Translating the Diabetes Prevention Program Lifestyle Intervention for Weight Loss into Primary Care: A Randomized Trial

Abstract

Background

The Diabetes Prevention Program (DPP) lifestyle intervention reduced the incidence of type 2 diabetes among high risk adults by 58%, with weight loss as the dominant predictor. However, it has not been adequately translated into primary care.

Methods

We evaluated two adapted DPP lifestyle interventions among overweight/obese adults who were recruited from one primary care clinic and had prediabetes and/or metabolic syndrome. Participants were randomized to (1) a coach-led group intervention (n=79), (2) a self-directed DVD intervention (n=81), or (3) usual care (n=81). During a 3-month intensive intervention phase, the DPP-based behavioral weight loss curriculum was delivered by lifestyle coach-led small groups or home-based DVD. During the maintenance phase, participants in both interventions received lifestyle change coaching and support remotely–through secure email within an electronic health record system and the American Heart Association Heart360 Web site for weight and physical activity goal setting and self-monitoring. The primary outcome was change in body mass index (BMI) from baseline to 15 months.

Results

At baseline, participants had a mean (±SD) age of 52.9±10.6 years and mean BMI 32.0±5.4 kg/m² with 47% female, 78% Non-Hispanic white, and 17% Asian/Pacific Islander. At month 15, the mean (±SE) change in BMI from baseline was −2.2±0.3 kg/m² in the coach-led group (vs. −0.9±0.3 kg/m² in the usual care group, P<0.001) and −1.6±0.3 kg/m² in the self-directed group (P=0.02 vs. usual care). The percentages of participants who achieved the 7% DPP-based weight loss goal were 37.0% (P=0.003) and 35.9% (P=0.004) in the coach-led and self-directed groups, respectively, versus 14.4% in the usual care group. Both interventions also achieved greater net improvements in waist circumference and fasting plasma glucose.

Conclusions

Proven effective in a primary care setting, the two DPP-based lifestyle interventions are readily scalable and exportable with potential for substantial clinical and public health impact.

Trial Registration

Clinicaltrials.gov identifier: NCT00842426

Introduction

An estimated 69% of U.S. adults are overweight or obese,¹ and those with modifiable cardiometabolic risk factors are a critical target group for intervention.^2,3 Lifestyle modification focused on modest (5–10%) weight loss and moderate-intensity physical activity can significantly reduce the incidence of type 2 diabetes (as much as 58% as shown in the Diabetes Prevention Program [DPP]) and cardiometabolic risk factors in high-risk individuals,^4–6 with benefits sustained for at least 10 years.⁷ Evidence-based guidelines therefore recommend effective lifestyle intervention for weight management and disease prevention.^8,9

However, national surveys reveal a continuing failure to incorporate weight management into clinical practice.¹⁰ Implementation of efficacious lifestyle interventions in the real world will require adaptation to improve generalizability and sustainability while maintaining intervention effectiveness. A meta-analysis of translation studies based on the DPP lifestyle intervention showed promising results, but most studies used a single-group design, few leveraged information technology (IT), and none had follow-up past 12 months.¹¹ Two recent trials provide further evidence on the effectiveness of alternative weight management models in primary care settings.^12,13

Evaluation of Lifestyle Interventions to Treat Elevated Cardiometabolic Risk in Primary Care (E-LITE) was a three-arm, primary-care–based randomized trial designed to evaluate the effectiveness of two adapted DPP lifestyle interventions among overweight or obese adults with prediabetes, metabolic syndrome, or both: (1) a coach-led, face-to-face group intervention and (2) a self-directed DVD intervention. We hypothesized that, compared with usual care, each intervention would result in greater mean reduction in body mass index (BMI) over 15 months.

Methods

The E-LITE protocol was approved by the Palo Alto Medical Foundation’s Institutional Review Board and was published previously.¹⁴ Some outcome data have been published in abstract form.¹⁵ All participants gave written informed consent.

Recruitment and Participants

Participants were recruited (7/2009–6/2010) from a single primary care clinic within the Silicon Valley (Los Altos, California) that is part of a large multispecialty group practice in the San Francisco Bay Area. All data collection and intervention visits occurred at the clinic. Inclusion criteria included age≥18 years, BMI≥25 kg/m², and the presence of pre-diabetes (defined by impaired fasting plasma glucose of 5.6 to 6.9 mmol/L) or metabolic syndrome (defined by 2005 joint criteria of the American Heart Association [AHA] and National Heart, Lung, and Blood Institute).¹⁶ Exclusion criteria included serious medical or psychiatric conditions (e.g., stroke, psychotic disorder) or special life circumstances (e.g., pregnancy, planned move). Of the 3,439 patients approved for study contact by their primary care provider, 1,057 were determined to be ineligible, 972 declined participation, 363 were unreachable, 806 were not screened because of recruitment success, and 241 were fully eligible and randomized (Figure 1).

Figure 1.

Screening, Randomization, and Assessments of Study Participants

Randomization and Allocation Concealment

We applied a covariate-adaptive, Efron’s biased coin method^17–20 to assure better than chance group balance across prognostic factors (age, sex, race/ethnicity, BMI, fasting blood glucose, waist circumference, and existing patient account to access personal electronic health record [EHR] system). Participants were randomized to (1) a coach-led, group-delivered intervention (n=79), (2) a self-directed DVD intervention (n=81), or (3) usual care (n=81). While study group assignment was identifiable to participants and interventionists, blinding was otherwise maintained for data collection, outcome adjudication, and data analysis.

Interventions

All participants continued to receive standard medical care. Participants’ primary care providers were not involved in the conduct of the study. The study provided no information about weight loss or weight loss goals to participants in the usual care group. Participants in both intervention groups completed a 3-month intensive intervention phase and a 12-month maintenance phase. During the intensive intervention phase, participants received an adapted, 12-session DPP lifestyle intervention curriculum, Group Lifestyle Balance (GLB)^™, that was developed by DPP investigators at the University of Pittsburgh after conclusion of the DPP trial.^21–23 The curriculum was delivered face-to-face in 12-weekly classes to coach-led intervention participants or via a home-based DVD to self-directed intervention participants. In addition to receiving GLB intervention materials, coach-led intervention participants had food tastings at check-in and 30–45 minutes of guided physical activity at the end of each weekly class. The E-LITE Lifestyle Coach, a registered dietitian certified to deliver the GLB program,²¹ and a contracted fitness instructor jointly taught all the classes at the participating clinic. We made no modifications to the GLB DVD, although self-directed intervention participants attended a single orientation class. During this class (class 1 in the coach-led intervention), participants were trained to use the AHA free Heart360 Web portal (www.heart360.org) for weight and physical activity goal setting and self-monitoring and were given a weight scale and pedometer. Via secure email embedded in the EHR and available to all intervention participants, the Lifestyle Coach sent standardized biweekly reminder messages about self-monitoring to self-directed intervention participants throughout the intensive and maintenance phases and standardized monthly motivational messages to participants in both interventions during the maintenance phase. Participants in both interventions could submit questions or concerns and received responses within 1–2 business days. Only coach-led intervention participants received personalized messages on at least a monthly basis that provided progress feedback and lifestyle coaching based on their Heart360 self-monitoring records during the maintenance phase. Table 1 shows key features of the coach-led intervention. (For more information on the interventions, see the protocol.¹⁴)

Table 1.

Features of the Coach-Led Intervention

	Intensive Intervention (Months 1–3)	Maintenance (Months 4–15)
Goal of Contact	Gradual weight loss associated with small, progressive changes in diet and physical activity and behavior change	Weight maintenance or continued gradual weight loss
Contact Schedule	Weekly (12 contacts, in the evenings)	Every 2–4 weeks depending on participant needs and preferences
Contact Mode	Group, in-person classes* (8–16 participants each; family members welcome)	Individual, secure email contacts (occasionally by phone)
Contact Duration	90–120 minutes	Variable
Contact Structure	Private weigh-in Food tasting† Group discussion of progress/barriers Review and sharing of progress Identification of personal barriers and potential solutions Presentation of the Group Lifestyle Balance (GLB) intervention session of the week‡ 30- to 45-minute guided physical activity§ Development of individual action plan/goals for the next week	Lifestyle coach review of individual participants’ weight and physical activity self-monitoring records on the Heart360 Web site Personalized progress feedback and lifestyle coaching on behavior change maintenance, problem solving, and relapse prevention via secure email within the electronic health record system

^*Classes were offered two evenings each week. Participants who missed their assigned class could attend the other one the same week. A registered dietitian working for the nutrition education department within the care delivery system and a contracted fitness instructor were hired, and they jointly taught all the classes. The dietitian received GLB certification training from the University of Pittsburgh Diabetes Prevention Support Center,²¹ including the two-day GLB Lifestyle Coach training workshop before the start of the intervention and the “train the trainer” workshop during the course of the intervention. The fitness instructor received no special training related to the study. All classes were audiotaped. However, consistent with a pragmatic approach in the context of this translational trial, we did not review the audiotapes or otherwise monitor interventionist adherence; nor did we utilize special strategies to maintain or improve adherence.

^†For food tasting, the dietitian purchased foods that did not require preparation (e.g., nuts, fruits, vegetables with reduced-fat, low-added-sugar dressings) and nonalcoholic beverages with no or low sugar content. Participants could volunteer to bring foods starting with class 7.

^‡The intervention manual for the 12 core sessions of the Group Lifestyle Balance (GLB) program can be found on the University of Pittsburgh Diabetes Prevention Support Center’s Web site.²²

^§Guided physical activity included timed walk in classes 1, 6, and 12 and circuit training with small, simple equipment (e.g., dumbbells, flexibility bands, yoga matts) provided by the study in the other sessions. Timed walks were conducted in an open area outside the clinic, weather permitting, or in a long hallway inside; the circuit activities were done in the clinic waiting area.

Outcome Measures

The primary outcome was change in BMI from baseline to 15 months. Trained research assistants who were unaware of participants’ group assignment performed anthropometric and blood pressure measurements using standard protocols¹⁴ at baseline and at 3, 6, and 15 months, except for height (measured at baseline only). At all time points except 3 months, blood samples were taken after an overnight fast. Possible adverse events were assessed by questionnaire at each follow-up visit and reviewed by a study physician per protocol.

Statistical Analysis

Between-group differences in primary and secondary outcomes were evaluated by intention-to-treat using tests of group by time interactions in repeated-measures mixed-effects linear (for continuous outcomes) or logistic models (for categorical outcomes). The fixed effects of each model consisted of the baseline value of the outcome of interest, randomization balancing factors, recruitment cohort, group, time point (3, 6, and 15 months), and group-by-time interaction. The random effects accounted for repeated measures with an unstructured covariance matrix and clustering of patients within primary care providers. Least-square means (±SE) were obtained from the models. We verified that mixed model-based results were not sensitive to violations of modeling assumptions with permutation and bootstrap resampling tests.^24,25

Of the 241 randomized participants, 205 (85.1%) had study-measured weights at 3 months, 201 (83.4%) at 6 months, and 194 (80.5%) at 15 months (see Figure 1). After replacing missing study weights with measurements obtained from the EHR (for 14 participants at 3 months, 18 at 6 months, and 24 at 15 months) and by self-report (for 5 participants at 6 months and 3 at 15 months), 22 (9.1%) participants at 3 months, 17 (7.1%) at 6 months, and 20 (8.3%) at 15 months had no weight measurement from any source. Similarly, missing blood pressure and laboratory values were replaced with EHR-recorded values. Primary analyses used all available data, but sensitivity analyses were performed that included only participants with studymeasured values. Missing data were handled directly through maximum likelihood estimation via mixed modeling.

Our primary aim was to compare change in BMI from baseline to 15 months between each intervention and the usual care control group. Our secondary aims were to 1) perform similar comparisons for secondary outcomes, 2) compare primary and secondary outcomes between the two interventions, and 3) evaluate sex as a pre-specified potential moderator. Clinical interest in weight loss outcomes by sex led to an analysis of intervention effects separately in women and men despite the absence of significant group-by-sex interaction.

The targeted sample size of 80 participants in each group was designed to provide 80% power to detect a 0.5-SD difference (medium effect by Cohen’s standards²⁶) in the primary outcome between each intervention and usual care, using t tests at 5% α (2-sided) and assuming up to a 20% loss to follow-up at 15 months. All analyses were conducted using SAS, version 9.2 (SAS Institute Inc., Cary, North Carolina).

Results

Study Participants

At baseline, participants had a mean (±SD) age of 52.9±10.6 years, a mean BMI of 32.0±5.4 kg/m² (weight 93.8±17.7 kg), 47% were female, 78% were Non-Hispanic white, 17% were Asian/Pacific Islander, and 4.1% were Hispanic/Latino. The majority of participants had high educational attainment and family annual income (Table 2). Approximately 54% of participants had pre-diabetes, 87% had metabolic syndrome, and 41% had both conditions.

Table 2.

Baseline Characteristics of the Study Participants^*

Characteristic	All Participants (n=241)	Usual Care (n=81)	Coach-Led (n=79)	Self-Directed (n=81)	P Value
Age, year	52.9 ± 10.6	52.5 ± 10.9	54.6 ± 11.0	51.8 ± 9.9	0.22
Female, %	46.5	45.7	48.1	45.7	0.94
Race/ethnicity, %					0.93
Non-Hispanic white	78.0	77.8	77.2	79.0
Asian/Pacific Islander	17.0	17.3	16.5	17.3
Latino/Hispanic	4.1	4.9	5.1	2.5
Income,%					0.12
<$75,000	12.0	11.5	14.3	10.3
$75,000–$124,999	26.6	28.2	32.5	19.2
$125,000–$149,999	13.3	6.4	15.6	17.9
$150,000+	48.1	53.9	37.7	52.6
College level or above, %	97.2	97.9	93.5	100.0	0.21
Weight, kg	93.8 ± 17.7	92.6 ± 18.1	95.3 ± 18.0	93.6 ± 17.1	0.61
Body mass index, kg/m2	32.0 ± 5.4	32.4 ± 6.3	31.8 ± 5.1	31.7 ± 4.7	0.63
Men	30.7 ± 4.5	30.8 ± 4.8	30.3 ± 3.7	31.1 ± 5.0	0.71
Women	33.4 ± 6.0	34.4 ± 7.2	33.4 ± 5.9	32.4 ± 4.3	0.34
Pre-diabetes, %	54.4	54.3	57.0	51.9	0.81
Metabolic syndrome, %	86.7	82.7	87.3	90.1	0.37
Pre-diabetes and metabolic syndrome, %	41.1	37.0	44.3	42.0	0.63
Fasting plasma glucose, mg/dL	99.9 ± 9.5	99.3 ± 9.0	100.5 ± 9.8	100.1 ± 9.7	0.71
Waist circumference, cm	106.3 ± 11.9	106.8 ± 12.7	106.2 ± 11.6	105.9 ± 11.5	0.88
Systolic blood pressure, mmHg	118.8 ± 11.7	118.4 ± 11.2	119.8 ± 12.5	118.2 ± 11.5	0.42
Diastolic blood pressure, mmHg	73.6 ± 8.3	72.5 ± 9.2	74.4 ± 8.4	73.9 ± 7.2	0.62
Triglycerides, mg/dL	171.1 ± 69.2	164.0 ± 65.8	174.8 ± 70.7	174.5 ± 71.2	0.53
High-density lipoprotein cholesterol, mg/dL	46.1 ± 12.4	46.7 ± 10.7	45.4 ± 13.3	46.2 ± 13.1	0.81
Men	41.8 ± 10.7	43.8 ± 9.2	40.1 ± 14.0	41.3 ± 8.1	0.27
Women	51.1 ± 12.4	50.1 ± 11.4	51.1 ± 9.9	52.0 ± 15.5	0.80
Low-density lipoprotein cholesterol, mg/dL	108.5 ± 31.0	108.9 ± 27.0	112.0 ± 37.7	104.6 ± 27.4	0.32
Total cholesterol, mg/dL	188.8 ± 35.4	188.4 ± 31.8	192.3 ± 42.5	185.7 ± 31.1	0.49
Triglycerides to high-density lipoprotein cholesterol ratio	4.1 ± 2.4	3.8 ± 2.2	4.4 ± 2.6	4.2 ± 2.2	0.27

^*Plus-minus values are means ± SD.

Weight Loss

At month 15, the mean (±SE) change in BMI from baseline was −2.2±0.3 kg/m² in the coach-led intervention (P<0.001 vs. usual care, P=0.03 vs. self-directed intervention), −1.6±0.3 kg/m² in the self-directed intervention (P=0.02 vs. usual care), and −0.9±0.3 kg/m² in the usual care group (Table 3). Results remained unchanged in sensitivity analyses using study-measured weights only (Supplementary Table 1).

Table 3.

Estimated Mean Change in Body Mass Index, Weight Change, Percent Weight Change over a 15-Month Period in the Intention-to-Treat Population^*

Outcome Variable	Usual Care (n=81)	Coach-led (n=79)	Self-directed (n=81)	P value
				Coach-led vs. Usual Care	Self-directed vs. Usual Care	Coach-led vs. Self-directed
BMI (kg/m2)
Baseline	32.0 ± 5.4	32.4 ± 6.3	31.8 ± 5.1	N/A	N/A	N/A
At month 3	31.5 ± 0.3	29.9 ± 0.2	30.2 ± 0.3	<0.001	<0.001	0.07
At month 6	31.5 ± 0.3	29.4 ± 0.3	30.2 ± 0.3	<0.001	<0.001	<0.001
At month 15	30.9 ± 0.3	29.6 ± 0.3	30.2 ± 0.3	<0.001†	0.02†	0.03
Change in BMI (kg/m2)
At month 3	−0.3 ± 0.3	−1.9 ± 0.3	−1.6 ± 0.3	<0.001	<0.001	0.07
At month 6	−0.3 ± 0.3	−2.4 ± 0.3	−1.5 ± 0.3	<0.001	<0.001	<0.001
At month 15	−0.9 ± 0.3	−2.2 ± 0.3	−1.6 ± 0.3	<0.001	0.02	0.03
Weight Change (kg)
At month 3	−0.7 ± 0.8	−5.4 ± 0.7	−4.5 ± 0.8	<0.001	<0.001	0.09
At month 6	−0.7 ± 0.9	−6.6 ± 0.8	−4.3 ± 0.8	<0.001	<0.001	<0.001
At month 15	−2.4 ± 0.9	−6.3 ± 0.9	−4.5 ± 0.9	<0.001	0.02	0.04
Percent Weight Change
At month 3	−0.7 ± 0.8	−5.8 ± 0.8	−4.9 ± 0.8	<0.001	<0.001	0.09
At month 6	−0.9 ± 0.9	−7.2 ± 0.9	−4.7 ± 0.9	<0.001	<0.001	0.001
At month 15	−2.6 ± 0.9	−6.6 ± 0.9	−5.0 ± 0.9	<0.001	0.008	0.07

^*Plus-minus values are means ± SE. The data for the three study groups are covariate-adjusted, mixed model-based estimates for the intention-to-treat population.

^†The primary comparisons between the two active interventions and usual care remain statistically significant (P<0.05) after the Bonferroni correction was applied by multiplying the calculated P values by a factor of 2.

At month 15, the mean change in weight from baseline was −6.3±0.9 kg in the coach-led intervention, −4.5 ± 0.9 kg in the self-directed intervention, −2.4 ± 0.9 kg in the usual care control group, corresponding to a weight change of −6.6%, −5.0%, and −2.6%, respectively (Table 3 and Figure 2). The percentage of participants who achieved the 7% DPP-based weight loss goal at 15 months was 37.0% (P=0.003) in the coach-led intervention and 35.9% (P=0.004) in the self-directed intervention versus 14.4% in the usual care group. Findings were similar for 5% and 10% weight loss goal cut-points (Figure 3). Complete fitted distributions of percent weight changes at 15 months are shown in the Supplementary Figure.

Figure 2.

Estimated Mean (±SE) Weight Change over a 15-Month Period in the Intention-to-Treat Population, Overall and by Sex

Figure 3.

Categorical Weight Loss at 6 and 15 Months in the Intention-to-Treat Population

During the trial period, 15 of 81 participants in the usual care group reported joining a weight loss program outside the study (12 used commercial programs, 2 used nutrition classes offered by the care delivery system, and 1 used a personal trainer), compared to 5 of 79 in the coach-led group (4 used personal trainers and 1 used a commercial program) and 3 of 81 in the self-directed group (2 used personal trainers and 1 used a commercial program; P=0.003). No participants reported undergoing pharmacological or surgical weight loss treatment.

For women, weight loss was significantly greater in the coach-led intervention than in the usual care control group (P=0.003) and the self-directed intervention (P=0.02) at 15 months, whereas the difference between the latter two groups was not statistically significant (Figure 2). For men, both the coach-led (P=0.002) and self-directed (P=0.007) interventions resulted in significantly greater weight loss than did usual care at 15 months, while the two interventions did not differ significantly. The difference by group between men and women was not statistically significant (P=0.31).

Estimates of number needed to treat (NNT) and area under the receiver operating characteristic curve (AUC)^27,28 for all participants and for women and men separately are shown in Supplementary Table 2.

Changes in Cardiometabolic Risk Factors

Compared with usual care controls, improvements reached statistical significance for waist circumference and fasting plasma glucose levels in both interventions and for diastolic blood pressure and TG/HDL in the coach-led intervention (Table 4). Total cholesterol levels increased significantly less in the self-directed intervention vs. the usual care group.

Table 4.

Estimated Mean Changes from Baseline to 15 Months in Cardiometabolic Risk Factors in the Intention-to-Treat Population^*

Outcome Variable	Usual care	Coach-led	Self-directed	P value
				Coach-led vs. usual care	Self-direct vs. usual care	Coach-led vs. Self-direct
Waist circumference, cm (n=218)	−2.2 ± 1.1	−5.8 ± 1.0	−4.9 ± 1.0	<0.001	<0.001	0.20
Systolic blood pressure, mmHg (n=237)	0.1 ± 1.6	−1.2 ± 1.5	−0.4 ± 1.5	0.21	0.61	0.45
Diastolic blood pressure, mmHg (n=237)	−0.3 ± 1.1	−1.9 ± 1.1	−1.1 ± 1.1	0.04	0.29	0.30
Fasting plasma glucose, mg/dL (n=209)	0.2 ± 1.7	−4.2 ± 1.6	−2.7 ± 1.6	<0.001	0.01	0.20
Triglycerides, mg/dL (n=208)	−18.8 ± 11.0	−31.2 ± 10.5	−28.8 ± 10.8	0.11	0.18	0.75
High-density lipoprotein cholesterol, mg/dL (n=218)	2.9 ± 1.4	4.4 ± 1.3	2.6 ± 1.3	0.11	0.79	0.06
Low-density lipoprotein cholesterol, mg/dL (n=217)	10.6 ± 5.0	4.5 ± 4.8	5.2 ± 4.9	0.08	0.11	0.85
Total cholesterol, mg/dL (n=218)	10.6 ± 5.5	3.9 ± 5.6	4.4 ± 5.6	0.05	0.04	0.98
Triglycerides to highdensity lipoprotein cholesterol ratio (n=218)	−0.5 ± 0.3	−1.0 ± 0.3	−0.8 ± 0.3	0.03	0.18	0.38

^*Plus-minus values are means ± SE. The data for the three study groups are covariate-adjusted, mixed model-based estimates for the intention-to-treat population.

Intervention Participation Rates

Participants in the coach-led intervention attended 75.1±25.6% (74.6±26.3% among men, 75.7±25.2% among women) of the 12 weekly group sessions (median number of sessions attended, 10; interquartile range [IQR], 9 to 11). Only 4 participants (1 man, 3 women) in the self-directed intervention did not attend the single group orientation session. Self-directed intervention participants had a median number of 31 secure email messages (IQR, 30 to 32) during the 15-month period, and coach-led intervention participants had 19 (IQR, 18 to 22) during the 12-month period after weekly classes were over.

Adverse Events

Five serious adverse events were detected in four coach-led intervention participants that may have been related to the intervention: three fractures and one case of chronic subdural hematoma requiring surgery several months following the participant’s syncopal episode during a group intervention session. Six other hospitalizations were reported, which were judged to be unrelated to the study (2 in the usual care group, 1 in the self-directed group, 3 in the coach-led group). One coach-led intervention participant and one usual care participant developed type 2 diabetes during the 15-month period. There were no deaths.

Discussion

This primary-care–based translational intervention trial demonstrated that two IT-supported, DPP-based lifestyle interventions both led to clinically significant reductions in body weight (measured as change in BMI), accompanied by improvements in waist circumference and fasting plasma glucose, as compared with usual care over a 15-month period.

Successful adaptation of proven lifestyle interventions such as the DPP for multiple channels of delivery, all populations at risk, and primary care settings will be critical to stem the tide of obesity and lessen its disease burden.²⁹ Until recently, rigorous trial evidence on effective, scalable treatment models in primary care practice has been lacking.³⁰ Newly published trials demonstrated the effectiveness of two primary care models, one involving in-person or remote (primarily by phone) professional weight management support¹² and the other combining lifestyle counseling with meal replacement or weight-loss medication.¹³ The E-LITE trial makes a unique contribution to this growing literature in that its interventions integrate standardized, packaged DPP translational programs (delivered in groups or by DVD) with existing health IT. Although these intervention components and delivery channels are not new, their integration into structured interventions for use in primary care is novel.

The maximum weight loss achieved within the coach-led intervention was substantial (6.3 kg, net loss of 3.9 kg relative to usual care) and similar in magnitude to that achieved by the DPP lifestyle intervention and other behavioral or drug-based weight loss trials.^6,12,13,31 Weight loss in the self-directed intervention was less pronounced (net loss of 2.1 kg) but noteworthy given its low resource requirements and high potential for dissemination. In this real-world translation study, we did not restrict participants from seeking other weight loss treatment. Nevertheless, the net intervention effects were significant even though a higher proportion of usual care participants reported attending outside weight loss programs during the study period. The fact that usual care participants lost some weight emphasizes the robustness of findings regarding the effectiveness of the interventions.

Women appeared to respond more favorably to the coach-led intervention compared to self-directed intervention, whereas men appeared to respond comparably to both. These sex-specific findings need to be confirmed in studies adequately powered to investigate sex differences. Future research also should investigate whether empirically-supported, gender-based intervention targeting strategies can improve the effectiveness of the interventions.³²

The present trial has several limitations. First, study participants were primarily of high socioeconomic status and from a single primary care clinic located within the Silicon Valley of the San Francisco Bay Area and within a parent health system that was one of the first in the nation to adopt a fully-functional EHR system. Therefore, its findings may not be directly generalizable to other populations and settings. Second, replacement of missing weights with clinical values recorded in the EHR or self-reported weights might have introduced bias, but sensitivity analyses using only study-measured weights did not change the results. Finally, the trial lasted only for 15 months, and was not designed to evaluate event-based outcomes (e.g., type 2 diabetes incidence) or cost effectiveness. Thus, the long-term effects and comparative cost-effectiveness of the two interventions await further investigation.

Independent efforts to broadly disseminate the DPP-based GLB in-person and DVD programs have been underway.^21,23 In E-LITE, these programs were integrated with health IT tools that have low-additive cost but high reach, in order to maximize translation potential and clinical and public health impact. Technology-based interventions need to be first evaluated where the technology is available, with broad dissemination following as adoption of the technology expands. Use of computers and the Internet has increased markedly in all population segments, including in lower socioeconomic groups.³³ The AHA’s Heart360 self-management Web portal is free, trustworthy, and easily accessible for patients and providers. Heart360-automated mobile texting for reminders and data transmission (without requiring Web logon after account set-up) is an enhancement that became available only after initiation of intervention with all participants in our study. It substantially extend the potential reach of the system.³⁴ Moreover, healthcare reform provisions have accelerated adoption of EHR systems.³⁵ Although the cost of acquiring an EHR system is substantial, use of a system already in place for disease management requires minimal additional investment (e.g., low time commitment by the lifestyle coach to communicate with participants via secure email).

The E-LITE interventions respond to the need for innovative, effective methods to manage obesity in primary care settings that do not overly burden practicing providers. These interventions are demonstrably beneficial, with potential for high clinical and public health impact, but they do not meet the current definitions of provider (primary care clinicians only) and delivery channel (face-to-face visits only) required to receive Centers for Medicare and Medicaid Services coverage of intensive behavioral therapy for obesity in primary care.³⁶ The Centers for Disease Control and Prevention also requires that lifestyle interventions be delivered in person as one of the standards for recognition in its National Diabetes Prevention Program.³⁷ Consideration of expanded program criteria in these national initiatives might encourage wider adoption of alternative and, potentially, more cost-effective lifestyle interventions such as the ones evaluated in this study, assuming that their effectiveness can be documented in more nationally representative populations.