RESEARCH AND PRACTICE
A Randomized Comparative Effectiveness Trial for Preventing Type 2 Diabetes Ronald T. Ackermann, MD, MPH, David T. Liss, PhD, Emily A. Finch, MA, Karen K. Schmidt, RN, MSN, Laura M. Hays, PhD, David G. Marrero, PhD, and Chandan Saha, PhD
More than 150 million Americans are overweight or obese and at risk for weight-related complications such as heart disease, stroke, high blood pressure, abnormal blood cholesterol, and diabetes.1,2 To the extent that obesity is a preventable or reversible health condition, many of these complications and their associated costs could be avoided. The US Diabetes Prevention Program (DPP) showed that resource-intensive lifestyle interventions supporting daily physical activity and modest weight loss can cut the rate of developing type 2 diabetes in half.3 Unfortunately, modest weight loss is challenging to achieve and maintain,4 particularly for younger, more obese, minority, and economically disadvantaged persons.5 Moreover, the cost of offering intensive programs imposes barriers for real-world implementation.6,7 Although the DPP and other trials have demonstrated the safety and efficacy of lifestyle interventions for diabetes prevention, a recent systematic review of studies evaluating the effectiveness of offering DPP-like interventions to broader clinical or community populations identified very few high-quality or even controlled studies8; none of these previous studies was large enough to evaluate effectiveness within minority subgroups. To address this research gap, the Reaching Out to Prevent Increases in Diabetes (RAPID) study used a randomized effectiveness trial design to evaluate a collaborative approach that links health system efforts for identifying prediabetes with community delivery of a lower-cost, group-based lifestyle intervention offered by the YMCA. This trial evaluated whether the YMCA model for translating the DPP intervention (YDPP) can reach large numbers of diverse, primarily low-income adults with prediabetes and, when compared with the existing standard of care, is more effective in achieving modest weight losses shown previously to lower the development of type 2 diabetes.9
2328 | Diabetes | Peer Reviewed | Ackermann et al.
Objectives. We evaluated the weight loss effectiveness of a YMCA model for the Diabetes Prevention Program (DPP) lifestyle intervention. Methods. Between July 2008 and November 2010, we individually randomized 509 overweight or obese, low-income, nondiabetic adults with elevated blood glucose in Indianapolis, Indiana, to receive standard care plus brief lifestyle counseling or be offered a group-based YMCA adaptation of the DPP (YDPP). Primary outcome was mean weight loss difference at 12 months. In our intention-to-treat analyses, we used longitudinal linear or logistic regression, multiply imputing missing observations. We used instrumental variables regression to estimate weight loss effectiveness among participants completing 9 or more intervention lessons. Results. In the YDPP arm, 161 (62.6%) participants attended ‡ 1 lesson and 103 (40.0%) completed 9 or more lessons. In intention-to-treat analysis, mean 12-month weight loss was 2.3 kilograms (95% confidence interval [CI] = 1.1, 3.4 kg) more for the YDPP arm than for standard care participants. In instrumental variable analyses, persons attending 9 or more lessons had a 5.3-kilogram (95% CI = 2.8, 7.9 kg) greater weight loss than did those with standard care alone. Conclusions. The YMCA model for DPP delivery achieves meaningful weight loss at 12 months among low-income adults. (Am J Public Health. 2015;105: 2328–2334. doi:10.2105/AJPH.2015.302641)
METHODS RAPID was a 2-arm comparative effectiveness trial that individually randomized each participant (1:1) to either (1) encouragement to take part in a group-based adaptation of the DPP lifestyle intervention offered free of charge (YDPP arm) or (2) receipt of usual clinical care plus brief counseling and information about existing community resources for lifestyle modification (standard care arm).
Participants Participants were aged 18 years or older and had a body mass index (BMI; defined as weight in kilograms divided by the square of height in meters) of 24 or greater, no prior diagnosis of diabetes, and at least 1 blood test indicating high risk for type 2 diabetes (fasting plasma glucose of 100---125 mg/dL; 2-hour postload plasma glucose of 140---199 mg/dL; or HbA1c [glycated hemoglobin] of 5.7%---6.9%). We excluded patients if they were unable to
provide informed consent, were unable to read English, were pregnant or planning pregnancy, were actively taking a medication known to alter glucose metabolism (e.g., oral steroids or select antipsychotic medications), had blood pressure of 180/105 millimeters of mercury or greater, or reported a comorbidity expected to limit life span to less than 3 years. At each of 9 urban primary care clinics in Indianapolis, Indiana, clinical data managers used electronic databases to identify patients who met inclusion criteria and had completed a blood test indicating prediabetes in the preceding 12 months.10 The data manager delivered the eligible patient list to primary care providers in each clinic to secure permission to approach their patients. A research assistant then telephoned and prescreened each patient for the study. All interested patients were offered a face-to-face visit with the research assistant at the clinic and were told they would receive additional information about ways to prevent type 2 diabetes. We enrolled all
American Journal of Public Health | November 2015, Vol 105, No. 11
RESEARCH AND PRACTICE
patients who attended this visit, were confirmed as eligible, and provided written informed consent. At enrollment, participants in both groups received information and encouragement to use local community resources and self-help diabetes prevention materials from the National Diabetes Education Program.11 This information was refreshed at each study visit. Research assistants also encouraged and assisted participants in both arms to complete a visit with a registered dietitian at the clinic to develop an action plan for dietary changes and weight loss. In addition to routine clinical care, these intervention components, common to both study arms, reflected published recommendations for adults with prediabetes and were considered more intensive than are levels of support currently being offered in most primary care settings. Participants randomized to the standard care arm were offered only these intervention components. This choice enabled us to evaluate the comparative effectiveness of standard care with that of standard care offered in conjunction with free-of-charge access to YDPP. Within 24 hours of enrollment, a study team member gave the YMCA intervention coordinator the name and contact information for each new participant randomized to the YDPP arm. The YMCA coordinator telephoned and offered each participant the opportunity to take part in the YDPP lifestyle intervention free of charge. Active participation in YDPP was encouraged but not required. Interested participants were assembled into groups of 8 to 12 persons who could meet at a mutually agreeable time and place, including both YMCA and non-YMCA locations. The group-based adaptation of the YDPP involved goal setting, self-monitoring, and participant-centered problem solving to achieve modest weight loss (5%---7% reduction from baseline) through a combination of moderate physical activity (150 minutes/week, equivalent to walking) and lower dietary fat and calorie consumption.12 The intervention began with 16 face-to-face, small-group lessons, each lasting 60 to 90 minutes and delivered over 16 to 24 weeks,13 followed by monthly support meetings, each lasting about 60 minutes, for the duration of the trial.14,15 Participants were also offered tools such as a step
counter, measuring cups, food scales, fat and calorie tracking tools, and recipe guides. YMCA leadership selected and hired YDPP instructors, who then completed an initial 2-day training program as well as annual refresher training that ensured core intervention competencies.16 The YMCA intervention coordinator supervised all instructors and monitored program fidelity and quality assurance.10,16 Research assistants collected data at baseline, 6 months, and 12 months via written survey and a short battery of biometric assessments and finger stick blood tests. The primary outcome was change in body weight over 12 months. Secondary outcomes included the percentage of participants who reached weight loss goals of 5% or greater as well as changes in blood pressure, total and high-density lipoprotein (HDL)-cholesterol, and glycated hemoglobin (A1c). We assessed blood pressures as the mean of 2 values taken at rest and in the seated position. We assessed A1c from a finger stick capillary blood sample using a DCA 2000/ Vantage portable bench-top analyzer (Siemens, Bellport, NY).17,18 We measured nonfasting cholesterol concentrations from capillary blood using an LDX lipid analyzer (Alere, Waltham, MA).19,20 We used the survey data collected at each visit to assess potential unanticipated harms and secondary clinical outcomes such as cardiovascular events. We computer generated randomization lists separately for each clinic site using 1:1 allocation with blocks of 2. We blinded intervention assignment to research staff using individually sealed opaque envelopes. We did not inform participants that the study involved randomization, but we did inform them that they would receive repeat diabetes risk testing and other risk factor assessments, brief lifestyle counseling, and ongoing updates about communitybased lifestyle intervention resources. Thus, although participants could not be blinded to the specific resources they were offered, they remained unaware that their resources might differ from those of other study participants.
Statistical Analysis The randomized encouragement trial design enabled us to observe YDPP participation levels when offered in a real-world context and to conduct 2 primary statistical analyses: (1) an assessment of overall, intent-to-treat (ITT)
November 2015, Vol 105, No. 11 | American Journal of Public Health
effects, regardless of YDPP participation; and (2) valid estimation of the average treatment effect among compliers. In ITT analysis, we analyzed repeated outcome measures using longitudinal linear regression (continuous outcomes) or logistic regression (dichotomous 5% weight loss), treating the participant as a random effect and using robust variance estimates. Regression models included 3 observations per participant (at baseline, 6 months, and 12 months). We estimated multivariable regression models using multiple imputation to account for loss to follow-up for some trial participants. Weight data were missing for 69 (13.6%) participants at 6 months and 79 (15.5%) participants at 12 months (P ‡ .31 for v2 test comparing proportions of missingness between intervention arms at 6 months and 12 months). Similar proportions of follow-up data were missing for secondary outcomes. We estimated missing weight observations using the predictive mean matching imputation method. We drew imputed values from the 5 nearest neighbor observations within the same study arm that had the most similar covariate distributions,21,22 including gender, annual household income, race/ ethnicity, age, intervention session attendance, baseline weight, study month, and clinic. We estimated differential intervention effects at 6 months through a group-by-time interaction term. The primary independent variable was the 12-month effect of random treatment assignment (YDPP or standard care). After model estimation, we predicted adjusted weight change and adjusted probability of 5% weight loss using the “mi predict” command in Stata version 12.1 (StataCorp LP, College Station, TX). In preplanned ITT subgroup analyses we assessed the adjusted, predicted weight changes among both African Americans and non-Hispanic Whites. In exploratory ITT analysis, we tested differences in weight changes between these 2 subgroups by introducing a race-by-treatment interaction term. Because of the encouragement trial design, we anticipated that a sizeable proportion of participants assigned to YDPP would not actually participate. Recent studies of referral from primary care to a free-of-charge community-based lifestyle intervention resource show that realistic estimates of uptake may be only about 30% for any participation
Ackermann et al. | Peer Reviewed | Diabetes | 2329
RESEARCH AND PRACTICE
Identified as High Risk by Primary Care Glucose Tests Over 29 Months n = 5201
Approval by PCP to Approach
Unable to obtain approval by PCP (n = 2137)
n = 3064
Successful Telephone Pre-Screen
Unreachable by phone (n = 927)
n = 2137
Attended Screening Visit
Not Interested (n = 779) Exclusion based on medication, comorbidity, or limited English proficiency (n = 718)
n = 640
Enrolled in RAPID Study
Not interested (n = 23) BMI
A Randomized Comparative Effectiveness Trial for Preventing Type 2 Diabetes Ronald T. Ackermann, MD, MPH, David T. Liss, PhD, Emily A. Finch, MA, Karen K. Schmidt, RN, MSN, Laura M. Hays, PhD, David G. Marrero, PhD, and Chandan Saha, PhD
More than 150 million Americans are overweight or obese and at risk for weight-related complications such as heart disease, stroke, high blood pressure, abnormal blood cholesterol, and diabetes.1,2 To the extent that obesity is a preventable or reversible health condition, many of these complications and their associated costs could be avoided. The US Diabetes Prevention Program (DPP) showed that resource-intensive lifestyle interventions supporting daily physical activity and modest weight loss can cut the rate of developing type 2 diabetes in half.3 Unfortunately, modest weight loss is challenging to achieve and maintain,4 particularly for younger, more obese, minority, and economically disadvantaged persons.5 Moreover, the cost of offering intensive programs imposes barriers for real-world implementation.6,7 Although the DPP and other trials have demonstrated the safety and efficacy of lifestyle interventions for diabetes prevention, a recent systematic review of studies evaluating the effectiveness of offering DPP-like interventions to broader clinical or community populations identified very few high-quality or even controlled studies8; none of these previous studies was large enough to evaluate effectiveness within minority subgroups. To address this research gap, the Reaching Out to Prevent Increases in Diabetes (RAPID) study used a randomized effectiveness trial design to evaluate a collaborative approach that links health system efforts for identifying prediabetes with community delivery of a lower-cost, group-based lifestyle intervention offered by the YMCA. This trial evaluated whether the YMCA model for translating the DPP intervention (YDPP) can reach large numbers of diverse, primarily low-income adults with prediabetes and, when compared with the existing standard of care, is more effective in achieving modest weight losses shown previously to lower the development of type 2 diabetes.9
2328 | Diabetes | Peer Reviewed | Ackermann et al.
Objectives. We evaluated the weight loss effectiveness of a YMCA model for the Diabetes Prevention Program (DPP) lifestyle intervention. Methods. Between July 2008 and November 2010, we individually randomized 509 overweight or obese, low-income, nondiabetic adults with elevated blood glucose in Indianapolis, Indiana, to receive standard care plus brief lifestyle counseling or be offered a group-based YMCA adaptation of the DPP (YDPP). Primary outcome was mean weight loss difference at 12 months. In our intention-to-treat analyses, we used longitudinal linear or logistic regression, multiply imputing missing observations. We used instrumental variables regression to estimate weight loss effectiveness among participants completing 9 or more intervention lessons. Results. In the YDPP arm, 161 (62.6%) participants attended ‡ 1 lesson and 103 (40.0%) completed 9 or more lessons. In intention-to-treat analysis, mean 12-month weight loss was 2.3 kilograms (95% confidence interval [CI] = 1.1, 3.4 kg) more for the YDPP arm than for standard care participants. In instrumental variable analyses, persons attending 9 or more lessons had a 5.3-kilogram (95% CI = 2.8, 7.9 kg) greater weight loss than did those with standard care alone. Conclusions. The YMCA model for DPP delivery achieves meaningful weight loss at 12 months among low-income adults. (Am J Public Health. 2015;105: 2328–2334. doi:10.2105/AJPH.2015.302641)
METHODS RAPID was a 2-arm comparative effectiveness trial that individually randomized each participant (1:1) to either (1) encouragement to take part in a group-based adaptation of the DPP lifestyle intervention offered free of charge (YDPP arm) or (2) receipt of usual clinical care plus brief counseling and information about existing community resources for lifestyle modification (standard care arm).
Participants Participants were aged 18 years or older and had a body mass index (BMI; defined as weight in kilograms divided by the square of height in meters) of 24 or greater, no prior diagnosis of diabetes, and at least 1 blood test indicating high risk for type 2 diabetes (fasting plasma glucose of 100---125 mg/dL; 2-hour postload plasma glucose of 140---199 mg/dL; or HbA1c [glycated hemoglobin] of 5.7%---6.9%). We excluded patients if they were unable to
provide informed consent, were unable to read English, were pregnant or planning pregnancy, were actively taking a medication known to alter glucose metabolism (e.g., oral steroids or select antipsychotic medications), had blood pressure of 180/105 millimeters of mercury or greater, or reported a comorbidity expected to limit life span to less than 3 years. At each of 9 urban primary care clinics in Indianapolis, Indiana, clinical data managers used electronic databases to identify patients who met inclusion criteria and had completed a blood test indicating prediabetes in the preceding 12 months.10 The data manager delivered the eligible patient list to primary care providers in each clinic to secure permission to approach their patients. A research assistant then telephoned and prescreened each patient for the study. All interested patients were offered a face-to-face visit with the research assistant at the clinic and were told they would receive additional information about ways to prevent type 2 diabetes. We enrolled all
American Journal of Public Health | November 2015, Vol 105, No. 11
RESEARCH AND PRACTICE
patients who attended this visit, were confirmed as eligible, and provided written informed consent. At enrollment, participants in both groups received information and encouragement to use local community resources and self-help diabetes prevention materials from the National Diabetes Education Program.11 This information was refreshed at each study visit. Research assistants also encouraged and assisted participants in both arms to complete a visit with a registered dietitian at the clinic to develop an action plan for dietary changes and weight loss. In addition to routine clinical care, these intervention components, common to both study arms, reflected published recommendations for adults with prediabetes and were considered more intensive than are levels of support currently being offered in most primary care settings. Participants randomized to the standard care arm were offered only these intervention components. This choice enabled us to evaluate the comparative effectiveness of standard care with that of standard care offered in conjunction with free-of-charge access to YDPP. Within 24 hours of enrollment, a study team member gave the YMCA intervention coordinator the name and contact information for each new participant randomized to the YDPP arm. The YMCA coordinator telephoned and offered each participant the opportunity to take part in the YDPP lifestyle intervention free of charge. Active participation in YDPP was encouraged but not required. Interested participants were assembled into groups of 8 to 12 persons who could meet at a mutually agreeable time and place, including both YMCA and non-YMCA locations. The group-based adaptation of the YDPP involved goal setting, self-monitoring, and participant-centered problem solving to achieve modest weight loss (5%---7% reduction from baseline) through a combination of moderate physical activity (150 minutes/week, equivalent to walking) and lower dietary fat and calorie consumption.12 The intervention began with 16 face-to-face, small-group lessons, each lasting 60 to 90 minutes and delivered over 16 to 24 weeks,13 followed by monthly support meetings, each lasting about 60 minutes, for the duration of the trial.14,15 Participants were also offered tools such as a step
counter, measuring cups, food scales, fat and calorie tracking tools, and recipe guides. YMCA leadership selected and hired YDPP instructors, who then completed an initial 2-day training program as well as annual refresher training that ensured core intervention competencies.16 The YMCA intervention coordinator supervised all instructors and monitored program fidelity and quality assurance.10,16 Research assistants collected data at baseline, 6 months, and 12 months via written survey and a short battery of biometric assessments and finger stick blood tests. The primary outcome was change in body weight over 12 months. Secondary outcomes included the percentage of participants who reached weight loss goals of 5% or greater as well as changes in blood pressure, total and high-density lipoprotein (HDL)-cholesterol, and glycated hemoglobin (A1c). We assessed blood pressures as the mean of 2 values taken at rest and in the seated position. We assessed A1c from a finger stick capillary blood sample using a DCA 2000/ Vantage portable bench-top analyzer (Siemens, Bellport, NY).17,18 We measured nonfasting cholesterol concentrations from capillary blood using an LDX lipid analyzer (Alere, Waltham, MA).19,20 We used the survey data collected at each visit to assess potential unanticipated harms and secondary clinical outcomes such as cardiovascular events. We computer generated randomization lists separately for each clinic site using 1:1 allocation with blocks of 2. We blinded intervention assignment to research staff using individually sealed opaque envelopes. We did not inform participants that the study involved randomization, but we did inform them that they would receive repeat diabetes risk testing and other risk factor assessments, brief lifestyle counseling, and ongoing updates about communitybased lifestyle intervention resources. Thus, although participants could not be blinded to the specific resources they were offered, they remained unaware that their resources might differ from those of other study participants.
Statistical Analysis The randomized encouragement trial design enabled us to observe YDPP participation levels when offered in a real-world context and to conduct 2 primary statistical analyses: (1) an assessment of overall, intent-to-treat (ITT)
November 2015, Vol 105, No. 11 | American Journal of Public Health
effects, regardless of YDPP participation; and (2) valid estimation of the average treatment effect among compliers. In ITT analysis, we analyzed repeated outcome measures using longitudinal linear regression (continuous outcomes) or logistic regression (dichotomous 5% weight loss), treating the participant as a random effect and using robust variance estimates. Regression models included 3 observations per participant (at baseline, 6 months, and 12 months). We estimated multivariable regression models using multiple imputation to account for loss to follow-up for some trial participants. Weight data were missing for 69 (13.6%) participants at 6 months and 79 (15.5%) participants at 12 months (P ‡ .31 for v2 test comparing proportions of missingness between intervention arms at 6 months and 12 months). Similar proportions of follow-up data were missing for secondary outcomes. We estimated missing weight observations using the predictive mean matching imputation method. We drew imputed values from the 5 nearest neighbor observations within the same study arm that had the most similar covariate distributions,21,22 including gender, annual household income, race/ ethnicity, age, intervention session attendance, baseline weight, study month, and clinic. We estimated differential intervention effects at 6 months through a group-by-time interaction term. The primary independent variable was the 12-month effect of random treatment assignment (YDPP or standard care). After model estimation, we predicted adjusted weight change and adjusted probability of 5% weight loss using the “mi predict” command in Stata version 12.1 (StataCorp LP, College Station, TX). In preplanned ITT subgroup analyses we assessed the adjusted, predicted weight changes among both African Americans and non-Hispanic Whites. In exploratory ITT analysis, we tested differences in weight changes between these 2 subgroups by introducing a race-by-treatment interaction term. Because of the encouragement trial design, we anticipated that a sizeable proportion of participants assigned to YDPP would not actually participate. Recent studies of referral from primary care to a free-of-charge community-based lifestyle intervention resource show that realistic estimates of uptake may be only about 30% for any participation
Ackermann et al. | Peer Reviewed | Diabetes | 2329
RESEARCH AND PRACTICE
Identified as High Risk by Primary Care Glucose Tests Over 29 Months n = 5201
Approval by PCP to Approach
Unable to obtain approval by PCP (n = 2137)
n = 3064
Successful Telephone Pre-Screen
Unreachable by phone (n = 927)
n = 2137
Attended Screening Visit
Not Interested (n = 779) Exclusion based on medication, comorbidity, or limited English proficiency (n = 718)
n = 640
Enrolled in RAPID Study
Not interested (n = 23) BMI
