVD risk fac-tors were defined using the NationalCholesterol Education Program’s AdultTreatment Panel III definition of meta-bolic syndrome (28). More specifically,hypertension was defined as systolicFigure 2dMean weight change by intervention group.

care.diabetesjournals.org DIABETES CARE, VOLUME 36, FEBRUARY 2013 205

Piatt and Associates

blood pressure $130 mmHg and/or dia-stolic blood pressure$85mmHg. Hyper-lipidemia was defined as triglyceridelevels $150 mg/dL.

During the in-person data collectionvisits, participants were provided with alaboratory requisition to have their blooddrawn at baseline and again at 3 and 6months after study enrollment. Bloodsamples were collected after an 8-h fastto determine glucose, triglycerides, andHDL cholesterol levels. Triglycerides andHDL cholesterol were measured by enzy-matic assays using the Dade Behring RXL.Blood glucose was measured by the hexo-kinase method using the Dade BehringRXL. All results were mailed to partic-ipants and their physicians. The overall3-month response rate was 60% (260 of434) and 6-month response rate was46.8% (203 of 434). Three-month re-sponse rates by group were as follows: FF80.7%, DVD 56.7%, INT 43.6%, and SS55.4%. Six-month response rates by in-tervention group were FF 72.2%, DVD38.1%, INT 35.6%, and SS 37.6%) (Fig. 1).Analyses. The primary outcome of thisstudy was change in weight from baselineto 3- and 6-month follow-up. Secondaryoutcomes included changes in CVD riskfactors, including glucose, triglyceridelevels, waist circumference, blood pres-sure values, and HDL cholesterol.

The primary analysis was based onthe intention-to-treat principle. Imputa-tion analyses that carried baseline valuesforward and the mean values forwardprovided no change in interpretation ofresults. The statistical analysis of the studyincorporated both descriptive and infer-ential techniques. Data are presentedusing descriptive statistics (e.g., propor-tions, means, SDs) by lifestyle group andby reassessment time. The analyses ex-amined whether there were within-groupdifferences from baseline to 3-monthfollow-up and whether improvementscould be maintained at 6-month follow-up. For examination of differences betweenstudy groups, a combined between- andwithin-group repeated-measures ANOVAwas performed for each outcome of in-terest. Statisticalmodelingwas then used toinvestigate whether there were differentialeffects on outcomes due to process ordemographic differences. For investigationof the possible effect of the clustering ofindividuals within communities, mixedmodeling (SAS PROC MIXED) was used.Models were adjusted for the baseline valueof the dependent variable, age, sex, com-munity, and the clustering of participants

within communities. The study was de-signed to have 80% power to detect abetween-group difference of 5% change inweight, with a two-sided significance levelof 0.05.

RESULTSdFive hundred and fifty-fiveindividuals were screened for BMI $25kg/m2 and abdominal obesity. Ofscreened individuals, 86.1% were femaleand 95.1% were non-Hispanic white, re-flecting the racial makeup of the studyarea. Mean weight and waist circumfer-ence were 209.5 6 45.6 lbs and 43.1 618.8 inches, respectively. Of those whowere screened, 493 (88.8%) were eligibleto participate in the intervention and 434(88%) enrolled (Fig. 1). When the screen-ing population (n = 555) was comparedwith the intervention population (n =434) to determine generalizability, therewere no statistical differences in meanweight (209.5 6 45.6 vs. 215.3 6 44.4lbs), waist circumference (43.16 18.8 vs.44.2 6 12.4 inches), sex distribution(86.1% female vs. 86.2% female), or race(95.1 vs. 96.8% non-Hispanic white).

Mean age of intervention participants(n = 434) was 51.1 years; 86.2% were fe-male and 96.8% non-Hispanic white.Mean weight, BMI, and waist circumfer-ence at baseline was 215.3 6 44.4 lbs,36.9 6 9.2 kg/m2, and 44.2 6 12.4 in-ches, respectively (Table 1). Baseline so-ciodemographic characteristics of theintervention participants are presentedin Table 1 overall and by interventiongroup. Average class attendance for FFwas 9.6 sessions. Participants in INTviewed an average of 6.8 sessions online,and participants in DVD attended an av-erage of 2.9 DVD debriefing sessions outof 4 possible. SS participants who choseFF attended an average of 8.4 classes,while those who chose INT viewed an av-erage of 9.2 sessions online (Table 1).There were no differences betweengroups in sex distribution, educationlevel, income, smoking status, family his-tory of diabetes, or whether participantsattempted weight loss during the 12months preceding the study (Table 1).

Weight lossThere was a marked decline in weightfrom baseline to 3 months in all interven-tion groups (FF212.5 lbs, P = 0.01; DVD212.2 lbs, P , 0.0001; INT 213.7 lbs,P, 0.0001; and SS214 lbs, P, 0.0001),with participants in SS experiencing thelargest average weight loss. At 6 months,the mean weight loss from baseline was

210.8 lbs in FF (P , 0.0001), 27.5 lbsin DVD (P , 0.0001), 26.8 lbs in INT(P , 0.0001), and 28.7 lbs in SS (P ,0.0001) (Fig. 2). Significant differences inmean weight loss between groups at 3 and6 months were observed, with SS havingthe greatest average weight loss at 3months (P = 0.03) and FF having thegreatest average loss at 6 months (P =0.05) compared with other groups. Whenthe effect of group was adjusted for theclustering of participants within communi-ties, baseline weight, age, and sex, themag-nitude of the association remained strongat 3 (P = 0.004) and 6 (P = 0.004) monthsin favor of FF.

When weight loss goals were exam-ined at 6 months, the proportion ofparticipants with a weight that was lowerthan their baseline weight was 71.4% inFF compared with 53.1, 42.6, and 55.5%in DVD, INT, and SS, respectively. Of theparticipants who achieved the 5% weightloss goal at 3 months (FF 57.2%, DVD56.7%, INT 62%, and SS 66.7%), .90%in each group sustained the weight loss at6 months (FF 90.7%, DVD 90.9%, INT92.1%, and SS 100%) (data not shown).

Cardiovascular risk factorsAll groups experienced significant im-provements in the proportion of partic-ipants with CVD risk factors after theintervention. The proportion of partici-pants with abdominal obesity signifi-cantly decreased in all groups frombaseline to 3 and 6 months (Fig. 3). Sim-ilarly, the proportion of participants withhypertension decreased in all groupsand significantly in FF (P , 0.01), DVD(P , 0.05), and INT (P , 0.0001). Tri-glyceride and HDL cholesterol levels alsoimproved but not in INT. Indeed, INTexperienced significant increases (P ,0.01) in the proportion of participantswith abnormal triglyceride levels overtime (Fig. 3). Additionally, the proportionof participants with impaired fasting glu-cose (IFG) decreased in all groups frombaseline to 3 and 6 months except forINT. SS had the largest decrease in theproportion of participants with IFG. Infact, of the 41 individuals who had IFGat baseline in SS, 26 no longer had IFG at3 months (data not shown). All associa-tions remained after adjustment for theeffect of clustering of participants withincommunities, age, sex, and baseline CVDrisk factors. The proportion of individualswith CVD risk factors remained steadybetween 3 and 6 months in all groupsand never returned back to baseline

206 DIABETES CARE, VOLUME 36, FEBRUARY 2013 care.diabetesjournals.org

Comparative effectiveness of lifestyle intervention

Figure 3dChanges in proportion of participants with cardiovascular risk factors across study groups after lifestyle intervention (baseline to 6-month follow-up). xP values represent within-group differences from baseline to 6-month follow-up. *P , 0.05, **P , 0.01, ***P , 0.0001.wAssociations remained after adjustment for the effect of clustering of participants within communities, age, sex, and baseline values. HDLc, HDLcholesterol.

care.diabetesjournals.org DIABETES CARE, VOLUME 36, FEBRUARY 2013 207

Piatt and Associates

(Fig. 3). When weight loss and CVD riskfactors were examined within the SSgroup and stratified by intervention mo-dality (FF versus INT), no differences inassociations were found.

CONCLUSIONSdIn this comparativeeffectiveness study, in which overweightand abdominally obese participants fromeight underserved rural communities wereenrolled, four modalities (FF, DVD, INT,and SS) of lifestyle intervention achievedstatistically and clinically significant weightloss, despite the modality. The observedweight loss was similar to weight loss ob-served in efficacy studies (3–8), with SSparticipants experiencing the largest aver-age weight loss at 3 months and FF partic-ipants experiencing the largest averageweight loss at 6 months. More than one-half of all participants lost at least 5% oftheir total body weight after the interven-tion, weighing on average 20 lbs less thanthey did at baseline.

Evidence demonstrates that 5% weightloss is clinically meaningful and has docu-mentedhealth benefits, including improveddiabetes and CVD risk factors (29). Allgroups experienced reductions in theproportion of individuals with impairedfasting glucose, abdominal obesity, andhypertension after the intervention. FFand SS were the only groups to achieve sig-nificant improvements in HDL cholesterol.

The data presented in this studysupport the work of others who designedand implemented lifestyle interventionsbased on the DPP in community settings.Numerous studies have reported initialweight loss and CVD risk reduction after alifestyle intervention (9,13,14,30) anddemonstrated that these types of inter-ventions are feasible and can be effectivelyimplemented in a number of settings andpopulations. However, few studies havemade head-to-head comparisons of mo-dalities of lifestyle intervention (31,32).We believe that the REACT study is thefirst scalable public health attempt at ex-amining the comparative effectiveness ofpreviously developed lifestyle interventionmodalities so that informed decisions canbe made regarding primary prevention ofdiabetes and CVD. Additionally, it is oneof the first studies to examine the effective-ness of empowering a group of parti-cipants to choose the interventionmodality that best suited their lifestyle.

In conducting community interven-tions, there are various limitations thatmay affect the study results. Although weimplemented standardized interventions

and data collection across all communi-ties and study groups, we had to tailorour approach to reflect the setting (e.g.,community-based hospitals, communitygathering places, community work sites,etc.). Additionally, because of the real-world setting of our study, we had toincrease our flexibility at times. For ex-ample, we encouraged all participants toattend the group sessions and the data-collection reassessments that were part oftheir study group; however, we could notmandate it. Although attendance washighly encouraged, average attendance/curriculum viewing was ~9 of 12 sessions.As attrition may be perceived as a limita-tion in our data with decreasing responserates at 3 and 6 months, all analyses werecarried out using intention-to-treat meth-odology and included the use of mixedmodels that addresses attrition while notoverestimating the intervention effect(33,34). We were powered at 80% to de-tect significant differences in the primaryand secondary outcomes. Therefore, it isunlikely that the reported findings aresubject to type II error. Indeed, the find-ings that showed statistically significantdifferences represent true differences.

The time frame of this REACT reportalso poses a limitation. A 6-month timeframe limits generalizability of the results,as many improvements in clinical out-comes cannot be determined in this shortperiod. A future report is in progress thatwill describe the 18-month results.

Although all REACT study groupsachieved notable improvements in out-comes, the largest and most significantimprovements were observed in the FFand SS groups. To that end, the REACTstudy provides a sort of template for whatmay be needed for successful community-based primary prevention efforts. Ourresults suggest that individuals who areattempting to lose weight and decreasetheir risk for diabetes and CVD benefitfrom accountability to their peers andhealth care providers in a group-basedformat and the option to choose thetype of intervention that best suits theirlifestyle. Future reports will focus on thelong-term maintenance of the observedimprovements and the comparative cost-effectiveness of the lifestyle-interventionmodalities.

Health professionals, payers, and pol-icy makers worldwide increasingly recog-nize the need for cost-effective, scalable,community-based primary prevention ef-forts. The REACT study provides the firstopportunity to compare the effectiveness

of several lifestyle interventions in multi-ple underserved rural communities. For ameaningful impact on public health pol-icy and clinical care, understanding thecomparative effectiveness of each primaryprevention modality in multiple commu-nity settings is critical. Efforts to makeprimary prevention a billable and reim-bursable service are ongoing at the localand national levels.

AcknowledgmentsdThis material is basedon research sponsored by the Air Force Sur-geon General’s office under agreement no.FA7014-08-2-0001.No potential conflicts of interest relevant to

this article were reported.G.A.P. wrote themanuscript and researched

data. M.C.S. and R.O.P. researched data andreviewed and edited the manuscript. J.C.Z.reviewed and edited the manuscript. G.A.P. isthe guarantor of this work and, as such, hadfull access to all the data in the study and takesresponsibility for the integrity of the data andthe accuracy of the data analysis.Parts of this studywere presented in abstract

form at the 71st Scientific Sessions of theAmerican Diabetes Association, San Diego,California, 24–28 June 2011.The authors acknowledge the University of

Pittsburgh Diabetes Institute for its supportwith the study. The authors also thank theircommunity partners (Indiana Regional MedicalCenter, L.R. Kimball & Associates, ConemaughMemorial Medical Center, UniontownHospital,Mon Valley Hospital, Highlands Hospital, andCenterville Clinics) for their support and effortin this project. The study would not have beenpossible without them.

References1. Ervin RB. Prevalence ofMetabolic Syndrome

Among Adults 20 Years of Age and Over,by Sex, Age, Race and Ethnicity, and BodyMass Index: United States, 2003–2006.Hyattsville, MD, National Center forHealth Statistics, 2009

2. United Health Foundation; AmericanPublic Health Association; Partnership forPrevention. The Future Costs of Obesity:National and State Estimates of the Impact ofObesity on Direct Health Care Expenses.Emory University, Atlanta, GA, 2009

3. Knowler WC, Barrett-Connor E, FowlerSE, et al.; Diabetes Prevention ProgramResearch Group. Reduction in the in-cidence of type 2 diabetes with lifestyleintervention or metformin. N Engl J Med2002;346:393–403

4. Pan X-R, Li G-W, Hu Y-H, et al. Effects ofdiet and exercise in preventing NIDDM inpeople with impaired glucose tolerance.The Da Qing IGT and Diabetes Study.Diabetes Care 1997;20:537–544

208 DIABETES CARE, VOLUME 36, FEBRUARY 2013 care.diabetesjournals.org

Comparative effectiveness of lifestyle intervention

5. Tuomilehto J, Lindstrom J, Eriksson JG,et al.; Finnish Diabetes Prevention StudyGroup. Prevention of type 2 diabetesmellitus by changes in lifestyle amongsubjects with impaired glucose tolerance.N Engl J Med 2001;344:1343–1350

6. Li G, Zhang P, Wang J, et al. The long-term effect of lifestyle interventions toprevent diabetes in the China Da QingDiabetes Prevention Study: a 20-yearfollow-up study. Lancet 2008;371:1783–1789

7. Lindström J, Ilanne-Parikka P, PeltonenM, et al.; Finnish Diabetes PreventionStudy Group. Sustained reduction in theincidence of type 2 diabetes by lifestyleintervention: follow-up of the FinnishDiabetes Prevention Study. Lancet 2006;368:1673–1679

8. Knowler WC, Fowler SE, Hamman RF,et al.; Diabetes Prevention Program Re-search Group. 10-year follow-up of di-abetes incidence and weight loss in theDiabetes Prevention Program OutcomesStudy. Lancet 2009;374:1677–1686

9. Colagiuri S, Vita P, Cardona-Morrell M,et al. The Sydney Diabetes PreventionProgram: a community-based translationalstudy. BMC Public Health 2010;10:328

10. Seidel MC, Powell RO, Zgibor JC,Siminerio LM, Piatt GA. Translating theDiabetes Prevention Program into anurban medically underserved commu-nity: a nonrandomized prospective in-tervention study. Diabetes Care 2008;31:684–689

11. Ackermann RT, Finch EA, Brizendine E,Zhou H, Marrero DG. Translating theDiabetes Prevention Program into thecommunity. The DEPLOY Pilot Study.Am J Prev Med 2008;35:357–363

12. Absetz P, Valve R, Oldenburg B, et al.Type 2 diabetes prevention in the “realworld”: one-year results of the GOALImplementation Trial. Diabetes Care2007;30:2465–2470

13. Amundson HA, Butcher MK, Gohdes D,et al.; Montana Cardiovascular Diseaseand Diabetes Prevention Program Work-group. Translating the diabetes preven-tion program into practice in the generalcommunity: findings from the MontanaCardiovascular Disease and Diabetes Pre-vention Program. Diabetes Educ 2009;35:209–210, 213–214, 216–220 passim

14. Kramer MK, Kriska AM, Venditti EM,et al. Translating the Diabetes Prevention

Program: a comprehensive model forprevention training and program delivery.Am J Prev Med 2009;37:505–511

15. Whittemore R, Melkus G, Wagner J,Dziura J, Northrup V, GreyM. Translatingthe diabetes prevention program to pri-mary care: a pilot study. Nurs Res 2009;58:2–12

16. Katula JA, Vitolins MZ, Rosenberger EL,et al. One-year results of a community-based translation of theDiabetes PreventionProgram: Healthy-Living Partnerships toPrevent Diabetes (HELP PD) Project. Di-abetes Care 2011;34:1451–1457

17. Norris SLZX, Avenell A, Gregg E, BrownT, Schmid CH, Lau J. Long-term non-pharmacological weight loss interventionsfor adults with type 2 diabetes mellitus.Cochrane Database Syst Rev 2005;CD005270

18. Wing RR; Look AHEAD Research Group.Long-term effects of a lifestyle inter-vention on weight and cardiovascular riskfactors in individuals with type 2 diabetesmellitus: four-year results of the LookAHEAD trial. Arch Intern Med 2010;170:1566–1575

19. The Trials of Hypertension PreventionCollaborative Research Group. Effects ofweight loss and sodium reduction in-tervention on blood pressure and hyper-tension incidence in overweight peoplewith high-normal blood pressure. TheTrials of Hypertension Prevention, phaseII. Arch Intern Med 1997;157:657–667

20. Appel LJ, Champagne CM, Harsha DW,et al.; Writing Group of the PREMIERCollaborative Research Group. Effects ofcomprehensive lifestyle modification onblood pressure control: main results of thePREMIER clinical trial. JAMA 2003;289:2083–2093

21. U.S. Census Bureau. State/county facts[article online], 2009. Available fromhttp://quickfacts.census.gov. Accessed 15March 2012

22. University of Pittsburgh. University ofPittsburgh Diabetes Prevention SupportCenter, 2006. Available from http://www.diabetesprevention.pitt.edu. Accessed 18April 2012

23. Millay JM. Accurately measuring bloodpressure. Accurate Blood Pressure Measure-ment for Medical Instrumentation. Availablefrom http://accuratebloodpressure.com.Accessed 18 April 2012

24. Mason C, Katzmarzyk PT. Variability inwaist circumference measurements ac-cording to anatomic measurement site.Obesity (Silver Spring) 2009;17:1789–1795

25. Ware JE Jr, Kosinski M, Keller SD. A 12-Item Short-Form Health Survey: con-struction of scales and preliminary tests ofreliability and validity.MedCare 1996;34:220–233

26. Fitzgerald JT, Davis WK, Connell CM,Hess GE, Funnell MM, Hiss RG. De-velopment and validation of the DiabetesCare Profile. Eval Health Prof 1996:19;208–230

27. Toobert DJ, Hampson SE, Glasgow RE.The summary of diabetes self-care activi-ties measure: results from 7 studies and arevised scale. Diabetes Care 2000;23:943–950

28. U.S. Department of Health and HumanServices. ATP III Guidelines at a GlanceQuick Desk Reference. Bethesda, MD, Na-tional Institutes of Health, 2001 (NIHpubl. no. 01-3305)

29. Franz MJ, VanWormer JJ, Crain AL, et al.Weight-loss outcomes: a systematic re-view and meta-analysis of weight-lossclinical trials with a minimum 1-yearfollow-up. J Am Diet Assoc 2007;107:1755–1767

30. Jackson L. Translating the Diabetes Pre-vention Program into practice: a review ofcommunity interventions. Diabetes Educ2009;35:309–320

31. Kramer MK, Kriska AM, Venditti EM,et al. A novel approach to diabetes pre-vention: evaluation of the Group LifestyleBalance program delivered via DVD. Di-abetes Res Clin Pract 2010;90:e60–e63

32. McTigue KM, Conroy MB, Hess R, et al.Using the internet to translate an evi-dence-based lifestyle intervention intopractice. Telemed J E Health 2009;15:851–858

33. Burton P, Gurrin L, Sly P. Extending thesimple linear regression model to accountfor correlated responses: an introduction togeneralized estimating equations andmulti-level mixed modelling. Stat Med1998;17:1261–1291

34. Ten Have TR, Kunselman AR, PulkstenisEP, Landis JR. Mixed effects logistic re-gression models for longitudinal binaryresponse data with informative drop-out.Biometrics 1998;54:367–383

care.diabetesjournals.org DIABETES CARE, VOLUME 36, FEBRUARY 2013 209

Piatt and Associates