An Automated Internet BehavioralWeight-Loss Program by PhysicianReferral: A RandomizedControlled TrialDiabetes Care 2015;38:9–15 | DOI: 10.2337/dc14-1474

OBJECTIVE

To evaluate 3- and 6-month weight-loss outcomes achieved when physicians referoverweight/obese patients to an automated 3-month Internet-based behavioralweight-loss intervention.

RESEARCH DESIGN AND METHODS

A total of 154 patients age 18–70 years with a BMI between 25 and 45 kg/m2 andaccess to a personal computer and the Internet were randomly assigned to 3months of Internet behavioral intervention (IBI; n = 77) with 12 weekly videosteaching behavioral weight-loss skills, a platform for submitting self-monitoreddata, and automated feedback or an education-only Internet-delivered eating andactivity control group (IDEA; n = 77). Outcome measures were weight loss after 3months (primary outcome) and 6months and changes inweight-control behaviors(secondary outcomes).

RESULTS

In intent-to-treat analyses with baseline weight carried forward for missing data,IBI produced significantly larger mean (SD) weight losses than IDEA at 3 months(5.5 kg [4.4] vs. 1.3 kg [2.1]) and 6 months (5.4 kg [5.6] vs. 1.3 kg [4.1]) (P < 0.001).Participants in IBI compared with IDEAwere also more likely to achieve a clinicallysignificant weight loss of 5%of initial bodyweight at 3months (53.3 vs. 9.1%) and 6months (48.1 vs. 15.6%) (P < 0.001) and reported more frequent use of weightcontrol–related strategies.

CONCLUSIONS

Physician referral to an Internet-based behavioral weight-loss intervention pro-duced clinically significantweight loss for over half of the patients studied. Furtherresearch is needed to determine the effectiveness of implementing this interven-tion more broadly within diverse health care settings.

Over two-thirds of U.S. adults are overweight or obese and therefore at increasedrisk of developing type 2 diabetes (1,2). Physicians are in an excellent position toidentify these individuals and advise them to lose weight, which has been shown toreduce the risk of developing diabetes and to improve a variety of health outcomes(3,4). However, few physicians provide their patients with counseling to achieveweight loss (5). The physician counseling literature suggests that having physiciansadvise their patients to lose weight, while referring them to an effective program or

1Department of Psychiatry and Human Behavior,Warren Alpert Medical School of Brown Univer-sity, The Miriam Hospital/Weight Control andDiabetes Research Center, Providence, RI2College of Agriculture and Natural ResourcesDepartment of Allied Health Sciences, Universityof Connecticut, Storrs, CT

Corresponding author: J. Graham Thomas,jthomas4@lifespan.org.

Received 16 June 2014 and accepted 30 Septem-ber 2014.

Clinical trial reg. no. NCT01222858, clinicaltrials.gov.

This article contains Supplementary Data onlineat http://care.diabetesjournals.org/lookup/suppl/doi:10.2337/dc14-1474/-/DC1.

A slide set summarizing this article is availableonline.

© 2015 by the American Diabetes Association.Readers may use this article as long as the workis properly cited, the use is educational and notfor profit, and the work is not altered.

J. Graham Thomas,1 TriciaM. Leahey,2 and

Rena R. Wing1

Diabetes Care Volume 38, January 2015 9

CLIN

CARE/ED

UCATIO

N/N

UTR

ITION/PSYC

HOSO

CIAL

allied health counselors rather than try-ing to offer counseling themselves, maybe the most effective model for obesitytreatment (3).Unfortunately, there are few options

for reimbursed weight-loss programs towhich physicians can refer their patients(6). Behavioral weight-loss programs(BWLs) such as the lifestyle interventionused in the Diabetes Prevention Pro-gram (DPP) produce weight losses of7–10%, reduce the risk of developingtype 2 diabetes, and improve cardiovas-cular disease risk factors (7–10). How-ever, these programs typically involvefrequent group or individual sessions,are costly to both patients and pro-viders, may not appeal to large numbersof patients seeking to lose weight, andare not widely available outside of re-search studies (8). Efforts to translatethe DPP for delivery by medical andallied health professionals (e.g., in pri-mary care practices) and in communitysettings such as the YMCA haveproduced a modest mean weight lossof 3.2–4.3% of initial body weight as de-termined by meta-analysis (11). Thereare also continued concerns about thecost, reach, accessibility, and long-termsustainability of these programs, whichhave relied heavily on in-person treat-ment delivery that is not well financedby the current health care system (11).Shortcomings associated with in-

person treatment delivery may be miti-gated by adapting BWLs for Internetdelivery (12,13), but there have beenfew randomized controlled trials testingthese programs as a resource for physi-cians to refer their overweight andobese patients. In early studies of pa-tients recruited through health care pro-viders, the Internet was used to providetailored or personal advice but not astructured BWL with a clearly definedcurriculum of training in empirically val-idated behavioral weight-loss strategies.Attrition was high in these studies, andweight losses were quite limited (14,15).More recent studies of physician referralto Internet programs have used strongerbehavioral programs but have providedsubstantial contact with health carecoaches in person, by phone, and/orvia e-mail (16–19). Although weight los-ses have been better in some of thesestudies [e.g., 4.8 kg at 12 months (16)and 5.1 kg at 24months (18)], the humaninteraction once again makes these

programs more expensive to deliverand less easily disseminated.

The primary objective of the RxWeight Loss Trial was to examine theweight losses produced by a nearly fullyautomated Internet BWL comparedwith a healthy eating and activity Inter-net newsletter control condition, aftera 3-month program and at 3-monthfollow-up, among patients who wererandomized after accepting a referralto the trial by their physicians.

RESEARCH DESIGN AND METHODS

Design OverviewParticipants were referred to the RxWeight Loss Trial by 35 physicians at 5primary care and 1 endocrinology prac-tice in Rhode Island. Physicians identifiedpotential participants during routine careand reviewed the study and inclusion/ex-clusion criteria with them. Physicianswere especially encouraged to refer indi-viduals with one or more obesity-relatedcomorbidities, such as diabetes, hy-pertension, or metabolic syndrome. A re-ferral form was signed by both thephysician and patient and faxed to theresearchers. Interested participantswere contacted by the research staff,phone screened for eligibility, and sched-uled for an orientation session, at whichtime further information was providedabout the study and informed consentcompleted. Participants were scheduledfor a baseline assessment and randomi-zation/program kickoff visit ;1 weeklater.

Participants were randomly assignedto a 3-month Internet behavior therapyprogram or Internet-Delivered Eatingand Activity (IDEA) control conditionwith information on healthy eating hab-its and physical activity. During both therandomization/program kickoff visit andin their lesson materials, participants inboth conditions received informationpertaining to a healthy rate of weightloss and safe engagement in physical ac-tivity. The importance of regular contactwith the referring physician for adjust-ments in medications was stressed. Ac-cess to the Internet interventions endedafter 12 weeks in both groups. Bothgroups were assessed at the end of the3-month program (posttreatment) andafter 3 additional months of follow-upand received $25 and $50, respectively,for completing the assessments. Studyprocedures were approved by the

Institutional Review Board of The MiriamHospital.

ParticipantsParticipants were 154 overweight orobese English-speaking individualswith a BMI of 25–45 kg/m2, age 18–70years old, with access to a personal com-puter and the Internet. Exclusion criteriaincluded a weight loss of $5% of initialbody weight within the last 6 months,current participation in another weight-loss program or current use of weight-loss medication, participation in aweight-loss program at the study clinicwithin the last 2 years, livingwith anotherstudy participant, pregnant or breast-feeding within the last 6 months, a planto become pregnant or move to a newgeographic region within 6 months, cur-rent chemotherapy or radiation treat-ment for cancer, untreated seriousmental illness, or hospitalization for men-tal illness within the last 12 months.

Randomization and InterventionsParticipants were randomly assigned toone of the two treatment conditionsusing a computer-generated permutedblocking procedure, stratified by sex. A1:1 randomization scheme was used. Theallocation sequence was concealed until apatient consented to participate and hadcompleted the baseline assessment.

Internet Behavioral Intervention

The Internet Behavioral Intervention(IBI) program included 12 weekly multi-media behavioral lessons, a website forsubmitting self-monitoring data, andweekly automated feedback providedto the participant on their progress todate. Participants were introduced tothe IBI program at the 60-min random-ization/program kickoff visit; they weregiven a goal of losing 1 to 2 lbs/week andachieving a total weight loss of$10% ofinitial body weight. To accomplish this,they were prescribed a calorie goal of1,200–1,500 kcal/day depending on ini-tial body weight with #30% caloriesfrom fat (40–60 g of fat/day) and a phys-ical activity goal that gradually increasedto 200 min/week of physical activity us-ing activities similar in intensity to briskwalking. The remainder of the random-ization/program kickoff visit was used toteach skills for self-monitoring dailyfood intake, physical activity minutes,and body weight. Participants receivedpaper diaries and a calorie referencebook for this purpose.

10 Internet Behavioral Weight-Loss Program Diabetes Care Volume 38, January 2015

Standard behavioral strategies forchanging their eating and activity weretaught to the participants by havingthem view weekly 10–15-min interac-tive multimedia behavioral weight-losslessons. In contrast to previous studiesusing static lessons with text andgraphics (20), interactive lessons incor-porating video, animation, audio, quiz-zes, and exercises for goal setting,planning, and problem-solving were de-veloped to improve patient engagement(21). These lessons were based on strat-egies used in the DPP and Look AHEADtrials and included topics such as restau-rant eating, changing the home environ-ment, and social support (9,22). Dailyvalues for body weight, caloric intake,fat intake (g), and physical activityminutes were submitted at least weeklyto the Rx Weight Loss website. In re-sponse to their submissions, partici-pants received a weekly automatedtailored feedback message. The feed-back messages were generated via analgorithm that compared participants’self-reported values to their goals forweekly and overall weight loss, caloricintake, and physical activity minutes.Participants received praise for meetinggoals and, when goals were not met, re-ceived specific recommendations forbehavioral strategies to implement,along with support and encouragement.Automated e-mail reminder messageswere sent weekly to participants whowere not using the study website. Al-though in our prior research we foundthat human feedback was associatedwith larger weight losses than automatedmessages (20), we used automated mes-sages in this trial because of the cost sav-ings and ease of dissemination.At weeks 4, 8, and 12, a letter was

sent to the referring physician (and cop-ied to the participant) indicating thenumber of weeks the participant hadsubmitted data and their weight lossat that point. The letters were intendedto provide physicians with informationon the outcome of their referral, facili-tate communication about health behav-iors between physicians and patients,and enhance patients’ sense of ac-countability for their health and healthbehaviors.

IDEA

Participants in this education-onlycontrol condition were seen at the

randomization/program kickoff visitand instructed to access the Rx WeightLoss website at least once weekly toview a new printable lesson with statictext and graphics in the Adobe portabledocument format. These newslettersprovided general educational informa-tion on the benefits of losing weightand healthy eating and physical activityhabits. Participants were not taught be-havioral weight-loss strategies and werenot instructed to self-monitor their be-haviors, but were encouraged to exploreonline resources such as choosemyplate.gov, which contains information and toolsfor weight loss. This condition mimickedwhat patients may receive during routinecare: a recommendation to lose weightaccompanied by general educational in-formation. IDEA also served as a methodfor controlling for access to Internet-based resources.

Outcomes and Follow-upThe primary outcome was change inbody weight (kg, percent of initial bodyweight). Weight was measured in theresearch setting at baseline, 3 months(end of treatment), and 6 months

(follow-up) in light street clothing, with-out shoes, and on a calibrated scale byblinded research staff. A secondaryoutcome was engagement with theInternet-based treatment system, whichwas measured by the number of weeks(out of 12) that participants logged in tothe website, and, in the IBI group only,the number of weeks that calories, phys-ical activity minutes, and body weightwere reported for at least 5 of the 7days. At baseline, participants reportedtheir demographic characteristics andphysician-diagnosed medical conditions.At baseline, 3, and 6 months, participantsreported the number of weeks in the past3 months that they had: 1) reduced thenumber of calories consumed; 2) de-creased fat intake; 3) increased fruitsand vegetables consumed; and 4) in-creased exercise levels. These itemswere drawn from the Weight ControlPractices questionnaire used in the LookAHEAD trial (23).

Statistical AnalysisPASW Statistics 19 (SPSS, Inc., 2009, Chi-cago, IL; http://www.spss.com) wasused for all analyses. Changes in weight

Figure 1—Flow of participants through the trial. The CONSORT flow diagram includes data onphysician referrals, patient enrollment, allocation to treatment groups, follow-up, and primaryanalysis.

care.diabetesjournals.org Thomas, Leahey, and Wing 11

and levelsof engagementwith the Internet-based treatment system were examinedvia separate repeated-measures ANOVAthat followed the intent-to-treat (ITT)principle in which missing data were re-placed by baseline weights (assumingno weight loss). A secondary ITT analysisusing multiple imputation by chainedequations was used to confirm the pat-tern of results from the initial ITT analysisgiven that replacing missing data withbaseline values could adversely affecttests of statistical significance (24). Givenno difference in the pattern of results,the more conservative analysis withmissing data replaced by baselineweights (assuming no weight loss) wasselected for detailed reporting. Correla-tions were used to evaluate associationsbetween engagement with the Internet-based treatment system and change inweight. Paired and independent samplest tests, respectively, were used to evalu-ate within-groups changes in rates ofweight-control behaviors and to com-pare IBI and IDEA on rates of weight-control behaviors at baseline, 3, and6 months. Unless noted otherwise, testsof significance were conducted at a =0.05; all tests were two-tailed. This trialwas designed to have 80% power to de-tect significant between-groups differ-ences in weight loss of $2 kg at 3 and6 months after randomization with 250participants. An interim analysis com-paring between-groups weight losswas conducted with 150 participants(4 more were in progress at the time),which resulted in early stopping of thetrial due to clear evidence of efficacy.The analysis, and decision to stop thetrial, were consistent with the a spend-ing function approach to interim dataanalysis in which a more stringent crite-rion for statistical significance is used, inthis case P , 0.004 (25).

RESULTS

Recruitment and RetentionThe CONSORT diagram is depicted in Fig.1. A total of 154 participants were ran-domized; 84.4% completed the assess-ment at the end of the 3 months, and79.2% completed the 6 month follow-up. All attrition was attributed to lossof contact with participants. Retentionrates were comparable in IBI and IDEA atboth 3 months (83.1% in IBI vs. 85.7% inIDEA) and 6 months (79.2% in IBI vs.79.2% in IDEA). There were also no

statistically significant differences inbaseline characteristics between partici-pants who did and did not attend a post-baseline assessment.

Sample CharacteristicsParticipants’ baseline characteristics arereported in Table 1. Eighty percent ofthe participants were women. Partici-pants averaged 53.2 (10.9) years ofage and had an entry weight of 94.9(16.4) kg. The most commonly endorsedphysician-diagnosed medical conditionswere hypertension (n = 80; 51.9% ofsample), arthritis (n = 53; 34.4% ofsample), cancer history (n = 22; 14.3%of sample), and type 2 diabetes (n = 20;13.0% of sample). At baseline, IDEA par-ticipants reported higher rates of de-creasing fat intake as a weight-lossstrategy; there were no other statisti-cally significant baseline differences be-tween IBI and IDEA.

ITT AnalysisGroup-specific changes in weight atmonths 3 and 6 are depicted in Table 2and Fig. 2. The IBI condition producedsignificantly greater weight loss thanIDEA at 3 months, 5.5 (4.4) vs. 1.3 (2.4) kg,and at 6 months, 5.4 (5.6) vs. 1.3 (4.1) kg

(P , 0.0001 for the overall effect ofgroup). Moreover, 53% of IBI lost at least5% of their body weight at 3 months,and 48% met the 5% weight goal at6 months. Of the individuals in IBI thatlost at least 5% of their body weight at3 months, 34% lost at least an additional1 kg from 3 to 6 months (mean [SD] lossof 2.9 [1.7] kg), 29% gained at least 1 kgfrom 3 to 6 months (mean [SD] gain of2.8 [1.2] kg), and 37% were within 1 kgof their 3-month weight at 6 months.Twenty-one percent of IBI participantslost at least 10% of their body weightat each assessment point. The propor-tion of IBI participants achieving clini-cally significant weight losses exceededthose in IDEA (Table 2). Nearly a third(25–30%) of participants in IDEA gainedweight at 3 and 6 months, respectively,whereas this occurred in only 2–7%of participants assigned to IBI (Supple-mentary Fig. 1). The secondary analysisusing multiple imputation produced thesame pattern of results and statisticallysignificant effects (not shown).

Engagement With the WebsiteIBI and IDEA did not differ significantlyin the number of weeks (out of 12) thatthe website was used during the initial

Table 1—Baseline characteristics of participants assigned to the IBI and IDEAcontrol conditions

Full sample(N = 154)

IBI condition(n = 77)

IDEA condition(n = 77)

Sex, no. (%)Men 31 (20.1) 15 (19.5) 16 (20.8)Women 123 (79.9) 62 (80.5) 61 (79.2)

Age, mean (SD), years 53.2 (10.9) 52.8 (10.2) 53.6 (11.6)

Race, no. (%)American Indian 4 (2.6) 1 (1.3) 3 (3.9)Asian 2 (1.3) 1 (1.3) 1 (1.3)Black 7 (4.6) 3 (3.9) 4 (5.2)White 136 (88.3) 70 (90.9) 66 (85.7)Other 10 (6.5) 5 (6.5) 5 (6.5)

Ethnicity, no. (%)Hispanic 8 (5.2) 5 (6.5) 3 (3.9)Non-Hispanic 146 (94.8) 72 (93.5) 74 (96.1)

Marital status, no. (%)Single 33 (21.4) 14 (18.2) 19 (24.7)Married 97 (63.0) 50 (64.9) 47 (61.0)Separated/divorced 24 (15.6) 13 (16.9) 11 (14.3)

Education, no. (%)High school or less 19 (12.3) 5 (6.5) 14 (18.2)Some college 44 (28.6) 27 (35.1) 17 (22.1)College or university degree 32 (20.8) 14 (18.2) 18 (23.4)Graduate degree 59 (38.3) 31 (40.2) 28 (36.3)

Weight, mean (SD), kg 94.9 (16.4) 95.5 (16.8) 94.3 (16.1)

BMI, mean (SD), kg/m2 34.9 (4.8) 34.9 (4.7) 34.9 (5.0)

There were no statistically significant differences between IBI and IDEA at baseline.

12 Internet Behavioral Weight-Loss Program Diabetes Care Volume 38, January 2015

3-month program (10.06 3.0 in IBI and9.5 6 3.1 in IDEA; P = 0.301). Using thewebsite for a greater number of weekswas associated with larger weight loss inthe IBI condition (r = 0.41; P , 0.001)but not in the IDEA condition (r = 0.18;P = 0.110).Participants in the IBI condition re-

ported their daily caloric intake, physicalactivity minutes, and daily weight on theintervention website at least 5 days outof the week on 6.7 (4.7) of the 12 weeks.Frequency of reporting was correlatedwith weight loss (r = 0.54; P , 0.001).

Weight-Control BehaviorsThe frequency of each weight-loss be-havior increased significantly from base-line to 3months in IBI and IDEA (P values

,0.001), but IBI participants reportedmaking changes during a greater num-ber of weeks during this 3-month period(Table 3). Although the frequency ofthese changes decreased over time (Ta-ble 3), at 6 months, IBI participants stillreported a greater frequency of reduc-ing calorie and fat intake than IDEA.

CONCLUSIONS

The Rx Weight Loss Trial found thatoverweight and obese patients referredby their physician to an Internet behav-ioral weight-loss program lost on aver-age 5.5 kg at the end of the 3-monthprogram and maintained this weightloss in full at 6 months. In addition,53% achieved a clinically significantweight loss of at least 5% of initialbody weight at 3 months and 48% at 6months. These results far exceeded theweight losses achieved by the patientswho were randomly assigned to an In-ternet newsletter control condition andprovide an effective, potentially low-cost option that physicians could usewith their overweight and obese pa-tients to reduce the risk of type 2 diabe-tes and other weight-related diseases.

Engagement with the IBI website wasquite high during treatment and corre-lated with weight-loss outcomes. Priorwork suggests that feedback messages,perhaps to an even greater degree thanonline lessons, contributed to improvedweight-loss outcomes in similar pro-grams (26). IBI participants also re-ported more frequent use of keybehaviors, such as reducing calorie andfat intake, than was seen in IDEA. This

suggests that successful skills acquisi-tion and implementation is at least par-tially responsible for the superior weightlosses in IBI. The use of key strategieswasgreatest during the 3-month treatmentand then appeared to decrease; this find-ing emphasizes the need to develop ap-proaches to maintaining adherence tothese behaviors longer term. Likewise,additional research could behelpful in de-termining the relative importance of IBIcomponents (e.g., goal-setting, lessons,self-monitoring, feedback, and reportsto referring physicians).

Weight losses in the IBI condition ap-pear to exceed those we achievedusing a similar 3-month interventionin a community setting (3.5 kg weightloss at 3 months) (26). This differencein outcome may reflect differences inthe two study groups. For example, par-ticipants referred by physicians mayhave more health concerns and there-fore greater motivation to lose weightcompared with individuals participatingin a community wellness campaign. Thedifference may also suggest that thereare added beneficial effects from physi-cians’ referrals and their involvement inthe process. For example, the letterssent monthly to physicians (and copiedto patients) reporting weight-loss prog-ress may have increased the patients’sense of accountability for their weight-loss success. Further research is neededto determine whether involving the phy-sician improves outcomes and how bestto maximize this effect.

Previous studies of weight-loss inter-ventions in primary care settings haveprimarily involved physician counseling,physician counseling combined withweight-loss medications, or delivery ofthe intervention by other office staff(27). In many cases, the interventionswere not based on empirically testedbehavioral weight-loss strategies, andthe results of these studies have beenvariable (27). We are aware of few stud-ies that used web-based approaches inprimary care, and as noted above, manyof these have involved substantial face-to-face (27,28) or remote counseling(16–19). Our results suggest that com-parable weight losses, at least initially,can be achieved with a nearly (i.e., withthe exception of the randomization/program kickoff visit) fully automatedInternet program. Likewise, our findingsappear consistent with efforts to

Figure 2—Change in weight at months 3 and6 in the ITT population for participants as-signed to the IBI and IDEA control condition.Means with 95% CI bars.

Table 2—Change in weight at months 3 and 6 (ITT) for participants assigned tothe IBI and IDEA control condition

IBI (n = 77) IDEA (n = 77)

Weight loss (kg), mean (SD)3 months 5.5 (4.4) 1.3 (2.4)***6 months 5.4 (5.6) 1.3 (4.1)***

Weight loss (% of initial weight), mean (SD)3 months 5.8 (4.4) 1.4 (2.7)***6 months 5.6 (5.6) 1.4 (4.6)***

Participants with no loss or weight gain, no. (%)3 months 1 (1.3) 20 (26.0)***6 months 5 (6.5) 23 (29.9)***

Participants with $5% weight loss, no. (%)3 months 41 (53.3) 7 (9.1)***6 months 37 (48.1) 12 (15.6)***

Participants with $10% weight loss, no. (%)3 months 16 (20.8) 1 (1.3)***6 months 16 (20.8) 4 (5.2)**

**P , 0.01, ***P , 0.001 for comparison of IBI and IDEA.

care.diabetesjournals.org Thomas, Leahey, and Wing 13

translate behavioral weight-loss inter-ventions, such as the DPP, for deliveryby medical and allied health profes-sionals and in community settings thathave achieved mean weight loss of3.2–4.3% of initial body weight after12 months (11). Compared with theseother efforts, the Rx Weight Loss ap-proach may incur fewer problems withcost, reach, accessibility, and long-termsustainability that have been a causefor concern with programs deliveredin-person (11).There are several limitations to this

study. First, the study lasted only 6months, and thus, it is not clear if theseweight losses will be maintained longterm. Since participants in behavioralweight-loss programs typically loseweight for the first 3–6 months of treat-ment (8) and then gradually regainweight, our study did not follow partic-ipants through the period during whichthey are at greatest risk of regain. Pre-vious studies have consistently shownthat continued contact is important formaintaining weight losses (8); providingfurther ongoing intervention and tailor-ing of the program to specific needs ofthe participant may promote long-termefficacy (29,30). Second, the study re-quired several visits to an academicmedical center that may have influencedoutcomes.At the randomization/programkickoff visit, both groups were seenface-to-face and introduced to the web-site and the key elements of the weight-loss intervention (self-monitoring andgoal-setting) were reviewed with partic-ipants in IBI. In the future, this informa-tion could be presented to participantsvia Internet-streamed video. Third, al-though some physicians referred manymore patients than others (data notshown), we do not know the factorsthat influenced this. Nor do we knowwhy some patients who agreed to a refer-ral decided not to enroll. An exploration

of factors affecting willingness of bothphysicians and patients to participate insuch a program could be important if theprogram were to be made widely avail-able. Fourth, while weight loss has beenreliably shown to improve cardiovasculardisease risk factors such as glucose control(31), we did not measure changes in thesephysiological indicators. Lastly, the samplehad low racial and ethnic minority repre-sentation and was generally well edu-cated; the generalizability of the resultsto other populations, including olderadults and those with disabilities shouldbe tested.

Strengths of the study include the sys-tem of referrals from physicians, whichrequired very little time on their part,and the fact that the approach weused is nearly fully automated and canthus be scaled for use with large num-bers of patients at minimal additionalcost. Moreover, we compared the IBIprogram to a control group that re-ceived weekly printable lessons via theInternet, thus controlling for engage-ment with the website and ongoing at-tention to weight control. The clinicallysignificant weight loss in IBI but not thecontrol group suggests that physicianadvice to lose weight combined withcommon diet and exercise informationis insufficient and that behavioral strat-egies and feedback are also required toproduce meaningful weight loss.

In conclusion, physician referral to abehavioral Internet intervention, whichincludes weekly behavioral lessons andweekly automated feedback to patientson their self-monitoring records, withperiodic written feedback to the physi-cian, can be an effective approach forweight loss in overweight and obese in-dividuals. Given that ;80% of U.S.adults use the Internet (32,33), thisapproach may provide a cost-effectivealternative or complement to the face-to-face counseling models of obesity

treatment, including that which is cur-rently reimbursable by the Centers forMedicare & Medicaid Services (34), andcould be paired with other Internet-based resources for diabetes manage-ment and risk reduction (35). Additionaltranslational research is needed to de-termine whether this intervention canbe fully implemented within currenthealth care settings and reimbursementmodels while preserving the positiveweight-loss outcomes.

Acknowledgments. The authors thank thestudy participants for their contribution to theresearch; the physicians for referring patientsto the trial; and Pam Coward, Katelyn Gettens,and Rachel Ogilvie, The Weight Control andDiabetes Research Center of The Miriam Hos-pital, for assistance with study implementation.Funding. This study was funded by a grant fromthe National Heart, Lung, and Blood Institute(RC1-HL-100002).Duality of Interest. No potential conflicts ofinterest relevant to this article were reported.Author Contributions. J.G.T. researched data,conducted the statistical analysis, and cowrotethe manuscript. T.M.L. contributed to thediscussion and reviewed and edited the manu-script. R.R.W. researched data and cowrotethe manuscript. J.G.T. is the guarantor of thiswork and, as such, had full access to all the data inthe study and takes responsibility for the integrityof the data and the accuracy of the data analysis.Prior Presentation. A portion of the findingsreported in this article was presented in posterform at the 30th Annual Scientific Meeting ofThe Obesity Society, San Antonio, TX, 20–24September 2012.

References1. Flegal KM, Carroll MD, Kit BK, Ogden CL.Prevalence of obesity and trends in the distri-bution of body mass index among US adults,1999-2010. JAMA 2012;307:491–4972. Bray GA. Obesity increases risk for diabetes.Int J Obes RelatMetabDisord 1992;16(Suppl. 4):S13–S173. KreuterMW, Chheda SG, Bull FC. Howdoes phy-sician advice influence patient behavior? Evidencefor a priming effect. Arch FamMed 2000;9:426–4334. Eden KB, Orleans CT, Mulrow CD, Pender NJ,Teutsch SM. Does counseling by clinicians im-prove physical activity? A summary of the

Table 3—IBI and IDEA control condition participants’ reports of their weight-control behaviors

Baseline 3 months 6 months

IBI (n = 77) IDEA (n = 77) P value IBI (n = 64) IDEA (n = 66) P value IBI (n = 61) IDEA (n = 61) P value

Reduce calories consumed 2.1 (3.7) 2.0 (3.1) 0.942 11.0 (2.5)* 5.9 (4.7)* ,0.001 6.6 (5.2)*‡ 4.3 (4.8)*‡ 0.013

Decrease fat intake 0.9 (2.4) 2.2 (3.6) 0.009 10.6 (2.9)* 5.9 (5.0)* ,0.001 7.2 (4.9)*‡ 5.1 (5.2)* 0.023

Increase fruits and vegetables 2.9 (4.0) 3.4 (4.4) 0.481 9.9 (3.9)* 7.6 (4.6)* 0.002 7.8 (5.0)*‡ 6.1 (4.9)*‡ 0.063

Increase exercise levels 1.9 (3.3) 2.8 (3.7) 0.882 8.7 (4.0)* 5.5 (4.9)* ,0.001 4.6 (5.1)*‡ 3.2 (4.6)‡ 0.096

Data representmean (SD) number of weeks in the past 3months that the behavior was endorsed. *Value is significantly different from baseline value(P , 0.05). ‡Value is significantly different from 3-month value (P , 0.05).

14 Internet Behavioral Weight-Loss Program Diabetes Care Volume 38, January 2015

evidence for the U.S. Preventive Services TaskForce. Ann Intern Med 2002;137:208–2155. Galuska DA, Will JC, Serdula MK, Ford ES. Arehealth care professionals advising obese patientsto lose weight? JAMA 1999;282:1576–15786. Tsai AG, Asch DA,Wadden TA. Insurance cov-erage for obesity treatment. J Am Diet Assoc2006;106:1651–16557. McTigue KM, Harris R, Hemphill B, et al.Screening and interventions for obesity inadults: summary of the evidence for the U.S.Preventive Services Task Force. Ann InternMed 2003;139:933–9498. Butryn ML, Webb V, Wadden TA. Behavioraltreatment of obesity. Psychiatr Clin North Am2011;34:841–8599. Diabetes Prevention Program (DPP) ResearchGroup. The Diabetes Prevention Program (DPP):description of lifestyle intervention. DiabetesCare 2002;25:2165–217110. Knowler WC, Fowler SE, Hamman RF, et al.;Diabetes Prevention Program Research Group.10-year follow-up of diabetes incidence andweight loss in the Diabetes Prevention ProgramOutcomes Study. Lancet 2009;374:1677–168611. Ali MK, Echouffo-Tcheugui J, WilliamsonDF. How effective were lifestyle interventionsin real-world settings that were modeled onthe Diabetes Prevention Program? Health Aff(Millwood) 2012;31:67–7512. Neve M, Morgan PJ, Jones PR, Collins CE.Effectiveness of web-based interventions inachieving weight loss and weight loss mainte-nance in overweight and obese adults: a system-atic review with meta-analysis. Obes Rev 2010;11:306–32113. Thomas JG, Bond DS. Review of innovationsin digital health technology to promote weightcontrol. Curr Diab Rep 2014;14:48514. Rothert K, Strecher VJ, Doyle LA, et al. Web-based weight management programs in an in-tegrated health care setting: a randomized,controlled trial. Obesity (Silver Spring) 2006;14:266–27215. McConnon A, Kirk SF, Cockroft JE, et al. TheInternet for weight control in an obese sample:

results of a randomised controlled trial. BMCHealth Serv Res 2007;7:20616. McTigue KM, Conroy MB, Hess R, et al. Us-ing the internet to translate an evidence-basedlifestyle intervention into practice. Telemed J EHealth 2009;15:851–85817. Bennett GG, Warner ET, Glasgow RE, et al.;Be Fit, Be Well Study Investigators. Obesitytreatment for socioeconomically disadvantagedpatients in primary care practice. Arch InternMed 2012;172:565–57418. Appel LJ, Clark JM, Yeh HC, et al. Compara-tive effectiveness of weight-loss interventionsin clinical practice. N Engl J Med 2011;365:1959–196819. Ma J, Yank V, Xiao L, et al. Translating theDiabetes Prevention Program lifestyle interven-tion for weight loss into primary care: a random-ized trial. JAMA Intern Med 2013;173:113–12120. Tate DF, Jackvony EH, Wing RR. A random-ized trial comparing human e-mail counseling,computer-automated tailored counseling, andno counseling in an Internet weight loss pro-gram. Arch Intern Med 2006;166:1620–162521. Norman GJ, Zabinski MF, Adams MA,Rosenberg DE, Yaroch AL, Atienza AA. A reviewof eHealth interventions for physical activityand dietary behavior change. Am J Prev Med2007;33:336–34522. Wadden TA, West DS, Delahanty L, et al.;Look AHEAD Research Group. The Look AHEADstudy: a description of the lifestyle interventionand the evidence supporting it [published cor-rection appears in Obesity (Silver Spring) 2007;15:1339]. Obesity (Silver Spring) 2006;14:737–75223. Raynor HA, Jeffery RW, Ruggiero AM, ClarkJM, Delahanty LM; Look AHEAD (Action forHealth in Diabetes) Research Group. Weightloss strategies associated with BMI in over-weight adults with type 2 diabetes at entry intothe Look AHEAD (Action for Health in Diabetes)trial. Diabetes Care 2008;31:1299–130424. Lachin JM. Statistical considerations in theintent-to-treat principle. Control Clin Trials2000;21:167–189

25. DeMets DL, Lan KK. Interim analysis: thealpha spending function approach. Stat Med1994;13:1341–1352; discussion 1353–135626. Wing RR, Crane MM, Thomas JG, Kumar R,Weinberg B. Improving weight loss outcomes ofcommunity interventions by incorporating be-havioral strategies. Am J Public Health 2010;100:2513–251927. Yoong SL, Carey M, Sanson-Fisher R, GradyA. A systematic reviewofbehaviouralweight-lossinterventions involving primary-care physicians inoverweight and obese primary-care patients(1999-2011). Public Health Nutr 2013;16:2083–209928. Bennett GG, Herring SJ, Puleo E, Stein EK,Emmons KM, Gillman MW. Web-based weightloss in primary care: a randomized controlledtrial. Obesity (Silver Spring) 2010;18:308–31329. Kreuter MW, Bull FC, Clark EM, Oswald DL.Understanding how people process health in-formation: a comparison of tailored and non-tailored weight-loss materials. Health Psychol1999;18:487–49430. Saperstein SL, Atkinson NL, Gold RS. Theimpact of Internet use for weight loss. ObesRev 2007;8:459–46531. Goldstein DJ. Beneficial health effects ofmodest weight loss. Int J Obes Relat Metab Dis-ord 1992;16:397–41532. Duggan M, Smith A. Cell internet use2013 [Internet], 2013. Available from http://pewinternet.org/Reports/2013/Cell-Internet.aspx. Accessed 31 July 201433. File T. Computer and Internet Use in theUnited States. Current Population Survey Re-ports. Washington, D.C., U.S. Census Bureau,2013, p. 20–56834. Centers for Medicare & Medicaid Services.Decisionmemo for intensive behavioral therapyfor obesity (CAG-00423N) [article online], 2013.Available from http://www.cms.gov/medicare-coverage-database/details/nca-decision-memo.aspx?&NCAId=253. Accessed 20 August 201335. Kaufman N. Internet and information tech-nology use in treatment of diabetes. Int J ClinPract Suppl 2010;166:41–46

care.diabetesjournals.org Thomas, Leahey, and Wing 15