A Randomized Trial to Compare Surgical and Medical Treatments for Type 2 Diabetes: The Triabetes Study

Abstract

IMPORTANCE

Address unanswered questions about the role of bariatric surgery for people with diabetes.

OBJECTIVE

Determine feasibility of a randomized controlled trial (RCT) and compare initial outcomes of bariatric surgery and a structured weight loss program for treating type 2 diabetes mellitus in grade 1 and 2 obese participants.

DESIGN, SETTING, PARTICIPANTS

A 12-month, 3-arm RCT at a single center including 69 participants age 25–55 years, BMI 30–40 with type 2 diabetes.

INTERVENTIONS

Two surgical procedures; Roux-en-Y gastric bypass (RYGB), laparoscopic adjustable gastric band (LAGB) and an intensive lifestyle weight loss intervention (LWLI).

MAIN OUTCOMES AND MEASURES

Primary outcomes in the intention to treat (ITT) cohort were feasibility and effectiveness measured by weight loss and improvements in glycemic control.

RESULTS

667 potential participants were screened of whom 69 (10.3%) were randomized, 30 (43.5%) with grade 1 obesity. Mean age was 47.3±6.4 years, 81% were women, and mean glycated hemoglobin was 7.9±2.0. After randomization, 7 (10%) participants refused to undergo their allocated intervention (3 RYGB, 1 LAGB, 3 LWLI) and 1 RYGB was excluded for current smoking. Twenty subjects underwent RYGB, 21 LAGB, and 20 LWLI with retention at 12 months of 90%, 86%, and 70%, respectively. In the ITT cohort with multiple imputation for missing data, RYGB participants had the greatest weight loss compared to LAGB and LWLI with average weight loss of 27%, 17%, 10% from baseline, respectively (p<.0001). Partial/complete remission of diabetes was 50%/17% in RYGB, 27%/23% in LAGB and 0%/0% in LWLI (p=.0005/.047, partial/complete) and there were significant reductions in medication usage in both surgical groups. There were no deaths and 3 serious adverse events; 1 RYGB ulcer was treated medically and 2 LAGB were re-hospitalized for dehydration.

CONCLUSIONS

This study highlights several potential challenges to successfully completing a larger RCT for diabetes and obesity treatment in those with BMI 30–40 kg/m², including the difficulties associated with recruiting and randomizing patients to surgical versus non-surgical interventions. Preliminary results show that RYGB was the most effective treatment followed by LAGB for both weight loss and diabetes outcomes at one year.

INTRODUCTION

Despite much interest and a growing body of literature there is a lack of enough level I evidence to support the use of bariatric surgery to treat type 2 diabetes (T2DM).^1–4 Bariatric surgery is very effective at inducing weight reduction and controlling obesity-related type 2 diabetes (T2DM) in the setting of grade 2 or 3 obesity.^2,3,5–9 However, there is insufficient evidence regarding its longer term outcomes resulting in less than uniform acceptance of these procedures as a primary diabetes treatment. This is especially true for T2DM in the setting of grade 1 obesity for which there is very limited evidence addressing even short-term surgical safety and outcomes.¹⁰

There are several key unanswered questions about the role of bariatric surgery in the treatment of T2DM that include the relative safety and efficacy of treatment options (surgical versus non-surgical), the modification of the future risk of micro and macro vascular complications, and the economic impact of these treatments.⁴ Answers to such questions could come from a new large, multi-center randomized clinical trial to compare bariatric surgery to the best available medical care, but such a trial would be costly in a difficult funding climate, time consuming when answers of comparative effectiveness are urgently needed, and potentially difficult to execute as two recent studies have shown.^1,3 The NIH funded seven prospective pilot and feasibility studies several years ago, possibly to consider a large, multi-center study to address these questions.^11–18

This is one of the first of these studies to be completed and addresses important feasibility and preliminary effectiveness aims. We report the results of a randomized, controlled clinical trial (RCT) examining the feasibility of a larger study and comparing the effectiveness of the two predominant types of bariatric surgery (RYGB, LAGB) and an intensive lifestyle intervention modeled after Look AHEAD¹⁹ in adults with grade 1 and 2 obesity and T2DM.

METHODS

STUDY DESIGN

This prospective, randomized clinical trial (RCT) was conducted at an academic medical center between October 2009 and April 2012. The local institutional review board approved the study protocol and informed consent was obtained. Participants were recruited by a variety of advertisement techniques (television, print-newspapers, internet-multiple, and other local ads) followed by an initial telephone screening and then an in-person, group informational session. Each participant underwent individual evaluation by the surgeon and after successful medical, nutritional and psychological assessments were completed, randomly assigned to one of three treatment arms (RYGB, LAGB, or LWLI). Stratified (gender, BMI) blocked randomization with equal allocation to each arm was performed using computer-generated random-numbers. The cost for the surgical procedures was subsidized by the medical center.

SUBJECTS

Adults were eligible for enrollment if they were between the ages of 25 and 55 years of age and had a BMI between 30 and 40 kg/m², as this is a high priority subgroup for comparative effectiveness studies.¹¹ T2DM was confirmed by either a documented fasting plasma glucose (FPG) ≥126 mg/dL and/or treatment with an anti-diabetic medication, to include a broad spectrum of T2DM severity. For those with grade 1 obesity, treatment with an anti-diabetic medication and permission from their treating physician were required to participate. Adults were excluded for prior weight loss surgery, impaired mental status, drug / alcohol addiction, current smoking, pregnancy or planned pregnancy, inability to tolerate general anesthesia due to poor health, Type I diabetes mellitus, failed nutrition or psychological assessment, unwillingness to be randomized, inability to provide informed consent, or if they were deemed unlikely to comply with study visits or procedures.

TREATMENTS

All surgical procedures were performed by the first author (APC). The RYGB was performed with a standard retrocolic, retrogastric technique using a linear stapled and hand sewn gastrojejunal anastomosis. The LAGB was performed using the Allergan 10 or the AP Standard LapBand^® with suture securing of the gastric cardia to prevent slippage and placement of the infusion port on the anterior rectus muscle. Surgical participants underwent clinical follow-up assessments consistent with current practice: RYGB at 2 weeks, 3 months, 6 months, 9 months, and 12 months post-operatively, LAGB at 2 weeks, and at 2, 4, 6, 8, 10 and 12 months or more frequently, as necessary, for band adjustment using a standard clinical assessment and questionnaire. Those undergoing surgical intervention were counseled on a diet program consistent with post-bariatric surgery recommendations and were encouraged to exercise a minimum of 3–4 times per week and to focus on weight-bearing, aerobic activity.

Subjects randomized to undergo the LWLI underwent a standard behavioral weight control program delivered in an in-person, individual, format based on the intervention developed for the Diabetes Prevention Program (DPP)²⁰ and the Look AHEAD Study²¹ and adapted into a 12-month program for subjects with grade 1–2 obesity. During the initial 6 months of treatment, LWLI participants attended weekly in-person intervention sessions. During months 7–12, they attended inperson sessions on the 1^st and 3^rd week of the month and received brief telephone contacts on the 2^nd and 4^th weeks. Each session focused on a specific behavioral topic related to weight loss, eating or exercise behaviors. Participants were provided supplemental written materials and were asked to self-monitor body weight, eating, and exercise. All LWLI participants were prescribed an energy restricted diet (1200–1800 kcal/day) and were provided meal plans, meal replacements, and calorie-counter books. Moderate-intensity, 5-day-per-week exercise was prescribed beginning at 20 minutes per day and gradually progressing to at least 60 minutes per day, bouts of activity encouraged to be >10 minutes in duration.

OUTCOMES AND END POINTS

The primary study endpoint was feasibility of performing a randomized trial involving surgical and non-surgical treatments. Feasibility was assessed by recruitment, randomization and retention rates. Secondary outcomes included effectiveness to induce weight loss and diabetes improvements. Changes in lipids and blood pressure are also reported. All participants were assessed at baseline (within 30 days prior to start of intervention) and 12-month (330–390 days) later. Weight and height were assessed using a digital scale (Tanita; TBF-300A^®) and a standard stadiometer, and blood pressure was measured twice at each visit. Serum measures of glycated hemoglobin, 12-hour fasting plasma glucose (FPG), total cholesterol, triglycerides, HDL, and LDL were obtained. Assessed at 12 months were weight loss (change in weight (kg), percentage weight loss from baseline and change in BMI), glycemic control (change in FPG and glycated hemoglobin), medication usage (categorized; none, oral/other medications, insulin use), and partial and complete remission of T2DM according to the ADA 2009 definitions.²² Change in blood pressure and serum lipids were also assessed at 12 months. Co-morbid health conditions (dyslipidemia/hypercholesterolemia, hypertension) were evaluated using a standardized comorbidity status form at baseline and follow-up. Participants completed a treatment preference questionnaire (TPQ) prior to but independent of randomization at baseline and follow-up²³ with the participants asked to rate from 1 to 6 (‘strongly prefer’ to ‘strongly do not prefer’) how they felt about each of the three treatments.

STATISTICAL ANALYSIS

Due to the small sample size of this feasibility trial, some of the analyses are descriptive in nature and results related to the efficacy outcomes should be considered as preliminary. Statistical analyses were performed using SAS (version 9.2), with the type I error rate fixed at 0.05 (two-tailed). Categorical variables are summarized by frequencies, continuous variables with normal distributions as mean ± standard deviation, and continuous variables with non-normal distributions as medians and interquartile ranges. Differences in baseline characteristics among the RYGB, LAGB and LWLI groups were examined using the Chi-square test or Fisher’s exact test for categorical variables and analysis of variance or Kruskal-Wallis test for continuous variables.

Changes from baseline to 12-months were analyzed using separate regression models. Each regression model included covariate adjustment for randomization stratification factors (gender and baseline BMI). Change in weight was further adjusted for baseline weight. Inferences were focused on the overall treatment effect along with three pairwise between group comparisons. Least-square means were obtained from the models along with their 95% confidence intervals. Intent-to-treat analyses were conducted using multiple imputation implemented using SAS procedures PROC MI and PROC MIANALYZE. For each outcome, ten datasets were imputed and results were then combined.

The Treatment Preference Questionnaire (TPQ) scores at baseline were categorized into 3 groups; congruent, no preference, incongruent (did not receive their treatment preference).²³ Partial remission of T2DM was defined as absence of any medications for diabetes at 12-months after surgery with glycated hemoglobin <6.5% and FPG ≤126 mg/dL. Complete remission of T2DM was defined as absence of medications with glycated hemoglobin <5.7% and FPG ≤ 100 mg/dL per ADA guidelines.²² For categorical data with missing values (e.g. T2DM remission, medication use, history of blood pressure or lipid abnormalities), we assumed no remission or no improvement for the condition at follow up.

RESULTS

Feasibility

Of the 667 patients assessed for eligibility, 463 (69.4%) were determined ineligible by way of telephone assessment. The most common reasons for ineligibility during phone screening were BMI out of range (n=170, 36.7%), lack of T2DM (n= 90 19.4%), not meeting age criteria (n=82, 17.7%), lack of interest (n=67, 14.5%), and not living locally (n=36, 7.8%). Other reasons (n=18, 3.9%) for exclusion included participating in another study, previous bariatric surgery, current smoking, unable to exercise, and significant medical conditions precluding general anesthesia. The 204 adults who were deemed eligible through phone assessment were invited to an informational session of whom sixty-eight (33.3%) did not attend. Later, 33 (16.2%) were excluded by medical assessment, 15 (7.4%) refused further participation, 12 (5.9%) had BMI out of range, 7 (3.4%) were unwilling to be randomized (1after gaining insurance approval for RYGB). As shown in Figure 1 for all screening (telephone and inperson), 74.7% of those excluded were due to ineligibility and 25.1% were due to lack of interest or unwillingness to be randomized. A total of 69 of 667 screened (10.3%) participants completed all requirements and were successfully randomized.

Figure 1.

Screening, Randomization, and Follow-up of Study Participants.

Of the 69 participants randomized into three treatment arms, 7 (10%) (3 RYGB, 1 LAGB, 3 LWLI) refused to undergo their allocated treatment after randomization and one participant was excluded the day of planned surgery (RYGB) due to current smoking. Sixty-one participants underwent their allocated treatment with 20 each undergoing RYGB and LWLI and 21 LAGB. Retention rates (defined as undergoing intervention and returning for 12-month assessment) in the RYGB, LAGB, and LWLI treatment arms were 90%, 86% and 70%, respectively. Overall ten participants withdrew or were lost to follow-up at the time of the 12-month assessment (83.6% overall retention) and one subject completed a partial visit with weight and health outcomes measured. Five of the participants were lost to follow-up and could not be contacted and 5 discontinued their participation in the LWLI (three less than one month into intervention and 2 prior to the 12-month assessment) due to lack of interest. Vital status was obtained and confirmed on 59 of the 61 treated subjects (96.7%)

Treatment preference incongruence by arm was similar (RYGB 4/24 (16.7%), LAGB 4/22 (18.2%), LWLI 3/23 (13%)) and the majority of participants were in equipoise prior to randomization. All 3 participants who refused RYGB and 3 LWLI preferred their respective arm and were not incongruent and the 1 LAGB treatment refusal was incongruent. Participants who were incongruent prior to randomization, but underwent treatment, were not less likely to be retained in the study.

For the entire sample, the mean age was 47.3±6.4, 81.2% were women, and mean BMI was 35.6±3.0. Mean HbA1c was 7.9±2.0 and the duration of T2DM at the time of baseline evaluation was 6.4±4.8 years and by treatment arm; 7.4±4.5 years RYGB, 6.1±4.3 years LAGB, 5.7±5.6 years LWLI (Table 1).

Table 1.

Characteristics of the Patients at Baseline (Intention-to-Treat Population)^*

	Intervention Assignment
Characteristic	RYGB (N=24)	LAGB (N=22)	LWLI (N=23)
Male sex– no. (%)	5 (20.8)	4 (18.2)	4 (17.4)
Age (yr)	46.3 ± 7.2	47.3 ± 7.0	48.3 ± 4.7
African American Race– no.(%)	8 (33.3)	3 (13.6)	4 (17.4)
Weight (kg)	99.8 ± 12.8	99.5 ± 14.1	102.6 ± 13.8
BMI (kg/m2)	35.5 ± 2.6	35.5 ±3.4	35.7 ± 3.3
Waist Circumference (cm)**	111.1 ± 8.2	114.5 ± 11.9	111.7 ± 9.5
Total Cholesterol (mg/dL)**	197.5 ± 39.3	189.5 ± 55.8	182.0 ± 39.0
HDL (mg/dL)**	42.0 ± 8.5	40.0 ± 9.3	44.1 ± 17.1
LDL (mg/dL)***	115.8 ± 45.7	90.6 ± 49.4	105.5 ± 33.3
Triglycerides (mg/dL)**– median(IQR)	119.0 (93.0–239.0)	139.0 (91.0–253.0)	119.5 (74.5–232.5)
Blood Pressure (mmHg):
Systolic**	139.4 ± 12.1	134.7 ± 17.0	132.0 ± 17.8
Diastolic**	81.1 ± 9.2	77.2 ± 8.7	76.3 ± 9.7
History of Hypertension– no./total no.(%)	12 (50.0)	13 (59.1)	16 (69.6)
History of Dyslipidemia/HChol– no./total (%)	14 (58.3)	16 (72.7)	15 (65.2)
Duration of T2DM (yr)****	7.4 ± 4.5	6.1 ± 4.3	5.7 ± 5.6
FPG (mg/dL)**	190.9 ± 78.1	180.0 ± 85.4	142.1 ± 28.0
Glycated Hemoglobin (%)**	8.7 ± 2.2	7.9 ± 2.2	7.0 ± 0.76
Diabetes Medication Usage:
None– no.(%)	0 (0.0)	2 (9.1)	1 (4.3)
Oral/Other Medications Only– no.(%)	12 (50.0)	12 (54.5)	16 (69.6)
Insulin Use– no.(%)	12 (50.0)	8 (36.4)	6 (26.1)

^*Mean ± Standard Deviation(SD) reported unless otherwise noted; HChol= hypercholesterolemia

^**N= 22, 21, 20 for RYGB, LAGB, and LWLI, respectively.

^***N= 22, 20, 20 for RYGB, LAGB, and LWLI, respectively

^****N= 21, 21, 20 for RYGB, LAGB, and LWLI, respectively

Weight Change

Both BMI and weight change differed significantly from zero within all treatment groups (Table 2). RYGB participants had the greatest weight change compared to both LAGB and LWLI (−27.0% for RYGB, −17.3% for LAGB, and −10.2% for LWLI) from baseline (P<.0001).

Table 2.

Measures of Weight Change and T2DM Improvement and Remission (Intention to Treat Population, Multiple Imputation)

	Intervention Assignment			P Values
Characteristic	RYGB (N=24)	LAGB (N=22)	LWLI (N=23)	Overall*	RYGB vs. LAGB	RYGB vs. LWLI	LAGB vs. LWLI
BMI (kg/m2)
Baseline	35.5 ± 2.6	35.5 ± 3.4	35.7 ± 3.3
Change in BMI	−9.68 (−11.0, −8.34)	−6.24 (−7.65, −4.84)	−3.62 (−5.36, −1.88)	<0.0001	0.0012	<0.0001	0.0167
Weight (kg)
Baseline	99.8 ± 12.8	99.5 ± 14.1	102.6 ± 13.8
Change in Weight	−26.8 (−30.6, −23.1)	−17.4 (−21.3, −13.6)	−10.3 (−15.1, −5.46)	<0.0001	0.0013	<0.0001	0.0179
% Change in Weight+	−27.0% (−30.8%, −23.3%)	−17.3% (−21.1%, −13.5%)	−10.2% (−14.8%, −5.61%)	<0.0001	0.0009	<0.001	0.0172
Waist Circumference (cm)
Baseline	111.1 ± 8.2	114.5 ± 11.9	111.7 ± 9.5
Change in Waist Circumference	−25.0 (−28.8, −21.1)	−16.8 (−20.2, −13.4)	−8.09 (−12.2, −3.96)	<0.0001	0.0020	<0.0001	0.0014
FPG (mg/dL)
Baseline	190.9 ± 78.1	180.0 ± 85.4	142.1 ± 28.0
Change in FPG	−79.7 (−110, −49.4)	−53.3 (−85.2, −21.5)	−7.20 (−41.5, 27.06)	0.0212	0.2375	0.0021	0.0450
Glycated Hemoglobin (%)
Baseline	8.7 ± 2.2	7.9 ± 2.2	7.0 ± 0.76
Change in Glycated Hemoglobin	−2.30 (−3.12, −1.48)	−1.05 (−1.88, −0.22)	−0.06(−0.93, 0.81)	0.0075	0.0359	0.0002	0.1030
Diabetes Medication Usage**
None– no.(%)	14 (58.3)	8 (36.4)	1 (4.4)
Oral/Other Medications Only– no.(%)	5 (20.8)	6 (27.3)	16 (69.6)	<0.0006	0.3108	0.0002	0.0055
Insulin Use– no.(%)	5 (20.8)	8 (36.4)	6 (26.1)
T2DM Remission
Partial Remission of T2DM– no.(%)′	12 (50.0)	6 (27.3)	0 (0.0)	0.0005	0.1146	<0.0001	0.0092
Complete Remission of T2DM– no.(%)″	4 (16.7)	5 (22.7)	0 (0.0)	0.047	0.7178	0.1092	0.0216

Baseline values: Mean ± Standard Deviation (SD), Change values: Mean (95% confidence interval)

⁺with adjustment for BMI, gender, and baseline weight

^*Represents the overall comparison of the change between RYGB, LAGB, and LWLI

^**Represents the 12 month data. For baseline data refer to Table 1

^′Partial Remission of T2DM defined as absence of medication for diabetes at 12-month after surgery with glycated hemoglobin <6.5% and FPG ≤ 125 mg/dL

^″Complete Remission of T2DM defined as absence of medication for diabetes at 12-month after surgery with glycated hemoglobin <5.7% and FPG ≤ 100 mg/dL

Diabetes Remission and Medications

As shown in Figure 2, no participants in the LWLI treatment arm experienced either partial or complete remission of T2DM at 12 months. For the surgical treatment arms, 12 of 24 (50%) undergoing RYGB were classified with partial T2DM remission at 12 months and 4 (16.7%) with complete remission, as compared to 6 of 22 (27.3%) and 5 (22.7%) of those undergoing LAGB, respectively p=.0005, p=.047. Pairwise treatment comparisons (see Table 2) were significant for RYGB vs. LWLI (p<.0001) for partial T2DM remission and for both partial and complete remission for LAGB vs. LWLI (p=0.0092, 0.0216). As shown in Table 2, there were significant reductions in diabetes medication usage at 12 months for both RYGB and LAGB compared to LWLI though no LAGB (n=8) (or LWLI (n=6)) participant on insulin at baseline was discontinued at 12 months. There were significant treatment differences for glycemic control at 12 months with RYGB vs. LWLI (p=0.0021) and LAGB vs. LWLI (p=0.0450) for FPG change and RYGB vs. LAGB (p=0.0359) and RYGB vs. LWLI (p=0.0002) for change in glycated hemoglobin. For the LWLI group, there were no significant improvements over the 12 month period in either FPG or glycated hemoglobin.

Figure 2. T2DM Remission at Month 12.

Partial Remission of T2DM =no medications for diabetes, glycated hemoglobin <6.5% and FPG ≤ 125 mg/dL.

Complete Remission of T2DM = no medications for diabetes, glycated hemoglobin <5.7% and FPG ≤ 100 mg/dL.

Missing data at follow up was assumed to be no remission.

Lipids and Blood Pressure Outcomes

The only change in cholesterol values that showed a significant treatment effect was change in HDL (p=0.0187, overall and RYGB vs. LWLI, p=0.0016, e-Table 1). Dyslipidemia or hypercholesterolemia at baseline was reported in 58%, 73%, 65%, of RYGB, LAGB, and LWLI, respectively and 29%, 55%, and 65%, respectively at 12 months. Hypertension was diagnosed at baseline for 50% RYGB, 59% LAGB, and 70% LWLI and for 38%, 32%, and 74%, respectively at 12 months. There were no significant reductions in blood pressure measurements. Among those with hypertension at baseline, antihypertensive medication therapy was reduced or discontinued in 58% of subjects undergoing RYGB, 54% of subjects undergoing LAGB, and 13% of LWLI.

Safety

Six surgical subjects stayed one additional night following their operation for either nausea or glucose medication management. One LAGB patient required a second procedure to replace a port that later detached from its position on the muscle and 3 other surgical participants had mild symptoms requiring clinical attention. There were no deaths and 3 serious adverse events. One RYGB participant developed an ulcer that was treated medically and 2 LAGB subjects were hospitalized for dehydration (Table 3).

Table 3.

Adverse Events^*

	Intervention Assignment
Adverse Event	RYGB (N=20)	LAGB (N=21)	LWLI (N=20)
Serious Adverse Event		No. of patients (%)
Requiring Hospitalization	0	2 (9.5)	0
Anastomotic Ulcer	1 (5.0)	0	0
Other Adverse Event
Prolonged Hospital Stay (+1 night)†	3 (15.0)	3 (14.3)	0
LAGB Port Malposition	0	1 (4.8)	0
Pruritis/Erythema at Incision Site	0	1 (4.8)	0
Abdominal Pain	0	1 (4.8)	0
Nausea and Emesis Requiring IV Hydration	1 (5.0)	0	0

^*Reflects events observed in all treated subjects regardless of follow-up status at 12-months; Patients may have had more than one event.

^†Reasons for prolonged hospital stay included elevated blood glucose, pain, and nausea prior to discharge.

DISCUSSION

This study highlights several potential challenges to successfully completing a larger RCT for diabetes and obesity treatment in those with BMI 30–40 kg/m² and helps to outline what options would be feasible to help answer the important question of comparative effectiveness in this target population. The feasibility aim was achieved and we show that RYGB was the most effective treatment followed by LAGB for both weight loss and diabetes outcomes at one year in a sample that included over 40% of subjects with BMI 30–35 kg/m² (low BMI) for which data are significantly lacking.¹⁰

Though we demonstrate feasibility, it required considerable study effort to recruit (only 10.3% of those screened were randomized) and retain subjects. Recently published RCTs demonstrate quite variable screening rates ranging from 4.5% to 83.3%^1–3 of those screened being randomized and each required several years to complete that phase. To reduce the number of people screened to meet recruitment goals future trials should target recruitment, e.g., from weight loss clinics, endocrinology clinics and other sources as opposed to the widespread advertising methods used for this study.

Retention is also a challenge. Despite our specific efforts to extensively assess willingness to undergo any of the three treatment arms (clinical equipoise) and, according to measured treatment preference achieving it in the majority of participants, 10% of those randomized dropped out after randomization, but prior to intervention. Interestingly, this could not be explained by stated treatment preference alone. Although retention efforts were equivalent across all three treatment arms, slightly higher rates of loss to follow-up in the LWLI arm were observed than in the surgical arms. There was not a significant relationship between treatment incongruence and retention status, i.e., those participants (16% overall) who were incongruent in undergoing treatment were not less likely to be retained in the study. However, the lower retention in the LWLI arm may have reflected unexpressed or unmeasured disappointment with treatment allocation for those that may have been seeking bariatric surgery. The retention rate in the LWLI arm is lower than that observed in Look AHEAD (94% at 4 years)¹⁹ indicating that more effective retention strategies than those employed in this trial (frequent contacts, relationship-building) need to be adopted. Furthermore, the differential retention by treatment arm and our inability to identify reasons based on our data indicates that there are potential inherent and unidentified difference(s) in motivation to be in a trial among subjects willing to undergo a surgical vs. non-surgical treatments.

Finally with regard to feasibility, funding for all the surgical procedures was significantly subsidized by the medical center where the study was conducted and those with low BMI would not have had their surgery covered by insurance that uses current criteria for bariatric surgery to determine reimbursement eligibility.²⁴ Substantial resources also appear to be necessary to maintain high levels of retention, with some new strategies geared to address drop-out prior to intervention implementation. Importantly, this study included only 1 year follow up and longer-term follow up is necessary for more definitive conclusions regarding the relative efficacy of treatments and are likely to require substantial resources.²⁵

In summary, these important feasibility issues; the need for targeted screening, the thorough assessment of clinical equipoise to reduce the likelihood of post-randomization drop-out, challenging retention in the lifestyle arm, and funding raise a number of concerns for the scalability of this type of study to a larger, multi-center trial. For this reason, it may be more reasonable to pursue alternative methods to obtain high order evidence regarding outcomes of surgical versus non-surgical treatment for people with relatively low BMI. One possibility is a pooled analysis of other similar feasibility RCTs^13–16,18 prior to undertaking a new, larger, multi-center trial.

With respect to the short-term effectiveness of weight loss and diabetes treatment outcomes, RYGB was the most successful, followed by LAGB, and then LWLI for weight loss outcomes only. In the LWLI arm, despite weight loss of 10.2% from baseline, which was more than what was achieved in the Look AHEAD Trial for the intensive lifestyle intervention (ILI) (8.6% average weight loss at 1 year),²⁶ and the shortest duration of disease there were no significant improvements in glycated hemoglobin, FPG, medication use, nor any partial or complete remission of T2DM. It is unclear why Triabetes LWLI participants did not achieve significant diabetes and other health improvements as seen in Look AHEAD, despite comparable weight loss and a similar magnitude of improvement in lipids and blood pressure. This may have to do with the relatively small sample size and variability in response, diabetes severity differences, medication adjustment differences that were carried out by non-study and different treating physicians, or the impact of fitness changes (not addressed in this paper).

RYGB resulted in average weight loss of 27% from baseline as well as 50% partial and 17% complete remission of T2DM at one year, which is fairly consistent other published results for weight and diabetes remission depending on the thresholds used for definition.^1–3,8,27 This is the first RCT in the U.S. to include a LAGB treatment option in this population for these endpoints and resulted in an average of 17% weight loss from baseline and 27% partial and 23% complete remission of diabetes. These results are somewhat lower for both weight and T2DM outcomes than previous reported studies.^7,28,29 A RCT by Dixon demonstrated an average 21% initial weight loss and 73% ‘remission’ of T2DM (using a different definition than current ADA guidelines) but the 60 subjects had mild diabetes with disease duration less than 2 years.⁷ The diabetes remission rates for LAGB in this study exceed those for intensive medical management alone in this trial (0%) and in Look AHEAD (11.5%)³⁰ but do not reach those observed with the gastric sleeve (37%, STAMPEDE)³, another restrictive procedure. In this trial the complete remission rates for T2DM are comparable in the two surgical treatment arms but the sample sizes are small, the study includes people with a broad range of diabetes severity, and follow up is limited to 1 year, so no definitive conclusions can be drawn. Nevertheless, the LWLI arm had no T2DM remission so these results suggest that there may be a place for the LAGB procedure, a low risk and potentially reversible procedure, as a treatment option for low BMI subjects with T2DM and further studies will be needed to confirm these results.

This study has some notable strengths, particularly the comparison of the two current and most common surgical procedures with an intensive nonsurgical treatment arm that has been standardized and extensively reported in the literature.^19,21,26 In addition, it is the first of 6 RCTs funded to be completed that address these specific issues of feasibility and comparative effectiveness.^13–16,18 Its limitations are the small sample size and the lack of generalizability of single site results. Nonetheless, the results of this study were intended to be a first step to inform a larger, more comprehensive, and more long-term multi-center trial to address critical unanswered questions. We conclude that the lessons learned here should be combined with those to come in the next two years from several other similar pilot studies^13–16,18 prior to planning a larger RCT. Such a study would require multiple sites and should compare the three most common bariatric surgical procedures in current use (RYGB, LAGB, Gastric Sleeve) to each other with or without an adjunctive lifestyle component or treatment arm over the longer term.

Finally, it is clear from this trial that at one year bariatric surgery is more effective to treat T2DM than non-surgical weight loss induced by diet and increased physical activity in people with T2DM and grade 1 and 2 obesity. What is not yet clear from this trial and others to date is whether the short-term T2DM improvements following bariatric surgery are related to the degree of weight loss, the type of surgical procedure, or other factors.